Professional Documents
Culture Documents
Manual of Transportation Engineering Studies
Manual of Transportation Engineering Studies
foreword
xxi
Acknowledgments
Chapter 1
Introduction
xxiii
1.0 Introduction
2.0
3.0 Organization
2
2
4.1 Definitions
5.0 Summary
Chapter 2
Glossary of Terms
Chapter 3
1.0 INTRODUOION
9
23
23
24
24
24
24
25
25
25
26
26
28
32
34
35
35
Table of Contents vii
36
37
37
37
I.
r
I
''
38
40
41
5.0 SUMMARY
41
6.0 REFERENCES
42
6. 7 Literature References
42
42
Chapter 4
Volume Studies
1.0 INTRODUCTION
2.0
43
43
TYPES OF STUDIES
44
44
49
58
3. 7 Manual Observation
58
62
67
67
68
69
70
5.0 SUMMARY
74
6.0 REFERENCES
74
Chapter 5
1.0
77
INTRODUCTION
77
7. 7 Safety
78
78
79
79
2. 7 Introduction
79
79
84
86
86
86
87
88
88
91
92
5.0 SUMMARY
95
6.0
95
REFERENCES
Chapter 6
1.0
INTRODUCTION
98
99
4.0
5.0
6.0
97
98
2.0 DELAY
3.0
97
99
QUEUE LENGTH
104
105
105
105
106
109
109
110
110
112
112
113
113
7.0
SUMMARY
115
8.0
REFERENCES
115
Chapter 7
1.0 INTRODUCTION
117
117
1.1 Purpose
117
118
118
119
123
124
Table of Contents lx
3.2 Signs
4.0 REMOVAL OF UNNECESSARY TRAFFIC CONTROL DEVICES
5.0 EFFECTIVENESS OF TRAFfiC CONTROL DEVICES
6.0
7.0
132
135
135
136
136
136
137
TCD CONDITION
137
137
139
140
SUMMARY
8.0 REFERENCES
140
140
140
141
Chapter 8
1.0 INTRODUCTION
143
143
J!
1.1 Purpose
144
~};
1.2 Applications
144
rt
1:
145
145
145
145
146
146
147
147
148
155
s_o
SUMMARY
156
6.0
REFERENCES
156
6. 1 Literature References
156
157
Chapter9
1.0 INTRODUCTION
159
1S9
1.1 Applications
160
160
160
2.1 Introduction
2.2 Data Collection Procedure
161
165
166
171
171
173
173
173
174
4. 1 Uterature References
174
174
174
160
177
INTRODUCTION
177
1. 1 Chapter Objectives
177
178
179
182
182
187
190
191
191
191
192
194
194
195
195
197
198
5.0 REFERENCES
Chapter 11
Simulation Studies
1.0 INTRODUCTION
199
201
201
201
202
202
Table of Contents
xi
203
1.5 Definitions
205
206
206
208
210
212
215
217
218
218
221
227
228
229
229
231
4.4 Documentation
232
232
233
REFERENCES
-
229
5.0 SUMMARY
6.0
223
234
"'
237
1.0 INTRODUCTION
237
238
238
246
246
250
251
253
257
4.0 REFERENCES
vii
250
3. 1 Introduction
259
259
261
261
. . "1111\1 f'IF
Chapter 13
263
264
264
265
265
265
265
265
268
271
271
272
273
281
283
285
285
285
286
288
288
5.0 REFERENCES
288
288
289
289
291
1.0 INTRODUCTION
291
291
3.0
293
298
299
302
304
REFERENCES
305
305
307
307
Chapter 15 Inventories
1.0 INTRODUCTION
31 0
310
l:i
309
1. 1 A Word of Caution
3.0
309
311
311
311
316
316
317
2.5 Storage
317
2.6 Retrieval
317
ESTABUSHING AN INVENTORY
317
317
318
318
318
320
320
320
4.1 Software
320
321
5.0 REFERENCES
321
323
1.0 INTRODUCTION
324
324
Study Locations
325
325
325
2.4 Method
326
329
330
332
332
33 6
337
338
339
4.0
. 5.0
6.0
Tabulation
339
339
SUMMARY
343
REFERENCES
345
345
345
345
347
1.0 INTRODUCTION
347
2.0
347
3.0
348
348
352
354
355
357
358
5.0
369
379
REFERENCES
379
383
1.0
INTRODU CTION
383
2.0
384
2.1 Introduction
384
384
385
388
4.0
390
3.1 Introduction
390
390
394
405
ADVISORY SPEEDS
406
4.1 Introduction
406
406
Table of Contents ~
5.0 REFERENCES
409
412
413
1.0 INTRODUCTION
413
414
414
415
415
415
415
419
3.5 Collisions
420
424
424
424
424
425
5.0 REFERENCES
426
427
427
429
1.0 INTRODUCTION
429
430
430
431
2.2 Inventories
432
437
438
438
438
446
426
5.1 References
\\,
423
448
448
448
5.0
450
REFERENCES
450
Chapter 21
Environmental Impacts of
Transportation Projects
451
1.0 INTRODUCTION
451
2.0
452
453
454
3.0
455
456
457
458
459
459
459
459
460
.
4. 7 References
461
461
461
462
463
1.0 INTRODUCTIO.N
463
Pr~paration
464
464
464
465
1.5 StudyComponents
466
473
473
477
478
479
3.2 Capacity
479
480
480
481
481
Table of Contents xvii
4.0 SUMMARY
4.1 Presentation
482
482
5.0 REFERENCES
483
485
1.0 INTRODUCTION
485
486
2.1 Definitions
486
2.2 Objectives
486
2.3 Statisticallnference
486
487
4.0
487
487
488
BEFORE-AND-AFTER EXPERIMENTS
489
489
490
492
492
494
494
6.0 REFERENCES
498
499
1.0 INTRODUCTION
499
500
500
2.0 METHODS
500
501
501
504
505
505
505
506
4.4 Genera/Tips
507
508
509
510
8.0 PRETESTS
510
510
512
11 .0 REFERENCES
514
515
1.0 INTRODUCTION
515
516
S16
517
S20
521
S22
522
3.2 Variab/1/ty
526
4.1 Estimation
s.o
528
528
529
532
533
CALCULATION AIDS
536
536
7.0 REFERENCES
538
539
1.0 INTRO.DUCTION
540
540
S40
540
2.3 Tables
542
543
549
549
550
4.0
550
550
550
552
PRESENTATIONS
554
Table of Contents ,...cix
554
555
555
557
557
558
5.0 REFERENCES
559
561
561
1.2 Coding
562
\i
561
1.1 Overview
564
Foreword
his second edition of the MIZ'miAI ofTrrmsportatitm Enginring Studies (MTES) is an updated and expanded version
to the 1994 MTES, which it3d.fwasa mision of the l976MIZ'miAiofTraffic EnginmingStvdUs, 4th Edition, by Box
and Oppenlander. Some original ooncent of these earlier editions is maintained throughout this publication.
This edition includes a reorganiurion of the clupters into six pans of related transportation studies. Fout new chapters
were added to this edition including a G/QSSJJry ofTn711S, a chapter on (;QmmunictZting Dlll4 to tht Public, a chapter
on ~=ay and Managtd Lann and a chapter on SimulatWn Studies. These new clupters represent new and emerging
trends in the transportation profession and are important additions to this publication.
Work on this second edition of the Institute of Transportation Engineers (ITE) MTES began in 2008 with the
confirmation of the project and author team. The team worked closdy with ITE staff and members of the Technical
Advisory Committee to draft an outline of the revised manual. Individual chapter~uthorship responsibilities were
based on areas of e:q>ertise and a detailed annotated bibliography was created to update references throughout the
manual. The revised oudine and annotated bibliography were n:viewed and approved by ITE and the project Technical
Advisory Committee before work on the final chapters oommenc:ed.
The ooncent of this handbook was finalized prior to the release of the 2009 Edition of the Manual on Uniform Traffic
Control Dmm (MUfCD). Therefore, although all general references to the MUTCD have been updated to 2009,
content with specific references to the 2003 MUTCD has been retained. Users are encouraged to consult the current
edition of the MUTCD when making technical determinacions. A free copy of the 2009 MUTCD (PDF or HTML
format) is available at htrp:f/mutcd.fhw.a..dot.gov. ITE, in oooperation with AASHTO and ATSSA publishes a hard
oopy version available via the ITE bookstore at www.ire.org/bookscore.
This publication is .designed to aid rransportacion professionals and communities in studying their transportation
JJWUICf, following procedures accepted by the profession.
system in a suucrurM
The MTES should be used in conjunction with ITE's Traf!ic EngiMmng Handbook, Transportation Pilznning Handbook
and other resources in order to prove most useful.
As we release the second edition of the MTES, I hope that it will continue to be improved in later edicions. Your
coqunents and suggestions are earnesdy solicited as a means of making those improvements.
..
,..._,
Acknowledgments
o I
any hours of concerted effort wenr inro the development of this second edition of che Institute of
Transportation Engineers (ITE) Manual ofTransportation Enginuring Studies (MTES), which would not
have been possible withouc the dediatced and thorough work by the individuals listed below.
My responsibilities as editor were shared by co-edicqrs Christopher M. Cunningham, P.E. and Daniel ). Findley,
P.E., who also served with me as authors of the overall manual, and as principal au~ors of individual chapters.
Appreciation also goes to chapter co-authors Mr. Robert S. Foyle, P.E. and Dr. Joseph E. Hummer, P E. (M), whose
professional expertise and technical insights made invaluable additions to the content of this manu al. As editor. I
gceady appreciate the synergy and professionalism of ch.is core project team, which led co the productio n of a coherent
and consistent publication.
The authors would further like to thank a group of reviewers, who offered valuable reviews and feedback during
chapter development including Mr. Loren D. Bloomberg (M), Dr. James A. Bonneson, P.E. (F), Dr. Darcy . M .
Bullock, P.E. (M), Dr. Ronald Hughes and Dr. Nagui Rouphail. The authors also excend their thanks co the members
of the Technical Advisory Coi:nmiccee ~ted below and ocher reviewers of draft chapters for the expert ise and bdpful
i~ight they provided.
lWp,b W. Boaz (M)
MichadE.Kti
Berty H. Tusrin,
(Lettm in parentheses jnJicate ITE member grad4: M-Mmsber, F-Fel/Qw, MS-Member Graduate Student.)
Acknowledgments XlC :5ii
___ ..
-.
Chapter 1
Introduction
OriginAl By:
H. Doug/iu !Wbmson, Ph.D., P.E.
joseph E. Hrnmner, Ph.D., P.E.
DtmnA C Nelson, Ph.D., P..
UptLzutl By:
BastiAn]. Schroetkr, Ph.D.
Cbrisfqpher M. Cun~injhmn. P.E.
Dtmiel]. Fitulky, P..
1.0
Introduction
3.0 Organization
4.1 Definitions
s.o Summary
1.0 INTRODUcriON
ransportarion ~tems ue the hckbone of a nation's infrastructure. They provide a means foe people to
travel to work, suppon home-based activities, enable vacation trips and serve as access to recreational and
leiswe activities. Swface transportation networks further ~ntinue to cury a majority of freight movement
and therefore directly affect economic prosperity. They enable rapid movement by fi.rst responders indudin~ fire,
police and ambulance traffic, and represent strategic infrastructure for evacuation routing and martcrs of national
secwity. The fields of transportation engineering and transponation planning have devised numerous methods
for describing traffic How, forecasting tcavd demands and predicting future operations of transportation facilities.
These methods are often focused on the automobile and truck modes, buc they also encompass nonmotorizcd
modes of transponation, as well as transit. They encompass a range of transportation facility types, including rural
roads, signalized anerial streets and limited access f.acilities. Many of these methods ue derived empirically and all
require some level of field data as inputs.
The nation's transportation infiastrucrure also f.aces challenges. Problems of congestion, delay, puking. po!IHtiQp and
safety are expected to dominate swface transportation foe the foreseeable future. Solutions to transportation problems
Introduction 1
and improvements to transportation facilities and services can only reasonably be developed after the magnitude, location and extent of the problems or the need for improvementS are well undemood. Such understanding comes from
factual information gathered in an unbiased, objective manner, and analyzed to present a clear, concise picrurc of the
narure of the problem.
3.0 ORGANIZATION
3.1 Organization of the Manual
Each chapter in this manual is written so thar it largely stands alone without the need for extensive cross-referencing.
However, to assure a con~e and efficient delivery of material some chapters will build on methods described in an
earlier chapter. The chapters are grouped into parts of related topics as follows:
Introduction
Pan I: Spot Locations
m;
Part
Muhimod41 and Network Data
Pan: III discusses alternate modes of uanspornrion, including a chapter on pedestrian and bicycle studies foUowed by
a chapter on public tranSportation studies. Both chapters emphasize concepts of user perception to describe the qu:Jity
of service of the transportation service. This parr concludes with a chapter on goods movement studies, an important
area of uansporration receiving increased attention in the profession.
Part IV: Assn Management Data
Part IV contains two chapters on asset management studies. The chapter on inventories contains detailed discussion
on automated data collection, including Global Positioning System (GPS) data and Geographical Information Syscem
(GIS)-based data management. This pan: also contains the chapter on parking srudies.
4.1 Definitions
This book is a mwual of transportation studies. In this context, a study is defined as the detailed examination
wd analysis of all or parts of a transportation system, supported by empirical data collection. The study statts
with the identification and definicion of a transportation problem, followed by the design and execution of
(field) data collection and the reduction and :malysis of the data in the office. A study Is typically performed to
explore a specific aspect of or question about a transportation system, and study results arc usually written up
in a ree,ort or similar document.
Transportation studies arc oftentimes supported by dtzta colkction to empirically gather data in support of the
study. Data collection traditionally is performed in the fie!~ using various study tuhniqu~ and titztll colkction
equipment that arc described in more detail throughout this mwual. A Hudy technique describes the stepby-step procedure used for gathering data. Data collection equipment refers to a technological device that is
specifically designed to measure one or more data elements (for example, volumes or speeds). Data collection
for transportation studies can also be performed without gathering data in the field, but by extracting information from a central system (for example, the U.S. Census or other database), or by modeling the trmsportation
system in a traffic simulation tool.
/ "WTTple, tiafli.c
'
C.
,rmancc of an
'!roblem the
: I
Transportation srudies are expensive and should not be conducted without considering tl{ / <ternatives. l "Cis no
general methodology for planning a study or selecting a srudy technique, since transpo! ; .on problems
n are
unique in scope and sensitive to regional differences in travel patterns. The following q ! tions are rdevan ben
deciding whether to use field <:bu. collection, and which srudy technique is appropriate./ i. walysr should cot 1er
each question carefully when planning a ttansportation study.
What is me purpose of the srudy and what measures are the desired outcome?
What analysis method will be used to solve the 'problem being faced? (Do not proceed until an analysis
method is selected.)
What input dara are needed for the analysis method?
&e there acceptable values &om previous work that can be used as input <:bra? (If"yes~ for all inpua, do
not use field <:bta collection.)
Are data available that can be mmipulated to become acceptable as input <:bta? For example, if ruming
movement counts are needed, ace estimates from available link volume <:bta and nerwork geometry
acceptable? (If "yes, do not use field dat2 collection.)
&e field srudy techniques available that will provide the input data needed? (If "no," do not use field
data collection.)
Ace the time, money, personnel and other resources needed to conduct the field study available? (If"no," do
not use field <:bta collection until the resources become available.)
i"
\\
\
Is there more dwl one 6dd study technique that will provide the needed input data with available
resources? (If"yes,. use the most cosr-efficient study technique.)
4 MANUAL OF TRANSPORTATlON ENGINEERING STUDIES. 2ND EOffiON
, The question whether to conduct fidd dat:a collection and the choice of a particular study technique arc driven by the
~ needs of the question that must be answered and the analysis that is planned. Ax any time before or during the study,
! the engineer has the option of canceling or re-scheduling data collection if there is a change in the conditions that led
fo the choice of study technique.
lntrodurtinn 5
.-,
.c.
-~~:t~f;l.,
"
. .' __
e . _..
-.:-o~
. ...
...
:~""f-F;:.,3
I.
To check che d2ta collection equipment. M3ke sure that ic records, scoces, and!or pcoduces oucpuc as you
rcquirt. Make sure that the equipment is c:alib1'2.ted pcoperly.
2.
To label the equipment as needed (e.g., turning movemenr counters can be labeled with the approaches
being watched).
3.
To bring the data collection equipment. Also. bring spares of equipment such as small, reliable
stopwatches.
4.
5.
6.
An abundance of forms. Also bring a clean copy of the form from which more copies can be made if needed.
7.
8.
Plenc:y of pens.
9.
10.
A letter &om the landownet or bis/hec contact information (if private propercy will be used) and/or from a
responsible agent of the highway aurhoric:y giving permission co collect data.
II.
A few business cards of the engineer supervising the study. The engineer's name and telephone number are
~etimes adequate.
12.
A short, simple answer co the question, "Whac are you doing here?"
13.
The telephone number where the supervising engineer can be reached on the day of the nudy, in case
questions arise.
14.
15.
Folding chairs.
16.
17.
18.
Sunburn protection.
19.
20.
Data collectors must typically arrive at least 15 minutes early at the site in order to assess conditions, distribute equipment, record crucial "header" informacion, assume positions and begin at the scheduled rime. The "header" informacion muse appear on each form or data record and include items such as (but not limited to) site name, date, time,
observer name and weather condition. Sample forms included in Appendix E of this manual use a consistent header
that assures all pertinent information is recorded. The same "header" informacion is also entered in decuonic count
boards and is crucial to assure that the dataset can be uniquely idenci.6ed among all the informacion stored in the device,
6 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
, even if electronic count boards, computers, or other advanced data collection equipment is used. A hand h eld clipboard
\is always a valuable asset to record any un usual events during the study or note general observations.
Header information should be completed just prior to beginning the srudy. Data collectors waiting unril the end of
che day to record header informacion on a stack of forms arc taking an unnecessary risk. Data collectot's should note
any unusual occurrences in the transportation system that could affect the data being collected. Any deviations from
accepted collection procedure should also be noted and probably should be cleared 6m by a responsible lead obser1fer
or engineer. Data should always be recorded in ink to prevent fading, smudging and erasing. A clipboard assures that
data collectors have a sturdy writing surface and that completed sheets are secured.
Data collectors on pubUc or private property will often be asked, in cones ranging from poUte to threatening. W'}lat
this question chat will satisfY most
members of the public without distracting too long from the data collection rask. A calm, professional approach and
a referral to the supervising engineer are usually enough to defuse even very suspicious inquirers. Ic is always useful co
carry the business cards of the supervisor whom people can call for questions. This is especially important if daa col
lectors are confronted by law enforcement personnel.
are you doing here?" Data collectors should be taught a short, standard response to
If data collection is to occur on private property, it is recommended char the engineer overseeing data collection contaCt
the property owner to inform them of the impending data collection, the nature and purpose of the data, the name of
the data collection firm, the anticipated dates of data collection and the narure of the data collection material. Providitlg
advanced notice to private property owners may a~id potencial conflicts in the field.
Local law enforcement should be made aware of data collection activities, especially if the study involves a laser speed
"gun", or a permanent data collection unit tied to a sign post that may look suspicious to the untrained eye. Similar
concerns may arise for any data collection chat uses video observations. Even standard" data collection (for cxample.
counting traffic volumes from a parked car) may appear suspicious to some citizens and a quick phone call to the laW
enforcement agency can prevent unnecessary complications.
In addition, it may be beneficial to inform local government agencies about when and where data collection is planned
and the method to be used, so local government staff can properly address any concerns raised by the public contacciflg
the agency. If data collection involves the use of automated or video equipment that requires the equipment ro be a:crached or fixed to ai.sting ruuctures in the field, such as Ughr poles or sign posts, a permit may be required in adv:lflce
from the local government agency. Consulting and informin.g that agency prior to data collection efforts can dctcnniP e
if any pe.rmits are required.
If da~ collection is performed &om within a moving vehicle, a second data collector (other than the driver;>
should perform all study activities, without distracting the driver.
Introduction II'
Data collectors should pay dose anention to the roadside environment, including holes, wires and potencial
poisonous wildlife (snakes, ants) and plants (poison ivy).
Data collection in cold climates or adverse weather requires adequate weatherproof clothing and ocher
accessories, such as hats and gloves. Daa collectors should not underestimate the potential health impacts
of cold and wee weather, especially when seated in a daa collection chair' for multiple hours.
Crime is also a threat to data collector safety during fidd studies. The: best defense: for data collc:ccors when criminal
behavior chrc:atens is usually to abandon the study and leave the area. Safety from crime can be enhanced by the following measures:
Minimize nighttime: data collection.
Collect data in teams of at least WO persons who remain in sight of each ocher at all times.
Alert the local police when a data collection c:ffort is under way.
o
f'
! .
Data collection often requires coordinating efforts with several members of a research team, who collect different data
items or may be: located in different locations. Data collection may also include the need to coordinate with equip
ment vendors or parmering municipalities for equipment installation or access to data sources. Considering these
different contributors, the engineer needs to bring interpersonal skills to conimunicating with these partners and co
supervise a pocentially large team of data collca:ors.
4
0
m
cc
cc
in
fc:
of
Data collection is subject to the uncertainty of weather and in some cases a project may actually require inclement
conditions. Delays due to inclement weather or malfunction of equipment have to be anticipated. Consequencly, data
collection efforts need to be well-scheduled and yet remain flexible to assure success.
In addition to manual and small-scale data collection c:fforu, advances in technology and autonomous data collection
methods may provide large quantities of data that can become unmanageable. Often, there is a need to prioritize data
from available resources and be efficient in both the collection and analysis of data, given financial and time considerations. Without careful planning. the analyst runs the risk of collecting too much data (both in terms of data dements
and guantity of data) that m;r.y result in inefficiency in data collection, as wd1 as in analysis time in the office.
wi
su
!
'
'
I
I
\:\~
\:
co
S.OSUMMARY
0
m
ti<
ol
Safe, efficient and c:ffective daa collection requires skill, artention to detail and common sense. The importance of
"good" data cannot be overstated. Important conclusions arc: drawn from fidd daa that form the basis for decisions
that a.ffcct the expenditure of large amounts of money and can have a significant effect on the safety of the public at
large. Data collection demands the same levd of professionalism as any other task undertaken by an engineer or engi
neering technician.
th
lx
6
Chapter 2
'
Glossary of Terms
ADT
Advisocy Spd
Algoridun
AMF
Approach
ATM
.Art2irunent Ara
Average Day
Ball-Bank Indicator
,~~~F""~
-_ -~ . :- . ;.;- -~ _;,,.;::~.-:-.--:
'.
_: ~:t~!t~~~~-~\~~i;;~~;~~~~~~:-.~~~;7~l$-_-.~~ ~~;
Beacon
A highway traffic control device wirh one or more signal sections that oper- 7,8
aces in a flashing mode.
Benefit-Cost Ratio
Bicycle
Bicycle Lane
A portion of a roadway that has been designated by signs and pavemem 7, 8,12
markings for preferential or aclusive use by l:iicydists.
BRT
[CAD
Calibration
Capacity
Car-Following
!
.I
Causal Chain
CBD
Cmtra/ businm district is the commercial and oftentimes also the geographic
center of a city. The CBD may often have unique transportation systems
incl11ding closely-spaced blocks and signals. It generally exhibits a high
developmental density and often features a mix of differenr land uses.
Centerline Markings
The yellow pavement marking line(s) chat delineates the separation of traffic 7,8
lanes that have opposite directions of travel on a roadway.
Checker
Chord
Any straight line measurement becween cwo points along a horiz.ontal curve, 18
typi~y taken from the centerline of the roadway.
Classi.6carion
A grouping of data into bins that describe different vehicle categories, such 4, 10
as passenger cars, trucks and buses.
Clearance Interval
Collision Diagram
Collision Report
The form used by the srate's department of motor vehicles (DMV) for 17
recording informacion about a collision.
Compliance
Concurrent Flow
HOY Lane
High occupancy vehicle (HOY) lane chat operates in the same direction as 7, 8, 10
the adjacent miud flow lanes, separated from the adjacent general pw-pose
freeway lanes by a standard lane stripe, painted buffer, or barrier.
I
I
17
.
8, 12
17
Condition Diagram
Confidence Level
Conflict
12, 18
Conuol Count
Cordon Coum
4, 13
Cosine Error
Count Expansion
n<"rtnnn<"tl
Countermeasure:
Covel'2ge Count
Crash
Crash Rate
CRF
Crash Severity
Critical Gap
17
The threshold gap 'rime used to determine whether vehicles, pedestrians, I 6, 11
or bicyclists at a minor approach enter or cross the major aaffic saearo. In
traffic flow theory, it is defined as the gap time where mi.norsaeet traffic is
equally lik.dy to accept or rejecr a gap. In the context of traffic simulation
models it is often lised synonymously with the concept of a minimum gap
where the minor-street vehicle will always accept gaps greater than the
critical 2:11>. and will alwavs reiecr shorter ones.
Crosswalk
The part of a roadway ac an intersection or at a midblock section that identifies the location for pedesuians to cross. The crosswalk is located within the
lateral connections of the sidewalks .on opposite sides of the highway, typi. ' or otherwise marked to be visible to drivers and
DDHV .
Delay
The difference berween acrual travel time and the theorecical uavd time at I 6, 10
free-flow oonditions. Ttm4-in--qunu tklay (TIQD) is the difference be~
the time a vehicle joins the rear of a queue and the time the vehicle clears
the intc.rsection. Control tklay is the component of delay that results when a
control signal causes a lane group co reduce speed or to stop; it is measured
by comparison with the uncontrolled cxindition. It is defined as the TIQD
plus time losses due to decderation &om and acceleration to fiee-Bow speed.
GrometrU Jelay is the component of delay that t'tStl.lts when geometric fewtts
cause users to reduce their speed in negotiating a facility. Tn:tvd-timl! Jelay
(TID) is the d.iffcrence berween the time a vehicle passes a point down.sttearn
of the incerscction, where it has regained normal speed, and the time it would
have passed thar point bad it been able to continue thro~ tbt intc.rsection at
This includes all control and
Delineator
Dapand
..
;
!~,
j:
Density
"
7,8
i
II
rl
,.!I
;j
:l
T
i
Disuibution
11
DMI
Doppler Effect
The physical principle behind speed measurements using a radar or Laser gun. I 5
The speed is inferred by measuring the fr~u~q $h.ift of a high frequency
laser/radar wave as iris reflected &om an aooroachirut or recedin2 vehicle.
DTA
Engineering Judgment I The ~uation of avaib.ble penincnt information and the application ofl7, 8
appropriate principles, standards, guidance and practices as contained in
the appropriate manual or other sources, for the purpose of deciding the
or installation of a device.
Engineering Srudy
EPDO
ERC
ET
Expms to,U lane is a managed lane freeway f.tcilicy that is open ro roll-paying I 10
vehicles only.
FHWA
multiple
Flow
Flow Rate
The number ofvehicles passing a pointin a period Je5s than aii hour expressed I 4, 10
in vehicles per hour. Operational analyses frequently usc a 15-min. flow rare,
which is the number of vehicles observed in 15 min. multiplied by four, to
obtain the hourly flow rate.
Gap
The available time in seconds between two successive vehicles at the same 17, 8, 12
point in space, measured &om the rear bumper of the lead vehicle to the
&om bumper of the followin2 vehicle.
AD. algorithm used in microsimulation models to describe the behavior of 1 11
a driver, pedestrian, or bicyclist entering or crossing a conBicting major
traffic scream.
3
Gap-Acceptance
GIS
GPLanes
Glossary of Terms 1J
~~;.
"";~~
0 .
:;:~;J~k;-
~w~~<TelJ:~:>~~k~~i~~: ~-~
GPS
Guide Sign
A sign char shows route designations, destinations, distances, ~rvices, points 7,8
of interesr, or other geographical, recreational, or cuirural informacion.
Highway Capacity Manual is aTransportation Research Board (TRB: see Iacer multiple
term) document providing computational methodologies for estimating the
operational performance of traffic Bow.
HCM
Headw:~.y
The time in seconds between the arrival of a common point on two succes- 6, 7, 8
sive vehicles (e.g. front bumper) co reach the same point in space. Headway
also describes the time in minutes between successive transit vehicle arrivals.
HOT
High occupancy and toll facilicy that is open to both HOY and drivers willing 10
to pay a toll to travel on the facilicy.
HOY
High occupancy v~hick is a motor vehicle carcying more than one person, 10
including carpools, van pools and buses.
HOY Lane
Any preferential lane designated for exclusive use by high occupancy 7,8, 10
vehicles for all or part of a day, including a designated lane on a freeway,
other highway, street, or independem roadway on a separate right of way.
HSM
Impact Score
Incident
1:
Interchange
Interrater Reliabilicy
ITE
Iteration
ITS
In~/ligtnt
Jay-walking
f'
,.
I
li
I
I
I "
~-1l.J.~--.;,.
. - .
~~:...,....... ';'i,l.\""i~
.
'
KABCO
Lag
Lane-Changing
Laser
Level
Levels of Service
18
Loa.d
13
Loop
See Smsor.
Macroscopic
Major Street
Managed Lanes
7. 8
A te;m encompassing alllimired access roadway &cilitics, including HOY, 10
HOT and Truck Only Toll (TOT) &cilitics. There are also some managed
lane &cilities along arterials.
Mean
The average of a sample calculated by the sum of all observations divided by multiple
the sample size.
Measwes of
Effectiveness
Measures or resrs which rdloct the dcgR:e of attainment ofparticular objectives. multiple
Mesoscopic
Michigan U-Turn
lntenection
Microscopic
Middle Ordinate
The measured distance from a known chord to the centerline of the roa.dway. 18
Minor Street
Model
7, 8
MOE
Multiple Threat
---
..
'
11
MUTCD
I ManUitl on Uniform
NCHRP
Netwark
NonattalnmentArea
I An area considered not to have met the U.S. Environmental Protection 121
Occupancy
0-D
cion network are compiled into an 0-D maaix that is used for planning and
analysis.
Online Mapping Tools IA desktop application for Mac, PC. or Limn computers that allows you to I 3
navigate around the world from multiple views. These tools combine satellite photos and maps with a search engine to allow the user to search for
directions and specific addresses or general locations and services. Multiple
service oroviders exist that orovide similar online maooin tools.
Operating Speed
speed might be defined with speed values such as the average, pace, or 85th
Optimization
0 nho-Rectification
The process of calibrating a video image for the pwpose of applying video I 4
image processing software for automated data collection from known
(true) dimensions such as pavement markings, lane widths, or the height
of the camera.
reoresents more I 5
typically
Pace Speed
13
Paratransit
Path-Based Count
A type of volume study that requires the analyst to trace a vehicle through I 4
an intersection such as a roundabout or
PCE
PwHour
~c
Performance Measure I A data clement used to quantify a particular aspect of cransportarion service I 13
and used to describe its quality ofservice. Examples ofperformance measures
include delay, travel time, or emissions.
Permissive Turn
I Left rurns may be made during the same signal phase as the opposing through I 4
movement. There is no exclusive signal indication (prorected phase), and
drivers must vidd to oncomiruz craffic and '
Platoon
12, 13
Queue
multiple
Queue Length
7, 8, 10
Raised Reflective
Pavement Marker
lUmp Metering
Reliability
10
Glossary of Terms 17
~~ti:.~"".l''~'
. .. " .... '''"''~. ;)';/' :,.:. .
.
-,~~lmr~>'ll ~ <.t't~~"lf~~~~!tt.~~-
~~;
i~~~~
Replication
Used in experimental design to describe how many units are evaluated or Appendi x
reseed for a particular treatment. For example, a study on a mfli.c-calming A
installation may include four replications, if the same treatment is applied
to four different incecsections. The term replication is also sometimes used
co describe an itnatirm in a simulation study.
Reversible Lanes
One or more Lanes char can be operated to carry traffic in either direction of uavel, 10
found on ~y .&cilicits, arterW streets and at sporting/event complexes. Traffie is managed through ITS treatments, variable signing and movable barriers.
RightofWay
RSA
Roundabout
Route
Run
See ltnation.
Running Speed
A measure of the average speed over a segment, which is calculated as the dis- 9
tance travded divided by che running time.
Running Time
The rime a vehicle is acruilly in motion (or moving faster than a predesignated 9
speed) while traversing a given segment ofstreet or highway.
SPF
Sample
A representative subset of the total population used for data anal)'3is. Scatis- multiple
tics (mean, st2ndard deviation) of the sample are used to approximate the
true parameters of che population.
A counr of ttaffic between two subareas performed by recording traffic flows 4'
at crossing points berween che two regions. A screen-line count is typically
performed along a narurai or human-made barrier wich limited crossing
points, such as an urban freeway or a river.
Screen-Line Count
11
Sensitivity Anal~is
Sensor
Service Measure
SbarCd-Use Path
A facility out'Side the main traveled way and ph)'3ically separated from motor- 7,8, 12
ized vehicular ttaffi.c by an open space or barrier and either within the highway
right of way or within an independent alignment. Shared-use pachs are also
used by pedestrians (including skaters, users of manual and motorized whedchairs and joggers) and other authorized motorized and nonmocorized usets.
Sign
Any aal:lic control device chat is intended to communicate speci.6c informa- 7,8
cion to road users chrough a word or symbol legend.
I
.
Signal Phase
Simulation
Speed
4,5,9,10
Speed Limit
Speed Trap
Spot Map
Spot Speed
SSAM
17
rs,
Statutory Speed
Steady-Scare
Superclcvacion
Superscreet
Synchronization
System
4, 12
TMS
TMC
unit that monitors traffic flow and ITS technology deployed on a freeway
or arterial network.
Tolerance
TOT
In a speed study, the tolerance may be the difference between before and
~e.r speed observations that is considered
I Truck only toll is a f.l.ciliry that is intended exdusivdv to process heavv 110
Traffic
Traffic Conflict
I Interactions between two or more vehicles or road users when one or more 118
Traffic Conuol Device I A sign, signal, marking, or other device used to ree:ulate, warn, or Uide I 7, 8
7, 8
11
Trajectoty
Treatment
-.A
'U.,II"\ CI"\IT1f'U.I
TSP
Units
User Perception
Validation
Vehicle
4, 5, 7,8
ed or dr.lwn
exclusive or
Vurual Earth
Visual Analytics
V LSUa.!iution
3,11
Volume
multiple
Warning Sign
7,8
Warrant
WIM
Glossary of Terms 21
Chapter 3
o 0
o o
o 0
o 0. 0
0.
o 0
o 0
o e
o 0
II
1.0 INTRODUCTION
1.1 Objective of This Chapter
1.2
Guidanc~
for.Other Readings
23
24
24
24
24
25
25
25
26
26
28
32
34
35
35
36
37
37
37
38
40
41
5.0 SUMMARY
41
6.0 REFERENCES
42
6. 1 Literature References
42
42
1.0 INTRODUCTION
b.is chapter gives an overview of modern techniques for data presentation and visualization for public involve ment in. \lllderstanding issues and bdping make decisions. With advances in computing technologies and >f'i--
sualization capabilities of modern software cools, new realms of dara display and ways to com.m,unicatedata c:4:i
the public and decision-makers have become possible. Ax the same time, people's levels of expectations of reporu an. .d
Communicating Data to the Public 2 ..::::3
public meetings have become elevated, and they have come to rely on sophisticated presentation techniques and visualization to convey traffic engineering data. The effectiveness of this presentation trend is further enhanced by simulation and animation techniques, providing the audience with four-dimensional representation nf traffic operations.
One question to be asked is, Where does 'data' stop, and 'information' begin?" (Transpornuion Research Board:
TRB, 2006). That is, when does the gathering of data and subsequent analyses everinW!y become informacion? For
the purposes of this chapter, the term data will mean some compilation of raw data into refined groups with the intent
of conveying information to the audience.
study? Are you looking to inform or reach consensus? Are you uying to entertain or share knowledge? Is this h ighly
technical information unfamiliar to the audience, or will the audience understand the terms being used wi thout
definitions? Asc maps or design drawings needed? These are some of the questions you need to ask yourself when
developing a report or presentation.
Public involvement is both mandated and necessary for many engineering projects. The designer should ~ow how
information will be wed; for example, small user group versus auditorium, and anticipated composition of the audience. A successful project will demand attention to these kinds of details (Kihl, 2008). Highly sensitive or controversial projects may require a public involvement officer to mediate any discussions to help achieve consensus on a
preferred design. In some instances, the legal department of an agency may be called in to present a legal opinion on
an issue or situation.
Kosslyn (2007) defines three goals for presentation material:
Goall: Connect with your audience.
Goal 2: Direct artd hold attention.
Goal 3: Promote understanding and memory.
These three goals seem straightforward at first glance, but there arc countless examples of violations in many presentation materials. A closer look at these goals is provided in Section 2.3.
Think about the appropriate ways to share data with your audience. What measures of effectiveness (MOE)
best relate to them? Engineers may expecr one rype, like average control delay, while rhe public may bener understand number of stops or reduction in average tr:~.vel speed or incrc:ased uavel time berween points A and B.
Sec Appendix D for additional details 6n this topic.
suu by silk, rather than sequentially, and especially not on separate pages.
3. Use smalJ multiple to facilitate comparistms. Show the same basic chart scve.ral times, with different data in each.
4. Show causality by linking varUtbks. For aample, do not just show vehicle miles cnvded (VMT) inc;reases
over cime; show what might be linked m them.
5. The world ~ seek to undnstami is mullivariate, and so should our displays. Again, VMf does not increase by
itsdf. what changes with it!because ofit? Show those as well in the same chart6. Int~grate word ami image. Almost all reporrs or presentations use both text and graphics. Make sur~ they
work together as closely as possible.
7. Content is luy. If your numbers are boring. you have the wrong numbers.
8. Don't throw out data, or "dequancify" by (for example) changing numbers to "yes" or "no.
These principles get at the heart of any presented data-the designer should spend the time developing appropriate
and compelling visual charts, graphs and figures to ensure audience comprehension.
em information, and so will not understand a message iftoo much information m ust he retained
or proctsud.
men
How
do we take these principles and apply them to create compcUing data displays? First, the designer n~ds to
know what kind of communication is relevant. Ifyour need is co communicare quantitative information, use graphs.
Charrs, diagrams, maps, ehocographs and clip arc are used in communicating qualitative informacion. Exhibit 3-1
shows the overall decision levels associated with choosing an appropriate display (adapted from Kosslyn, 2007) along
wich the intended P.urp?se(~).
Desired
Communication
Qu.amiative
Grapbs
Bar Graphs
(Vertical,
Horizon tal,
Saclced)
Maps
Qualitative
Photographs
and c1ipan
e-1
j
Approach
~ Rlghttums
-Through
~ Leftturns
Grouped bar graphs can be vertical, as shown, or horizontal. Vertical is the most common. If titles for bars an: used
and they are too long to fit above each bar, use a horizontal bar graph. In this gnph of traffic volumes at an intersection, the bars are shown with a third dimension into the paper. The reader naturally wants to discern what the third
dimension represents, and in this case will waste time uying to figure out chat there is no additional informacion.
Unless the third dimension represents a data dement, then present bars in two dimensions only.
The.designer chooses the shading, color, or pattern within each bar. For color presentations or reproductions, colored
bars are effective. However, Kosslyn (2007) emphasius char competing colors, such as red and blue, should not be
plated next to each other. Because of the way our eyes process the wavdengths of these colors, a blurred transition is
created where the colors join edges. The colors red and cream, or blue and cream, better complement each other and
make it easier for the reader to process the visual information. The opposite diagonal cross-hatching for right rums
and left rums in Exhibit 3.2 is acceptable, but using a different pattern for one of them might be more pleasing. Also,
as each grouping represents one approach to the intersection, it makes more sense to orient the turning counts as
left turns, through, and right rums, just as a driver would approach the intersection. Note: Uu ofcolor for shading or
iJmtifying sptcific dat41114J not trans/.au ~tt ifth~ 1114teri4/ wilt be "P~~d in blaclt and white. Test a printed wmpk
ofJI'Ur work for clarity bifrm using it in a public docummt.
In general, data groupings using up to four data series are easiest for the reader to track and follow (Kosslyn, 2007).
If the groupings total more than four series, consider breaking the data into rwo graphs. The reader should have no
problem understanding the data in this ahibit.
The horizontal lines at 50-interval points on the volume axis in Exhibit 3-2 hdp the reader see the relative values of
the tops of each bar. Ifspecific values are most important, then these should be labded directly above the bar. Do not
make the horizontal lines too chlck; they will become a distraction to comprehending the message sh.own by the bars.
Also, as discussed above, the third dimension is difficult to process vi.!ually.
Legends should be used when it is too complicated to place the information directly on the bars. Make sure the legend
is described in the same order as the bar3 Ocft to right in~~ or clockwise for a pie graph).
'"'"rntrrnr rTllf""'U"'f"
.,~,n ~niTinM
J
;
-~
350
250
300
200
1150
100
150
0
NB-ftrst
Approach
~ Right turns
c=J
Through
~ Left turns
Exhibit 3-3 shows the same traffic count informacion presented as a stacked bar graph. In this case, the designer was more
inrercmd in showing the differences in the roral approach volume than individual movement councs. Hence. the sta.eked
bar graph clearly shows the relative differences in roral volume. There is some additional information robe gathered from
the individual movement counts compared across each approach, bur the reader must really search out !his information,
moSt easily seen for right rums bea.wc each bar component starts at the same lc:vd on the horizontal axis.
Cera 3%
\
~
Pic grapbs such as Exhibit 3-4 are hdpful for showing relative proportions of items. Colored sections of the pie, versw
patterns, are useful in presentations, because patterns can become distracting: The reader must look at too many diagonal lines. In this graph, each pie slice is labded along with its proportion to the whole. This is more hdpful to the
reader than trying to match slices up with a legend. The 3D view of the pie gives an added dimension to the standard
20 view that is not associated with any content. One 6.nai point is that a portion of the pie has been exploded away
from the rest. It is hard to tell if this is trying to highlight the largest percentage of vehicle mix (uti.liry vehicles at 65
percent) or that the remaining 35 percent is comprised of four other vehicle types. Regardless, either an oral or wtit
ten aplanation of the graph should point out why a portion was exploded, as the reader is led ro believe the design
aspect is associated with some content.
Communicating Data to tli~ Public 29
35
30
25
Ill
~
~
20
.~
s...
u..
16
10
10,000 vehicles
5
0
1910
1920
1930
1940
1950
1980
1970
1980
1990
The line graph in Exhibit 3-5 combines multiple dimensions of data in one visual display. A long time line for the
X-axis indicates the designer is interested in looking at fatality rates (Y-axis) since the beginning of vehicle production.
Line graphs arc especially useful when data suggest a continuous element.
One line in the graph depicts the number of vehicles (I 0,000 vehicles), another depicts vehicle miles of uavd (I million VM1} and the third. line depicts population (100,000 population). Perhaps labeling the lines with per~ before
each label would aid viewer understanding of fatality rates per these three measures.
While these lines arc weful, it seems some other relationships arc missing or not c:xplicicly shown in the graph, for c:xamplc, the sharp decline in rate per population during WWII. Also, why is there a steady declining rrend in the lower
two lines when population rate is Huetuating from 1960 to 1985? Some of these related causes should be c:xplained.
60~
50
:a
Q.
40
:::>
30
-o4>
4>
Q.
(/)
20
10
0
50
Density, k (vpm)
100
150
XY plots like the one shown in Exhibit 3-6 are very helpful in discovering if a mathematical relationshi p exists between
rwo variables, in this case speed and density. A plot without a fitted line becomes a scatter plot. The firred line, often
a straight regression line, helps define the relationship of the data because an equation for the line can be determined.
With data points plotted, the lit of the line (R2) can be determined as well, providing a descriptor of how well the line
represents the data points. If the line equation and R2 value are irnpocrant for the reader, these should be shown on
the graph. One item missing is an explanation of what the triangles and squares represent. A legend is needed here
to explain the clliference in these two elements. Also, data points best show up using plus(+), open circle (0), solid
reiangle (A) and solid circle ( ) (Kosslyn, 2007). Certainly other symbols, coloring and line hashing can be used, but
care needs to be taken that the viewer can see the differences, especially if being viewed in an auditorium presentation,
or if printed in a black and white publication.
2o-40 D40-60
.o ..o
o-.~
ii--~ _! l
&:: ,, I 1:5opm
1~,
l l.:SOpm
u:s oam
10!SO;arn
J!
-.so.rn
&:SO.m
=
-': .
; I_!-=-
.---
~f .
'-'
<>;
-51
lili
C51
""''
4>i
='
g'1
C..i
.,!
..,
F=
ol
:X:i
-. -.:
!i
II
c
.,!:.;
!<!'
b
li
CQ
..:s
7:50 am
I I
=~
c;
,::...
0
:;:;
"'-l
10
I I
Cl
; .:5!
-_,;
..g!
f:ii
"""'
Exhibit 3-7 gives an example of a 3D graph of 6-hour pedestrian flows along.an arteri~ street (Schroeder ec al., 200 9).
In this case, the legend ac the top of the graph, just under the title, uses color to depict a range of pedestrian volulles
that correspond to plotted "contours" that add a third dimension to the graph. Often, these types of graphs are b e:st
viewed in color as lower values can be darker colors and higher values can be lighter colors which catch the eye quicker.
While there is a lot of information here, the value of this graph is :in having tht viewer quickly see that Pogue St. to
Logan Ct. is the section with high pedestrian flows, and that Hoine St. and Chamberlain St. have the highest peak
levels during lunch time (11:30 a.m.-1:30 p.m.). In addition co the dimensions oflocation, time period and pedestrian volume, the graph gives additional dimensions of road names, and the distinction between crosswalks (numbers)
and rnidblock crossing locations {letters). The exhibit would be improved if the mid-block locations were S(a)ed to
match the actual distance between intersections. However, this is one example where the Excel software package is
limited in its abilities.
;::31
.. 90.o-100.0
100.0
c8o.o-9o.o
90.0
70.080.0
80.0
::ii
::::.
!!.
GZ60.0 70.0
70.0
... 50.0-60.0
60.0
a4o.o-so.o
so.o
30.0-40.0
40.0
20.0 30.0
30.0
1010.0 -20.0
20.0
0.0-10.0
10.0
0 .0
nmelnt erval
10 11 12 13 14 1S 16 17
18
Sepnent Number
The graph in Exhibit 3-8 depict.S &ceway congestion (de~ity, Y-a.xis) for 12, 15-min. time imerva.Ls (right side, Z..
axis} along 18 freeway segments (X-axis), some segmenrs being basic segments, some on ramps, some off ramps and
some weaving sections. The view clearly shows the third dimension (time) going into the paper. Similar to the other
3D c:xhibit above, shaded ribbo~ are wed to emphasize different density interva.ls, where the vertical height at any
point tepresenrs the density in that segment at that point in time. These kinds ofp2phs are atrerndy weful for
viewing congestion along a highway section to determine bottleneck locations and the effects of queue buildup and
dissipation (queue discltarge) in both time and space. In this case, a density of 45 vehicles per mile per lane (veh/mi/
In) corresponds to capacity and the Levds of Service p threshold as defined by the Highway Capacity ManwJ (TRB,
2000}. lt is directly evident from the graph that congestion and queuing are present during the analysis period, but
the actual onset and extent of congestion would likely be better estimated in a Bat" contour plot like the one shown
in Exhibit 3-7.
1. 1
sidewalk:
parking; and
pc:desuian access or possible: adjacent path.
Construction plans would also include the:
cc:mc:rlinc: alignment;
drainage features;
storm water detention;
sediment conuol;
construction phasing and work zone: safety;
traffic co~uol (signs, signals and pavement markings);
vertical profile;
earthwork;
a North arrow; and
a scale.
The: small section of an engino:ring plan in Exhibit 3-9 shows many of the: forures d=ibed above. Note: that by show
ing information rdc:vant to the road realignment and new stormwacer collecrion system (for cxarnpk. station informacion,
catch basin nwnhc:rs, supcrdc:va.tion slope, c:tc.) along with uecs, the: drawing appears too cluttered for a public prescn12rion.
Again, dtcidc what informacion your audience: needs, and the: outcome you want at me end of me meeting or present:~rion.
This section of an acrual plan sheer is part of the consuuaion drawings and, as such, all the information is rdc:vant for: the
intended putpase. Color ~ of existing~ new features would be: appropriate for a drawing presented at a public
hearing. such as described in the nc:xt ahibiL
Communicatinq Data to the Public
33
...
0;
f&~~
~-- -~--~
Exhibit 3-10 is a computer rendering of a proposed streecscape redevelopment project along Hillsborough Sueet in Raleigh,
NC, USA. Note that decailed information about cenreriine alignment curvarure, drainage consuuction elements, etc.,
would clutter the drawing. This additional dccail is not needed given the intent md argec audience of the graphic. This
plan view drawing was created for a public presentation of the project; the entire layout required a 160-in. x 36-in. printing.
5. Use figures (graphs) and cables ro portray the most important results.
6. Gee to the point. .
34 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
7. These principles are amplified in the sections r.hat follow. Remember, good engineering must be accompanied by good writing. Bad writing can trump good research .
. See Appendix D for additional derails on this topic.
List of ligures
Summary or executive summary
Body of the report
List of references or bibliography
Appendix
Glossary
Index
This sequence follows normal conventions bur may be altered to tic a given siruacion or local preferences.
A report should be artractive to the reader. lts appearance should be neat, the tone authoritative and professional (but
not dry and overly formal) and the content accurate. It should be free of spelling and grammatical errors, which tend
to distract the reader from the content. Break lengthy seccions by using headings and subheadings liberally and by
inserting tables, figures, photographs, or lists to portray results or present data. Make the report a convenient size (for
example, 8.5 by 11 in.); a standard page size is easier and cheaper to reproduce. Where possible; design tables and
figures so they can be read without rurning the report sideways. However, rotated pages or even Z-folded inserts (11
by 17 in. pages that are folded to fit into an 8.5 by 11 in. report) may be necessary. Your audience needs to be able to
read figures and tables in the report; thus, they need to be adequatdy sized.
Font size and scyle are a personal preference. However, font size should be 11 or 12 point for the text portions. Using
Times New Roman or something similar is suitable for a wide range of readers. Further, one and one-~alfline spacing
is easier to read than single-line spacing. In general, double-line spacing does not add clariry.
methods used;
data collected (for example, rype, arnowu, when, where, etc.);
analyses performed;
findings;
conclusions; and
recommendations.
Other subjectS that may be appropriate for a tra..Osportation engineering study report include:
alternatives devdoped or camined;
sdecrion of alternatives, TCDs, or routes;
evaluation results;
cost analysis or financial impact;
\ I
\.
Detailed material that supports but is not essential to the body of the report should be placed in an appendix.
Appropriate appendix materials include
supporting data;
detailed explanations of methodologies or procedures;
derivations of formulas;
conversion factors;
lists of symbols;
data collection formats;
data collection protocols; and
checklists.
Appendices are an effective means of fully documenting and suppqrting the results of a study without cluttering the
body of the report. The main report should remain an independent and free-standing document that provides sufficient decill co understand the study objecrive, findings and recommendations. Appendices are strictly supplementary
and can be longer chan the succincr main report. In some cases, appendices may even be electronic or online documents that supplement the printed report. They are therefore readily availa.ble to the reader if needed, but resou.rces
and paper are not spent on a document few readers will use.
2. Limited scope
3. Personal presentation
Communitaling Data to the Public 37
Organize and rehearse the presentation for the established cime limit. The average person speaks at a rate of I 00 to 150
words per minute. (A reader can comprehend up to 600 words per minute, but a listener cannot). Talk at a pace that
holds the listener's attention. You can check your riming by practicing with audio, video, or an audience of coUeagues.
Video is the most effective. It gives the presenter the opportunity to see how he/she is coming across and how to improve his/her presentation skills. Simply speaking aloud helps to check your organization; however, when practicing
in an "empty room" people tend to speak much fasrer than when they're recording or speaking to a live audience.
Rehc-.usal will hdp you achieve a relaxed, at-ease posture, and a smooth, confident delivery.
Include only those pieces of information relevant to the presentation's purpose and what is appropriate for the intended audience. Data interesting co the engineer, but unnecessary for understanding and potentially confusing to
the audience should be simplified or left out of the presentation. Conversely, if the infonnation is necessary for the
presentation, make sure ic is clearly understandable and can be easily commurticated during the presentation.
4.1.2 Vuwd Aids tnul Hmulouts
Some presentations are more effective if the audience has a copy of the presentation or if there are maps or other
props used during the presentation. Examples include brochures or Syers on a design concept, plans for a project,
the presentation slides in handout format and samples of products, devices, or equipment needed co perform a job.
VISual aids and handouts should be developed and included if audience comprehension is co be improved, or if ic is
imponanc to have audience members leave with something presentation-related in hand.
Handouts are c:speciaUy useful if complicated or detailed graphs and charts are used. The frequent use of projectors
in presentations lends itself co "squeezing" a lot of content on the displayed slide. However, what may be clear on a
computer screen can be harder to process when projected. A handout on 8.5 by 11 in. or even 11 by 17 in. paper can
overcome the limitations of projected media. As Tufte (2007) points our, a well-designed handout or "super graphic"
can also keep the audience engaged as they arc actively involved in exploring the presentation content.
Prese.nters should not be afraid the audience might "read ahead" ifthey're given handours. On the contrary, a presenter
should feel a sense of reward and accomplishment if the audience is interacting with his or her material. A handout
stimularc:s curiosicy and ke~ps the audience engaged in the presentacion.
38 MANUAl OF TRANSPORTATION ENGINEERING STUDIES. 2ND EDITION
Exhibit 3-11 shows the existing view of the srreet lanes, cemer rum lane, sidewalk, ov<;.rhead urilicy lines, utility poles,
buildings and traffic signals.
Exhibit 3-12 is a computer-generated view of the new srreetscape with one lane, curbside parking, underground udlicy lines, a ne_w modern roundabout at the downstream intersection, no left turns at this lim intersection, a plan red
median and new light poles and decorative banners.
i
By using this kind of visualization, the audience can see what is being planned and is in a better position to react ~d
comment.
By adding a time dimension to the visualization, proposed projects can be presented in "recorded fly-through" vie-ws,
&om both an observer's and a driver's viewpoint. The following four images are screen shots from a ttaffic simulation
model of the proposed project and give a static sense of what this could look like.
The simulation and visualization in Exhibit 3-13 were created as pan of the Hillsborough St. streetscape project
introduced previously. Additional details on creating these kinds of By-through views are provided in Chapter 11
on simulation studies.
4.1.4 Inter4ctWe Wib Fetzturu
A new Web tool is emerging called data visualization (or visual analytics). This kind of Web applic;arion performs
queries on databases and then displays the ~ula, often in a graphical map. Because of the graphical nature of the
outcome displays, daCl visualization applications are often developed within a GIS program and architecture.
VJSUal analytics bas a strong data mining~ component and the expectation of the discovery of n.:w knowledge or
relationships within the data. For aample, suppose a traruportuion agency could query a county to locate high crash
rates associateci with specific criteria, such as speeding vehicles or alignment curvature. The agency could then target
resources for remedial measures or enforcement where the payback-reciuced crashes-would be most noticeable.
The agency could print out a map showing these locations for usc in meetings to discuss possible actions.
When creating a poster for a research conference or a public meeting, the presentation material should be printed in
large font and simplifi~ to the essentials. The viewers will be uncomfortable and become unen~ed if they must
scand roo close to the display or if they have to read long paragraphs. Graphics and tables should be large, and simple,
and self-explanatory. Alternatively, short captions can be used to explain the display or illusuate irs key points. Poster
presentations can be supplemented with handoutS, as discussed above, to provide additional information.
S.OSUMMARY
This chapter has explored some cognitive learning principles as they apply to developing graphs. Focus points for both
oral presen.tations and written reportS or displays outlined the key components for describing and conveying technical
information to the public. Some of the major poinrs discussed include:
know and connect with your target audience;
be judicious in selecting appropriate graphics for inclusion in a report or presentation;
pay attention to default gcaphic styles and question if the desired message is dear. If not, change or modify
the gcaphic style;
pay attention to default coloring or shading and question~ this adds or distraets from the message. Know if
the information will be produ~ in color or black and white and adjust accordingly;
direct the reader/viewer to the information of most importance in a graphic;
.i
I
promote understanding in all graphics with the intent of having the viewer remember specific information;
3D and 40 visualization techniques can enhance viewer understanding of choices for a complex project;
pi:'Cpare repo~ with the same intensity and thoughtfulness .as a presentation; and
In addition, using the Web can both inform the public and solicit irs fe~back on major projectS. The Web can also
provide a mobile environment for conducting queries on complex databases, resulting in seeing physical and topographical relationships that could not be cap~ in any other environment.
The reader is encourag~ to find additional details on these topics from the following publications.
6.0 REFERENCES
6. 1 Literature References
Editorial stafl' of th~ University of Chicago Peas. Tht Chicago Manual ofStyk, 13th ed. Chicago, IL: Univ~rsity of Chicago
Pr~.l982.
Hodg~s.
J. C., M. E. Whircen, W. B. Horn~r. S. S. Webb and R. K Miller. Harbrau Colkge Handbook, 11th ed. San Di~go,
!Ghl, M. and K. Kat. "Involving the Public in Project Evaluation-A Case Study in Arizona."ITE]oumal (F~bruary 2008):
18-23.
Kosslyn, S.M. CLEAR atui to THE POINT, New York: Oxford University Pr~. 2007.
Kroner, M. G., J. W. Presley and D. C. Rigg. Prtntiu-Ha/1 WOrkbook for Wrirm, 4th ed. Englewood Cliffs, NJ: Prentice-Hall,
1985.
R.obenson, H. D. and D. C. Nelson. Manual ofTnmportation Enginuring StutUes, Washingron, DC: Inscirure ofTranspomtion
Engineers, 1994.
Schroed~r, B.)., N. M. R.ouphail and B. A. Lehan. "Observational Srudy ofPedesuian Behavior Along a Signalized Urban
Corridor.~ 88th Annual Meeting ofthe Traruportation Research &ard. Washingron, DC, 2009.
Transportation Res=ch Board. NCHRP Project 08-36, Task 51: Primer on Information Daign for Effi~ DOT Decision-Making.
Washington, DC: Transportation Res=ch Board.2006.
Tufte, E. R. The Visual Display of~ntitlltive Information, 2nd ed. Cheshire, CT: Gr:apbics Press, 2007.
www.crb.org/AnnualMeeting2010/AM2010PanicipamGuides.aspL
"Creating Effective Poster Prescncarioru.~ North Carolina State University Web sire, www.ncsu.edu/project:lposter:s/NewSire.
Chapter 4
o o
o o o
o o
o o o
o o
o o
o o
o o o
Volume Studies
Original by:
INTRODUCTION
44
44
49
4.0
43
58
sa
62
67
67
68
69
70
5.0 SUMMARY
74
6.0 REFERENCES
74
1.0 INTRODUCTION
Engineus often~ counts of the number ofvehicles, bicycles, or pedestrians pusing a poinr, entering an intersection,
or using a particular facility such as a travd lane, crosswalk, or sidc:walk. Counts are usually samples ofactual vo]uro.es,
although continuous councing is increasingly performed for certain sitW~cions or circumstances. Modern automaced
count stations are found along signalized arterials and on freeway facilities, and are Standard features in combination
with weigh-in-motion (WIM) stations and automated toll facilities. Sampling periods may range from a few minu. -ces
to a month or more. Despite advances in modern data coUeccion technologies, the length of the sampling period :remains a function ofthe type of count being taken and the use co which the volume data will be put.
lp this chapter the focus is on the common methods for counting tra8ic in the field; how volume data are sampled,
expanded and analyz.ed; and how count prognms are established. Brief descriptions of specific studies are prese~t
ed along with references containing more dttail. The chapter distinguishes and contrasts manual versus auromaced
counts and real-time versus post-processed data, and discusses tradeoff's berween data accuracy and data coUection eficiency using examples. Special emphasis is given to data collection challenges at geometric configurations that reqll.i.re
path-based vehicle counts, including roundabouts, and ocher intersections (for example, superst.ttets), whert soO'le
or all traffic movements are not observed in isolation, bur at all times share lanes with ocher movements. Appendix:: E
presents sUmmary forms suitable for copying.
stop configurations are most common. In other countries, yield~ntrotled intersections are more common, where
me minor movement is concrotled by a YIELD sign. Many European countries further feature unsigned intersections,
where the right-band-rule governs traffic operations (drivers yield to me vehicle on their righr). Yield~ncroUed
roundabout intersections are addressed in a separate section below.
Before counting an unsignalized intersection it is important to understand the operations at the site, as well as me
anticipated volume levels. Generally, a four-way stop or right-hand-rule intersection has low-enough traffic volumes
that a single observer can comfortably count all movements, including heavy vehicles. Ar busier two-way nop or yieldcontrolled intersections, a second observer may be necessary, depending on me volume levels on the major approaches.
2.1.2 Sigrudiutl lnunectimu
At a signalized inrc.rscction, opposing movements arc conrroUed by signal phasing and therefore move at scheduled
and alternating times. Since all approaches do not have the right of way simultaneously, an observer may alternate
counting movements in rwo directions (for example, eastbound and southbound) as the signal phase changes.
Low-voluinc signalized intersections may therefore be counted by a single weU-traincd observer. However, since the
majority of volume studies arc performed during peak flow periods, it is likely that multiple observers are necessary
in most cases.
Counts at larger and busier signalized intersections are complicated because one or more movements occur during
each phase, because each signal cycle contains rwo or more phases and because the green time for one phase often is
not equal to char of other, opposing phases. In order to not bias the count toward any parti~ set of movements,
the count interval should be an even multiple of the signal cycle length. This effect is important to consider at short
rime intervals (and long cycle lengths), but tends to average out if longer count intervals are uscd.lt is good practice
to select count intervals that capture at least five cycles.
Actuated signals fUrther complicate counci,ng, because both the cycle lengths and the green times vary from cycle co
cycle. One rule of thumb is to select counting incc.rva!s that will include at least five cycles, using the maximum cycle
length to determine the interval. Assuming the sigll21 concroller is responding to demand,
counts should be representa~C of the demand, despite the variations in timing (Roess, Prassas and McShane 2004).
me
'.'. I
Special challenges at signalized intersections are permissive turning movements that do not move consistently during
their (permissive} green phase. Permissive right turns and right-turn-on-red movements are easy to miss among busy
through movements and special attention is required.
Tunc Period
Aaival VoltUDe
(Vehicles)
Volume Studies 45
For both of these nonmotorizcd modes of transportation it is important to understand behavioral patterns, since they
can affect the volume counts. Pedestrians frequently cross outside of marked crossw2lk areas and away from intersections. The observed pedestrian count at a crosswalk is therefore oftentimes less than the acrual pedestrian demand
volumes. Bicyclists may travel on the roadway with motorized traffic, or may decide to dismount and we the sidewalks for certain maneuvers. Chapter 12 describes procedures for performing pedestrian and bicycle volume counts.
2.1.4 PAth-Based Counts
Some modern intersection configurations require the analyst to perform path-based counts. Counts at traditional intersections (signalized or unsignalized) are typically done at a point at which the vehicle path is uniquely defined (for
aa.rnple, the stop bar for an exclusive lane at an intersection). However, several modern intersection configurations
combine multiple movements into one or more shared lanes, and the count i.s a function of both origin and destination of the vehicle, that is, the vehicle path. Eumples of these intersections include modern roundabouts, supcrstrcets
and Michigan U-Turn intersections. At these intersections, the volume from any one approach (origin) mixes with
other traffic (from other origins) before exiting at a common destination. As a result, individual turning movements
(for example, eastbound left tum) are never observed in isolation but are at all times mhed with one or more other
\
movements.
,~....,.~,.$'d.i\'~
.-:"71&..r..JJ::~:.JJ,..
....~~""!";
~~-~(!ll~
a)
~
~-t
i)oo
0
C:*
~
.............!..
% .........)-
......
lt>-
...
b)
C>>e '
at::)
......)o
...........
-==::J
The .sampling method is appliable to all intersections that require path-based counr:s. The 6dd observer lim estimates typical 0-D matrices with turning percentages for different approaches and different times of day. This can be
done froQI. video or in the field. The 0-D matrix is chen applied to the demand flows at the four approaches to the
intersection to estimate turning movementS.
An alternative way co perfOrm a path-based count ac a large intersection such as a supe.rsuecc is co use a li<Xll.SC-place
matching approach. Multiple observers can be positioned at the minor approaches and U-rum bays to record the last
three digirs ofvehicle licenses plates. These data can later be analyzed in the office and converted to an 0-0 matrix, or
full turning movement count. This approach is labor intensive, but is a reliable approach to obtain a detailed count at
these intersections. Modem licease plate recognition technology shows a lot of promise that may gready assist pathbased count applications in the future.
.'
~ ,
.. o,
-
.,
. :r.t1ft!:. ,
, ....,.
"'-"'. :~.--~
-~ ,
,
-. ~'
.
'
...
Approach
U-Turn
Left
North
U-Tum
15
Left
33
100%
0%
24%
48%
28%
ff7
100%
6
24
4%
14%
35%
47%
100%
Riglu
Total
U-Turn
Left
..
Through
Right
60
Total
171
81
~~'
~ ~.
~-
..
'
Mo~ment
Deman~
Demand
(~hides/hour)
(~hides/hour)
. 37
I ll
578
307
123
59
130
39
243
0
21
42
24
Left
~;
- '>--
578
!56
Through
~-
6%
14%
64%
16%
Right
'
.,.....
- ' - -~ -
100%
. Through
U-Tum
West
6%
19%
53%
21%
75
Total
South
9
27
30
141
Total
East
Percent
Right
_,,._~
:.:-:-;~t="'\ :.~~;lt
. --
Appi'!Jach
Count
Through
958
615
154
958
0
331
678
80
160
91
331
24
95
238
321
678
Volume Studies 49
Agencies use cordon counts moSt commonly as part of an 0-D survey as a basis for expanding imerview data. The
counts are taken in conjunction with the interviews. 0-D studies are described in Chapter 20. Cordon counts may
also be taken for trend analysis purposes. For this application, agencies count one weekday each year, during a month
with an average daily traffic (ADn that is dose to the annual ADT. The counts are made at the same time each year.
The cordon is established by following several guiding principles d.efinc:d by Roc:ss, Prassas and McShane (2004):
The: cordoned area must be large enough to define the full area of interest, yet small enough so that
aca.unulation estimates will be useful for par~g and ocher traffic planning pmpose.s.
The cordon is established to cross all sucets and highways at midblock locations, to avoid the complexity of
establishing whether turning vehicles are entering or leaving the cordoned area.
The cordon should be established to minimize the number of crossing points wherever p~ible. Natural or
manmade barriers (such as rivers, cailroads, limited access highways and similar features) can be used as part
of the cordon.
Cordon areas should have relatively uniform land use. Accumulation estimates arc used to eStimate ~tteet
capaciry and parking needs. Large cordons encompassing different land-use activities will not be focused
enough for these purposes.
The objective of :1. cordon count is to capture 90-95 percent ofADT entering the cordoned area. The most heavily traveled roads are observed for a full 24-hour period, less heavily traveled roads for l().. hour periods, md minot roads for
9-12 hours. Alleys and very low-volume streets may be ignored, if the aggregate loss is less than 3-4 pcrcenc of the toral.
Count stations on the cordon boundary are always located at midblock locations. Agencies can keep the number of
stations to a minimum by taking advantage of natural or human-made barriers. Counts should be made on the same
day. However, if the agency maintains a set of control stations, cordon counts made on different days can be adjusted
using the control station data (discussed later in this chapter). Short counts covering only peak-hour flows should
nor be used because the distribution of traffic at each crossing location is critical to the determination of accumulation. Counts of transit passengers should be available from the local transit agency or can be made using the methods
presented in Chapter 13. Exhibit 4-4 shows typicalfidd and summary sheets for a cordon count. Full-size versions of
these sheets for copying are available in Appendix E.
~~~~~~~~~~~~~
.
- ""
- ..
. --
, .. ---y:..,.,._. _ _...-, .
.~
~-
.:..p:.p~w -"--~
Traffic
Street
From
W eather
To
Numbers of
Passenger Vehicles
Including Taxicabs
Other Vehicle
Excl. Bus &; Str. Can
Trucks
----Recorder
Date
Pedestrians
Side of
Traffic
Street
Hours From
Weather
To
l.obound
Oulbound
- -- --
--Recorder
Da.te
Date
Street
Weather
Street
Street
Compiled by
Street
St
St
i/4Hour
Sw1lng
Side
Ia
0...
Sick
lo
0...
'IOoal
Ia
o..
Side
lo
o..
Side
lo
0...
St
Tooal
lo
o..
Side
Ia
Oal
Sick
Ia
St
Tooal
Ia
o..
Side
lo
Oat
Side
lo
o.,,
Total
lo Oac
Total
A--ae
..... }loot
~~1!--,~
Passenger Vehicles, Trucks and Misc:dlaneoos Vehicles Summary Sheet
Ihte
Traffic on
Weather
Compiled by
Sueer
Namlov lolooaad
1/4 Ho..,
.........&<'
Scardaa
Vehicla
Num.bu OutboUDd
ToQJ
ra-p
Truda
TraDJit
MJ.c
ToQJ
lD .....
Vebldes
'l'racb
Tramir
MJ.c
o...
ToQJ
Total
Average
Peak Hoa.r
'
All Penon SWIUIW'f Sheet
Date
Traffic on
Weather
Compiled by
J/4
Hoa.r
SanloJ
..........
v.blda
-..lahoaadv.
Tracb
Tamil
WUda
1\Wc
VCiolda
S=r
,.......o..m.....dV"oa
Toal
Wo.IJdac lahoood
PuocDp
Ve!Uda
Tta<b
y-a
~
.....
Hoat
Ttu.oia
VCiolda
Mloc
WUd.
Wolkios
Teal
TocaiiD
Owl>oa...t PluO..I
Vehicle accumulations within the cordon arc found by summing the entering and leaving counts at all COI,Int statiQns
by time interval. The counts usually begin when the street system is at its lowest B.ow. Agencies can count the number
of vehicles parked on-street and off-street to estimate the number of vehicles inside the cordon when the study begins.
'.Exhibit 4-5 illustrates the procedure for computing accumulation. Exhibit 4-6 illustrates a method of displaying cordon count data summaries.
A
LAgtnd
_ .......,.~
c:J.__
~
.......... ......
~
pr-..q AUMu.nt~el""""ar,an
eut
111
,,
2.2.3 Omtrol.Counts
Daily and seasonal (monthly) volume variation patterns are established and monitored using control counts in an
areawide program. Counts are made either continuously or petiodically throughout the year. The most useful counts
are made at pcrmancm-count stations, which operate 24 hours a day, 365 days a year. Control-count stations supplement the data obtained from permanenc-count stations to obtain estimates of seasonal and monthly volume variations
at addicionallocarions in the transportation network.
'.i
Control-count sracions are distributed across the cransporcation network and placed at strategic locations. They can
be divided into major and minor concrolcouar stations, where the major Stations are placed at key corridors in the
network. Typically, at major control-count srarions, a 7-day continuous count is performed during each month of
the year. At minor control-count stations, a 5-<iay weekday continuous count is made every other month. There are
:j
;i
I
generally twice as many minor as major conrrol-counr stations. Concrol counts in urban areas may be refe~red co as
key counts. For every 20 to 25 coverage srations (discussed below}, there should be a permanent- or concrol..:oum
station.
=Vcu,.,..,.t(l +g)"
Equation 4- 1
where
Typically, a running average of 5 or more years of traffic data is used from a control station; however, engineering
judgrnem must be used to make good decisions about the exact time period that should be used for the growth rate.
For instance, a small town that has rece.ncly begun heavily sprawling should consider using a smaller time frame to
calculate the growth rate if sprawl is expected to happen for many years co come.
To calculate the growth race from just two volume observations of volumes (current and future) chat were !aken n
years apart, the growth factor gcan be computed directly by rearranging Equation 4-1 (all rimes as defined b~fore) :
Equation 4-2
The computation ofseasonal or monthly variation factors follows a similar procedure. The ADT for each month is
the monthly volume from the permanenr-count mcion divided by the number of days in the month. The AADT is
then computed as the average of the 12 monthly ADTs. The monthly adjustment factors are obtained by dividing
each monthly ADT by the AADT. Exhibit 4-9 illusttates the computation of monthly variation factors. Daily and
seasonal factors can be computed in a similar way from control-count data. Since control counts are samples rather
than continuous counts, the ri:wgin for error is greater. However, carefully planned control counts will produce reli-
able estimates. For further discussion, see Roess, Pcasw and McShane (2004).
\ '
Equati on 4-3
where
The Highway CAptz&ity Ma~ual (HCM) offm typical values for K and 0 that can be used for anal)'3is. If an a.gency
desites to develop its own estimates of these proportions, a conuol-count station can be wed, provided that da.ta. are
available in hourly bins and separated by direction. The basic computational procedure mirrors the development of
daily and monthly factors described above.
HCM computations use the peak 15-rninute Bow rate expressed in vehicles per hour. It is obtained by dividing the
DDHVby th.e peak hour factor. (PHF). The HCM and many agencies offer default values for the PHF chat can be
used in the absence of detailed (15-iilln.). 6.dd data. If 15-rnin. volumes are available, the HCM methodologies directly we the highest 15-rninute flow rate (expressed in vehicles per hour). The PHF can be estimated from 15-rnin.
volume observations a.s follows:
PHF=~
Equation 4-4
4*V, .m
where:
PHF
V,....-
V, ..w
The peak how ranges from a theoretical minimum of 0.25 (all peak hour vehicles arrive in the peak 15 min.) to 1.0,
in which case the vehicles arc evenly distributed over the peak hour. ~mmended HCM default values arc 0.88 for
rural areas and 0.92 for urban area.s (TRB, 2000) .
. 2.2.4 Conr.p Collllb
A coverage count is a relatively short-term but continuow count performed at one location over a period of 24 to 72
hours. To adjust a 24-hour coverage count for a given location to an estimate of AADT, the count is multiplied by
the appropriate daily and monthly variation factors obtained from conuol-count stations. The factors wed should
be ~m a permanent- or conuol-count station !ocation similar in geometry and t:raflic characreriscics co the location
of the coverage count. Guidance on how to match up the: appropriate factors for a site are given in the literature (for
cumple, Min-Tang ct. al, 2006).
Volume Studies 57
3.1.2 Equipment
3.1.2.1 Taliy Shun
The traditionally simplest approach foe conducring manual counts is co record each observed vehicle with a tick mark
on a prepared field form. Exhibit 4-10 shows a field sheet for a vehicle turning movement count. Pedesuian '!lld bicycle counts may require separate sheets if volumes of those modes are high. For many in[crsections low pedestrian
and bicycle movements can be added manually on the existing form. The form allows for whatever classifications may
be desired. A watch or stop~tch is required ro cue the observer to the desired count interval, and a new form is used
at the stan of each interval. The raw counts are tallied, summarized, or keyed into a computer upon return O the office. This method is low-cost and is easily adaprable to different geometries and count types. However, its application
is less common today with the availability of clecaonic count boards and laptop computers, which are convenient to
use and reduce analysis time.
to
Olote
NIS S tr. .t
E IW Str . . t
. p -
Dy
weath er
Obaerver
~
p
. IT
I
I
.,
I
,.
I
I
--
I
I
I
I
I
IT
........
..1.
I
I
I
I
I
I
I
- .
IT
- - -I T
IT
jp
I
I
I
I
I
I
IT
II
lp_
;-.:
I
!T
I
I
I
I
I
I
;'
~
I
I
.I
I
I
I
I
..
!:!
::
}
)
i
3.1.2.3 Laptop ComputnA battery-efficient laptop computer can be substituted for a handheld count board in many applications. Oftentimes,
laptop computers ate already available to the analyst, making this an attractive alternative to new hatdware. Commonly available spreadsheet software can be used to record time stamps of different types of events using a macro
routine. Appendix E contains discussion of how to code a time-stamp macro in Microsoft Excel~ and VISual Basic~.
The benefit of using a macroenabled spreadsheet to collecr volumes (or other temporal event data) is that it can be
customized to the specific needs of the user. Also, many commercially available count boards are ~cted to output
aggregated data.. Alternatively, this approach allows the analyst to obtain individual time-stamped events. The downside is that some software coding and post-processing analysis are required. Consequently, this approach may be more
interesting in research settings and other special applications.
:;
;i;
'!
;,;
:~~,,
Manual counts can be performed in real time in the field or in a post-processing operation from video observations
in the office. The use of videos may be practical if video is already available from a previous effott, or if it is easy to
obtain. For example, modern traffic management centers often have live video feeds from permanent field cameras (on
freeways or signalized intersections) to a central office, where they can be recorded. For manual freeway counts this
is frequently a safer approach than positioning an observer on a bridge srruccure or other location in close proximity
to fast-moving traffic. Video-based counts can also min.i.tnizc staff requirements, if the same analyst can replay the
video to count different movements. A well-chosen camera angle, ideally from some overhead vantage point, is critical to cnswe the video is useful. With adequate light conditions and a good vantage point, one camera can capcure
all curning movements at a typical intersection. A digital clock in the video image is an excellent way to note the end
of intervals. However, notations by the camera operator or the camera operator speaking into the microphone arc
a,dequate substitutes for the clock.
Modem video-image processing software (discussed below) may auromatically provide volume data. Alternatively,
observers can record their counts with a handhdd count board, tick macks on a tally sheet, or directly into a computer.
For most applications, this way of obtaining manual counts from video effectively doubles the analysis, making it
Jess practical if video observaciom aren't already available. One benefit of video is that the observations can be easily
error-checked by a second analyst or can be harvested for additional data op vehicle dassificacio.n, dday, or queues.
If necessary, observers can view the video in slow motion for additional scrutiny. The ability to conduce other stud. ies with video, including the intersection and driveway studies discussed in Chapter 6, can lead to very efficient uses
of the labor of fidd crews. Usually, other studies require a higher-quality video than a turning movement count, S<?
engineers should consider the needs of the other studies first when specifying video use.
If no pre-existing video exists, and unless additional data dements arc exuacted from the video, it is usually i nefficient
to usc video for a simple volume study. The costs of video equipment, additional set-up time and the need to keep up
with batteries and recording media are unnecessary extra steps for obtaining traffic volumes in many c:a.ses. If video
equipment is set up ac an intersection, the data collector typically needs to stay dose to the equipment co prevent
vandalism or theft. In this case, it is more time efficient to simply perform the volume study at that time, rather than
investing additional time copying and post-processing the video.
3.1..3 Perromui.Requ~J
Manual traffic counting requires trained observers. They must be relieved periodically to avoid fatigue and degraded
performance. Breaks of 10 to 15 min. should be sclleduled at lca.st every 2 hours. If the data collection perio d is ~ore
than 8 hours, breaks of 30 to 45 min. should be allowed every 4 hours. The size c;>f the data coUcction ccam depends
on the length of the counting period, the type of count being performed, the number of lanes or crosswalks being
observed and the volume level of traffic. One observer can easily count turning movements at a four-way, low-volwne,
signalized intersection with one-lane approaches as long as special classifications and/or vehicle occupancy are not
required. As any or all of the foregoing variables increase, the complexity of the counting cask increases and additional
observers will be needed.
Duties may be divided among observers in various ways. At a signalized intersection, one observer may record the
north and west approaches while the ot:h~ watches the south and cast approaches. In that way, only one ap proach is
moving for each observer at any given time. Another way to divide duties is for one observer to record occupancy or
certain classes of vehicles, while the other observer counts total volumes. At complex sites, individual lanes, crosswalks,
or classifications may be assigned to individual observers. Also at complex sites, one observer may have the sole job of
rdieving me other observers on a rotating schedule basis.
3.1.4 Field Procedure
3.1.4.1 PrtparatifJn
An accurate and reliable manual traffic count begins in the office. A locally devdoped checklist is a valuable aid, even
to experienced teams, co ensure that all preparations for the field study have been completed before the team arrives
at the site. Exhibit 1-1 is a general checklist that can serve as the beginning of a locally devdoped checklist. Pre~
tions should start with a review of the purpose and type of count to be performed, the count period and time intervals
required, and any information known about the site (such as geomeuic layout, volwne levels by time of day, signat
timing, etc.):This information will hdp determine the type of eq~pment to be used, the fidd procedures to follow
and the number of observers required. Online mapping and visualization tools may hdp identify good vantage points,
but local knowledge or a site visit to the location is usually necessary.
The selection of equipment will dictate the types of data forms needed, if any. Header information should be filled in
to the extent possible in the office, and the forms should be arranged in the order in which they will be used by each
observer in the fidd. When using handhdd count boards it is very important that a naming convention for streets and
orientation be agreed upon beforehand. Without such convention, it can be difficult to matcll up multiple output
6Jes in the analysis step. The checklist should also include equipment items, such as pens, batteries, stopwatchd"and
blank media (video tapes, discs, Hash media, or hard drives), as appropriate. Having to rerum to the office to retrieve
forgotten itemS may delay the start of the st1;1dy or cause it to be postponed. An inadequate number of forms could
also invalidate the study and waste resources.
Equipment must opcrare properly to ensure accurate counting. Good counting boards have firm keys that provide the
observer with tactile and audible confirmation when a key has been pressed successfully. Units with "soft~ keys should
be repaired or discarded. The lack of tactile or audible confirmation when a button has been pushed poses a challenge
when using laptop computers over long count periods. A commercially available count board may thus be prefera~lc
for long coi.wts. An office review of the procedures to be foUowcd and a check of the proper operation ofall c;quip111ent
complete the preparation stage of the study.
Volume Studies 61
Modem technologies for automated counts can principally be divided into on-road technology and roadside technology. On-road technology includes pneumatic tubes, piezoelectric scrips and various forms of magnetic inductance
technology. Roadside technology can utiliz.c video, radar, infrared, or lasc.r technology. Roadside technology can also
be combined with in-W!hick technology which typically takes the form of deccronic roll transponders that can communicate with roadside readers.
62 MANUAL Of TRANSPORTATION ENGINEERING STUDIES. 2ND EDinON
Automatic count technologies are frequently used at intersections, where chey provide turning movement councs.
. They are also increasingly applied ro freeway segments and tolled facilities to provide lane-by-lane infor mation on vol. umes and other data. Generally, all of the above technologies can automatically count vehicle classifications, and may
in some cases provide additional output on vehicle speeds, headways, density and even travel time from one count
location co the next by using vehicle identification technology (for example from toll transponders).
3.2.2 Equipment
Boch in-road and roadside technologies typically consist of rwo basic components: a data recorder, and sensors to
detect the presence of vehicles and/or pedestrians. Some equipment also has the capability of communi cating the collected data to a cenual facility for processing.
Space does not permit a.discussion of che wide range of equipment available today. Besides, advancing technology
is causing continuing change chat would quickly render such a discussion obsolete. This therefore fo cuses on basic
characteristics of in-road .and roadside technologies, and tradeoff's co consider.
Permanent traffic monitoring s~tions 3!e common for both signalized nerworks and freeway facilities. In signalized
n~rworks, the vehicle detection technology (most commonly magnetic inductance loops or video detection technology) is routinely used to get traffic volwne data. Often, special system loops are installed for the sole purpose of trafii~
monicoring and data collection. On freeways, in-road detection technology is used to collect traffic volumes and other
traffic parameters. More information is provided in Chaptet 10. For permanent data collection equipment, durability
and reliability are of central importance.
In addition to permanendy installed equipment, me analyst has me choice between several portable or temporar}'
options for data collection equipment. Most frequently, agencies use pneumatic rubes or magnetic inductance technology mounted directly on the pavement. Data are stored in built-in memory and can be downloaded via USB con
nection, or even wireless transmission. Battery life in modern equipment allows chese devices to record several weeks
of data on a single charge, although applications this long are rare. Exhibit 4-12 shows a sdection of modem in-ro~d
data collection equipment.
Volume Studies 63
d) Sensys Detectors
:~
Source: Images by a) Jamar Technologies Inc., b) City of Richmond, CA, c) Vaisala Inc., d) 2009 Sensys Networks Inc.
; ,
~:~
The aforementioned in-road technologies are limited in their ability to count pedestrians and bicycles. Further, special
care and attention is needed when installing and removing these devices from lanes of moving traffic. Depending on
the intersection configuration, these technologies may require a lot of equipment to capture lane-by-lane data. The
additional cost and set-up time make these devices more applicable for longer duration counts. For short-term counts
(up to 8-12 hours), a manual count is typically more cost-effective.
~:.
::
~~
:t!
.j(
;!.
}'
:~
Just as with in-road equipment, roadside count technologies can be permanent installations or can take the form of
portable or temporary equipment. Permanent installations include video detection systems at signalized intersections. The use of permanent equipment based on other technologies (radar, microwave, etc.) is more common at
freeway applications and tolled facilities. More detail on this is provided in Chapter 10. Portable roadside equipmentis available, but used less frequently than in-road equipment. Exhibit 4-13 shows a selection of modern
roadside data collection equipment.
'.
::1
l
\
\ ,
-I
J
!
'
Videe>-based automated counts are u.sed at signalized intersections, where video detection cameras can double as data
collection technology. The positions of these cameras are typically st:a.tic and the c:a.meras are c:a.librar:ed to record diffen;nt movements on one approach of the intersection. Because of sight-angle restrictions, multiple c:a.meras :a.re used
if more than one :a.pproach to the intersection is evaluated. Once the video detection cainera is installed, the an:a.lyst
predefines virtual daecton using computer softwa.re at locations on the video image where movements :a.re to be recorded. The data aggregation interval is :a.lso a software input. Once the vide<\ image is c:a.librared and configured, data
are collected and aggregared in the specified time intervals. In statc-of-the-:a.rt signal,systems, individu:a.l intersections
are interconnected with each other and a central proccsser in a traffic managemenc center (TMC), where volumes and
other dat:a. arc being stored and an:a.lyz.ed.
Video-image processing can also be applied to offiine video recor~ in the office. The ch:a.llenge for offiine video
analysis is that the video detection softwa.re needs precise information of the c:a.mera loca.tion to a.cau:a.tdy process the
video. In the c:a.libcation step, the an:a.lyst has to enter the camera height and rdative distance to a known reference
point on the video. This process is known as orthortctifoJZtion or imagt ca/ilnution. Without calibration inform:a.tion,
the video typically c:a.nnot be used for offiine image processing. Due to the levd of effort involved, oflline autom:a.ted
video counts are likdy to be a highly inefficient way to collect simple count data. The set-up effort may be justified if
:a.ddition:a.l data are collected or if the goal is to obtain path-b:a.sed countS.
need
In addition to the
for c:a.librarion, all video-b:a.sed detection. and count technology are susceptible to camera
movement'(wind), lighting changes (daylight, clouds) and occlusion by call vertical objects (trucks). The reader is
referred to documentation provided by the equipment manufacturer to overcome these issues.
3.2.3 Pn'$0tmel Required
The only pcrsonnd required for :a.utomatic countS are those needed to inscall and recover the equipment. Crew siz.cs of
two or three are usually sufficient to deploy most portable counting equipment. Depending on the type of equipment,
the inscall:a.tion crew may have to be in the a:a.vded ~y and it may be preferable and s:a.kst to temporarily close Lanes
or inscall equipment during periods oflow traffic. The recording component can be handled by one person; however,
one or rwo persons will be needed to install ro:a.d rubes or magnetic sensors, while an addition:a.l person w:atches for
traffic. Roc.overy of the equipment can usually be performed by one or two persons. The inscallariofl of permanent
counters with in-pavement sensors rn:a.y require a l:a.rger crew and the closure of travd lanes.
Volume Studies 65
.
.!
,.
Do not place sensors across marked or unmarked parking lanes, where a parked vehicle could activate the
sensor continuously.
Deploy sensors at right angles to the traffic How.
For directional counts, keep sufficient space between the sensor and the centerline of the roadway.
At intersections or near driveways, place sensors where double counting of corning vehicles can be :~:voided.
Record sensor placement by noting the physical location on a condition diagram sketch.
Use a test vehicle to ensure that bidirectional counters are recording the proper direction.
Avoid locations where frequent queuing occurs.
Set the count interval to ensure that totals will occur on the hour to make the data compatible with other
counts.
Note the time that counter operation begins.
Additional guldeJ4les for in-road counters include:
Avoid placing sensors on pavement expansion joints, sharp pavement edges, or curves.
Fasten the sensor securely to the pavement with nails, clamps, tape and/or adhesives made specifically for
this purpose. Loose sensors will prevent the collection of data and may pose a hazard to motorists and
pedestrians.
Locate the count recorder near a sign pOSt or tree and secure it with a lock and chain, or place it in a locked
signal control cabinet to prevem vandalism.
Keep the cable or cube that conneCts the sensor to the recorder as short as possible.
Check the installation periodically to ensure it is in place and functioning properly. In cold-cl.imate areas,
agenCies should check sensors whenever it snows to ensure snow plows have not removed the sensors from
the road.
3.2.4.3lmtallationand Retrieval
The primary concern during installation and retrieval operations is the safety of the field crew. The crew's vehicle
should be clearly visible to traffic and should be parked away from the travded way. All crew members should wear
reflective clothing at all times. Deployments and recoveries should be accomplished during periods of low traffic volume and good visibility. If nighttime operations are necessary, the crew should employ extra safety measures (such as
lights, cones, warning signs). Anytime that crew members must enter the roadway, at least one crew member should
have the sole duty of watching for traffic and warning the rest of the crew. Details of installation techniques vary and
in many cases are product-specific. Information of this nature is generally available from the product manufacturer. In
some eases, police assistance may be required to ensure the safety of the crew and the public.
66 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
_J)~
n
StiniOil
LU 12:11
PM 8:00
110.:1.:..:.:15;:.___ __
110..:"'=._ __
WAIIMrCoDtlllloM
AN~
w~
PM~
0
c-ar<>
- - -- -
tt
:l'
il'
City of Raleigll
Pullllc Worb
"i
J,,
ii
:
:~
Automatic data collection equipment typically records data on built-in or removable memory. Depending on the
product, cbe data need to be downloade.d in the office or may be mnsfc.md directly over modem, fiber, or wireless
connection. As with modern count boards, automatic data recorders typically come with analym software that will
automati.cally reduce and tabulate data at a user-specified aggregation inccrval.
;.,
:;
...
;i":
.,,
;;
ll
lit
.:.
'~
.}
ri
,:P.
This computer software accomplishes the reductions and produces summaries of the data and desired calculated
values (such as ADT, peak-hour factors and percent turns). These tools save analysrs time and often eliminate errors.
\J
'I
\l
for a given location depends on the planned use of me data and me methods available
'for collecting the data. The count periods should be representative of me time of day, day of week, or month of year
that is of interest in the study. For example, Mondays and Fridays are usually not typical weekdays. Engineers and
planners rarely need turning movement counts, vehicle classifications, or pedestrian counts from nights, Sundays, or
holidays. Saturday counts are sometimes needed for shopping areas. The count period should avoid special events and
adverse weather unless the purpose is to study such phenomena. Count periods may range from an hour t o a year.
Manual counts are usually for periods less than 1 day. Typical count periods for curnlng movements, sample counts;
vehicle dassHicarions, pedestrians and bicycles include
2 hours; peak period
Short counts may be expanded by use of a control station. Ifa number ofsample counts are needed in a relatively small
area, analysts sdect one location representative of the area streets to be sampled. It is important that !he control seacion service me same type of Street, and variations of traffic being sampled, on me Orner streetS. The COntrol Station
is counted continuously during the entice sampling period using the same count interval (for example, 15 tnln.) as
on the sampled streets. The counts taken at the sampling location are called covmzgt counts. Bod! the coverage and
control counts are taken at midblock to avoid the complexity of~ movements. Each link or street segment to
be sampled should be counted at least once during the sampling period. The counts may be made manually or 'ith
automatic counters.
The control-count data establishes me volume variation partern for the entire sampling period. The partern is quantified by calculating, for the control-count data, me proportion of the tocal sampling period volume occurring during each count interval. Assuming that this pattern applies to all of the sampled locations in the study area, the full
sampling period volume for a coverage-count location is obtained by dividing the sample count by the control-count
proportion for the corresponding count interval. Exhibit 4-16 shows an illusttation of me procedure. In the example,
a control station was used and four coverage counts were made over a 2-hour sampling period. The count interval was
15 min. beginning at each quarter hour. Roess, Prassas and McShane (2004) provide a more extensive ~ion of
this count expansion procedure with examples.
Volume Studies 69
fl'
.. ---.... - ;'~W~
~ .. ~ . ~- j./."
-~~!:f'~lt~1f~~~~~t'"~':~1~-~t""
.... ~,~~tt~1!e~~,~~~
~ . . ' '!. ~ ~ -~
! ,
. .- :,.,....~~
~-- ,,::~ .~_:.-,..~~- ...... ~. ~~tl-:: ~~- s .,... : .. 'f ~~;~~!/ .... ~,.~-!.~t-v ... -.
Control Station Data
Cou.nt (vehicles)
7:00-7:1SAM
720
72016,864 .. 0.105
7:15-7:30AM
nG
776/6,864 a 0.113
7:30-7:45AM
837
837/6,864 ~ 0.!22
7:45-8:00AM
95 1
951/6,864 = 0.138
8:00-8:15AM
1,022
1,022/6,864 . 0.149
8:15-8:30AM
986
986/6,864 = 0.144
8:30-8:45AM
874
874/6,864 =0.127
8:45-9:00AM
698
698/6,864 0.102
Tune
Count (vehicles)
7: 15-7:30AM
784
784/0.113. 6.938
7:45-8:00AM
1,192
1,19210.138 =8!638
8:15-8:30AM
863
863/0.144 . 5,993
8:45-9:00AM
532
532/0.102 = 5,216
Location
~-
Dr.
~'r!Z. . ~.--
~ 1)0
"' 160
~
j
140
120
130
~0
JIO
'o
120
100
90
110
l.
": 100
90
:>'
80
?0
.s
.Dt.ain St JWirold St
l lO
I
I
I I
I I
~F.l~f~~"'g:lJr.>!:l
- ~ ~ < ~ - ~ < ~ ~ ~ 5
80
70
r--
rr--
f-
r-----
~
-
I ro
":' 0
l!
!!i
...
E
~
f-
f-
r--
t-
f-
,_
t-
fl
~
~' ~
.,~
l>o.IolWotk
MH.IholYeoc
Mool.bl11'rallk Varildou
9.0 8.0 _
7.0
t--
f-
Dalli1nllic Variatloos
DMainSt
Railroad St
6.0
s.o
4.0
3.0
2.0
1.0
0.0
..
n.
~
- ...
~a ..
.... ~ ~ ~
'T
,.:.
:t
... ;;;"'
-6 - U..Ptriod
.
....
.. ..
~~
Howty1nllic Volunats
An intersection Bow diagram (Exhibit 4-18) provides a similar picture. The width of the flow band indicates the level
of traffic volume. Traffic flows may be depicted in the same way on a route map (Exhibit 4-19). The thlckness of tb.e
bands provide,s a visualization of the relative volumes found on thesn:eets of a Highway network. Another common
means of presenting volume data for an area or system of streets is a traffic count map. Exhibit 4-20 shows a nup wich
total volumes in vehicles per day (vpd) in both directions for a city street system. These traffic count maps are typically
maintained by and available through the state department of cranspottacion.
29
~~
I
;\.
'Ii .
,,
;
:r1:
seleCt riinet .
. To . .j 05:45 PM
.
:......~.:_ -
2~
-~:. .
'.
::J
. .,.
. ~
1.. t!~ 1
CITY OF RENTON
2008
TRAFFIC FLOW MAP
DO'M1l4)1T I1F "--/IMUllll/I'UIIUC MR<S
lRAIISI'OifTATIQN smEllS llNSIClH
CII'OA1lCIICS !iECIICJI
8CAI: ,.
NOT:
IKTEIISTA'tt -
~000
IEHilD
~ I1F 1-4011}
:i
S.OSUMMARY
In this chapter the various methods used to perform traffic volume councs have been described. Issues rdated to data
reduction and analysis and specific types of counting studies have been discussed. For further information and more
details on traffic volume studies, refer to Roess, Prassas and McShane (2004).
6.0 REFERENCES
Banks, J. H. Eval=rWn ofPorusbk Dara GJiltion Thno!IJgit:s: Fmal &pon. San Diego, CA University of California-Berkdcy;
Partners for Advanced Transit and Highways (PATH), 2008.
Bonsall, P. W., F. Gbari-Saremi, M. R. Tight and N. W. Marier. The Performance of Handheld Data-Capture Devices in Traffic
and Transport Surveys, Traffi~ Engineering and Contro/29, No. 1 (1988).
Box, P. C. and J. C. Oppenlander. Manual ofTraffi~ Engineering Studia, 4th ed. Washington, DC: Institute ofTransponacion
Engineers, 1976: 13.
Dixon, M. P.-R. "Fidd Evaluation ofRoun<bbout Twning Movement Estimation Procedures." ASCEjoumal ofTransporr.atiqn
Engineering 133, No. 2.
Federal Highway Administration. Virginia State Route 1 V"uko Deuction ~stem hrfoT7711t11u Assessmmt. FHWA. Washington,
DC,_2006.
Garber, N.J. and LA. Hoe!. Traf!U & Highway Engineering. Pacific Grove, CA Brooks/Cole, 2002.
Kyre, M. and M. Lines. "Development ofVideo-B:ued and Other Automated Traffic Data. CoUection Methods. '' Traffic Track(r
Manual. Retrieved April2010. National ln.stitute for Advanced Transpormion Technology, Universicyofldaho, Moscow, ID2004. http://www.websl.uidaho.edu/nian/ruearch/Final_Reports/KLK203_N04-02.pdf.
Li.sr. G. F. Identifying Vehicle T rajectorie.s and Turning Movements ar R.oundabqucs." 5siJ International Sympoium on Highway
CapariiJ and Quality ofSn-ttu~. Yokohama, Japan, Japan Society ofTraffic Engineers (2006): 44_9-458.
Min-Tang. L F.-F. "A.s.signmeoc ofSe:uonal Factor Categories to Urban Coverage Counr Srations Using a Fuzzy Decision Tree:
ASCE]ourruzi ofTransportarion Enginuring, 132, No. 8: 654-662.
RobertsOn, H. D. and D. C. Nelson. Manual ofTrr:nsportarion Enginuring Studi~, W:uhiogton, DC: ITE, 1994.
R.odegerdts, L. B. c!. al. NCHRP Report 572: Roundabouts in the United States. Washington, DC: National Coop erative
Highway Resea.rch Program, 2007.
Roe.ss, R. P., E. S. Prassas and W. R. McShane, Traffic Enginming, 3rd eci Upper Saddle River, NJ: Pearson Pren tice Hall, 2004.
Sm:~.d.i, A. B. "Advantages of Using lnnova~ve Traffic Data Collection Techniques." Applkations ofAduanud TechnobJgi11 in
Transportation. The Ninth Inrernational Conference. Re5ron, VA; ASCE Publications, 2006.
Tarlw, A. and R. Lyles. ~/opmmt ofa Portabk Vuleo Dmcrion Syffnn for Caunring Turning Vthicla tU lntmmions.
Purd~
Universicy Repon FHWNTNIJTRP-2001/18. hnp://doc:s.lib.pwdue,eduljap/62/. West Lafayette, IN: Purdue Universicy, 2002.
Transporration R=ach Board. "Data, Survey Methods, Traffic Monitoring. and A.s.set Mmagernent. Transpor111rion &rarch
&cord:jourruzi ofthe Transporrmion &much Board 1993 (2007).
Zou, N. and J. Wang. "A Video-Based Method for Evaluacing Traffic Dara from De rectors." App/.Wltions ofAdvanced
Ttchnologi6t in Transportation. The Ninth International Conference. Reston, VA; ASCE Publications (2006): 232-237.
Chapter 5
....................................................................................
j
2.0
INTRODUCT19N
77
1.1 Safety
78
78
79
79
2. 1 Introduction
79
79
84
86
3.1 Introduction
86
86
87
88
88
91
92
S.O
SUMMARY
95
6.0
REFERENCES
95
1.0 INTRODUCTION
Speed is an important measure for tta.ffic operations, because highway users relate speed to economics, safety, time,
comfort and convc:n.iencc. Speed is a basic measure of tta.ffic perfOrmance. Thus spot speed data have a number of
applications, including detennin.i.ng tta.ffic operation and control parameters, establishing highway design demencs,
analyzing highway capacity, assessing highway safety, monitoring speed trends and measuring effectiveness of controls
or programs. Spot speed studies are designed to measure speeds at specific locations under the traffic and envirorunental conditions pteVailing at the time of the srudy.
This chapter is divided into two fundamentally d.i.lferent approaches to collecting vehicle speeds at spot locations. The
first is the individwJ whick ~!mUm method where a subset of vehicles in the mHic stream are sampkd using predominatdy manual speed measurement techrUques. Altem2tivdy, the all-vehicle sampling method records (almost) aJJ whick
spmls using automated in-road or roadside measurement equipment. While the first method is wgeted to short-t~
speed measurcmencs, the second is appropriate for system performance monitoring systems that rdy on. continuous estimation. lbis chapttc presents study design, methods and analysis techniques for both methods, following a discussion
ofsome geoaa.l speed study concepts.
~nnt
C:noor4 <:t,.4io<-
,,
1.1 Safety
As with all fidd studies, safety is the paramount consideration in conducting spot speed studies. The measurement of
speed involves having workers in the proximity of the roadway, whether it is ro install and recover in-road detection
devices or to operate roadside data collection equipment. Workers must use care and vigilance at all times while: working near the roadway. Workers should park their vehicles off the traveled way. wear reflective vestS and act in a manncr that does not disuact motorists or influence their driving behavior. Workers should place and recover detection
devices in the roadway under low-volume conditions. Standard lane closure procedures and warning devices should
be used if extended time in the roadway is required. Study teams may need police assistance co direct traffic during the
deployment and recovery of equipment.
Approach
Lm
. .
~
Typical
Eqaipment
.. .
'~.
~~i~~,'
\~'t~~~~~{>3f.
.' " lF.lfli
--~ ._:/~ !_'i_'1~.-.-id.jt!.
-~.-,
. ~-
<.. ; ..
TMS
.Jll
Variabla
Spot
Location
Laser!R.adar,
Short Speed
Trap
v1 Individual Speed
n II of vehicles
Segment
Stopwatch,
GPS,
Overhead
Video
TMS
SMS
2)
Ll'.o(v,)
n
Spot -
.I:r. n(!f)
1
= y----r
iiL~ v-
_(4)
_13
TMSseg
SMS.rpoc
.I:r cd,)
nL
SMS,.g = .I:f(tt) = tr. o(t,)
Exhibit 5-1 shows four potencial ways to calculate speed, labeled (1) through (4). For the purpose of chis chapter, (l)
is the m~t common application, as individual spot speed measurements are averaged to obtain TMS. However, for
some applications, including freeway and arterial segment operations, SMS is needed and equation (4) is appropriate. In the absence of segment speed or cravd time data, equation (2) can be used to estimare SMS from spot speed
measurements (Roess, McShane and Prassas, 2004). The remaining equation (3) is rarely used, since it is unrypical to
use TMS in a ,segment evaluation.
Since chis chapter focilses on spot speeds, the resulting measure is typically the spot time-mean speed of traffic (Equation 1). This measure is appropriate: to sample approach speeds to a signalized intersection, to measure the speeds in a
horizontal curve, or to quantify che dfect of a traffic-al.ming treatment on vehicle speeds at that location. However,
it is not an appropriate measure to use in segrnenr-based speed evaluation related to segment travel times (Chapter
9) <?r freeway studies (Chapter 10). Freeway performance is described by the speed-flow-density relationship, which
should always use SMS.
Additional discussion of the definitions of speeds are given in Chapter 2: Glossary.
i Spot speed data are collected by one of two general approaches: direct and indirect measurements. D irect m~1uure
. mmts of speed are made using permanent or handheld technology that apply the Doppler principle via, for example
'radar or laser measurement technology. If a direct measurement is used, the resulting metric is typically the TMS,
although SMS can be esiimated as discussed above. Indirect measurements of spot speeds actually calcula.te speed from
time measurements of a vehicle traveling a known (short) distance, such as the distance between two closely spaced
magnetic induccance loops. By m.inimizing the length of the spud trap the difference between the indire ctly measured
space speeds and the desired spot speeds is negligible.
A!; discussed in Chapter 9, speeds can also be inferred from vehicle travel times ove.r a longer segment. With increasing
segment length the difference between TMS and SMS can be significant as discussed in Roess, Prassas and McSh~e
(2004) or ocher sources.
Two basic methods of data collection are rhe individual vehick sekction method and the a/1-vehick sampling method.
Both merhods can use direct measurement or indirect measurement. Each is discussed below in terms of study pur"
pose, location, time period, personnel and equipment, 'sample size and field procedures ..
.!1
l
:l
if,,
.,,.
The srimtific difftrme~ b<:rween laser and radar technologies is telated to the wavelength of the wave beam, with radar wave$ gen
erally having a higher wo.velength and thus lower frequency. The
strength of a radar signal decreases with increasing distances. Effective rangc;s vary from a few hundred feet to rwo miles (approxi
mately 100m to 3 km). The tranSmitted beam b<:comes wider as
it uavds out from the unit, so it may cover more than one traffic
lane. Most units allow for adjustments in range and beam width.
Check the manufacturer's specilications when selecting the radar
device to ensure it will meet your srudy requirements. Laser b<:ams
:. . ~~
are more concentrated with a shorter wavelength and higher fre. So
CiS ock b
m/Dan'd
LoiscU
1
quency. They generally have a greater range than radar, allow the
uroe:
t
P oto.co
e
amlyst to directly aim ar a selected vehicle and arc less likely to pick up adjacent vehicles. Most modem bser devices
record both the vehicle speed and relative distance to the observer, which can be useful in some applications to filrer
vehicle speeds for a certain segment.
.
I
L
'IJ,
I
;
The practWJ applklllion of the two measurement technologies also difers. A radar measwement device generally
transmits a continuous spectrum of wo.vcs thac can re8ect off multiple objeets. In fact, radar units used in police
cruisen commonly give multiple speed estimates of vehicles directly in front of and behind the cruiser, as well as the
speed of the fastest vehicle. The scatter resulting from the continuous radar signal can b<: picked up by in-vehicle radar
detectors, and analysts must recognize the bias associated with driven slowing in advance of a radar measwement
station. Many of these detectors are visible to the observer as the vehicle passes, which allows chose observations to be
discounted.
' .
A laser beam is typically not continuous, but is triggered by the analyse (or police officer) after focusing on a vehicle.
It therefore is le.ss likely to be picked up by drivers in advance, even if an in-vehicle laser detectOr is installed. Due co
the ooncenuated ll2.turc of the laser beam, there is less signal scatter that could b<: picked up by upStream drivers approaching the measurement station. A laser device is therefore more likely to result in an unbiased estimate of speed.
The accwacy of laser and radar units is affected by two errors: round off error and cosine angle error. Radar units
typically display the measwed speed in digital form rounded down to the nearest whole unit of speed. For a.a.mple,
a reading of 55 mph (88.5 km/h) would mean this estimate was actually b<:rwcen 55 and 56 mph (88.5 and 90.1
km/h). Laser unirs typically provide speeds with greater accuracy to one decinu1 point, but differences may still aist
b<:rwccn manufacturers.
The cosine angie error occurs b<:cause the angle of incidence of the beam to me navel direction of the target vehicle
produces a reading on the unit that is less than the acrual speed. As shown in Exhibit 5-3, the measwement is a function of me ~inc of the incidence angle. While in law enforcement this error provides a margin in favor of the target
vehicle, other applica?ons of spot speed data may require a concction to the reading to ensure accuracy.
iI
J\l
I.
.P lan
Radar
meter
v
l.
Vsin a.
~~
b.---vcos a
Vector
diagra-m
Eth.ibit 5-4 illu.suata the effect of the aine angle error on true speed. Because of the absolute nature of thae two
error sources, the relative error decreasa as speed inaeasa. Some units have a built-in correction for angle error based
on praet angla of incidence (Taylor and Young. 1988).
A radar or laser unit is easily operated by one person. Many modern units have built-in data storage using USB or Hash
media. Alcernatively, operators can write down the digical readings displayed on the unit or can record the readings
verbally on a voice recorder for transfer Iacer co paper or computer for analysis. Even ifbuilt-in data storage is available,
it is useful to take notes on paper or a voice recorder, since it is otherwise difficult co reconstruct erroneous measurements from the elecuonic dawet. For c::xample, a study focused on free-flow speed measurement may inadvertently
include some vehicles in foUowing mode chat need co be excluded from the analysis. If traffic is heavy or the sampling
strategy is complex, rwo observers may be needed: one to call out the readings for the vehicles of interest and one to
record the speeds on an appropriate form.
l
!
I'
;}
i!
N = (s~f
if
where
'I
':;''
<
~ ~
Equation 5-1
<
1 '
LI' 'i
''~ .l'
Analyses can estimateS for this equation from previous speed srudies under similar conditions of srudy or from speed
monitoring data at a nearby location. In the absence of these data, Exhibit 5-5 presents estimated values of average
standard deviations (S) as a function of traffic area and highway rype.
q
''
!; !'
11
'I
!I
II
I!
.! ;
I I
I I
!I
I I
- -~----
-~.
..._.... .
; ~l*~~.w,.._.,~ " e .
. . .. ....
--~-. ~.
,.
--~
-~~-D-~~~-:
~-~-~~~~~~H~~~.:~~-: :~
Highway Type
km/h
Two-lane
5.3
8.5
Four-lane
4.2
6.8
Two-lane
5.3
8.5
Four-Jane
5.3
4.8
8.5
4.9
5.0
7.9
Two-lane
Urban
mph
Four-lane
Rounded value:
7.7
8.0
For the greatest accuracy, analysts can conduct the study, calculate the acrual standard deviation of the data and check
to see if the sample size was adequate. If not, additional data would have to be collected under the same conditions
as the first study. Another technique is to use a calculator to update continuously a running toea!, average speed and
standard deviation. When the standard deviation becomes stable, an adequate sample size has been obtained.
The confidence level expressed by constant Kis the probability that the difference becween the calculated mean speed
from the sample and the true average speed at the study location is less than the permitted error. This concept is discussed in greater detail in Appendix C. ~e corresponding constant K values for selected confidence levels are shown
in _Exhibit 5-6 and are valid for any sample size:s greater than I 00 measurements. For smaller sample sizes, the cot'~
re-sponding K value can be obtained from the tdi.mibution table given in Appendix C (Exhibit C-12).
The permitted error, E, reflects the precision required in estimating the mean speed. This parameter is an absoluc:c
tol~rance and is expressed as plus and minus a specified value. Typical permitted errors range from :1: 1.0 to :t 5.0 mp.b
(:1: 1.6 to :1: 8.0 kmlh).
For example, the analyst may be interested in the necessary sample size for a speed study on a rural tw(}-lane higbwa::Y
at a 95 percent confidence levd and a permitted error of :t 1.0 mph (1.6 km/h). In other words, the analyst want$
to show with 95 percent certainty that the true population average speed is within plus or minus 1 mph (1.6 kmfb.)
of the observed .sample mean. The estimate standard deviation, S, from Exhibit 5-5 is 5.3 mph (8.5 kmlh) and th.~
constant Kis L96 per Exhibit S-6 (for .samples greater than 100 measurements). Applying Equation 5-1, the neces-.
sary .sample size is:
-
= ( 5.3 1.96)
- =108 observanons
1
Once the sample of speeds has been collected, the anal~t should recompute the sample size requirement using the
uue field-observed standard deviation, to ensure the sample is still sufficient.
If the statistic of interest is some percentile speed, such as the 85th percentile, Equation 5-2 is appropriate for determining the sample size required.
S 2K2 (2 + U2 )
N=
22
Equation 5-2
where N, S, K and E are as defined for Equation 5-1 and U is the constant corresponding to the desired percentile
speed. Exhibit 5-7 presents constants corresponding to percentile speeds. As a general rule, the' minimum sample size
should never be less than 30 measured spot speeds to satisfy underlying assumptions of the statistical disuibutions in
Equations 5-5 and 5-6.
I
I
!
.1
!; I
\I
f,l
. on the spatial position of the observer the angle of incidence and associated cosine error may occur in the horizon~
; or vertical direction.
______
--
~~~~ I F--~------- - ~
------ - ---
~~
2.3.1.3 Docummtation
The final layout of the data collection site should be fully described in any report of speed data. Observers should
make an accurate sketch of the site showing the number of lanes, the position of the measurement unit and the x., y
and z dimensions as shown in Exhibit 5-8. The dimensions permit calculation of the angle of incidence so that a cosine error correction may be applied, ifdesired. Observers should record the start time, end time, any downtime and
the conditions prevailing during the study. Photographs of the layout may also prove useful.
2.3.1.4 UdibratWn i
The radar and laser manufacturer's recommended calibration tests should be made before the start and again at the
end of data collection. The results should be included in any report of speed data. It is also wise to make an initial test
in the office to ensure the radar unit is operable before traveling to the site.
2.3.2 Mznlllll SJHed Trap1 atul V"uleo
As indicated in a preceding section, analysts seldom use manual speed traps, and video is generally less efficient chan
direct speed measurement with a laser or radar unit..If video is used, general set-up considerations are consistent with
the d.iscuss!on offered in Chapter 4. lf video is co be used for speed studies, it is further necessary to establish known
distances in the video field of view to use for reference in measurement. The dashed lane striping provid~ ~ gooc,i ref.ccence in many cases, as do closely-spaced roadside fearures such as barriers, signs,,or trees. When no existing features
arc usable, the analyst can retrofit the site using temporary markings or roadside cones, but care should be taken that
the installation does not affect driver behavior.
Spot Speed Studies 85
For the study, the analyse measures the time it takes a vehicle co cravd the known distance between two points. This is
most simply done using a stopwatch, bur can further be performed using automated equipment including handheld
eleccronic count boards or a laptop computer. More information about rhe procedures for conducting spot speed
studies using manual speed reaps or video is given in Roess, Prassas and McShane (2004).
' !
!,,!
:i .
'~
'I.
j .
I,
I .
I .
~ ~
i
i
~I
j
.l I
3.2.2.1 Pmonnel
One of the advantages of automatic speed data collection is chat personnel are needed only during installation and
recovery of the data collection equipment. Once installed, the equipment can operate unattended for several days or
may be installed permanently. Agencies normally use two or three people to deploy and recover sensors and recorders.
Safety dictate.s that one person never attempt this task alone. During installation and recovery, the crew and its vehicle
should be highly visible co motorists. The crew should use cones and warning signs. One person should watch for
traffic anytime a crew member is in the roadway. Permanent, long-rerm installations are typically done by professional
work crews and involve lane closures to ensure safety.
3.2.2.2 Equipment fqr Alltcmaud Speed Mearuremmts
The most commonly used devices for measuring speed are in-road sensors in the form of pneumatic tubes, standard
induction loops, or point loops. These devices are normally deployed in paits. Two measurement units are placed a
short, measured distance apart to form a speed crap that measures the time it takes the vehicle to travel from one detector co the next. Agencies may place these sensors in saw cuts or bore holes in the pavement, sealed for protection from
the environment, in the same manner that sensors are installed on approaches to signalized intersections.
The devices used to collect automated speed data are generally the same that were introduced in Chapter 4 for automated volume studies. The technologies include permanent in-pavemem installations, and temporary on-road
measurement technology. For most technology, the manufacrurer specifies installation requirements, including the
distance berween' two sensors. These should be followed closely to ensure that the unit is well-calibrated. To measure
f
;
speeds on multilane roads, study teams must place sensors with separate inputS to the recorder in each. lane. Exhibit
tT~:l-~'"l_q1lJ.111l_tf~f-~~
f.
n
cl
Sl
3.3.1.2 DqJ/oymmt
Safery is the 6rst consideration during deployment. It is preferable to close each lane to traffic while work is under
way. The sensors should be prepared on the roadside to minimize: the time each lane is closed. Workers then place each
set of loops, rubes, or other equipment prescribed in the data collection plan and in accordance with manufacturer
specifications. The proper spacing ofsensors determines the accuracy of data and the crew must be careful. to install
them as dmcted. After placing, the lead wires are connected to the recorder and ihe sensors are checked for proper
functioning. After any needed repairs arc made, the crew can secure the sensors to the pavement. The field crew should
have some leeway to select the exact position to deploy the data collection system, to avoid broken pavement and to
locate the recorder near some fixed object to which it can be secured.
3.3.1.3 QzljbrlltWn
With the sensors and recorders in place, the next step is to check the accuracy of the equipment in measuring the
counts and speeds of the traffic stream. A calibrated radar or laser gun is used to measure vehicle speeds, which arc
compared to the speeds of those same vehicles monitored by the data collection equipment. If necessary the crew can
adjust the recorder until the speeds are within a :t 1.0 mph (1.6 kmlh) tolerance. For temporary, in-road installations
it is ~dvisable to check the functions and accuracy of the equipment at lease once during every 24-hour data collection
period. Permanent sensor installations should als() be checked at regular intervals.
3.3.1.4 &cowry
After the data collection period, the recorded data are cliecked for accuracy by computing ~e ratio of usable vehicle
data to total vehicle data. To compute this ratio, compare the total vehicle count to the usable vehicle count. A usable
vehicle is a vehicle with both a length and a speed. Sometimes a vehicle will be sensed by only one of the loops in the
pair. This occurs when the vehicle path does not cros:s both sensors or when one of the sensors &il.s. Such vehicles are
stored in the total vehicle count but not in the usable vehicle count. The ratio of usable to total should exceed 0.75.
If this criterion is not met, the team should collect data during another period.
Crews recover data collection equipment by reversing the process they used to deploy it. Again, safety should be the
principal concern. If the equipment is to be deployed at another site, the data should be transferred from the recorder
either at the site or via modem or fiber connection co the office. If the site is being monitored for an extended period
of time, the data should be dumped" periodically.
..
3.3.1.5 Docummt4lion
The fuu.llayout of the data collection sice should be fully described in any report of speed data. The crew should
make an accurate sketch of the site, showing the number of lanes, the position of the sensors and the location of the
recorders. The crew should record the start time, end time, any downtime and the conditions prevailing during the
study. Equipment malfunctions and repairs and the results of calibration and accuracy checks should be recorded.
.Photographs of the layout may also prove useful.
I,
'\I.
I' !II
'.
l \
I:
)
! \
t~
i!, !
H
1[i~ ~
A typical spot speed nudy analysis has three parts. Data rcduaion is the first pact; that is simply the arrangement of the measured speeds, or "raw data, inro a convenient tabular or gr.1phical form. The second part is the calculation and presentation
of~pcive statistics which illustrate the collection ofspeed data by means of a few representative values or variables. The
third part of a cypical analysis is statistical i.nfuenec, which permits the devdopment ofstatistical estima.tes and the testing
ofStatistical hypotheses. Appendix C presents detailed information on field srudy analyses and uses many examples from
spot speed data. Therefore. the disrussion in this section is brief, and the reader is referred to Appendix C for greater detail.
~ I
ij
l!
,,
IL
.,~~ :
~i
I
i;
l!;
;[
~
!
~ i
ii
>I
f.
.
.
I'
11
i
...
_.,."'
~
1:
J"
8
..
l
,.,
,.,
VtHdS~
..
......
"'
I'
40
~g
000
20
00
OS
00
"'
60
SO
o..:.,..... ,.
I
....
t ..
~
..
~~----.--.~.~--~--.~.~--~--~~--_:~~.~.v.ru~
SJioled
c~l
49m. .
)O
lO
:
0
00
~~---L--~.-~--~.t..-~--~~;---L-~~
2. The before and after means (i ~and i...) and standard deviations (s~ and s.p) are calculated independently. The before and after sample sizes Should be as calculated earlier in this chapter. The following test
should only be used for sample sizes greater than 30 observations for each period.
3. The difference in means, idlff is calculated from the two estimated means.
Xdllf
= Xafter -
ibefore
Equation 5-3
4. The standard error of the combined sample estimate of the difference in means is calculated:
2
SE = ~~ + S~ntor
NA/tttr N"''"''
5
Equation 5-4
5. The hypothesis in this statistical test is that the distribution has a zero mean-cluu is, there is no difference
between mean speed before and mean speed after the change.
Xafter - Xbefore
=0
Equation 5-5
6. The test sarisric is calculared &om the observed~ and hypotbesized (0) mean diffm:nce, the standard error of the
combined sample estimate of tbe difference in means (SE) and tbe t-value obtained from Exhibit 5-16 fur the userspecified significance level (~ or confidence level (1- ~ and the before and after study degltes of fi=lom 0. The
before and alter samples are sufficient to establish that che difftrcnce in spcds is significant if.
-0
Equation 5-6
~~t
SE
idlff - 0
It is important to note that a before and after study that yidds S E < t does not establish that the difference .in
speeds is not significant. Rather this result means only that the sample sizes (N+ and N ~ were not large enough to establish the statistical significance of the ~ d.i.B.i:rcnce. As with any study. the praaiCif implications should always be
considered since even a statistically significant difference in speeds rnxt not be meaningful if the observed difference idlfl
is small. This is espocially true as modem aummated means of collecting speed data read.ily result in very high sample sizes,
which reduces che srandard error of the estimate and increases the test statistic in Equation 5-6.
lI
~.
Sample calculations for the procedure done on a spreadsheet arc shown in Exhibit 5-17.
92 MANUAL OF TRANSPORTATION ENGINEERING STUOIES, 2ND EDmON
!1Z'litlt~r.lf!
Ous Boundaries
:-
.. . , . .
..
'
O...limits
Before (f,)
28-29
2B.5
.30-31
30.)
32-33
32.5
34-35
34.5
14
36-37
36.5
38-39
38.5
20
40-41
40.5
38
25
42..(3
41.5
29
34
#-45
44.5
35
44
~7
-46.5
15
29
48-49
48.5
12
12
50-51
50.5
13
52-53
52.5
II
>4-SS
54.5
~~~
27.5
29. 5
31.5
33.5
35.5
37.5
39.5
41.5
43.5
'
45. 5
47.5
49.5
Sl.5
53.5
'
55 ..5
56-57
56.5
57..5
To..J
:E
Man _
X=
Scaadatd Deviuiom
f ,u,
~~
$.IL
f,.' - ~i~:]~J ~
Xdi
SE =
N.
.:!L.+i
-0
z- ----s
'Dilertru:e ill Jpttds wilh 95% one-ailed con.&Wtce inaMI, +- aitic:ol (s 0.05, df~ 200 . 1.64).
1&6
198
.8clore
After
42.35
44.67
4.50
4.68
2.32
0.4685
..
b57<l.6-4'
,.,.
!.
'' '
.
I
1.,
I
,, ..
!:
,,ij
'.\,
1
1
.,
l
.~
l
(
".
[:
~~
!
i
!
.l
. I
.l
I ..
il
,,i;
S.OSUMMARY
In this chapter various methods used co perform traffic speed studies at spot locations have been described. The chapter distinguished between individual vehicle sampling mechods, where data collectors randomly select a subset of che
vehicle stream using mosdy handhel~ or manual data collection equipment or video, and all-vehicle sampling, where
automated speed data collection equipment records speeds for the entire traffic stream. The chapter further discussed
is5ues related to data reduction and analysis, including before-and-after statistical tests for speed studies. For further
information and more details on speed studies, refer to Roess, Prassas and McShane (2004).
6.0 REFERENCES
Banks, J. H. EvalUAtion ofPortable Data Colkction uchnologier: Final R.tport. San Diego, CA: University of California-Berkeley;
Partners for Advanced Transit and Highways (PATH), 2008.
Bonsall, P. W., F. Gbari-Saremi, M. R. Ttght and N. W. Marier. "The Performance of Handheld Dara-Caprure Devices in
Traffic and Transport Surveys." Traffic Engineering and ContrOl29, No. I.
Box, P. C and J. C. Oppenlander. MantuJ! ofTraffo Enginming Studies, 4th ed. Washington, DC: Institute ofTransportation
Engineers: p. 13.
Dixon, M. P.-R. "Field Evaluation of Roundabout Turning Movement Estimation Procedures. ASCE journal ofTransportatiofi
Engineering 133, No.2. Reston, VA; American Society of Civil Engineers.
Federal Highway Administration. Virginia State Route 7 Vuleo Detection System Pnfonnance Astessmmt. Washington, DC:
Federal Highway Admioisuacion, 2006.
Garber, N.J. and L A. Hoe!. Traffic & Highway Engineering. Pacific Grove, CA: Brooks/Cole, 2002.
Gates, T. S. "Comparison of Portable Speed Measurement Devices. Transportation Retearrh Record: journal ofthe Transportatit? ?'l
&search Board 1870 (2004): 139- 145.
Grecne-Roesel, R. D. "Effectiveneness of a Commercially Available Automated Pedestrian Counting Device in Urban
Environments: Comparison with Manual Couors." Procmiin~ ofthe 87th AnnUAl Muting ofthe Transportation Retearrh Board.
Washington, DC: Transportation Research Board, 2008.
Hummer, J. S. ~Recent Superrueet lmplementacion and Research." Confemue Proem/in~ ofthe 3rd Urban Street Sympptium.
Seattle, WA. Washington, DC: Transportation Research Board, 2007.
Spot Speed Studies 91' 5
JAMAR Technologies. Data Collection Equipmnrt. Retrieved December 29, 2008. JAMAR Technologies: ja.martech.com/index.
hun I.
Kyte, M. wd M. Lines. "Development ofVideo-Ba.sed and Other Automared Traffic Data Collection Mcthock " Traffic Trrulur
ManU4L Retrieved April 20I 0. Nacionallostirutc for Advanced Transporw:ion Technology, Un~ity of Idaho, Moscow, lD.
2004. www.wcbs1.uidaho.edulnianlr=rcb1Finai_Repons/Kl.K203_N04-02.pdf.
Laser Atlanta. Spud Mtasumnent Equipment. Retrieved December 29, 2008, Laser Adanta: www.laseradanta.com.
List, G. F. "Identifying Vehicle Trajectories and Turning Movements at Roundabouu. 5th lnremational Symposium on
Highway Capacity wd Quality of Service. Yokohama, Japan: Japan Society ofTraffic Engineers (2006): 449-458.
McCarthy. J. TnzjJU J!naipis Toolbte.
~ashington,
Min-Tang, L F.-F. "Assignmenr of Seasonal Faaor Categories to Urban Coverage Count Stations Using a Fuzzy Decision Tree.
ASCE]ou1714i ofTrrwport41ion Enginming 132, No.8: 654--662.
Robertson., H. D. Manlllli ofTransportation Enginming SnuUn. Washington. DC: Institute ofTran.sporration Engineers, 1994.
Rodegeru, L B. NCHRP Repon 572: RowuiAb~IS in tht Umttd Statts. Washington, DC: Transponation Rc.scuch Board,
2007.
R.oes.t, R. er al. TnzjJU Enginuring, 3rd ed. Upper Saddle River, NJ: Pearson Prentice Hall, 2004.
Smadi, A. B. "Advantages of Using Innovative Traffic Data Collection Techniques. Appliurums ofAAwntd Thrw/4gia in
Transporwion. The Ninth International Confewtce. Reston, VA:. ASCE Publications, 2006.
'
i~
Tarlto, A. and R. Lyles. Dtwlopment oftJ PmtJble Yliko fRttiqn System for Counting Turning Vthicks at lntmectio'!l. Purdue
Taylor, M.A. P. and W. Young. Trttjfic Analysis: Ntu1 Ttclmo{ggy aNI. Ntu1 Solutions. North Mdbourne, Viaoria, Australia:
Hargrctn Publishing (1988): 150--157.
University Repon FHWA/IN/jTRP-2001/18 doa.lib.purdue.edu/jup/62/. West Lafayette, IN: Purdue University, 2002.
Transponacion Resea.ch Board. "Data, Survey Methods, Traffic Monitoring, and Asset Ma!ugcment. Tramport41ion RestJZrch
&cord.: ]~1714i oftht Trrwportlltion kurch Board 1993 (2007).
Zou, N. and J. Wang. "A Video-Based Method for Evaluating Traffic Data from Derea:ors." .Applicatioru ofAJvanctd
Ttch":"login in Transporwion. The Ninth International Conference. Reston, VA:. ASCE Publications (2006): 232- 237.
;~
.;
.;!
.:
I..
l
[
~
d
\
Chapt er 6
97
2.0 DELAY
98
2. 7 Equipment Needs
98
99
104
10S
99
!
6.0
10S
105
106
109
5. 7 Equipment Needs
109
110
110
112
6. 1 Equipment Needs
112
113
113
7.0 SUMMARY
115
8.0 REfERENCES
115
1.0 INTRODUCTION
ntersection and driveway studies are among !he most common studies in transportation engineering. In particular,
many agencies routinely count turning mOvem.ents and study intersection delay. Other intersection and driveway
studies include queue length, saturation Bow and lost time, gap and gap-acceptance, and intersection sight distance
studies. Analysts perform lhese studies less often, but find !hem useful for special locations or when calibrating basic
traffic Bow relationships. Analysts use the results of intersection and driveway stwiics to determine what kind of traffic
control devices (TCDs) arc warranted and to determine intersection capacity,.uaffic signal timing, site development
impacts, ~e speeds, driveway locations and oilier important parameters.
The basic data needs from intersection and driveway studies have changed very little lhrough the years.' However,
the equipment used to conduct the studies has changed greatly in recent years and is likdy to keep changing. In this
Intersection and Driveway Studies 97
chapter, the basic procedure for each srudy is described together with the equipment that is most commonly used by
agencies in North America for dday; queue length; saturation How and lost rime; gap and gap acceptance; and intersection sight distance srudies. The procedures for intersection turning movement counrs were derailed in Chapter 4.
Analysrs should note this chapter contains methods for calculating u:Ufic measures developed for U.S.-bascd srudies.
Different methods developed in other countries were extensive, and therefore were outside the scope of this chapter.
Therefore, if outside the U.S., the analyst should consult specific guidance on procedures for intersection and driveway srudies.
~ ;
'!'
2.0 DELAY
:i'
:} ~
'
ij .
I
~i
!':
lmersecrion delay srudies are very common. Intersection delay data have many US<:S, including measurement of the
quality of traffic Bow and evaluation of the need for traffic signals. Analysts can estimate intersection delay with equations or simulation modds. However, the inpurs co the equations and models can be extensive, and the results arc only
approximations of actual traffic operations. Therefore, field srudies of delay arc often used at operating interSections for
greater accuracy or to validate theoretical delay prediction. One of the major problems with interSection delay srudies is
the definicion of delay. There are many types of delay, and using terms casually can lead to error. The following are among
the most useful terms describing delay at intersections (TRB, 2000, FHWA Trt1jjic Signa/ Timing Manual) .
' I!
7irM-in-queue tklay (TIQD) is rhe difference berween rhe time a vehicle joins the rear of a queue and the
time the vehicle dears the imersection.
,.,
'
.; .
!
Ctmtrol tklay is the component of delay that results when a conuol signal causes a lane group to reduce
speed or to stop; it is measured by comparison with the uncontrolled condition. lt is defined as the TIQD
plus time losses due to deceleration from and acceleration to free-Row speed.
Gro=tric Delay is the component of delay that results when geometric features cause users to reduce their
speed in negotiating a facility.
d
i
'I
Travel-ti~ tklay (lTD) is the difference between the time 3. vehicle passes a point downstream of the
intersection where it has regained normal speed and the rime it would have passed that point had it been
able to continue through the intersection at its approach speed. This includes all conuol and geometric
delay.
Analysts use control delay most often because it is the easiest to measure and because the Highway Capacity Manual
(HCM) (TRB, 2000) bases its definition of intersection level of service on conuol delay, among other reasons.
i.;; .
;
.~ !,
Observers can collect delay data manually or by electronic means. Data sheetS are often helpful for collecting del2y
data at intersections, and allow data to be collected manually while in the field. Ho~r, more often the use of electronic counting boards, such as those de~ibed in Chapter 4, are used to collect delay. Built-in software allows the
user to quickly obtain outputs. If electronic counting boards arc used, the user needs to make sure they understand
the output delay measure. For instance, many count boards collect orne-in-queue" delay, which is closely related to
co nuol delay bur is lower based on the deceleration and acceleration rates and the length of the queue discharging.
j :
.!.,I
,,
:q
'I!
i:
Becoming more popular in recent years is the use of a laptop computer. Computer software can be manipulated by
writing *code" to use the key pad with the internal clock, allowing time scamps of various events to be recorded in real
time. This manual provides an example code in Appendix E. This form of data collection is typically more flexible,
and spreadsheers can be developed to calculate various intersection parameters.
SCleral srudies have used video with a time stamp overlay to measure some types of delay. Video provides a permanent
record of the srudy period that may be used for further review of the delay data or for other srudies. Video may also
reduce the num.ber of field personnel needed for a delay srudy. However, video recordings frequently suffer from poor
lighting conditions and vantage points. Long queues are especially difficult to capture. When possible, an overhead
vantage point should be Ud to prevent occlusion. Many times, surveillance or video detection cameras arc already
jI
.
I
I
II
pu~poses.
. For a delay study, it is often necessary to determine the free-flow speed (FFS) of vehicles prior ro the intersection or
interesr. This is typically done using a radar or laser speed gun; however, the space-mean speed (SMS) could be used
by determining the run rime between known points for a large number of vehicles. Chapter 5 provides guidance on
rhe use of various speed devices and methods for determining the average speed.
{:
I'
[i
t ~.
fi
l~
1!it
rJ:i
r,:
~~
:
.i
'!
~
!!:
..
;
!:
Equation 6- 1
i::)
0.9
Equation 6-1
where
i,.
~
IV
IV
'f
Equation 6-3
~~
,.--;"' _. .. . .. ' . ...
Sl~
lonalysl
lnllaectlon
Ala Type
C CBO
Judsdlctlm
f'l,(
~11-Periad
---- -
1899
'-~)~sis Year
Input fnttllll,
Numlzral loles, N
~- ~. Fl'S (mllh)
40
,,
~C"~ ~-cl~C
SU.-...ycountlnllrwl.l,(s)
Cyclt
-.-llumbot
llrra
'I
2
I
4t:)4
2m
'
"
_j
--;------:-cl.:...
~-s - ~----.iiiii;,=JF"~-:-- 1
- - - - - . - - - -
......
Clock
letvlh. c (s)
!5.!10
D f a l " l tel t z l z l o 1 2
61 121 t:> I 16 1 6 I o I o I 2
1 o 1 o
2
2
"''"
Total
($
?::
&
II
l!Y7
16
o. ~) 0.9
~Z!
...
I 41 .
,...,...,oreyc~es~., .
v.
it'f
~
0.42
1.7
ff-2
Source: High1INIJ Ozp4LjJy M411Vid 2000. Copyriglu, Nationa!Acadt.m.y of Sciences, Washington, DC. Exhibit A16-l , p. 16-89.
Reproduced with permission of the Transporotion Research Board.
Vehicle-in-queue counts in
--
'
Before starcing the srudy, rhe observer should record any genc:ral information, especially the input par.imerers. These
will be usc:d in many of the calculations to determine the control delay. FFS is determined as the unimpeded speed
through the intersection if it were green for an extended period of time. This can be determined by driving th rough
during the green indication when vehicles are not impeding Bow, or by recordi ng the speed at a midblock location
away from the intersection under study.
Ideally, rhe survey will begin at the start of a red phase with no overflowing queued vehicles from the previous cycle.
If rhis is nor possible, the overflowing vehicles should be excluded from the analysis. Two observers should be used,
with the following tasks for each observer:
Observer 1:
!:
I. Keep track of the queues for each cycle in the survey period by observing the last vehicle in every lane that
stops due to the control device. This includes queuing vehicles during the green phase.
il.'
'=:1
2. At some predetermined intc:rval {between I 0-20 sec.), observer should record the numbc:r of queued vehicles. Again, a repeated time intc:rval (using a stop warch or othc:r device) is necessary to do chis srudy and
the time interval should not be divisible inro the cycle length.
-~I
~I
,"
3. Enter the vehicle-in-queue counts in the appropriate box in the worksheet. The survey period should be
assumed prior to conducting the study. The clock rime should be recorded at the beginning of every fifth
cycle.
.i '
ji
.1 ,
G;
4. At the end of the srudy period, the observer should continue following any queued vehicles prior to the
end of the srudy period until they exit the intersection.
'
)j'
I:
Ji
Observer 2:
1. During the entire survey period, conducts a continuous vehicle count and a count of vehicles stopping one
or more rimes. Vehicles stopping multiple rimes should only be counted once.
,q
!:1,:.
9
it ;
I
,.,
l'
'I;
11
I'
;j
"' i
~\
:i
H
' I
ji
li
ll
Once a control delay is calculated, the level of service {LOS) can be determined using the appropriate facility type
(signalized or unsignal.ized ime~ection) . For signals, when control delays for the minor street are high, the analyst
should be aware it may not mean there is really a problem. Since control delay is measured in seconds per vehicle, if
the volumes are low and the cycle length is long, a vehicle or two may have to wait a significant amount of rime before
receiving a green signal indication. As an alternative, the volume-co-capacity ratio (vic) would be a better measure of
the available throughput (or lack of throughput) for the given movement.
1!
G7'T1
liof'.
"'
FFS
'I
l:Ii
iJ
Equation 6-4
where
. GTT
,
assumed geometric travel rime had rhe geometric fearure nor been installed
d2
d,
Next, the analyst should record the travel rimes of a large sample (usually 100 or more) of unimpeded vehicles (or
other modes) to determine the travel rime (GTI2) between the same two pointS of interest. It may be h elpful to use a
test vehicle to emer the uaffic stream at random times and determining the time between the two points. The average
of the actual individual uavel times is then used to determine the geomwic delay (g) by Equation 6-5.
g = GTJ1 - GTJ1
Equation 6 -5
It is important to point out again that geometric delay DOES NOT include the effects of any con uol device or
conflicting traffic. It merely measwes the delay that the geometric feature causes by comparing it to the "no change"
condition.
If the geometric dday were utilized, it would be calculated by comparing any unimpeded rravd time of any single
movement through the intersection to the similar unimpeded movemenr using the constructed geomerry in the fidd,
each under free flowing conditions (no TCDs) with no conflicting traffic ar the same beginning and ending poinrs.
However1 TID would be differenr because it would also include the effect of the TCD employed u this rype of inte.rsection. For the mainline thru movements, the unimpeded travel time is easy ro determine because the vehicle's FFS
is assumed to be consranr through rhe signal. However, the FFS cannot be used for mainline thru movements, or any
of rhe U-twn or left twn movements. Therefore, one of two merhods can be utilized.
I!
\i
.I
f'
lJ
Firsr, the travel time through any of the movemenrs could be collected using actual driver data through the corridor
during low traffic volumes when the signalized inrersection movement under analysis has a green phase. However, this
could be complicated. Another possible option is to use a simulation program (such as those discussed in Chapter 11)
and assume the default (or field-a.librated) acceleration and deceleration rates. The analyst could apply volumes only
to the movemenr of interests and derermine the unimpeded travd time. This would be TI1
~I
;
;
',.
.i
Next, the actual travel time is derermincd by conducting a fidd or simulation study and tracking vehicles through
the supersueet intersection at the same two points: the beginning and end of the section of interest. This is TT2 It
may be helpful to use a test vehicle, sometimes called a Heating car," to enter the traffic stream at random times and
determining the time between the two points of interest. AgWt, a fairly Large sample should be used to gain confidence
in the result. It is best to use 100 or more measwements for each movement; however, it may be impractical to collect
this much data and engineering judgment should be used.
l:
,r
!I'
I
Equation 6-6
where
'
!
1T1
1T1
Another option for determining TTD chat is becoming more popular is the use of GPS devices to track vehicle
trajecrories through intersections using a time-space diagram. Chapter 9 describes this technique, along wirh other
technologies, that could be utilized. Last, TTD is used more often for corridor stutlies where the effecr of multiple
coortlinatcd intersections is of interest. For instance, a corridor may have new timing plans or signal p)'wing that has
been employed along the corridor. A travel time study through rhe corridor could be used to determine if the new
coortlinuion plans bad a positive effecr. Again, these corridor srutlies are described in much more detail in Chapter 9.
; Field investigations are the best method for determining the :tctual qyeue lengths at an :tpproach. Macroscopic models
: are used frequently for signal riming and can often be wed co determine a queue's percentile {e.g. 50th, 90th, etc.).
i However, macroscopic models are equ:ttion-based and therefore do not cake into account the actual s cor:tge bay
"'.lengths. Instead, they :lSSume an infinite length for each lane group; therefore, simulation is usually employed co determine if a stor:tge bay is long enough or if :t proposed drivew.~.y would be blocked for a significant amount of time.
Further informacion on simulation srudies can be found in Chapter 11.
phase:
StArtup lost time occurs between the time the green signal begins wd the queue begins moving efficiently.
Ckaranct lost time occurs between the time the last vehicle crosses the stop bar and the next sign:l! phase
begins.
Many agencies usc standard constant values for saruration flow and lose time in analyses. However, saruration Bow
and lost time vary significantly between intersections and between different times of day. To avoid errors caused by
in:tppropriace usc of a stmdard value, some agencies measure saruration flow and lose time directly before performing
o~er analyses. More often, agencies sample saturation Bow and lose time periodically at several sires in an area and
calibrate their equations based 9n those samples. The procedures for measuring saruration flow and lost time are described in this section. The procedures are rel:ttivdy simple and one can use a variety of equipment to perform them.
'
The video inethod requir~ a clear van~ point and good light conditions. In the office, a technician must stop the
video and record the time on the on-screen qock as the vehicles of interest cross the stop bar.
have good reflexes and understand the exact data collection methodology prior ro going in rhe field . Any errors, even
small errors, could have a significant effect on the values (especially lost time).
:~
The hardest part :~.bout conducting one of these two studies is uying to be inconspicuous. Rardy is the analyst able to
obcain a good field of view wichouc ~ing out of a vehicle and in full view of approaching vehicles.
i1:
ti.
J'
'0:
~:
iii
H
!I
I'
\I
.
~!
I
:!
~~
:;
I
I
~
'
'
.l
j'
.'!
,,,,
'
t:
I
!l ,
j lt
J
i
~I
L
,.I
' 1.
. I.
r.
n
11:
:~ .~-.~~
....
-,
I
FJEJ.D SHEET- SAn.NtA nON JILOW' S1l.JDY
~ 8 oundTtl f'flcO~to:
Une! ---ca.!'__.~~clr:u;._.-A~"'
CLC.QO"'"'!i~~t...~'----:--:-T!mv.
!.f::oof'D!!='~ Wf!!.\:t'.
Y=la/92.. Tiy:cs
ObMNera~
Qrado:'
1\1. H
I
Qbs
2!o
Weattor.~C..""';I~p~.....:f~":;j=?;:-::"!f:o.__ _
Lan.e Wicnh~
I :Z.tl
Areo;
.tw.bkV'&e.,.:,
"" ....h.
81h ..... ,.
8lh voh.
l ,Olh -h.
1r- :z..
1/!.9::._
1l '2..
10.1
h~
10.1..
-:r."
.
. . /0
, II-
..
.
-
--
20
S' ~
!!!i'
113
' ( .'I-
~0
Ttl\ v~ .
.._,~h .
Mhvo h ,
lO th. Vol'\,
It-S"
/ 1 -1
nl
S-2-
c .~
s~
I.
.c:;,
CfO
/0.,
"" I
S.'&
::I
I~.~
I 0 -2-
:13
:M
""o
~
- -- --
22
2AO
' 10
c;,..,
- -- --
-n I
5-3
12
.,.
:s l
20 I
., ,
'-3
..
Ot:hor. -
22
_1,1-_4
<o.eA.:Y""'t;"c
Obo
No.
.eoee
~'S:==t ~ =~u
Sa.,o.cnn"
Looallon:
&/.'2-
sa
~
oo
Column..,..,.
I 5? 0
J
I
....
I
I
I
lf-S.tlI 4-7.0jiO'J.S
a..
""
ld>
~ ~
If -
The factors that affect saruration flow rates are grade, lane width, :intersection location (CBD versus other), type of
lane and presence of adjacenc packing lanes (TRB, 2000). Therefore, the analyst muse carefully select approaches to
measure saturation How co ensure an unbiased result. Do not use a saruration flow estimate from a steep approach 1:0
analyz.e a fiat approach, for instance. Heavy vehicles also affect saruration flow rates, so observers should not reco .rd
data if a heavy vehicle is in one of the first seven positions in the queue. If a heavy vehicle is in position 8, the obserVer
can record the time becween the fourth and seventh vehicles, and so on. Also, do not record data during a signal ph~.se
in which traffic How is interrupted by buses, by left-turning traffic waiting for opposing traffic to clear, or by rigb-tcurning traffic waiting for pedestrians to clear. Analysts can calculate interrupted saruration How from id eal sa~~~r.~ti on
How by the methods of the HCM (2000). The procedure for studying saturation flow in an exclusive lefrrurn or
right-turn lane with a protected signal phase is the same as the basic procedure for a through lane.
For agencies that have difficulty finding sites unaffected by the factors mentioned above, Roess, Prassas and McSha.$'le
(2004) suggest a procedure for estimating ideal saruration flow from measurements at nonideal sites. The analyst c;:;;afi
solve the saturation flow equation in the HCM for the ideal saruration How given the measured oonideal saruratiOD
flow and the standard adjuscment factors for the nonideal conditions. Trme of day, wearher, and evenrs that affe ct
driver populations or behavior also affect saruration How. Measure ideal saturation flows during peak hours, in cl..JY
weather, and during times when no special events are affecting drivers. It may be difficult to collect saruracion flow
daca during nonpeak hours in any case, due to small queues.
-
Intersection and Driveway Studies 101
One can calculate desirable sample sizes for a saturation flow study from a standard sample size equation. Usually,
analysts have some knowledge of the precision of the saturation flow estimate they desire. For instance, an analyst may
not want the mean estimated saturation Bow rate to differ from the true saturation Bow rate by more than d vehicles
per hour. The analyst can find the necessary sample sitt n by Equation 6-7.
n = (z~J
Equation6-7
where
n
constant from the standard normal disuibution corresponding to a certain confidence level
(see Exhibit 6-5)
= estimate of the standard deviation of the population of saturation flow rates
A I)'Pic:al value for sis 140 vehicles per how (ITE Technical Committee 5P-5, 1991). If the analyst is willing to use
this !)'Pica! standard deviation and wants an estimated mean saruration flow rate within 50 vehicles per hour of the
true rate with 95 percent confidence, the analyst would have to observe n = (1.96(140/50))1 = 30 valid queues. A
peak period at a moderately busy intersection usually produces at least 30 valid queues.
Ii. .
Once the data have been collected on the form in Exhibit 6-4, one can calculate the mean saturation flow rate using
the equation on the bottom of that form. Basically, a mean saruration Bow rate is estimated by calculating an average
numb.er of seconds consumed per vehicle (headway) and converting that into a number of vehicles per hour. For the
sample data shown in Exhibit 6-4, mean saruration Bow in vehicles per hour, SF in Equation 6-8, is estimated from
the equation on the bortom of the form as shown in Equation 6-8.
SF=
3600n
~+~+~+~+
3 4 5 6
SF=
~;:
~
t .
Equation 6-8
3600 31
57.0 + 45.1 + 47.0 + 109.8
3
4
5
6
I.
i::
.,,
:j .
.~
.i,
I
I
T
l
\!I :
~
SF .. 1,925 veh1hr
In Equation 6-8, n is the total number of observations, and a, b, c and dare the times in seconds between the fourth
vehicle and 7th, 8th, 9th and lOth vehicles, respectively.
4.3.21Ast Tmu
Lost time is more d.iflicult to study than saturation flow for several reasons. Fust, lost times are short, so accurate measurements require quick reflaes. Second, observers can measure clearance lost time only during completely saturated
green phases. FuWly. many of the variables that affect .tarurarion Bow affect lost time, plus others, including signal
head position and lens size. The analyst must be cacefu.l when applying a lost-time estimate from one lane to other
l:mes, approa.ches, or inrersections. Observers record lost-time data with a stopwatch, l.aprop computer, audiotape and
108 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
a computer back at the office, or with video that has an on-screen clock. Since analysts need an estimate o( sa turation
flow to compute startup lost time (as described below), they often gather data for the two studies simultaneo usly.
The majoricy of uncertainty in lost-time studies is where to establish the reference point for timing. In othe r words,
"Where is the vehicle considered to be in the intersection?" Previous studies have used:
the front or rear tires as they crossed the position that had been occupied by the front tires of the first
veh.icle in the queue;
the stop bar;
the crosswalk line;
the extension of the curb line of the intersecting sueet; or
other points.
Berry (1976) showed that .these different reference points dramatically affected the startup i~st-time estimate"{an
almost 3 sec difference in some cases). For capacity analysis, Berry recommended recording the time when the from
bumpers of the vehicles crossed the extension of the nearside curb line of the intersecting screet.
The total lost time (tL) is the sum of the average start up (t) and clearance lost times (t), shown in Equation 6-9.
IL = t,1 +lei
Equation 6-9
The data needed to compute Startup lost time (t) are the times when the green signal begins and when the third vehicle in a standing queue passes the ref~ence point. Most studies use the third vehicle because that is the last vehicle
that commonly experiences any measurable lost rime. Analysts can compute startup lost time for a phase by computing the difference between the .two recorded times (green signal and third vehicle) and subcracting three times the
average headway for the lane (in seconds per vehicle) found during the saturation flow srudy. For instance, an average
headway value for a signalized intersection is approximately 1.9 sec per vehicle. If the time between the green signal
starting and the third vehicle passing a known reference point (such as a near curb line) were 6.1 sec, the start-up lost
time would be 0.4 sec [6.1-(3*1.9)]. As mentioned earlier, the headway values are determined using a stopwatch and
a predetermined reference point (where the times between ftonr bumpers of consecutive vehicles over the reference
point are recorded) with sufficient queues of four-1 0 vehicles in length. The result from the startup lost time calculation can be below zero for a particular phase. In that case, anal.ysts assume a Value of zero when calculating statistical
parameters based on the results.
Observers measure clearance lost times (t) directly at the end of a sarurated green phase. This is a signal phase whcie
heavy volumes (and consequently tight, ftee..flowing queues) were o:insistendy flowing during the entire allotted phase
time. They record the time when the last vehicle through during the phase crosses the reference point and the time when.
the signal rums green for the next phase. The difference between these times is the clearance lost time. Observers need to
lind a location where they can observe both the reference point for timing and the signal indication for the next phase.
5.1 Equipment'Needs
G;tp and gap acceptance studies collect data using count boards, laptop computers, certain types of automatic vehicle
detectors, audiotapes in combination with oomputers in the office (as described above for saturation flow studies),
video, or stopwatches. With automated detectors, analysts must ensure that only the lanes of interest are being measured. Laptop computers and audiotapes with computers require special computer programs, such as the one provided
in Appendix E-13.
If voice recorders are used to collect data, it is advisable to use a headset or microphone which clips to a shirt {Bonsall
et al., 1988). Acceptable headsets are inexpensive and quite useful.
:jj
"
I
!'
Observers can collect gap data during weather rhar does not affect normal traffic volumes. Observers need good visibility ro the reference point but also need to be inconspicuous to avoid influencing driver behavior. It is advantageous
for many reasons, including inconspicuousness, for the observer to sir in an auromobile during a gap acceptance
study, Buses and trucks usuilly do nor block the view at unsignalized intersections, so usually observers can sir in an
automobile.
:l
~:
H
i!
:ii
1!
:j!
~
!i.
~.
1!.
lij;h
~! I
~~ I
~
~
The size of gaps in a traffic stream depends on the traffic volume, speed on the major approach, grade on the side
street (minor approach), number of lanes to cross and the median width. Because volumes change over any given day,
an analyst must sample gaps during each period of interest that has a volume different from those of adjacent periods.
The mean gap bas ~mly marginal meaning in analyses using gap data. Stuistics that describe the shape of the gap
distribution, such as percentiles, are more useful.
1:
~;
~;
;.
) '
''
f
~:
!~
H
}i
11
':j!
~
Lags precede gaps because a gap is measured between two consecutive main street vehicles, whereas a lag is only concerned about the rime before the first main street vehicle arrives. Gap acceprance srudies are conducted at locations
such as two-way stop controlled intersections or roundabouts co determine the critical gap (or minimum gap) for
capacity calculaciom or for calibration of simulation models.
The simplest procedure for collecting gap acceptance clara with typical agency equipment requires an observer with a
count board, laptop. PDA, or video. If a PDA or video is used in the fidd, a technician with a computer in the office
would need to record the data into a computer so it can be easily manipulated during analysis (adapted &om Shanteau, 1988). The observer would strike a key indicating (or say "major" in the recording device) when a major street
vehicle passed the reference point and strike another key indicating when a minor street vehicle appeared at the stop
bar (not at the back of a queue). Between the appearances of major and minor vehicles, the observer would also key
whether the gap or lag was "accepted," "rejected," or untested" using other specified keys for the specific data entry.
The observer should record "header" informacion at the beginning of the 6.le or speak into the recording device being
used periodically during the session. If a recording device such as a PDA or video is used, the technician in the office
must listen to the tape at the same speed at which it was recorded. A relatively simple program determines whether a
gap or lag was observed, computes the gap or lag time and swnmarizes the distributions of accepted and rejected gaps
and lags. The tape recorder must run u a comtant speed, the computer clock must be accurate and the technician
must manually edit the data (removing discrepancies by replaying the tape) before the analysis. For this reason, many
analystS feel more comfortable coUecring the data in real time, supplemented with a video camera to account for any
110 MANUAL OF TRANSPORTAnON ENGINEERING STUDIES, 2ND EDmON
. potential errors. In chis way, were the analyse unsure whether che data was collected correctly, it could b e error-checked
. with the video camera record.
' Analysts may require more information from the gap acceptance srudy besides that described above. For example,
the gap acceptance behavior of truck drivers may be ofincerest. Analysts may alter the audiotape proc edure described
above to include such variations.
A$ wich gap srudies, data collected in 2-sec bins are adequate for most gap acceptance srudies. Ramsey and Routledge
(1 973) suggest that 2-sec bins requ.ire a sample of200 acceptances, and 1-sec bins require a sample of 5 00 acceptances
(with a somewhat higher-quality result for the 1-sec bins). Observers can also collect gap acceptance clara with laptop
computers at the intersection or with videotape that has an on-screen dock. At intersections with low volumes, two
observers with a stopwatch and a form can usually collect gap acceptance data successfully.
Tht Rmnuy and &utkdgt mtthod hq;i1u with convmion oftht acctpranc~ + rtjtctions data abtwt f4 a ptruntagt and ml:rt1114t
int4 column 1 ofTabk A, h~ltJw.
Table A
'
Colwan Number
30.0
0.0
0.0
0.0
0.0
J
j
25.0
35.7
0.0
0.0
0.0
22.5
32.1
50.0
0.0
0.0
20.0
28.6
44.4
88.9
0.0
2.5
3.6
5.6
11.1
100.0
100.0
100.0
100.0
100.0
100.0
TOTAL(%)
Tht pmmtaga in column 1 ofTabk A trprt1mt tht distrifxaiqn ofaccq>~dgaps that would he ohrervedifaU driwrs httd a critic.J
gap ofO stconds. Ifall drivm httd criJiadgaps of2 stconds, tht first mtry in column 2 would ht 0% htcaust noru ofrhem would
accq>r a 1-s gap, the stt:ond mtry would he 25/(1()()..30).; 35.7%, t& third mtry would he 22.31(100-30) 32.196, etc.
Table B contains the number of drivers with each criliol gap size that accept a gap of a given size. First, column S
ofTable B is filled out with the number of accepted gaps from the raw data. Next, the first entry in column 1 can be
made; since 5 acceptances of a 3-sec gap were recorded, all of those drivers mwt have had critiol gaps of2 sec. Next,
the total of column I can be derived from this first entry and Table A; since 35.7% of all drivers with a critiol gap of
2 sec saw and accepted a gap of3 sec, the cow of column 1 is 5 x 100/35.7" 14.0. Then,-the rest of column I can
be completed by wing the percentages from Table A on the total in column 1, and Table 8 looks like:
TableB
Cola.m.o Nanibu
Total
5.0
5.0
4.5
10.5
15.0
4.0
9.3
11.7
25.0
1.2
1.5
1.9
5.0
21.0
13.1
0.5
9
I
TOTAL
14.0
-,u-",~r.s.:?:.i!~r<'l.
1.9
50.0
-~'26'!!( ~-}~l~-,~~t, -I -_.. - ~00% ~&i
~...... -.,l
' ' , "" to( 1 '.;,.\.t.O.."'-:. :~ ~.~-i,._.,. . .'_
'. ...:f -~ ---.-, ii -7. ~-...J~
-~ -3?~ -~~
.I .
. ..
. ._.-- ~..; :
+--~:~----
'The fim mtry in colum 2 is 0, siN:t drivers with criticalgapt of4 ~con4s Jo 11JJt ~etpt gapt qf3 ttron4s. The ttanui mJTy ~
bt tht tqta/ number of~ctpttdgaps of5 ~Nmds rmmu tht number ofJrivm with criticalgapt o/2 stN!nds who acctpttdgaps of
5 ttcon4s, of15- 4.5 e 10.5. Ntxt. tht tot41 ofcols.m 2 can be tltriwd just Ill tht total for cols.mn I Will dtriwd tmce 1111 mtry
Wlll/moum:. 10.5 X 100/50 21. 0. The mrutinJn ofcofum11 2 is compfmJ UU tht rmuzinJn in coJum 1 Will DN:t tht total Will
!miiiV1L Columnt 3 and 4 llTr complmdjust like cols.mn 2. TIN t<JW rtn11 ofTable B is tht distribuJion qfmric4J gaps 11m1111g tht
50 tiinJtrJ m1M lll11fPle and C/111 be wwJe into percmtaga The procedun C/111 JISt inttnlais orkr thtm 2 uconds. CII1IIINI}yu lag
r1Z1her than gap tmd Cll1l txamint tUrJ sizes qfltl/1 and lags.
It should be noted that this method will not work if the proportions of accepted gaps do not increase. This should not
be a problem since, theoretically, pedestrians and drivers should accept larger gaps more frequendy than smaller gaps.
If this is problematic, it is probably due to error in data collection, or there is an insufficient sample size. A reference
sprea~heet is provided in Appendix E-45 which is very hdpful when using this method to determine the critiol gap.
~- :
[:
6.0 INTERSECTION SIGHT DISTANCE
:t
I
i
.t '
Sight distances on approaches are aitiol to safe interseetion operations. The American As;;ociation of Scm Highway and
Transportation Officials (AASHTO) (2004) provides recommendations fur minimum sight distances ar in.tersections. Provisions should be made to account fur proper interseaion sight distance (lSD) continuously along each highway and meet
so that sufficient time to srop is av:Wable. The ISD is very imporcmt at all interseetion approaches and is aitiol to intcnc:ction operation and safety. The provisions provided in AASIITO have many cli.H'amt scenarios fur olculating ISD; the most
common are prcsenco:l in this chapcer. This scaion should be used as a tool fuc undctst:anding ISD and olcul.uing the
necessary daea for calaloring necessary sight distance requimnents. Ifaltcm.a.IM scenarios are necessary, the reader should
ronsult AASIITO.
-1
''
i;~~
!'
j
j
!.
\! !
~AAMI
lAI
l"'\~
TI)4MC:Pf"\ATAT1tit..l Ct..lt";t,_tCtOit\lt:
C'TIIf"\lt'~
~f\U"\
Cf"'rTir'ltd
6-7. The no-conaol or yield-control sight triangles are depicted in the upper diagram and depict the approach sight tri
angle. Stop control is showri in the lower diagram and depicts the departure sight triangle. Each is defined below.
Approach Sight Triangle: The uiangular area that should be dear of obstructions for any approaching
vehicles so drivers can see potcntiaUy conflicting vehicles in sufficient time co stop before colliwng within
the intersection.
Departure Sight Triangle: The uiangular area that should be clear of obstructions for any departing
vehicles stopped at the intersection and trying to enter or cross the major road.
ce-land ll
No control- yfeld control on min or road
Mefo rroed
_::;;
---...._
a;--- .
'
'11/)tq';;;---
I
I
I lllllnor roed
I
I
I
I
I
1-'
t::
I
I
I
d,
. I
::sa:
--.... ....
0
Cl>
f&~"''~ .
d:z
ca- m
_..,..-
jEF=
~:
F
li,.
d = ad.
b
d. -b
ii
~;
'
Equarion 6-10
If the speed corresponding to d from Exhibit 6-8 is lower than the current design speed of the minor road, the.sight
distance is inadequate.
~\I'
,,,~
!~
~i
1:
lh :
'
,,I
;I '
II
t''
Sourc~: Crearod
:!
1!
ISO = 1.47
~.! ;
p
J
I
'I'
91
by author using information from AASHTO's A Policy on Geommic Dnign ofHighwtiJS and Strem, 2004.
l;
j f
.'
'
v...,;... t1
Equation 6-11
Where:
ISD
= inrersectioo sight distance (length of the leg of sight triangle along the major road)
v...",
tg
= time gap for the minor road vehicle to enter the major road(s)
Typical time gaps recommended by AASHTO for passenger cars, single-unit trucks and combination trucks are
7.5 sec, 9.5 sec, and 11.5 sec, respectively. This assumes a stopped vehicle rurning right or left onto a rwo-lane highway with no median and grades of3 percem or less. However, the time gap value for the minor road vehicle (tg} varies
depending on the type of control used, the type of vehicle analyzed (passenger car, single-unir truck, combinarion
truck), approach grades, the number of opposing lanes, and the type of movement (left, right, or thru moverneor).
114 MANUAL Of TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
;Because the material is discussed in detail in AASHTOs "Greenbook," it should be consulted for the exact rime gap
icalcularion so the min imum recommended ISO can be determined (MSHTO, 2004). Sight distances sh ould be mea:sured in the field to compare against the minimum lSD values determined using the AASHTO "Green book. Jf che
}neasured sight distance is found to be lower than the recommended sight distance, :~.gencies should consider removing
sight obstacles, reducing approach speeds, changing TCDs, or taking other action.s. Usually, rwo observers on foot
using some form of distance measuring equipment record interseCtion sight distance data for each of the movemenrs.
AASHTO recommends the observers use a height of3.5 ft. (I m) off the ground and step back 8 to 10 ft. (2 to 3m)
from the stop bar on the minor approach for departing vehicles.
7.0 SUMMARY
In this chapter, methods used to perform intersection and driveway studies were discussed at length. Topics including delay, queue length, saturation Bow and lost time, gap and gap acceptance studies and sighr distance studies were
discussed. Specifically, various equipment needs md personnel requirements were given for each study, and field procedures used to perform the srudies were discussed in derail. For further information and derails on a specific studY
type, the reader should refer co references in the specified section in question.
8.0 REFERENCES
American Association of Srace Highway and Transportation Officials. A Policy on ~ommic Design ofHighways and Strms,
Washington, DC: AASHTO, 2004.
Berry. D. S. "Discussion: Rdatioruhip of,Signal Design to Discharge Headway, Approach Capacil)', and Delay," Transportation
Rn_earch &cord: journal ofthe Tra!IJjJortation '&uarch Board 615 (1976).
Bonsall, P. W., F. Gbahri-Sarcmi, M. R. Ttght and N. W. Marier. "The Performance of Handheld Dara-Caprure Devices in
Traffic and Transport Surveys, Traffic EngiMmng and Conrrol No. 1: 10.
Box, P. C., and J. Oppenlander. Manual ofTra.ffic Enginemng Srudier, 4th ed. Washington, DC: Institute of Transportation
Engineers, 1976.
Buchler, M. G., T. J. Hicks, and D. S. Berry. "Measuring Delay by Sampling Queue Backup," Tranportlltion Rnearrh Rmrd:
journal ofthe Tramportation Rnearch Board 615 (1976).
Federal Highway Administration. Traffic Signal Timing Manual. FHWA-hOP-08-024. Washington, DC: FHWA, 2008.
Hewitt, R. H. "A Comparison Becween Some Methods of Measuring Critical G~ps, Traffic EngiNmng and Omtroi 26, No. 1:
13-22.
:
Institute ofTrmsportarion Engineers TE Technical Committee 5P-5A, "An lnfomutional Report: Capacities ofTriple Left
Tum Lanes," ITEjoul7llli, p. 37. Washington, DC: ITE. May 1995.
Institme ofTransportation Engineers. Traffic Enginming Handbook, 6th Edition. Washington, DC: IT, 2009.
Ramsey, ]. B. H and I. W. Routledge. "A New Approach ro Analysis of Gap Acceptance T uncs. Traffo Enginemng Conm/15,
No.7 (1973).
Reilly, W. R., C. C. Gardner, and J. H. Kell. A Technique for Measurnnem ofDelay at lntermtiont, Vol. 3, User's Manual,
H{WA-RD-76-137. Washington, DC: Federal Highway Administration, September, 1976.
Roess, R. P., E. S. Prassa.s and W. R. McShane. Traffo EngiMmng, 3rd ed. Englewood Cliffs, NJ: Prentice Hall, 2004.
Shanteau, R. M. "Using Cumulative Curves to Measure Saturation Flow and Lose Tune." ITE]ouma/58, No. 10 (October
1998).
Teply, S. "Accuracy of Delay Surveys at Signalized Intersections, Transportalion Rnearch &cord:jounud ofthe Tra!IJjJortarU!n
Rnearch Board 1225 (1989).
lnte~ection and Driveway Studies
11S!i
Teply, S. and G. D. Evaru. "Evaluarion of the Quality of Signal Progression by Dday Dlsrributioru." Transpor141ion &uarch
&cord: Joumal ofthe Transportarion &search Boart/1225 (1989).
Tran.spon:acion Research Board, Highway Czpacily Manwzl, Washington. DC: TRB, 2000.
.,.
;i
i'
I
Chapter 7
Dontu~
EJiutlby:
<:brimlpher M. c-nmglunn, MCE, P.E.
1.0 INTRODUCTION
117
118
~.0
4.0
5.0
117
1.1 Purpose
118
119
123
124
3.2 Signs
132
135
135
136
136
136
137
137
137
139
140
7.0
SUMMARY
140
8.0
REFERE NCES
140
140
141
1.0 INTRODUCTION
1.1 Purpose
(MUTCD) defines trafJic control tkvic~s (TCDs) as a11 signs,
signals, markings, and other device$ used to regulate, warn, or guide traffic, placed on, over, or adjacent to a
ruec:t, highway, pedestrian f:aci.ljcy, or bikeway by authority of a public agency having juri.!diccion (FHWA.
2003). The: general purpose ofTCDs is to provide visual information to the road user. TCDs are used to. help ensure
the safe, orderly and efficient movement ofall-types of traffic.
-
Devices are classified inco groups that regulate, guide, or warn craflic. &gulaiQry tkvim inform the road user of regulations that
are in force, instruct the road user to rake some action, prohibit or permit the road user from making certain maneuvers, or assign
the right of way. Gttuu tkvicn typically are used to identify routes; provide rravder directions; delineate the roadway; and provide
information on facilities, services, points of interest and political boundaries. "Wtming devic/!1 provide notice of unexpected conditions. They draw auention to the presence of geometric fearures with potemial hazards, major changes in roadway character,
obstructions or other physical hazards in or near the roadway and areas where bazard.s may exist under certain conditions. They
inform the motorist of regulatory controls ahead and advise drivers of appropriate actions.
I\
_,ll :
('
i
;!( i
In the United States, the MUTCD defines the basic principles that govern the design and use ofTCDs. The MUTCD presents
TCD standards for streets and hlghways open to public uavel, rc:gardless of che cype. class, or govemmencal agency having jwisdic.tion. While the MUTCD is not a starute, it carries the power of a statute in ddining national standards. Many jurisdictions adopr
the MUTCD without revision; others modify or eliminate SJXdhc designs, applications, or requitemen!S by state l~larive action .
Frequendy, modifications rdiea mote stringent requitemen!S than the minimum expressed in the MUTCD. Equivalent state and
local manu& that meet or exceed the MUTCDS minimum requiremen!S also carry the power of a starute (FHWA, 2003).
.I' .;'
'
1t;
:;, :
.
~I :
~ I
.fll.l '
l~
. ;!
'
~. .
. I
Readers of this chapter should note that similar manuals exist in other counaies; however, the sheer volume of other studies and differences in each manual are not able to be presented as part of thls chapter. Therefote, if outside the United
Scates, analyses should seek specific guidance on procedures for installing various TCDs.
The content of this handbook was finalized prior to rhe release of the 2009 Edition of the ManUAl on Uniform Traffic Control Deuim (MUTCD). Therefore, although all general refeccnces to the MlJfCD have been updated to 2009, content
with specific references to the 2003 MUTCD has been retained. Users are encouraged to consult the current edition of the
MUTCD when making technical determinations. A free copy of the 2009 MUTCD (POF or HTML format) is available
at http:l/mutcd.fhwa.dot.gov. ITE, in cooperation withAASHTO andATSSA publishes a hard copy version available via
the lTE bookstore at www.ite.orgfbookstore.
!, ,_
The MUI'CD setS out general requirements for the design, placement, operation and maintenance ofeffective TCDs (ITE, 2009).
it
D(sign: TCDs must be designed with the combination of physical features (size, color and shape) needed ro
command attention and convey the correct message.
:.~o~
li t
PIAcm~mt: Devices are placed to fall within the road user's cones of vision so the devices are able to command
attention and allow time for driver response.
tl !
.i
,;,.
~f '
j
J,
~~
d' '
r~
Operation: Devices must be employed in a way that meers traffic requirements in a uniform and ~nsistent
manner, ful1i.Us a need, commands respect and allows time for response.
Maintmant:(: Devices must be maintained to retain Iegibiliry and visibility. Devices that are obsolete or are no
longer needed should be removed.
Application and operation oiTCDs should be uniform. Similar devices should be used for similar situations and in similar
locations to minimize road users' confusion and gain their confidence. The inappropriate or overuse ofTCDs can lead to
a number of problems including driver disregard, increased dday, excess fuel consumption, increased vehlcle emissions
and increased crashes. Concrol devices should supplement each other by providing a meaningful message to motorists and
should be designed and placed so they srand out from the environment..
J~ ~
g I
~.
:\ .
~' ;
::
;i
. 1;
~ ,'
.'
l
~ I
ri
J~ j
.TCD studies may be conducted at any location where excessive delays, excessive speed, or other traffic problems have
'; been observed; where there have been citizen commentS or complaintS; or where analysis has indicated that rraffic
!contwl will be needed to accommodate furure demand. Most state agencies have devdoped their own procedur<S an d
guidelines for conducting engineering studies based on state policies. In addition, both federal and state funding programs carefully define the cype of project studies required to be eligible to receive financial assistance in implemtnriog
traffic control and safety improvements (lTE, 200 I).
.
Most traffic studies require a_ thorough d10scription of the study sire. The extent and detail of the information nee4ed
depends on the analysis to be performed. A conditions diagram and location plan show the derails o f the physical
layout, including such features as intersection geometries, channeliv.tion, grades, sighr-<iistance restrictions, bus srops
and routes, parking conditions, pavement markings, signs, signals, street lighting. driveways, location of nearby rail
road crossings, distance to nearest signals, ucility poles and fixtures and adjacent land use. An example is shown in
Chapter 17, Exhibit 17-10.
2.1.2 Collmon Studies
Crash information is used to iq~nrify and analyz.e high collision locations, conduct before-and-after studies, evalua.cc
requests for additional craffic contiol; cv.ai!Jate roadway features, identify and rank improvement projects, establish
and maintain traffic regulatory devices and identify need for poliee surveillance and enforcement (ITE, 200!).1'he
information cypically required includes collision dia.grams, spot crash location maps and collision rates. Collision diagrams summarize crash experience of a specific location by crash type, location, direction of vehicle movemenr, sev~r
ity, time of day, dare and day of week. A single collision diagram should summarize data for at least 1 year (preferablY
3 ro.5 years) in a format similar co that of Exhibit 17-19.
Spot collision location maps are used to identify high crash locations. A spoc map is a quick visual method ofide.J:l
tifying locations and collisions that may warrant derailed analysis. Crash lo~tions are marked on a map. This is ~
excellent application for GIS technology. The rype, severity and time of collision may be coded using colo~ and or
symbols.
2.1.3 Volll1.ile Studies
Volwne data :ire required for most TCD studies. The specific volun?e data required and details of data collection a.e
determined by the purpose of the study. Traffic counts are an important factor in cv.aluating improvements and recocrJ.mendations. Vehicular volumes may be collected and grouped by movement or by approach. Studies may be limircd
to specific time periods (such as peak hour), direction of travel, or geographic location. Studies may be conducted
specifically to establish axle counts or vehicle cypes. Chapter 4 contains derailed information on the collection, reduCtion and presentation of volume data.
When n~ int=tctions and roadways are being planned or where major construction projects will be implemented i .1:1
the near future, actual traffic volumes cannot be counted and therefore muse be estimated. The ITE publication Trij
GmertltUJn OTE, 2008) may be hdpful in deriving trip estimates for this purpose. Additional methods for esrima~g
future traffic volumes or projecting sample ~untS are described in Chapter 4 of this book, the Traffic Engintm?"$$
Handb()(J/t (TEH) OTE, 2009) and the Transport4tion Planning Handboo!t (fPH) (ITE, 2009). Roadway volun?-e
projections arc generally expressed in rerrns of average daily traffic (ADT). Although peak-hour Bows can also be e5
timared, the hourly d.iscribution is not normally available. Consequently; minimum requirements based on estimate~
ADTs (EADTs) may be obtained from values given in Exhibir7-1. These values are based on the assumption that th-e
eight highest hours will each exceed 6.25 percent of the ADT (equivalent to 500 vehicles per hour (vph)). This~
vary by agency (for example, Texas uses 5.6 percent of the ADT, which results in ADT values 11 percent higher tha..$1
those shown) (Kdl and Fullerton, 1982).
TRAI'FIC SIGNAL
I~ ~n
URBAN
WARA~S
' RURAL _ _ _ _ __
EADT
t. Minimum Vahieular
Not Satlofled - - - - -
S.tlofied
Number of -
--o-1
direction only)
--o.ch
0#
M IY\oOre .............. ...... .
Minor Street
1 ... ................" .. - ....... .
1 ................... ............ ..
2 or mOte ...... ............. ~.
1 .........: .......................
2 or more .....................
11-'Jor Street
1 ............................... ..
2
2
Urt>a1
8,000
8.800
8.600
11,000
Rurel
Urban
Ru.-.1
15,1100
8,720
8,720
6,1100
2,400
2,400
3,200
3.200
1,880
1,880
2,240
2,240
Not Slltllfled - - - - -
. - (toul of both
appro~hesJ
direction only)
Minor Street
Urban
1 .... ............................ .
2 ot more .................. . .
2or more ................... .
1 ............................... ..
12,000
14,400
14.400
12.000
1 .................................
2 or more ....................
2 M tnOre ................. ..... .......
3. Coll'blnatlon
Slltitfied
Not S.tllfled - - - -
Rural
&,400
10,080
10,0110
8,400
2 WR"anta
Urban
Ru,..l
1,200
1,200
1,.800
1,.800
1,120
1,120
860 '
8150
2 WarrenU
2
NOTE ;
.
1. Left. turn movements from the mJor nreet may be lnctuded with minor nrwet volum.. If seperete
aignal ph... is 10 be provldi for the feft..turn movement.
2 . To be uoed only for NEW INTE ASECTIONS or other locetlona where actual traffic volumes Cllnnot
be counted.
Exhibit 7-1 provides three possible signal warrants for new intersections where aaual traffic volumes are not known:
1.) Minimum vehicular, 2.) Interruption of continuous traffic ft.ow and 3.) Combination warrant. For instance, the
minimwn vehicular warrant has four possible combinations of major and minor roads to chQO.Ic from. If the major
and minor roads only have one lane each, and arc located in a rural area, the minimum nwnber of vehicles that must
be present on each approach are 5,600 and 1,680.
Pedestrian volumes.may be recorded with vehicular volumes or on a sepa=e sheet. For signal stUdies, counts should be
taken on each crosswalk during the same periods as the vehicular counts and al.so during hours of highest pedesaian volume.
For other signal stUdies, pedestrian volumes crossing the major street should be sufficient. Where young or dderly persons
need special consi<kntion, the pedestrians may be chssified by general observation and rc:cordcd by age groups as rollows:
Under 13 years
13 to 60 years
Over 60 years
\I
mr~
FmT1mo
fhe collection of pedestrian volume data is described in more detail in Chapter 12.
:n
flW:dftifi?~l~ll:~t~
J ~
43
58
46
47
45
;;j '
45
;l:
l.,l .
'i! .
I' '
l'l
I
57
48
51
44
46
53
52
36
57
'I'
!
~
;p'"
,,~ ;
u
. IJ)I..
j;
: I
!I'
Sam
II
Category (I)
Range
II
<35 kph
35.1-40
40.1-45
45.1- 50
50.1- 55
55.1-60
>60.1
6
12
12
17
13
7
8
75
35
40
45
50
55
60
11%
60+
II%
r~
...5 10
,1 '
""
lli
.
L
l
8
6
4
2
"'v
M
~
;6
......,"'0
'I'
"'.
0
It)
'I'
I
1:
Frequcocy
Distribution
8%
24%
40%
61%
78%
89%
100%
16
14
12
'~1'.
i
63
38
37
44
49
36
43
59
47
37
47
32
54
64
53
Cumulati....:
Frequency
Distribution
8%
16%
16%
21%
17%
Upper Limit of
Total
iji -
61
43
31
46
48
37
62
37
42
39
51
32
47
61
59
Samples (n) 75
47.44
3,558
Mean
~~
~I
54
58
34
53
47
61
54
62
48
54
47
37
50
42
48
53
41
63
52
42
32
36
39
55
34
43
50
47
37
58
ttl
0
ttl
...~
It)
10
CD
A
Travel-time delay studies look at the amount of time lost by a vehicle through a roadway system by comparing che
time a vehicle travels through a roadway system to the rime it would have taken if it had traveled at the desired speed
wirh no delays. Three common studies are che test vehick, vehick observation and probe vehick methods. These methods can use various techniques for collecting data, from simple daca collection forms, count boards and softwar(b3Sed
programs using an internal clock to more complex methods used in probe studies that rely on transponders or cellular
devices. Delay srudies are discussed furcher in Chapters 6 and 9.
2.1.6 Gap Distributions
The ability of vehicles to enter a major scree~ from a side screet or driveway, and often the ability of pedestriaJJS. to
cross at an unsignalized location, depends on the distribution of gaps in the traffic .stream. Ifgaps of adequate lwgth
are infrequent, it may result in unacceptable delay for vehicles arrempting co enter the scream. Some road counters
can record the gap distribution as well as the axle count, by time of day. Gap distribution can also be observed in the
field manually by a person with cem..in eleccconic count boards or using a portable computer with che software-based
code using the internal clock of the computer. Gap acceptance and gap studies are discussed further in Chapter 6.
2. that driver response to the influences of a TCD has previo~ly been identilied by observation, fidd experi~
ence and laboratory test under a variety of traffic and driver conditions.
Th~ Ml.JI'CD presents warrants as a series of guidelines that should be used co help evaluate the situation ar hand. ~
not as absolute v..Iues. The satisfaction of a warrant does not guarantee that a TCD is needed. Similarly, failure tO
fully sawfy a specific warrant is not positive proof that the device could not serve a useful purpose. The application o:
warrants is effective only wb.en combined with knowledgeable engineering judgment. The MUTCD describe war-rants, design and placement criteria for a wide range ofTCDs. Warrants for signals and signs are summarized btlow-In addition, state and local jurisdictions may have developed their own warrants for applications not included in ch~
MUTCD, including loading zones and speed bumps.
12~
It should be noted that traffic signals do not always have a positive effect on roadway operations. A:n. unjustified, poorly designed, improperly operated, or poorly maintained traffic signal may result in increased collision
frequencies, excessive delay, motorist disregard, decreased capaciry and circuitous travel by alternative routes.
Experience has indicated thac although the imtallatiOP of signals may result in a decrease in the number of rightangle collisions, there may be increases in the number of rear-end collisions. Consequcncly, a th9rough study
of uaffic and roadway conditions should precede the inscallation and the selcct.i on of signal control equipment.
.!
The MUTCD describes eight warrants for the installation of traffic signals. In many instances, state or local
jurisdictions have supplemental warrants that should be consulted in addition to the MUTCD. The MUTCD
contains the minimum rcco~endations for installation of a traffic signal. The MUTCD signalized intersection
warrant requirements arc summarized below. Traffic signals should be considered only if one or more of the signal warrants are met, or if good engineering judgment suggests a signal should be wed to serve a weful purpose.
The analyst sho uld keep in mind the MUTCD states "The satisfaction of a traffic signal warrant or warr:mts .
shall not in itself require the installation of a uaffic control signal (FHWA, 2003):
'
Condition B is intended for application where Condition A is not satisfied and applies to operati ng
conditions where the traffic volume on a major street is so heavy that uaffic on a minor intersecting
meet suffers excessive delay or hazard in entering or crossing the major street.
:~
.I
.' I
,!\
i!
arc present at an
To meet the requirements for this warrant, the vehicles per hour for Condition A orB in the 100 percent columns for major street and higher-volume minor suect approaches mwt be met. In applying the major and
minor street volumes, the same 8-hour time period should be used. If the posted speed limit or 85th percentile
speed exceeds 40 mph (64 kmlh}, or is located in an isolated communiry with a population less than 10,000,
the 70 percenc columns for both conditions should be used in a similar manner.
Alternatively, a combination Condition A antfB could be considered after all other remedial measures have been
reviewed. This measure is similar to the single condition analysis; however, it only requires chat the 80 percent
vehicles per hour measure apply to Conditions A and B. As before, in applying the major and minor street volumes, the same 8-hour time period should be used. If the posted speed limit or 85th percentile speed exceeds
1 'M
MANilA! O F TRANSPORTATION
40 mph (64 km/h), or is located in an isolated community with population le.ss than 10,000, the 56 percent
columns for both conditions should be used in a similar manner.
'.Warrant 1 is satisfied if either Condition A or B is satisfied. In extreme cases, if A or B are not satisfied, then the
Combination Warrant can be used.
~~~~~i(~.J~,.
Major Street
1 .... .... ....
2or more .
2or more ..
1 .................
1 ...........
. 1 ........ .......
2 or mora ..
2 or more ....
80%.
70%.
56%4
600
480
350
420
280
336
600
480
420
336
!500
400
350
2 80
500 \ -400
!2!!: 56o/ocl
120 . 105
120 105
84
84
160
140 112
160
140
112
Minor Stree t
1 .................
1 .................
2 o r more ...
2 or more ....
60%.
!16%.
600
720
720
800
!125
630
630
525
420
504
504
420
60
60
80
60
53
53
70
70
42
42
56
56
-1<1
u Md IIOf combrtaoon Of Condliol\a A and 8 lltler adequate trlal ot ocnr rernedtal meuurft. .
MayDa UH<IWI\<In the mlljor-Mtape<od .,..,... . 70 lcm/ll o r - 40 mpn or In en
conununiiY wt!h a
populdon oC -than 10.000.
May De uNCI for oomDinallan or OondlllonS A and e alter adequate 111a1 or .,....., maaJures w1>an tn m<o~<>r
SIJ. .t
70 1an1t1 Ot <40 ms:>t> or tn en loOiat<od ccmmtA'IIIy wttn populallon or less Chan 10,000.
- <><! .,__
Source: 2003 Manual on Uniform TTlljfo Control Dtvicc. Table 4C-l. Sect.4C.02. Page 4C-3.
n
-If
'
:c
a..
>
:c
I .
~.
<..:>
500
I:
1: i
tue5
wCC
400
I-
a..
300
~~
~3
200
a:
w
:c
(!)
:X:
100
.;:I
If.
I
lj
a;O..
en<
.,,s
~?
:,liI :
,:,r,!
so
400
300
500
600
700
800
900
1000
1100
1200
1300
1400
I ' *
_I.I I '
',f
'i~
I
ll
Sowce: 2()()3 Mamuzl on Uniform Traffic Control Dwices. Figure 4C I. Sect. 4C.04. Page 4C5.
~~
~
0
8~
~~
~~
300
200
z3
2 ~
~~
(!)
1:
!I <.
~~ l
~:I
~~~I
!l l
~l
100
200
300
400
I ==---~ S?SI
500
600
700
Sowu: 2003 Manual tm Uniform Traffic Control Dnnets. F.gure 4C2. Sea. 4C.04. Page 4C-5.
,, I
'
.l j
I
!
900
~!I
1,
800
.. '
'
1=:~
1'000
3.1.3 Wan-ant3:Peak-Hour
'; Warrant 3 is intended for application where traffic conditions are such that for a minimum of I hour of the day, the
',minor street traffic suffers undue delay in entering or crossing rhe major street. The warrant is intended to be applied
'to unusual cases where high volumes of traffic are occurring during a very limi red time period. These could include
office complexes, manufacturing plants, industrial facilities, ere. The peak-hour warrant is satisfied when one of rhe
cwo categories below is mer:
Category A is satisfied when all circumstances given below exist for 1 hour (any four consecutive IS-min.
periods) of an average weekday. All rhree provisions must be met.
1. The total stop time delay experienced by the uaffic on one minor street approach (one direction only)
controlled by a stop sign equals or exceeds 4 vehicle-hours for a one-lane approach and five vehiclehours for a cwo-lane approach tmJ
2. the volume on rhe same minor street approach (one direction only) equals or exceeds 100 vph for one
movi~g lane of traffic or 150 vph for cwo ?loving lanes and
3. the total entering volume during the hour equals or exceeds 800 vph for intersections with four (or
more) app roaches or 650 vph for intersections with three approaches.
Category B is satisfied when the plotted point representing the vehicles per hour on the major street (total
of both approaches) and the corresponding vehicles per hour of the higher volume minor street approach
(one direction only) for 1 hour (any four consecutive 15-min. periods) of an average day falu above the
curve in Exhibit 7-6 for the existing combination of approach lanes.
Al ternatively, if the posted speed limit or 85th percentile speed of major srreet traffic exceeds 40 mph (65 kmlh). or
if the intersection lies within a'built-up ea of an isolated community having a population less than I 0 ,000, Exhibit
7-7 should be used in place of Exhibit 7-6.
It should be noted that the peak-hour warrant is not accepted by all states, so checking with the proper authority or
reference is iroporranr if considering a signal based on this warrant alone.
iE
eoo
>
eoo
i~
400
5~
300
~~
200
~
~
' 150
'100
100
400
600
600
700
600
QOO
1000
1100
1200
1300 1400
1 eGO
1600
1700
1600
Source: 2003 Manual on Uniform Traffic ContrrJI Devim. Figure 4C-3. Sect. 4C.06. Page 4C-7.
~
:X:
ii
~~
200
~
ffi
100
:e-'
:X:
300
400
I. I
C)
=r-
I T
I:~~o
s:
500
600
700
800
goo
tooo
1100
1200
1soo
Source: 2003 Mmua/ 1111 Uniform Tmfo Control Dwices. FJgUrc 4C-4. Sect. 4C.06. Page 4C.7.
!, G=R
+ W/3.5 +2(N I}
Equation 7-1
\ where
G = adequate gap time, sec
Equation 7-1 uses a walking time based on 3.5 fps. The walking speed may be changed to more accurately reflect
observed conditions. R is the perception reaction time: the time required to look both ways, make a decision and commence the walk. A commonly used value is 3 sec. The term 2(N - 1) represents the pedesuian platoon time. Children
are assumed to cross a street in rows of five with an interval of2 sec btr.veen each row (ITE, 1972).
Next, the total survey time (7) of the vehicular gap survey is converted from minutes to seconds:
Finally, the percentage of the study time during which the gaps are of adequate size (D) is calculated.
D=
(T-t)
-r-too
.. ,,-IJ''"-'-'"-"""
'''";b;t~,;,.- ''!';;t~J,j,iJ-<-
.'~-~~;-. -~~~~~~)-~~>~f~);-_;;1~~- - ::1~~~;):-~<~,- .~ :-_..:.:,. ~~i'(:-
~~~~~~
- : e. _ ~
+_.
~----.
c _ ~-
__
.. __
. , _ , , - -
Survey Dace
Location
Crosswalk across
Number ofG"' s
Tally
Total
Multiply by ~ Size
11
12
13
14
15
116
117
18
19
20
21
22
23
1
ComJ>..uu.rion.s
G=
R + (W/3.5) + 2(N-1)
GG
_ _ _sc:c
Ta
to121 survey
timex60
Ta
_ _ _sec
o.
r(T-1)rn x 1oo
_ _ _%
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
Totals
Source: KdJ and Fullerton, 1982.
130 MANUAL Of TRANSPORTATION ENGINEERING STUDIES. 2ND EDITION
I=
Sul'Veydate
Crosswalk surveyed
Crosswalk across
Divided Roadway?
Yes
No
Number of rows
5 or fewer
6to 10
15
16 tO 20
11
21 to 25
26 to30
5
6
31 ro 35
(0
36 to40
41 tO 45
46 to 50
10.. ,
Tally
Total
Cumulative
Comp utations
Sour= ITE. A Programfor School Crossing Pro~ction: A R.tcommnukd Practice, 3rd Edition, 1971.
fa.r
2. On a cwo-way street, adjacent signals do not provide the necessary degree of platooning and speed conrro 1
and the proposed and adjacent signals could constitute a progressive signal system.
According to this warrant, the installation of a signal should not be considered where the resulting signal spacio.g
would be less than 1,000 ft. (300 m).
Five or more reported coUisions of types susceptible to correction by traffic signal control have occwrtd
within a 12-month period, each crash involving pe.rsonal injury or property damage exceeding the reportable crash thresholds in the state or local jurisdiction.
3. There existS a volume of vehicular traffic not less than 80 percent of the requirements speci.6ed in wanant6
1 and 4:
Traffic Control Device Studies 13-1
3.2 Signs
The MUTCD prescribes standards for the uaffic signing within the right of way of all classes of public highways.
Traffic signs &11 into three broad functional classifications according to use. These include regulatory signs, warning
signs and guide or informational signs. ~ch is defined as follows:
'
Regulacocy signs give notice of traffic laws and regulations.
Warning signs give notice of a situation that might not be readily apparent.
Guide signs show route designations, destinations, directions, disunces, services, points of interest and
other geographical, recrcacion.al, or culrural informacion.
Signs should be used only where warranted by facts and field srudies. Regulatory and warning signs should be used
conservatively because these signs, if used in excess, rend to lose their effectiveness. Route signs and directional signs
should be used frequendy because they promote reasonably safe and efficient operations by !=ping road users informed of their location. Signs arc essential where special regulations apply at specified places or :u specific rimes only,
or where haz.a.rds uc not self-evident (lTE, 200 1).
The MUTCD contains the complete, current warrants for a wide range of reguktory, warning and guide signs
(FHWA. 2003). The summaries below arc presented for convenience only and do not represent the entirety of available signagc options. The MUTCD or the equivalent local manual must be consulted for explicit requirements for
specific signs.
3.2.1 IUgrJtmn, s;xru
Regulatory signs inform road users of traffic laws and regulations. They arc to be installed near the location .at
which the traffic regulation applies and with adequate visibility to obtain compliance. Rl:croreHectivicy or illumination sca.nduds should be followed to allow similar day and nighttime viewing ofshape and color. Regulatory signs
uc used in many applications; however, this section focuses on stop sign, mulciway stop control, yield control and
speed limit warcanu.
Stop sign wamtn~ at an intersection require an examination of collision diagrams and knowledge of operating conditions at the site. A stop sign may be warranted at an intersection where one or more of the following conditions exi!t
based on engineering judgment.
I. inrecseccion of a less important road with a main road, where application of the normal right-of-way rule
is Wlduly hazardous;
4. unsignaliu:d intersection where a combination of high speed, restricted view and serious collision records
indicate a need for control by the stop sign.
Stop signs cannot be erected at intersections where traffic control signals are present, and they should nor be installed
for the sole pwpose of controlling the speeds of the motoristS.
Multiway slfJp control sign warrant.i incorporate numeric:al criteria for collision occurrence, vehicular volumes and approach speeds. Four-way or all-way stop sign installations can be used as a safety measure at some locations where the
volume on the intersecting roads is approxirnatdy equal and the following conditions have been established.
I. Where traffic signals are warranted and urgently needed, the multiway stop conuol is an interim measure
that can be installed quicldy to conrrol traffic while arrangementS are being made for the traffic signal
installation.
2. The:re is a.collision problem; as indicated by five or more reponed t~hes in a 12-mortth period of a rype
susceptible to correction by a multiway stop installation. Crashes would likdy include such types as righ~
and left-turn collisions or right-angle collisions.
3. Minimum traffic volumes must meet the following criteria:
a.
The total vehicular volume: entering the intersection from all approaches must average at least 300
vehicles/hour for any 8 hours of an average day; and
b. the combined vehicular and pedestrian volume from the: minor street or highway must average at least
200 units/hour for ~ ~arne 8 hours, with an average delay to minor street vehicular traffic of at least
30 sec per vehicle during die: maximum hour; but
c. when the: 85th pc:rcencile approach speed of the major street traffic exceeds 40 mph (65 km/h), the
minimum vehicular volume warrant is 70 percent of the foregoing requirements.
Alternative criteria could be considered if they exist, including:
l. problematic left-turn confliCts;
3. sight-distance problems at the site causing the driver to take inherent risk negotiating the intersection unless the other approach is required to stop; and
,
4.
two residential roads inrcrsect, each having similar design tnd operating characteristics, where mulciway
stop control would improve the ch.aracreri.stics of the intersection.
Yield sign w<tmmts may be used instead of stop signs to assign right of way to traffic on certain approaches. The
MtiTCD recommends yidd signs may be used ifengineering judgment indicates one or more of the following conditions exist.
1. Given the available sight lines, enough time is available for. a road user to come to a safe stop driving at the
posted speed_. 85th pc:rcentile speed, or statutory speed.
2. A merging maneuver in the entering. roadway docs not provide adequate accderacion geometry and/or
sight distance for merging maneuvers.
3. Within an intersection with a divided highway, where a stop sign is present at the entrance to the first
roadway and further control is necessary at the entrance to the second roadway, and where the median
width bcrween the two roadways exceeds 3~ ft.
4. M intersections where a special problem exists and where an engineering scudy indicates the problem to be
susceptible to correction by the use of the yield sign.
Typically, yield sign applications take place at two-way srop-concrolled intersections or channelized turn lane applications; however, it should be nored yield signs should also be used at the entering approaches of all modern roundabouts, which are becoming more common across the coumry.
Spud limit rign warrantr $hould only be wed after an engineering study has been conducted based on established
engineering practice, regulation, or as adopted by the aurhoriz.M agency. The MUTCD recommends posted speed
limits should adhere to the following guidelines.
I. Speed limits should be posted in multiples of5 mph (10 km/h).
2. State and local agencies should re-evaluate nonstatutory speed limits every 5 years if roadways have undergone significant changes in land use or the roadway.
3. When a speed limit is posted, it should be within 5 mph (1 0 krnlh) of the 85th percentile speed of freeBowing conditions.
4. Other options exist that can be considered when establishing the posted speed limit. These include: roadway characteristics, shoulder condition, grade, alignment and sight distance;
5. the 'pace' speed;
6. development along the roadway;
7. parking practices;
8. pedestrian movement and activity; and
9. 12-month crash ~perience.
3.2.2 Warning Signs
Warning signs give notice of a situation that might nor be readily apparent. Care should be taken to use warning signs
sparingly and only when necessary so compliance is nor degraded. If warning signs are used to provide informacion
during seasonal events, they should be removed or covered when the condition no longer exists. Typical applications
of warning signs include:
, and towns with directions and distances. The MUTCD provides very basic instruction on the types of guide signs;
',how chey can be assembled; background and lettering colors/siu:s; and limiced information on location of very specific
'signs such as "distance" signs. The engineer should consulr rhe MUTCD directly for this basic informacion on rhc:
'guide sign in question.
lAAPJ.IIAl
r\~ T~AM<Pn~TA.Tif'\f\.1
, dirca measurements with count tubes, radar or laser speed guns, or calculating the rime a vd:ticle travels over some shon
\ (predetermined) distance. TCD studies usually use individual vd:ticle methods because a relatively small sample ofspeeds is
' enough to infer if the device had any dfca-positive or ncgative-<>n speeds. Otaprer 5 lays out the basic &amewodc for
'conducting a spot speed study and should be consulred if the TCD is expected ro have some effect on speed.
~~~~wt~
.'<.
t1$~~:$-.~.
Sheeting Typo ASTM 0495604)
Sign Color .
White on GrGQn
Black on Yellow
or
B l ack on OranQe
White on Red
Black on White
(!)
(!)
Q)
(!)
Beaded Sheeting
I
W':G0!:7
W';G2:
Y';O'
v:o
Prismatic Sheeting
II
Ill, IV VI VII, VIII, IX X
Ill
W':G2: 151 W';G2:25
W:t250:G2:25
W2:120;Glt15
Y2:50;02:50
Y<!:75;0<!:75
W2:35;R2:7
YV;tS
Additional
Criteria
:Nerhead
Ground-mount a
Q)
(!)
(!)
The minimum maintained retrorellecllvlty l evels shown In lhls !Able are In units of cdllxlm' measured al an
observation angle ol 0.2 and an entrance angle of -4,0.
For texl and line symbol signs measurng at least 1200 m m (48 In) and for an sizes ot bold symbol signs
For text and line symbol signs m easuring len lhan 1200 mm (48ln)
Mlnlf!'lum Sign Contrast RaUo 2: 3:1 (while retroreflecUvlty .. red retroreftecllvlty)
This sheetlng type should not be used for this color lOr this application
Sign re~ordlectivicy standards will be increasingly important to highway agencies because each agency will be required to implement a sign assessment or management program by January 2012, with sign compliance by January
2015. Thi.s section covers assessment and management methods used to evaluate when signs should be replaced, and
includes visual inspection, taking actual reuordlectivicy measurements, service life, fixed time period and a control
sample. Other methods can be used as long as they are documented in an engineering srudy and meet the thresholds
provided in the M'UTCD. The: only signs that may be: excluded from recroreflectlyicy standards are parking; standing, walking, hitchhiking, crossing, adopt-a-highway, any sign with a blue or brown ~ackground and bikeway signs
intended for exclusive use by bicycliSts or pedestrians.
6.1.1 Assessment Muhods
The actual assessment of traflic signs can be done by visual or measured methods. VISual inspection is the preferred
method because it can be done much faster and is generally easier. Visual assessment is conducted by a trained inspector
from a moving vehicle at posted speeds. The inspection should only be done during nighttime: conditions with low-beam
headlights. The inspection could be done in a number of ways; however, the typical method of assessment requires the
inspectOr to use a set of signs that meet minintum threshold requirements for rettoreflectivicy as a base for nuking decisions about 6dd measured signs. These signs arc: viewed prior to conducting a nighttime fidd srudy as a "calibration" for
. each inspector. These calibration signs could be insralled in a maintenance yard or at known locations along a corridor
where various signs are located. Trpically, the inspector uses the amount of time he or she can read the sign at the posted
speed as the threshold for replacing the sign.
An alternate method of assessment is to measure acrual retroreflectivicy using a retrorefiecromerer. This method is
more time consuming; however, the advantages are that assessments can be conducted during daytime or nighttime
conditions and the actual measurement is more exact
than a visual inspection. A direct comparison is made
to minimum retroreflectivicy levels for various sign types
noted in the MI.JI'CD. Trpically, a portable rettoreflectometer is used to conduct the srudy, such as the one
shown in Exhibit 7-11. Newer methods are cunencly under devdopment to try to collect actual measurements
from a moving vehicle. At the time of this publication,
there was limited success using this method of measurement; however, fururc efforts may have more success.
6.1.2 Mantlgemmt Methods
In many cases, it is easier for an agency to have a management plan in place for replacing signs. In this case, e:~.ch
individual sign is not assessed, so signs are frequently replaced before minimum rctroreflecrivity measurements arc
reached. Three methods for managing sign maintenance are
brieRy covered in this section..
Source: DELTA.
: FllSr, signs ace often replaced based on their e)Cpected service life, usually indicated by the sheeting warra my period, rnea~ surements from field-insralled signs over some period of time, or an alternate method. This method re<juires a d3tabase
to be kept on signs' age and placement so they can be removed after the appropriare period of time. Another method is
. to replace all signs of a certain rype, or in a specified area, after a specific rime period has elapsed (such as rhe warr3llry
period). Thil is commonly called the blanlut m~thod. This means signs that wm recently installed would be replaced if
they fell in the identified sign category or area. Last, a conrrol sample ofsigns could be used to represenr all of the agency's
field-installed signs. This sample could come from a randomly selected sample of signs in the field, of various rypes or
from signs installed in a maintenance yard. The retrorefiectiviry should be monitored using one of the assessment methods mentioned in the previous section.
_,.i!~
_1 [-;l.,
~tt~~
......., ..
~:!:mfl~~~~~~~".<"~-~1
- ~ -~
!\.-~~~~.. ,'li.:,'t~:...*r~~:pg, .h~
1s
."
Without R.R.Pl\:ls
!5 50 milh
55--45 ml/!a
'\Yhb
~ 70
JD1Ib
RRP:\l.s
60
90
40
2 50
S7S
50
Pavement markings can be evaluated using similar methods discussed in Section 6.1. However, when conducting ~
assessment of a pavement marlcing, the equipment used is different. In many cases, a pavemenr rerroreflectomcret' 15
wed such as the one shown in Exhibit 7-13. lr is important to take multiple readings along a shon segment of rhe
marking and average the readings to account for variation in paint thickness, as well as the proponions of beads ::;1ft~
Source: DELTA.
to the paint. Becoming more popular are reuoreBectomerers which are attached
co vehicles coUecring asset inventory daca at highway speeds. An example is shown in Exhibit 17-14. This method is
gaining acceptance as measurements are becoming increasingly accurate.
7.0SUMMARY
[riscallacion ofTCDs is a common task, especially as it relates co traflic signals and signs. However, TCDs should not
be inStalled if they are not warranted. The MUTCD is a signi1icant resource that provides a foundation of warrants
applicable to the installation of such devices, especially signals and signs. IfTCDs are inscalled char are not wah-anted
or have nor operated as imcndcd, guidance for removing the devices is provided in this chapter. However, the amount
of work to remove a TCD further necessitates the need to think through whether. it should have been ~ed in the
first place. New or controversial TCDs often need studying to determine if the device a.ctually worked as intended.
That subject is covered in this chapter, with references co the appropriate chapter. Many times, TCDs need to be
ftequendy inaintained to provide the performance intended when they were inscalled. Users of this chapter wishing
to obtain relevant TCD-related information should consult the MUTCD or ITE's Traffic C1111trol Devic~s Handbook.
Other resources are provided in the reference Section of this chapter.
8.0 REFERENCES
8.1 Literature References
Box, P. C. and J. Oppenlander. Manlllli oJTrriffic Engirumng Studia, 4th ed. Washington, DC: lnstirute ofTranspo~tion
Engineers, 1976.
Fcden1 Highway Admi.nistntion. MlllwtJ on Uniform TtttjJU unml DMas. Washington, DC: FHWA, 2003.
Federal Highway Admini.!cration. UpJtztes to &s~=h on Rrcommmdtd Minirnwm Lewis for Pa~~m~mt Mar/ring &tro+aivity tQ
Mm Driwr Night Vt.fibiJjt.J Nmis. Washington, DC: FHWA, 2007.
Institute ofTransporcation Engineers. A Prognmt for SchJl Crotring ProttiQn: A Rr~J Pruniu, 3rd ed. Washington:
, Institute ofTransporcuion Engineers. Trip Gmerarion, 8th Edition: An ITE Informational Report. Washington, DC; ITE,
\2008.
'-JHK and Alsociates. A Uur GuUkforth~ &moval ofNot Nmud Traffic Sigllais, FHWA-IP-80-1 Z. Washington, DC: U.S.
Depanmcnr ofTraruportacion, Federal Highway Administration, 1980.
McShane, Wand R. RDess. Traffic EngitUmng, Englewood Oiffi, NJ: Prentice Hall, 1990.
Chapter 8
o o o o o
'I
o o
o <> > o
t.
o 0
o 0
<
'
o o 0
10
o o,. 0
<> o
o o o 0
<1
o o o o o o o o 0
o o 0
o o o o o o o o 0
o 0
'
0 0 0
o 0
Compliance with
Traffic Control Devices
Original by:
Donna C. Nelson, Ph.D., P.E.
Edited by:
2.0
3.0
INTRODUCTION
143
1.1 Purpose
144
1.2 Applications
144
TYPES OF STUDIES
145
145
145
145
146
146
147
147
148
4.0
155
WMMA~
1~
6.0
REFERENCES
156
156
157
1.0 INTRODUqiON
ser disregard ofTraffic Control Devices (TCOs) is of increasing concern to !:hose involved in traffic engU!cering and roadway safety. The level of public olxdience is one test of the effectiveness of regulations and devices
the adequacy of publicity and education dealing with these controls and the level of enforcement (Humme.s::'
2009). This chapter focuses on the srudy of driver and pedestrian compliance with TCOs. The speciJic srud.ics addressed in this chapter include driver compliance wit:h stop signs, traffic signals, no-turn restrictions and right tur.O. .s
on red. While these studies focus on compliance at intersections, the techniques presented can be adapted to &rud.y
compliance with most traffic regulations and control devices at a wide variety of locations. Bicyclist and pedesaia.J:"'=l
compliance with control devices and regulations may readily be measured by these field procedures as well. Additiona-l
background information can be found in FHWA Publication RD-89-103 Motorist Compliance with Slllndar~ Tr4f.t:::
Control Devices (Pierrucha, Opiela, Knoblauch and Cringler, 1989).
-
Devices
1~
1.1 Purpose
Road-wer compliance studies are conducted to evaluate driver, pedestrian, or bicyclist conformity with a specific
TCD or regulation. An engineer conducting an analysis that indicated unacceptable compliance with a regulation or
device would rethink whether the device was necessary, determine if an additional improv~ment were needed (made
larger, brighter, or supplemented with another device), consider enforcing the regulation through added police enforcement, or explore ways to better enforce the regulation, such as automated enforcement (Hummer, 2009).
1.2 Applications
There are two basic methods for conducting a compliance study on traffic regulations or TCDs: a fidd study or a
controlled study in a lab setting (ITE, 2001). The most desirable study is conducted in the fidd at any roadway or
intersection location for which information is desired. However, it may be more feasible or desirable to conduct the
study through a questionnaire using pictUres or video in a lab setting. The fidd study is typically considered a better
option beca.use the actual observed reaction to TCDs is recorded.
The Manwd on Uniform Traffic Control Droica (MUTCD) defines traffic control tkvica as "all signs, signals, markings,
and other devices used co regulate, warn, or guide traffic, placed on, over, or adjacent to a street, highway, pedestrian
&cility, or bikeway by authority of a public agency having jurisdiction" (FHWA, 2009). Compliance studies may be
conducted to:
evaluate the effectiveness of traffic control devices;
develop edua.rion:U progwhs for drivers, schoolchildren and the general public;
determine critical locations for sdecrive enforcement efforts; and
2.2
Oat~
Needs
4. practically stopped;
5. stopped by traffic (forced stop); and
6. voluntary full stop.
Categories 1 and 2 represent total and putial noncompliance; categories 3 and 4 represent compliance. The choice
berwecn categories 1 md 2 is at the discretion of the observer and may produce confusion and inconsistencies among
data collected ?Y different observers.
Turning movement
Turning movcmenr
Driver action
Vehicle r:ypc
Vehicle type
Cycle lengrh
Temporal restrictions
Speed limit
Approach volume
Approach volume
location
Peak hour
Location
Improper rigbc tum on red
Queuing conditions
Turning movemem
Queuing conditions
Location
Vehicle type
Vehicle type
Approach volume
Approach volume
Pedestrian paths
Pedestrian paths
Pedestrian signals
Location
~.2
!he time and conditions under which compliance srudies u e conducted can affect resulu. These studies ue usually
~rformed in good weather and under "normal" uaffic conditions (when there ue no circumsrances presc:nr chat rnay
affect the results). However, it is possible the TCD Wider investigation may be intended for use in an ab n ormal event,
such as various pavement markings developed for nigh crime use in rainy weather. Samples ue taken tO cover all applicable periods of the day unless it is imperative that a specific time period or weather evenr be taking place during
the analysis. Data are commonly CQilected in 5- or 15-min. intervals over a period that allows the collec tion ofdata
for more than the minimum sample size as described below.
Data should be collected under the conditions during which the problem has been observed or is likely to be most evident. For example, traffic delays and crashes generally occur more ofren during peak traffic periods; therefore, traffic
data should be collected during these periods. Excessive speeds on some streets may occur only when tra ffic volumes
are light to moderate. Compliance with TCDs in school zones is generally of interest during school hours , specificallY
when srudenrs arc traveling to and from school. Compliance studies may be performed in off-peak periods to provide
a comparative analysis of the violation problem. If results arc to be comparable for before-and-after analysis, simila r
conditions must exist during both periods of data colleccion.
\~\.<Mi'i""
~t!j
a_te-\..\.(O.'~ .. "'
ft~.-.:~~~~:-:-"~~~.~~~""4
Y"
Coo.rtant, K
Coolidenc:e ~I (%)
1.00
68.3
1.50
86.3
1.64
90.0
1.96
2.58
95.0
99.0
What is the difference in sample sizes necessary if stop sign violations are assumed to be 20 percent and 50 percent.
So/IliUm: Assume a 20 percent violation rate (p 0.8, q =0.2), a 5 percent permined error and K a 1.96:
{0.8)(0.2)(1.96)
{0.05)2
Equauon 8-1
N-246
If the violation rate is assumed to be 50 percent (p = q = 0.5), then
N. (0.5)(0.5)(1.96)
(0.05) 2
N-385
Approximately 140 more vehicles are necessary for the sw-vey if the violation race were an equal 50/50 split.
Exhibit 8-3 shows conservative (p = q 0.5) sample size requirements for permined errors of 5 percent and 10 percent,
with a confidence level of90 percem or 95 percent. For a more dcta.i.led explanation of K. sec Appendix C. Box(1984)
suggests samples of 100 are ofu:n adoquaa: to indiact: compliance with TCDs, except when violations are rare.
., ...
~<:
location ____________________________________________________
~--
nme
to
Weather-----------------------------Non-Stopping
Stopped by Traffic
I
Voluntary Full Stop
"'0
?i
..;
vi
z
Right
Straight
Left
Straight
Voluntary Full Stop
Right
Stopped by Traffic
Non-Stopping
"'0
?i
..;
vi
z
Pate
Recorder _ _ _ _ _ _ _ _ _ _ _ _ _ _ __
Swp-sign c-ompl.imrce dau are recorded on the form in Exhibit 8-4. For a four-way stop controlled intersecti.on, it will
be necessary co use two forms, one for the north-south direction and one for the east-west direction. An observer
for this study sb.ou1d be positioned to see the study vehicles as they arrive :1.t the stop sign and the aoss traffic on the
through street. Dau are recorded by movemeac for each driver action: full stop, almost stopped, forced stop and no
stop. A foil stop is defined as a "complete cessation of movement, however brief." Nearly tf(Jpped is defined most commonly as ~ 3 mph (5 kmlh). Afo~ui lfQP occurs when the motorist is cequ.Ued co stop because of conflict with cross
traffic and pederuiansi no stop is commonly defined as"> 3 mph (5 km!h)." If more than one observer is used, m~
sure all observers are categorizing vehicles the same way. Cars and trucks (or large vehicles) can be differentiated by
cmering tick marks in the cells for car or truck.
Compliance with Traffic Control Devices 149
~~~~~r~i~~~~t~~::.~~--:f1E~t~~-~i~:~~t~~l:~-::~
~uon
_______________________________________________________________________________
Oiceccion ofTmel
Ttme
Dace
Area Type.___________________________________
ro _ _ _ Weather ----------------------------------------------Observer
Left
Total
Lefc
To cal
c
A
R
T
R
Total Cars
To cal
Trucks
Total
Noces:
Source: Adapted from Mttt4rist Onnpli4nu with St4ndttrd Traffic Cmrrol Dn~ices, FHWA-RD-89-103.
No-f4t-tum mnpllanu daca may be recorded on the form in Exhibit 8-5. The observer is positioned to observe the
study vehicles as they approach the no-left-turn location. The nwnber of vehicles passing the study site that make an
illegal left rum are counted along with the total number of vehicles. Through and right-turning cars and trucks may
also be recorded.
:~
~~ frfcl@r'[.'fd!)if~
-.:-.7-.-.::-: ~ .-.-. ---~--~ - -"~
lo~tion
nme
_______________________________________________________________________
to
Weather ________________________________
N.S. E.
- W.
,,, V H
,., -
1:
1:~
leu3!S padwnr
1.!g
~
c
.,3
pa~
"'Q.
.."'
Q.
~!!!.
5'
.
.."
....
"
Cl
;;
~
Ual~J9
....... .
.... - ...-
~
- ...
N. s. . w. on
~"
I
j
Green
10
..
.,3
a.
"'
"!!!.
~-
..~
"'
ID
0.
.""'
~
"'
~1r
Red
Jumped Signal
I"<:
&i
:;::::
i
z
R~rder
Date-- --------------------
The drivtr compUan. with traJfo: rignals study is particularly concerned with the response of drivers approaching the sign~
during the clearance interval and red phase. Dara may be recorded using the form shown in Exhibit~- The form Nsspace
to record drivtt actions on four incersection approaches. D!Mr behavior is rallied by direction of crave! as the vdllde le:r.'t! lb.e:
inccrseaion. The observed beh:Mor is related ro the traffic signal indication seen by the c:lrMr when the vehicle enterS the imej:"scaion. Entry is usually defined as crossing the near-side curb line. Tbe signal indications arc listed on the d.2,ra form ~
"green," "yellow after green, "red" and "jumped signal. The forms have space ro cecord driver actions. Care must be take~
to consider peraption-r=:tion times in classifying violations of craffic signals. Signal compl.iance studies are appropri~c.O::::
during both p~ and off-peak hours.
.
"'
,.:-
...
--
Location
Direction ofTravd
Tune
Date
~ ~! . !~.---~lim.:~~~:.
...,.;
"-'
-;~~J!U1.(~~ ~>G-.;1'
1-:: ~-
-_~.~~ -vl!...
!"
Ast2Typc
Weather
Observer
to
Indication
Action
Cars
Trucks
Total Can
Total Trucks
Turned on G or Y
On Green
lkhind a W~ter (above) Turned
onG
Attempted to rum on R, rumed
onG
Full Srop
'
!
No Stop
Full Stop
On Red Queue
Stopped by Cross Traflic
Stopped by Pede$trian Crossing
Toal!
~'-------------------------------------------------------------
Source: Adapted from Motorist CompliAnce with S1"4111i4rtl T.mffic Control Devka. FHWA-RD-89-103.
Right-tum-onmJ..aftu-ftOJJ compliAn and data on the proper yielding ofsuch drivers ro pedestrians and other vchicub.r traffic can be oondua:cd using the fOrm shown in Exhibit 8-7. The measure of violation is the ob.ser-I slowing or brake light
application of through aaffic, or interference with pedestrians. Right cum on red lll2Y be identified by sevaal oonditions:
failing to stop oompletdy at the stop line or before aossing the crosswalks a.t an urban incetsection if
pcde$trlans a.ce present;
interfering with a pede$trian in either crosswalk being traversed; or
causing sudden slowing of a vehicle on the cross street which has the green indication.
I:
\ ,
The observer should be able to see the study vehicles as they arrive at the traffic signal and the cross traffic o n the
through street. Observe eVery right-turning vehicle that passes through the intersection during the study. Begin a new
data sheet every 15 min., recording the following observations.
When the signal is green, record the number of study vehlcles that:
turned on green or yeUow;
stopped on red, waited for green before turning;
stopped on red behlnd a vehicle waiting for the green indication before turning right (rhe turn was
executed on green); and
attempted ro turn on ted, signal rurned to green before turn was c:xecured (the turn was completed on
green).
When the sisnaJ is red:
Determine if srudy vehlcle arrived as a single vehicle or as part of a queue waiting at the signal.
Determine if the vehlcle made a full stop (a brief cessation of movement), was stopped by vehlcular or
pedestrian cross traffic, or did nor stop at all before entering the intersection.
Location
Tome
Wuthet
to
Pedestrians crouna
of
St. In
Cre>mdStnfiht
Stpped from Curb on
(cronw1llc)
Slde
Toto!
~ ~
11!~
~f
ilL
"1~
<:roued o;''""''lv
I'
111
l ~f!H
Tol>l
oor.
Recordtr
Pedestrian compliance with traffic 1ignals is recorded on the field sheet shown in Exhibit 8-8. Pedestrian behavior is
noted as the signal indication on which the person steps from the curb onto the pavem~nt of the inrersecrion. Signal
indications are observed for that intersection a.pproach which the pedesuian is crossing and include ~en," "yellow" and "red." If signal indications arc separately provided for pedestrians, the appropriate indications arc "walk or
countdown," "Hashing don't walk"and "steady don't walk." The field sheet is designed to check on crosswalks with a
provision for recording diagonal crossings. If a bicycle compliance study is not conducted separately, the compliance
of bicyclistS can also be recorded by using the lcner "B" for bicycle riders and p for pedestrians.
It is important the analyst undetStand that any compliance srudy could be conducted using similar methods and data
oollection sheers. For instance, a local jurisdiaion may want to check speed oompliance along a certain roadway segment
with a 45 mph (72 km/h) posted speed. A threshold could be used to determine the compliance of drivers on the roadway.
Alternatively, a city ll)ay want to study pedestrian compliance ar traffic signals as shown above. However, there may be an
issue, such as jaywalking, that needs to lx srudied as weU. This additional compliance daca could easily be added co the data
shttt. The point here is that the methodology and data collection method is the same for any cype of comp liance issl.>e th<
.agency wishes to study. In effecc, all that is needed is whether drivers were compliant or not, which is simply "yes/no" data.
~
"'rn-
Equation 8-2
where E is rhe standard error, n is the actual sample size and p, q and K, as previously defined in Section 3.3. The K
values corresponding to the desired levels of confidence are provided in Exhibit 8-2. Since compliance will be measured as a percent, the standard error musr also be converted to a percent by multiplying by 100. If standard error
abour the mean does not include zero, the analyst says the findings were significant ar the levd of confidence used in
the study to determine the "k-value."
Although summary statistics fot compliance studies are usually expressed as a proportion or percentage values, other descriptors of central tendency can be developed. Data analysis and statistical summaries are descnbed in deta.i.l in Appendix C.
Therefore, the analyst could state that he or she was 95 percent confident that drivers came to a voluntary full stop
at stop signs 54.9 :1: 6.2 percent. Because a "kvalue" of 1.96 was chosen {based on a 95 percent confidence level), the
analyst could say he or she was confident the findings were Jignificant at the 95 percent confidence level because the
range does not include zero (48.7 !> x;,;: 61.1). The summary Statistics for each of the four compliance rates are shown
bdow.
Voluntary Full Stop:
Stopped by Traffic:
Practically Stopped:
Nonstopping:
The perctntage of practically stopped, stopped by traffic and voluntary full stop vehicles are typically summed to
represent the percentage of traffic "yielding. which is considered safe in most jurisdictions. Therefore, the total
percentage of drivers compliant with stop signs could be determined as 85.4 percent :1: 4.4 percent. Most jurisdictions
would likdy agree that an 85 percent compliance rate would be acceptable; however, no refermce is a".-ailable on
typical stop sign compliance rates.
S.OSUMMARY
Disregard for TCDs is a serious problem that is frcquendy addressed by transportation analyst. Compliano! with such
devices is one way to measure the effectiveness of a device intenclcd to enforce laws and regulations, education and
publicity programs, or law enforcement efforrs aimed at making drivers more compliant with laws. Many types of
common compliance srudies were addressed in the chapter pertaining to drivelS and pedestrians, many of which could
be adapted for less-&equencly conducted studies. Users of this chapter wishing to find additional information related
to TCD compliance can consult FHWA Publication RD-89-103 Mowrist Complilznc~ with St4ndarri Traffic Control
DMces (Piecrucha, Opida, Knoblauch and Cringler, 1989).
6.0 REFERENCES
6.1 literature References
Box, P. C. Transportation and Tr~ EngiMmng Handbook. "Traffic Studies. Washington, DC: !Jutitute of Transportation
Engineers, 1984: pp. 546-547.
Box, P. C. and J. Oppenlander. Manual ofTraffic EnginJ:mng Studies, 4ch ed. Washington, DC: !Jutitute ofTransporcation
Engineers, 1976.
Federal Highway Administration. Manual on Uniform Traffic Omtrol Dnices. Washington, DC: FHWA, 2009.
Hicks, T. Traffic CDntrol Dntke CDmplimue-Summary ofAASHTO At::Wns. AASHTO Highway Subcommittee on Traffic
Engineering. Seattle, WA: American Association of State Highway and Transportation Officials, 1985.
Hummer, J. Trttf!U &ginuring Hantlboolr, 6th ed. "Ciu.pter 8: Traffic Engineering Srudies." Washington, DC: Institute of
Transportation Engineen, 2009: pp. 8(}-107.
Institute ofTransportation Engineers, Traffo Omml Dnias Handhoolr. Washington, DC: ITE. 2009.
Pietrueha, M., K. Opiela, R. Knoblauch and K. Cring!cr. Mowrirt CDmplimue with St41ldarJ Trrzffic Omtrol Dntices.
Washington, DC: Federal Highway Administration, 1989.
.... .. - - - - n,..
'""T'
~rr
""ta.U"\
c"~Tln~l
Flotida Department ofTransportation. Operations-Manll41 on Uniform Traffic Studies. Tallahassee, FL: Depmment of
Transporcation, 2000. hr:tp:l/dot.state.B.us/TrafficOperuions/Operatioru/Srudies/MliTS/MUTS.shtm.
i
~
..
::
s::
}>
z
~
r-
-n
.....
~
zVl
(3
""};'
:::1
z
m
z
Cl
"'
z
Cl
VI
-I
;:;:;
"'
"'
z
Cha pter 9
....... . ... .. .... . .... ......... ..... . ... ........................ ... ....... .. . ............ .
~
2.0
INTROOUaiON
159
1.1 Applications
160
160
160
2. 1 Introduction
160
161
165
166
171
171
173
173
173
4.0 REFERENCES
174
174
174
174
1.0 INTRODUCTION
ravel time and delay are two of the principal measures ofhigh"A"aY system performance wed by traffic engince:_.:-s,
planners and analym. Vehicle speed is directly related ro uavel rime and delay and is also used ro evaluate u-af-6-c
and highw..y systems. There are two types of average speed: rime-mean speed (TMS) (or mean spot speed) ~~
space-mean speed (SMS) (or mean travel speed). Measuring TMS is described in Chapter 5. SMS is covered in tt:;a.J.S
chapter by examining the mean uavcl time in relation to the segment length.
Travel time varies inversely with travel speed. A travel-time study provides data on the amount of time it takes cot!"~
verse a section of sueet or highway. These data, combined with the length of the section under study, produce m~
uavcl speed. Travel-time and delay studies are conducted when the sources and amounts of delay occurring wich..i- .n
the section are also noted. This chapter treats the measurement of delay along a roadway segment. Intersection dd~Y
studies are addressed in Chapter 6.
1.1 Applications
Engineers and planners usc data from travel-time and delay studies in a number of casks, including:
determining the efficiency of a route with respect to its ability to carry traffic relative to other routes
through the use ofsufficiency ratings or congestion indices;
Vehicle observation
Probe vehicle
The first requires the analyst to perform measurements while in a moving vehicle in the traffic scream, while the others methods do not. The choice of method depends on the purpose of the study; the type of roadway segment under
srudy; the length of the segment; the time of day of interest; and the personnel, equipment and resources available.
The mOSt common methods (the test vehicle methods) are presented in detail, whereas the others are qescribed only
briefiy. Appendix E contains data forms that are suitable for copying.
rravel time;
running time;
distance traveled;
',
'I
..,
(
SMS ; ~d
hourly volume.
Agencies usually study travel time ~d delay during the peak hours in the directions of heaviest traffic Row. It may also
be desirable to compare travel times, speeds ~d delays berween peak ~d off-peak periods or berween sets of other
conditions. Some of these other conditions include ideal versus adverse weather ~d commuter versuS special-eyent
traffic. During the peak period, speeds will probably vary, which may make it necessary to conduct separate traveltime and delay studies for differe.nt portions of the peak period.
Before test runs begin, observers select the starr point, end point ~d comrol point locations along the route whe re
they will record rime mcaswes. On arterial ~d oPter types ofsurface sueets, these locations are usually at major signali:z.ed intersections or other easily identifiable control points. The choice of the ncar curb, far cwb, or center of the
intersection :as the control point should be consistent throughout the study route. On freeways, control point locations~ include milepost markers, bridge overpasses, bridge underpasses, exit ramps, cn~ce ramps, or other ea,sily
identifiable points. The choice of the bridge edge or center and ramp location (such as the gore) as the cont rol point
should be consistent throughout the srupy route.
2.21 Dritting Teclmiqua
The driver of the test vehicle proceeds along the study route in accor~ce with one of the following techniques:
Avaage.Car Technique: test vehicle travels according to the driver's judgment of the average speed of the
traffic sueam.
Floating-Car Technique: driver '"floats"' with the traffic by attempting to safely pass as many vehicles as
pass the test vehicle.
Maximum-Car Technique: test vehicle is driven at the posted speed limit unless impeded by actual traffic
conditions or safety considerations.
The selection of test vehicle technique is based on the purpose of the study and the study team's judgment of the
technique that best reflects the traffic stream being investigated. Most study reams prefer the ~verage~ technique.
However, the average~ and Boating-car techniques function similarly in real world traffic situations. The travel
lane should be consciously chosen, especially since the outside lane may be affected by slow-turning vehicles, parking
maneuvers, or transit stops. In freeway studies, travel lane might be analyzed as ~ variable.
2.22 Penomuland Equipmmt
The test vehicle method requires a test vehicle and the m= to record time and distance. These ~ be recorded
manually oc automatically. A voice recorde.r can be useful to take nores of queues or other incidents, particularly in
heavy traffic situations. Also, ~still or video camera can be useful to take photographs or videos of unusual events that
will help explain the data resulrs.
Manual daea collection requires~ driver and observer/recorder, two stopwatches and data collection forms. The srudy
~ be conducted with only one stopwatch. The distances between conuol points and the length of the total route
may be obtained from ~ccurate, drawn-co-scale plans or maps. or from the vehicle odometer.
Automatic data collection equipment i.s ~vailable to aid with data collection. Global Positioning System (GPS) unirs
can be used to measure the test vehicle's position and speed along the corridor. The term GPS is used in gener_al in thi.s
chapter to describe the standard technology which can be applied or can be augmented through diffcrcnoa1 methods
or ocher advancements in the technology. Numerous varieties of sofTWare programs for use wirh GPS units or lapcop/
handheld computers arc available to perform travel-time and delay srudies. These prooucu can caprure fearures including speed, distance, stopped delay and number of stops. A stopwatch or other manual equipmem should be used
on at lease one run to error-check the data collection instruments.
2.2.3 Stzmple Siu Requirements
The purpose of the uavel-time and delay srudy dictates the site of sample (in terms of number of test runs) required.
The following arc suggested ranges of permitted errors in the estimate of the mean travel speed related to study purpose.
transportation planning and highway needs srudies: t 3.0 to 5.0 mph( 4.8 to 8.0 km/h).
traffic operations, crend analysis and economic evaluations: 2.0 tO 4.0 mph ( 3.2 to 6.5 km/h).
before-and-after srudies: 1.0 co 3.0 mph( 1.6 to 4.8 km/h).
Engineers can apply these suggestions to similar studies as well.
Exhibit 9-1 suggests numbers of test runs to perform travel-time and delay studies (Quiroga and Bullock, 1998). These
values have a confidence level of 99.73 percent, 95 percent, 85 percent, or 75 percent and arc based on the difference
between minimum and maximum speed measured in the inicial study. Multiple confidence levels arc presented to provide the analyse with the option to select the Je..cl consistent with their specific needs. Upon completion of the inicial test
runs, calculate the difference bcrween minimum and maximum speeds of the test runs (11), as shown in Equation 9-1.
..
1-1
11
R = maxv1 -minv1
F.q~on9-l
where
m
v1
Select the approximate minimum sample size from Exhibit 9-1 fur the computed difference in minimum and maximum speeds and the desired error permitted(&). If the required number of test runs is greater than the initial number
of test runs, make additional runs under the same conditions as the initial runs ro complete the sample required. Determine a sample size for each direction of travel and for each set of traffic and/or environmental conditions of interest.
For ~ample, assume that five runs were made to obtain a difference in maximum and minimum running speeds of
10 mph (16 kmlh) and that the desired permitted error is 3 mph (4.8 kmlh). Exhibit 9-1 indicates that 8 is the
minimum number of runs required, assuming a con1idence level of95 percent. Therefore, three additional runs would
have co be made to achieve the required sample site. However, if the additional runs result in a larger difference, additional runs wiU be required in most instances.
The obscrver(s) must be sensirlve to changes in the traffic or environmental conditions. The sample number of runs
represents a single sec of conditions. For example, speeds will probably vary during a peak pcrioo. Therefore, it may
be necessary to conduct separate travel-time and delay studies for different portions of the peak perioo. Srudying
speed changes during the peak period can be useful and important for agencies. Also, signal progression can introduce
travel-time variability during data collection. Signal timing along the corridor should be understoO<! before starring
the study so the analyst can account for changes in progression over the study period.
2.2.4 Manual Data ColiectUm
Position the test vehicle a shore distance upstream of the begin poinc. Zero~ both stopwatches and complete the header
infurmarion on the data forms. Exhibit 9-2 shows a sample field sheet. Review the duties of the driver and observer. Make
a couple of dry runs to measure the distances between checkpoints and to rehearse the procedure prior to the desired start
time of the study. The study team can measure the distance from a variety of tools including: plans, maps, online mapping
services, vehicle-mounted distance measuring instrument (DMI), GPS receiver, or (with less accuracy) from the vehicle
ooometer. The tool used to measure the distance should be able to obtain a measurement precision of within 1 percent or
2 percent of the actual length.
.: ~;- ,~
:il:W~~itf4tcWdi{!f&t~~:.~.::--
ii
{mph)
1 mph
2mph
3mph
4 mph
5mph
1 m ph
2mph
3 mph
4 m ph
13
4
5
3
3
4
5
17
10
21
11
12
4
4
26
13
15
32
15
10
18
37
17
12
10
43
19
13
10
24
11
so
21
14
11
8
9
9
21
9
10
27
12
II
57
24
IS
12
10
31
13
12
64
72
26
17
13
11
34
13
29
18
14
11
38
14
80
32
20
15
12
43
15
89
34
21
16
13
47
20
50 .
17
71
2S
68
30
39.
21
27
21
30
89
so
34
26
5mph
3
3
3
4
4
4
5
5
5
5
5
6
6
7
7
15
10
16
11
18
11
12
9
9
19
27
17
12
10
99
36
22
15
12
47
27
19
15
6'
I
I
I
8
!
R.
5mpb
1 mph
2mph
(mph)
1 mph
3 mph
4mp h
,6
5
6
i ]
5mpb
10
7
. 8
12
14
II
3
3
16
4
5
5
12
14
4
5
15
7
8
5
6
6
4
4
5
4
4
4
4
4
5
10
18
11
20
6
7
12
23
10
17
13
25
28
5
5
19
14
11
12
20
15
30
13
22
20
45
18
11
33
9
10
14
25
62
. 81
24
IS
18
11
13
44
18
11
10
58
22
14
10
. 8
30
30
Source: Quiroga, C. A. a.nd D. Bullock (1998). Determination of Sample Sizes for Travd Ttme Scudies." lTE]ournal on k
'Web. Institute ofTrao.spon:ation Engineers. Washington, DC.
Travel-Time and Delay Studies 16 ..::3
-~--
.
.
l\..11':-'--
lOlL,.'""~.
--
:J. .
.,_
..
~
. --- -
';
~:$ ~~~~-.
:;-. _}t~wit7~ft~~~
Date
Weather
Route
Dira:cion
Trip Stalled at
At
Trip Ended at
. . .:
..,11.
it:IO
Trip#
n.caoonl
(Mqc)
(Loarion)
(Milaa<)
At
Control Points
Location
Sto}l-' or Slows
lime
Location
Type
Delay
(second.)
Trip Length:
Trip Tune:
Running Tune:
Stopped T L111C!
ITravel Speed:
I Running Sp=i:
Symbols or Delay cause: S-Traffic Sigoa.ls, SS-Stop Sign, LT-Lcft Turns, PK-Parked Cars, DP-Double Parking. T-Gcneral,
Pcd-Pedcstrians, BP-Bus Pa.s.scngcr Loa.ding or Unloading
CommeniS
Ob=vu
If the rehearsal runs go smoothly, count them as test runs. One round trip (one run in each direction) through the
test section constitutes a test run. Test runs should begin promptly at the beginning of the desired study period so
as to complete the required sample of runs before conditions along the route change. If the required sample cannot be obtained in a single period, complete the remaining runs on another day under the same conditions or add
additional teSt vehicle teams to the peale period. As the test vehicle passes the begin point, the driver starts the first
stopwatch. The test vehicle proceeds through the srudy route according to the driving technique selected. The observer records time readings from the first stopwatch as the vehicle passes each control point. When the test vehicle
stops or is forced to rravd slowly (that is, 5 mph [8 kmlh] or less), the observer uses the second stopwatch to measure the amount of delay and noteS the location, duration and cause of each delay on the field sheet. Alternatively,
the study could be conducted with only one stopwatch. The observer would stan the stopwatch at the beginning
point and note the start and end times of any delays during the run.
As the test vehicle passes the end point of the srudy route, the driver reads the first stopwatch, and the obscrvc:r
notes the total time of the run on the field sheet. The rest vehicle then rerurns to the begin point, the driver and
observ~r reset the stopwatches, the observer prepares a new fidd sheet and the next run begins. This procedure is
164 MANUAL OF TRANSPORTAnON ENGINEERING STIJniF<; ?Nn Ff'liT11'1N
repeated until the required number of sample runs is reached or until conditions that could affect the study change.
If the reverse direction is also being studied, the same procedure governs the return trip, with the data recorded on
.~ separate field sheet.
A laptop computer software program with a rime-stamp procedure can be used to replace stopwatches. An o bserver
using a laptop can reduce the additional tasks required of the.driver and capture the same information with predefined keyscrokes to represent locations or delays.
Calibrate DMls before arriving at the study site. Train the driver (and observer, if used) to operate the automati c
equipment before staning a study. Safecy is a primary consideration in conducting this type of study with a driver
only. Make a dey run and enter into the data recorder the location of the beginning, ending and control points
for the route under study. Begin the test runs prompdy at the start of the desired study period, and complete the
required number of runs before conditions change. If the required sample cannot be obtained in a single peri~d,
complete the remaining runs on another day under the same conditions.
As. !he test vehicle passes the begin poirit, the driver activates the data recorder. The test vehicle proceeds through
the study route according to the; driving technique selected. As each control point is passed, the driver activates the
appropriate switch (this provides redundancy because the swcing point can be used to derive any control point
along the route). When the test vehicle stops or is forced co travel slowly (the DMI will automatically sense chis),
the driver pr= the appropriate button to enter the cause of each delay. The DMI automatically records the location and duration of the delay. Upon reaching the end point, the driver activates the appropriate switch to note that
fact. The test vehicle then returns to the begin point, the driver resets the data recorder and the next run begins.
This procedure is repeated until the required number of sample runs is reached or until conditions that could affect
the study change. Some runs might need to be discarded if unusual events affect the run-for example, a passing
emergency vehicle requires vehicles to pull over to the side of the road, a collision, or other atypical' circumstance.
s = 3600 E.T
where
Equation 9-2
The equation for SMS or mean travel speed of all travel time runs is
3600ND
S,w=~
Equation 9-3
where
space-mean speed or mean cravd speed, mph
SA;'.!:
LT
Use Equations 9-2 and 9-3 co calculate running speed per run and mean running speed of all runs, except use running
speed in place of crave! speed and running cime in place of uavel rime. Mean speed may be calculated for each section
of the study route in addition to the mean speed of the total route. Field notes can be used to justifY the removal of
inconsistent values from the analysis due to an unusual siruation.
Since the srudy team measures delay directly, it can summariz.e the delays of each type (refer to Ch~pter 6 for detailed
informacion on rypes of delay) for each section and for each run. Total delay (the d.ilkrence between the recorded crave!
time and the free-Bow travel time) is a useful and popular delay calculation. Mean delays are calculated by dividing the
sucns of delays by the number of runs. Each of the resulting speed and delay measures may be ueated to statistical analyses, such as those described in Appendix C.
Graphical summaries are useful in displaying the various travel-speed and delay measures. Exhibit 9-3 is an example
of an average speed graph. For areawide
studies, time contours overlaid on a map
of the city's streets (as shown in Exhibit
9-4) offer a practical graphical display of
travel times. Exhibits 9-5 and 9-6 present
examples of typical travel time study output..Computer sofrware that accompanies
automatic data collecrion equipment offers a powerful range of analysis and data
summary capabilities.
LT~,-JW~m.tao
l~~~.-~~
-nt
45r---------------------------------A-oSpoedll-lntarMCIIono
Feny RA>ed-nd (A.II Pule)
I 30
..j
f1s
"'
I I
A::
~-
0 ~------~----r---~--~--------~--~
2000
4000
~00
8000
lllol.once(looq
-Lht:
Con~oM~e~"- 2 v...oy5
f TI)<O-
-!liM.
Clallt- 4 . --
(I)
'*-(fwtl
~
t:l:l
A.
1!1
'"0
'"0
~'i
g0..
"G
sf
'-J
9'
"?
;;
.......... (fHI,
I
I
I
II
I
I
i
I
/I
IN l/!
;I
1!1
t.-J! [J
~l
i
I
. I
~:r-rr-~
~
~~~
1[
, if
[ . il
I
lj f
iz
a
s
8
~;
Il
'
t..
,.
-'
..
..
,..
-~\-
~jl>;.'>
'i'H~ .
..
...
:;,
~
~":t.
'
~- .
. ...
.f'""
&
. . . ""
~"'
<
...
l' '
-~
0
l .
'tt
-"'
. .... .
.
<
;,.,
>
.o ~. ~ .
Dare
Route
End Point
Start Point
Vehicles
Stan Tune
Run
Fmish T.unc
Travel Tune
Vehicles Met
<>vena.king
Vchlcles
Passed
Bound
1
2
3
4
5
6
7
8
9
10
Total
Average
Stan Tunc
Run
FtnbhTune
TnvdTune
Vchlclcs Met
Bound
1
4
5
6
7
8
9
10
Total
Average
Comments
Obse~r
Vehicles
Owrtak.ing
Vchlcles
Passed
, Both directions may be studied simulra': neously. ln this situation, one observer
f counts the vehicles passed and overta.kc:n, while a second observer counts the
vehicles traveling in the opposite direction. The team must ensure the counts
are propc.rly recorded on the handheld
counters. If traflic is heavy in the opposing direccion, it may be necessary to
weight the count with a multiplier factor
(for example, one count equals five vehicles). Studies on freeways with more than
three lanes in each direction are difficult
and may require an additional observer.
Safety is a primary consideration in conducting these studies. V'JSual obstructions
from trucks and a wide median can compromise the accuracy. If traffic is heavy,
usc additional observers so the driver can
concentrate: on the driving wk.
EndPoint
I
I
I
I
I
I
I
I
I
I
I
I
I
:~
I
I
I
Start Point
Automatic data collection can utilize specially designed data-recording hardware to perform travel-time and delay
studies using the volume: extension (Robertson and Courage, 1973). Handheld or laptop computers can be configured and software developed to serve this purpose. Because of safety concerns, automation would not eliminate
the need for an obsecver/recorder.l,he procedure is the same as for manual data collection, except the observer uses
a handheld or laptop computer to rerord time: and traffic count data. The procedure requires the same number of
personnel. Automated data reduction saves time and money during the analysis.
Equation 9-4
where
~
M, opposing count of vehicles met when the test v~hicle was traveling south
0. = number of vehicles overtaking the test vehicle as it traveled north
Travel-Time and Delay Studies 169
..
Northbound Trips
T. (min)
M.
...
IN
2.65
85
2N
2.70
83
3N
2.35
77
4N
3.00
85
0
I
5N
2.42
90
6N
2.54
84
Tow
15.66
504
Average
2.6!
84.0
1.5
1.0
Southbound Trips
T (min)
P,
IS
2.33
112
2$
2.30
113
3$
2.71
119
45
2.16
120
5S
2.54
105
6S
2.48
100
Total
14.52
669
A-verage
2.42
111.5
0.5
1.0
The average travel rime for one..<firectional flow is calculated using the foUowing formula:
r.a.. = r.-
eoco"-'"))
v.
Equation 9-5
where:
,.
Equation 9-6
where S. is the SMS northbound, in mph; and dis the length of test section, in miles.
EXAMPLE9-I
Exhibit 9-9 shows data for a 1.25-mile (2.01-km) length of rest section. From Equation 9-4, the ho urly volumes in
the northbound and southbound directions are, respectively,
Vn = 60 (
V. = 60 (
From Equation (9-5), the average travd times for the northbound and southbound directions ace, respectively,
1.0))
r,.,.... = 2.61- ( 60 (1.51336
= 2.59 min
T,,_ =
2.42- (
60 (0.5- 1.0))
= 2.45 min.
996
From Equation 9-6, the space-mean speeds for the northbound and southbound directions are, respectively,
. 60(1.25)
s, =
60(1.25)
~=
This method produces travel time only, from which average travel speed may be calculated once the distance betw~en
observation poincs is measured. A test vehicle is not required. Observers/recorders posicion themselves at the emra.oce
and ait to the test section and at other major intersections or pointsofinterest along the route. There are four methods
that can be implemeiued for license plate marching: manual, portable computer, video with manual transcription, or
video with chacoa.ccer recognition (Turner, Eisele, Bem and Holdener, 1998).
For manual license place marching, as vehicles pass the observers at each location, the observers record the last three or
four digits of the license tag along with the time from a stopwatch, with pen and paper or on an audio tape. Using; a
portable computer, an observer can record license plate numbers accompanied with an automatic time stamp. Lice fl.=' e
plate numbers can also be collected using video. The video can be post-processed using manual uansaipcion; tP..i.s
involves an analyst in the office watching che video and recording the license plate numbers. Character recogniriO ..J:I
.
Travel-Time and Delay Studies 17 --1
software could be implcmcmcd on the video <kra which would seck to recognize the characters on the license plates
and record them automatically.
Using one of these four <Uta collection methods, the license plate numbers and times arc then matched in the office,
either manually or by computer, to obtain travel times. The matching is laborious if done manually. A sample size
of 50 License number matches usually provides adequate accuracy (Box and Oppcnlandcr, 1976). lkfer to Exhibit
9-1 for a specific sample size estimate, where the minimum number of o~rvations wiU replace the number of runs.
Computer software for matching license plate numbers is available.
3 .1.2lntnvintl Method
Selected individuals who are wiUing to cooperate may provide a satisfactory sample from which to obtain travel times
and delays without the use of a test vehicle or observers. These persons are asked to record their start and end times
for designated routes. They also record the times and durations of delay. This is a variacion on the average-car method,
except that in place of a single test vehicle there arc multiple test vehicles. Employees wbo drive on the job, truck drivers and raxi drivers often make good subjects. This method is useful when a large amount of <Uta is needed in a short
time (Pignataro, 1973). The interview method requires some training and equipment (such as stopwatches) for the
subjects. The reliability of the results may not equal that of methods that employ better-trained data collectors. lkfer
to Appendix B for additional information.
3.1.3 W'~JTless Teclmoklgy Metlxld
Wireless technology is used to connect devices including cellular phones, car radios and computers. Bluctooth is
a common technology cunendy in use to emit unique signals which can be collected by monitoring units placed
alongside the roadway to collect travel time and origin and destination <Uta (Tarnoff, Wasson, Young. ~nig,
J3ullock and Sturdevant, 2008). Initial testing showed an inJiltration rate of '5-7 percent, which provides adequate
sample sizes for volumes greater than 600 vehicles per hour.
3.1.4 CelhJ4r Te/.ephotu ObsertJIIIUm Mnhod
Cdlular telephones can be used as an observational travel-time study. Geolocating of cellular telephones can be captured through the discreet tracking of telephone calls (Turner, Eisele, Benz and Holdener, 1998). The system tracks
telephone calls that initiate from vehicles and triangulates the location of the vehicle. The data is encrypted and sene
to a central office for analysis.
3.1.5 Extrapolation Method
The extrapolation method estimates travel times by applying spot speeds across shore segments (Turner, Eisele, Benz
and Holdener, 1998). Spot speeds can be collected using a varicry of methods including:
inductance loop detectors;
piewdectric sensors;
active and passive infrared sensors;
magnetic sensors;
video tracking and tripline systems;
doppler microwave;
passive acoustic sensors; and
pulse ultrasonic detectors.
The extrapolation me!hod can be used for planning applications, but should be avoided in most scudies. Inaccuracies
can result from dtis method due to &cility type, detector spacing. traffic flow conditions, or individual device accuracies.
Signpost-based uansponder systems a.re typically used by transit agencies for monitoring the srarus of the fleet vehicles
(Tumer, Eisele, Benz and Holdener, 1998). This technology operat~ by Beet vehicles receiving timNramped identificarion data from signpost mounted decrronic uansmirtcrs. The ideqtification data is transmitted to a central dispatch
ficility which can monitor schedule adherence.
3.4.2 AW TrtmiJH>IIn't
Electronic toll-road tags a.re a primary application of automatic vehicle identification (AVI) uan;sponders. These
uansponden can also be used for .real-time traffic monitoring. incident management, travder information and performance measuring (Turner, Eisele, Benz and Holdener, 1998). T1me of arrival at each booth is .recorded on each tag.
The toll-road rags provide data coUecrprs with a method to uniqu~ identify vehicles and measure the uavel times
over a facility between toll booths. Knowing the distance between booths will allow the calculation of overall ttavd
speeds.
3.4..3 GrounJ.btuetJ llAJio Nllfliguion
Ground-ba.sed radio navigation wes a receiving antenna netWork and demonic uan;sponders within vehicles
(Turner, Eisele, Beru. and Holdener, 1998). Transit agencies and companies with large fleers are co=only we this
technology. A cenual office computer can request vehicle location information; this results in a transmission from
the vehicle to the radio towers to the central office. The location information can be determined by triangulation
from multiple towers.
4.0 REFERENCES
4.1 Literature References
Box, P. C. ind ). C, Oppenlander. Manual ofTrrzffic Enginming Studies, 4th ed. Washington, DC: lnsricuce ofTransporution
Engineers, 1976: pp. 93-105.
JAMAR Sales Company, Inc. PC-Trawl Sampit &portJ. Ivyland, PA: JAMAR, 1990.
Krygsman, S. and P. Schmitz. "The Use ofCcllphone Technology in Activity and Travel Data CoUection." 24m Southern
African Transport Conference. Pretoria, South Africa: Document Transformation Technologies, 2005: pp. 696-705.
Pignataro, L. Traffic Enginming Theory and Prrutice. Englewood Cliffs, NJ: Prentice Hall, 1973.
Quiroga, C. A. and D. Bullock. "Determination of Sample Sizes for Travd Tune Studies. IT foumal 68, No. 8 (Augusr
1998): 92-98.
Quiroga, C. A. and D. Bullock. "Measuring Control Delay ar Signalized Intersections. ASCEJoumal ofTransportation
Enginuring 125, No.4: 27 1-280.
Roberrson, H. D. and K. G. Courage. "Traflic Signal Srud.ies Using l Digiral Tape Recorder TnzfJU &searrh &pon, June 1973.
T:unoff, P. J. et al. The Continuing Evolution ofTravel Tune Data Information Colltion and Procwing. Washington, DC:
Transportation Rcse:arch Bo:ud, 2008.
Turner, S.M., W. L. Eisele, R. J. Benz lnd D.). Holdener. Tnzw/ Time Data Co/kaion Handhook. Report No. FHWAPL-98-035. Washington, DC: Federal Highway Adminstrarion, 1998.
; Eisele, W. L., J. A. Crawford and R. L. Stensrud. Mtasum, Mttbods, and Applications ofa Mobility Monitoring Prowsfor SfflalL
; to Mtdium-Siud Commrmitit1. CoUege Stacion, TX: Te:u.s Transportation lnstirute, Texas A&.\1 Univc:rsiry System, 2007.
f.iscJe. W. L. and L
Rileu. "Travel-Time Estimates Obtained from lnteHigent Tra<Uporration Sym ms and Instrumented Test
Vehicles: Statistical Comparison." Transportalion &starch Record: journal ofthe Transportation Rt1earch Board 1804 (2002).
Garber, N.). and L. A. Hoe!. Traffic & Highway Enginming. Pacific Grove, CA: Brooks/Cole, 2002.
Guindon, M. E. "2002 Congestion Management Report: Travel Time and Delay Study. Ninth National Conference on
Transportation Planning for Small and Medium-Si:r.ed Communities. Colorado Springs, CO: Transportation Research
Board, 2004.
Guo, B. and A. Poling. "Geographic Information Synems/Giob;al Positioning Systems Design for Network Tr..vd Time SrudY
Transportation Research & cord journal oftht Transportation Restart:h Board 1497 (1995).
Mortimer, W "Moving Vehicle Method of Estimating Volumes and Speeds. Transportation Rntarrh Record: journal ofrht
Roess, R. et al. Traffic Enginmin~ 3rd ed. Upper Saddle River, NJ: Pearson Prentice Hall, 2004.
Srinivasan, K. and P. Jovanis. "Determinacion of Number of Probe Vehicles Required for Reliable Travel Tune Measurement in
Urban Nerwork." Tnznsport4tion Rntarr:h &cord: journal ofthe Tranrport.uion Rnearr:h Board 1537 (1996).
Transportarion Research Board. "Traffic and Urban Dara. Transportation Restarr:h Record: Journal oftht Transporration ResetlfCh
Resurrh&ardl55I ( 1996).
W~g, z_ "Using Floating Cars to Measure Travel Tune Delay: How Accurate is the Method?" Transp0rt4tion Research Record:
journal ofthe Tnznsport4tion &search Board 1870 (2004).
Chapter 10
................................. ................................................. .
177
1. 7 Chapter Objectives
177
178
179
182
182
187
190
191
191
191
192
194
194
195
195
197
I
i 198
199
1.0 INTRODUCTION
The chapte; is-organized into four parts. The introduction section provides a general overview and de6.ni;iQAS of the.
different facility types. Ncn, the types of studies and data items of interest for freeways ue presented; followed by a
Freeway and Managed Lanes Studies 177
discussion of the data collection procedures necessary co obtain che data. The chapcer concludes with a section on data
reduction and analysis of freeway facilities.
The data collection procedures on general purpose (GP) freeway facilities and on managed lanes (ML) are concepcually similar. In most parts of the chapter both are therefore discussed concurrently and differences are emphasized
where necessary.
,Due co high volumes and operacing speeds on freeway facilities, data collection on freeways frequencly involves auicomated data collection technology, although nonautomaced methods exist. Freeways represent the backbone of chc
fransporcacion infrastructure, and as such, it is common to lind permanently inscalled sensing equipment co continuously monitor the traffic operations. Section 2 of this chapter highlights the most commonly collected data items on
freeways. It further emphasizes some special freeway studies related co incident management or freeway. work zones.
HOY, HOT, ET and TOT lanes are ML treatments geared cowards pr!oviding more efficient use of the available ~ciCY"
of a freeway. An HOY !we may require different vehicle occupancy lCvds, expressed as two or mote (2+), three OtJOO.-e
(3+), or even four or more (4+) persons per vehicle. HOY facilities consisr ofone or more dedicated lwes that give1rio$ity co HOV vehicles, which is intended to increase the actual person trips along a freeway segment without increas~ c.Pe
number of vehicles using the facility (Kuhn et al., 2005). HOV lanes aim co reduce vehicle congestion by redu~ t:Pe
number of vehicles on the highway relative to the numbec of pecsons carried. In other words, HOY lwes maint.ae ~c
same (or greacec) person-movements while reducing vehicle-movements. HOY users are provided with lwes dur 1pe~
ate at lowec congestion levels and thus higher speeds. They therefore (should) have lowec travel times associated wh ~
average trip compared co the general purpose (GP) lanes, which gives"w incentive for their use.
HOT lane facilities ue similar co HOV lane facilities, in that they allow HOV vehicles to travel on separate heS
However, they further include pricing strategies chat allow vehicles not meeting HOV oa;upancy rcquiremeq; ~ ~
purchase access to the lanes thtough electronic rolling. This strategy typically uses a time-of-day or dynamic pt:i.~~
strategy that is directly tied to the demand wdcapacityin the HOT lwes. HOT lwes aim co (paccly) 611 excess~~
ity of undcrutilizcd HOV lanes as illuscrated in Exhibit 10-2.
:rc..,-
Z:J:c. .s C..podry:Tiae
to B1cJ> Oc._acy Toll J..oDOs
(from ColorD.SO VDJu. Eqlftss x..a-s :Feaolbillty S~: 1 - 11199)
teO'!
90'/o
""'"
i ..,.
"'"
,.,..
I
a
""~
2<1'.
toe
.,...
s.oo
Alof
""
A3.l
7:'10
8:GO
H)O
Alr4
~t
A!ol
S!OO
PM
laii!OVVo- -...
Source: FHWA mm~IIJ Opmuions Handbook.
2::a:l
ftt
-4.'00
nt
.SA
n&
6IDO
ft.S
?M
&.<0
P3r.(
.I'.W.
Exhibit 10-2 makes evident HOY lanes can be undcrutilized, and improved utilization can be achieved through
dynamic pricing. This is also helpful from a public perception point of view since HOY lanes are often perceived to
carry few vehicle-ttips, even though the acrual person-ttips in the managed lane inay be equal to or greater than the
GP lanes (due to higher in-vehicle occupancy). Finally, the revenue generated from HOT lanes supports maintenance:
activities and construction of furure ML facilities, which is an important factor as agencies try to identify new funding
sources for frc:c:way construction. The capaciry and operations of any tolled faciliry are critically linked to the means
of toll collection, which can range from manual payment to an attendee, to semiautomared coin collection, to fully
automated electronic toll collection.
HOY/HOT lanes are wually constructed wing one of three possible methods, described below. The dlfferent con
figurations generally vary in appearance and capaciry, and are listed here since they may affect data collection practices.
Concurrent Flow Lane: a freeway lane, typically the innermost lane, which is designated for HOY/HOT
we only. This lane is not physically separated from the other lanes, but is instead designated through the
use of pavement markings or a painted buffer.
_,_---
Contraflow Lane: peak-direction use of an off-peak direction freeway lane. Traffic is separared from
adjacent mainline traffic in the off-peak direction by some form of barrier, including movahle concrete
barriers, pylons, or cones.
&dasi:ft HOV/HOT Fa.cility: HOV lanes that are separated from All GP lanes (both directions) wing
~physical barrier. These lanes are cxclwivdy used for HOY/HOT during all or a portion of the day.
Exdwive lanes can be operarc:d as two-directional or reversible. Reversible lanes typically operate With the
inbound and outbound traffic patterns associated with commuting to a CBD.
One of the key differences between HOY and HOT lanes in terms of data collection is that HOT lanes are almost
always fuUy instrumented to monitor traffic operations, while some older HOY facilities may not have broad (or
any) sensor coverage. However, it is anticipated that most ML facilities will become increasingly instrumenrc:d in
the furure. ET and TOT lanes are exclusive toU-only facilities for passenger cars and heavy vehicles, respectively. For
purpose of data collection they are similar to HOT facilities, and will therefore not be discussed separately. One dif.
ference worth emphaSizing is that toll-only facilities are typically les$ likdy to involve vehicle occupancy studies, which
is common for both HOY and HOT &cilities.
Kuhn ec al. (2005) provides more detail on all ML &ciliry types. More detailed informa.tion on HOY and HOT lanes
can be found in NCHRP Report 414: HOV Syrtmas Manwzl (Taas Transportation Institute, 1998) and A GuiM for
TOI\~I('Of"\OTI\Tif"'M
~f"'ICTit"'lf\1
.
Ramp metering is an effective suategy for improving traffic flow along the aWnllne by controlling the rate of vehicles
entering from the on-ramp into the mainline traffic. This improves the ovcraiJ capaciry along the &ccway. cspecially
during congested conditions. Platoons of vehicles attempting to enter the mainline traffic stream can cawe a loc.gf
turbulence during high demand periods, because gaps in traffic an; less &equent. The friction crcared &om merging
craflic can slo"w the mainline considerably, especially in the outside :ttavd lanes.
The metering rate is typically set as a function of the upstream freeway mainline volumes and the capaciry of rhe
freeway section downstream of the ramp. Since the &ccway mainline carries a higha &action of overall traffic, the
on-ramp vehicles are delayed to minimize the overall delay of the system. In some cases, the introduced on-ramp
delay can be capped by the available queue storage on the on-ramp. The metaing strategy is employed by means of
traffic signalization and signing wing three increasingly complex methods: I) a pretimed meter co provide a consistent
pattern of vehicles ~cering the traffic ~cam; 2) time-of-day mere{ing plans along with logic to determine the most
optimal metering pattern based on sensor information related to the dcnsiry of traffic along the freeway; or 3) more
complex ramp metering concepts wing sensors inst:allcd along the freeway mainline to detect available gaps in traflic,
which are then wed by vehicles exiting the ramp.
An cxampk of a ramp meter is shown in Exhibit 10-4. The RAmp MAiwgmJmt tmJ C6ntrol H.anJboolt Oacobson et
al., 2006) should be wed as a primary U.S. reference if ramp metering is being considered for implementation. The
reader could furtha refer to guidance produced by the European Ramp Metaing Project (EURAMP, 2007).
~P9r:t
Freeway Entering
RampEntering
::c:
Freeway-toFreeway Flow
Ramp-to-Ramp
Flow
Freeway Exiting
-Ramp - Exiting
Flows at spot locations arc readily collected using automated detectors or can be obtained &om manual counts. Automated sensors can readily be installed on the freeway mainline or on- and off-ramps to the facility. More challenging
are studies on freeway weaving segments. A weaving segment describes a situation where an on-ramp is closely followed by a downsueam off-ramp, and the two are connected by an auxiliary lane. The auxiliary lane is used by both
movements to "weave in and out of ~e main uaffic stream.
For freeway weaving segments, the flow pattern is nor fully described by counts on the mainline, on- and off-=p;
an additional estimate of weaving behavior is needed. This is most easily done by performing a supplemental manual
count to estimate the ramp-to-ramp flow. With that value known, the on-ramp-to-freeway flow can be estimart::d
by subtracting the ramp-to-ramp flow from the on-ramp demand. The freeway-to-off-ramp flow is calculated b y
subuacting the ramp-to-ramp flow from the off-ramp demand. The remaining freeway-to-freeway Bow is similarlY
estimated by arithmetic. Exhibit 10-5 shows a typical weaving diagram for illustrative purpose with four movement::s:
freeway-to-freeway flow, freeway-to-ramp Bow, ramp-to-ramp How and ramp-to-freeway Bow.
2.1.2Speed
Average vehicle speeds at a spot location can be obtained from a "speed uap~ using two dosdy-spaccd detectors or
by other automated data collection equipment as presented in Chapter 5. The speed used in the theoretical ~y
uaffic Row relationships is always the space-mean speed (SMS). As <Uscussed in Chapter 5, the SMS is best estimated
from vehicle uavd time over an extended distance, but can also be calculated from spot-speed measurements using rb-e
harmonic mean speed. A simple arithmetical average will result in the TMS, which in some cases can be significantlY
different from the required SMS (Roess, McShane and Prassas, 2004). Chapter 5 gives additional details and ex.unple
calculations for both measures. If the objective of the srudy is to calibrace a speed-flow relationship, speeds shoul d
be observed over a range of volume (and congestion) levels. Exhibit 10-6 shows the theoretical rdationshlp berwee:: P
freeway speed and How from the HCM. Exhibit 10-7 shows a sample freeway data sec of 15-min. speed measuremen -a:s
on a basic freeway segment obtained from automated traffic sensors that reflect the theoretical HCM relationship.
80
Mmllh
60
55 ml/h
li!
~50
,,
40
.::;:.~.,'
&;
20
,
.,."'
.,."
'~~,~~,. .,."
,:''
.L>
~ ~~
.-
~"
.. f:~~"~
.,.
,
1--
10
,
, ,
-i"Mn:::::::::::::::
1750
lOS A
~ 30
7 ' - - - /_~a-1450
,
,
,
R
C':
10 mi/h
65 milh
70
fae:f!P!..Spdl _ffs.=_7~!l!!.
, .,.
.. ., ,.. .. -- ,
-- -- -
, ,
--- -
~<-,'
~
__ .,..
-~---
--
,.,..
"'~-
-:..~-
400
800
1200
1600
2000
2400
Sou=: Higlrwtzy Czpacil) M1111UAI 2000. Copyright, National Academy of Sciences, Wa.shington, DC. Exhibit 233, p.l35.
Reproduced with petmission of the Tn.n.sporuuoo Rcscarch Board.
70
i~
-a
J-50
~40
>30
f20
<
ObserwdSpeeds
10
0
400
800
1200
Traffic flow
1600
2000
2400
Exhibit 10-7 ~evident pans of the field-observed data match the theoretical HCM speed-flow cdatiooship quire
wdl. The fidd-observcd values that ace shown signilicandy below the speed-flow curve likdy cornspond to unstable or
congested flow periods. The 6dd-obscrvcd data further show only one data point~ a. flow ofl,l 00 pdmilln. Ptcfcra.bly, more data would need to be coUected a.t this loca.rion to validate that capacity is in fact 2,400 passenger ca.rs, per
mile, per la.ne (pelmi/ln).
2.1.3 lkcupamy
Density, the third component of the fundamental flow relationship, is idca.lly meuured aver a segment of roadway as
described below. However, with the presence of permanent traffic detectors, it is often easier to estimate the density
from the average vehicle occupancy on a detector. Detector occupancy is different from person occupancy within
a vehicle. It is defin.cd as "the proportion of time that a deteCtor is occupied by a vehicle in a defined time perioda
(Rocss ct al., 2004). Density can be estimated &om deteetor occupancy using the following equation (and assuming
an ~vc:rage vehicle le~, L).
1 tiA
5280.0
L,+Ld
'0=-I
Equation 10-l
~ere
D .. average density (vehicle/mile/lane (vehlmillnl)
5280.0.150
+
= 24.75 veh/mL/ln
26 6
The densiry is then related th,rough the theoretical speed-How rdationship in the HCM to a LOS. Sina: the HCM
is b2Sed on passenger-car equivalents, the ~cnsity needs to be adjusted for the pcesena= of heavy vehicles in the traffic
~. Assuming the observed measurement does not contain any trucks and following the radial density thresholds in
Exhibit 10-6, a density of24.8 vehlmi!ln results in LOSC (greater than 18 and less than 26).
2.1.4~
~ooo.---------------------------,
25000
20000
E 1sooo
i
,1;
10000
5000
.IS'
----------
lt
I
400
200
0
Mon
Tues
Wed
Thurs
Fri
Sat
Sun
Day of Week
,,-.,.
;r;,,~~'"
50
Legend
40
30
20
10
0
---------
10
15
20
Source: Highwll] Capacity MIZIIuai 2000. Copyrighr, National Academy of Sciences, Washington, DC. Exhibir 8-28, p. 8-26.
Reproduced with permission of the Transponation Research Board.
For HCM-cype analyses, all types of heavy vehicles are commonly aggregated imo a percentage of heavy vd\icles relative to the overall traffic B.ow. For ocher applications, including planning-levd studies and pavement-mwagement
programs, the exact classilicarion type may be of interest. Modern detectors and associated software have built-in
algorithms that convert axle patterns to bins of heavy vehicles. Manual classifications can be performed foliowing
conventions in the AASHTO (2004) Green Book" or Chapter 14.
A headway study can be useful in traffic flow theory applications to evaluate the arrival distribution of vehicles, which
can, for .e-xample, be used as input in simulation models (see Chapter 11). A srudy of gaps in freeway traffic can be
useful for analysis of ramp metering Strategies, where on-ramp demand is released as a function of gaps in the outside
freeway lane(s). A freeway gap study can also give insight into driver-following behavior, where short gaps (at high
speed) are evidence of an aggressive (and risky) driver behavior. Both headway and gap studies are more special.iz.ed
than speed, .flow and occupancy studies. While detectors can readily be used to obtain chese data, they are less fre...
quencly used in nonacademic applications. An example of results &om a freeway headway study is shown in Exhibit
10-9. The exhibit clearly distinguishes headways on the right lane (large variability) and the twO left lanes, where
headways are grouped very tighcly.
2.1.6 Onnp/Umee
Compliance studies are typically conducted at or ncar intersections with regard ro various TCDs. However, compliance can be studied at any location where disregard for a traffic regulation or law may take place. The concept of
compliance studies is simple since they essentially answer a "yes/no" question. However, the acrual data coUeccion can
be challenging on busy high-speed freeways, especially if the intent is co identify violators in an HOY lane. HOYcompliance studies require the analyst to estimate the number of persons within a vehicle, which can.ooc be done
186 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
through automated sensors. Instead, roadside or overhead video recordings or manual observations may be needed
:sample driver behavior.
to
Other compliance studies on freeways are more straightforward. Violators on tolled facilities arc identified automatically if che vehicle doesn't communicate with che toll tag reader. Similarly, violations of no-truck restricted lanes are
obtained automatically from classification data. Chapter 8, "Compliance with Traffic Control Devices,M provides useful material regarding sample size calculations and relevant study methods.
2.1.7 Person Occupancy llnd Bus/Transit
In some cases, a freeway study may include data collection on special vehicle classes, including high-occupancy artd
transit vehicles. Most performance measures for freeways are based on vph mettics, which tend to underestimate
che benefits of HOY and transit modes on the total person throughput on a facility. Person-based measure of cffcctivcsness (MOE) are recommended for these modes, because the multioccupancy vehicle presumably replaces several
(drive-alone) vehicles. To escimau the person throughput on a transportation f.aciliry, a study of vehicle occupancy
needs co be conducted.
Obtaining data on ~chide Occupancy for aa.osit vehicles is most easily obtained from the tranSit agency and usipg
data collection methodologies laid out in Chapter 13 or TRB Transit CapaciJy and; Q}lality of&rviu Manual (TRJ3,
2003). This approa<;h is likely to be more efficient and result in greater accuracy data than any transit occu pancy study
in the lidd.
The person occupancy of HOY vehicles on the ocher hand, needs co be performed manually in the field, and as
discussed in section 2.1.6, is challenging. !>. manual occupancy study is both time and labor intensive, but is mote
reliable and feasible than automated methods that rely on video image processing. Person occupancy is best measured
in transition zones, such as ramps or access points co an ML, since vehicle speeds are expected to be lower. It is u.nrealistic to collect data on every vel:ricle apd instead, analysts should employ a random sampling scheme by sdecti.ng
every n"' vehicle. The value of n depends onthe volume, Bow rod speed of traffic, and the ability to observe and record
occupancy. A constant sampling frequency (for example, 11 10) asSures an unbiased study. The reader should refer co
Appendix C for details on random sampling and statistical tests of significance.
Alternatively, density ~ be inferred from occupancy measurements at spot locations as discussed above. Densiry is
further related to the traffic stream parameters of 8ow {v, in vehicles per hour per lane) and SMS (s, in miles per hour)
through the fundamental equation
d=~
s
Equation 10-:Z.
Density can therefore be computed from measurements of the other cwo parameters (Dowling, 2007). In the HCM.
density is the measure used co define LOS for freeway operations as was shown in Exhibit I 0-6. A visual representation.
of different deruitycondicions and the corresponding LOS caregory is shown in Exhibit 10-10.
Sour= Hjtfnuay CapflriiJ Mllmi4L Copyright, N~tional Academy of Sciences, Washington, DC. Exhibia 13-5 chrough 1310, pages 3-8 chrough 3-10. ~roduo:d wich pennlssion of che Transpon:uion RA:scarc:b Bo:ud.
.._Ur" '""''""'"'''
time delay (TIT>), it is a useful performance measure. Agencies commonly use delay to estimate a congestion use r cost by
. multiplying the total delay (in vehicle hours) by me average vehicle ocrupancy (in persons per vchicle),and men by an eco. nomic estimate of me value of time (in dollars per person-how). The resulting user cost (in dollars) gives a rough estimate
of me economic impact of uaffic congestion, and similarly, the benefits obtained from congestion mitigation strategies.
2.2.3Qunus
Queuing studies are among the most important study types for freeways; however, they are also some of the most
difficult studies to conduct. Queues are a key performance meuic repoHed from the analysis of freeway facilities in
the HCM, as well as simulation-based studies. They are also commonly reported by the media to alert drivers of the
spatial extent of freeway congestion.
Freeway queues are difficult to study due to their sometimes long and variable spacial extent and a lack of reliable
automated means of measurement. There are three basic methods for estimating freeway queues in the field.
1. Th.e analyst can estimate me spatial extent of queues from speed drops on permanent detectOr stations. The
problem with this method is the deployment of mffic sensors is not continuous, md even on very advanced systems, half..mile spacing is common. The agency therefore only has "spotty" data on vehicle queues and further
has to assume that a certain speed or density (occupancy) threshold in fuct corresponds to a queued siruation.
2. The analyst can direaly examine queues in me fidd or from video using visual observation at fixed vantage
points. Since the spatial extent of queues varies, me use of preinstalled freeway monitoring video cameras.is
typically more.rd.iable than observations from a roadside or bridge observer location. Many modem freeway
TMCs have access to a broad deployment of video cameras that can be used to monitor queues from within
the TMC. The collection ofqueuing data'in this fashion is not automatic and requires some judgment about
when a vehicle is considered to be in a queued state. The precise location of the back-<>f-queue can also be difficult to see depending on the video angle. Chapter 6 offered a methodology for performing a queuing srudy for
the approach to a signalized intersection. The same general methodology applies here, with the caveat that the
qurues are much larger in both spatial extent and temporal duration. For this reason it is acceptable to increase
the analysis intervalli:om 1 to 5 min. and use the distribution of 5-min. back-of-queues to describe the queuing
system.
3. The analyst can record queues from within a vehicle traveling in the opposite direction of the congested freeway
facility (typically driving slowly along the shoulder with hazard lights on). Peak how traffic patterns are typically
di.rectional and conditions in the opposite direction are therefore less congested. Given fast freeway speeds and
the cognitive task involved, the ai:lalyst should not be the one operating the vehicle and should not disttaet the
vehicle operator from the driving task. Queue lengths can be related to mileposts at the side of the freeway or
through waypoints in GPS equipment. This fOrm of queue srudy can also easily be combined with a travel-time
srudy of the congestod facility on the return trip.
The selection of an appropriate queue study methodology ultimately ilepends on detector coverage, the presence of overhead video cameras and the ability to see the spatial menc of the queue &om video or the opposite ttavd din:ction. Sample
size is also ~ important consideration, since travding co and from the 6dd can be timcxonsuming.
Another important point in relarion to freeway queuing is that queuing patterns are highly sensitive to the source ofcoop
cion. For cx.ample, a queue may be caused by a 6=1 boaleneck where demand =ds the (constant) capacity. In this case
the queue will grow upstream of the bottleneck location. As demand drops, the quelic will begin to clear from the back
towards the bottlene4: location. On the other hand, if the queue is caused by an incident such as a collision or cemporary
lane dos~ the segmtnc demand may be constant, while the capacity changed. The reduced capacity will once again cause
a queue that grows upstream of the bottleneck location. However, once the incident is cleared and assuming constant demand, the queue will noW' dear &om the front, which may cause some data collection challenges. In another caveat, a free"
way queue may be caused by a moving congested source such as a slow-moving heavy vehicle or a military convoy. In these
cases the low-capacity bottleneck is mobile and the resulting queues roll with the congtstion source. The analyst needs to
be aware of the particular queuing paaems in the studied segment and needs to adapt the data collection strategy accordingly, possibly utilizing a second moving vehicle in the opposite direction of aavd collecting similar data at set time intervals.
Queue srudies are also performed in rdation to ramp metering strategies, to evaluate the queuing patterns on the metered ramp. These are conceptually similar to queue studies at intersections as discussed in Chapter 6. Ftnally. queue
studies are common in simulation applications presented in Chapter 11.
2.2.4Safay
Fre~vay operations arc substa.ntially affected by crashes and other incidentS. A significant amount of con~tion in the United Stares is the result of incidents resulting in huge annual productivity losses, in addition to the potential loss of human life
and personal injury (PB Famdyne, 2000). Safety sru.dies on instrumented freeways are aided by
availability of cmtera
systems. While these systems typically are not used to record continuous observations (and therefore won't have a record of
the incident), they do allow the agency to quickly identifY the location of a crash. This is especially hdpful in reducing the
amount of incident clearance time, which can significandy affect the operations of a freeway facility. In tenns of user-cost
elfeas on nonrecurring freeway congestion, reducing incident response time is a key target for many agencies. More generally, safety studies are used on freeways to evaluate specific design features (barrier type, median width, cuCve radii) or policy
strategies, such as reduced speed limit, truck restrictions, or enhanced enforcement. Chapters 17 and 18 provide details on
collision stUdies and alternate safety sru.dies that are also applicable to freeways.
me
Work zone stUdies are gmeralty not difli:rcnt from other freeway sru.dies, as me same pcral measures are of interest: volumes, speeds, rravd rimes and queues. An important difference h~ is the potential lack of automated data sources, as
sensors are often 1m1<rnd for resurf.cing projects or lanes are temporally shifted away &om aisting sensors.
The 'lb.r* ~ OpmuWns Best Pr:zctias Guilkboolt (FHWA. 2000) is a resource designed to give State ~d local tranSportation agmcies, construction contru:tors, transportation planners and others with interest in work zone operations access
to information and points of contact about current best practices for achieving work zone mobility and safety. Additional
resou.rccS are available from the FHWA Office of Opcracioos WoJk Zone Mobility and Safety Web sire at http://ops.fhwa.
dor.gov/wzfanda.asp
2.3.3 ~ &nurin
In recent yeats, much national attention has been devoted to c:rnergency evacuation strategies in response to narural disasters
(hurricanes) or th.tea1:s to national security. Whatever the scenario, transportation engineers and planners have devdoped
evacuation t:raflic plans that commonly involve special lane rcversal.s on interstates with lanes in both directions moving
traffic away from the disaster area. While these reversal facilities typically fall into the freeway functional class, the aitical
congestion points are 1oca.a:d at me transition points to and &om the freeway lanes (Kuhn er al., 2005). Oftentimes, Law
enforcement has to employ special temporary traffic control Strategies to direct drivers to the evacuation &cility. As a result
ofthis cemporuy aaffic control and c:xpectcdly high demands, the capacity of the transition points is expcctedly much lower
than the evacuation freeway capacity and should be the focus of any sru.dy of evacuation scenarios.
Since data coUeaion during a na.rural disaster or in a national case of emcrgcncy is unlikely to be performed manually, permanendy installed sensors are the best source ofdata. .Alternativdy, many evacuation strategies and scenarios are tested using
simulation computer models; Chapter 11 provides more detailed discussion of those.
190 MANUAL OF TRANSPORTATION ENGINEERING STUDIES. 2ND EDITION
This chapter addresses data collected on two types of &ecways: GP and ML facilities. While chc traffic operations on both
facility types are generally similar, the dara coUeccion praaices differ in two ways. FllSdy. ML facilities are often outfitted
with more sophistiC!ted data coUeccion equipment, allowing fOr more automated and more detailed srudies. In addition,
ML facilities have additional and unique performance measures associared with them that are clifferent &om general freeWay
facilities. These measures are rdared to vehicle occupancy, intelligent cransporcation systems (ITS) straregies and real-ciJne
~eed measurement for dynamic pricing in HOT lanes.
This fundamental rdationship is-the underlying basis of freeway analysis methodologies in the HCM (TRI3, 2000) and als 0
finds apptiC!cion in simulation analysis :IS diScussed in Chapter ll. Speod, flow and density are key performance measu.te5
on freeways and are therefore common measures included in freeway srudies. Today they are most commonly measured
using in-road or roadside sensors; however, it is somerimes necessary co collea data manually or chrough nonpcrmanc:Pt
means. Since density is difficult to measure in the fidd, detector ocrupancy is a common substitution. Detcct:ors are also
capable of providing analysts with an estimate of gaps in c:raffic, which may be important in some circumstances.
In addition co the above fundamental measures, freeway studies are often interested in the composition of the c:raffic ~
and the classi.fi.Cltion of vehicles. The presence of heavy vehicles directly influences the op=.tioos of chc traffic sueafll
especially in mountainous terrain. The vast majority of detectors are capable of providing.vehicle classilicarion ifthis dlJ. ta
dement is necessary.
Fmally, the safcc:y of motorists is a primary concern to engineers. Safety can be comP.romlsed in many ways. For insranC~:
during congested conditions, c:raffic often becomes unpredictable as ~des begin to accelerate and decelerate em.ticallY
Facility improvementS such as HOV/Har lanes or reversible lanes could alleviate a large portion of the u:ailic collisioll.l tb.:lJ.C
cause craffic to act unpredictably. Since many of the MI. aeaonents are installed co prevent unnecessary aaffic coUisiooS
sa6:ty srudies should-be condUcted and updated frequendy co determine the long-term sa6:ty dfcctiveness of the facili t:Y
Chapter 17 provid.-more information on how to ~cifically conduct safety studies.
against data collected by automated means, or along segments that do not provide consistent and accurate data using an
automated format. For instance, it may be important to collect data at twO closely spaced interchanges where the weaving
segment is very shon. The weaving narure of traffic ar this locarion would nor lend i!Sdf to automated dara collection methods because vchides would fRquendy be double-counted while changing lanes, not to mention that traffic may be traveling
coo slowly during rush hour through one or more of the lanes. Lastly, it is also po55ible thar an agency might not have mearu
or funds to c:Ollcct data in an automated fashion, and uses more rudimentary methods to updare outdated data or as a spot
check for assumed growth patterns. Typically, data collected in this manner are for traffic counts, speeds and classification.
In addition, ~en safety studies are conducted, it is nea:ssazy tO collect collision dara for the segment of intcn:st. Typically.
the sate department of cranspon:arion or local municipality can provide the ncccssary data; however, it is po55ible thar the
. analysts will neal to oollect the data on their own. Segment analysis methods would likdy need ro be employed using specific mileposts along the segment. Chapter 17 should be consulted for sample siz.e calculations and further reading on how
t<Hibnduct a collision srudy along a segment.
"
l..asdy, it is often ncccssary to conduct surveys of the users of a new freeway or ML &cility to detennine the perceived effectiveness. This is especially useful when comparing aaual quantitative results to more qualitative results. Information on
: }iow..to condua surveys can be found in Appendix B, survey Design.
Freeway monitoring is typically done through the use of a permanent count station. Typically. a data collection location
takes the form of a small roadside station called an auromatic traffic recorder (!UR) or is part of a larger weigh-in-motion
(WlM) station.
are wually collectrd 24 bows per day, 365 days per~
':Thea
An ATR will usually consist of a cabinet cont:aining a traffic recording unit which stores dara from one or more roadside
or in-road detictors as described in Chapters 4 and 5. Many times, the data from an ATR is collected at a centralized
19C3(ion via aninternet connection on an as-needed basis. The detection type and daca collected could cake many forms
depending on the type of detection used. Detection at ATRs usually uti.l.iza miccowavi:, point detectors, video detection,
or surveillance cameras.
Permanent data recording stations are also integrated in WlM stations. WIMs (Exhibit 10-11) are primarily used ro enfo1ce
heavy vehicle axle weighcs; however, many times othc:r forms of detection are installed at the adjacent mainline &ciliry to
acquire other forms of data. AI. ATR and WlM stations, dara can be downloaded in a raw form, but in many cases sofiware
will.automarically reduce and tahulare dara in the format the user .rtqu.estS.
.
There is an array of detection devices available for collecting many types oftraffic data. Most are described in Chapters 4 and
. 5, and include inductance loops, point detectors, video, infrared. miccowave, radar, acoustic and semipermanent deteaors
thar can be removed for use ar othc:r sites. Alccmatively, auromated dara collection equipment, iri the form ofemissions and
pavement monitoring. is becoming ioaeasingfy popular along freeways. The most common use of dara &om continuous
'. monitoring sWions is traffic counts for det.cnnining ~annual daily traffic (AADT), typical daily traffic ~lurnc patrems. seasonality f2aors and to estimate vehicle miles cravdlcd (VMT). Each detection device has trade-offi; for instance,
- 5ome are less acrurare or do not perform certain types of detection, some require increased maintenance, others allow more
flexibility to change detection locarion, while others are more intrusive or cequire special mounting angles so vehicles do not
visually occlude adjacent detection mnes. For continuous moniroring purposes. the detection types used are more likdy ro
be mounted 01 installed in a predetc:rmined locarion ro allow for more stabk daca cx:dlcction. For more information on the
tt.ide-ofl's between various detectors, users shoukl consult the FHWA Trrzffic lkttor HIINiboo!t (Klein et al., 2006).
Although ma.!!r of the same types of data collected. by automattd means can be collected by manual or nonpermanent
.methods, the.cquipment used is much more restricted. Manual data collection can be done by handheld count boards or
. by using alaprap with a time-based maao such as the one described in Appendix E (Exhibits E-1 and E2). Often, a video
camera is used to record a specific rime period of events so many types of dara can be collected in the office at a later time.
, Nonpermanent methods of data collection commonly use detection through pneumatic tubes or movable point detectors
communicating ro a roadside data collection unit. These are described in detail in Oupter 4, volume Srudies." A small
number of~encies and unMrsities havt alluircd equipment for video dctcaion tba.c can C3$ily be mm~ tO any necessary site. H~. these methods are typically used for research purposes since dal3 may be needed ar a specific location
where continuous monitoring stations are not located. Exhibit 10-12 shows an example of equipment and a schematic of
fidd deployment ar a freeway work moe.
.,,
I
<.
is a useful method for obtaining such a vantage poinc;, it is also much safer to gather the data in the office chan nor the
., roadway. Surveillance videos should be considered if observation cameras are already installed along the roadside.
' During the installation of ci:lta collection equipment on freeways (including nonincrusive roadside units), data collectors should ensure activities do nor interfere with traffic operaLions nor cause significant distraction to drivers.
This sort of capacity-reducing "friction" from driver distraction is common for freeway work wnes, wh ere additional
measures are taken (such as advanced signing, barrier placement) to minimize the impact on the uaveling public. For
freeway data collection activities, the data collectors should be sensitive to these issues, and minimiu interference
with traffic operations. This aspect is particularly important, since the focus of data CQIJection is cypically to measure
"standard" operating conditions, without the preseno: of additional driver distractions.
Lasdy, evaluating the effectiveness of various freeway strategies should not be con~ideced a one-rime accivicy, but parr
of a periodic review of the component or strategy employed, as well as the overaiJ system performance. This is especially true just after a freeway management strategy is fust employed. It is recommended formal evaluations of a new
facility take place several times during the first year of operation. EvaluationS at 1 month, 6 months and 1 year ~.fter
insrall:uion would be ideal.
li.
i
i
700
600
+---------------------------------------~~--------------
500
+-----------------~------------------~
:\/.!.
400
1
J
AIITraffl<
a c ars
300 +---------------~
Trucks
200
100
I.
.5 6
10 11 12 13 14 15 16 17 18 19 20 21 22 2.3
l
:
lS
1
l 20
lllane4
.E
lS
.I
10
II lane 3
~
'S
Dlane 2
Blanel
s
0
..
....
9.
....
t1 t1
"' ...
.,
...
t1
....
r:'!
"'
'!'!
'!'!
0
~
....
~
'!'!
.... ~
.;;
9.
t1
t1
t1
t1
.~
'!'!
'!':
....
'!':
'!'!
....~
With the availability of lane-by-lane sensors, it is further possible tO evaluate traffic patterns by lane as is shown in
Exhibit 10-15. The Exhibit shows the observed 5-rnln. counts of heavy vehicles per lane over a two-hour period.
In addition to craffic volumes, analysts arc frequently interested in dau. of freeway flow and the associated SMS. The
paired speed-flow obsclvations are commonly displayed in a graph as was shown in Exhibit 10-7.
The performance of an ovctall facility over time and space can be represented in a 3-dimensional contour plot. Exhibit
10-1~ shows the predicted freeway density and SMS over 18 analysis segments and 10 time periods. The graphs in
the Exhibit were generated from the FREEYAL compuu.cional engine for the HCM 2000 freeway facilities chapter
(fRB, 2000), but could easily be created for acrualfidd dau. as well.
__
.,...._~
-e......nCON4Il-l
...
.......
-.........
...
.......
.........
.......
.....
.......
.....
.........
....
.....
....
"
,.,.
~-
II
..
,,,~
\
'
""
u~IIIAI n~ TIIAN~PI'1RTAT11'1N
--
IAAuuu.,u.,uu'
--
~ 4.2
System Monitoring
JFor &ec:way systems with broad and continuous sensor coverage, comprehensive freeway performance monitoring
~rograms are employed to trade system performance (Ncudorff ct al., 2003). For this purpose, FHWA defines a series
of system-wide performance measures that are derived &om various field data sources. The performance mctrics are
conccptw.lly divided into 1111Jbili1J avtrages and rruabiliiJ mtasum. The following definitions (in italics) arc given by
FHWA.
.
Average Mobility Measures (Definitions &om Neudorff et al., 2003)
Tram-time index - This a rttlio ofITIIvtl conditioru in tht pea! perUxi to a target or accrptabk lfflvtl condition (typically
foe-flow conditioru art used). The trawl-timt indcc indicates how much longer a trip will take during a pta! time. For
tx4mpk, a travJ ti71ll! indcc of1.3 indicaw tht trip wili talte 30 ptrctnt longer (1.3 timeJ longer).
Pcn:ent of congested tram- This is primarily a syttnn 7111!4SUrt that quantifies the exttnt ofcongestion. A fot-.Jiow spud
is used as a congestion "bmch11141'1t" and any trawl on a road stction for a time ptriod that is 1m than the fot-.Jiow spud is
tktnmirwi to he congested. 'J.:ht congested trawl is sum71ll!d and thm divided by totiJi trawl esti1114UI.
Travel in this case, is defined as a measure that quantitatively describes the amountof traffic flow. Common examples
include VMT or vehicle-hourS:-tcaveled.
Delay per penon - Exprmed in pmon-hours peryear, this 71ll!asurt is used to nduce the tofiJi travel tklay va/ut to a figure
that is mort rrlatabk to user txptrimct. It also normalizn the impact ofmobility projtcts that handk much higher ekmand
Buffer index- This measurrtxpmm the a1111Juntofextra "buffer" nmkd to be on-time 95 pm:tnt ofthe rime (/au I day
per month). Trawkrs could multiply thtir average trip time by the buffer irukx, thtn add that buffer ti71ll! to their trip tQ
msu~ thty will be on-ti71ll! 95 pm:mt ofa// trip1. An advantage oftxprming the n/iabilil] (or la+k thenoj) in thu way is
that a pm:tnt valut is distance- and time-ntutral.
Percent "Yariation - Also !mown as the cotjficimt ofvariation, this is the lt1111Junt ofvariabiuty in relation to avtrage travel
conditions. It is calculauJ as the sran4ard deviation Jjvifkd by the mean. Trawkrs cou/J muhiply their avtrage travel tinu
by the pm:qzt variation, thm add that product to thtir avtrage trip ti71ll! to gtt the .time ntetkd to be on-time about 85
pm:mt ofthe time (one stan4ard deviation above the 71ll!an). Higher vdiuts indicau kss rtliabiuty.
Misery index- This 71ll!a.!Urt atttmpts to quantijj the inttnsily oftklay for only the worst trips. The avtrage travel rau is
subtraaedfrom the upptr 20 pm:tnt oftravel rafts to get tht A1111Junt ofti71ll! btyQnd the avtrage for some 41111Junt ofthe
slowtst trips.
When deciding on a system-wide performance evaluation scheme, an agency needs to decide which of these measures
it wants to usc to track performance. It then needs to ensure the sensor instrumentation and detector coverage are
sufficient to deliver the desired data. NCHRP Synthesis 311 (Shaw, 2003) identifies additional performance measures
to supplement the ones discussed above and shown in Exhibit 10-17.
., (- _ m
o;-
Person-miles travded
Persons moved
Bridge condirion
Safety
Toll revenue
Incidents
Truck-miles rraveled
Trucks moved
V/Cratio
Pavement condition
Vehicles moved
VMT
: 5.0 REFERENCES
~ American Associ arion of State Highw:~y and Transportation Officials. A Polil) 011 Gtommic D~sigr. ofHighways and Smrtr.
,Washingto n, DC: AASHTO. 2004.
Dowling, R. Traffic Analysis Toolbox Volumt \11: Dpnition, lnttrpretatilln, and Calculation ofTraf!ic A1111/ysiJ Tools M~asurn of
Efficrivmm. Report FHWA-HOP-08-054. Washington, DC: Federal Highway Administration, 2007.
EURAMP Handbook of Ramp Metering." Euro~an Ramp Mtttring Projm (EURAMP) &porr, September 2007.
Federal Highway Adminisuacion. Fruway OptTatillnr Handbook. Report FHWA-OP-04-003. Wasbingcon, DC: FHWA, 2003.
Federal Highway Administration. WOrk Zone Operations Best Practicer Guiekbook. Report FHWA-OP-00-010. Washington, pC:
FHWA, 2000.
Federal Highway Administration. Work Zone Saftty and Mobility R11k. Federal Regisrer 69, No. 17 (September 2004).
Hauer, E. Ob$tT'IIarionallkforr-AjitT Studies in Raad S4foty. Amsterdam, Netherlands: Elsevier Publishing.1997.
Jacobson, L., J. Suibiak, L. Nelson and D. Sallman. Ram; Manag~ment and Control Hllndbook. Report FHWA-HOP-06-00 1WashingtOn, DC: Federal Highway Adminisuacion, 2006.
Klein, L, M. Mills and D. Gibson. Traffic Detector Handbook. Report No FHWA-HRT-06-108. McLean, VA: Federal HighwaY
Adminisuacion, 2006.
Kuhn, B. er al. Managtd lAnes Handbook. FHWA!fX-06/0/4160/24. Auscin, TX: Texas Department ofTransporcacion, 2005
Neudorff, L. G., J. E. Ran<WI, R. Rei5.l and R Gordon. Fruway Managemmt and Op~onr Handbook. Performed by Siemens
lTS. Report FHWA-OP-04-003. Washingcon, DC: Federal Highway Administration, 2003.
PB Farradyne, Tmf!ic ltuitlmt M=g=mt HandbDDk. Washington, DC: Federal Highway Adminisuacion, 2000.
Perez., B. G. and G. C. Sciara. A Guitk for HOT laM Dntlopmmt. Report No. FHWA-OP-03-009. Washington, DC: Federal
Highway Adminisuacion, U.S. Deparunen.t ofTransportacion, 2003.
Rness, R., E. Prassas and W. McShane. Traffic Enginuring, 3rd ed. Upper Saddle River, NJ: Pearson Prentice Hall, 2004.
Shaw, T. NCHRP Synthesis Report 311: Pnformanct Mt/Jfllm ofOptTafiona/ Effittivmm for Highway Stgmmt> and Systtms.
Washington, DC: National Cooperative Highway Research Program, 2003.
Texas Transportation I.nstirute, Parsons Brinckerhoff Quade & Douglas, Inc. and Pacific Rim Resources, Inc. NCHRP Rcpor1:
414: HOY Systtms ManU4I. Washington, DC: Transportation Research Board, National Research Council, 1998.
'
Transportation Research Board. Highway Capacily ManuaL Washington, DC: TRB, 2000.
Transportation Research Board. TCRP Report 1100: Tmnrit Capanty and ~lity ofSmtU-t ManU4L Washington, DC: TRB.
2003.
19~
--
Chapter 11
... .. ........ .. ... .. .. ... ... ..... .... ...... ................. ...... .........
...... . .. ..
Simulation Studies
Bam-]. &brodn; Ph.D.
1.0 INTRODUCTION
201
201
202
202
203
1.5 Definitions
205
206
2. I Sensitivity Analyses
206
208
210
212
215
217
218
218
221
223
227
228
229
229
231
4.4 Documentation
232
232
5.0 SUMMARY
233
6.0 REFERENCES
234
1.0 INTRODUCTION
1. to provide guidance on how to design studies using simulation tools as an alternative tO field data coUeccion;
2. to provide guidance to simulation-tool users on how to march field data to simulation inputs and outputs
through calibration and validation; and
3. to provide nonusers of simulation tools a basic undemanding of assumptions that go inro a simulation
study and the relationship berween simulated and field-measured data.
The focus of thi.l chapter i5 primarily on microscopic simulation, which is a commonly used form of simulation in
uaffic engineering and planning practice today. The chapter will reference mesoscopic and macroscopic simulation
models as necessary, especially in cases where their applications differ from microscopic models. The differences berween these three cypes of simulation models arc discussed below.
The chapter focuses on stochastic uaffic simulation models, which represent complex representations of traffic systems
in a computing environment. A stochastic traffic simulation analysis typically involves one or more of the foUowing:
the aim ro evaluate a complex uaffic condition that cannot be readily addressed through deterministic
(equation-based) analysis techniques;
the availability of very detailed data to properly calibrate and validate the model;
an initialization period to populate the simulated network before extracting any data;
the use of multiple simulation runs and the statistical analysis ofstochastic outputs; and
input parameters that are preoptimized using an external tool. At the time-of this writing, simulatio~ tools
(for example, cycle lengths, splits, offsets) typically do not optimize; other tools or methods are needed to
obtain the ~t possible set of input parameters.
Mesoscopic simulation also models individual vehicle movements, but predicts vehicle behavior in the next time-step
based on macroscopic uaffic stream models (speed-flow-density relationship). Computations in a mesoscopic model
are performed at a more aggregated level and are thus more efficient. These models lack the necessary detail for microscopic behavioral modeling (for example, gap acceptance), but have broad application for nerwork-levd analyses.
Mc:Soscopic models are commonly used in relation to dynamic route choice algorithms that predict driver response co
short-term or long-term sources of congestion.
Macroscopic simulation models aggteg-4te vehicle flow (not ind.ividual vehicles) based on macroscopic traffic Bow
principles. Macroscopic models predict traffic conditions for segments rather than individual vehicles. They offer
202 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
a less-detailed analysis, but are more computationally efficient than the former rwo
cat~ories. Macroscopic simulation is
. based on modeling the facility as a sceadysr:ace system and describing its behavior
using a long-run average representation of
an analysis period measured in minutes or
hours. This approach is fundamentally different from microscopic and mesoscopic
simulation, which modd individual vehicles and driving events on a second-bysecond (or even subsecond) basis. Exhibit
Il- l shows a schematic representation of
the three simulation analysis levels.
It is also import~t co !iistinguish all three
types ofsimulation from deterministic traffic analysis tools. Simulation is a stochastic
process char randomly generates rraffic
and driver behavior. Simulation models
estimate traffic behavior based on distributions of many behavioral parameters. Since
simulations are stochastic, multiple mode;!
nms with different random number seeds
will result in different solutio!?s that need Source: TSS-TC20Sp0rt Simulation Syscem.s S.L
to be averaged or analyzed statistitally before reporting. Deterministic tools are based on empirically or theoretically-derived equations that give a unique and
constant solution from a common set of input parameters. Examples of deterministic models are those in the Highway
Capacity Manual (HCM) (TRB, 2000), and software packages that are based on HCM theory. The stochastic nacure
ofsimulation and principles for reporting simulation cesulcs are discussed in more detail in Section 4 of this chapcer.
assessment of opemtional issues that go beyond the limitations of deterministic analytical tools, including
oversaturarion, queue spillback and blocking of rum lanes;
multiple cransportatioo facility typs, such as a freeway intersecting with swface streets;
different intn"sl!ction configuraJions, such as a roundabout, in a signalized arterial corridor;
variow mode oftransportation such as cars, transit, pedestrians and bicycles;
'ir.t'iTflliilffin~:J.o.-11.~..;.~.......,.., ,
NGSIM Algorjthm
Development Process
~~ext
Generation S t~ulation
So wee: FHWA. NGSIM TllS!t E. I: <Art Alprithm A=ummt, FHWAFHWA Traffic Anaiyris Toolhox Wllum.! m HOP-06-0009, 2004.
(Dowling. 2004) contains guidance for selecting a particular class of analysis roo!. The discussion distinguishes among various software categories, including
the different simulation classes, decermin.i.stic tools, and uavel demand modeling software. With the help of a publicly
available spreadsheet, the user can enter deta.i.ls about the intended analysis and receives a recommendation for the
model category (hue not a specific software package) that is most appropriate. Exhibit 11-3 highlights the criteria used
i.o the FHWA Toolhox to sdecr a traffic analysis cool.
,\J:_t:.,~;~ Crn~t.~,l:
..........
looWo<l
s._.,.
CDnYJ/
Smlll-\
SOV
HOY
f.-.yWpt
.-.,.
11&0
Tn4
ln<i<knllllp!l
HOVI.oM
~
8icydo
In-
Mtriol
~.,.,
.....
(2.3.:1<)
-... .,
~ew-
o A..oliouyl.oM
Lerw
Moria! "'""'
"=r""'
lone
Wed
SpcE..,..
APlS
ATIS
DKwcriic:
,.,_,.
r-nctt..w
lluot.erw
QX
ToG ftu&
Uptii&Ol.lne
AVC!I>
w.tt.r,......
CVO
TOM
....
Dmnloo
r..Trip
..
.~
!of'!'i<$iolll
..__,
.,..,.,_
~
a..,.
r=::
---
Tool Q;>iool
Coot
611oo1{Coot/
IW<nh!p
......,_,w.o.
LOS
Speod
TdftiTimt
VolwN
Tnt't'C1 Diltlnct
AVO
/cRAtio .
Domlty
YMT/I'WT
VHI'/l'liT
Dolay
~~
so.,.
~
n RdlabiUty
._,c..t
t.lod.eSpllt
Tnl~Ws)
e-c~u..
-....TNIOid
a..-..
Do
..
~
~
u
... s.ppc.t
~=-:
.......,_
Oelaalt vaN.
So)lt'CC: FHWA TftZjfo .ANJyris TIJ(JIJJIJX VtllutM I. Trttjfo ANtJysis Tools Primn, FHWA-HRT-04-038, page 17, figure 2.
1.5 Definitions
The following terminology is used in this c:b2ptcr co describe demerits of a. simulation study. A simulated 1Jffem describes
the extent of the analysis dolmin in space, level of compkxity a.nd temporal duntion. The nnworlc rcfecs to a collection of
stm:ts, highw.ays a.nd inr.erscctions dw defines che physical c:haractcristics included in thc simulation srudy. The network
carries simulated automobile, transit, &eight, bicycle and pedestrian traffic, depending on che scope of the analysis. A simulation ti!ol refets co a. piece ofsofiw:ue that is a.wilable (for pwchase) to perform simulation mulies. Each simulation tool has
underlying~. which are sets of rules char describe a particular component of driver, pc:dcscrian, or bicyclist beha.vior
{for example, car-foUowing, Ian~ or gap-acceptance). A mcJJ generally ref= to a computational approach for
~ traffic How, which includes dmnninistic {equation-based} procedures found in the HCM, as wdl as sthastic
{probabilistic) ~ods found in simulation. In the context of this chapter, a. 1710CklWill be used to describe a simulation
srudy lile, including ill dements of the simulated network. traffic inpua, beha.viora.l rules and con.figuntion ofoutput.
The concepts of calibration and valida.tion a.re central tO the quality of a simulation srudy . While chese are cliscussed
in greater detail in Section 3.0, ba.sic explanations a.re given here. Calibration represents the underuking of specialized field studies to meuure specific traffic How parameters (or cliscributions) that are used as in~tJ in a simulation
model to replica.te driver behavior. The pa.runeters needed for calibration, and the method of fidd measurement, are
ofun specific to the model being calibrated a.nd can vary by geographic location, facility cype, or rime of day. Once
measured by a.n agency a.nd provided to the model, the agency can use the model at vuious locations in its jurisdiction
knowing the simulation model should replicate driver behavioc. IXfotdl pa:ram= tmd distributions are often a.dded
to a model in case the user cannot obtain calibration data. Modd defaults should only be used as lase resorts,' or if
prior analyses have established defaults are acceptable for the simulated purpose. Ana.lysts should not blindly rely on
defaults, a.nd should scrutinize ea.ch modeling pa.ramecer, especia.lly if they are new users of a. particular software tool.
Validation represents the comparison of estimates obtained from the simulation model (t>Utputs) with similar measures
obtained from fidd data for a common location (for example, intersection, street, highwa.y, etc.). The lcvd of agreement obuined from this comparison SetVeS as evidence of how a.ccuntdy the model reflects the observed conditions.
If the a.ccuracy is not acceptable, then calibration is needed. If the valida.tion a.ccuncy aftet calibration is not accept
able, the model is repla.ced a.nd an alternative ev;aluarion method is sought.
For more general deli.nicions of other tertns, the rea.der is referred to the Glossary {Chapter 2).
2. Evaluating Alccmatives
3. Predicting Behavior
~~\-:p:_~
oU!l-t
2400.-------------------------------------~
~ 2200r------------------------~~~--_J
~ 2000 r-----~----------~~~~::==0------1
! 1800r--------------~~~~~======~-----1
1::
'S 1600
,------"')i;;!!!~~~::;;;:::~--=-=-===;
1400 +------~
:I
1200 12:'.....
..... 1000
1000
..c:
Car Following
Se 1'15itivity
M.Jitiplier Values:
1500
2000
2500
1oo
.... - 175
-74E--2oOJ
Source: Holm et al. FHWA Traffic .Analysis Toolbrc \.11/unu IV: Guitkli= for Applying CORSIMMicrosimulazion Modeling
Software. FHWA-HOP-07-079, 2007.
2.1.2 AmJysis Steps
1. Define objUYe: The objective should be clearly defined at the ou~et. It is hdpful to lay out whar me
goals of the srudy are and how the study will be undertaken. The analyst should think about the types of
output and performance n=.sures chat will be needed. Ar this stage the malyst also lays out data needs
and, if necessary, plans supplemental field data collection to use for model calibration and validation. Th~
objective should be defined in light of an understanding how MOEs are expected to vary as model inpuc
parameters are modified.
2. Set 11p base model; Every simulation srudy should use a bau 11UJtU/ that reflects existing or baseline con eLitions. Later analysis steps then modifythe base model in some fashion. The base model is what the analyst:
uses for calibration and validation. The base model generally includes network geometry. traffic volllllle.s
(all modes), traffic composition, routing information, behavior rules including speeds, traffic conrrol stra..:-.
egies and data collection demenrs (such as detectors) that~ provide the actual simulation results. Modc:::l
serup is discussed in more derail in Section 3.1 and gui<iaJ'ICC for collecting data from simulation is given
in Section 3.2.
3. Calibrate and validate base model: Once the base model is completed, it needs to be calibrated to ens~~
that traffic behavior and initial resultS are reasonable. This is an essential step in any simulation study,
since it ensures and documents the correct application of the simulation resulr.s. Calibration and validation can be conducted wing field dara, theoretical craffic flaw relationships, other software models and ro a
limited extent visual inspection. This is discussed in detail _in Sections 3.3 and 3.4.
4. Adjust sensitivity parameters: In a sensitivity analysis, the analyst next varies traffic patterns, control
strategies, geometric conditions, or parameters in behavioral algorithms to test the underlying hypothesis.
Scenarios in the sensitivity may include a general percentage increase or decrease in traffic volumes to rest:
the margin of error in a traffic forecasting srudy, or changes in driver routing behavior. The specific tC!t
conditions depend on the srudy objectives.
5. Cbcck reasonableness: It is important to consider if the simulation analysis and result~ are reasonable. For
=ple, traffic volumes should only be increased to a levd that is consistent with the capacity of the simulated.
link or node. If the base model already operates close to capacity, a doubling of traffic volumes willlikdy not
prodtiec any meaningful resulrs, since extensive queuing will nor provide real.i.scic or useful resula)be analyst
generally learns more about bottlenecks and solutions in a system that is only somewhat oversarurared.
Simulation Studies 20-;;11""
The reasonableness check should be performed in light of known or expected capacity constraints a.nd
include a visual inspeaion of the model prior to performing multiple iterations for analysis. Severely oversaturated simulation models sometimes do not record accurate performance measures since congestion may
prevent a portion of vehicles from entering the network. While these dmied tntry whicks are typically recorded in an output file, they are generally not included in delay and queue length calculations.
6. Perform multiple runs: Simulation is a stochastic process and all results are subject to random variability
in vehicle generation, driver behavior and other variables that are drawn from a distribution. It is therefore
critical to perform multiple iterations or runs of a simulation model. Results arc then reponed as averages
of all runs with a measure of the variability between runs. More discussion on the number of runs and
sample size considerations for simulation are presented in Section 4.0.
7. Analyu resulu: At the conclusion of a simulation study, results are compiled and a.nalyzed. Depending
on the specific software applications, some analysis may be performed in a post-processing operation using spreadsheet software. Sections 3.0 and 4.0 provide additional detail on the types of results that can be
obtained from simulati.on and the means of reporting and displaying results in tables, figures and animations of the simulation model. When conducting a sensitivity analysis, it is sometimes useful to compute
the change r11tio as the change in an output measure divided by the change in an input measure of interest.
In econometrics, this ratio is also called elasticity. In calculus, it is a diffn-rntial.
An imponant caveat in the evaluation of alternatives is related to traffic conttol strategies that require some form of
optimization .(for example, signalized intersections, or ramp metering on freeways}. To ensure different alternatives
are compared adcqua~ely, each should be simulated in an opti.mal manner. If the signalization is suboptimal, then the
interpretation of the differences between alternatives is confounded by unknown differences in signalization quality,
as discussed in NCHRP Repon 457: E1111/uating Intmtctii)1J ImpT1111t7M11t:s: .An Enginming Study Guide (Bonnesoo,
2001). Since many simulation tools do not include optimization routines, analysu often have to perform a separate
optimization in a deterministic macroscopic model and then use these optimal parameters in the simulation-based
coq~parison of alccmativcs.
In a variation of this srudy, the analyst may perform a strrztegy optimizfuion. The objective is to identify the operational
strategy that provides the best operation at the subject facility. The analyst applies traffic operational factors (such as
speed limit, signal timing, ramp metering) and varies inputs in an iterative manner to find the optimal operating conditions. The analysis may include simple design changes (fur example, add a turn bay, drop a lane, lengthen a ramp).
Person Throughput
K St Connecticut Ave to 18th St: WB
'.'.=/,'~,
,"{'c~
I'l _:z:,
<.'
:.};~
i:{..
),
}~
,-
-;--'
Source: Kindson, W. FHWA Tmffo Analytis Toolhox ~lume V: Tmffo AtuJysis Tools Cue Stwlks: &Mjiu tm4Applications.
FHWA-HOP-C~05. Figure 4.3, page 16, 2004.
Rather than comparing competing alternatives for a corridor, the objective is to optimize ics operations with a mix of
strategies. Note that simulation tools typically do nor automatically optimize in the way some deterministic (signal
timing) tools do. Optimization is achieved by iteratively applying and modifying parameters (that is, tweaking) until
the objcctive function is optiQ:llzed.
..
2.2.2 AnAJ.ym Sups
1. Define objective: (For derails, sec step 1 of Section 2.1.2.)
2. Define alternatives: The analyst defines the speci.fic altcni.ativcs to be tested in the simulation srudy. Alternatives may be conceptualized, but typically arc not fully defined. For example, the analysis may compare
a modern roundabout to a signalized intersection, but only early design plans arc available for the CWO options. The simulation study can be used to test both alternatives and make a recommendation about which
one should proceed to firu.l design.
In this type of study, the analyst commonly has to make assumptions about the specific configurations fur
the alternatives (for example, diameter of the roundabout; cycle length of the signal). It is critical these assumptions arc docwncnted. In some cases, a sensitivity analysis (as presented in Section 2.1) can be~ to
test variations of certain parameters. In other examples, a design analysis or deterministic operational tool
may be used to prcscrecn altcmativci prior to implementing them in simulation.
3. Define aoalpis scope: Once the alternatives arc defined, a study scope needs to be defined that matches
the objectives for doing a simulation study and the hypothesized effects of any variation in sensitivity parameters. This is important, since ir has implications for the types of measures to be extracted, the physical
extents of the model and the number of simulation runs required.
4. Detme tole.nlllce: A study evaluating alr~rnatives considers tradeotfs between different scenapos. including
cost, del:a.y and other measures defined in the hypothesis. For a successful simi;Liation study, it is im~ru.nc
to define the tolerance that would be considered a notable impact. For example, a study may investigate
freeway operations with and without HOY lanes with the hypothesis: The HOY configuration will reduce
travel time for all travelers. The effect size of interest then would be defined as the reduction in travel time
that would be considered a noteworthy impact. The specilied effect size has direct implications on sample
size calculations co determine the required number of simulation runs, as discussed in Stion 4.0.
5. Set up base model; (For derails, see step 2 of Section 2.1.2.)
6. Calibrate and validate base model: (For details, see step 3 of Section 2.1.2.)
7. Code alternative models: Once the calibrated base model is in place, the analyst codes the alternatives.
While data may not be available to fully calibrate and validate all al~mative models, reasonableness checks
will need to be performed on all scenarios by visual inspections of the results and animarions, or by comparing the performance to other sites and sources in the literarure.
When coding alternative models, it is important to consistendy define any data collecrion dements. Ideally,
the data collection dements used in the base model will .remain unchanged in the scenarios. However, in
some cases differences in model geometry may result in slight differences in ddinitions. For aample, intersection approach delay may be measured from the stop bar at a signal, but from the yield line at a roundabout. Further, if either configuration is modeled with a free-How right-tum movement, neither a stop nor a
yield line exists for that movement. It i.s of critical importance the user understands how a simulation model
defines "performance measures~ and how that definicion may affect alternatives. The 2010 HCM offers a
detailed discussion of simulation performance measures in comparison to deterministic methods.
8. Check reasonablenes.s: (For details, see step 5 ofSection 2.1.2.)
9. Perform multiple runs: (For details, see step 6 of Section 2.1.2.)
10. Analyze results: (For details, see step 7 ofSection 2.1.2.)
Driver behavior and resporue to these messages have a big effect on operacioru in areas where an alternate route is
available. In some cases, the alternate route may be a coUed section of the same facility, such as a HOT lane. In these
applications, simulation may be used to predict driver response to travder informacion and dynamic pricing schemes.
Since the focus of chis type of srudy is on the broad-level nerwork impact, it is generally not necessary co conduce amicroscopic simulation study. While many microsimulation tools have dynamic traffic assignment fearures, mesoscopic
simulation has more computational efficiency.
In modeling driver behavior, it is important to distinguish between a one-time solution and steady-state optimization.
In the first case, drivers react to an incident on the network based on predefined parameters. These include access to
traveler informacion, responsiveness to the informacion and a measure of driver patience and value of time. The simulation estimates the behavior of all drivers based on stochastic distributions of these parameters.
In a steady-state optimization, the simulation accounts for a long-term karning ejfict in individual drivers. lnitially,
drivers still respond co predefined parameters. However, they chen further adjust their behavior based on their driving
(delay) experience on that initial trip and all subsequent trips. Through repeated simulation, a steady-state solution is
obtained that presumably optimizes the overall network performance. For example, chis cype of study was conducted
to evaluate strategies for a downtown event in Minneapolis, MN, USA (Kwon, 2005). Exhibit 11-6 shows the mesoscopic network used to eval_uate evacuation from a sellout crowd at a downtown uena.
210 MANUAL OF TRANSPORTAnON ENGINEERING STUDIES, 2ND EDITION
..
.:-~
~~
II
Source: Kwon, E. and S. Pitt. Ev2luation of Emergency Evacuation Strategies for Downtown Event Traffic Using a Dynamic
Nerwork Model: From Transportation Rmarr:h &corrfjournal ofthe Tramportation Rcearch BotzTd. No. 1922, Figure I, p.
150. Copyright, National Academy of Sciences, Wa.shington, DC, 2005. Reproduced with ~rmission of the Transportation
Resea.rth Board.
--
2. Focmalate incidents and traffi<: management strategies: As part of the srudy scope, the analyst defines
the specific features that drivers react co, ranging from a shore-term incident in one lane on a freeway to a.
full-scale and long-term road-closure. Additionally, traffic management suaregies that affecr rou ting beh2.~
ior are implc;mented. These could take the form of variable message signs, or pretrip traveler information
alerting driv~rs of congested routes. The analyst then defines the spatial and temporal extentS of the event
and identifieS potential diversion routes. These determine the extents of the base model, so it is importan c
to select them carefully. When evalu.atin.g the effects of a work zone, the analyst needs to selecr which
construction stage of the work zone is of interest. Over the life of a work zone, conscruction activiciCli = Y'
include multiple traffic configurations, among them narrow lanes, speed reductions, lane closures and
crossovers. Modeling all potential configurations is time-<:onsuming and may not be necessary depending on the analysis question. The focus of the srudy is on the effect of bottlenecks and incidents on driver
behavior. The objective is to let the simulation model predict changes in behavior using dynamic craffic
assignment (DTA) algorithms and user-defined informacion sources. DTA refers to a class of route choice
algorithms, in which the simulation tool dynmnica/Jy assigns traffic (individual drivers) to d.iJ!erem routCli
in response to congestion on the simulated network. It is important to note the same congestion effects
can also be analyz.ed using the approach discussed in Section 2.2, if the analyst is willing to make assumptions about how drivers may react.
Simulation Studies 21f'
3. Identify coding approach for behavior: The specifics of coding driver behavior differ by software tool,
but different strategies arc typically available. Many simulation tools allow the analyst to define 1tatic :md
dynamic routes and may have specialized features to modd Mls :md road closures. Static routes arc fixed
path assignmentS between an origin :md a destination. Dynamic routes are flexible, and allow a simulated
driver to select a route dynamically as a function of uaffic conditions. In a driver behavior analysis, traffic volumes arc typically entered through an origin-destination (0-D) matrix that leaves the actual path
choice Acx.ible. The reader should refer to the specific software manual for details. Other sources in the
literature give more discussion on driver behavior with respect to routing (Zh:mg, 2008).
5. Calibrate and vali.UU base modd: (For details, see seep 3 of Section 2.1.2.)
6. Perform one-time solution: As discu.ssc:d above, a one-time solution modds driver ~havior in response
to the ru1es specified by the user. These may include the percentage of drivers that respond to ITS treat
ments, the penetration rate of on-board navigation systems, or the willingness of drivers to detour from
their preferred routes. A one-rime solution may still include multiple simulation runs (see step 9), but all
using the same base input for behavioral parameters. In ocher words, a one-time solution is used to obtain
average driver behavior in response to the same (shon-term) incident.
7. Perfonn .repeated iterations for steady-state solution: Depending on the application, the' congestion effect
in question may cx.ist for an extended period of time. It is therefore expected drivers will alter their behavior
over time. Starting with an initial set of behaviors (step 6) drivers will continually test different routes until
they find one that optimizes their trips. The overall system (all drivers on all links) will eventually r~ a
steady-sate solution where all drivers have settled on a new preferred rou~. AgW, this can be done multiple
times to obtain average behavior. The imponant distinction from step 6 is that here drivers rest and learn new
behavior over time. The steady-state solution may therefore differ from the initial behavior set identified by
the user. The iteration process toward a steady-state system solution is computationally inrcnsc and not all
software tools may be capable of performing this step. This step is only necessary to evaluate new and (semi)
permanent sources of congestion such as a work zone or road and bridge clo$ures.
8. Check reasonableness: (For details, sec step 5 of Section 2.1.2.)
9. Perfonn multiple runs: (For details, see step 6 of Section 2.1 .2.)
10. Analyze .results: (For details, see step 7 ofSection 2.1.2.)
Source: Brown. C., W. White, C. vanS~ and J. Benson. Dcvdopmcnt of a Strategic Hurricane Evacua.tion-Dyna.mic Traffic
Assignment Modd for the Howton, Texa,s, Region: From Trruupttrtarion Res~arch Record: journal oftM Transporralion &search
Boarrl, No. 2137, Figure 4, p. 52. Copyrighc;National Academy of Sciences, Washington, DC, 2009. Reproduced with
peim_ission of the Transportation Research Board.
213
::.;;..:t.~&~'-~~~ ..~-
~~~~J!~~~!#i-t'"~~~~ --
-
~L~-- -"-~.
,::I
-
I I I _.. -1
80 .0
.
~
TO.O
,'
60.0
..
<40 0
0..
'
/
lJ
,'
,
:
,
.
/
30 .0
/
20 .0
,'' ,""
,............... ~
10.0
.,,._,.
_,.
+-/l/l/1
,/
'
.
/
at eo .o
'
'
,""'
"
.,..--"'
"
~/
,/
,-"
,"'
LOng "-apon. .
Medum R oaponea
----
------
~..,
-1
------- --
0
Legnd
,,.,.....
I_?~
0.0
:s
"'
11
e'
Rapid Reapon7
Source: Apalache~ Bay &gicn H~trricaN Ew=atWn StuJr-Thnical Da/4 Rrport. Homeland Seau-ity Digital Library. www.
hsdl.org. Figure 4-1, 2007.
3. Identify evacuation scenari05: The analyst needs to identifY the evacuation scenarios to be modeled. For
c:xample, many hurricane evacuation plans in the United States use freeway lane-reversal plans, where borh
travel directions are used to evacuate. Research on simulation studies of evacuation behavior determined
the critical bottlenecks of an evacuation corridor cypically lie at its starting and end points (Tagliaferri,
2006). The analyst needs to have a dear understanding of the geometric configuration and the type of uaffic conuol at the crossover, and may need to test different alternatives.
4. Set up base model: (For details, see step 2 of Section 2.1.2.)
S. Calibrate and validate base model: (For details, see step 3 of Section 2.1.2.)
6. Identify approach for driver behavior: Depending on rhe cype of analysis, driver behavior may be an
input in the simulation from external forecast methods, or it may be derived from within the simulation
through merhods discussed in Section 2.3. The analyst also needs to consider that drivel'$ will likely behave
differently in an evacuation and that driver behavioral algorithms may need to be adjusted.
7. Code evacuation/emergency scena.ri05: With assumptions about the evacuation configuration and driver
behavior in place, rhe analyst codes the actual evacuation model. Given the sensitive nature of this srudy
and its implications regarding human life, the analyst should be very comfortable in all model assumptions and perform sensitivity analyses on critical parameters. Analysts should also consider an emergency
evacUation may be associated with an increased likelihood for crashes that inuoduce unforeseen sources of
delay. By inoorporating some incident likelihood in the simulation study, a more conservative estimate of
evacuation time may be obtained. For example, analysts could design a sensitivity analysis (see Section 2. 1)
of varying frequency and duration of incidents (broken-down vehicles) on the evacuation routes. Considerations for projected incident-dearing times should consider difficulties for first responders accessing the
incident location on a congested network.
8. Check reasonableness: (For details, see step 5 ofSection 2.1.2.)
9. Perform multiple runs: (For details, see step 6 of Section 2.1.2.)
I 0. Analyze results: (For details, see step 7 of Section 2.1.2.)
214 MANUAL Of TRANSPORTAllON ENGINEERING S1\JDIES. 2ND EDITION
One potential surrogate safecy measure is the "time-to-collision" (TTC), which is rhe (theoretical) rime until rWO
vehicles (or other simulated entities) would have occupied the same space at the same time without evasive action by
one or more of the entities. In simulation practice, the TIC and other surrogate safecy measures are calculated from
simulation trajecUJry dAta. This is a detailed record of vehicle position and speed for each simulation time step, which
can be post-processed to extract the safecy measures. One example of this post-processing approach is the Surrogate
Safety Assessment Methodology (SSAM), which was developed by FHWA (FHWA 2008, Gettl)'lan 2008).
Using.the detailed vehicle rra:jecrory data output, SSAM or other post-processing applications develop a vector representation of vehicle movements. A confliCt is registered if the veCtors of two vehicles (nearly) overlap, depending
on user-definable safety thresholds. At the time of this writing, the process of surrogate safery analysis from simulation as well as the SSAM application are new and need further validation. However, the approach generally suggests
simulation has the ahil.icy to be a proaCtive safecy assessment tool. It is presented here as an approach to simulation
safety analysis under the assumption that the analyst has properly calibrated simulation inputs and safery performance
thresholds. Exhibit 11-9 shows a screenshor from the SSAM user guide showing a conflict analysis at a fourlegged
signalized intersection.
Before setting up a safecy analysis i~ simulation it is vital to recognize that .the validicy of the results is a direct funcrion of die quality of the input. While this is true for all simulation studies, it is especially important here since rhe
resultS are very difficult to validate. By de.6nition, conJ\iets occur at intecsections and merge poinu. These locations
are controlled by software-internal algoriduns, which in some cases are designed to prevent noncomp~ancc (and chus
conRicu). For example, while real-world signalized intersections have an elevated risk of high-speed :mgle collisions
Source: FHWA. Surroga~ Sizftty Assmmmt Motkl (SSA.i\1). TechBriefNo. FHWAHIIT-08-049, 2008.
Simulation Studies 21!$
due ro red-lighr running. a simulated signal typically results in perfect driver compliance. It is theorerially possible
to also code some noncompliant drivers, bur then the safety performance of the intersection becomes a function of
analysis assumptions and not necessarily of the true safety of the intersection. The safety outputs are limited by the underlying model assumptions, and behavior calibration and validation are essential to assure the validity of the rcsulcs.
rninimwn TIC:
minimum post-encroachment;
initial deceleration rate;
maximum deceleration rate;
maximum speed;
maximum speed differential;
classification as lan~ge, rear~d. or path-crossing evenr type; and
vehicle velocity change h.ad the event proceeded to a crash
{Source: http:// amc.gov/safety/ pubs/08049/index.htm).
The above measures are used to specify user-definable dutsholds ror vehicle conB.ica &om tbc: simulation output.
The partia.dar measures used vary by application and the sdeaion should be done according to SSAM documentation. The reader should carefully review lirerarure on d.ilkrcnt measures, thresholds and calibration and validation ofsimulation-based safety analyses, which were largdy unavailable at the time of publication.
6. Ron sample and post-process resulu: With a completed and calibrated base model and a conflict analysis Strategy, it is recommended a sample simulation with trajectory output is performed and the rC$ulCS are
post-processed to obtain safety statistics. & indicated, a trajectory-output simulation run can be time.consuming and the analyst should thus make sure the results are reasonable.
?6 Environmental Studies
~.6.1 Struly Overview
Simulation moc:lds have application co environmenal sruclics. R=rch has linked vehicle emissions to cravd patterns and
lW shown emissions an be described as a function of vdllcle sped and acceleration patterns for different vdlldc types (Frey,
2002). The input variables these emissions models need arc avaiWJle from microscopic simulation: vdllclc ~ speed, acceleration and displacem.cnt over time. Many simulation tools estimate emissions (C01 , NO and others) as a d.iJ:ecc output.
The reader should arcfiilly review d!e software documentation to understand the empirical basis for the emissions model
and its limitations. Often, dc.&ult emissions models may be based on limited or outdated vehicle fleet samples, or may have
been developed based on fuel and emissions scanda.rds from other countries and arc dicrefore not univcrsalJy applicable.
When trajeaory data are avaiWJle from simulation, other emissions models can be applied in a post-proc=ing seep wing
spreadsheets. Given increasing focus on emissions nxluaions and the requ.irements lOr quantifying gnxnhousc gas emissions,
simulation an be used to evaluate emissions benefits ofdifferent traffic engineuing strategies (lOr cample, modal shift, signal
progression, roundabouts). Research has dcmonscratod (Frey. 2002) emissions arc not evenly distribuccd over time or space,
but are conccnttaccd during peak periods and areas of rapid 3.f(deration. A:; traffic demand and operations change througho!lc
the day. so will-emissions. Exhibit 11-10 shows an cample df an environmental srudy perfonnc4 in simulation.
, ,..,-'
;
_,..---~
..
_.,.. _., ,.
--~
Source: Sartin, B.; S. Mudigonda and I<. Ozbay. Impact of Electronic Toll Collection on Air Pollution Levels. Estimation
Using Microscopic Simulation Model of Largc.Scak Transportation NetWOrk: From Tnznsporttaion &st4n:h Rt=J: j~~Urnai
oftht Tnznsportatio11 &se4n:h &tml, No. 2011, Figure I, p. 70 and Ftgure 2, p. 74. Copyright, National .Academy of Sciences,
Washington, DC, 2007. Reproduced with permission of the Transportation Research Board.
~impl~tif'\n <:tt~~4iac
')1 "t
In a simulation study the analyst must clearly define the problem scope and collect any necessary field data required
for coding the model. In the absence of field data, lheoretical traffic Bow relationships or expert judgment may be
used. It is 'of critical importance all assumptions and data sources are properly documented and reflected in the
projecr report.
Next, an initial base model is developed, representing a working simulation model that has yet to be comp~d to
actual data. Following base model development, the essential calibration and validation step is performed, in which
simulation algorithms and inputs are calibrated until simulation outputs are validated. Only if the results from this
effort demonstrate an acceptable match between the simulation and field data or theory should the simulation srudy
proceed. The calibration and validation effort should be repeated and model (input) parameters adjusted until a
match is obtained.
With a fully calib rated and validated simulation model, the actual simulation study can proceed to meet any
of the objectives discussed in Section 2.0. The model development process is generally consisten t for all types
of simulation models, including chose ac the microscopic, mesoscopic and macroscopic level. T he difference
'between these types of models lies only in the level of derail of the model and calibration/validation parameters.
Q3
Q 27
azs+----------------1._----------------
1: r
Q 21
QU
I
Q04
1400, ' :1 , I,
~
111
Qll
a a
at l
QOS
Q2l.
Q()1
I, I, I ,1,1,1, 4! 1 ,ooo,
E
s.,...._*
;lf~~ir9d,Sp~~ Oh\;iburlon
.. .
'", ~
,.,
~ I 45 1Gm~ObtrbJion
No.:E3J
:?1
1:-,.-~....- j
.::
'CJ~.
.-.;,.--- .
X~
;.,;
: ..
. -
:. .~.
';' i. :v;:.,
~~}<_~
.,
- :. ::. :~:
~~.: . ~. :
..:..
:"'::
- ~.~~~i ...:
,., ,
;-.
,/'!I
.:lkldo
: ..: ..l .. .. ....
,;
4 '
,(
~....
'~.
'
'
ior for specific poin!S, segments, nodes, or the network as a whole. These demen!S are necessary to obtain customized
output from simulation. The definition of data collection demen!S is conceptually similar to designing a volwne,
speed, travd time, or other field study with automatic data collection equipment.
In the final step, the simulation model is executed to obtain resui!S. Typical simulation resul!S include visual observation of traffic that can be recorded as animation for distribution to clients and public stakeholders. The analyst .also
obtains data from data collection elements in caw form or aggregated in a selected time interval. Depending on the
type of model, aggregation and analysis of results in tables and charts may occur auromuically, or may require some
manual post-processing in spreadsheet sofrware.
Through this discussion, it is evident any simulation model has a lot of Ba.ibiliry for custom-user input, as network
geomerry, volumes, routes, speeds and traffic control can be calibrated and compared to field data. In some cases,
the analyst can choose to rely on model defaults for increased coding efficiency, bur great care should be taken when
doing so. The analyst needs to fully understand the implications of different default parameters and should pay close
acrentiori to the validation of model outputs when defaul!S are used. In general, simulation defaults are designed by
the sofrware developer to provide a reasonable coding baseline. But since many tools are used internationally and for
a variery ofspecial applications, default values ofrentimes will not satisfy local or regional traffic behavior. The discussion below gives more detail regarding when defaul!S are acceptable and when additional calibration is necessary.
3.1.3 Systnn BoutU!my
A common challenge in designing a simulation study is to define the system boundary. The system boundary decision
affects both the spatial extent of segments and nodes, and the level of detail included in the analysis. Both of these
are important considerations; they directly affect the validity of the simulation results as well as the computational efficiency of the model. Larger and more detailed models increase the computatio11al requiremen!S. On the other hand,
a system that is too small may not caprure all facets of the analysis.
Asimulation system should be large enough ro contain all queues within its boundaries ar all times. Ifat any time a queue spills
beyond the spacial cxrents of the network. the analysis results are no longer coma, since only vehicles that enter the nerwork are
included in delay calculations. All queues should also be ront3lned within the remporal.limits of the analysis. In other words,
all queuing etfects and overcongested operarions should be rontained within the boundaries ofthe time-space analysis domain.
The network also needs tO be large enough 10 assure traffic behaves realislically at points ofinterest. For example, ifoperations at
a two-way stop-controlled inrerseaion are afkaed by vehide platoons from an upsa= signal, that signal needs to be included
in the analysis (or the platoon effect represented through customized vehicle inputs).
The level ofdetail needed for the
simulation network is a function
of the relative amount of rraffic contained on minor links.
Chapter 4 recommended areawide counts can ignore minor
links on the network. provided
the total flow on all these minot
links is less than 5 percent of
the overall traffic demand. Following this same guideline, the
modeled Simulation network
should contaln enough detail tO
include 95 percent of demand
traffic. The remaining 5 percent
of ttaffic from minor generatOrs
is not ignored, but allocated to
bigger links. This guidance is
approximate. In face, it can be
very challenging to get link-specific volwne data from simulation, since volume inputS are
Source: Tagliaferri. TFHRC Newsletter on NGSIM. Pages 49-50, figures 3-ll and 3-12;
Tagliaferri, A. W. 1-40 l'..ll1u RnmJ Tl'lijftt MA/ytit. FHWA-NC-2005-14. bnps:/lapps.
dot.state.nc.us/dot!directory/authenticared/UnitPage.aspx?ide8781. 2006.
typically 0 -D based. In general, the network should include enough detail to include all links that have a un ique and
' significant dfect on network performance. If a minor link has some sort of unique effect on the overall system it needs
to be included. Examples of a unique effect include significant queuing from a small 4riveway or a minor link that
' cawes spillover at a signalized intersection. & a rule of thumb, all links that are cawes of ~urrlng congestio n need to
be included. in the analysis. Exhibit I L-12 shows a simulation network as represented in two different analysis tools.
3.1.4 SimulAtUm Resolution
All simulation cools .w e an underlying time step to up<hte vehicle position and signal timing, and recalculate driving
rules wing the behavioral algorithms. In some older simulation tools, the minimum time step si:z.e was 1.0 sec, bur
most modern cools allow a simulation resolution as low as 0.1 sec. However, even modern tools typically allow the user
co change the resolution. A higher resolution simulation (lower time seep) will result in more rigorow computations
and therefo.rc may slow down the processing time (depending on the computer wed and the complexity of the simulated system). It may therefore seem appealing to switch the simula.tion to a lower resolution (bigger time seeps) and
thw improve computatiqnal efficiency. This practice is discouraged, since it will change driver behavior and therefore
the results. For example, a simulation resolution of 1.0 sec means that (simulated) drivers can only react (*o pen their
eyes") once every .second, which results in jerky reacc;ion time in a microscopic modd. A 10\Ver resolu.tion may be
acceptable for a macroscopic simulation, where less derail is provided. But as a general guidance, microsimula~jon
models should be executed with the highest possible resolution, which is typically 0.1 sec.
i
i
I
I
.I
In the following sections, common measures arc discussed for the poim, segroem, node and network levels. The variow input data, calibration parameters and output variables have predse definitions in each simulation tool and these
definitions tend to yary from product to product. The HCM attempts to provide general guidance co simulation
developers.to encourage consistency across tools. However, in the absence of national standards, the analyst must be
aware of the precise definicion of each variable and parameter used in the tool chosen for the analysis. Coo.sequencly,
any data that is input to the model mwt be consistent with these definitions. The interpretacion of output variables
should be based on a complete understanding of what the measure is (and is not) quantifying.
3.2.1 Poial DllliZ Co/Jeaion
Similar to 6dd studies dcsaibcd in Cbaptca 4 through 6 of this manual, simulation studies can be performed at a single-point location in the necwork. The ability to define wee-specific data collection points varies between simulation
tools and ic is recognized chat some can only aggregate measures to the segment, node, or necwork levels as disawed
below. Point data collection gathers data on traffic volumes, spot speed estimates, or vehicle delays ac_an approach ro
an intersection, or at a midpoint location along a segment. Simulation models typically allow the analyst to code a
Simulation Studies 221
data collection point that will record these and other data items at the selected location. The user also specifies a level
of aggregation that tells the program the bin size at which data should be stored.
Point data from simulation can be especially useful for model validation, since .field data are usually also collected at a
spot location. In this case, the ~rcgarion interval needs ro match the field data. For delay studies iris important chat
the analyst uses the same definitions, since results will be different for stopped delay, queue delay, or approach control
delay. For definitions of these terms, refer to Chapter 6 or the Chapter 2 Glossary.
3.2.2 Segment DatA Collection
Some traffic performance measures require that data are collected over a segment of roadway. These measures include
travel rimes, delays measured over a distance or route and queue lengths on a freeway or on approaches to an intersection. While the field studies co obtain these measures can be extensive, in simulation the analyst selects the appropriate measure from the menu of available performance measures. As with point clara collection, the analyst needs to
carefully select the se.gmenc length and clara aggregation interval to match the study objective and the corresponding
field study if applicable.
In a freeway application, simulation segment data collection can estimate me traffic SMS, flow and de.nsiry along a freeway
segment. Ar. a shore analysis interval (for
example, 5 min.), a simulation model can
quickly generate many data points to plot
a speed-flow-density relationship for the
freeway segment. Since segment capacicy ls not an input in simulation, but a
result of traffic operations in the model,
a visual plot of the speed-flow-density is
an important validation parameter. If the
resulting relationship does not march field
obserntions (or traffic How theory) the
analyst can adjust driver behavior parameters in me car-following, lane-changing,
or gap acceptance algoricltrns to calibrate
the speed-Bow-densicy relationship. Exhibit 11-13 shows an example of a speed
Bow curve for a freeway segment from
field measurements. and as modeled in
simulation. The twO clouds" of data
show comparable free-Bow speeds during
uncongested flow periods (less cltan 8,000
vehlb in this case) and a reasonable represenracion of segment capa.cicy (at approximately 11,500 vehlh). Analysis methods
beyond visual inspection are statistically
rigorous and are discussed elsewhere (for
example, Hollander ct al., 2008, or Zhang
er al., 2008 I and ll).
Measurement
:..:.....,::~=....-::...:.J!9.9!:.!!SI~-~~..g.V;~..J~I!~~L-=-.: ;:c::;:;:~:.
:~~~--~----~~~
I .
70
:<:' 60
Q.
..
50
..
"i 40
_:.
i- 30
20
10
0+---~--~----r---~--~----r---~
2000
4QOO
6000
8000
Simulation
- ! .. . . .
r:--:::-::::~ . ~-~.~.,_::-_=:-~
.... ---=...-...:.::..-:~{.......'1.-J. - -
90--------------------~
80
70
..-
J:' 60
~50
.....
40
30
20
10
l . ...... y= ~~~2000
4000
8000
now
[Ve'.'fhJ
I.
6000
If a simulation scudy is performed to compare incersection alternatives (for example, a signal versus roundabout) ~
node-aggregated clara analysis measure is useful. A node-aggregate measure can give, for example, the total control
delay for an intersection under different traffic control strategies, or in a volume sensitivity analysis.
Node-aggregated data are more difficult to compare co field observations, since a field study for a node performallce is
very involved and usually not cosc~ffeccive. It is therefore more common that the opera.cions of individual appr~ches
ace evaluated in the model calibration and validation efforr before resulcs are aggregated to the node level.
3.2.4 Network- Wule Measures
When a simulation study is performed for a larger network it is sometimes necessary to gee an overall network-wide
performance measure under different system configurations. For example, a simulation srudy may evaluate systemwide impactS of ITS strategies and changes in driver routing behavior, or the overall reduction of vehicle emissio~s
for a corridor after installing a coordinated signal system. For the evaluation of these types of large-scale scenarioS. It
is helpful to look at net:wOrk-wide aggregate measures.
The rwo major problems with nerwork-widemeasures.are chat operational derails are los.t with extensive aggcegati~n,
and that any model comparison with field scudy resulcs is virrually impossible. Therefore, while a study may rcqlure
the analyst to report network-wide measures, it is always important to look at the 'operational performance of subelemencs in the system. When reporting network-wide resulcs it is therefore common to also report node-level resultS and
highlight the worst intersection or spot location in the network.
tion to traffic Bow theory relationships (such as the HCM speed-Bow curves) should only be done if field escimll.tes
are impossible to obtain, or would add unreasonable expense and time to the modeling dfon. Calibration ensures t:he
sp~ed rules for driver behavior and other computational algorithms match user expectations based on field d:;~.ta
or resources in the literature. A national guidance document on calibration procedures 'for simulation lists five core
components (Zhang. 2008).
I. project scoping and error checking
2.
~lobal
parameter calibration
In general, global calibration refers to parameters chat apply to the entire modeled system. For example, in some cools
the composition of the vehicle Beet or driver behavior (car-following. ere.) are global parametetli. Local calibratiOn
parameters are those that are manipulated at the individual intersection, segment, or approach level, and can inclucle
speed distributions, signal timing, or gap accepcance pacameters.
Calibration of model parameters is an imponant aspect of any simulation scudy, since reliance on default values is
likely to ignore regional or international differences in driver behavior. The following paragraphs present the m:J. n
simulation inputs that are cypicalJy calibrated in more detail: volumes, speed, gap acceptance, signal control anddriv-e=r
behavior.
uolurms if the count location is downstream of a botrleneck. The input demands $hould therefore always be mea$uted
upmeam of any congC$cion and similarly coded to enter the simulation network on uncongested links. The calibra
cion of input demand volume$ is fundamentally different &om the validation of $C!Vcd traffic volume$ internal to the
network discussed in Section 3.4.1.
3.3.2. sp~~Js
The distribution of vehicle speeds on a segment in the simulation modd is another means of calibrating the model.
The methods in Chapter 5 can provide fidd C$timatC$ of the speed distribution. The analyst calibrate$ speeds by delining target speed distributions in the simulation. Global calibration changes vehicle input speed distribution, while
local calibration changes speed behavior on specific segments in response to sharp turns or steep grades. Some simulation tools are limited in the extent to which they allow speed distributions to be user-defined, while others allow full
flexibility as to the type of distribution used.
Speed calibration should only be used to reflect speed patterns caused by (speed limit) si~ng and geometry, never
congC$tiOn. In other words, analysts Th(IU/.J only C4librau fm-fow speeds on the segment based on corresponding .field
measurements. Speed calibration should never be used to force drivers to slow down in congested operation or to force
botdenecks. Ifa real-world segment shows speed reduction due to high demand the simulation model should predict
this congC$rlon {and the resulting drop of speed) as an output. If necessary, the analyst can calibrate driver behavior
as described below to achieve more or less aggressive behavior (and thus higher or lower capacity). In this case, the
field-measured congested speed becomes a validation parameter as discussed in Section 3.4.2.
3.3.3. SigruJiuti Control
Modern simulation models offer a variety of options for coding signalized intersecQons, ranging from built-in controller logic, to so&ware-41-theloop (SIL) and lwdwarcinthcloop (HIL) emulation of aaual signal controll~.
In SIL modeling, a separate program handles the signal operations.
program
runs paraJid" to the simulation and the two
programs exchange information during every time step. In particular, the simulation
passes signal detector calls to the SIL program. The SIL then interprets the effect of
these calls on signal operations and (if neccs
sary) passes a signal phase change back to the
simulation. The updated signal phasing in
rurn affects vehicle operations in the simulation, which again affects the detector call in
an iterative process.
nus
Befora Calibration
80
70
eo
-
!50
140
J:
a.- CF Panometers
HCM FFS 60 Curve
10
0
0
1 000
2000
3000
Flow Rllte [pc:JMnJ
4000
After Calibration
eo
70
eo
~eo
..40
\.!
~ k>?~" I
0
1000
'
Modlfted CF
HCM FFS 60mph
'
2000
3000
Flow R - [pc::IMnJ
'
4000
: Given the complexity of signalized control, many sowces of error may affect the operations including detector settings, field hardware delays, and coding errors.
Because driver-behavior ~ritluns are the underlying rules of the simulation, they are an important calibration roo!.
If an analyst observes unusulll queuing at a freeway merge area, for example, an adjustment in lane-changing behavior
may be necesury. Similarly, if a freeway segment appears to carry too much traffic before reaching capacity, the carfoUowing algoridun may need to be adjusted to make drivers more conservative. However, the relationship becween
the driver behavior parpnecers and calibration measw;es (for example, capacity) is usually indirect, so some trial and
error is needed.
3.3. 4. I Car-Folklwing
In their most basic form, car-following models predict a mponu (the acceleration or deceleration rate of the driver)
as the function of some stimulus. In the original car-foUowing models, the stimulus was strictly the relative difference of two (foUowing) vehicles' speeds. The stimulus is further multiplied by a smritivity rerm that prescribes more
or less aggressive foUowing behavior. This original class of stimulus-mponst car-foUowing models was later expanded
to what are known as saftty-dirt4nct mod'els. Safety-distance models further refine the following relationship in .~at
the sensitivity of the desire<! safety teem (for instance, the relative speed) varies with distance. For example, a relative
difference in vehicle speeds oflO mph {16 km/h) elicits a more drastic driver response when the vehicles are closely
spaced than. if they are very far apart. Today, many simulation tools apply what is known as a psycho-physical model,
which combines the psychological car-foUowing aspectS (desired speed differential and safety distance) with physical
lirnicarion.s on driver perception and vehicle throttle control.
In terms of calibration, the stimulus-response class of car-following models is relatively easily calibrated, because
the only calibration tcnn is a single sensitivity teem. Safety-distance models cypicaUy have multiple (rwo or three)
sensitivity terms to describe variOIL~ aspects of the interaction. Psycho-physical models arc d.ifficulc co calibrate since
several sensitivity parameters exist. Calibration is fucchcr complicated if the car-following models employ multiple
driving rtgi~for example, to distinguish free driving (away from other traffic), from car-following. and from an
emergency-braking regime. Consequendy, modem simulation tools may offer 10 or more calibration rerms.
The calibr:ttion of car-following models will primarily affect the spacing berwcc!_l vehicles and a good calibration W
get is therefore the capacity (for a freeway segment) or the sacuratiqn flow rate (for a signalized intersection). Exhibit
11-14 shows a freeway car-foUowing calibration example, where p:iramerers were adjusted to march the HCM (TRB,
2000) speed-flow curve for a 60 mph (96 km/h) free-Bow-speed facility. The before calibration" chart shows the
simulation daca appear to over-predict the (theoretical) capacity of2,300 passenger cars per how per lane (pclhlln),
and the mer calibration provides a better visual fit. More rigorous mathematical methods for sirnula.tion calibration
arc discussed in Zhang ct al. (2008 I and 10.
The details of simulation tool-specific car-following calibration are found in software user guides. A good discussion
of car-foUowing models, including a discussion and comparison ofsome commerciaUy available software products, is
given in Olscrom and Tapani (2004).
.
Lane-dwtging algorithms are related to car-foUowing algorithms, in that a vehicle with a high desired speed may wish
to pass a slower-moving preceding vehicle in a voluntary lane change. Lane changes may also be rtquiml if a vehicle
has co shift lane to, for example, turn left at an intersection or nit a freeway, in which cases lane changing depends
also on a simulation route choice algorithm.
In its most basic form, a lane-changing algorithm involves some driver reaction time, and safety paramerers that describe aUowable headway to the preceding vehicle, as weU as the acceptable size of gaps in c.raHic in the ~jacc:nt lane(s).
A lane-<:ha.nge maneuver fu.rther includes some acceleration and deceleration behavior as cars speed up or slow down
::I:
~
Source: Dowliog, R. S. FHWA Tniffic An4lysis Toolbox V.,/ume III: Guilkli= for Applying Trtrj}ic
Mi=simuiAtion Motkling Softwan. FHWA-HRT~4-040. Figure 18, page 117, 2004.
to make the lane change, or to allow another vehicle to change into their lane (coopmztive behavior). The availability
of these paramerers varies by simulation tool and the user guides should be checked for more detailed guidance.
The goal of a lane-change algorithm calibration can, for example, focus on operations at a freeway merge area (or
other bottleneck), and may~ be compared to field dara or theoretical uaflic flow relationships. Exhibit ll-15
shows a dual-objective calibration exercise to get a model ro predict a mean freeway bottleneck speed of35 mph (2
mph) (56 krnlh 3.2 km/h), while processing 2,200 veh/hour/lane (:tlOO velticles) through the bottleneck (Dowling.
2005). The calibration parameters are mean reaction time and mean headway in a lane changing algorithm.
The aample in Exhibit 11-15 illusuaces that, often, multiple calibration goals (in this case, speed and capacity) are competing. For aample, as driver reaction time increases (slower reactions), the observed velticle speed decreases, as does the
observed capacity. The solution rtgion represents the realm where both calibration rargets are wit:hi.n an acceptable region.
3.3.4.3 Gapkceptance
Gap acceptance algorithms in simulation describe operations at stop- and yield-conuolled intersections, including
modern roundabouts, freeway merge sections and the interaction between drivers and pedesuians. Gap acceptance
theory in deterministic models uses the parameters csirical gap and follow-up time; Chapter 6 described methods to
obtain these parameters &om field data. Gap acceptance behavior in simulation typically employs a minimum gap
threshold. The distinction between critical and minimum gap is important, because it makes it difficult to directly
calibrate simulation.gap acceptance behavior in simulation. Furthermore, the follow-up time parameter does not exist
explicitly in microsimulation, since the likelihood of subsequent vehicles accepting the same gap is an implicit function of the car-following algorithm. A more aggressive car-following model results in the equivalent ofshorter followup time, but the calibration cannot rely on a field estimate of that parameter alone. For mesoscopic and macroscopic
simulation the saturation headway can approximate the follow-up time for gap acceptance.
Given the d.ifficulties of calibrating gap acceptance behavior &om field-measured da~a, this particular calibration is genctally
closely linked to validation of other measures. For aample, a frequently used approach to calibrating the gap acceptance
behavior ac the enuy leg of a modem roundabour is to start with an assumed initial ser of gap acceptance parameters. Using
this initial ser, the analyst runs the model and validates the roundabout enuy capacity under a range ofvolumes. In an iterative ptocess, the analyse canthen adjusr the gap acceptance behavior to achieve higher or lower enuy capacities.
226 MANUAl OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
Diffcrcnc simulation tools allow varying degrees of calibrating gap acceptance algorithms. Some allow full Aexibiliry to
modify gap acceptance parameters for different locations and different driver types. Ochers constrain gap acceptance
to a global set of parameters, rhat may limir che analyst to differentiating parameters by the type of maneuvec (such as
'. right turn, left rurn, or through movement), and not a specific location.
Volumes in simulation can be an input or an output. Section 3.3.1 discussed the calibration of demand volumes as
simulation inputs. Volumes internal to the simulated network are a result of many simulation parameters and algorithms and should therefore be treated as outputs. They can only be observed and measured after the analyst elects
to run the simulation. Network-internal volumes can be validated against field measurements of prevailing nerwork
conditions, which may include congested Row periods (discharge from a signal, or Bows downstream of a freeway
bottleneck). The validation of-network-internal volumes is also important in relation to driver behavior srudies (Section 2.3) that apply dynamic traffic assignment.
3.4.2 Operating Speeds
Similar ro volumes, speeds can be a calibration or validation parameter. Section 3.3.2 discussed the calibration of uncongested, free-flow speeds from field data. Opetating speeds are outputs of simulation after the analyst elects w run
the.simulation and are a function of many parameters and algorithms. Operating speeds are an important validation
pararneter for congested Bow conditions on freeways and arterial streets. The measurement and aggregation of speeds
should mirror whatever field srudy was performed to estimate the benchmark speed daca. In this regard, speed validation can be performed using the TMS at a point location (see Chapter 5) or the SMS measured over a segment (see
Chapter 9). Similarly, speed validation may be performed across all lanes on a segment, or may focus on a particular
element, such as speeds through a channelized turn lane at a signalized intersection. Sample size calculations for spec: d
comparisons and rests of significance are consistent with those discussed in Chapter 5.
3.4.3 Point Delays
A common validation parameter is a delay measurement at a point location. The analyst may want to validate the
delay at an intersection or at a freeway ramp. It is important to note simulation models almost always define delays
as the difference between desired and acrual travel time. In comparing delays to field measurements the analyst needs
to be mindful whether the delay includes geometric speed constraints or not. For a.ample, a single vehicle tnvelin.g
through a modern roundabout experiences geometric delay because the driver is forced ro slow down to navigate the
circle. However, in simulation the geometric delay is typically included in the estiiiillte of desired or free-How uav-el
time and the delay is ,therefore zero. This distinction determines whether the field data used for model val.idation is t):l.e
rot:al delay (including geometric delay) or the control delay only. Details on performing an intersection or point dcl:;;J.Y
srudy are given in Chapter 6. The issue of delay dcfin.jtion is also common when comparing simulation to outp~ cs
from deterministic anaJysis models. Some simulation tools may not allow user-defined point data collection elemen a
to estimate delays, but may be constrained to measures reported to the link or node level.
3.4.4 Segment Travel Tnnes
Travel time over a segment or through a facility is a good validation measure because it takes into account a 'IUieCY
of factors inBuencing the model operations (for eumple, signals, traffic congestion and speeds). Travel rimes are
relatively easy to obtain ftom the field using in-vehicle stopwatch measurements or GPS travel time logs as describe d
in Chapter 9: GPS data have the added advantage of allowing the analyst to not only validate travel time,.but a.ls .co
compare average running speed, delay times and accderarionldeceleration behavior between the model and realiry.
Simulation Studies 22?
simulation inputs and v.didation ofsimulation outputs in relation to field data. Exhibit 11-16 relates the ~bration
and validation parameters to common data sources for each measure. For field data, the cable gives the chapter references for thi.s manual. Other common data sources arc in the literature, including theoretical traffic Bow relationships
and lessons learned from prior simulation analyses for similar problems statements and using the same analysis cool.
Validation parameters ofsimulation analyses are also often compared to estimates obtained from other analysis tools,
including well-established deterministic models.
Point Deb.ys
Queue Lengths
2.2a
60
-~1
-pteeea-002
-pnx:ea-003
-proc:ooss-004
-proce$$-005
lo-008
50
40
sE
30
20
10
<
o~~TMTM~TM~~~~~~~~~~~~TM~~~~
ro'!<...~ro't~ro-Y~ro~~r;;it?~r;;1'~"''"'~ro~~.,..Pro~.,.~
...r;;<t ro~:\<$'.~:\~~-~~':-"'-~
.-!':\'t'~:\1'':\.Y~.~
~ ~ ~ ~ ~ ~ '
~ ~ '
Simulation T1me
Source: Dowling. R. S. FHWA Tr4JU Ana/ytis Toolhux ~fume lll: Guitklines for Appfyjng
Traffic Mimlsimularion Modeling Software. FHWA-HRT-04-040. F~gUCC 14, page 106, 200i.
The effeq size of inceresr: How much of a difference is considered noteworthy? Consider a study that
compares different alternatives for traffic contwl on an arterial street. The intuitive measure of e.ffectiveness
is the average travel time on the faciliry. The question of the effect size of interest then depends on what
difference in travel time is considered a noteworthy difference. For example, a 5-mile (8-k.m) segment
posted at 50 mph (80 km/h), has a theoretical free-flow travel time of 6 min. (360 sec), but due to signals
and congestion the ex.isting average travel time is approximuely I0 min. For this example, a noteworthy
effea size may be a travel time reduction of 20 percent to 8 min., or a 2-min. difference.
The variabilitY berw:cn runs: The stochastic nature of simulation means che performance measures of
interest vary &om run to run with differenr random number seeds. The improved average travel time of
8 min. in the example above may therefore vary between 7 and 9 min. across different runs. The measure
of variability is cbe standard deviation of the travel time estimate. In principle, if the travel time is .F.Urly
constant across runs, a lower sample siz.e is needed. However, if random effects in the simulation (signals,
parking) result in large differences across runs, more iterations are needed.
The significance !eye!: As with all statistical testS, the significance level is specified by the user depending
on how much confidence he/she wanes to have in cbe estimate. For example, the analyst may want to be 95
percent sure that the travel time difference is at least 2 min.
Given cbe above factors, it is impossible to give a general requirement for the number of simulation runs required. The
effect size of interest will vary depending on the goal of cbe analysis. The variability between runs ean under some circumstances be estimated, but more often requires the analyst to perform some test runs. And finally, the significance
level may also vary depending on the problem at band.
The rdationsbip between the factors above is given mathematically by equation I 1-1 .
X > tXSp
fi
Equation 11-1
where
X -= effect size of interest, defined as the absolute estimated difference in means of a variable between cwo alternatives
n ., the number of model runs per alternative tested
t .. t-statistic for a confidence levd of [(1-alpha)/2] and (rn-2) degrees of freedom (froin Exhibit C-12 in Appendix C)
the pooled standard deviation of the variable csti.mace from multipk modd runs of the two sa:narios to be compared
"
The values oft for confidence levels are determined &om cables of the e.-distribution rbar areincluded in most statistical tc:nbooks (for example, Washington et al., 2003) and repeated in Appendix C (Exhibit C12). For an estimate
with 95 percent confidence, and a cwo-sidc:d significance: test, the column for alpha- 0.025 should be used.
The standard deviation, s , is the pooled standard deviation of all runs for both scenarios in the comparison and is
calculated from Equation'll-2, assuming an equal number of iterations in each of cwo scenarios x andy.
Sp
Jsi ;s~
Equation 11-2
where
s,
the pooled standard deviation of the variable estimate from multiple model runs of the two scenarios to be
compared
n>2x~
x'
Equation ll -3
> 2 X (2.57l)'(t.S)'
{LO)'
TI1e value of t>-2.571 was obtained fi;()m Exhibit C-12 assuming a sample size of five runs per scenario. Since the escimace
of n changed. to 30 runs per scenario (after rounding up), the new estimate oft is 2.042 for a 95 percent confideflce
lcvd. Recomputing n with the revised tyic!ds :l.rt estimate ofl8.8 (or 19 runs), which in cum changes the estimate oft co
2.093. In the third iteration of Equation 11-3, the estimate of runs is now 19.7 (or 20 runs), at which time the solution
converges (t no longer changes significandy enough to impact the result of n). Consequendy, the analyst needs to perform an addirionall5 runs for each scenario to add to the original five runs each. After perfonning all40 simulation rufl.S
(20 per scenario) the analyst should recompute rhe pooled standard deviation to ensure th~t the estimate didn't change
Ifthe analyst strictly foUows the sample size procedure outlined above, the consequence may be a large number of ruf'lS
are required to fully satisfy the criteria. Multicore computers and batch run processing can reduce the time requir<!:d
to perform multiple runs. However, making multiple runs for large models may still be a time-conswning process
Therefore, the model inputs and data collection configuration must be finalized prior to starting the compur:atiofl.:a.l
iterations. To facilitate the analysis, it is helpful to establish a pose-processing approach to quicldy analyze the outp~t:s
of many different runs and calculate average resulrs. Many conunerc;ial softw:ue p(ogratru have some form of an an3-lyzer module that will facilitate this process.
:
hi&
! ! I
~
35
30
25
20
15
10
5
0
Scenario1
Scenario2
Scenario3
Scenario4
ScenarioS
4.4 Documentation
Bcause of the inherent variability of stochastic simulation results and due to !,he ability to "tweak simulation rc
suits duough calibration efforts, prop~f dcxumentation of$imularion analyses is csscncial. It is fair to say simulation
outputs are highly sensitive to the sdected inputs and calibration parameters in behavioral algorithms, and almost
any conceivable result can be obWr\ed by adjusting the parameters, even if the a.djustmcnts are not sensible! Proper
documentation therefore ensures the integrity and validity .o f the submitted analysis.
The ducc key components to proper modd documentation are:
1. a detailed listing of input variables (such as geomeety, volumes, signal timing) in the form of screen shots,
input tables, or other format;
2. a record of behavioral algorithms and any modifications done for purpose of calibration (for example, car
following, lane changing, gap acceptance); and
3. a write-up of modd validation results, demonstrating the modded base scenario matches field observations
or theoretical performance metrics.
With these items in place, an cxpeticnced user and reviewer of simulation work can attest to the validity of the analysis. This requires that reviewers at the funding agency arc familiar with simulation practice and have experience with
the particular modd in question. In the absence of a qualified reviewer, a peer-Teview process with a third-party con
:performance of traffic Aow and to compare different scenarios. Therefore, visual animation should only be used as a
~lupplementary form of reporting simulation output.
i
If animation is used, it is common to record a predefined "Hy-duough" course duough the network to highlight key
points. Some guiddines for generating such video dips arc:
Start with a network overview to allow the audience to become familiar with the scope of the model.
Slowly room lnro any sub~ections that are to be shown in more detail. A lay audience is less used to
watching animation and may not be able to follow too-Ease transitions.
Rest on any desired detail long enough to show operations. Consider whatever narrative may be desired by
the presenter.
Assure an overall dip length that is :tppropriate to show the intended results, without being overly long. For
public podium presentations, dips longer than approxim:ttdy 1 min. are discouraged as,too long to hold .
the attention of the audience.
The recommend:ttion for a dip length of approximately 1 min. is given in light of temporal constraints for many
presentations. Clearly, longer dips ue feasible if time permits. Especially in a walk-in type public presencuion of a
project, longer dips arc feasible md in lice desirable to give the audience more time to take in the information presented. Longer dips also reduce the :tbilicy to mask or hide problems that may occur at other times in a simulation.
The general recommend:ttion is to tailor dip length to show what needs to be shown to fully communicate the craflic
operational issues under consideration.
Exhibit 11-19 shows screenshots
from a sample simulation By-thrOugh
for .an urban roundabout and screet
redevelopment project. Sdected 3D
bull~ were created to hdp the
audience recognize fUniliar features.
The arrows were added to trace a university bus duough the network.
S.OSUMMARY
Thil chapter presented an overview
of studies performed using cra.ffic
simulation tools and was intended
as guidance for users of simulation
tools, as well ~ those charged with
reviewing results of simulation studies. There is much additional detail
on simulation studies, and espe. dally the underlying simulation algorithms, that go bey~nd the scope
of this chapter and the reader is encou.raged to refer to other material in
the licerarwe, such as FHWA Traf
fie Analysis TotJibox (http://ops.fhwa.
dot.gov/ trafficanalysistools/index.
hem), for funher discussion.
. . . . . ...
6.0 REFERENCES
Alsnih, R. "Review of Procedures Associated with Devising Emergency Evacuation Plans. Tmnrportation Rmarch Record:
journal ofthe Transportation RescardJ Buard 1865 (2004): 89-97.
Bartin, B., et 2!. Estimation of the llhpacc of Electronic Toll Colleccion on Air Poilu cion Levels Using Microscopic Simulation
Modd of a Large-Scale Transportation Nerwork." Tmnsportlltibn &March &cord: journal oftM Transportati4n &uarrh &ard
2011 (2007): 68-77.
Bonneson, J. ct al. NCHRP Report 457: Enginming Study Guitk for Evaluating lmmecti4n lmproiJCTimtJ. Washington, DC:
Transporcation Research Board, 200 I.
Brown, C. ec a!. "Devdopmenc of a Stttcegic Hurricane Evacuation Dynamic: Traffic Assignment Model for the Houston
Region." Transportation R.escacch Board 88th Annual Meeting. Washington. DC: Tcansportation Research Board, 2009.
Courage, K. NCHRP Project 03-85: Guwna on the Uu ofAlumativt Traffic Ana(rsis Tooi.J in Highway Capacity Amtiym.
Washington, DC: Transportation Research Board, 2008.
Dowling, R. S. Tmffic Analysis Toolbox Volume Ill: Guitklines for Applying Traffic MicroJimulation Motk/ing Softwarr.
Washington, DC: Federal Highway Adminisuation, 2004.
Federal Highway Administration. NGS!M Task E.l: Corr Algorithm AJrwmmt. Publication No. #FHWA-HOP-06-0009.
Washington, DC: U.S. Deparunc:nt ofTransporcation, FHWA. 2004.
Federal Highway Administration. SurrogauSafttyAJsmmmt Motkl (SSAM). Tc:chBriefNo. FHWA-HRT-08-049. Washington ,
'
DC: FHWA, 2008.
Frey, H. C., N. M. Rouphail, A. Unal and J. D. Colyar. Emimon Redudions Through &na Traffic Managanmr: An Empirical
Evaluati4n Bastd Upon On-Road Mt4Sumnmts. FHWY/NC/2002-00 1. Raleigh, NC: Department of Civil Engineering, North
Carolina Scare University for North Carolina Department ofTransporcation, 2002.
Gerunan, D. P. Surrogau Sttfay Assmrot Motkl and Vaiid4sion: Final &port. Washington, DC: Federal Highway Administration,
2008.
Hidas, P. and P. Wagner. "Review of Daca Collection Methods for Microscopic Traffic Simulation. IOtb World Con~ on
Transport R.esearch. 4'on, France: World Conference on Transport Research Sociery, 2004.
Hollander, Y. and R.. Liu. "The Principles of Calibrating Traffic Microsimulation Models." TranJportlltion: Plannng, Policy,
Rntarch, Pmdict 35, No.3. springerlink.com/contem/40387v6350097010/.
Holm, Peter, Danid Tomich, Jaimie Sloboden and Cheryl Lowrance. Traffic AsuJysis Toolbox Volume IV: Guitklin6 for Applying
CORSJM Microsimultrtion MotklingSoftwarr. Federal Highway Adm.ini.stration. Publication No. FHWA-HOP-07-079.
Washington, DC, 2007.
Jcannone, K. C. Trttffic Analyr# "[qq/box Volume JJ; Deciwm Suppi!Tt MahotkJ!JJgy for Sekcting Traffic Anlllysis Too/;. Washington,
DC: Federal Highway Administration, 2004.
1Gm, T. E. "Operational and Safety Performance of a Nontraditional Intersection Design: The Supersueet. Transportation
Research Board 86th Annual Meeting. Washington, DC: Transportation Research Board, 2007.
Kictdson, W. Traffic Analysis Toolbox Volume V: Traffic Analyris Tooi.J Cast Snuiks: &ntfitt andApplicarioru. Washington, DC:
Federal H ighw:i.y Administration, 2004.
Kwon, E. et al. "Evaluation of Emetgency Evacuation Strategies for Downtown Event Traffic U$lng a Dynamic Network
ModeL Transport41Wn Rnearch Record: founus/ ofth Trtmrport4tion Rntarrh &ard 1922 (2005): 149-155.
McCanhy,J. Traffic Ana/:pis Toolbox. Washington, DC: Federal Highway Administration, 2005.
Olscrom, J. and A. Tapani. Omparison ofCar-Fol/Qwing Motkls. Publication No: VTI meddelande 960A. ISSN 0347-6049.
Linkoepig. Sweden: Swedish National Road and Transport Research lnstiruce, 2004.
I.
Schroeder, B. and N. Roup hail. ''A Framework for Evaluating Pedesuian-Vehicle lmera,cdons ac Unsignalized Crossing
Facilities in a Microscopic Modeling Environment." Proceedings ofthe 86th Annual Meeting oftbe Transportation Research Board.
: Washington, DC: Transponacion Research Board, 2007.
Schroeder, B., N. Rouphail and R. Hughes. "Towards Roundabout Accessibility- Exploring the Operational Im pact of
Pedescrian Signalization Options at Modern Roundabours. ASCE journal ofTransportation Engineering. 2008.
Tagliaferri, A. W 1-40 Lane &venal Traffic Analysis. Report# FHWNNC/2005-14. Raleigh, NC: North Caroli na Deparcmem
ofTransporcadon, 2006.
Theodoulou, G. and B. Wolshon. "Aicernadve Mechods to Increase the Effectiveness of FreewayConuaflow Evac uation."
Tramportation Rmarch Record journal ofthe Tramportation Research Board 1865 (2004): 4&-65.
Transpon:adon Research Board. Highway Capacity Manual, HCM2000. Washington, DC: TRB. 2000.
Transportation Research Board. "Traffic Signal Systems and Regional Systems Management 2006." journal ofthe Tramportation
Rerearch Board No 1978. Washington, DC: TRB, 2006.
U.S. Environmental Protection Agency. Greenhouse Gas Emissions from the U.S. Transportation Stor, 1990-2003 . Washington,
DC: U.S. EPA Office ofTransportation and Air Quality, 2006.
U.S. Environmental Protection Agency. Prifmedl&commmtkdMotkls. Washington, DC: Technology Transfer Nerwork
Support Center for Regulacory Aunospheric Modeling. U.S. EPA Web site: www.epa.gov/scraroOO!Idispersion_prefrec.hcm.
Washingron, S.P., M.G. Karlaftis and F.l.. Mannering. Statistical and &onometric Methods for Transportation Data Analyris. Boca
Racoo, FL: Chapman & Hall, CRC Press LLC, 2003.
Zhang, M., J. Ma and H. Dong. Droeloping Calibr41ion Tools for Microscopic Traffic Simulation Final Report Part 1: Overview
Methods and Guidelines on Project Scoping and Data Co/kerion. Berkdey, CA: Partners for Advanced Transit and Highways
(PATH) Working P<~.p; O!Jifomi<~. Dep<~.nmem ofTramponation, 2008.
Zhang, M. M. Droe/oping Calibration Tools for Microscopic Traffic Simulation Final Report Part II: Calibration Framework a
Calibr41ion ofLocaUG/obal Driving Behavior and Departure/Route Choic~ Model Paramtm. Berkeley. CA: Farmers for Advanced
Transit and Highways (PATH); University of California; California Department ofTransporcation, 2008.
Zhang, M. M. Droeloping Calibration Tools for Microscopic Traffic Simulation Final Report Part Ill: Global Callibration-0/D/
Estimation, Traffic Signal Enhancement and a Cas~ Study. Berkeley, CA: Partners for Advanced Traosit and Highways (PATH);
University of California; California Departmenr ofTransporcacion, 2008.
Chapter 12
237
238
238
246
246
250
3.1 Introduction
2SO
251
253
257
4.0 REFERENCES
259
2S9
261
261
1.0 INTRODUCTION
here are several types of pedestrian studies designed to capture some aspect of pedestrian and bicycle beha.vio.r
or pe.rform.ancc. Engineers use the resulcs of these studies to determine if traffic signals are warranted, devdop
exposure cb.ta for c:alcu.l.acing nonmotorized crash rates, locate and design sidewalks, crosswalks and trails,
design and implement ~ty improvements, and analyz.e roadway cro6Sings to derermine appropriate controls and
control operations for ex.ample, school crossing protection or signal timing. In recent years, measures describing userperceived quality of service of the pedestrian and bicycle modes have.received increasing attention. Methodologies are
now available IX> preditc a wer-perceived pedestrian and bicycle LOS at crossings and along trails and sidewalks. Srudy
methods apply to bicycles and pedestrians in many situations, particu.l.arly on shared-use paths. However, bicycles are
typically required co operate in the roadway a,nd follow the same laws as vehicles; this affects the ability to conduct
some of the studies included in this chapter.
The behavior or performance of the pedestrian and cyclist is generally evaluated by one or more of the following:
volume
walking or travd speed
gaps in traffic
.., ~l~~
''ofi:j{Q~.
-~
~ -------------------------------------------T~
w~---------------------
~--------------------
Da
AdWU
0\i'*-1
tN
II
I I .
I I U
II
--
.. . . . I
-"'""""'
IJ
and count types. However, its application is less common today with the availability of electronic count boards afld
laptop computers.
Battery-operated, handheld, electronic count boards are currendy the most common device to aid in c:he collcctiotl
of traffic count da~a, which can include pedestrians and bicyclists. Electronic count boards arc compact, lightweigl:1~
handheld computers with different buttons allocated to different movements at ~intersection and the abiliry to claSsify groups separately (such as pedestrians, bicyclists, etc.). They~ much simpler in design and visual display thaft
a laptop computer, and feature rugged casing and long battery life. Electronic count boards con!ain a.n internal dock
that separates the da~a by a specified interval, so no field forms are needed. Many electronic count boards are capable
of supporting crosswalk, classibcation, gap, gap acceptance a.nd pedestrian behavior studies. For agencies requiriog
more than occasional manual pedestrian and traffic counts, the electronic count board or handheld computer is a
cost-effective, labor-saving tool.
A battery-efficient lapcop computer can be substinned for a handheld count board in many applications. Often, lapcop computers are ~dy available co the analyst, making this alternative more aruactive than the purchase of nev<-'
hardware. Commonly available spreadsheet software can be used to record time stamps of different types of events uSing a macro routine. Appendix E contains an example and discussion of how to code a time-stamp maao in Miaosof1:
Excel a.nd V!sual Basic. The benefit of using a macroenabled spreadsheet to collect volumes (or ocher temporal M:r:L -c:
data) is ic can be customized co the specibc needs of the user. Also, ma.ny commercially available count boards are
resoicted to output aggregated d:ua. Alternatively, this approach allows the analyst co obtain individual time-st2.Dlpec:l
events. The downside is some software coding and post-processing analysis are required. Consequencly, this approacP
may be more interesting in research settings and other special applications.
Manual counts_ can be performed in real-time in the field or in a post-processing operation from video observationS: .
in the office. The use of videos may be practical if video is already available, or if it is easy co ob!ain, For example,
modem traffic ma.nagement centers often have live video feeds from permanent field cameras (at signalized intcaecPedestrian and Bicycle Studies 139
lions or other locations) to a central office, where they can be recorded. Video-based counts can also minimiu staff
requirements, if the same analyst can replay the video to count different movements. A well-chosen camera angle, ideally from some overhead vantage poinc, is critical to ensure usefulness of the video. Video image-processing sofrwue
can in some cases automatically provide volume data even for the harder to detect pedesuian and bicycle movements
discussed later in the chapter. Alternatively, observers can record their counts with a handheld count board, with tick
marks on a tally sheet, or directly into a computer. The abilicy to conduct other studies with video, including the
intersection and driveway studies discussed in Chapter 6, can lead to very efficient uses of the labor of field crews.
A more detailed discussion of the advantages and disadvantages of various counting equipment, as well as exhibits
featuring commonly usd devices, is in Chapter 4.
2.1.1.3 Penonn~l &quirtti
Trained observers are required co perform accurate manual pedesuian counting. They muse be relieved periodically to
avoid fatigue and degraded performance. Breaks of 10-15 min. should be scheduled at least every 2 hours.lf the data
collection period is more than 8 hours, breaks of30 to 45 min. should be allowed every 4 hours.
The siu of the data collection team depends on the length of the counting period, the type of count being performed,
the number of crosswalks or bike lanes being observed and the volume level of pedestrians and bicycles. One observer
can easily handle a four-way signalized intersection with single approach lanes and low volumes as long as special classifications and/or directional counts are not required. As any or all of the foregoing parameters increase, the complexity
of the counting task increases and additional observers will be needed. The c:xact number needed can be determined
by conducting pilot studies at the locations of interest.
Duties may be divided among observers in different ways. At a signalized intersection, one may record the nonh and
west crossmlh while a second watchC$ the south and east crosswalks. In chat way, only one crosswalk is active for each
observer at any given time. Another way co divide duties is for one observer to record certain classes of pedestrians,
while the other counts coral pedestrian volumes. At complex sites, individual crosswalks or classifications may be as
signed to individual observers. Also at complex sites, one ob~rver may have the sole job of relieving the other observers on a rotating schedule basis.
- !
2.1.1.4 ~arruion
An accurate and reliable pedestrian count begins in the office. A locally developed checklist is a valuable aid, even to experienced teams, to ensure all preparations for the field study have been completed before the team arrives at the site to
be counted. Exhibit 1-1 is an overall checklist analyses can modify or add to for local conditions. Preparations should
start with a review of the purpose and type of count to be performed, the count period and time intervals required,
any information known about the site {for example, geometric layout, vehicle or pedestrian volume levels by time of
day, signal timing. etc.). This information will help determine the type of equipment to be used, the field procedures
to foUow and the number of observers required. If the purpose of the study requires ideal weather conditions, analysts
must prepare criteria for canceling the count or procedures for dealing with inclement weather.
The selection of equipment will dictate the type of data forms noeded, if any. Header information should be filled in
co the extent possible in the office and the forms arranged in the order they will be used by each observer inthe field.
The preparation checklist should include equipment items such as pencils, batteries, stopwatches and blank videotapes. Returning to the office to reuieve forgotten items may delay the Start of the srudy or cause it to be poscponed.
An inadequate number of forms to complete the study could also invalidate !he study, resulting in wasted resources.
An office review of the study procedures and a check of the proper operation of all equipment completes the prcpara
cion stage.
Bicycle counts shoul<l further distinguish whether the bicyclist is traveling in the roadway and is thus treated as avehicle, or on the sidewalk. The Highway C4pad'!J Mant141 (HCM) bicycle levd of service (LOS) methodology {TRB,
2000) is currendy strictly based on an estimare of in-road bicycle volumes. Studies on multiuse paths may furth.er
include a more detailed distinction of different road users including bikes, baby suollers, skaters, or joggers. Other
attributes may also be coUecred, including gender, age category. or the use of helmetS. Exhibit 12-2 gives an example
dai2 collection form.
II ~!
Weather_ _ _ __ _ _ _ _ _ _ _ _ _ __
StlrtTimt
EndTim---------------~
~ti~----------------------------------------------- Mode
u.. Grovp
Oi..alon
Ale Gonder Htlmtt
l.oation
Wlleeled-Oevic:o
Pedestrian
lime
~
r
Li
.____
...
!~ 1
l
"
0
ili Li li l' 1 h1 ! I ~ I! I ~ li
-'--
;) I ~
..,
~
' ,__
the conditions under which the counts were made. The form itself should clearly indicate the movements, classifications and time intervals. Location descriptors are critical for later data reduction, including a north arrow, sneer names
and landmarks.
The observer must concentrate his or her attenrion on accuratdy recording each counr in the proptr place on the
form or with the proper button. Special care muse be raken with dearonic count boards or handheld computers to
ensure chey arc properly oriented co the geographic and geometric layout of the intersection. lime intervals must be
accurately maintained and coordinated when rwo or more observers arc working together. When mechanical count
boards are used, the observers must have time ro record the accumulated counts and reset the counters at the end of
each interval. Two procedures may be used to accomplish this: the short-break and alternating-count procedures and
are described in detail in Traffo Enginemng by Rocss, Prassas and McShane (2004).
2.1.2Auto71Ultic Counts
2. 1.2.1 Purpor~ tJnd Application
.
There arc some applications of pcdesuian and bicycle volume dara that do nor require complex classi.ficarions, or on
the other extreme are so complex they must be recorded for slow- or still-motion analysis. The simple counts may be
needed for extmfkdpmods oftim~ (such as days, weeks, or even months) at busy intersections or in remote locations
on paths and trails. The use of observers for such purposes would be cost prohibitive. When complex behavior classifications are required, the actions {for example, head movements) may be too quick for observers to see and record.
Automatic video recording provides a means of gathering these pedestrian or bicycle data at a reasonable expenditure
of time and resources.
2.1.2.2 Equipmmt
. There are several types and models of automatic volume data collection equipment. This equipment generally 'includes
rwo basic components: sensors to detect the presence of pedestrians or bicycles, and a data recorder. Sensors may
employ active or passive infrared light transmission and detection, Piez.o 61m, time-lapse video, in-pavement loop detectocs and pneumatic tubes (Schneider, eta!., 2005). These technologies can reduce labor costs compared to manual
counting methods. Classification of user cypes can be difficult with automated techniques, but can provide extended
counting periods.
N.ew technologies can further be used to record more detailed travel activicy of pedestrians and bicyclists, including
pedometers, acceleromete.rs, GPS transponders, location-tracking mobile telephones and laser counters suitable for
measuring traffic on paths and trails. These technologies allow the analyst to collect much more than just spot volume
data, and other studies are discussed later in this chapter.
2. 1.2.3 Pmonnel &quired
The only personnel required for making automatic counts are chose needed to install, calibrate and recov~ the equipment. Crew sizes of one to two are usually sufficient to deploy and recover most counting equipment. For complex
video studies, personnel are needed to process video recordings in the office.
2. 1.2.4 PrqJaratWn
fu mentioned previously, 6.eld work $hould never be undertaken without proper preparation in the office. A locally
prepared checklist is an invaluable aid even for the most routine task. The purpose of the count will drive the type of
equipment to be used and the deployment procedures. AU equipment should be checked to see it is functioning properly and appropriately positioned. An ample supply of accessory items (such as nails, clamps, tapes, adhesive, chains,
locks and batteries) and aU necessary tools should be provided. Field personnel should be prepared to provide business
cards of the principal investigator of the project and an authorization letter from the sponsoring agency with contact
information of the appropriare official. Analysts also cannot forget weatherproofing for cameras. If night photos or
video is desired, a camera applicable to nighttime settings should be used. Interference of the equipment with pedestrians and bicyclists should be minimized.
2.1.2.5 Selecting the Count Location
The street or highway on which the count will be made and the generallocation-midblock or incerseccion-where
the counters or cameras will be placed is decided in the office and is a function of the type of study being performed.
The exact locations of the cameras, count recorders and sensors are usually determined in the field. In the case of
cameras, the most important factor is the field of view. The location of the camera should take into account adverse
weather and reduced visibilicy from shadows. For studies that investigate compliance at signalized intersections, a
supple!Jlental camera may ~e needed to capture signal changes. Multiple cameras can be synduonized in the office
242 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
.using a video multiplexer device. Additional informacion on count location selection, installation and .rerrieval of
'counting devices and count periods can be found in Chapter 4. Chapter 7 contains informacion about pedestrian
rolume warrants for traffic signals.
.
Seep 2.
Seep 3.
Seep 4.
Step 5.
Step 6.
Step 7.
The second step is to select the count interval. The choices are 5, 10, 15, or 30 min. of counting in the middle of each
time period selected in Step 1. The uade-off in the choice of count interval is between economy and accuracy. The accuracy of the estimate incrases with the length of the count interval, whlch in turn increases the cost. If, for example,
a likely outcome is being verified, a shorter count interval may be satisfactory.
In Step 3, the analyst chooses the order of counting and the specific time periods for each location. In Step 4, data
are collected. One observer is usually sufficient depending on the volume level. At an intersection, the observer simply
conducts short counts of each crosswalk or bike lane in turn according to the schedule. Check the user manual for
specifics on dealing with signalized locations.
Step 5 is tci select the expansion modd coefficient and exponent. From Exhibit 12-3, choose the proper values, based
on time period and c:ount interval, of a and b for the expansion model:
volume - a X counf'
Equation 12-1
where:
vo/umt : estimate of pedestrian volume for the 1-, 2-, 3-, or 4-hour period of interest
. Count
Seep 6 is co compute estimated volumes. For example, if 25 pedestrians were counted during a 10-min. interval in the
middle of the time period from 7:00 and 9:00 a.m., the estimated volume for the 2-hour period would be:
volume 20.9 x 25o.au =296 pedestrians
Since the numbers calculaced from the expansion models are estimates, the range of values within which the acruaJ
volumes are likely co fall may be established with Seep 7, determine estimated volumeTanges. Exhibits 12-4 through
12..7 contain range factors (in percent) by pedestrian volume level and counc interval for each of the four dme periods,
respectively. For the example above, the factor would be (from Exhibit 12-5) :t 32 percent. Thus the actual volume for
the 2-hour time period would likely lie between 201 and 391 pedestrians. Thls range may seem large; however, there
are many siruations where dais levd of accuracy is sufficient. Simply !mowing whether a crossing location has a low,
moderate, or high pedestrian volume may be adequate to select the proper pedestrian conuol or accommodation.
f.
i
\;
Source: Min&Q er al. Mt4Niring Pt~n Yofwnes; A UHrl Uznu4 Yolui7U' IlL
FHWA-IP-88-0~0.
Sowce: Mingo ct a!. Mus11ring P~rlenrUm W.lwna: A Usn-'s Manll41, W.illmr 1/J. FHW.A-IP-88-030. FHWA. Table I, 1988a.
Source: Mingo et al. M~llJuring Pdestrian W.luma: A User's Manll41, W.lumr 1/J. FHWA-IP-88-030. FHWA. Table 2. 1988a.
Sowcc: Mingo et a!. M~IIJIIring P~destriim W.illma: A Usn's Manll41, W.lsmu 1/J. FHW.A-IP-8~30. FHWA. Table 3, l 988a.
Source: Mingo et al. M~llSUring Ptdotrian W.illmn: A Usn-'s ManiUil. W.lwnt 1/J. FHWA-IP-88-030. FHWA. Table 4, 1988a.
Walking speeds typically range &om 2.8 ro 5.7 feer persecond (fps) for fully abled pedestrians and from 2.0 to 3.7 fps for disabled pedestrians (Dewar 2007). Recent rese:uch regarding pedesrri.an walking speeds has found that slower wa!J<iog speeds
ace needed in the calrularion of pedesman clearance times to accommodate older and slower pedestrians. The recommended
walking speed for calculating the pedestrian clearance time is reduced from 4 fps to 3.5 fPs, except where extended pushbutton presses or passive pedestrian dercction has ~n installed for slower pedestrians to rtquesr additional crossing time.
'I
Ii
Also, based on the same research, guidance is added that the rota! of the walk phase and pedestrian clearance time should be
long enough to allow a pedestrian to walk &om the pedestrian detector ro the opposite edge of the tnvded way at a speed
of3.0 fPs. This change will ensure that slower pedesuians can be accommodated at longer crosswalks ifthey starr crossing at
the beginning of the walk phase. If this calculation finds that sufficient crossing time is not available, additional time should
be added to the walk interval.
The srudy should be performed at the locacion ofinreresr under the conditions of interest. One or more o~servers may
be used based on how much the conditions vary over time and the number of classes of data desired. The observers
should be positioned where they have a clear field of view and do not distract passing pedestrians. Observers mark a
measured distance along the path tnveled by the pedestrians and then simply time individual pedestrians through the
speed trap. A sample of 100 observations is generally adequate. Analyz.e the data by first calculating each individual
average walking speed by dividing the uap distance by the observed time, then classifying the observed speeds, and
6nally plotting the cumulative percentage of observations by class. This will produce a cumulative speed curve from
which values of various speed percentiles may be derived (see Chapter 5). The 15th percentile speed is a generally accepted value to use in timing signals for pedestrians (KeU, 1991). Bicycle travel speed studies are performed similarly,
by measuring the time required to travel a predetermined distance.
,Gap studies consist of measuring the predominant pedestrian group size, d etermi ning the length o f a minimum
~dequare gap. measuring the gap sizes in the uaffic strea m and derermining the quantity of adequate gaps. The
principal application of the study results is in analyzing roadway crossings by pedestrians to d etermine appropri
:1.: traffic controls and safety improvements. Applications for bicycle traffic are rare, but may be adop ted from the
foUowing discussion as necessary. The results of gap studies are used in traffic signal warrant analyses and school
crossing studies. In addition to the techniques described bdow, the procedures for determining gap acceptance
characreristics for drivers of vehicles entering or crossing roadways described in Chapter 6 may be ad apted tope
desrrian gap and bicycle applications.
L
tC =s-+1I
Equation 12,2
where
t, = critical gap for a single pedestrian (sec)
sp ..
3.5 ft/sec
t, : 3sec
,I :F.ii1!1mnv~:
.. ..
~.
r.~
Location
Date
Crossing Distance
..... .
'
.!~.
Time
Walking Speed
Critical Gap
Tally
Total
10
11
12
13
Discard Gaps less Than Critical Gap
14
15
16
17 1111
18 J.lii')IHII
19 Jilt' I
20 Jilt 1111
21 1111
22 lJIIt
23
24 II
8:15AM to 9:00AM
3.5 feet/sec
16.7 sec
4
12
6
9
4
5
0
2
25
26
27
28
29
30
. ......
.,..,. .
.
'
- ~~~~~~
lUI
I
Ill
1
3
0
1
51
I
Total Adequate Gaps
'
If p~atooning of pedestrians is observed in the fidd, the spacial disoibucion of pedestrians should be computed using Equation 12-3, to determine group critical gap (TRB, 2000}. Othetwi.sc, Equation 12-4 ca.n be used to estimate
the platoon size for usc in Equation 12-3. Group critical gap is determined using Equation 12-5. If no platooning is
observed, spatial distribution of pedestrians is assumed to be one.
NP=INT[
0(::-l)]+l
Equation 12-3
where:
N,
N.
WE
8.0 = dcf.wlt dear effective width used by a single pedestrian to avoid interfetence when passing other pedestrians
J!c= - - --
Equation 12-4
'ol(here:
\
N,
'
vP
r.
tc &t.+ 2(Np-l)
Equation 12-5..
where:
tc
t,
N,
The measure ofdfecriveness for a pedestrian crossing is delay. The average delay per pedestrian is a function ofcritical gap
and the vehicular Bow rate as shown in Equation 12-6. Exhibit 12-9 relates the calculated average delay per pedestrian
to an appropriate level ofservice (LOS) and the likdihood of risk-taking behavior following the HCM (fRB, 2000).
Equation 12-6
where:
a,
t0
Measured gaps arc rounded to the nearest second. A rick mark is placed in the rally column corresponding co rhe
measured gap size char equals or exceeds rhe minimum adequate gap. The tally marks arc then totaled for each gap
size. The sum of these corals is the number of gaps of sufficient length to accommodate rhesafe crossing of8) percent
of the pedesrrian groups using the crossing at a day and time and under the conditions similar co chose of the srudy.
The example shown in Exhibit 12-8 indicates chat a total of 51 adequate gaps were recorded.
Gaps may also be measured using electronic count boards or laptop computers in place of the stopwatch and tally
sheer. The observation procedure is essentially the same as described above. [merna! clocks in the computer record the
times. Observers push the appropriate burrons co record gaps in the traffic. The primary advantage with this technique
is chat computer software reduces the data, rhus saving time. This approach can also be adapted co record accepted and
rejected gaps and from those data field-estimate the critical gap using methods described in Chapter 6.
To evaluate the study results, analysts compare the number of gaps equal co or exceeding the critical gap co che number of minutes the gap measurement study is conducted. The appropriate criteria are chen applied co che result. The
length of che srudy depends on the type of application for which che gap srudy results are being used. For example, in
the MUTCD warrant 4 for traffic signals requires that in addition to the scared minimum pedestrian volumes, there
shall be fewer chan 60 gaps per hour in the traffic stream of adequate length for pedestrians to cross during the same
period when the pedestrian volume criterion is satisfied. Another MUTCD criterion (warrant 5) states a traffic signal
may be warranted when the number of adequate gaps in the traffic stream when school children are crossing is less
than the number of minutes in the same period (FHWA. 2003). If the analyse applied chis criterion co the data shown
in Exhibit 12-8, the signal would not be warranted since the number of adequate gaps (51) exceeded the number of
minutes in the srudy {45).
This chapter gives an overview of both and refers the reader to ocher sources for details on the devdopmcnt and application of very elaborate user-perception based models. NCHRP Report 616 (Dowiing ec a!., 2008) gives a comprehensive overview of literarure on user-perception measures and model development.
A pedesrrian/bicyde/vehicle con.Bict occurs when one of the interacting agents has to cake some action, such ~ a
change in direction, speed, or bodl, in order co avoid a collision. In the definition of a conRict it is key char a colliSIOn
was imminent in the absence of such action. Assertive pedestrians and bicyclists may sometimes exhibit behavior char
may appear "risky~ co an observer, but that is still wit!Un their conuol. Enmples.include delibcratdy "forced yi~d
events (in which case the pedestrian could have stepped back if the driver had not reacted) or imcntional accepting
of a shore gap in traffic by rurtning.
The most commonly applied conflict scudy is on pedestrian/vehicle conflicts. Bicycle/vehicle conRicr st udies are rare.
Researchers have mer with difficulty in establishing a causal rdacionship between pedestrian/vehicle conHicts and
acrual crashes. While some .evidence of such a relationship has been uncovered, the complexity and rdativdy rare occurrence of pedestrian-vehicle and bicycle-vehicle crashes at a given location has to date prevented a clear conclusion.
Despite the difficulty in predicting crashes, pedestrian/vehicle conBicrs remain a useful measure of relative differences
among pedestrian safety alternatives. A number of studies have used conflicts as a measure of effectiveness for idencifri ng
pedestrian safety problems, evaluating TCDs, and comparing pedestrian accommodation designs {Campbell er a!.,
Zegeer et al., 2005). With devdopment of the Transportation Research Board (TRB) Highway Safety Manual, much forus !w been on developing pedestrian safety prediction modds (TRB, 2008), the details of which are beyond the scope
of this chapter. Exhibit 12-1 0 shows common pedestrian/vehicle crash and conflict types (Harkey and Zegeer, 2004):
zoo4.
Source: Harkey and Zegeer. PEDSAFE: Ptdmritm Safoy Guitk and Cormln7MilS1m! Stkmon Syrrnn. FHWA-SA.C4-003.
Chapter 3, 2004.
Pedestrians crossing a roadway may encoWiter through vehicles, right-ruming vehicles and left-turning vehicles. Th~ bas.i.::::;
conflicrs ocrur when the projected paths of the pedestrian and the vehicle intetsect and either the pedestrian or the \lhi:l- ~
muse change ~on and/oc speed co avoid a collision. Variations and rdinernents to these basic conBicrs have been~ -.
In addition, the severity of the conflict, as determined by the strength of the decderation or accderation, th~.speed diHe~
rW and how closely spaced the involved parties are, has been used efecrivdy in further de.6ning conflia: MOEs.
Pedestrian and Bicycle Studies 2S~
A particularly common and risky conBict is known as the multipk threat confl.ict, where a yielding vehicle in the nar
lane blocks the pedestrians' view of a vehicle in the far lane and vice versa. The situations are dangerous, because the
pedestrian alrady made the decision to cross and may not be paying attention anymore. The situation is especially
risky at a multilane two-directional crossing, where the pedestrian may already be screening traffic in the opposite
direction. This case is very common at crosswallcs at multilane roundabouts.
t. ' . . .
,,I A'
J
\
I
l
: al., 2008), but a detailed d.iscwsion goes beyond the scope of this chapter. Special attention should be given to sueet
\crossing barriers (Wdlar, 1998, Litman, 2009), as well as to end-of-trip facilities, such as (secure) bike parlci ng and
:.even showers and clothes-changing areas for commuting traffic.
3.2.5 Other &hibiutl Beh11,U,r
In addition to coolliccs and compliana with TCDs, other pedestrian and bicyclist behaviors luve proven reliable to
varying degrees in identifying problems and evaluating safety countermeasures. Examples of these behaviors for pedestrians include failure to look left and right before and while crossing, hesitating in the roadway, running, jaywalking,
use of the signal pushbutton and n:turnlng to the curb after starting to cross. Examples of these beluviors for bicyclists
are the use of hand signals, bicycle helmets and nighttime lights, or bicycle speed, positioning, use of the signal pushbutton and gencnl driving style on the roadway.
These types of behaviors represent undesirable or unique actions that can reflect some degree of threat to the pedestrian or bicyclist. Acro~odations and/or TCDs that reduce these behaviors are generally regarded to be safer. Use
of these measwu is documented in a nwnber of research studies (Harkey and Zcgeer, 2004, Fiapatric.k et al., 2005,
Hunter cc al., 2006, R.o!legercs ec al., 2007).
3.2.6 Usn-Percei11d Qruzlity ofService Me~
User-perceivod QqS measures represent a o~ paradigm in the transportation field. These QOS measures shift away
ftom traditionally used performana measures (such as delay and uavel time) and towards measures describing the
aperiena of the user-the customer. These meas11res are popular with nonmotorized road users because they directly
target their travel experiences. They are also gaining acccptana in the engineering community and are anticipated to
actually rcplaa some methodologies in the 2010 f<:lease of the HCM to describe the levels of servia for nonmotoriied road users. User-peraption-based QOS methodologies are available for the pedestrian (Dowling ct a!., 2008,
Peuitsch ec al., 2008) and bicycle modes (landis et al., 2003, Pcuitsch et al., 2007, Dowling et al., 2008), :md have
recently been calibratctl to u.s:"national.dacascts in an NCHRP project (Dowling Ct al., 2008) for pedestrian and bicycle experieoa on urban meets. Another 'FHWA research project has developed corresponding models for off-street
shared-use paths and cra.ils (Hummer et al., 2006).
The QOS methodologies are commonly developed from ratings .o f video clips showing different travel experiences.
The racers are accual pedestrians and bicyclists who assign a letter score to each clip based on their perception of the
portrayed quality of service. The research team then correlates participant ratings with. variables shown in the clip,
inclu<iing the adjaant vehicle volwne, pedesuianlbicycle delay, the number of passing events, the presence of trees
and the presena of crosswalks and bike lanes. The exact process of developing user-peraption-based QOS models is
beyond the ~pe of this manual, but is described in detail in the literature (Dowling et al., 2008).
The application of QOS measwes typically uses operational data (adjacent traffic volwnes, speeds, signal timing)
and inventory data such as the width of bike lanes and sidewallc3, the presen~ of buffers between motori.wi ancl
norunotori:z.ed transportation and landscaping. Operational data is jneasured in the field using methods discu.ssed in
Chapters 4, 5 and 6. Inventory data is d.iscwsed in Clupter 15, and can be measured in the fidd or estimated in the
offia using GIS databases. Invenrories arc often supplemented with photo or video records of different pedestrian
and bicycle facilities.
The following discussion highlights special considerations for behavioral and user-perception studies for the pedesuian and bicycle modes.
. The aforementioned example illumates a common problem of behavioral studies. Depending on the p roblem sClltC:
\ ments, it may be the case that one observer cannot reliably record all necessary factors, either bccal!se of a high co~r1 l
_ tive load or the observational field of view. In addition to the use of multiple observers, it is possible to use mult'ple
video cameras co capture different angles of the same event. For example, Exhibit 12-11 shows a screenshm from a
picrure-in-piccurc view of a pedestrian signal compliance study (Schroeder ct al., 2009). One camera captures che entire crosswalk and pedestrian behavior, while a second camera is zoomed into the signal phase. The picrure-in-piccure
(PIP) display can be created in the office using a video multiplexer device that is common to security camera systems.
The experimental design and sample size considerations of controlled experiments follow principles discUS!~ ;..n
Appendices A and C. The guiding principle is to control for only one or two factors at a time, and control for ctb- e::r
souxces of variability.
3.3.5 D~~t~~for Usn--Perception MnhotlologUs
The input data needed for the QOS methodologies describe aspects of comfort, convenience and safety of nollJil..c:::>toriz.ed road users on a facility. Typical variables include traffic flow factors (vehicle volumes, speeds, heavy vehicks :::>
roadway design characteristics (crossing width, number of lanes, presence of bike lanes), characteristics of TCc::::-' 5
(signal timing) and aesthetic appeal (presence of trees, pavement condition). Exhibit 12- 12 shows t!te list of iap'-!-t
variables in the auto, tranSit, pedestrian and bicycle QOS models from NCHRP Report 616 (left column) . Tbcpo..Jf'Pedestrian and Bicycle Studies
zs; !!f!!!S
models for bicycle, pedesuian, transit and auto modes. While only the firn
two are directly rdated to this chapter, variables for transit and auto modes are included for comparative purposes..
The "X)OC markings indicate interactions between modd input variables and factors on the facility like design, traffic
control, or volumes.
-.
-.-.- ...-.-. -
_-- -- -
.-
-_"
-.
..
..
.. - --"- ,_
' '~-
~~~l<~~~~~
_,__
1_ ........ _
Foolllly.,....
,ocllly Coolrol
TrtMit~
.....
'-LO.lOU<
-~
_.....,._.......
--
"""-
~~
lOU<
XXX
"""
lOCI
XXX
XXX
...........
"""
"""
XXX
"""
XXX
-~
"""
_.,_....,
-c..-
XlCX
lOCI
XXX
XXX
XXX
-LOS__
~,...,....~
XXX
0..--
XXX
-Oorollr
_.,
XXX
-~
XXX
XXX
XXX
0>.-,..,.,.~
XlDt
Prllencl oiT,_
XXX
-__
--eo--o---
XXX
lOCI
XXX
XXX
_,.,,_.......
OloiMco1li-U..
XXX
XlDt
DX
XXX
NuN- On Red
XXX
XXX
DX
XXX
XXX
lOOI
XXX
"""
XXX
"""
XXX
XXX
XXX
XXX
XXX
IX1
XXX
IX1
xu
lOOI
XXX
XXX
XXX
lOCI
lOUt
XXX
xu
"""
"""
XXX
xu
o-.Doloy
......-~
XXX
"""
-lMo'll
S9*Ctdo fAoGI\
lQQI
XXX
)00[
XXX
0..-U..
DX
.....a.-n..o
"""
XXX
XXX
.... LOellrodllt
--llua-
......... (orlleloyt
LoftlVnU..
XXX
-LOS llodoil 2
IT-1.011-lOS
I -l)oo
.....
..-...,_
XXX
lluolpood
"""
XXX
JCa
JCa
"""
JOCX
xu
XXX
JCa
"""
"""
Source: Dowling et al. Interaction of Modal LOS Modd Inputs. NCHRP Report 616, MultitriiHW Ln!e/ ofSmtict for Urb~tn
Strttts. Project 3-70, page 93, exhibit 99, 2008.
The datt clements listed in the first column of Exhibit 12-12 are addressed by datt collection methodologies described
el.scwbere in this manual. Chapter 4 provides methodologies for volume nudies, Chapter 5 describes speed nudies, a.nd
delay-based measwes are discussed in Chapters 6 and 9. The reader should also refer to Chapter 15 for guidance on
roadway inventories and the collection of asset management data.
A3 an alternative to the formal QOS methodologies discussed above, agencies commonly develop more qualittcive
survey- or interview-based methods ro assess the user-perception on, for c:xample, a panicular multiuse path or a
downtown area. Exhibit 1213 shows a walkability checldist {FHWA, 2009) that can be used as a basis for designing
custom survey forms (www.walkinginfo.org/library/details.cfm?id"l2). Researchers b.ave conducted large-scale applications of on-street surveys and daca collection with many volunteers who walked or bicycled a particular course with
intersectioru, sidewalks and crossings to rate their experience (for c:xample, Peuitsch eta!., 2005, Landis ct a!., 2003).
A comparable bikeability checklist is also available (www.bicyclinginfo.org/libruy/details.cfm?id3).
- -
"""'lr\ Cr'\ITII""'._I
,.
. ..,....,...
.
........
Rating Scale:
Location of w a l k - - - - - - - --
_,,,
0 Soo'w ptd;llcat.J;
D N.-..lb Of' poW la.I'Ud and cop~d
0 Sldrwalb we~ brcbo OC' cracbd
SW.W.Ib ....... b&ocbd 'kh pokof,
c
0
0
"*"'
No ~,....,orr thouldltn
"lOo M~Kil cnJik
~~~-~~~~~~~~~
Locod... ot.,-1....~
Jladllcl(cla<;if.,.> '
0
12J4S"
Son. p.obtn..:
[] lloa.4 ..-u roo wW4I
Cl T~ . . . . m..de u _ . -.o kJoscw 4jd
ar-
_.,...acatPiiiDI>tDaoou
.-co
0YC'S:
ONo
C.O.tat.~lborwt.youc:ould
ONo
ov..
=:t~~=~na:.::: dt~
CNo
ov..
ONo
=~:::.:c:~~
... .anctt.c...,.by;.en'P
l..oatioaa oC~c
_._
. --,..-
Some,__,
Dftwoft..
0 - - . , . _ , . wlcho _...,.
oo......--
o~
a.~,
(ebr:J. oa.)
1 2 . . . 5~
k4
1 2. .J <f I '
0 Y
c - . , . - __
ce;:: ~".,
--~"''
"''"' e:
ov...
~~
.-.d
..
- -..-
....a.:~
0 Pad4 can bJocb4 . - vMw ~trdc
0 N.-..t ... -
O-n..
_,..
..... _ _
t.ocoo-.ofpooblcaoc
OSo.._~..,_
Radacfl:l k-)
l1J "I '
'--
--:t. _ . _
~--
ToiOI _ _
2'-".JO
a.- .
.....,._._ .,._,..
Cddnae.l ,.,_ bn'e
1t- 1 S
5-tO
~ k.c
k liMed& ..-k.
l t.,...&.ocoi"""'''k.,....._......_
bcclt..rdua ct.c.
l('tadilaltb~
Source: FHWA Wallcabilicy Checklist. FHWA. Pedestrians and Bicycle Information Center. www.walkinginfo.org. Accessed
August 2009.
KlllsboroUJIISII'III, Ralelch.NCSouthbounc!Podutrloi\Croulnas
C.. . . ~ D . . - . _ CIIIO-UII t.J.e ~JM at-..~ . . . . .
~l
~I
II lj ,
I ~ ~,;:~ '
A
f~ --~
. ,, l i-~' I~L~~ , _
'. 41 ,! 'I'. Ir 1~~
Jl~
'11
1
"":.:'ll':: ~~l
II'V ' \
'
r----t ~ 'r-''r-':r--TT---T'.. ..
"'
J
J 1
.3
..
User-perception studies are commonly combined with spatial GIS analysis to provide levels-of-service maps of the
studied regions. Different colors can illustrate regions of particular good or bad LOS scores. The same approach can
be taken fur other QOS measures. Exhibit 12-15 shows a map of "bicycle ease of use for downtown Austin, TX,
USA. Different shading represents Low, moderate and high ease of use, as weU as barriers to bicycle connectiviry.
~TmFAcunes
~-- --.=""'L_
...,_,~
__:._~'_
....:...
.. ..., ~----
4.0 REFERENCES
4.1 Literature References
Campbell, B. J., C. V. Zegccr, H. H. Huang and M. J. Cynecki. Report FHWA-RD-03-042: A RrvitW ofPedestrian Saftty
!WMreh in tht Uniud St4Rs and Abrolld. Washingron, DC: Federal Highway A.dminstruion, 2004.
Dewa(, R. E. and P. Olson. Pedemians and BU:yclist~, Human Facton in Traffic Saf~. Tucson, AZ: Lawyers&: Judges Publishin~
Company, Inc., 2007: pp. 568, 571.
Dowling. R. F. NCHRP Report 616: Muhimlldai Ltutl ofSmtict Analysis for Urban Sutm. Washingwn, DC: Natioru.l
Cooperative Highway Research Program, Federal Highway Administration, 2008.
~ckral
Highway Administration. Manual on Uniform Traffic Control Dnlicn. Washingwo, DC: FHWA, 2003.
Federal Highway Administration. Manual on Uniform Traffic Control Devicl!. Washingron, DC: FHWA, 2009.
Federal Highway Administration. "Walkabiliry CheckliSt. Pedestrians and Bicycle Informacion Cenrer. www. walkinginfo.org/.
Accessed August, 2009.
Fitzpatrick. K. et a!. TCRP Report 112/NCHRP Repon 562: Improving Ptdertrian Safiry at UMgnalizul lntmections.
Washington, DC: Transportation Research Board, 2006.
Gerwchar, D. R. and S.E. Hassan. "Ddver Behavior in Yidding to Sighted and Blind Pedestrians at Roundabouts." journal of
Vuuai Impairment and Blint.Wss 99, No. 5: 28~302.
.
Harkey D. L., and C. V. ZegeeL PEDSAFE: Ptdertrian Safoty Guidt and Counttnn4a.:s-t= Stkction Syrttm. Report FHWA,SA-04-003. Washingron, DC: ~decal Highway Administration, 2004.
Pedestrian and Bkyde Studies
25~
Design. tznd Opmttion ofShartd-Ust Ptuhs FiMI R.tpon. Report PHWAHRT-05-137. Mclean, VA: Federal Highway Adminisuarion, 2006.
Hunter, W. W., L Thomas and J.C. Stutts. BIKESAFE: Bicyck Coun~1'7MIZSUrr &kctU!n Systnn. Report FHWA-SA-05-006.
Washington, DC: Fedenl Highway Adminim:uion, 2006.
Iacono, M., K Kriuk and A. EI-Geneidy. kess to Destiruuions: How Close is Close EMugb? .&timllting.A=rate Distana DeCAy
Funailms for Multipk Modn tznd DijformJ Purposts. Report 2008-1I. Minneapolis, MN: University of Minnesota Center for
Truuporration Studies, 2008. www.ets.umn.edu/Publication.s/ResearchRqx>rts/pdfdownload.pl?id=916.
Kdl, J. H. Tnmsportlllion Plmming Harulbook. Chapter 2: Transportation Planning Studies. Englewood Cliffs, NJ: Prentice Hall,
1991.
landis, B. W. et al. "Inteneetion l..eve! ofSetvice for the BiqdeThrough Mo-anent." TI'IINJ'Drtlllion &starch Rwml:founlill of
tht Transportlltion &wuch Bocrl1828 {2003): 101- 106.
Litman, T. EwdU4ling .Acctstibilityfor Transport4tion Plllnning. Victoria, British Columbia: Victoria Transport Policy ln.sticuce,
2008. www.vcpi.org/aa:ess.pdf.
Litman, T. ~ Ejfts, Tnmspvrtlllion Cost tzndjknrfo.ANJysis. Victoria, British Columbia: Vicroria Transport Policy
Institute, 2009. www.vtpi.org/rc:a.
Mingo, R., H. D. Roberaon and S. E. DavU. Mt11111ring htiettria-tr VOlumes: .A Usui M~~IUI!, VOL IlL
Wasjllngton, DC: Federal Highway Administration, 1988.
FHWA/IP-88~30.
Petritseh, T. A. ec a!. "Level-of-Service Model for Pedestrians at Signalized Intersections. T11Wpol14tion Restllrrh P.=rrl.]ownud
o/tht T'rtmSportlllion Resum:h Botmi 1939 {2005): 5~2.
Petritseh, T. A., et al. Bicycle l..eve! of Service for Arterials. Tnmsportlllion Restllrrh &corrl: ]()1jrn41 vftht Tnmsportl#ivn
&surch Boart/2031 (2007): 34-42.
Petritxh, T. A., ec al. "Pedestrian Level-of-Service for Arterials." Transportlllion &starch Recortl: joUrn4/ oftht Transportation
&stJtrrh Botmi2073 (2008): 58-68.
Pline, J. L Trtifjit Enginming Htuulhook. Chapter 2: Traffic Studies. Englewood Cliffs, NJ: Prentice Hall, 1992.
Rodegcn:s. L ec aL NCHRP Report 572: ~in tht lJnitl,J States. Washington, DC: Transportation Resean:h Board, 2007.
Roess, R. et al. Tmffic Engintning. 3rd ed. Upper Saddle River, NJ: Pearson Pre.tltice Hall, 2004.
Schneider, R., R. Patton, J. Toole and C. Raborn. Ptdatrian tznd Bicyck Data Colltion in United St4tn Communidn.
Washington, DC: ~enl Highway Administration, 2005.
Schroeder, B. J., N. M. Rouphail and B. A. I..ehan. Observational Study of Pedestrian Behavior Along a Signaliud Urban
Corridor. Transportation Research Boaro 88th Annual Meeting. Washington, DC: Transportation Research Board, 2009.
Transportation Research Board. Highrv4y CAptzcity Mmrwal. Washington, DC: TRB. 2000.
Transportation Research Board. NCHRP Web-Only Document 129: Phast III Pttlestri4n S4fity Prrdiaion Mttholoo. Report
FHWA-HRT-04-100. Washington, DC: National Cooperative Highway Research Project. McLean, VA: Transportation
Research BoarQ. fedenl Highway Administration, 2008.
VtaOria Transport Policy lnstilll!e. UnivmJ Dmgn-T~ ~that~ .AlJ Usm.lndwlint Ptttpk wiJh
Disabi/itks, anJ Other Spma/Nd. VICCOria. British Columbia: Vicroria Truuporc Policy Institute. www.vcpi.org/tdm/aim69.han,
Accessed January 5, 2010.
Wdlar, B. W.Jbnt Surity IIUiex. FiNd Rrpon, Ottawa. Ontario: University of Ottawa Geography Oqlarunent, 1998.
hccp:l/aix1.uottawa.cal~wdlarb/Dcsign%20Swdy%20Publications.br:rn.
Zegeer, C. Y. et a!. S4foty Effms ofMarJud vmus UnmarkM Crosswalh at UncontrDlled LoctUionJ: Final &port anJ &commmdtd
Guuirlina. McL:an, VA: Federal Highway Administtatioo, 2005. www.tfhrc.gov/sali:cy/pubs/04100/inda.htm.
Flannery, A., D. McLeod, N.J. Pedersen. cuswmer-Based Measures ofl.evd of Service." ITE]ournal76, No.5 (2006): 17-21.
Patten, R. S., R J. Schneider, J. L Toole, J. E. Hummer and N. M. Roupbail. Shared-Use Plllh Ltw/ of&rvi&t Calcuiatot-A
lhn-s Guitk. Report FHWA-HRT,.05-~38. Washington, DC: Federal Highway Adminsuation, 2006.
Vandehey, M. NCHRP 03-92: Production oflhe Year 2010 Highway Capacity Manual. Washington, DC: Transportation
Research Board, 2008.
Chapter 13
....
...
..
. . . . .
...
. . . . . .
. . . . . . . . . . . . . . . . . .
..
..
. . . . . .
. . .
..
...
]os~ph E.
EJit~dBy:
INTRODUGION
264
264
265
265
265
265
265
268
271
271
272
273
281
283
285
285
4. t Sampling
285
286
288
288
5.0 REFERENCES
288
5. t Literature References
288
289
289
1.0 INTRODUCTION
ublic transportation studies provide operat.ors with the information they need to make inteUigent choices about
services. Studies provide the numbers and trends whlch may indicate changes in operations are needed. Studies
indicate whether patrons have responded to changes. Studies also provide data for comparisons berween agencies, to measure the quality ofservice and to provide the basis for funding decisions. It is' important that public transportation studies be conducted properly. A poorly designed or executed study may be worse than no study because the
numbers may provide operators with c:xrra confidence in a mistaken course of action. In addition, studies may consume large quantities ofscarce funds. A mff of over 30 full-rime data collection personnel (referred to as "checkers" by
public transportation professionals) is not unusual in a large agency. Automatic data collection equipment can be expensive, but agencies increasingly make the investment to obtain more complete data on transit system performances.
Thls chapter applies primarily to srudies of fixed-route bus transportation. Many of the techniques discussed also apply for other fixed-route modes such as streetcars. Demand-responsive (for example, parauansit) servit;es have different purposes and vasdy different ways ofoperating than do fixed-route services, so their study techniques are generally
different. Those interested in studies of paratransit operations are referred to resources provided by the Easter Seals
Project ACITON (Accessibk Community Tranportation in our Nation) and specifically its bus stop checklist (Project
Action, 2008).
The studies discussed in this chapter are primarily those whlch are useful for operators studying their own services.
Studies are also conducted to meet the needs of outside funding sources, investors, the media and the public. In the
United States, agencies that receive federal funding conduct studies to provide National Transit Database (NTD) data
to the FTA. The types of data agencies are required to report include (mdprogram.gov):
Operational Characteristics: vehicle revenue hours and miles, wilin.ked passenger trips and passenger miles, etc.
Service Characteristics: service reliability and safety, etc.
Capital ~enues and Assets- Sources and uses of capital, Beet size and age and fixed guideways, etc.
Fmancial Operating Statistics: revenues, federal, state and local funding, costs, etc.
Transit operators conduct srudies under vecy rigid sets of rules that are nor repeated in this chapter. Thls chapter further focuses on the conduct of the study rather than the establishment of data collection programs or the manipulation of study results.
Fmally, the chapter is limited to studies requiring actual field data collection and does not include any discussion of
the coUeccion and analysis of routine management informacion. Counting revenue, counting transfers, conducting
an inventory of equipment and parts, collecting personnel data, compiling cost informacion and other continuous
management informacion efforts are outside the scope of this chapter.
.,.:A a
lAAMt l.t\1
1"'1~
"U,Ir\ t:r.I"TI/"'1.._1
The customer-service-based approach to transit performance measur~ is reBected in the TCQSM user-peretption
measures. That manual emphasizes the use of measur~ that are useful to the operating agency, but also reBect the
travel (or waiting) experience of the transit user. The TCQSM defines transit LOS for three levels of analysis: the
transit stop, the route segment and the overall sysrcm (definitions &om TRB, 2003).
Transit Stops: measures addressing transit availability and comforc and convenience at a single location.
Performance measure values in chis catego ry will rend co vary from one location co another, since these
measures depend on passenger volumes, scheduling, routing, and stop and station design.
Roucc Segment:;/Corridors: measures chat address availability and comfort and convenience along a
portion of a transit route, a roadway, or a set of parallel transportation facilitie.o> serving common origins
and de.o;tinarions. The.o>e measure values will tend to have less variation over the length of a route segment,
regardless of conditions at an individual stop.
Systems: measures of availability and comfort and convenience for more than one transit route operating
within a specified area (that is, a district, city, or metropolitan area). System measures can also address doorto-door travel.
In an effort to incorporare user-perception into the evaluation of transit systems, the TCQSM defines smnct mttzJures
for each level and for each of two QOS categories: availability, and comfort & convenience. Service measures are chose
mecrics char are used to define transit LOS. The TCQSM also provides ocher performance measures that are useful ro
the agency, but chat are not directly used to derive the A-F letter score for uansir operations.
The TCQSM recognizes both service categories are important: Passengers expect transit services chat have broad
temporal and spatial coverage (availability), and services that are comfortable, convenient and safe (comfort & convenience). Exhibit 13-1 shows the service measures used in the T CQSM to define LOS for the thiee analysis levels,
as well as other performance measure:; that can be useful to transit agencie.o>. Exhibit 13-2 gives definitions for the six
service measures as defined in the TCQSM. Note that information in Exhibits 13-1 and 13-2 applies to fixed-route
bus service and that the TCQSM defines many additional measures for ocher forms of public transportation,
Availability
Other Measures
Reliability: on-time
Service Measu.re
Reliability: headway
adherence
Missed Trips
Comfort and
Comoe.o..icnce
Other Measures
Measwes adopted from Trrtn!it Capacity and Quality ofStroict Manual, TRB, 2003.
Complicated and requires spatial GIS analysis.
~~~.
' ~
-.
p~~~-
~~-~~~~~
~
...: 1 o. .
- - ' .. -""''. '' ' ~jto ~ . . f.l..- :1o.....
t ' J. ~ ~ . .
~ ..... "
'
~..-~
-~~...........
" ........~;-~'';i.;i!';
,~.:.::-4:~!
".!{:tr~f~\~~
. . . . . . ~.Jo.a:.
.. "1- ..
,.p; ..-,. ,
'
'
Service Me:a.sure
Definition
Service frequency
Hows of service
I. The number of hours during the day betwee n the start and
end of service on a transit route, also known as the service
span. 2. For calculating transit level of service, the number of
hours during a day when service is provided ar least hourly
on a uansit route.
Area, coverage
Passenger load
~liability
Travel time
d.ifferen~
-----
-----
------
Note: The equations wed 10 estimate some of the service measures in Exhibit 13 1 are estimated from a range of input vari
ables that go beyond the scope of this chapter. The reader is referred co the TCQSM for dewled discussion of the methodology. The focus of the following discussion is on performing field studies that can be wed to evaluate fixed-route bus services,
including srudies to validate the service measures predicted by the TCQSM.
Source: Kittelson & AssociateS Inc. et al. TCRP R.tporr 100: Transit Capacity and Quality ofSnvice Manual, 2nd Edition.
Put 8 "Gio.Ssuy", p. 8-i, 2003.
9. Survey (3.3)
The different srudies are described in detail in Section 3 of dU.s chapter. Exhibit 13-3 relates the performance measures
to the applicable 6dd studies.
"tea
I:Mt:.l.._li:::~Ofto.lt":
C'TIInlf:C'
"'U.In
~r"\ITII"'to"- 1
r~rn.n~o ..
-. ~
::,;,:_~.
;o'i~::ii~~~:
'
..
~.
-"'t<:i -~'1!..">!1
<
-?r-"i""'
.'j
~-
Study Method
Dmu
Study
Data Item
Point
Ride
Chk
Chk
Trail Car
Survey
AVL
Office
Observation
TCQ.SM Senice
Measures
Service frequency
Hours of service
,;
,;
.J
..J
.J
..J
..J
,;
..J
..J
..J
..J
.J
.J
A=., coverage
.J
Passenger load
Rdiability
,;
,;
..J
Othec Measures
Customer Satisfaction
..J
.J
,;
Missed Trips
Freq. of Mechanical
Brcakdown.s
.J
..J
..J
..J
..J
"
..J
Wait Assessment
..J
..J
.J
.J
.J
. ..f
..J
..J
..J
Passenger miles
..J
.J
Pedestrian Crossing
Difliculty
..J
' .J
.J
Transit-Auto Travel
Tune Ratio"
Boarding and
Alighting (by fare
category)
..J
.J
..J
Pedestrian Alx.ess
..J
Bicycle A=
.J
Parle. &
Ride~
..J
Disabilities
'
Passcoger Loading
Ease
Transit Stop Amenities
..J
"
..J
These performance measures represent a sample of the most commonly used transit attributes. TCRP &port 88 (Kitte~on, 20Q3) idenrilies additional transit service attributes that are sometimes used by agencies as shown in Exhibit 13-4.
Source: TCR.P Report 47: A HAndbook for Measuring CustQmer Stuisfaction and Qflality, 1999.
It is a challenge for an agency to select a subset of these attributes to use for evaluation of their transit service. Oearly,
the ones required by the NTD for reporting are critical. Measures used in rhe TCQSM to define transit LOS are useful
and may be required by region. But other measures may be of particular interest to a transit agency. Fmally, measures
describing the accessibility of transit systems to persons with disabilities are useful to ensure the service is in compliance with the Americans with Disabilities Act (ADA) legislation.
One approach for selecting additional transit service attributes is to survey the users of rhe system about their perceived preferences. TCRP &port 47 (TRB, 1999) presenrs a methodology to assign an impaa scOt? to each attribute
based on resulrs of a customer sat:isfaction survey. The survey principally asks respondents to rate transit performance
for a range ofattribute$ and also asks whether a person experienced a problem with each attribute during the trip. The
approach rhen calculates a gap sco", which is defined as the difference between the average attribute rating for those
who experienced a problem and for those who did nor.
While the impact score approach can be helpful, it also requires signilicant (survey) data collection prior to collecting
other items. In a more general guidance, rhe selection of a data item for study is a function of the srudy objective and
in som~ cases may be deterrp.ined by a specific complaint or issue with the service. Ifseveral data items are relevant to
270 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDI110N
! the
problem, it may be possible co collect them simultaneously. Analysts must take the rime co understand the prob\lem thoroughly, however, or they may overlook a relevant data item. For example, analysts evaluating adve rtisements
:for public transporration services will usually collect hoardings, and may be interested in hoardings by fare catego ry.
~ measures of behavior. Since advertising usually has the objective of changing attitudes and awareness as wdl as behavior, analysts should also collect data on passenger attitudes and awareness in chis case. Data items selected for srudy
must be closely related to the problem or to the objectives of the program being evaluated.
Whichever performance measures-are selected by an agency, it is critical that recording of these measures is done accurately and consistently to be able to reliably track performance over time. Iris useful for agencies to develop their own
data collection standards and best practices affecting specific data elements or certain aspects of their tra nsit service-
vals. Data required by NTD need to be updated annually, but agencies may wish to collect these data more fr~uendY
To reduce the need for ad hoc" data collection e.lforts, some agencies develop continuous data collection prograr11s.
These programs often consist of a baseline phase and a monitoring phase. The baseline phase provides a comprehensive one-time snapshot" of system operations. Data items collected simultaneously during the baseline phase provide all the data necessary for long-term planning and other infrequent major efforts, provide conversion factors for
auxiliary items and provide the beginning points for trend analyses. During the monitoring phase, key data it~ms are
updated as needed, using conversion factors whenever possible. During the monitoring phase, operators should alsO
make periodic checks co reveal problems that require attention.
Trail Car and ObstrVaMn: Are sufficient numbers of observers and uail vehicles available? Can data items
be combined to ma1te most efficient use of resources?
Office: Is the software and expertise available in-house? Are funds available to outsource parts of the
analysis?
If the resources are not available to conduct the study using a particular method, analysts must explore alternativ-e
methods, change the data item to be collected, or abandon the scudy.
Measurement errors occur with every study method. For driver studies, increasing passenger loads and heavy uaflic
are major factors in causing error. For point checks, many standees and vehicles with tinted windows lead to measurement error. For ride checks, simultaneous boarding and alighting, two or more doors in use, a multitude of fare
categories and crowded vehicles cause error. Automatic data collection eqwpmem can fail systematically (due to large
crowds, passengers with large packages blocking sensors, etc.) or can fail fiom mechanical or electrical problems. Finally, surveys are extremely sensitive in the choice ofsample, format, wording ofquestions and many other ways, as
described in Appendix B. When considering a study method, analysts must consider any problem particular to that
method and the data item needed. For a boarding count, which is the most common type of public ttansporracion
srudy, automatic data collection eqwpmenr will generally provide the most accurate data. The ride check method
will be close to automatic data collection equipment in accuracy, and point check data will be less accurate (UMTA,
1985b. Deibel and Zumwalt, 1984).
Cost is more important than measurement error in choosing a study method. This is because all the methods discussed provide reasonably accurate data and because increased sample sizes can reduce the effects of many measurement errors. The choice of a srudy method is easy when costs are considered, since driver srudies are generally the least
expensive srudy method and point checks are the next le;m expensive. Asuggesred prioritization of study methods is:
1. automatic data collection systems, if available
The following discussion is divided into ~anual methods (driver study, point checks, ride checks and observational
srudies), a.utomatic methods (wing APC and AVL technologies), surveys and office srudies for planning applications.
Having drivers count boarding passengers or bouding passengers by fue category is a very efficient and :~.ccurate study
method. The method is efficient because no labor coscs for checkers ue incurred. Driver counts ue a.ccurate becawe
they ace kept simple, becawe drivers are monitoring boarding passengers anyway and because drivers ace Wn.il.iac with
the vuiow fare categories being recorded. Driver studies ue impossible on unmanned vehicles and when p:.I.SSCngers
can board the vehicle tluough more than one door. In these cases, transit agencies must we other study methods.
The driver cao keep the count on a mechanical counter or on a counting system integrated into an dectronic fare box.
Usually the count boa,rd presents no more than six keys to the driverso some fare categories may have to be recorded
together. The driver oould write the total count or the count per fare category on a summary form, like that given in
Exhibit 13-6, at the end of each trip, or the driver could enter the code for end of trip" into an electronic count board..
Appendix E provides blank data. collection forms suitable for copying.
Statistics produced from driver srudies include the mean number of passengers per ttip and the proportions of passengers by fare category. The analysis of statistics produced in public transportation studies is di.scwsed later in the
chapter.
~~~~~:~~1}~~~!::~~~~~~-:-;l~~---h:~;;~tJ~~~~~~t4:~~~JY!f~~~
Boarding Count field Sbeet
Block Number
Route
Date
Day
Weather
O bserver
Route Segment
From
Boarding Passengers
To
Foil Fare
Reduced
Fatt
Transfer
Full +
Transfer
Reduced
+ Transfer All Passes
Point checks primarily provide load co unt:; and schedule adherence. Exhibit 13-7 provides a convenient data form
for these primary items. Checkers will need an accurate warch ro record schedule adherence. Secondary purposes of
est.
:;;:
.;
--~,.,.,
... .,.
.. .
-~
.. ~~~~W~'"f~-:~~~~!~~'*-~~i;~~~-.~~
Point Check Field Sheet
Route(s)
Day
0 Departing Load
Weather
Observer
0 Arriving Load
Block
Vehicle
Route
Number Direaion Number Capacity
Arriving Tune
Scheduled
Actual
Passens ers
-During most point checks, the checker estjmates the load while standing ouuide the vehicle. An efficient technique
is to count the passengers if the load is light, count empty seats if the load is moderate and count standees if the Jo:<ld
is heavy. Checkers should know. the searing and standing capacities of all vehicles subject to data collection. Checkers
may have to board the vehicle for a quick count if timed glass, heavy loads, obstructed views, or other factors preclude
a count from the roadside. Drivers should be made aware that a srudy ~ under way if boarding is anticipated. Load
counts are difficult to conduct for both directions on a wide street. However, recording schedule adherence on bot:h
directions of a street is nor a problem at any point where rhe view is not obstructed.
Operators usually conduct point checks at the peak load point of a route (that is, the point where the maximum n~
ber of passengers is expected). Other key points along a route may be of interest, such as transfer points and poi.o.1:5
where vehicles may rurn around short of the end of rhe route. It is very efficient to conduct a point check where sever :;;a.!
routes overlap, even if that point does not happen to be a peak load or key p~inr for each route. Sometimes, ope~
tors conduct several poinr checks simultaneously on one route. Furth (1989) provides 'a methodology for esrimatiO$
the passenger origin and destination pattern along a route from several simultaneous point checks and previow ricJ.e
check or survey data.
Office per~nnel usually enter data into a computer manually from P<>im check forms. However, in some cases pc:rsonal data assistanu or some types ofsmart phones" may be custom.iud for data collection and allow for time savin~ 5
in data processing. Optical scanning may be useful for other surveys as well.
Standard point check data provide several useful plors and statistics. Exhibits 13-8 and 13-9 show cwo of the mor.e
useful plots with load data and with schedule adherence and load data, respectively. The vertical axis in Exhibit 13-B
is the loadfoetor in percentage of seating capacity, rather chan the acrual number of passengers, co remove chc dfect:5
of different sizes of vehicles. In Exhibit 13-9, both passenger load and schedule adherence are shown on the saJP e
graph. Later in this chapter the analysis of relevant sta.tistics, such as 'rhe mean load and the proportion of vehicles o~
schedule, is d.iscussed.
I:
~:'-1 II ~
Pc&ylDad
t; : ::::::::
~
OJ{.
ll
VeHde Nu'ri;)er
A.M. P.ak
I
I
I
I
I
I
I
.I
I
Pa. . .nor
toad-
I
I
I
I
I
I
I
I
I
I
I
I
I
II
:' Il . II
I
I
II
I
I
I
I
I
I
. I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
II III .
II
I
I
I
I
I
7:00
7:10
7:ZO
7:30
7:.c>
7il0
8:.10
Tl""'
a
Ii
i
:00
ll:20
8;30
0
e:.c>
I
I
I
I
I
I
I
I
I
I
I
I
II
II
I
I
I
I
1
I
I
e:ao
eili<
8ctledu1ed amval
Actuar am..,.l
0 :
\.,.
Ride chcc.ks are Ba.ible and provide seven! useful dau. items. Ride checks most ofttn provide boarding and alighting volume~ the vehicle arrival time at each stop and at other key points. Exhibit 13-10 shows a fonn suitable for recording
771: '" Ml1NIIo11
<Ti U'\IC<
-..u-. <"OTt"'
~~~er
Day
Date - - - - - - - - - - --
Observer - - - - - - - - -- - - - -- - -
Scheduled SwtTune - - - - - - - - - -- -
Location
On
Off
Load
Tune Check
these basic items. A checker must wear an accurate watch to record arrival times. Progr.uns are available that allow data
to be recorded on a handheld or portable computer and later tranSferred to a personal computer for analysis.
Ride checks sometimes provide more detailed data on vehicle movements than simply arrival time at stops. Ride
checkers can also record deceleration times, dwell times at stops, acceleration times and dwations and causes of delays.
Exhibit 13-11 shows a form for these speed and dday items. If they are busy, checkers df) not need to compute column
7 for dday time in the field. Colwnn 2 for the time at conuol points allows schedule adherence and various speed
measwc:s to be calculated. A ride checker using a form such as the one shown in Exhibit 13-11 can carely collect data
on passenger boarding and alighting at the same time.
.,..,.,
""!
"~::~ ~
D::l6i'?l
o.
'""''~'
-. ;~;7~ ., . ',: ' ~''J
....
'' ::..u.;:
:..:;;<':~~
~- ~i;ltk~;:>.;;;~:,4.~'c.;
s .- . '
. ...
__ ,.
"
-_
':1-~~-
. . .
)'~~
Date
Route
Direcrion ofTrip
Weather
Vehicle Type
Observer
Method
Trip Number
(2)
Location
Tune at
Control
Point
(3)
(5)
(7)
(6)
Tune
Time
Jlowerthan
faster than
walking
wal.ldng
speed
StopTwe
speed
Delay Cause
(5)- (3)
Delay time
(sec)
Ride checks can also provide many of the data items that driver studies and point checks provide. Checkers can count
hoardings by face category during a ride check. Load counts and boarding counts per trip can be made direccly during
a ride check or can easily be derived from a ride checker's boarding- and alighting-by-stop data. Ride checkers usually
occupy the seat direccly behind the driver. Boarding and alighting from the front door, rhe causes of delay and the rype
of face paid are all best observed from that sear. Boarding and/or alighting through a rear door, with m:my standees,
may require a change in position or may mean the checker stands to conduce the counts. In rare instances, more than
one ride checker may be needed on a vehicle. Checkers must be familiar wirh the causes of delay or the possible fare
categories, depending on the data items being collecred.
Ride check data are analyzed in sevetal ways. Data are mosr commonly displayed in a simple bar chan showing boarding
and alighting by stop. The bar graph can further be modified to vary the width of the bars by the distance between scops.
Ic i.s also common with boarding- and alighting-by-stop data to calculate the passenger miles per trip. The p=ger load
lxtween twO stops is multiplied by the distance between the scops; the number ofpassenger miles per trip is then the sum
of these quantities for the entire route. Operators examine tranSit vehicle time and delay data by producing a space and time
plot as in Exhibit 13-12 or a bar graph with magnitudes ofdelay by type as in Exhibit 13-13. Useful statistics from ride check
data include the mean number of passengen boarding at stop x, mean p=nger miles per trip, mean delay ar sigJul y, mean
ovu.ill ttavd speed and the proportion of delay due to ruming vehicles.ln a Iacer section the analysis of typical statistics from
public transportation studies will be discussed. With new visualization techniques, analysts have been developing new and
innovative means of displaying transit data. Exhibit 13-14 shows an example where transit load, boarding and alighting and
searing capacity are all displayed on a spatial plot of the route in Google Earth.
~.
..
..
Each of the four ride check methods of determining origin a.nd dcscina.tion patterns by stop on a. route bas adva.nta.ges
a.nd cl.isa.dva.n~es. The first twO involve more direct measurement, so they a.rc generally more a.ccurate but more apensive tha.n methods three a.nd four. Method one is appropriate only on trips with no sca.ndees and limited seating
turnover. Method two bas a. high response rate but may ca.use initial confusion among passengers. The checker must
be able to c:xplain quickly. possibly in more tha.n one la.ngua.ge, the purpose of the card a.nd wba.t the passenger is to
do with it (Simon and Funh, 1985). Instructions can a.lso be printed directly on the card, but it shouldn't be assumed
all riders can qUickly read a.nd understa.nd printed (English) material. If a.na.lysu can tolerate somewha.t less accuracy
280 MANUAl OF TRANSPORTAnON ENGINEERING STUDIES. 2ND EDffiON
and have the needed ride check or point check data on hand, methods three and four arc the best ways to estimate
o1gin and destination patterns.
Operators also usc surveys to estimate origin and destination patterns, particularly when they need uue" doo r-to
door origins and destinations. On-board surveys arc discussed in Section 3.3, and general survey methods are described in greater detail in Appendix B.
3.1.4 Trail Car
A Checker in an auwm~bile trailing the transit vehicle of interest can also collect the speed and delay items on the
form shown abcve in Exhibit 13-11 for the ride check. The trail car method is advantageous be<;ause the checker is
(perhaps) out of sight of the transit vehicle driver, so no bias is introduced by changed driver habits. However, the trail
car method is more expensive than the ride check method because it always requires an auwmobile and under heavy.
traffic conditions may require a second checker. The trail car method also docs not allow a full view of the causes of
ddays. Because the disadvantages of the trail car method rudy outweigh the advantages, operators use the ride check
method more often for speed and dday data items. With the advances in AVL equipment discussed in the next section, the mil car med_lod ~ rudy liSM in practice.
While the cost of i.nsta1lacion is significant, oftentimes APC and AVL technologies arc installed as part of a broader
ITS deployment in a city, which reduces the relative cost of APC technology. Usc of both technologies requires good
communication among municipal departmentS, because the technological infrastructure spans across jurisdictions.
TCRP Synth~is 77 (Boyle, 2008) provides a more detailed overview of advantages and disadvantages of these technologies, as well as guidance for good practices and lessons learned with installation.
3.2.1 AuUima#c Pasmrger Counters
.
APC equipment includes data acquisition, recording, transfer and reporting componenu. In most installations, a
series of photoelectric beams near the door or pressure-sensitive pads in the stairs provides boarding and alighting
counu by stop. Computer algoritluns decipher the activity recorded by the sensors and decide, for aarnplc:, whether a
sequence of signals represents a boarding or alighting passenger. The equipment records tra'ld distances (through the
vehicle odometer) and times to index the countS and ro provide schedji.(c adherence and similar data items.
Agencies may usc APC technology to obtain both ridership and travd time data along routes. The main advantage
of APC systems is that data are available in a permanent form allowing the transit agency to track performance on a
continuous basis. The primary benefits of APCs cited in a recent survey of transit agencies (Boyle, 2008) were data
disaggrcgated at the srop. segment and trip lcvds: better quality ridership data; availability of running time data to
adjust schedules; and a better basis for decision-making. The main problems with APC systems included issues with
reporting software. data. processing and analysis, data validation and h;udware problems. Approximately onequarter
of all respondents reported either no problems or only minor startup issues (Boyle, 2008).
3.2.2~J Ko.bi& lACAIWft
With increasing availability and cost-dfectiven~ of GPS technology, AVL insall.ations arc being used more commonly by transit agencies. Through on-board GPS installations, an agency can automatically gather a lot of data including uavd times, schedule adherence and ddays. With a continuous data stream the agency can readily implement
a performance monitoring system and can identify common sources of recurring congestion (for example, a particular
intersection or a particular time interval). An agency pn usc these data to improve upon the service provided and
enhance the qansit experience for customers.
AVL dara can also d irecdy benefi t customers through t:he use of real-rime transit data, disuibured to rhe ridership
through displays at transit srarions, Web sites and even integrated in applications for cellular telephones. The technology investment therefore serves the combined need of collecting data and providing a service to customers. Exhlbir
13-15 shows an example of an online route visualization system for a major university. Exhibit 13-16 shows a snapshot
of data (five trips) ob r.Uned from one of the transit routes indicating what segments of rhe route are most correlated
wirh travel rime vuiabiliry.
O r----~-----~----.-~--------.-~------~.-~-------,------~
--~r-----------------------------------------------------~----~R~oundabout
0 .5
Gym
C heck 2
::E 1. 5
Che ck 3
2.5
0
2000
4000
T ime (sec)
eooo
8000
Automatic data collection equipment is expensive, but ir can be inscalled on only a representative fraction of the vehicles in a fleet. Deibel and Zumwalr (1984) concluded a useful, valid, reliable database can be main tained if :tboUl
10 percent of the vehicle fleer is equipped for automatic data collection. Scheduling an equipped vehicle to a trip on
which data are to be collected is challenging due to a variety of problems, including vehicle servicing a nd incompatible vehicle sizes. For example, operators cannot schedule a data coUeceion vehicle with a capacity of 40 passengers
on a route char nee9-l 60-passenger vehicles. However, equipping I 0 percent of the fleet providcl an adequate mugin
for most of these problems." If the objectives of technology installation include public visualization of transir data. as
discussed above, mosr ifnot all of che vehlcles need to be equipped with AVL technology.
Automatic data collection equipment produces much more raw data than manual study merhods. The data processing and report generating capabilities of the system must be sufficient to handle this load. With lase microcomputers
and other advances in computer technology, though, this is more an administrative than a technical challenge. The
reports themselves are no different from ch.e types of reports produced with rrianual methods and described elsewhere
iil this chapter.
Data collected automatically aie usually more accurare than data collected manually, but bias can be a problem. ALlto(llacic systems tend to undercounr for several reasons, including mechanical or eleccrical failures, environmen~ factors and passengers blocking sensors. If bias is suspecred, the operators can compare the automatic count to acarefu.].,
trusted manual count. If the operators .find a bias, they can repair the system or correct for the bias machernaricallY
during data processing.
Beciuse the costs ofAPC inscallations vary so widely, it is impossible to State exaccly which public transportation ageocies would fi.nd the equipment cosr-dfeccive and which would not. However, experience and cost analyses indicate
for larger agencies (over 400 vehicles) the equipment will usually be i:ost-effecrive
for midsized agencies (100 to 400 vehicles) the equipment may be cosc-effeccive depending on data needs
for small agencies (under 100 vehlcles) the equipment will wuaUy not be cosc..effeccive
When the availability of funds, the data accuracy needed, the data rurnaround time and other faceors are considered
with the cost-effectiveness, deciding whether to invest in aucomaric data collection equipment becomes very compl~
ingly popular. Online surveys rely on riders completing the survey after completing the .trip, once they have access too
Public Transportation Studies 283
a computer (although many may be able to access the Web while riding via laptops and smart phones). The resulting
response rates may therefore be lower than traditional in-vehicle surveys, buc the cost for administering the survey and
printing material is much reduced. Furthermore, much of the survey analysis is automated &om online database tools.
A survey done via text messaging uses readily available cdl phone technology to vote" on a specific transit survey
question, much like the viewers of variow tdcvision shows and polls.
Analysts gencrally we two types of traditional" on-board surveys: hand-in and mail-in. For both survey types, survey
workers distribute the forms with a quick word of explanation to passengers upon boarding. Hand-in surveys are collected &om passengers as they alight. Hand-in surveys arc short, consisting of only a few questions, because there is
limited time for passengers to complete them. Passengers answer hand-in surveys c:xclusivdy on the vehicle, so only
information passengers remember readily can be asked for. Hand-in surveys are printed on a card that can be written
on and are distributed with pencils.
Passengers may complete mail-in surveys on the vehicle or elsewhere. Thus they do not need; to be as short as hand-in
surveys. They can pose more difficult questions, since passengers have time to ponder questions and seek assistance for
answers. Mail-in surveys arc printed on ordinary paper and do not have to be dist.ributed with pencils. The forms can
be stamped and addressed on the reverse side, or a stamped and addressed envelope can be provided with each form .
The choice of survey type depends primarily on the number of questions the operator must ask. Shorter surveys arc
generally berter, but on-board surveys can be longer than postcard size if the questions provide needed data items.
Response rare is another factor that in.fiuences the choice betwttn hand-in and mail-in surveys. Hand-in surveys have
higher response rates, which could approach 90 percent usable returns, while mail-in surveys generally achieve 30-60
percent usable responses. However, a high response rate does not necessarily mean an unbiased survey {Brog and Meyburg, 1981, Doxscy, 1983). Fmally, if passengers are on board for a very short t(ip or if there is a signi6caht number
ofstandees, scv~ bi;~ses will occur with hand-in surveys.
An intriguing option that combines the best features of hand-in and mail-in surveys is the two-part suivey (Stopher,
1985). A two-pact survey consists of a short hand-in form and a longer mail-in form that are handed out together.
The hand-in form asks for very basic informacion on passeriger characteristics and should have a high return rate. The
mail-in form asks the detailed questions that arc truly of interest to the survey sponsor as well as the same questions
askai in the hand-in form. The amlysr can compare the characteristics of passengers who rerumed both parts to the
characteristics of the passengers who only returned the hand-in pact, and will thereby gain insight into biases in the
sample that answered the important questions.
Bamford et al. (1984) described a useful innovation for hand-in surveys. Instead of distributing pencils with a hand-in
survey, they created a survey form on which passengers could rub off responses with a fingernail or coin, similar to the
instant lottery tickets sold in some states. They reported a high response rate and that the eight-question survey took
only 30 to 60 sec to complete. The main disadvantage they noted was the need to prccode all possible answers, so a
survey with detailed questions could not use the format.
Stopher (1985) offers several exceUent suggestions for conducting on-board surveys. First, he recommends a count be
maintained of the number of survey forms distributed for each trip. This count is essential for measuring the response
rate. Second, Stopher argues drivers should never be asked to distribute survey forms. Distribution of forms by drivers
may bias questions about the driving. may not allow the control of forms noted above to be maintained and may not
provide passengers with the assistance they need to complete the survey. Next, he states that separate forms should
be available for each predominant language in the survey area. A quick greeting by the checker distributing survey
forms should be enough to indicare which language a particular passenger can respond in. Stopher also advises boxes
should be placed at each exit door for the return of forms, with large signs posted nearby to draw actention to them.
The chccket distributing survey forms may be busy with boarding passengers and may not be available to coUect every completed form. Finally, Stopher urges every fare-paying passenger be given a survey form, since there are many
problems with sampling a subset of passengers on a vehicle.
4 .1 Sampling
A sample is necessary for almost every public tl:lnsportation study because the cost of a complete census is pro hibitive.
Public transportation agencies face increasing financial pressure, especially for expenditures perceived as nonessential,
such as data collection, and cannot tolerate waste. Forrun.ately. the ability to generaliz.c: about an entire population
from a sample allows useful data to be collected and analyz,ed for a reasonable cost.
..
After determining !:he study methOd to be used, the analyst conducting a public transportation study must determine
the sampling unit. The sunpling unit is the entity tha.t represents one data point. For most publ.ic rransportacion scudies, the analyst chooses me trip-one vehicle in revenue service traveling from the start to the end of a route-as the
sampling unit. For example, in determining whether hoardings have risen due to the introduction of a new fare pass,
driver scudies are to be conducted. Each driver participating in the study reports the number of passengers on each
trip. From these data the analyst can calculate mean boudings per trip. In tum, multiplying this mean by the total
number of trips will produee the desired quantity for comparison.
Once a sampling unit is chosen, the analyst devises a sunpling plan. Histotically, analysts have used simple random
sunples in most public tr:~.nsportation studies. With a simple random sunplc, each unit in the entire population of
units has an equal chance ofbeing selected for measurement. The sample size and analysis formulas for simple random
samples are straighcforwud. However, the plan sometimes causes inefficient use of resources. A simple random sample
of trips measured by ride checks, for example, may mean that check~ will be scrunbling aU over the route necwo.rk
to be on their assigned trips. A simple random sample of passcngeis measured using an on-board survey may mean
distributing survey forms to only a few passengers on each vehicle.
For more efficiency, analysts o.n conduct many studies using stratified or multistage sampling plans. Strarifod plans
require the analyst to divide the population of units into several identi6.:~-ble groups and then sample randomly from
within those groups. For example, analysts could distinguish radial routes from crosstown routes and could sample
a ips from those groups. Another common stratification is by time of day: One could draw separately from A.M. peak,
P.M. peak, between-.(leak, evening and overnight "owl" services.
MultistJJg~ puns are more common and usually more efficient than stratified plans. Most multistage plans require the
random selection of.
<t~ l nr
..
,.c
into the data collection process, assure a more accurate estimate of the statistic ofinrerest and can result in a reduced
sample siz.e. The sample siz.e and analysis formulas for suatified and multistage sampling plans ate more complex than
for simple random samples. Refer tO Cochran (1977) and other sampling texu for details. Analysts contemplating
stratified or mulrisrage sampling plans should seek advice from a professional statistician.
(t~)
Equation 13-1
where
t
Analysu can conservativdy assume values fort as 1.8 with 90 percent confidence and 2.2 with 95 percent confidence,
based on Urban Mass Transponation Association (UMTA) (1985). A tolerance, d, of 0.15 with a 90 percent confidence level means that 90 percent of samples of the size n will have mean values that fall within 15 percent of the estimated mean. To apply Equation 13-1 and to estimate a sample size, advanced knowledge of the coefficient of variation
is necessary. Exhibit 13-17 therefore prov]des codlicients of vamtion (v) for common data items if estimates are nor
available locally. Tolerances recommended by UMTA (1985) range from 10-30 percent for most common data items.
Load
Boardings, passenger-miles
RWlning time
"Owl (late night) default values are the same for weekdays and weekends.
Source: Urban Mass Transportation Association, 1985b.
Equation t3-t above is use& for data that are in the form of a simpje random sample with a sample mean andsom.~
variation about that mean. However, many data items collected in public transportation studies are proportions, such
as the proportion of passengers paying special student fares and the proportion of trips more than 5 min. behind
schedule. The sample size, n, for proportion data using a simple random sample, can be estimate using
n-
t 1 p(l- p)
dZ
Equation 13-:Z
where
t
Values of in use with Equation 13-2 are approximately 3.0 for 90 percent confidence and 4.0 for 95 percent confi..dence. If no prior estimate of pis available, analysts can use a conservative value ofp =0.5. In public uansporncion&srudies, a cominon value of tolerance for estimating proportions is 0.1. for values of p near zero or 1, analystS oeecf-
asymmetric tolerance ranges (see UMTA, 1985).
Public Transportation Studies 287
""
Equation 13-3
where the variables are as defined for Equation 13-1. One can compute the coefficient of variation, v, for.Equatlon
13-3 from the sample (that is, sample sta.n<Wd deviation divided by the sample mean). Values for t depend on the
confidence level and the sample size. For 90 and 95 percent confidence levels and large samples (more than 30 units),
analym should usc t values of 1.7 and 2.0, respectively. MOSt statistic tc:xts contain tables of t-<iistribution values for
smaller samples and ocher levels of confidence.
Equation 13-4
where the variables are as defined for Equation 13-2. The sample provides the value ofp, and the value oft is as described for Equation 13-3. Presenting estimated means or proportions and the tolerances on those estimates graphically is usually of great assisrance to decision makers viewing the study results.
5.0 REFERENCES
5.1 Literature References
Bamford, C. G., R. J. Carrick and R. MacDonald. "Public Transport Surveys: A New Effective Technique of Data Collection."
Trtrjfic Enginuring and ConiTOINo. 6 Ounc, 1984): 318 ct seq.
Boyle, D. Passmgtr Counting SJ11nN-.A. S]nthais ofPnzaUt. TCRP Synthesis 77. Washington, DC: Transportation Rc:search
Board, 2008.
Brag. W. and A. H. Meyburg. "Coruidcration ofNoruespoo.sc EffectS in Large-Scale Mobility Surveys" Tranrportlllion Rnearrh
&cord: journal ofthe Trrmsport:aJion RntarGh Botm/807 (1981).
Career, R. T. ~ .A.uvnmar and Quaiity ConiTOI GuiJelma. Washingroo, DC: Feckral Transit Adminisuation, 1987.
Cochran, W. G. Sampling Ttclmiques, 3ni ed. New York: Wiley, 1977.
Deibel. L. E. and B. Zumwalt Modular App~h 111 On-BoarrJ AutonltttU Dllla Colition ~ National Cooperative Transit
Rcscarcli and ~opmcnt Program Report 9. Washingron. DC: Traruportation Research Board, 1984.
Doxscy, L Rnpondmt Trip Frrqumcy Bias in On-Board Surveys. Transport4tion Researrh &cord: Journal ofthe Tramportuion
: Rnearrh &art/944 (1983).
Furth, P. G. Updating RUU Chedu with Mulrip/4 Point Chec!ts. Transportation Rnearch &cord: journal ofthe Transportation
Rnearch Board 1209 (1989).
Kittelson and Associucs, Inc. A Guitkboolt for Dtvtloping a TT111Uit Pirfonnanu Me~ZSUmnmt Systnn. TCRF &port 88.
Washington, DC: Transp0rt4tion.Rnearrh Board, 2003.
l.ede, N., L Yu, C. A. Lewis and K. Godui. A Manwzlfor Evaluating Pmonaliud Public Trzznsit SyJtnns. Dallas, TX: Texas
Southern Univcnity, Federal Transit Admini.stntion, 1998.
Schwenk, J. Ewzillation Guitk/ina for Bus RApid Transit Dnnonstration Program. Washington, DC: Federal Tr.uuit
Administralion, 2002.
Simon, J.. and P. G. Furth. "Generating a Bus Route 0 -D marrix from On-<:~ff Data.. ASCE Journal ofTrzznsportalion
Engineering 111, No. 6 (November 1985).
Sropher, P. R. "The Design and .Execution of On-boanl Bus Surveys: Some Case Studies. New Surwy Methods in Trrvuport.
Furth, P. M. "Designing Automated Vehicle Location Systems for Atchived Data Analysis." Trtzruportation Rntarrh &cord:
journal ofuu Trzznsportation Rnearrh &ani 1887 (2002)._
Golani, H. "Use of Archived Bus Location, Dispatch, and Ridership Dau for Transit Analysis." Tramportation &uarch &cord:
journal oftlu Tr1Z11lportation Board 1992 (2002).
Kimpel, "f. Data Vi111ahzation aJ a Toolfor Improved Decision-Malcing within Transi& Agencies. Seattle, WA: University of
Washington, 2007.
Kimpel, T. S. "Automatic Passenger Counter Evaluation: Implications for Nacional Transit Dacabase Reporting." Transportation
&uarr:h &cord journal ofthe Tranrportation &search Board 1835 (2003).
Perk, V. and N. Kamp. Handboolt ofAutomated Data Collection Mtthods for the National Transit Database. Tampa, FL: National
Center for Tr:msit Research at the University of South Florida, 2003.
Robenson, H. D. Manual ofTranrportation Enginming Srudm. Washington, DC: lnstirute ofTransporracion Engineers, 1994.
Urban Mass Transportation Administration. &vemu Based Sampling Proudum for Obtaining Fixed Route Bus Operating Dat4
&quirtd uiukr the Section 15 Reporting Synnn, Circular UMTA-C-2710.4. Washington, DC: Urban Mass Transporracion
Adminisuuion, 1985.
Chapter 14
........................... .............................................. . ..........
INTRODUCTION
291
291
293
298
299
302
304
3.0 REFERENCES
. 3.1 Literature References
305
305
307
307
1.0 INTRODUCTION
oods movement, the distribution ofraw materials and finished produCtS, is most frequently handled by cruck, train
and :Urplane. Some conunod.ities move by ship or pipeline, but these modes tend to be less visible to the ~e
cran.spornuion planner. Incorporating &eight planning into general transportation planning activities is bco>rniPg
increasingly important as the regional and 10<:21 road necwod<s becom~ more congested, funds to apand physical apaci.q
continue to dc:crea.sc:, air quality concerns heighten and healthy economic activity is essential. Goods movemenr is an enormous enterprise. In 2007, frnght movement among all modes transported almost $12 trillion worth of cargo, with mu $8
trillion by truek (RITA, 2007). Approx:imatdy 120,000 establishments were involved in truck mnspocation c:mployit'1g
over 1.5 million people (U.S. Census Bweau, 2007).
Many views are represented with respect to freight movement: those ofshlpp~ and carriCI'S, public agencies such as ckpa.-ecments of cranspornuion at alllevds, plann~ and c:ngino:rs, public officials, citizens and the general motoring public. '[b.e
role each plays when ~y study is being conducted will inspire the study design and therefore the results. The question of
whether one is looking for short-term, praccical solutions to ccisting problems or projections for long-rerm planningpu..1"
poses will also influence the types of srudies conducted. lnvestigacioos of.&eight movement may lead to changes in o~ra.-ring practices, wning or devdopment regulations, design scand.ards, fo:s, tariffs, or caxes. It may also influence the timing o f
public or private projects and the dollars expended. In this chapter we address the process of coileaing d:ua. the impaas o f
route characrcristics,loadinglunloading studies, vehicle weight studies and hazardous ma.ce.rials movement.
mitigation options or policy implementation. An examination as part ofNCHRP Report 320 described the following
commwtity issues from the various freight modes (Suauss-Weider, 2003).
Trucking community issues include:
congestion generated on local roads, highways and at customer facilities;
large tractot-uailers making deliveries ro customer facilities-inslifficient loading dock space, leading to
double parking and meet congestion;
movement of heavier trucks on roadways adversely affecting automobile speeds;
damage caused to pavement, especially from heavier trucks and more frequent truck movements on local
roads;
hazardous materials spills and accidents caused by truck movements;
collisions involving trucks;
diesel emissions (impact on ait quality) derived from truck operations;
truck hours of operation affecting peak period traffic flows;
noise and vibrations generated by trucks;
potentially negative impaccs on property values from truck activity;
lack of available truck parking and rest stops resulting in trucks parking on shoulders and along roads,
causing potential safety concerns;
light and noise pollution generated by nighttime operations at loading docks and truck terminals;
potential new development on cxlsting truck terminal properties;
inadequate truck access to matitime and air cargo terminals affecting the competitiveness of these facilities;
and
on adequate road geometries, turning radii, grades and turning lanes to accommodate trucks.
1'\.- Tn A ltrnr\nTITI,..
.,_. , , . , . , , . . , . , ,
.,_, - -
Access co waterfront-access to the waterfront for nonfreight uses is increasingly desirable, particula rly in
wban areas. This creates competition for land in these areas.
Air cargo community issues inClude:
Hours of operation and noise-Air cargo operations rend to occur during nighttime hours. A:; a result,
noise issues are more pronounced.
Truck traffic on access roads--Similar to maritime cargo facilities, the volume of truck traffic typically
increases as cargo activity grows at an airport.
Theft and security-Criminal and terrorist activity can occur at air cargo operations.
tr~~::~i~iilt?F-i.'.r~:t~idt~'!~~,,~~J'.-~~:\~~t"
DataSet
Description
Commodity movements among states and metropolitan areas by value, weight and
mode for 1997,2002, and forecastS for 2010 through 2035.
www.ops.fhwa.dot.gov/freightlfreighc_analysis/faf
North American merchandise trade data by commodity and mode. Geographic detail
for U.S. exports to and importS from Canada and Mexico. www.brs.gov/transborder
Number of trucks, uuck containers, train and rail concaine11 crossing imo the U.S.
through land ports on U.S.-Canadian and U.S.-Mexican borders.
www.bts.gov/programs/internarional/border_crossing_cntry_data
TonJUge and uips by commodity for major poru/waruways and origin and destination
data on warerborne cargo movements by waterways and harbors.
www.ndc.iwr.usace.army.mill/wcsc/wcsc.hun#2008%20Waterborne%20
Commerce%20of%20tbe%20Unired%20Staces%20%28WCUS%29
Value and weight of cargo by type of service for U.S. waterborne imports and expom.
www.iwr.usace.army.mill nddusforeign
Oil Pipeline
Air tnflic, connage and revenue ton-miles data for large air carriers and by airport.
www.bcs.gov/prograrn.slairline_information
ln&astructw"C
National Transportation Adas
Database (NTAD)
Railroad-Highway Grade Crossings Grade crossing location and safety data. Other rail network characteristics in NTAD.
hrrp://safetydata.fra.dot.gov/Officeof5afety/
U.S. Pores and Waterway Facilities
Database
Physical characteristics of coastal, Great Lakes and inland U.S. portS, terminals and
locks. www.iwr.USi.ce.army.mil/ndc
Fteight Vehicles
FAF Highway Capacity Database
Truck Bows at highway segment level for 1998 and forecasts for 2010 and 2020.
www.ops.fhwa.dor.gov/freightl&eighr_analysis/faf
VTRIS-W
Number of trucks weighed and vehicle weight information by type of vehicle and
highway functional class. www.fhwa.dot.gov/ohim/ohimvtis.cfm
Highway Statistics
State truck registrations, moror vehicle and motor carrier ux receiptS, and disposition
of rax receipts. www.lhwa.dor.gov/policy/ohim/hs02/mv.htm
>[~~~
. _-. -' ~iiL~cY~_tt""'~ _ ''w~
,_.. ~r&'i.mt4t
It
4
._.
.......
~ \~t.;;~~~:~r~~~~-.:~'~'1~~~;~,
.. ,.. ~ ~
.!~~':)~~,~~~,
,,.,/ .~.~:~'.
'"':0..<.;!'~'-- .. ~?~ , ;.;:~~~~::~-
' Economy
U.S. Economic Census
Gross state product, personal income, population and employment at sruc: levd.
www.bea.govlbealregionalldaca.htm
Productivity
Compreheosi.-e Tabulations
Volume and value of freight flows, network characteristics, economic, safety and energy
usc data, and' environmental effects. www.ops.fhwa.dot.gov/frcight
Information on. transportation and related activities in Canada, the Unitc:d States and
Mexico and berwccn the three counuics.
www.brs.gov/publications/north amcrican_transportation_atlas_data__
--
Numerous additional W.ca sees, from W.ca service bureaus or agencies specializing in the topic of inccrcsr, in addition
to those listed in Exhibit 14-1, may be helpful for goods movement sturues. Affiliate state trucking association organizations also publish deWied data on activities in their regions, such as the largest manufacturing fuciliries, number s
of vehicles by type and taxes paid, agriculrural p roduction and magnitude of freight movement by mode. Exhibit 14-2
presentS an example of the information presented on truck Aeet mixes from American Trucking Association {1987) .
The Arizona F~t NetWOrk Analysis Decision Support
System (Rul.wan, et al., 1988) was devdoped as a result of
inCXJrporaring the &eight movement and highway carrier
attitude survey findings intO the planning process, and is
represented by Emibit 14-3. The major steps in the process include mOdal decisions, forecasting and simulations
and management and strategic planning capabilities.
Several effon:s have resulted in the development of databases relating truck trip generation to land use (Fischer
and Han, 2001). TheauthoNhavecreatedasynthesis of
available truck trip generation data so~m:es, data characteristics, ex:unples ofstudies and the current state of the
practice. Data coUea:i9n was done in many cities in the
United States, with the rcsulrs reported in proportion
to acre2Se, employees, or 10,000 square feet (sq./fi:.).
of development. An ex:unple of truck srop generation
is sum.marized in Exhibit 14-4. Other trip generation
data is available in the ITE Trip Gmntl~Um Handhook
(ITE, 2008).
Special activjcy centers have aucking ncc:ds that may be
atypical, and standard trip generation rues or weight
c:x.pcctacions may not apply. A study done by the Texas
Transportation Institute (Middleton, 1990) used man-
,;11,.; .
WI
- ~~
~cs- swr
Flatbe d
45%
Dump
11%
Tank
9%
Van
7%
Uvectock
6%
Between 1982 end 1987, the number .of
large Wyoming r~tstered trucks engaged Iii
commerc:lal activities decreased by 4,000;
this Is 15% reduction trom t he 27,000
ve hic les registered In 1982.
( J) MO":clo.l~CTOAY
rJ.-Tc'm\'J.ON-=.1
'
~COUNT
OATA
Jr-.I...'W'~"::.
Source: Radwan, A. Essam, el. a!. A !Rcision Support Systnn for Frright T1111Up0rt in Ari.uma. Prepared for the Arlwna
Department ofTransportation. Transjx>rution Research Record #1179, 1988.
.
Source: Ad.aptt<l from Dallis CBD Gooc:!s and Services Di.suibution Project, 1980.
296 MANUAL Of TRANSPORTATION ENGINEERING STUDIES, 2ND EDITlON
: ual counts and poruble weigh-in-motion (WlM) scales co assess the chan.cteristics of truck mffic for timber, produce,
~ grain, beef cattle, limestone and sand/gravel. Interviews were used as a followup to quantify the typical conditions.
!
'Exhibit 14-5 presents the specified permissible weights and vehicle dimensions for the countries sucveyed (Nagl, 2007).
The range of gross vehicle maximum weights ranged from 36 metric tons in Japan co 62.5 metric cons in Canada.
The Commociity Flow Survey, conducted on a 5-year cycle as part of the economic census by U.S. Cernus Bureau,
provides data on the discribution of products by class and shipper group. A sample of manu&.ccuring establishmentS
and a sample of their shipping documents are used to develop the database. The data elements in the survey include
weight and value shipped by North American Industry Oassificarion System (NAlCS), dace, destination and means
of transportation. The Annual Survey of Manufacturers, along with Cu17T1Jt Jntiustrial &ports and other publ icacions
by the U.S. Census Bureau, provide measures of total sales, market share, quantity and cost of materials consumed,
production hours and transportation modes and location information. Over 100,000 establishmentswere sampled
during the 2007 Commo4ity Flow Survey (RITA, 2007). Reports generated include a series by industry, mode of
uansporcation and distance, weight and commodity shipped.
Australia
Brazil"'
Canada
China
Germany"
India
Japan
Russia
Sweden
Thailand
.UK
USA:u
Permissible
slnglt axle
wdi'ht .
{tonn,_;f'
9.0
10.0
9.1
10.0
10.0
10.2
10.0
10.0
10.0
9.1
10.0
9.1
Permissible groSJ
weight tractor+
uml-trnl/er
letlglh
Height
ll'idth
{tonnes]
45.5 (125.2)
45.0 (74.0)
62.5
40.0 44.0
44.0
36.0
.44.0
44.0 (60.0)
37.4
44.0
36.3 (S9.45)
19.0 (53.5)
22.40 (30.00)
23.0 {38.1)
18.0
18.75
18.0
18.0
20.0
18.75 (25.25)
10.0
:
18.75
19.8 (35.20)
4.3
4.40
4.12
4.2
4.0
4.2
3.8
4.0
4.0
4.0
4.0
4.1
2.~
2.6
2.6
2 .~
2.6
2.7
2.5
2.5~
2.S5
2.~
2.S~
2.6
11
Sources: Asia: ESCAP (2002), p . 7f; Brazil: RESOLU<;A.O No 68, DE 23; Canada: Schulman
(2003), p. 13f; Sweden, Gmnany: ECMI" (05-12..01): UK: "The Road Vehicles (Construction and Use)
ResuJations 1986" and "The Road Vehicles (Authorised Weight) Regulations 1998". Europ~
regulations can be found in EC Directive 96153/EC; USA: Code of Federal Regulations Title 23 Part
6S8 (Re'<ised as. of200S-04-01); A-ustralia: National TransportCommission (National vehicle
dimensions). '
1' Weight per bearing axle (dual tires) in tonnes (t), in some coUlltries the weight per drive ule is
~ermined to be one to two tonnes higher:
The weight and dimension limits in parenthesis are only allowed on certain routes.
21 Germany allows semi-trailers engaged in intermodal container transport a maximum weight of 44.0t.
22 In the US truck weight and truck length is determined by every State individually. The nnmbers in
this table show the minimum values regulated by Federal law. The highest values among tbe State-laws
appear in parenthesis.
Source: Nag!. Philipp. WP/07/02. UNESCAP Working Paper, longer Combination Vehicles (LCV) for Asia and ,!he Pacific
Region: Some Economic Implications, January 2007.
r .~. ..... ....... ... . ,. .... _. . . ..
In Vancouver, British Columbia, Canada, truck route rules are applied only to vehicles with more than rwo axles
or greater than 32,000 lb. gross vehicle weight (Swan, 1979), A test was run to compare the existing route nerwork
to a highly concentrated scheme of four north-south and four east-west arterials. The evaluation looked at the additional cost to the trucking industry in terms of driving time increases, vehicle miles traveled and noise impact.
The data collection dforc was an 0-D study, from which travel patterns were derived. For coding purposes the
study area was divided up into 70 zones. The classifications used for the rest of the coding detailed the body types,
gross vehicle weight, fuel cype, use of vehicle, land use classifications and commodity classifications. The Standard
International Trade Classification of the United Nations provides the youpinS$ for motor transport traffic in Canada. A computer model was used to rest the original alrernatives and subsequent variations, using aU-or-nothing
assignments.
Route srudies are sensitive to roadway capacities and in particular the ability of trucks to navigate roadways and intersections. To determine the suitability of a particular roadway or intersection to handle trucks of different sizes, field
observations may be made. A srudy conducted for FWHA (Hummer, er al., 1988) involved more than l ,100 trucks.
The characteristics inventoried included number of legs in each app[oach, average lane widths by approa~, width
of lane from which turn was made, average curb radius, signalization at intersection and phasing of signal, where appropriate. The data were then stratified based on truck cype (semi 40-ft., 45-ft., 48-ft. and double 28-ft.). Turn times
were measured as well as encroachments and conflicts caused by truck maneuvers.
Wisconsin applied the Caltrans turning template software, together with overhead photography and ground-level
photo-triangulation, to determine the adequacy of various ramps and road segments to handle erucic traffic. Computer simulations were run for each srudy site and compared to field observations. Truclc-rdared factors examined
included angle of turn, radius of rurn, tractor wheelbase, trailer lengrh, trailer width and axle width. Standard roadway
geometries were examined.
Modeling similar to that done for road nerworks can be done for railroads, with ownership being added as one of the
critical variables for each link in the system. The impact of tail line abandonments and changes in ownership can be
evaluated by modeling freight Bows and alternative path choices. The intermodal implications and highway issues are
tied directly to the railroad component of the commodity movement.
Systems/Network
Traffic counts with detailtd
vehicle classification
______ __ ___
________ -- - .--
Operational
C. Moveable
StreetS
operators
Economic
C. Water
D. Air
External Impaccs
. .
Morutonng
A. Air
B. Noise
swrounwng area
M . .
oruronng
Monitoring
A. Air_
B. NoiSe
C W:
. arer
Source: Gcndell, DavidS., et. al., Urban Goods Movnnmt Comidmttiom ;,. UrbiVI Transporratum Plannmg
for the U.S. Deparunenc ofTransporcacion, Engineering Foundation Conference, 1974.
Goods Movement Studies 2~ ~
If the ddivery location is not located conveniently, the dwell time can be excessive and there can be a measurable
impact on the adjacent road network. The typ~ of vehicles used for urban deliveries include everything from private
passenger cars, taXis, pickup trucks and so on, through tractor-trailer combinations. According to a cenrral business
district (CBD) study in Dallas, TX, USA (Christiansen, 1979), the majority of delivery vehicles are single-unit ttllcks
(40 percent), with vans accounting for an additional 27 percent and passenger cars providing another 18 percent.
Tractor-trailers were only 3 percent of the delivery vehicle population.
Sample data collection forms are included as Exhibits 14-7 through 14-9 for individual trucks, single. loading docks,
or curbside pickup and delivery operations. Srudies of single-truck operations are conducted ro determine such f.u:rors
as trip-length characteristics and dwell times by land use type. The investigation of an alsting loading dock allows the
derration of the number of trucks per day, peak period and peak-period percentage md dwell rimes, all for a given
land use. Curbside studies also allow the determination of space requirements by land use.
Gty/ke:'---------------------- -----
Stop
Location
Arrival
Time
Depanwe
Time
Notes
"AllNIIlt.l
"'"'n ,.,...,..,...'"',..
. ~lli1iTit.B1_.:
..
0 I
'
" ~
Location
Size of Buildin~
sq. ft.
und Use
Company Name
Tune
In
Tune
Out
Vehicle
'JYpe
Commodity Deliftmi
Type
Q.uanticy
Commod.ity Pic:lad-Up
Q.uaaticy
Type
Not~
Commencs
Observer
Str=L----------------------=
Company
Name
A.rriftl
TUDe
Depamue
Tune
Da~----------------------------
__________________________________
Cicy/Aro
N o ta
-'',..r.
.,.
~-
1~:3'1iTic!Il..'II~\Olll
... . .
"..tl('
:..
""- D
CommerciaUindusrrial
HotcUMotd
Food and Beverage Secvices
55 ft.
35ft.
20-25 ft.
40%
60%
60%
40%
25%
25%
50%
40%
60%
1 space
75% I space
25%
40%
60%
Source: Walters, Carol A., 1980. CBD DallaJ: A Cau Study in DnJekJpmmt ofUrban Goods Movnnnu Rrgula:Wns. Prepared for
the City of Dallas, Texas, Office ofTranspornuion Programs, Dallas, Texas.
Past srudies indicate most pickup and delivery trucking activities occur between 6 a.m. and 6 p.m. (Christiansen, 1979).
Activities tend to peak between 10 a.m. and 2 p.m. (Habib, 1980), with dwdl times ranging from 11.5 to almost 20 min.
Ocher studies have supported these observations of a peak period between I0 a.m. and 2 p.m. with an additional afternoon
peak around 5 p.m. (DDOT, 2004), with food and beverage ddivery services experiencing an average parl<ed time of 12
min. and courier services with an average of 20 min. parked. Some types of esmblishments have different parr:erns. For example, restaurants ocpcrience heightened early morning operation, with food and supplies arriving prior to the start of the
regular day. There are selected seasonal variations in freight movement, including the response ro commercial requirements
for the winter holiday seasons; there are also agrirulrural peaks which move throughout the country.
Calculati ng loading dock requirements includes the number of expected truck arrivals per day and an estima.te of the
peaking characteristics. Based on the processing time anticipated or measured, a service rate can be calculated. One
can determine the number of docks needed to achieve the desired level of operation. Analysis can also allow for testing assumptions, such as the impact on waiting time if one less dock is provided. The final srep in this process is the
prediction of the mix of the vehicle types and therefore the design requirements for the loading dock space. The Dallas
study provides a reasonable estimate by land use type, as shown in Exhibit l4-10.
The use of static scales is required for the certification of truck weightS in the case of a cited violation. The accuracy
of !hese scales, when calibrated, makes the data defensible, if necessary, in a coun of law. Typical Pc:rmanenr stations
are often costly to construct, equip and operate. From a transportation network perspective, having to bring the truck
ro a complete stop often causes backup and safety problems, as a weigh station may be prooessing many thousands
of trucks in a single day. The data collected at permanent weigh stations is often biased, due to general knowledge of
their location and operation. Oversized and/or overweight vehicles are suspected of circumventing the stations. As a
result, weight enforcement programs often focus on known detour routes around weigh stations.
302 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2N O EOrrlON
.......
-y~~-~,.;;~-~1\:~-~...>(
"'"''"'-'
.. -'
. ..
,_.,._
'. .
'
Technology
Quam
Piewdecrric
Sensor
Performance
Can Meet
.. -.
...
-.,o~l
. --
Polymeric
and Ceramic
Piezoelectric
- ..~~,... ~,
Bending Plates
.',.J,' - .
Loadulls
BridgeWIM
Can Meet
More Researcll
Needed to Verify
Accuracy
Snull Road
Cuts I Day to
Complete
Small Road
Cuis I Day to
Complete
Significant
Road Cut.with
Proper Drainage
Required.
Multiple Days to
Complete
Maintenance
Must maintain
surface
smoorhness and
seal properly .
to achieve
satisfaCtOry
performance
Musrmaint:ain
surface
smoothness and
seal properly
to achieve
satisfactory
performance
Inadequate
Corrosion ofload
cell if nor scaled
drunage can
cause bending
correctly
p~ tO come out
of rhe roadway.
Required
6-month checks
and annual inroad in.specrion
Minimal
Safety Issues
One Day
for System
Insu.Uation and
During Periods
ofin-R.oad
Maintenance
One Day
for System
Installation and
During Periods
ofln-Road
Maintenance
Significant
safety issue if
bending plate
comes our of the
roadway. Multiple
Day System
In.scallation and
During Periods
ofln-ROad
Mainre~ance
Multiple Days
for System
Installation and
During Periods
ofln-ROad
Maintenance
None
Cost of System
Low-Medium
Including Installation Cosr
Low Cost
Medium-!:Ugh
Cost
High Cost
Low Cost
Marure I Proven
Technology
Yes
Yes
Yes
Not in United
S12res
Senso~
Enforcement
~uiremencs
Enforcement
~uirements
wilh 2 rows of
sensors. Bcuer
accuracy can be
achieved wirh
3 rows rhrough
averaging out of
vehicle dynamics
In.sta!lacion
Yes
Significant
Road Cut wich
Proper Drunage
Required.
Multiple Days to
Complete
Non -Intrusive
ln.strumenr:a.tioo
2.4.2 Weigb-in-Moi'Um
WlM is usd in many areas to detesmine if a truclc is traveling within a reas9nable range of the legal limits as adopted ~
a local government. The dara collected may include all or some of the following: gross vehicle weight, axle weight anC:f.
tandem axle weight. In the United Srates, local and federal departments of tranSportation or other responsible agmcie~
monitors rhe data collectt<l, and together with an enforcement agency is further charged with penalizing violacorn.
Several technologies can be utilized for WIM sensing inducting bending plates, piezoelectric sensors (qu.arcz,' polymeC"'
and ceramic), load cells and bridge WIM (Pines and Fang, 2008). The Wh\1 scales are often found ar permanent::='
Goods Movement Studies 303
truck weigh station sites, but are also used by some agencies in a roving mode. Exhibit 14-11 shows a comparison of
the various WIM technologies.
The accun.cy of the weightS collected by a WIM device are affected by many factors. The characteristics that affect
the accuracy include roadway factors, such as longitudinal profile, cransverse profile, grade, cross slope and curvature;
vehicular factors, such as speed, acceleration, axle con.6guration, body type, suspension system, tires, load, load shifts,
aerodynamic characteristics and center of gravicy; and environmental facrors, such as wind, temperarure and icc (1TE
Technical Committee 5D-6, 1986). The final repon of the ITE Technical Committee on Truck Weighing-in-Motion
presentS the findings of many test studies conducted around the United Scates with respect to the factors lisred above.
In addition, scale calibration is a significant factor in WIM readings.
6,061
.9
.6
l-',241
2.342
.7
57
KEY: - Repr=nts ari estimate equal to zero or less than 1 unit of measure.
S = Estimate does not meet publicacion standa.tds because ofhigh wnpling variability or poor response quality.
~Truck"
as a single mode includes shipments tharwae made by ooly private truck, only for-hire uuck, or a combination of
prmu: and for-hire uuck.
2 Escimates for Diodine e:zclude shipments of crude
Sou.rc.c: U.S. J)qlartment ofTransponacion, Raearch and Innovative Technology Adm.i.n.istratioo, Bureau ofTran.sportacioo
Statistics and U.S. Census Bureau. 2002 Commodity Flow Surwy. H.u.!mt Daca. Table la. December 2004.
.
~""
,..,....,,.,.,
(Shipments should be routed to optimize r.raruit time and reliability, minimize accident probability and minimize the
<impact on poople, property and the environment in the event of an incident. These goals can sometimes be in conflict.
1<:;omputerizi files exist for analyzing rail movement ofh:uardous materials. This is done py combining network data,
Federal Railroad Administration accident/incident Iiles and demographics ncar system links. To determine the primuy
and badcup routes for moving h:uardous materials, a detailed srudy needs to be conduaed (for example. McMillen et al.,
1984). Existing conditions are inventoried, similar to a typical traffic impaa srudy inventory; however, more in-depth
information has to be gathered on collisions (by location, type and vehicles involved) and surrounding land uses. The
areas to be avoided as much as possible include hospitals, schools, historical areas, runnels and parks. Proximity to public
safety agencies and emergency medical services is desirable. The National Hazardous Materials Route Rq;isuy exists to
determine which routes arc designated or prohibited for hazardous material transportation (FMCSA, 2009).
3.0 REFERENCES
3.1 Literature References
Aiura, N. and E. Taniguchi. pbnP.ing On-street Loading-unloading Spaces Considering the Behavior of Pick-up Delivery
Vehicles." ]DIITIIIIl Dftht E.astnn AriA Sti~IJfor TtrtnspDTt41itm Snu/Us 6 (2005): 2,963-2,974.
American Trucking Association, Wyoming Truclcing Association. Trudang in lr)ooming. Burlington, MA: American Trucking
Ass~tion and Wyoming Trucking Association, 1987.
Bureau ofTransportation Statistics. 2002 EctmtJmic Cnuus- TnznsportllliM: H4Z4rdous Maurials. Washington, DC: Bureau of
Transportation Statistics. U.S. Department ofTransponation, 2004.
Christiansen, D. Urban Tnmsportatitm Planningfor Goods and Smtiar. Washington, DC: Fed=! Highway Administration, 1979.
Christiansen, D. DttJJ.u CBD C.txis 11Ni Smtices Dimibu:Um ~jm. Dallas, TX: Gcy of Dallas Office ofTransportation
Programs, 1980.
DC (District.of Columbia) Department ofTransportation. Di.ttrict DfColumbia Mbt11r Ctrrin- Mmwgrmmt and Thro# Ass~t .
St...ty. Washington, DC: Volpe National Transportation Syscans wccr: U.s; DepartDWtt ofTranspon:uion, 2004.
Federal Highway Admittisaation. Frright A1ralyrit ~ft. Washington, DC: Federal Highway Administration. Freight
Management and Operations, November 19, 2008. http://ops.fhwa.dor.gov/freightlfreight_analysis/faf/index.hun. Reuieved
January 5, 2009.
Federal Highway Administration. Fmtln ~t and OpmuiDm. Washington, DC: Federal Highway Administration.
http://ops.lhwa.dot.gov/freightlindcx.cfm. Retrieved January 7, 2009.
Federal Motor Carrier Wecy Administration TIN NiUilm41 H~lr4n~Dws Maurials Rouu &gistry. Washington, DC: Federal Motor
Carrier Safety Administration, June 10,1009. www.fmcsa.dot.gov/safery-security/l=mat/oational-hnmat-rout~aspL
FISCher, M. ]. ar~d M. Han. NCHRP Synthesis 29S: T~!t Trip GmmzliJJn. Wa,dllngron, DC: Transporration Raearc:h Board,
2001.
Gardner, W. D. "Truck Weight Study Sampling Plan in WISCOruin." Trrzmporr.tllilm kuTch &ami: foliTIIIIl Dftht T~"
~tmh IJolfnJ 920 (1983): 12-18.
Gcndell, D. S., ct al. Urb1t11 Gotxis Mtn~m~mt C4mi4n#Wm ill Urban Trtl11Sport41Wn Pl4nning Studin, FHWA 32-01-23 RFP
397. Washington, DC: U.S. Department ofTransportation, Engineering Foundation Conference:, 1974.
Habib, P. "Transportation System Management Opcions for Downcown Curbside Pickup and Delivery of Freighr. "Tramporrarion
&search Record: Joumal ofth~ TTl111!porration <arrh Board 758 (1980): 63-69.
Hummer, J., C. Zegccr, and E Hanscom "Effects ofTurns by Larger Trucks at Urban lncersecrions." Transponarion &srarch
Rmrd:joumal oftht Tmns~rttttion &uarrh Board 1195 (1988); 64-74.
lnstiwre ofTransportation Engineers. Trip Gmtration, 8th ed. Washington, DC: ITE, 2008.
lnstiture ofTransponation Engineers Technical Commlrree 5D-6. Trucking ~ighing-in Momm, Informational Repon.
Washington, DC: ITE. 1986: p. 7.
McMillen, R., M. Anderson, and C. Cerbone. Traf!U and Transportation Analyris: H=rus Wastt Disposal Facility. How ron,
TX: Nacional Conference on Hazardous Wastes and Environmencal Emergencies, 1984.
Middleron, D. Rnults ofSptcial-Ust Truclt Dlllll CollNrion, FHWArfX-420-3F. Washington, DC: U.S. Pcpanment of
Transportacion, Federal Highway Admlnimarion, 1990.
Nagl, P. Langer Combination Vth~ks (LCV) for Ana and tf1t Padfo &gion: ~=Economic lmplicalions. Bangkok, Thailand:
United Nations Economic and Social Commission for Asia and the Pacific, 2007.
Pines, D. and C. Fang. A Study ofWtigh Swion Ttchnolot,Us and Practim. Hanford, CT: Con.necticut Academy of Science and
Engineering. 2008.
Radwan, A. E., J. Cochran and M. Farris. A Decision Support Sysum for Freight Tramport in Ari.7Jlna 1179 (1988): 2}-30.
Radwan, A. E., M. Rahman and S. Kalevcla. Freight Flow and Attitudinal Survey for Arizona. Transportation Reuarr:h R~rd:
journal oftht Transportation Restarch Board 1179 (1988): 16-22.
Research and Innovative Technology Administration. Commodity F/Qw Survey. Washington, DC: Research and Innovative
Technology Adminisuation, Bu=u of Transportation Sratisrics, U.S. Census Bureau, U.S. Depanment of Commerce, 2007.
Stammer, R., C. Wright and J. Donaldson. "Conducting Truck Routing Studies from a New Perspective. Transportalion
Res~arch Record:,]oumal ofthe Transportation Restarrh Board 1038 (1985): 59-63.
Scrauss-Weider, A. NCHRP Repon 320: Jn~grating Frtight Facilities and Operations with Community Goah. Washington. DC:
Transportation Research Board., 2003.
Swan Wooster Engintering, Co. Led. Evaluation ofUrban Trucking &aionalization in Vizncouvtr. Phases/ and 2, W>L 5. Montreal,
Quebec: Urban Transportation Research Branch of Canadian Sumce Transportation Administtacion: pp. 2-15.
Transportation Consulting, MDA,lnc., d/b/a Street Smans, Duluth, GA. 1992.
Tnnsponation Research Boatd. Mtasuring Pmonal Tra~l and Goods Movtmm~ Special Report 11277. Washington, DC:
Transponation R=aJch Boud, 2003.
U.S. Census Bureau. Econam~ Cnuus. Washington, DC: U.S. Depattmenc of Commerce. Bureau of the Census, 2007.
U.S. Department ofTranspomtion. Comprthmsivt Truck Sizt and weight Study. Washington, DC: United Stares Depattment of
Transportation, 2000.
Walrers, C. A. CBD Dallas: A Cast Study in Dtvtlopmmt ofUrban Goods Movtmmt Regulations. Dallas, TX: City of Dallas
Office ofTransportation Programs, 1980.
Winfrey, R., P. D. Howdl, and P. M. KenL Truck Traffic Volume and Wnght Dara for 1971 and Thtir Evaluation. Washington,
DC: Federal Highway Administration, 1976.
Saito, M. and T. G. Jin. E~~t~fuating rh~ At:curacy Lewl ofTrudt Traffic Data on Stat~ HighwllJS. Provo, liT: Brigham Young
Univcrsiry, 2009.
Teal, R. Estim.aJing tiN P,(J Economic Costs ofTruclt lnaamts on Urban FruwllJS. Falls Church, VA; AAA Foundation for Traffic
~ery,l988.
TEE Consulting Services, Inc. FraT1UWOr!t for Urban Goods Movnnml lnfomliltion in Catimia. Vol. 9. Urban Goods Movement
Rcpon Series. Montreal, Quebec: Urban Transportation Research Branch, Transpon Canada, 1979.
U.S. Department ofTransporution. Urban Goods Movnnmt Input to NawMI Trrmtporr41wn Ptm.. FiMf &porr. Washington,
DC: U.S. Department ofTransportation, Office of the Secretary, 1976.
Chapte r 15
.........................
...........................................................
'
Inventories
OriginAl by:
H. DougfM Robertson, Ph.D., P.E.
EJiuJby:
Dtmulj. FUu/Jey, P.E.
1.0 INTRODUCTION
309
1. 1 Word of caution
310
310
311
3.0
311
316
316
317
2.5 Storage
317
2.6 Retrieval
317
ESTABLISHING AN INVENTORY
317
317
318
318
4.0
311
318
i 320
320
MAINTAINING AN INVENTORY
320
4.1 Software
320
4.2 FrequencyofUpdates
5.0 REFERENCES
321
321
1.0 INTRODUCTION
n inventory is a catalog. lisring. accounting, record, or display of &ctua1 information that describes aisting
con<l.itions. Agencies store information in manual or automated (compute.r or video) dat2hascs, from which
they gen=.te meaningful presentations for various analpes. The type of information included in an inventory
is <l.iaaa:d by the purpose for which the information will be used. Some inventOry data remain constant, while other
data change rapidly over time and must be updated frcquendy. To be useful, inventories must be accessible and contain relevant data dements. Common types of transportation-related inventory data categories include:
lnH.c.nttvioc
traffic generators
A general procedure for conducting street and highway inventories is presented in this chapter. Discussions pertaining to performing ocher types of inventories can be found in the chapters devoted co those subject areas. Additional
information on inventories is available in the Trrmsport41ion PlAnning Handbook (ITE, 2009), the Traffic Enginming
Handbook (ITE, 2009) and NCHRP Report 437 Colkction and Pmrot4tion of Roadway lnvrotory Dat4 (Karimi,
Hummer and Khattak, 2000).
Can those data be obtained more effectively by means other than an inventory?
Does the informacion ilieady exist in another form?
How large is the srudy uca and what will be the extent of the effon to complete the invenrory?
Who will collect, enter and analyz.e the dua, and what is their level of expertise?
Agencies must identify the intended uses of inventory information so they can collect relevant dara in a suitable form.
For example, if the inventory is to be used co schedule maintenance, a sec of data elements should be identified that
relate to condition and service life. On the other hand, if the inventory is to be used co crack maintenance costs, elements related to time, labor, equipment and materials will be required. Some typical uses of inventory data ue:
illustrate street classifications
locate TCDs
specify condition o~ service life of devices
310 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
schedule maintenance
manage costs
depict application oflaws and ordinances
assist in evaluation of traffic operations
provide baseline conditions for use in other studies
locate traffic generators
compare measures of effectiveness with before and after conditions
track deterioration and service life of transportation assetS
Of course, the inventory may sene more than one purpose. Agency or governmem accounting procedures might
require the storage of cer~ inventory characceriscics for the calculation of infiascrucrure costs. Agencies must avoid
collecting data dements simply because they are there and might be useful in the future. If after careful srudy :an
agency cannot identify a use for a data element, the element is nor needed. Many inventory efforts fail because dJ.ey
attempt to collect too much information.
having those data in che inventory Will preclude a visir to the field co obtain che information at a later time. Agencies m!J-St
weigh th.i.s savings against how frequendy' and quickly they will need the data dement and whecher a field trip would
be 'necessa.ry anyway to collect or verify other data. Occasional field trips may be more cost-effective than mainwruo.g
certain data elements in an inventory. Inventory data collection and maintenance are costly. Although computers a$e
valuable and necessary tools for coostructing inventories, they are a4o associated with the possibility that coo muchcfat:a
are collected and the pracrical utility of the inventory is compromised. Extensive amounts of electronic data can be difficult O format, update and access. The fundamental rule is: Ketp it simpk and to th( point. Elements that are crirital a!J.d
are used frequently should be inventoried. Ocher data dements should only be collected "as needed."
Inventories 311
~"
~
.
.
. -:-:~t'!::....:~!!!'::~-:...~-.- .
.
=
'''
I'
''
I
I
'
I
I
I
'
I
I
~:t!tt
I
I
I
I
I
I
'
I
I
I
I
I
I
LEGEND:
======= .
30 MPH
35 MPH
- - - 40MPH
- - - 45 MPH
Sour= Bor. and Oppcnlandcr, 1976.
""
0
" .
~ ~
~
.....
1\'
~' q4 1
""
""
'"!"
....
>-I
9
. <>
r .
rJ
(:
"'
"'
"'
~
~
-+
e
I a
.
:
tS
~~ @l[,gl~, ~w
0~
ri ~
.
.
----
0
(X)
ground~
remote sensing
. The importance of accxss to inventory data has already been mentioned. The means of accessing data and information
. must be a primary consideration when establishing any inventory. The method ofstorage and access must be tailored
to the frequency and type of usc to be made of the data. The cost ofsofrwate licenses can affect the total price of che
inventory and should be considered carefully when selecting a sofrwate package and when considering how many
access poincs to the invenrory need to be provided to users. Sofrware training expenses and the technica,l expertise
required for analysis should also ~ considered.
il
I:
2.5 Storage
Inventories ace stored either manually or in a computer. Manual systems may be in the form of maps, card files, file
folders, boUlld listings, ~cro6cb.e, photologs, or videos. These systems are kept in 6ling cabinets or on boo kshelves.
They work reasonably well with smaller inventories, but larger systems become unwieldy and consume space. Computer inventories may be stored on hard drives, computer storage media, or on a server. Computer systems consume
less space than manual systems do, so they are better- suited to large inventory systems, including picrures and vi4eo.
However, computer systems must be appropriately archived on a regular basis to reduce the potential loss of valuable
data in the event the operating system crashes, 6les become corrupted, or users unintentionally delete data. The use
ofshared systems across departments can provide a reduction in storage costs.
2.6 Retrieval
Nearly all inventory systems are stored by location (that is, intersection, street segment, or GPS coordinates). Therefore, one musc know the location of incerest to access inventory data on that location. In some manual systems, all
inventory data for a given location may be kept in the same 6le. Other manual systems store data first by type of
inventory and second by location. Computer data systems may be organized in a similar fashion. These systems are
capable of searching and sorting data by data elements other than location. For example, if there is a need to examine
the installation dates of all stop signs in the system for maintenance purposes, each intersection in the manual :file
would have to be checked to see if stop signs were present. In the computer system, the computer seeks out stop sign
locations. Additionally, the computer could sort and retrieve only those intersections with stop sign installation dates
prior to a given cutoff date, thus saving both time and effort.
The comp11ter can also generate repons or listings containing only the information desired and therefore avoids the
need for manual extraction of data. This capability applies to periodic replacement ofdevices, spare part inventory
control, stock replacement and ordering, dispatch of work crews, maintenance of history files and possible liabair:y
defense. In addition, work orders may be generated automatically. Agencies typically design work order formats to
enhance inventory updating upon completion of the work.
and recorded, how ro handle unusual ciccumsranccs and che rules ro follow rhar wiU ensure consisrency in measure:
menrs and dara recording. If agencies collect dara manually, rhey may design forms or program laptop compu;us so
.observers can record data in the field directly. These actions preclude the daca reduction ~teps required in the office if
video is used. Exhibit 15-7 shows a sample of a sign inventory data collection form.
, "'';
City
Street
between
Location
l;-
~ ~
~~
Size
and
Condition
. ?:
;;
~I I
~
.=f!ef! ;;;"8~
--c
t!
~c:
"EijV.:~~ .. ~l;~~;. ~~ ~c o!1.,
o."D
8~E c...,.,:::."'"'u"ii2"E
f:~i!l:!l'~of.c"il ~ ~ ~-~~ ~... ~ 8
g v
~~~Vi::i!~
X~~ a:.. .. z::>E z::>E
41
a:
\1
4'
::r
-----------------
Post
>
2:~
Sign Type
Direction,_ __ __ __ _
l;
....
l;
"0
"0
~I
Notes
""c: 1:
1 "'~
0
..
Q.
IJ
a:
Inventories 31 9
4.1 Software
Numerow computer software packages are available to aid in both establishing and maintaining traffic and highway
inventories. Tcaffic maintenance software focuses on pavement markings, signs, signals and street lighting. Highway
maintenance software encompasses a road inventory system of highway and bridge segments. Typical data elements in
highway maintenance software include pavement, culvert, guardrail, road and sign elements. The strategy evaluation
capabilities QfRoadSoft GIS (Michigan Technological University, 2009) can provide the ability to anal}'le the impact
of policy and fun~g decisions on the roadway network. This analysis can provide planning and reporting details for
maintenance spending over the lifetime of roadway assets.
~cral factors need co be considered based on the specific needs of the agency. The number of users who need access
to the sofrware will affect the cost of software licenses; this should be considered in the total price. The synchroniza-
tion of the software files must be considered if multiple users plan to si.multaneowly edit the files. Software data must
be appropriately archived on a ~ basis.
5.0 REFERENCES
Cunard, R.. A. NCHRP Synthesisof Highway Practice 157: M.Untmanc~ MaMgmsmt DfStrut and Hifbway Signs. Washington,
D~: Trw.sportar:ion R=an:h Board, Nar:ionil R.c::search Council, 1990: pp. 28-43.
Dana, T. K, K H. Tsuehiyarna, and K S. Opida. CDn-Effictiw lnvmtJJry Promium for Highw4J lnvm~~~ry D<aa. DTFH6183-C00043. Washington, DC: U.S. Department ofTran.sportadon, Fedeta! Highway Administration, 1985.
Edwards, J. D. Transportation Planning HandbDolt. Chapter 11. Englewood Cliffs, NJ: Prentice Hall, 1991.
Institute ofTransportation Engineers. Traffic Engin~mng Han4JJDD!t, 6th ed. Washington, DC: ITE, 2009.
Institute ofTranspottation Engineers. TranspDrtation Planning Handhoolt, 3rd ed. Washington, DC: ITE, 2009.
Karimi, H. A., J. E. Hummet and A. J. Khattak. NCHRP R.:pon 437: Colkction and [>rQmtlltion Df&adw4J lnvmtDry Data.
Washington, DC: Transportation Research Board, 2000.
McShane, W. R.. and R.. P. R.ocss. Trrtjfo EngiMmng. Englewood Qiffs, NJ: Prentice Hall, 1990: pp. 75-79.
Michigan Tec:hnologic:al University. RoaJSqfi GIS. Houghton, MI: Michigan Tec:hnologic:al University, Technology
Devclopmenc Group, 2009.
Pline, J. L Traffo Enginming HitndbDDit. Englewood Cliffs, NJ: Prentice Hall, 1992: pp. 59~.
Sabria, F., B. Bonnet and E. Sullivan. Evaluanon ofTraffic Conrrol Dnnu lnvmtDry Progrwns for Microcomputn'S, UCB-ITSTD-85-2. BeYkdey. Ck University of California, Institute ofTransportation Studies, 1985.
lnuDntnriDc:.
'1.,1
Cha p ter 16
...... . ........... ...... ....... ... ... .... .... ... ....... ............. ........ ......... . ......... .
Parking Studies
OriginAl by:
L EOis Ki11g; D. Eng., P.E.
Edited by:
324
324
325
325
325
2.4 Method
326
329
330
332
332
336
33 Postcard/Flyer Interviews
337
338
339
339
339
5.0 SUMMARY
343
6.0 REFERENCES
345
345
345
345
1.0 INTRODUCTION
arking is one of the essential elements in urban transportation. With few exceptions, automobiles and trucks
must be parked at least temporarily at each end of a vehicular trip. Even in areas served by public transit, rhe
automobile is a common means of transportation. This chapter focuses on vehicular parking, but many of these
concepts and methods can be applied to specialized applications, such as motorcycle or bicycle parking. There are rwo
general types of parking:
1. owner-supplied facilities ar homes, apartments, retail centers, industrial buildings, institutions and offices,
including public street curbs where non-metered parking is allowed;
2. commercial parking. including private lots or garages where fees are collected; and charge parking. such as
meters at curbs or in public off-street lots or garages.
Parking s~dies arc typically conducted to check physical needs of the existing parking supply and location or to establish parking policy and parking regulations, such as zoning codes or specific developments. The majoriry of parking
studies are performed to determine the need to expand existing parking or adjust management techniques by comparing parking demands to parking space supply. Parking studies can involve central business districts (CBDs), industrial
developments, office parks, densely populated apartment/condominium areas, hospitals, universities, sports arenas,
cultural facilities and special events.
The provision of parking spaces has important economic and environmental effects. According to a study in Tippecanoe Counry, IN, USA, parking spaces consume 6.57 percent of the total urban footprint in the counry and use
approximately 20 percent more area than the buildings they service. With appr.oximately 1.7 parking spaces per
. registered vehicle in the counry or 83,000 potentially unused spaces, the study found an environmental iinpact of
$22.5 million for the additional spaces (Davis c:r al., 2010).
The presence of on-street parking reduces the capaciry and flow of roadways, while increasing the potential for
collisions (AASHTO, 2004). However, parallel and angle op-street parking can be useful in specific situations with
context-sensitive designs by encouraging pedestrian activity, increasing packing supply and traffic calming (ITE,
2009). On-street parking is generally more favorable for low-speed and low-to-moderate-volume roadways.
Advanced parking management systems such as pretrip planning information, quantiry available (by lot, floor,
aisle, or space), reservation systems and navigarion systems can ease parking supply issues (FHWA, 2007). Examples of these techniques ace present in many countries. The public dissemination of information about parking
space a.vailabiliry can increase the utilization, efficiency and consumer satisfaction of a packing system. The occurrence of illegal parking increases significandy when the packing lor reaches 80 percent of capaciry and is generally
perceived by users to be full when 90 percent of capaciry is reached (ITE, 2004).
Parking studies look at supply and demand. On the supply side, inventories ace conducted to detail the current
parking conditions. Demand is commonly measured in terms of usage and turnover. However, accumulation and
usage studies measure where drivers actually park, while real demand is the location where drivers would prefer to
pack- Usage examines number of cars while turnover looks at the speci.c vehicle using a space. Additional data may
be collc~ed ro determine packing generation which is important in calculating future needs.
.j
:!
:
1.1
:l';,
\ :
\;
~24
If the srudy involves a congested area, it is desirable to sdcct natural boundaries, such as rivers, railroads, or major routes
along which changes in land use oc.aJI. For a major route, it is generally necess:uy to include a distance of 300 to 500 fi:. (90
to 150 m) along each cross street since these areas are potential locations for curb parkers unable to park on the major route.
cp 0
I
l) 2~2
.(D
2.4 Method
Before beginning the invenrory. a master coding system is
established. Each block is given an identification number
and excra-long blocks may be subdivided and given cwo
numbers. If there is a prior tr.tnsportation or origin-destination (0-D) srudy for che area, which includes a separate
number for each block of the parking study area, it may be
convenient ro use the same numbers.
.
i'
G)
3
3
I
3
I
.I
24(j)
...
CD
:D 402 4
3
I
2~
2~.
2
4
I.EOENOo
< DBLOCt< NUMOEA
3 .CUA8 F ACf!. NUM8EA
11,1.11"
PARK~
FACIUTY NUMBERING
~t
41
ST.
@j~l
~
91
1::;
12
lf individual curb parking stalls do not have pavement markings, it is necessary to determine the approximare length
of each curb parking secrion. This length is the distance from near edge to near edge of crosswalk in a block that
contains no driveways, fire hydrants, or other parking restrictions. However, che usual block contains several areas of
restricted parking and measurements are needed within each separate section. If parking restrictions are readily identi~ fied on a scale aerial photograph, much of the field measuring can be eliminated.
Estimates can be made of the available number of parking spaces in a given unmarked discance by utilizing the following figures:
parallel packing
angle parking
Ii
90 parking
These dimensions for unmarked stalls reflecc a lower operating capacity than if markings were in place.
Counts of individual stalls or linear measurements may be used to determine the number of parking spaces available
in a lor or garage. If the facility is a private commercial cype with attendant parking. determination of capacity is more
&fficulr because ve~ides may be stored in the aisles. In general, the operating capacity of an attendant-parking (v-aJer)
facility is the peak-condition loading, when all space is filled except for what is needed to allow for off-street man~u
vers and temporary parking of vehicles (which must be moved to allow access to parked vehicles).
The number of parking spaces in each block of the survey area and the number of available parking spaces in similar
facilities are tabulated and summarized for the entire study area. Typical categories for the overall summary are
1. curb and alley parking
a.
metered
b.
nonmetered
c.
special zones
d.
2. parking lou
a.
public
b.
private
3. parking garages
a.
public
b. private
The block-by-block summaries are tabulated on an inventory form such as that shown in Exhibit 16-3 . Under c.he
"special facility cype" columns, individual types of facilities, such as those under a given time-limir restriction or a
special regulation such as loading zone: may be identified separately. In a parking report, maps are frequently used
to illustrare invcnrorr. data, as shown in Exhibit 16-4.
Parking Srudies
12?
?;:~~wv'."'~'
t-
-~~
..
-.
: '>ric~:,
.. 0
...;-":'i'-'~;~w~
fi~~"'-~w~~~~~'3<~~;:
, ..
~r:.;:
. .
: ~ !-'"
~~....,;.:
.... ~
..,
~r~..
" . ~~- . .
Date
Area/Ciry
Special Fadliry Type
Block
Facility
Comments
Observer
//
Off-Suut Parking
Private
Public
Total
Stalls
qrivate lots or garages are those restricted to the use of employees, residents, or tenants of a building; r:U:icab
d[ truck storage area, or other facilities not open to the general public. Lots and garages open to some public
parking, as weU as weekly or monthly rentals, are classed as public facilities. Where space is rented to jnd ividuab, it is classified as public space because the space may be rented in the future by other members of the genen.l
public. If this study is for public parking only, employee spaces are not included since they are not usable b y the
general public. Other restricted spaces include those for named employees, visitors, compact cars, motorcycles,
or carpools.
When appropriate, the pa.rking inventory may also gather data on assessed valuations of vacant lots or old buildings that might be replaced by parking facilities. Certain elements of traffic control, including one-way s treetS
and alleys and restrictions on turning movements at intersections, should also be included. Such controls affect
routing ofwindshield surveys, as weU as access to potential locations for development of new parlcing facilities.
Inventories of smaU areas and along major traffic routes foUow the general format of the CBD inventories except
that aerial photographs and/or st.r ip maps are used to a greater extent.
Source: Litm:~.n, T. Parlting MaMgnnnu: Stmugin. E~n and Plmrning. Victoria Trall.'lport Policy Institute, 2008.
Vehicle-Hours of
Block I
Parking
Accumulation, or parking occupancy checks, plus rurnover and duration studies, are useful in determining what (if
any) curb-parking improvemenrs can be made to increase parking capaciry. The duration analysis indicates where
long-time parkers are using space inefficiently, and a relative measure of curb-use efficiency ~s provided by comparing hourly parking turnover rates with desirable rates to indicate efficient usage. Parking practices or regulations thar
contribute to uneconomical and inequitable use of street space are revealed in this analysis. For example:
1. The srudy may show enforcement of existing time limits is needed to stop overtime parking.
2: It may indicate existing time limits are too long or coo short. Consistent 30-min/parking in a 15-min. z.one
mighr indicate a need to revise the regulation. Conversely, a preponderance of 30-min. parking in a twohour z.one would indicate the desirability of reducing the limit to 60 min. or less if an increased rumover
is desired. Permit parking during restricted hours (7 a.m. to 3 p.m., for example) could displace parking to
adjacent areas.
3. H azardous or illegal parking may be revealed. The faccual evidence of such practices may prove helpful in
bringing about needed enforcement or voluntary public compliance with regulations.
Occupancy checks are useful in determining needed improvements for cruck pickup and delivery service. Consistent double parking by commercial vehicles may indicate the need for additional loading z.ones or for berter enforcement of regulations if available loading z.one space is adequate. Police action may also be required if passenger
vehicles are observed to be blocking the existing zones. Where it is proposed to prohibit parking. occupancy checks
are helpful in determining the parking demand at the hours of the day during which the prohibition would be in
effect. When considering changes from angle to p:uallel parking, installation of parking meters, or installation of
curb passenger and freight loading z.ones, a study of the curb parking activity data is useful. These studies are also
performed to determine parking efficiency at older industrial plants, and parking lots or garages in business areas.
To represent typical conditions, a study should consider any underlying schedules, such as nearby school class
schedules, holidays, or special event schedules, to either capture or avoid those conditions as desired by the study
needs.
If the accur;nulation studies are being made at a particular generator, such as an office building, studies should
begin prior to the morning opening of offices and immediately after closing hours. At industrial plants operating
on more than one shift, parking checks are needed at time of shift overlap. It is typical for office and administrative personnel to have different hours &om licrory workers, and separate checks may be needed to determine these
demands for parking.
330 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
Along major traffic routes, checks are made during the morning peak hours from 6:30 to 9:30 a.m. , during sever:tl
: midday times, and across the evening rush period from 3:30 to 6:00p.m. Tf residential properties a b ut the route ,
it is essemial checks be made during the lace-night period since for residemial parking, peak accumulation occurs
berween 1:00 to 5:00 a.m. Demand is relatively constant during this period and a single overnight check should be
adequate. The traffic engineer is generally most interested in the peak-parking-demand times as rela ted co currenr
supply. ln the case of a major traffic route study, where the incenc may be to impose only morning or evening rushhour resrrictions, checks should be made during these time periods.
It is often desirable to prohibit all parking along a major route since this is the most effective way co provide for the
total transportation of persons and goods, and co maximize public safety. Therefore, parking checks along maj~r
routes usually include hours ocher chan just che rush periods. In all parking checks, it is important tO avoid condttions of tempocuy parking restrictions, such as foe street sweeping. screet repairs, or snow plowing. Parking checks
in retail areas are usually not taken during periods of abnormal demand such as the day following Thanksgiving or
national holidays. I n SOf!le cases, seasonal parking demand, such as the weeks before Chriscrnas, may be the focus
of a parking study. The effect of major sports activities should be considered in the area of such generators, and
special checks ace 9ften made to determine peak parking needs.
If the accumulation studies arc made co dete.r mine the peak parking demands of specific generator types in cider
to develop wning code specifications, peaking characteristic information muse first be secured from cooperadve
administrators or other reliable sources. The heaviest hours, days and months muse be determined, and factors s1.1ch
as varying visitor hours at h ospitals and nursing homes or shift overlap periods must be ascertained.
In the generation-type study, it is essential to know the occupancy or facility usage at the time of the checks. In
hospitals and nursing homes, this includes the number of occupied beds as a percentage of total available b~ds,
and in the case of subdivisions or apartment buildings, it involves the number of occupied dwellings. In offices,
industrial plants and retail ceiuers, rhe .amount of occupied leasable square footage of Boor area as a percemage of
che total leasable floor area is used.
A s~cond precaution in the performance of generation-type srudies is the need for identification of parked vehicles
as related to specific establishments. When patron parking is cominglcd with ocher parking, it may be nccessa.r)'
to conduct interview-type studies or make specific observations of patron activities to secure reliable data. T h e
parking needs of specific generators often vary among distriets within an urban area, and the availability of pub lie
transit is a major factor. An office building in a CBD served by transit will have a lower parking demand than chat
of a similar building located in an outlying area. Parking-generation figures calculated for one district should pot
be applied arbitrarily to a dissimilar area. Representative trip and parking generation races for vario us land uses
and building types are readily available (ITE, 2003, ITE, 2004). However, these average rates should be used wi. t:~
care when applied co a particular situation. The interaction and coordination of parking for multiple land uses .s
a consideration for mixed-use developments (ULI, 2005). Mixed-use developments provide the opponunity or
complementary land uses to share parking facilities due to variatio;ns in hourly, daily, or seasonal vehicle accumulation and walking trips within the developme11t to different land uses.
Parking-accumulation studies can be performed using maps that show the oudine of the different parking facilities. In larger areas, a number of sectional maps may be required. The occupancy studies work well using a vehicl e
with driver and observer. Colored pencils are used to note the number of parked cars found in each facility, with. a
different color used for each time period. In larger facilities it may be easier to note the empty spaces rachCI th~ll
the occupied spaces.
Since pare of the accumulation checks are made during periods of traffic congestion, it is important to plan c.b-e
study route carefuUy. This includes preliminary checks of driving time required to cover the different ar~ a~d
establishment of the most efficient travel pattern. In more complex CBD accumulation studies, where the 101..1 ce
may be involved and congested with cf'l!flic, it is desirable tO have the survey crews make several practice flllll. J: .D
some of the more congested areas, it may be necessary to conduct the accumulation checks on foot. When it _i.s
impractical to count accumula_rion within garages by driving through them, the required data are gathered fro~
garage operators or by walking through the facilities.
In some facilities, an accumulation analysis may be made by continuous counting of in-and-out vehicular movc:::=.ments. This requires knowledge of the number of vehicles within the facility at the beginning of the cciu.iies, plus ::a
check of the number remaining at the end of the count, in order to verify accuracy of the counting. ln-and-outdac- ~
Parking Studies 33 ......,
for facilities wim automated gates or attendant-controlled gates may be recorded automatically and made available
by the facility operator. This type of study may also include notation of license plate numbers as described in me
nc:xt section.
Accumulation studies of residential parking are unable to determine whether vehicles are parked inside private
garages. Checks are usually limited to vehicles observed in driveways and ac curbs, which makes it impom.nr to
alert data users to the incompleteness of me check. In larger areas, personnel limitations may make it impractical to perform all accumulation studies during a single day. As a general rule, it is desirable to conduct additional
portions of the checks on the same day of succeeding weeks, provided weather conditions and ocher in8uencing
factors are the same.
Whenever possible, me complete area should be covered in me initial occupancy checks, even though this may
mean a less frequent interval for each time check, to establish a contrOl base. Supplemental studies can then be
made at intermediate hours. The accumulation data from studies within a specific area are normally plotted on an
hourly basis to show the changing demand. Exhibit 16-8 shows a typical accumulation curve. It is also customary
co indicate the number of available parking spaces, including any hourly variations that may occur due co differing
operating schedules and regulations. Accumulations may also be shown in tabular form for individl121 blocks and
types of &cili ties.
800
100%
700
90%
80%
1600
&!
I!!
70%
500
60%
....
"400
..
SO%
ll
E
::s
40%
300
30%
200
~
~
u
a.
c
":.
II
1:!
20%
100
10%
0%
6:00
8:00
10:00
12:00
16:00
14:00
18:00
20:00
22:00
Time of Day
While accumulation srudies provide information on total numbers of vehicles by location, they yidd no information
on the length of time each vehicle is parked or on driver destinations. Information on length of time parked and turnover, or number of times each parking space is used during the day, is determined from a license plate check.
i' :
f:
(: ;
J,
I
'"""'"' r-niTII"'\'1
l;l}:li_liOJk(llj .. ',
..
..
. . !:\7~:-
-~
' ....~~~:tt~~~~,(~":;rt~::
~ .. ~.~!tii;Jljil~<
Weather
Cicy
Side of Sattt
between
Sueet
.
Spa-ce and
Regula.do.o
and
Codes: ###: last three digits oflicc:nsc plate number for first obscrnrion of a vehicle
{ Check mark for repeat vehicle from a prior c.Ucuit
- : Dash mark for empcy spaces
Comments
Observer
= _N_umb-:-:-er_
. -:o..:..f_D....:If;"f::-er_ent77'"_Car-::--s-P_ar_k_td_
Number of Parking Spaces
Equation 16-1
Other reasons for license plate checks include accuring data on length of stay, acru.mu..btion, illegal packing and enforcement which requires special notation as to whether a ticket is found on an illegally parked car.
The study hours arc sdect~ to suit the purpose of the study and the liccnsc plate check is most often performed by
a person on foot. Because the study is rdarivdy expensive, a sampling technique is normally used. Several block faces
are sdeaed which typically represent the different time limits of parking found in the study area. Thus two or three
block faces having 1-hour parking, twO or three others with 2-bour 'parking and so on, are checked.
The pcrsonnd requirement is dependent on the frequency of checking. which in cum depends on the parking rumover. Since cumover is unknown at the beginning of the study, it must be estimated from known or assumed characteristics of the situation. For example, if a given curb &cc is regulated by 1-hour parking. checks arc needed at 20- or
30-min. intervals. If a curb face has 2-hour parking. checks would nonnally not be needed more often than every 30
min. but no greater than 1 hour. Very short parking. such as a 15-mia.li.mic, requires checks at 5-min. intetvals. In
the fringe aceas, where unrestricted all-day parking occurs, hourly checks arc generally sufficient.
Data from thc.acru.mulation checks on numbers of cars packed per block arc useful in setting up routes and determining pcrsonnd requirements for license plate checks. Voice recorders have been used succcssfUI.ly in license plite check~
ing, although there arc ~ere limitations and large-lClie opportunities for error in their use. The handhdd computer is
Parking Studies 333
also weful in conducting license place surveys. Data recorded in the field can be downloaded into a patking database
after retu rning to the office.
A field form for use at typical curbs is shown in Exhibit 16-9. The essential components include the identification
of che facility by both code number and srreer names. The column headed "space and regulation" muse have provi
sion for illegal parking as well as legal spaces. This column can also be wed as a curb inventory form. As indicated
on the form, a ponion of the license number, usually the lase three digits, of each parked vehicle is enteted in the
appropriate column. When the same vehicle is found at subsequent checks, the number is not repeated but rather
a check.mark is wed. This simplifies the calculation of turnover, since the number of different cars parked in the
block is being determined.
Other notations typically used in license plate checks include the indication of expired meters and color notes to
idemify vehicles rhat have no license plate. Trucks should be noted. Double-parked vehicles ate identified by puccing
a diagonal line across the particular box in the column. The license plate of the curb-parked vehicle is written in the
upper part of the box, and the license number of the double-patked vehicle is wrirten in the lower pan of the box.
It is sometimes desinble to note full license plate numbers so owners of vehicles may be identified. This information
is wed to check the extent to which store owners or employees are wing curb space to patk their private vehicles. It
may also be desinble to designate commercial and out-of-State vehicles. The license plate numbers of trucks may be
underlined iri areas where the state plate does not give a separate identifying symbol such as a T.nSpccialized notes
may also be made on size of truck and type of deliveries being made. Knowledge of frequency and location of truck
loading and unloading activities are needed in the planning of curb loading z.ones as well as for development of truck
terminals, and the degree of truek activity during variow hours is a faeroe when considering restrictions on such activity to increase street capacity (see Chapter 14 for more discussion on goods movement StUdies).
'
,I..
I
'!
One person can conduce license plate checks for approximately 60 parking spaces each 15 min. This perm!ts coverage
of two to four block faces on each trip after making the initial round, which generally requires additional rime. The
degree of park,ing turnover is a major factor inHuencing the acrual time required and consideration mwc also be given
co sections the checker has to traverse where no parking is allowed. Other administrative dements include need for
rest and food breaks for the checkers, personal security in some areas and need for checkers co act in an unobtrwive
fashion. While ic is difficult co conceal the fact that a parking check is being made, the survey results may be biased
ifmecer feeders" ate aware of the field study. When questions ate asked, the checker should answer them courreowly
and give assurance the study has no relation to law enforcement activicy.
Where widespread license plate checks are being made, the wual method is to start routes at a common point to facilitate supervision. The field checker walks down one side of the street and back on the opposite side, covering all curb
spaces along the route. In a rectangular street layout, routes may be started ac one intcrscccion and extend out around
the block with the supervisor Stacioncd at the starting point. Fidd checkers often work a 5-bour shift, but this length
is dependent on the tot:allengch of the parking survey period.
The license place check is most often used at the curb; however, it can also be used in lots or garages. ln th~e ~~.
fidd observers go through each lot or garage at regular intervals, recording license numbers. Care mUSt be-taken co
follow the same route within the facilicy co simplify comparison of license numbers during the summarizing process.
If the off-street check is made as a part of the curb check, the intervals between trips arc the same as in the curb check.
If the off-scrcct study is made separately, intervals can be increased by 20 to 30 min. or longer since parking durations
in off-street locations are commonly longer chan at the curb. Any underlying schedules, such as nearby school class
schedules, church service hours, or movie theater schedules, should be taken into consideracion.
Where off-street facilities arc too large to check every vehicle in the time allotted, or arc ope.ratcd by people who
will not cooperate in allowing the observer to make the necessary observations, the in-and-out" study is used.
This study method requires observers at all entrances and exits of the facility and gives the maximum accuracy
of turnover and duration data since every vehicle entering and exiting is observed and the license number recorded. In cases where the volumes of entering and/or leaving vehicles arc unwually high, several observers or
video recording may be required.
The cime limits of this StUdy should coincide with those of the curb license place check. Ax suitable intervals, the time
of day is noted following the lase recorded license place number, on both the "in" and "out" sides of the form. When
crafiic volumes are low, the ~c of acrual entry or exit may be recorded for each vehicle. During high-activicy periods,
lfll~ hmrn_o
"' te-
I!''''
,.:-~
., -~"~TI;;!"
~ - -~1r?I~~~\~':;
. .~. ./~ :~:. .-:<'"rA-~~~~-t-.. ,:t"';_.~~' ~-:. - -ittf-Jt:.--r:;
Weather
Scm lime
End1ime
Location
Passenger
Vehicles
Truckt
Passenger
Vehicles
Truckt
Passenger
Vehicles
Truckt
Passenger
Vehicles
Trucks
Number
Pera:D.t
Number
Percent
Nu.mbcr
Percent
Number
Percdlt
Total Vehicles
Total Vehicle-Hours
Average DIIIlltion
Percent Overtime
Comments
Observer
time should be noted at 5- co 10-min. intervals. When possible, the number and license plates of the vehicles in. a
facility at the beginning and at the end of the study should be checked.
Office surqmary and analysis of the license plate field check can provide information on parking accumulatio ri.
The peak accu mulation for each block face is determined directly by a co unt of the parked vehicles. Exhibit 1 610 shows a typical dwation summary form. The column headings may classify data by blocks; by type of parkiPg
zone (such as !-how curb parking, 2-hour, etc.); or by curb, lots, blocks; and grand total for the area. Park.iog
durations may be estimated from the round-trip times of the parking activity checks. For example, if trips a..re
made at 15-min. frequencies, a vehicle found on only one trip is assumed to be packed for 15 min. Ifitis seen
on two successive trips, the vehicle is assigned w a 30-min. duration group, and so on. When the vehicle is seen
on the first or last trip of the observer, its parking dwation is unknown. All durations should be consid(red :as
starting within the time limits of the srudy.
Total vehicle hours arc calculated by dividing the total number of vehicles seen on each trip by the number of chcclo
made each hour. Thus if 10 vehicles were observed three times on 30-min. check intervals, the nwnber of vehicle
hours is (3 x I 0/2) or 15 vehicle hours dunng the 1.5 hours of the study. The average duration is found by divi~g
total vehicle hours by the total number of different vehicles observed (15/10 = 1.5 hours). This average is slight:J:Y
higher than the true mean value, because all packed vehicles are not observed. The percentage of overtime parkers J.S
found by adding the figwcs foe all durations in excess of the legal parking time limit.
Exhibit 16-11 shows a. typical rucnovec summary form tha.t may be used to swnmarize the turnover for eac::ll
block; for each classification of parking zone according to time limit; for lots, garages and cwb space; and for
different areas. A summary is made for individual blocks; for classification of time zones; for each lot or garage;
Parking Studies 33 .!5
vv~mer
_______________________________________________
_ _ _ _ _ _ _ _ _ _ _ _ _ _ __
End lime _ _ _ _ _ _ _ _ _ _ _ _ __ __
Location
Total
Facility
Block
Number
of Stalls
Tum.nu
Vehicles
Hourly
Turnove.r Rate
(vehicles/staWhour)
Conunents __________________________________________________________________
Obset'Rr
for all curbs combined, all lots combined and all garages combined; and finally, for a grand total. The left-hand
column of Exhibit 16-11 is used to indicate classilicacion. The next column contains the number oflegal vehicle
parking stalls that are being used all or part of the time on each normal weekday. The number of vehicles observed making use of the parking stalls is entered in the third column. Any vehicle making use of a space whether
or not it entered or left during the period of the study is included in the turnover. count. The fourth column is
used to record the hourly turnover rate per stall. For instance, if 50 stalls had an observed total of 100 different
vehicles in 8 hours, the rate would be 0.25 vehicles per stall per hour [100/(50 x 8)].
The three remaining blank columns in Exhibit 16-11 may be used for various purposes. For example, it may be
desired to show probable future number of spaces, allowing for new off-sueet facilities and prohibition of curb parking. and resulting possible turnover. Or it may be decided to show what turnover would be if the turnover rate were
increaSed through enforcement of existing time limits. Column headings might be, in this case, "estimated turnover
rate assuming compliance with time limits." "total turnover" and "increase in turnover compared to pr~ent." Daily
or study period turnover is usually shown. Continuing the example, 100 cars parked in 50 stalls would yidd a daily
turnover of 2.0 for the 8-hour period.
~'2~
.a
.,ll..lr"'\ C:t'\ITII"'\Al
)The postcard interview is used to determine charaaeristics only at the peak parking hour and the handout o r place'{nenr of cards u.nder windshield wipers shou.ld be done in a short time span. The time required for card ~istribucion
sbou.ld be determined experimentally before the stu.dy begins. Newspaper publicicy is essential for any cypc of inter
view stu.dy, especially when postcards arc being used, and every effort shou.ld be made to obtain maximwn coopcra
tion from the public. The personnel rcqu.ircments for the postcard survey arc much lower than those for the p ersonal
interview; however, the low rate Q.f returns and the unknown degree to which dara are biased are nevtive aspects of
card usage. Exhibit 16-12 shows a cypical parking-survey postcard.
The
Department ofTransportation is conducting an analysis of the parking conditions in this area. -we
are asking each driver to complete the following questionnaire. Your answers will be kept anonymous and da~ will be
summarized ~o that .the anal~is and n:pott will in no way dw.wc da~ on ind.ividuals. We thank you in advance for 6.lling
this brief survey. Your time is valuable, and the answers vou l!ivc will hdp us imorovc parkine: in the area.
l. What is the primary 'purpose of your trip? (check only one)
OWork
Q Tourism/R.caeacion
Social, eu:.:)
QSchool
QShopping
QOther
!~cion?
---------------===
hou.rs
---------------------------------- minutes
4. Where did your trip stut (where did you come from just prioi to parking ~ere)?
Nearest street intctsoetion:
in this area:
Typically. 30-50 percent of the cards distriburelar~ returned. To expand the sample, it is necessary to have a record
of the nwnber of cards placed, by locarion and rime. For example, if five cards were returned from a given b lock h.~
indicating these parkers shopped at a particular store, and the rerur;n sample from this block &ce was 50 percent, it
cou.ld be estimated there were a toeal of 10 parkers at the peak rime ,..ho actually went to the particu.lar store. Because
the sample of rctu.rn.s wou.ld be expcaed to vary from facility to facility, a separate calcularion of =pie size and ex
pansion &ctor shou.ld be made for each parking fa.cil..ity.
One method of maximizing the return is to take special care in selecting qucsriotts so as not to ask for informacion
some drive~ wou.ld be rductant to reveal. The best illustrarion concerns the home address, which should never be
asked since .many persons do not want to give out this information and it is of no direct value in the parlci.ng survey.
The needed locarion is some general part of the city or the name of.the cicy for parkers from subwbs or other nearby
communities. Appcnaix B shou.ld be cottsu.ltcd for additional suggcstiotts.
To maximize the return of the postcard interyicws, pastage shou.ld be included on the postcard. To avoid paying for
postcards that are not recucned, business reply mail postage can be used for the nudy.
Another method of contacting drivers is to record license plate numbers of parked vehicles, trace mailing addresses on
the basis of these numbers and mail return postcards or quesrionnaires to the owners. However, a rime delay of several
days is caused by lookup of nwnbers, mailing and ~ccipt of card by the vehicle owner. This resu.lts in fewer returtts
and less a.a;:ucacy.
P;u~inn
'\tuniP< ::1::17
":,!'-
~t.~;p~fii,
N-
Wather
Dace
SwtTime
End Tune
Location
1ime
Start (Parked)
Trip Purpose
(Work, Scltool,
Shopping,
Personal, Other)
Trip D estluation
s~t
Walking
address or
Bwiness Name
Hours
Minutes
Distance
Commenu
O~r
Flyers can 21so be used co direct drivers to a Web-based survey. A Web-based survey can reduce data collection costS.
However, the survey might suffer from sdection bias from drivers. An incentive for participation can increase the
quantity of completed surveys.
~.
Personal interviews of parkers can be conducted at the curb or in a parking facility. They can also be conducted at the
entrance or exit of a particular generator, such as department store, office building, or hospiral. The exact questioru;
vary on whether the interview is conducted at the origin or the destination of the parker, at the time of parking. or
at the time of unpacking. Exhibit 16-13 shows a typic21 form that is used when the interview is being conducted at
either time of parking or unpacking. Questions include uip purpose and destination of the parker. Interviews at an
apartment garage would add "reside here" as a trip purpose. The parlter may be asked to give an estimate of length
of time parked. The form shown as Exhibit 16-13 docs not include space for trip origin; however, this can be added.
h
:;.
Person21 interviews
be used on a day-long basis, during sever21 hours of the day, or only during the peak hour.
The work is not only C:xpensive bur also personally demanding, a.nd careful selection a.nd training of interviewers are
essenti21. Persons with previous c:xperience in census taking are especially valuable for this activity. Ea.ch interviewer
should wear a distincr armband, ribbon, reflective vest, or other identifying insignia ro minimize confusion, apprehension, or antipathy on the pact of paclters being approached for questioning. The words "Traffic Survey" or other
suitable information should appear in large letters on the insignia.
'
1'
...
li
!;i;
.: .
: The interviewer should be on location 15 min. in advance of the starting time. One person can handle aboor 15
: spaces, provided rurnover is not too rapid. Thus one or two persons are usually needed to handle one block face. In
'. the case of a large off-street F.!cility, it may be necessary to sample parkers by interviewing every second or third pe.r
'son. The study normally does not include vehicles stopping momentarily to discharge passengers, taxicabs parkin!? tn
reserved curb zones, buses and emergency vehicles. Truck drivers may be interviewed, depending on th e informatiOn
that is desi.red. Interviews are directed only to the driver of each vehicle.
Vehicles parked when the interviewer first arrives are recorded on the field form with an "inn time 5 min. prior tO
beginning the official survey. The drivers are interviewed if they return to their vehicles during the survey period, and
it is usually then possible to determine the approximate actual time of parking. The original entry is then deleted and
replaced with the correct time.
Normal procedure calls for the interviewer to approach a vehicle quickly as it puUs into a parking space and record
the "time parkedn to the nearest minute. The interviewer then asks the driver the selected questions and noces ch~
answers on the 6dd form. When the driver rerurns and begins to unpack. the interviewer records the "time leaving
on the form if the survey is being made along a block face or :a a point where the interviewer can see the unpulcin g
activity. For such cases the 'interview form should also include a column for license plate numbers. For curb checks
the interviewer records the time leaving only on the same line as the time parked if the license number can be found
quickly on the form. If not, only the license number and the time leaving column are filled out on the next empty
line, and the cross-matching is done later in the office.
If the same driver parks several times in the same block during the study, he or she must be interviewed separately each
time. If the driver refuses to cooperate or the interview is missed for other reasons, the interviewer writes "refused" or
missed" in the purpose" column. Sometimes, trip destination and purpose can be determined by observing wbc:r e
the drivers walk and which building they enter.
Other ~abula.r information, such as ttip purpose related to length of time pa.rked and distance walked related to both
trip purpose and facility where parked, will also be needed. Depending on the applio.tion of the study, such as the
dcvdoprncnt of a &asibility report for the construction of new revenue-producing pa.rking &.cilitics, it will be neccssasy to determine the number of parlrus by both trip purpose and parlcing duntion for each block. A va.ricty ofspecial
forms may be wed, as shown in Exhibit 16-15.
..
!> .. "
.. .
""
..
.~~~~ ~~~~~-..,"::'~;~~~~
Type of Facility
..
Non-Metered Zones
Tourisaal
RKreatloo
School
Shopping
Pcnonal
BIISioCSS
Otbu
Unknown
Metered
(30-Minutc Zones)
Metered
( 1-Hour Zo.nes)
Etc.
Type of Facil.ity
< 100
1,300 1,700
lOOto SOOto 900to
to
to
499
899
1,299 1,699 2,000 >2,000
AYC.-.,e
Unknown
Total
Di.staace
Non-Metered Zones
M=rcd
..
(30-Minutc Zones)
Metered
(l-Hour Zones)
Etc.
Durarioo
< 100
1,300
to
1,699
1,700
to
2,000
Avaase
> 2,000
Uob.cnm
Tocal
D istaace
0 to 14 Minutes
15 to 29 Miitutt:S
..
'
30 ro 60 Minutes
Etc.
Walldoa
Distance
(Feet)
I
Tocariam/
Work
Recreation
Shoppins
School
Pcnooal
Bu.sincsa
Other
Uobon
Ecc.
Cmtinud on nca fllfl!
-
. .. .
,,
s:-r.. .. -;:,
.:...a . . .
'<l
..
Trip Purpose
Duration
Work
Tourism/
Recreation
School
Shopping
Personal
Business
Other
Unknown
Total
Unknown
Total
0 to 14
Minutes
15 to 29
Minutes
30 [0 59
Minures
.Etc.
Block .
Work
Tourism/
Recreation
School
Shopping
Penonal
Bwiacss
Other
Vehicle Stalls
Maximum Expected
Total Turnover
Number of Driver
Ttunover Observed
Desti.aations
Continu.!d on nextpage
Source: Box and Oppenlander, 1976.
Block Number
Numbers
of Vehicles
Space-Hours
Parked
Used
Driver
Destination
Space-Hours
for this Driver
Destination
5.0 SUMMARY
The analysis of parlcing studies should lead co the evaluation of current parking techniques and the potencial mitigation of parking supply issues with various management techniques. A wide variety of parking management techniques
are available and must be carefuUy selected for each unique situation or parking issue. A lise of strategies and their
respective typical reduction in required amount of parking, and whether traffic volumes are reduced as a result, is
shown in Exhibit 16-16.
Shared Parking
10-30%
Parking Regulations
10-30%
10-30%
./
Walking and Cycling
Improvements
~15%
Increase Capacity of
Existing Facilities
~15%
Mobility Management
10-30%
./
./
Parking Pricing
10-30%
Improve Pricing
Methods
Varies
Financiallncemives
10-30%
{/nbundle Parking
10-30%
~15%
Improve Information
and Mark.eting
5-15%
Improve Enforcement
Varies
Transport Management
Associatio'n
Varies
./
./
./
./
./
./
./
Varies
Varies
Varies
6.0 REFERENCES
~-1 Literature References
American Association of Stare Highway and Transportation Officials. A Policy on Gtommic Design ofHighw4J$ and Street:s.
Washington, DC: AASHTO, 2004.
Davis, A.Y., B. C. Pijanowski, K. Robinson and B. Engel. "The Environmental and Economic Costs of Sprawling Parlcing
Lots in the Uniced Scares. LanJ Use Policy fournJl/27, No. 2 (2010): 255-261.
Federal Highway Administration. Advanmi Parking Management Syrtems: A Cr11u-Cutting Study-Talting the Stms Out if
Parking. Washington, DC: FHWA, 2007.
Federal Highway Administration. Manual on Uniform Traffic Contr11/ Devicesfor Strtet:s and Highways. Washington, DC:
FHWA.2009.
Federal Highway Administration. Origin-DestinatWn Suroep. ~ashingron, DC: FHWA, 1970.
Insticure ofTran.sportation Engineers. Parking Gmtrtztion, 4th ed. Washington, DC: ITE, 2010.
Instirute ofTransportation Engineers. Traffo Engineering Handboolt, 6th ed. Washington, DC: ITE, 2009.
Instirute ofTran.sportation Engineers. Trip Genenuion, 8th Edition. Washington, DC: ITE, 2008.
Litman, T. Parlting Management: Srraugies, Ewtlwztion and Planning. Victoria, BC: Viccoria Transport Policy Institute, 2008.
Syrakis, T. A. and J. R. Plan. "Aer~. Photographic Parking Srudy Techniques. Highway Research &cord 267 (1969): 15-28.
Urban Land Insrirute. Shartd Parlting. Washingoon, DC: Urban Land lnstiruce, 2005.
Chapter 17
~
e ...
.. e
..... .
.....
e ... ...
ltiTRODUCTION
'
347
347
348
3. 1 Collision Reports
348
352
354
355
356
357
358
5.0
369
379
REFERENCES
379
1.0 INTRODUCTION
raffic collisions exact a terrible toll in the United Scates and ocher countries. In 2008, 5.8 million motor vehicle
traffic crashes were reporred in the United States alone (NHTSA, 2008). The collection and analysis of data o n
traffic collisions is fundamental to the design of programs to reduce chat toll. AJWyscs use collision data ro bel p
understand why collisions occur, co hdp identify collision-prone locacions, to aid in deciding which safety progran"ls
or countermeasures should be implemented and to assist evalua.cions of countermeasure effectiveness. In addiciot'l
certain programs of the U.S. government require that parcicipacing agencies have a program of collision data analysis.
This chapter discussd techniques for conducting studies with traffic collision data. Fl.rst, the collision report forms are
described along with sources of collision daca available co the analyst. Further discussion is provided on how to reduc~
and merge data sees to prepare for analysis, along with typical problems and limitations that come up when prcparin.g
data. Next, studies are discussed with particular emphasis on uends, hazardous site identification and the possibl~
countermeasures implemented at those sites. Finally, specific types of analyses ue briefiy described.
n=s in the United States are d.iscusst:d in this cbapter. Some analysis techniques are nOt mentioned in this chapltt because tbcy
require cxpecti.sc, dfon and/or equipment nor readily available to most jurisdictions and engineers. The premise of the
chapter is engineers who wish to analyze uaffic collision data that have been collected by others, principally the police.
Thus, the chapter does not cover techniques for investigating individual crashes. However, engineers working with
coUision data should become familiar with crash investigation and reconsuuction to provide greater understanding
of the data.
Analysis of trends in crash types, time of day and weathe.r conditions should be distinguished from analysis of detailed
crash causes, which is often a secondary exercise for practitioners who may be involved in an expert witness role. Cer
tain generic causal factors, such as driver inexperience (younger drivers), slow reaction times (older drivers), drinking
or drug impaired driving and distracted driving such as ceU phones, tcxting, global positioning system (GPS) unit use,
etc., should be reviewed if compiled. Standard crash reports, at least at present, do not require all necessary data. Most
of the core causal factors must be debated and resolved on a larger policy level. Traffic engineers' predominant role is
to design the most forgiving transportation system they can within budgetary constraints. The engineers' mission is
mobiUty for all and their mandate is safety. However, when a significant causal factor is present in a large proportion of
crashes (for example, distracted driving due to electropic devices or an increase in elderly crash rates), engineers have
an obligation to nuke recommendations to polieymakers th:u could reduce their occurrence.
Several new traffic collision analysis tools have been prepared in recent years. A primary safety resource in coming
years wiU likely be the American Association of State Highway and Transportation Officials (AASHTO's) Highway
Stlftty Manual (HSM, 2010). The analysis methodologies presented in the HSM are robust and will likely become
srandard for safety analysts in the future. In addition to the HSM, the Federal Highway Administration (FHWA)
distributes a useful software tool for safety practitioners called "Safety Analyst," which is based on the m~ules pre
sented in the HSM (FHWA, 2009). Analysts should become familiar with these new tools, as weU as the availabl~
tools presented in this chapter, when preparing to conduct safety studies.
!Jt:~~~~=-::mw~~OETERIIINAT10N$0f
__,_
(1=T:-
__......_
-"------------------o:=r:__......._
.J
Contin~d on .nexJ pag~
s'
..
'ifiil'll ~"
~ ,>.;.~,,,_.. ,. " , ,
:'I)U-,j.
, ..... ...
o -~
t; 011 Tli
~ ., , . __ r 'l'- , _ .
~-
.. _,
: . .
--Uo)
'ltiUCLEINFO.
.,..,
A ;
__. _ . .
...,
:"'''!~ZOflfRELATEli '.'
ROAOWAYINFQ,
w..i;.i:iioi';r;,:;-~--" >~-- ~
. .. . ,
-,;
,. .... a..-
"-~~:{:-:): ""'
u r.. ~eatn--C.'l.)
lt,..,O.k.lb
n Pbal~ Tp
....
u ....-tl<G!fo~Pfb'-'
04 llillloT, _....,_fP.l
n-~
lltmllf9. _R-UlWdforNIW
~v.at:t~ U.
nAa:~essc-d
~~~i~:t._,;
,. r!d:c.."'r'"
jcRASH SEQUENCE
-.vc.ta~
tUt:Mti~~
IUokiA'II~ ~. . . . .
f
I
--~,,
~-..-
. .. ; o
.r.P~'f!t!. 4
, -~'! . ~.
a truH~StcDidEwt'll
.. <mOs._...-~~
to l!sioolielo.iolwlf"'""Spud
"~or--arqaa
PG~tC.fhtl'Y'ct.clt.b:l3
0
0
0
os-e.;.r-alwro
UOW.!--A-~[\ft
~;'j,;ifi:f ~~;_) ~
n li*CtalfO,.
"..,...,,.,.,_'(>._
u~DIIIctl
:::~"~~~d~A~:?r~~~..B.:~:~~:~i\~~;..,: 1;
+.
, ' . . ,
,
...: ;
UrDODD
~-.o -r..., r w
u.u
......................
MIRATM. ,....,...._._.....,_
u-
-o - r"*"-
o'
o'
---
'<
..::_.
0
DODO
'".. --------------~
Qu
...:.
::::.:::~:
..;;;~~- r$_. ..
-t;~.,t:; ::.:-;~-?1 .;~:,
n~ .. l.a"Ja
0
0
0
.r .. .... ""-: -! . :
.~
_,_
't>il**"
Oiiol-
~~~?k~~tfuMt<k.~Qf,J}~{1~~tf,<:~:'"~:a;?',N.<'
'' ~~;; .,. !;;!f' ,, ,,;;;.;ll
'--~- . _ '= .~ ~~--~ -~ "'.".! 1y 0 ,:_ 1... .., ~~i-;:._
:r'Jo:.. .. ~.j.<:-:;:~;..:,;_;4:;.%~." -=?"..':-:~~:~- ~:::t~~--~-.- . ~~-;--.:i3:~
' -. ~
Collision Type
Non-Collision
S~cific Collision
Code Number
I
J:u:kknif~
Overturn/Rollover
Orner Non-Collision
Pedestrian
Ped~cydist
Railway Train/Engine
10
Animal
11
Movable Object
12
Fixed Ojea
13
14
15
16
17
18
19
20
21
22
Angle Collision
23
Backing Up
24
__
_J
----
Crash type is one of the most important fields on the report form f~r traffic engineers. Various jurisdictions code collision type in various ways. Some states co<le the collision type when the report i.l being computerized in the cenc.t":al
office, based on the diagram and narrative on the form. However, most states rely on the police offu:er to co<le the collision type. The collision co<les used by police in North Carolina ace given in Exhibit 17-2. Police in North Cacoli.c:1 3
record coUision type in terms of the first harmful event and the mosr harmful event for each of the vehicles involved.
The person filling out the report has 24 choices, including several choices of other." They could co<le the collision ~s
a noncoUision, collision of a motor vehicle with some other object than another vehicle, or a coUi.lion between mot:Oc
vehicles in the traffic "Stream. For motor vehicles that collided in the traffic stream, most states employ a typical co Cling system for collision cype, including right turn, left rucn, rear-end, head-on, sideswipe, angle and backing. So~ e
collision report forms include separate co<les for sideswipe coUisions between vehicles moving in the same directio J1
and vehicles moving in opposire directions. However, some do not have co<les for different categories of left- ao-.::!
right-tum cotli.lions.
The driver and passenger injury co<les ace very important for many studies. The most common coding scheme [C" $
the extent of injuries incurred by panidpants in a crash is the "FA.BCO" or "KABco scale, which includes 6.,.,..-.c
categories:
1. F (htaliry) or K (killed): The person died within 30 days of the collision as a direa result of injuries received
during the collision.
2. A:. The person experienced serious, incapacitating, non&tal injuries during the collision. Broken bones, massive
losses of blood, or more serious injuries are rated A.
3. B: The person experienced a visible but not serious or incapacitating injury during the collision.
4. C: The person complained of pain or momentary loss of consciousness due to an injury during the collision,
but no visible sign of injury was evident to the investigator.
5. 0: No injury, which includes "PDO-Property Damage Onll collisions. These can be significandy underreported since they are often handled between the drivers of the vehicles (or by the driver striking the obstacle)
without the assistance of the police.
Researchers have developed other scales of injury with finer gradations than the FABCO/KABCO scale; however, this
scale has gained mainstream acceptance on standard collision report forms. An c:nrople of another injury scale is rhe
Muimum Abbreviated Injury Scale, discussed by Miller et al. (1985), which rates injuries from "1" minor injury
through "5" =critical injury and "6" unueatable, immediately htal injury.
Another interesting and useful code on most collision report forms is the vehicle identification number (VlN). The
YIN is a 17-digit number unique to each vehicle that is assigned to most automobaes and other motor vehicles.
Coded within the YIN is information such as vehicle weight, wheelbase and engine size. The VIN provides information about the vehicles involved in collisions that is not otherwise recorded on the report forms. Computer software
is available to decode the YIN for further a.nalysis.
.
'.
In some jurisdiaions, other data are available in addition to, or instead of, a police collision report, such .as individual
collision repons of drivers and/or witnesses. A driver or witness form is usually shorter than a police collision report
form. For some collisions, police or other investigators collea supplemental information. Photographs, sketches, calculations, notes and other information may be on file, especially for more severe collisions.
Police officers usually complete collision repons during the shifr they investigated the collision. Once an officer su'bmits a report, a supervisor or other personnel may check it and rerum an incomplete or obviously inaccurate report
to the officer for correction. The proportion of police collision reports completed on a computer and submitted
elearonically grows each year. Although every attempt is made to correa ertors in the collision reporting, some can
and do still occur. Analysts should review these reportS critically and with caution as they have been prepared using
obseiY.lcions and statements of wimesses who are prone to error and often subjective.
Although not considered part of the collision report, another useful tool becoming more popular for crash analysis in
many countries, and particularly accident reconsuuaion and investigation, is the "black box. This hardwue device
is on most new model vehicles and stores data on events just prior to the collision event (SWb, 2008). In the event
of a crash, the boxes will give clues as to why the collision happened. However, many other countries have not moved
further into this arena because of fears the box could infringe on drivers' civil liberties.
already time-consuming collision rep<~rting process, many entities require separate filing of collision rep<~rts for insu ranee companies, adding additional constraints.
'Most agencies can quickly provide analysts with copies of, or electronic access to, requested collision data via software
designed to poll various parameters noted by the analyst. Stares will erase data items identifying individuals to prot~r
privacy when providing data to outside agencies, a private company, or an individual. Collision report forms are cypically available 4 to 6 months following a collision, but can be available earlier in many jurisdictions depending on the
method of data entry.
There are various ways in which collision data in a collision report form relate location information. In most jurisdictions, police and motorist report forms ask foe the name of the highway where the crash occurred and the distance
and direction of that point from some known reference p<~int, such as an intersection or milepost. The fo rm may
also request the address of the collision location. The informacion on the form is usually translated into the standard
location sysum of the state at the time the data ace entered into the computer. Instead of a highway name, the agency
may enter a code number identifying the highway. Instead of the distance from the intersection nearest the collision,
the agency may r~rd ~e distance from the beginning of the highway or the distance from an established reference
point such as a major interSection. Many police agencies are now geo-locacing crashes wi.th global p<ISitioning system
(GPS) devices installed in police vehicles.
The U.S. Department ofTransporration (US DOT) maintains electronic collision databases that are useful for various
studies. Bdow is a list of commonly-used databases along with descriptions.
FARS: The National Highway Traffic Safety Administration (NHTSA) maintains the Fatality Analysis
Reporting System (FARS). FARS'is an ahnost complete census since 1975 of all collisions in the United
Stares that have res(!!ted in one or more fatalities.
GES: The General Estimar~ S}'Stem, which is also maintained by NHTSA. contains records from detailed
independent investigat!ons of a small, statistically selected sample of collisions across the United States.
HSlS: The Highway Safety Wormation System (HSIS), developed by FHWA. consists of nine relatively
high-quality databases from geographically scattered states based on the quality of data available, and states
ability to merge data from various files. The collision data from the police report have been processed to
correct some errors and to allow easy merges with rra.ffic, roadway and other .files also maintained within
HS1S.
MMIRE: Modd Minimum Inventory of Roadway Elements (MMIRE) is an FHWA undertaking, which
includes a listing of roadway inventory and traffic dements critical to safety management, and proposes
standardized coding for each. MMIRE serves as the companion to the Model Minimum Uniform Crash ..
Criteria (MMUCC) which was previously developed to sca,ndardize crash data variables used by states and
local'jurisdicrions to improve crash data systems.
SAFETYNET: The Federal Motor Carrier Safety Administration (FMCSA) designed SAFETYNET to
manage and provide appropriate access to crash data, roadside inspection history and data, and motor
carrier and shipper identification informacion. Dara records include, but are not limited to, truck/bus
driver name, driver social security number, driver license number, driver date of birth, driver and company
contacr information and VIN.
In addition to state and federally funded collision databases, other useful supplemental data sources could be used.
These could include:
DMV: State Department of Motor Vehicles databases typically collect records ofVIN, license plate number
and owner's names.
Judicial Records: Many jurisdictions keep records of criminal infractions for some minimum time period,
cypically "r7 years. Data could be used to identify patterns of driving and vehicle infractions.
Hospital Records: Most emergency medical service agencies and/or hospitals keep data on every patient
serve<l. These data ace typically subjecr to stricr privacy policies.
Mapping: Most agencies have geographical information system (GIS) programs in place that have various
layers of data that could be useful co the safety analyse. ln the GIS, basic data on roadway geometry may be
linked co roadway inventory d.:ua on speed limits, signagc, signals, barriers, etc., as described in Chapter I 5.
Traffic Volum~: Tht rnajority of agencies collect traffic dara using various methods such as permanent
auwmacic traffic counters, pneumatic tubes, or manual turning-movement counts as described in Chapter
4. Traffic volumes are used in collision studies co capture uexposure," or the chance that a particular entity is
involved in a crash.
One of the lim steps in this reduction process is to choose an analysis time frame and discard oudying data. In some
studies the choice of a time frame is self-evident (for instance, the study is to see whether the collision rate in 2010 cliffered from the rare in 2009). When analysts must choose a rime frame, 3 years is the most common choice. rtu"ee years
represents a compromise becwtW the desire for larger samples and the desire for cirneframes within which conclitions
were unlikdy co have changed a great deal. Also, a 3-year cimcfrarne eliminates many problems with cliscarded daca.
Timeframes of up to 5 years are common, but for periods longer than 5 years, analysts must use special care co ensure
changes in background conclitions (traffic volume, land use, traffic control, geometry, ere.) can be tolerated within the
scope of the study. T uneframes ofle:ss than 2 years may be necessary, but a smaller sample siu could constrain the study.
It is good practice to check that collision data for parycular locations have not been biased by construction or ocher
major but temporary traffic events during the cirnefraine selected. Analyses can consult highway agency records for
construction and/or project lectings, but such records are often incomplete or inaccurate and cartnot be trusted alone.
Supervisory and/or experienced area highway personnel can estimate whether major construction projects or other
aaffic events have taken place, or at least give some hint of construction which can be further investigated. However,
those estimates should not inspire great confidence without corresponding documentation. A scan of the collision
data co see whether any reports mention construction or work zones is often reassuring.
3.3.2 Analyzing Specific Locations
Many sruclies focus on one location or a limited set of locations on the highway nerwork. One of the analyse's important casks in these srudies is to reduce the database to crashes chat occurred at locations of interest. Analysts usually
summarize collliions into those that occurred at spors and those that occurred in roadway sections. Spots are short
segments ofh.igbways that help identify problem "point locations such as intersections, curves and short bridges. The
highway cross-section and ocher fearures at a spot should be noticeably different from surrounding spots. Sections are
longer, relativdy homogeneous segments of highway convenient for studying cross-sections, pavement surfaces and
other longitudinal fearures. Roadway sections will rypicalJy correspond to tangent sections of roadways. Spot lengths
of 0.2 to 0.3 miles (.32 to .48 kilomece~ [km]) and section lengths of 1 co 2 miles (1.6 ro 3.2 km) are recommended
for most agencies (TRB, 2010).
Spots and sections that Boat are usually more desirable than chose that begin at fi..wl increments. For example, suppose
that many collisions occurred near a highway fearure at the 0.6-mile (.97-km) mark. Ic is likely the police would code
some of these collisions as occurring at the 0.5-mile (.8-km) mark and some as occurring at the 0.7-mile (1.13 km-)
mark. If the agency used a fixed spot length of 0.3 mile (.48 km), a spot would end at the 0.6-mile (.97-km) mark, so
SO!Jle of these collisions would fall in one spot and some would fall in the next. By contrast, a Boating spot (sometimes
referred to the sliding window") definition might have 0.3-milllong (.48-km-long) spots beginning every 0.1 (.16
km) mile, overlapping each other. Thus there would be a spot in uding all collisions coded as occurring at the 0.5-,
0.6- and 0.7-mile (.8-, .97- and 1.13-km) marks, and the agen would obtain a more accurate estimate of the uue
number of collisions at the feature location (TRB, 2010).
354 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDffiON
Am.lyz.ing crashes ac incerseaions requires a precise defini tion of the limits of rhe intersection. Some agencies include
coLlisions within an intersection if they were coded as occurring on che ap proaches, up ro a fixed distance usually In
the range of 150- 500 feet (ft.) (46--152 (meters [m)) from the incerseccion depending on such factors as the :unoum
of congestion, queue spillback, the length of rurn bays, or the length of the approach legs in relation to nearby iote~
secrions. A more rime-consuming buc technically defensible definition, especially in areas with chronic congestion. IS
to examine each collision on an approach co determine whether the intersection was an influencing factor. With the
latter definicion, the intersection size varies and basically extends co the back of the queue on the appro aches.
Many agencies now provide a geo-coded reference point for every crash location if the exact location is known .In
addition, in many agencies, Incerner and reporc-based informacion is provided on various regional crash rates for stgnalized and unsignalized imerscccions and roadway links by functional class. If data is provided in this format io your
district, it may be easily uti.lized; however, caution should be used to make sure rhe method for ranking is knowll up
&one. For instance, if a floating method is noc used along links, it might not be advisable to use chis type of data for
choices such as countermeasure deployment.
entire
...
")~If'\ S:niTlt"'U\1
delete the offending collision report or delete the offending value. If the error occurs in an important variable or if a
limited sample of collisions is available, the analyst may need to consult a hard copy of the collision report and read
the narrative or consult the drawing. Consider deleting the entice collision tep<lct only if the error is so gross that the
'collision is most likely irrelevant to the analysis (for example, the collision occurred in an'other city) or if the prograrri
Bags many errors in the same record.
Some errors ace important for many analyses and should be checked routinely. Errors in collision location ace common; however, GPS units are helping eliminate this type of error as they ace becoming more commonly used in police
cruisers. Distance estimates ace a prime culprit, but misspelled street names, confusion involving multiple street names
and numbers for a given facility, and unusual geometric features such as freeway inccrchanges also cause location errors. Manual double-checking of collision locations is sometimes necessary. The collision-type Variable is very useful
but often erroneous. At intersections, vehicle intent codes may be more accurate than collision-type codes because they
ace less ambiguous (Hauer et al., 1988). Using vehicle intent codes at inccrsections provides another advantage as well,
which will be discussed in a later section of this chapter. Variables on the collision ICp<lrt form where police officers,
drivers, or wi messes describe highway features ace also notoriously erroneous. If analysts need more than very simple
information on rugpway features for a srudy, they shoulc;llook for it from a source other than the collision record, such
as a visit to the collision site or througb a supplemental database such as in a GIS.
3.5.4 Collision Ozuse
Engineers often think of the events leading to or causing a traffic collision as a causal chain, where the removal of a link
in the chain would have prevented the collision. Links in collision-ause chains might include the decision to make
a trip, the choice of a route, vehicle defects such as worn tires, slick pavements due to rain or snow, objectS blocking
driver vision, distractions in or outside the vehicle, or the presence of fixed objects on the roadside. Vehicle and highway factors ace usually present on cause chains, but driver faqors ace the most common links.
Most crash reports contain oruy the last ~d most obvious links of the collision cause chain. Police who ace investigating collisions do not have the time or, in mmy cases, the training to seek links farther up the chain. Police often use
favorite standbys such as driver inattention when other causes ace obscure. Most analysts believe driver and witness
reports ace virrually worthless for providing crcdlble information on collision causes. Indeed, it is very difficult for
even highly trained investigators with sufficient time to construct m accurate chain. Thus, in the case of a singlevehicle, run-off-road collision, the object that was struck on .the roadside will be prominent in the report:. Analysts
may tend to seek treatments for this "cause" while ignoring driver, vehicle and highway factors that preceded the
vehicle leaving the road. That tendency can be resisted by searching the collision n:p<~rt for other clues to links on the
chain that are not prominent. Another way to attack the collision-cause problem is to malyze smaller samples of more
detailed collision data when designing ways to break the chain.
'
3.5.5 limit4tions ofthe Report Form
Collision r~port forms have resulted from balancing competing interests. One pf the major interestS is the police;
whose time isrequired to complete the report. The report form thC:CCfore is likdy to be missing dements that traffic
engineers would like to use. Usually, traffic engineers consider infor1nation on the highway to be insufficiem on collision report forrns. Information on the vehicle is also lacking for some analyses, especially analyses involving trucks
and other large vehicles and their contents. Driver and passenger information on collision report forms is usually more
than adequate for traffic engineering uses. Analysts should check the standard form md the coded collision files to
ensure the informacion needed has been collected before launching a study.
correcting crash-prone drivers or vehicles can, in most cases, use the same methods while substicuting the words
"driver" or "vehiden for "location.~ Engineers arc typically more interested in focusing on countermeasures that can
reduce the likelihood of future collisions in the physid environment (for <:xample, sight lines, grades, speed limits,
ere.) and nor collisions focused on things char cannot necessarily be controlled (such as state of mind, cell phone use,
etc.) by typical coumerrneasures. In addition, countermeasures focused on the physical surroundings are more easily
anal}'led because collision records typically focus on causes in the physical environment.
RSEC
= 1,000,000 * A
365 * T V L
Equation 17-1
where:
RSEC
For highway sections, collision rates arc typically expressed in collisions per 100 million vehicle miles. In this instance,
RSEC should be calculated using rhe constant 100,000,000 instead of 1,000,000.
For pot:s, agencies usually calculate the collision rate in terms of collisions per million emcring vehicles using Equa
tion 17-2.
Equation 17-2
where:
= AAOT, vehicles per day (for intersections, this is the sum of the average daily approa.ch volumes)
An example calculation is provided in Example 17-4 (sec page 366), which references Exhibit 17-8, where the rates
arc calculated for each intersection ("spm" location).
Obviously, ranlcing locations by collision race requires traffic volume dau. The time period of the volume qau should
match the time period of the collision data being analyzed. Analysts may use volume dara somewhat outside the collision data time period if they adjust the volumes for temporal variation (growth) and seisorulity (ste Chapter 4). For
high growth areas, analysts should use volume data that arc 5 years or more removed from the time period of the: collision data with extreme caution, if at all. Like the frequency method, RTM could cause a problem with th.is method
as well. Analysts should note the n.te method of identifying hazardous sites is likely biased in &.vor of sites with low
exposure (the opposite bias of the frequency method) because only a couple of "unlucky" collisions on a spot or section with low exposure will produce a relatively high rare. l..astly, there is no apparent threshold value to determine
if the site is indeed luz.ardous; however, the HSM docs provide a methodology for calculating a "critical rate" which
can be used to provide a threshold.
cc,
W=-' CC,00
Equation 17-3
where:
W1
"'
cc;.
CC,00
Equation 17-4
, where:
; WA: I. PDO"' weighing factors for each crash type
PDQ
Weighting factors are typica11y based on crash cost estimates developed and upda~d by FHWA (FHWA, 200 5) or regionally calibrated estimates (such as those collected at the state level) based on specific region.allevd collision and C!JSt
data. For this reason, different agencies may use different equivalency factors. For example, in 2008 North Carolina
used wcighcing factors of76.8 PDO collisions per fatal or A-injury collision and 8.4 PDQ collisions per B-injury or
Cinjwy collision. The arialystshould note some agencies may separate collision severity types differently than others;
therefore, the equation can be easily manipulated.
Analysts should keep in mind that chis methOd is generally biased towards locations that' have a higher proportion of
severe crashes (such as low volume rural locations). RTM is again problematic using this method because sit es identified as having a high EPDO rate would likely rerum to a lower rate even if no coumermeasure were implemented.
Last, there is no obvious threshold value to determine if the site is indeed hazardous.
EXAMPLE 17-ls Application of the EPDO Method
State A has provided the crash cost estimates by severity in Exhibit 17-3. Determine the EPDO scores for the foU!
signalized intersections describ~ in Exhibit 17-4.
s~
lntaaudon
Total
8:
EPDO
Solution: The weight factor and EPDO rates will need to be calculated for each of the four sites. Calculations of
weights for fatal and injury collisions are:
$97,500
-:t .. $9,300
10
The EPDO value for Intersection 1, in equivalent property damage only collisions over the time period of interest, is
calcubted as:
tRSI1
RSl = J.:!___
Equation 17-5
NO.y
J'
where:
RS~
RS~
tRSI,
RSI, =-7---
Equation 17-6
LNo,,
,..
where:
RS~
RS~
Signaliz.ed
Intersection I Rear End
'
<'
.
,
.i!":~~~~,-~. . :~F
-;._ $
.
.
r . . .,~o -~ ,
"'
;~1 .
... ."1:, .. ,
__
:~ -.,rifir.;.,........-..-:--;r-
...~
.,A~tj~'f~'f'\n'11t\~~~-~~~~i~i;~
Rear End
Sideswipe
$ 12,000
$9,500
Faed
Angie
Pcd.Bicycle
Head On
Object
Other
$74,000 '
$115,000
$72,000
$83,000
$47,800
Total
RSl _
"'ol'ti'!"'"''"'" .,. .
,. ,.;ji
~-~s~~~.t --~~~--~
$60,000
$19,000
$740,000
$230,000
$72,000
$249,000
$239,000
$1,609.000
$57,46~
$168,000
$47,500
$296,000
$0
$332,000
$191,200
$1,034,700
$33.377~
$72,000
$0
$370,000
so
so
$216,000
$166,000
$334,600
$1,158,600
$50.374
$132,000
$9.500
$148,000
$1 15.000
$0
$498,000
$143,400
$1,045,900
$43.579
Tow RSI
$4,848,200
AvecageRSJ
$45,738
Solution: Calculations of the R.Sl for Intersection 1 and the RSI fo.r the reference population are shown below, wich
the solutions 'for the R.Sls for Intersect.ions 2-4 shown in Exhibit 17-6.
RS!t
$1,609,000 =$57,464
28
. 106
Exhibit 17-6 shows the R.Sls for Intersections I and 3 are greater than the average RSI and should b; consideredfc:?r
further evaluation.
'
Traffic Collision Studies 3& 3
RQC can apply to spots or sections. For spots, analysts use collisions per million vehicles. For stctions, analysts use collisions per million vehicle miles (mvm) or per 100 mvm.
The RQC method Bags a location as hazardous if it satisfies the following inequality.
xs)o.s 1
( v, +-2V,-
Equation 17-7
XS
= mean crash race for locations with characteristics simil2r co those oflocation i
volume of traffic at location i, in the same units as the crash rates are given
Agencies commonly use 90, 95 and 99 percent levels of confidence, which correspond to K values of l.282, 1.645
and 2.327, respectively. The question of which locations ue similar enough to include in the computation of XS is a
difficult one. Generally, agencies have wed relatively broad definitions of similarity to compute XS. For example, one
agency used statewide average rates based only on intersection type (for instance, arterial meeting collector in an urban
area) and traffic volume to provide XS.
EXAMPLE 17-3: Application of the Rate Quality Control Method
Roadway Section AA had 40 reported collisions in 3 years, and the agency responsible for the ~ection estimated travel
on the section was 19 mvm during that time. The mean collision rate for all sections in the jurisdiction similar to Section AA was 140 per 100 mvm. Should an analyst fbg Section AA as haza.tdous with 95 percent confidence?
So/mum: The RQC method requires the same units for each variable in the inequ.ality, so the analyst should convert
XS to 1.4 eollisions per mvm to be consistent with the units given for V, OBR is 40 collisions/19 mvm s 2.1 collision.s
per mvm. For these daca, the rate qu.ality concrol inequality {Equation 17-7) above holds crue as follows.
XS}o.s I
OBR> XS+ K
( - +-
V, .
2~
4)o.s 1
1. +-2(19)
2.1>1.4+1.645(19
2.1 > 1.9
Therefore, the agency should consider Section AA hazardous with 95 percent confidence using the rate quality concrol
method.
'\11:4
FOITI(\~1
Equation 17-8
(}'F
This condition should be checked shorrly after new or rebuilt sites are constructed to correct deficiencies. Usc
of all the available collision data possible in the after period is important. Typically, very shott periods of time
are all an analyst has in identifying a sire; however, whole years of data are bestfor proper usc of this method.
If shorter time periods (< 1 year) are all the analyst has available, it is advisable the collisions be multiplied by a
time period ratio for proper comparison. For instance, ifonly 4 months of collision data were available, a ratio of
3 (there are three 4-month periods in a year) would be used. If 10 collisio~ were reported during this 4-mont:h
period, the proper comparison would be 30 collisions (10 colJision.s x 3 =30 collisions/year). It is important to
note using this ratio method has inherent flaws (the 4 monthS of data may be during the worst collision period
of the year); therefore, if the newly constructed site is considered deficient, use caution and good engineering
judgment. I..asdy, it should be noted forming groups could be difficult if sites were not previously defined.
Condition C: Ate there sites that have rapidly deteriorated in recent years?" All sites should be checked for deterioration, not just a subser. The analyst should look for spikes in the collision frequency (F) on a regular basis,
say every year. Usc as much crash data as reasonably possible, preferably 10 or more years.
Condition D: ;.Are there potentially hazardous sites we might be missing because they are low-volume! lowcollision sites?" Thj.s criterion accounts for exposure and makes sure all sites have an opportunity to be identified
(not just large, high-volume sites). Only2-3 years of collision data are necessary; however, exposure data in the
form of traffic volumes will be necessary to calculate a rate. This check should be done on each site every )-10
years. Refer co tate calculations for spots and segmentS in Section 4.2.3.
Condition E: Are there newly constructed or rebuilt sires that may have some deficiencies present but could
be missed because they are low-volume, low-collision sites?" This criterion is similar to Condition B but instead
looks at <;ollision rates. Only sites that are recendy constructed or rebuilt would be considered in this calculation.
The rate (R) at each individual location is compared to the mean rate (R.) of similar sites throl!gh thecaJtula;
cion bdow:
Tr~Ui ...
r n.llidnn C:.tllriiDC:
~f:.lt
Exhibit 17-9
R, - R.,
(JR
This criterion should be checked shordy after new or rebuilt sites are consuucred ro correct deficiencies. The
same basic principles applied in Conrlilion B can be used for short lime periods of collision dara. Ir should be
noted forming groups could be difficult if sires were not previously defined.
Analysts should keep in mind the SWP method does not account for KI'M. Its basic premise is that it uses the combined
strengths of the frequency and rare methods to determine sires where countermeasures could be most productive.
EXAMPLE 17-4: Application of"Sites with Promise"
In an attempt to decrease overall collisions, a city has decided to screen for hazardous signaliied intersection locations
by looking at angle, left-rurn, rear-end and total collisions. The list of intersections shown in Exhibit 17-7 was pulled
from a larger database of collisions the city populates. The two newly constructed sires include a full year's worth of
data. The city has acquired safety funds foe use in fixing hazardous locations. Although the city is considering any
possible countermeasure solutions for decreasing crasheS, they are considering insraUing additional red light running
cameras (RLCs) as a part of the program they started 7 years ago. Which sires the city should consider further evaluating and what other measures shoufd be taken once the sires are determined haz.ardous?
11B'f.il~
-<-;
..,._
. ...
..
...
-.-~...::
..
~----
...,,
, -~-
.....
. .....
..::-
-~b~V;fr]t...'JC"'~o.4-""'~~~..rt-~j-c""-"~~
"~
: ~~ .: ~'J~i ~~~ '-!~ . -~~:~~--.,~~--:
Site
Site Type
Angle
Collisions
30
Left Tam
Collisions
Rear-End
Collisions
10
7
Site rebuilt in summer 2009
Sig.lna..
Sig. In a..
Sig. Ina.
16
22
27
Sig. Ina.
10
5
G
Sig. Ina..
18
12
12
27
Sig.lntx.
11
Sig. Ina..
25
18
18
Sig. Ina..
24
9
10
Sig. Ina.
S"tg. Ina.
II
11
Sig. Ina..
l1
17
16
7
Site constructed in lace 2009
13
15
II
24
17
ADT
39
32
13
Total CoUisioos
Other
Collisions
2008
2009
2010
42,000
27
36
26
60,000
19,000
..
50
37
27
34
28,000
33
21
36,000
31,000
24
18
30
32,000
26,000
11
16
24
27
17
17
20
14
28
18
25,000
24
24
35,000
17
19
23
15
~1,000__
18
16
L_ _
15
'
-~
-=
~~~~.tllit'~'i~~ffili.i'm#-1i1't":."-~~4ir'~~~
..... {~:~:~
- ,_,_. ,.,:it_'flti:
A . ~- h!!.. . .. ;+: _ .. -~ >o"u _ !f-=.~~- _:.: :__ o_ .w .. _(
< o~~ -~'il~-,~~.~i-:~~M.,-..~~':S''"1'
.:.:; ~,~
- ~-~,J~::f~~~;,::;t,-.;:.'--~~
Site
Angle+ Left
55
29.7
1.94
50.0
2.28
2
3
38
34.7
5.00
28
24.0
2.35
37
27.0
2.05
27
17.3
1.53
27
18.0
1.54
32
17.7
1.86
24.0
2.63
lO
24
19.7
1.54
11
26
16.3
2.13
22.7
2.2
6.5
1.1
----
_,___
,___
Solution: The solutions f?r each of the five conditions are shown bdow and refer to calculations shown in Exhibit 17'-7
cicr
S.iiU
50-22.7 - 4.2
sm..2
6:5
Both sites are above average because their resulrs are greater than ttro. Site 2looks most promising.
Condition C: Looking at "Total Collisions" in Exhibir 17-7, Site 8 shows there was a rapid deterioration d.v.J:'ing the 3-year period, with nearly 3 times the nll;!llber of collisions in the third year compared to the first yc;~
Although the site appears to have rapidly deteriorated, it would be a good idea to look at the long-term frequet1 cY
of collisions (say 10 years or more).
Condition D: Excluding both new sites and looking only at collision rates for the other nine intersections, Sire .3
stands out with a rate of five collisions per million vehicles, shown in Exhibit 17-8. Note that rates arc calcuJace~
for "spors" and not segments, and that total collisions are used in the calculation of"R.are per million vdticles
Condition E: Compare Sires 2 and 9 with similar sites. Since all other sites are "signalized intersections," ~
comparison should be made using the mean and standard deviations of the rates for each of the other nine sire&
Again, total collisions are used. Refer to Exhibit 17-8.
SitU
2.28-2.2 = 0.73
1.1
S&.2
2.63-2.2-0.39
l.l
Both sites are above average because their results are greater than u:ro. However, neither site is promising since their
value is not greater than 1 (one standard deviation).
Summary: Sire 1 appears promising for implementation of a Rl..C. Sites 2, 3 and 8 appear to be good sites for implementation of some form of countermeasure; however, the exact countermeasure(s) has yet to be determined. Since
Site 8 appeared to rapidly deteriorate in 201 0, it would be a good idea to look at the long-term history to confirm this
excessive increase in collisions was not random. Lastly, a condition and collision diagram for each site would be useful
for determining crash causes (discussed further in Sections 4.3.1.1 and 4.3.1.2).
...
A.MF/t
Equation 17-1 0
Using the predicted and observed crash frequency, the expected number of collisions could be calculated. The basic
premise of this methodology uses a weighting factor based on the overdispersion factor of the SPF. The more reliable
the SPF used to predict the number of collisions, the higher the overdispersion factor, and thus the more reliance
given to the predicted collisions versus the observed collisions. The HSM provides methodology for determining the
expected collision frequency, N~The expected number of collisions would then be used for ranking in lieu of the
actual collision frequency (Section 4.2.2). The expectation is that RTM is no longer an issue because of the reliance
on predicted collision frequency.
The EB methodology is a much more rigorous method for determining hazardous sites; however, the downfall is that
it is muCh more cumbersome. In addition, SPFs are nor available for all facility types. The first edition of the HSM
includes SPFs for rural two-lane two-way coads, rural multilane highways and urban and suburban arterials. It's hoped
SPFs for other facil.icy types will be added to the HSM as research is completed.
(Methods to account for severity can supplement ocher methods and reveal locations chat experience extreme numbersof
~ere collisions. However, severity methods such as EPDO and RSI introduce yet another arbitrary judgment a nd vola1pte source ofvariation into the analysis. Also, since underrcporting levels vaty by severity, the choice oflogical EPDO values is difficult. Therefore, severity methods should n()( serve as the only means of identifying locations for further review.
The RQC method is generally superior to simple frequency and rate in that it identifies truly l=ardouslocations correctly more often. However, the rate quality conuol method n:quires many mon: resources than ocher, simpler methods
because agencies need average rates for different classes oflocations. If an agency has a rdiable source for avenge rates
or sufficient resources to collect such data. RQC could be effective.
The SWP method looks for sites where countermeasures could be most useful, tries to be pro:lCtive, is more efficient
in its usc of data and looks at potential sites that may be able to take advantage of proven treatments, chw usi ng safecy
funds in a manner that takes advantage of proven countermeasures. The method is simple since it builds on the basic
building block of collision. frequency, with some emphasis on collision rates co account for exposure.
The EB techniques .o f identifying collision-prone locations hold great promise and agencies should consider them..At
this time, SPFs provided in'the HSM only include three facility types (rural two-lane ~way roads, rural multilane
highways and urban and suburban arterials}. Futun: facility types will be added as n:scarch is completed in other areas.
C+"'l""'
..
.,.c.a
ounD
~
Road#l
-r
...
,.-~ , ~.
IIJiliiDl
~t:,Cttod:
~01..-m~
DD.,U&f*
M-Dwt,NoUII'!t
-- """"
1-CNsk
P.. rto'l
If the diagram becomes too crowded when each collision is shown separately, analysts can use symbols to represent
sets of collisions of a particular kind. A bold arrow representing 10 collisions of a particular kind is common, for ex
ample. The diagn.ro loses interesting derails by summarizing the data in this way, however. Twenty to 30 collisions fit
comfortably on an 8.5- by 11-in. sheet, and analysts can usc oversize sheets for spots with more collisions or longer
periods. Labels on the arrows could indicate the dace of the collision, the day of the week, the time of the collision,
the road surface condition at the time of the collision and the lighting condition at the time of the collision. Collision
diagrams should show military time for clarity and brevity. Abbreviations such as "OD" and "WN~ can represent "dry,
daY and "wet, night without lighting. for example, w show important information concisdy.
Light condition and road condition are two of the most important variables used to idel'ttify overrepresenteq dusters
of collisions. Day and night conditiotls are assumed to be correctly entered by the officer that produced the report. A
cluster of night collisions, for example, might indicate the need for countermeasures such as reflectors, delineators and
street lighting. Light conditiotl is a reasonably accur.i.te data item analysts can check against the time of the collision
to find errots. Analysts often check the night-to-day collision ratio to ~uge wltether a location or duster of collisions
is overrepresented at night or during the day. Analysts must he cautious, however, because the night-to-day ratio will
fluctuate a great deal With low numhen of collisions. Most collision report forms include codes for "dawn" or dusk"
light condition at the. time of the collision. These codes present a dilemma since the samples of dawn or dusk collisions are often too small co analyze separately. It is customary in many agencies to eliminate dawn and dusk collisions
when computing night-to-day ratios and when necessary during other analyses. FmalJy, since the amount of daylight
per day varies throughout the year, light condition measures, such as night-eo-day ratio, will also vary throughout the
year even if other faaors are constant. Analysts making comparisons using light condition should therefore use the
same months of the year for all locations being compared.
CoUision diagrams contain very few details about the particular location. Street names, outlines of the edges of the
pavement and a direction orientation are all that are necessary besides the symbols for collisions and fixed ohjeets.
Sometimes, collision diagn.ros contain a table or supplementary diagrams for summarizing the collisions by type, se370 MANUAL OF TRANSPORTATION EN GINEERING STUDIES. 2ND EDITION
~eriry, light condition, or road condition. This can be particularly hdpful when a collision diagram beco mes crowded
~nd clusters are hard to identifY. Computer programs are available to help prepare collision diagrams, and an atrial
'i.mage is nice for overlaying che curb lines. The programs are often flexible and allow the user to specifY which co\Jisions to include and which variables to show on the drawing. The flexibiliry of computer collision diagrams is a major
advantage. For example, analysts can quickly obtain separate diagrams of night collisions, injury collisions, or :lilY
subset of the total reported collision picture. The output from computer-generated collision diagrains can be ploned or
shown on the moniror and computer plots are usually easier ro read chan manual drawings. Use caution co ensure rhe
information on computer-generated collision diagrams is complete and accurate-some collision derai ls may have to
be added manually.
t6
"""'"""
\>1
01
,....
..
~~.
~
.@
~";~
:::~;., ~~t
.,,< :
~ :~ ~; ~.
s::-o
:I:
.Jil.
~:
u:u
I
..
:::c
~~~
:iu l~
:
1l... ::
! !
'-1 1 oil
o~
. ,. __
Road Ill
!~
!Iii
----~-----~-;:----~-------
- -----~--~;------r--------
--------------r------..,._.s;____. . .__-:::~'-~:..;:..::.....-:
-'-----------------u;---------------
r
--------, \
o.. str.~~on
t.SQI&on
t:rRM<I'l
f*Tin ll
I
I
I
Hottt:
'
u:tl'
I
I
SIP\IIq'd~tlc~lti,OMOIUII
.,.II~Unstt"tlctdii'ISW,.,...tstiN't
-o.d...,JVIIf:IC'i.t,.nsdll"''tfMdrt~ on
I
I
41Mrdn..ar ~ n.otHdu/d.,.
ol4t412AOI' 7,010vtl'lltlu/W.,
Gos stllllon(4Pumps)
& COnvenience store
I
I
I
I
The site diagram should be approxi.matdy to scale and include all the physical features of the study site (for example,
sidewalks, crosswalks, bicycle lanes, vegetation, drivewa~, embankments, signs, traffic signals, markings, roa~way
shoulders, abutting land uses, bus stops and other characteristics or conditions that may affect visibility and/or sight
distances). All streets or highways should be labeled by their official names and/or numbers. The diagram should indicate estimated measurements for roadway widths, shoulder widths and lane widths. A checklist of items commonly
required for a complete site description is shown in Exhibit 17-11. All these items may not be required for every study.
\,
Countermeasure selection is typically done based on the contributing factors associated witb crashes and the e:xp<:rience of the analyst. For instance, an intersection may have a high number of angle and rear-end collisions on an approach that rounds a curve. Using the collision and condition diagram, along with collision summaries by type and
"severity, the analyst identilied in the likdy causal ciWn a sight distance problem at the site during all hours of the day.
The analyst recommended landscaping be removed bea.use it was encroaching on the ability of the drivers to see the
imersection signal displays as they rounded the curve.
Highway loca cions vary gready, tke state-of-the-art in collision countermeasures changes rapidly and there are limits
on the quantity md quality of collision data at many locations. Therefore, generating lists of possible countermeasures
for particular sites is, in some ways, more art than science. Ideas for countermeasures come from e:xperience and a
thorough knowledge of the technical literature more than other f.tctors.
The major sowce for countermeasure ideas is the technicalliterarure. Many references arc available to the eng.ineer to
aid in countermeasure selection for certain highway situations. A few primary references used by traffic engineers are
discussed bdow.
Exhibit 17:12 provides lists of countermeasureS for several typical collision patte.rns and contributing
facto.rs (FHWA, 1991). This countermeasure list is certainly not exhaustive and does not include newer
technologies. However, the collision patterns and probable a.uses are summarized nieely for the user and
most of the traditional countermeasures that inspire confidence are listed in the exhibit.
Another good source of countermeasure information is NCHRP Report 500: Guidance for Implnnml4rion
ofthe AASHTO Strrz~ Highway Safrty Plan (NCHRP 2003,2004: Volumes 3-10 and ll-14). The
volumes focus on various highway crashes and contributing f.tctors, and aid analysts in treatment and/or
prevention of these crashes through recommended coumermeaswe ideas.
of
The HSM provid~ a list crash contributing f.tctors for various emphasis areas in "Chapter 6: Sdect
Coumermeaswest however, the concurrent lists of potential countermeasures are nor provided (TR.B,
20 10). HSM chapters pertaining to individual types of roadway features have countermeasure ideas and
accident modification f.tctor (AMF) estimates for each.
-_.
.,,.
'~w&:trr~'~:>',..,~
~
""-~~~-~;~r
-~~:~~-.:.-<c;~.--.~,, ,...,,
Bi" ' !': - ~';'0 ~~
:;a~or~
~ .. 'I ~
~
7~-;. .... ___t-:~:!;-=:i;.:.:i'.;,::;'_~
..
-. . -. .... .
. .
'. '
Type oflmprovement to
Construction Classification
Intersection and craffic control
. '..
..
37
15
15
..
' *' . -
;;oo:. ~
--~
....
Benefit!
Fatal + Injury Cost Rlttio
$18.7
4
2.8
$297.7
48
23
24
$510.9
$19.7
$7.8
44
31
32
$371.4
$22.8
3.6
Traffic signs
$19.6
34
$59.3
$15.7
20.9
$~3.7
15
(I )
(I)'
$751.8
1.6
10.3
Illumination
$,13.2
45
$122.6
$16.8
$63.1
40
22
22
$412. 1
$8.6
$127.3
49
21
21
$344.5
$10.5
5.1
$307.7
50
28
29
$752.2
$92.1
1.7
$79.8
49
22
23
$103.8
1.2
Bridge replacement
$156.9
72
47
49
$1,201.6
$159.3
1.1
$26.2
77
40
43
$1,637.8
$223.2
0.8
$39.0
36
20
21
$277. 1
$39.8
4.5
$5.6
41
45
$189.5
$39.4
6.5
$1,971.3
31
13
13
$722.2
$54.0
1.8
$511.0
9"
$4,041.3
$174.2
0.4
Lanes added
$212.9
-2
13
13
$66.8
0.1
$56.8
73
17
19
$382.4
$72.8
3.2
$88.3
28
11
12
$497.9
$37.9
2.6
Roadway realignment
$329.9
61
32
34
$1,193.2
$111.1
1.1
$468.4
26
18
19
$837.0
$30.3
1.8
Pavement grooving
$12.6
34
15
15
$377.3
$14.5
3.8
Upgraded guardrail
$149.7
42
$151.7
$31.5
8.1
StmctuJ'CS
$1,0n.9
'
'
$7.4
45*
28
29
$192.8
$10.8
$58.9
62
3*
$224.7
$213.6
5.4
Impact attenuators
$10.7
31*
36
36
$390.0
$8.0
$36.5
(2.5)'
8*
$102.2
$4.8
61 *
44
45
$200.9
$25.8
6.3
$15.0
49
22
23
$202.5
$16.0
6.4
$331.2
89
63
67
$570.0
$114.2
2.2
$55.0
91
74
77
$551.4
$103.5
2.2
$138.1
84
86
$591.8
$115.3
2.1
80
$432.8
$96.7
2.8
Obstacle removal
Rallioad-highway crossings
$63.2
90
78
Equacion 17-1 1
Equation 17-1-:2-
The basic method for applying AMPs a.nd CRFs is the same, and they can be used interchangeably as long as all cb. te
analyst undetSt.ands this cli1ference in calculation a.nd meaning. An AMF of 1.0 (CRF =0 percent) would be the "basG
condition: which essentially means there would be no cha.oge in the expected number of collisions because ther~
would be no change in the conditions related to collision causation.
Often, the standard error is provided with an AM.F. The standard eccor can be used to provide the analyst an estimace
of the effective range of collision change, as shown il! Equation 17-13:
Cl (X"/o) -
Equation 17-13
when::
CI (X pe.rccnt)
= confidence in.terval, which would snre mar one is X percent sure the AMF value falls within the
inrerval provided
AMP
SE
MSE
&
AMFs cannot be applied directly when the original condition is nor the same as the AMF base case. AJ an example,
consider lane width, when: the base condition AMF (AMF = 1.0) would be for a 12-ft. (3.7 m) lane. U an c:xisting
roadway has 9-ft.(2.7 m) lanes (AMF = 1.50) and the agency is considering expanding to 11-ft. (3.4 m) lanes (AMF =
1.05), an analyst would need to take the ntio of the updated condition tO the existing condition AMFs, or 1.05/1.50
to get an AMF for the treatment of 0.70.
One major ftaw with applying AMFs or CRFs is the method assumes countermeasures act independendy of one
another. At this time, the combined effects of multiple tn:a.tmenrs an: generally not known and further research is
needed to fully understand how multiple treatments interact. A.t this point, two methods can be wed co make an
estimate, if multiple treatments an: being considered for a roadway facility. Firs.t, the analyst could try to sepa.nre the
collisions into those types that would be affected by a treatment and those types that would not be affected. For instance, an intersection has been ide.ntified that has high numbers ofleft-turn and right-rum crashes. The analyse may
be considering cwo ueacmenrs: 1) changing protected/permissive left rums to protected only and 2) adding channdized right-rum lanes. In this case, it would be a good idea to separate collisions by movement for left-rurning vehicles
and right-turning vehicles and applying each AMF direaly to its rdated crashes instead of the entire intersection.
The second mecbod should only be considered if it i.s not possible to separate collisions, such as on a roadway with
no shoulder and small lane widths. If the analyse was considering increasing both shoulder and lane widths, it would
be nearly impossible ro determine how collisions should be separated for those treatments. lns.tcad, the analyst could
multiply the AMFs together to get the combined effect. & mentioned earlier, one mus.t we caution and unders.tand
the limitations of this second method.
A second 8aw in applying AMFs is the estimate of the number of collisions expected under the do nothint condition. The recommended method is tO use the long-term average number of collisions as the base levd because it is
simple and easy to understand. Usually, 4 years or more of collision data are necessa.ry to reduce spiking tendencies.
The AMF is then applied din:aly to the long-term average number of collisions to determine the apeaed collision
savings. Shorter durations of time could seriously jeopardize the estimate if the collisions spiked up or down during
those shorter time periods. This method iJ flawed because it doesn't account for RTM. A better method would be to
use an SPF ro predict the long-run expected number of collisions without the treatment. However, this method is
fhwed because it assumes the predicted collision frequency is similar to that of the Location the SPF was based upon.
Therefore, the best method would be to calculate the expected number of collisions using an EB methodology which
is based on a long-term observed crash history and a prediction from an SPF. See the HSM for a thorough explanation
of these methods.
~.,e
iAII~IIt/\ 1
()CTDA~Ic"nl"'\nTATII"'\t.l
,..,,,.. ,,,,.,..""',...
2. If lane widths were increased to 11-ft., one wolild first need to calculate the AMF from ia emling 9-ft.
lane width condition. The AMF i5 calculated as:
AMFu-/t 1.05
AMFrreacm~mt = AMF9-/t = Ls = 0.70
The expeCted collision reduction would be 30 percent {(1-0.70) x 100}, or a reduction of2.4 collisions per
year {{8 x (1-0.70)}.
3. If both treatments were installed, it would be nearly impossible to separate collisions related to each
treatment. Lacking a trustworthy SPF, the analyst would therefore need to use the combination method
described above. The AMFs for RPMs and 11-ft.lanes are 0.85 and 0.70, as noted above. The comp utation for the combined AMF i5
AMF....w...~ "' 0.85 x 0.70 = 0.60
Assuming the two treatments act independendy, the expecred colli5ion reduction from installing both treatments would be 40 percent {(1-0.60) x 100}, or a reduction of3.2 collisions per year ({8 x (1-0.60)}.
Benefit =
.
i:. (CC
11
N1 )
Equation 17-14
where:
CC1
N,
Next, the annwl cosrs must be identified for each countermeasure. Annual eosr is based on the annualized installation costS,
AG;, and the annual maintenance cosrs, C,. The calculation of the estimated acrual COSts is shown in Equati~n 17-15.
Equation 17-15
The annualiud installation costs, Ac;, are based on the total installation cost, C? discoum rate, R, and the expected
service life in years, N. The di.lcount rate is cypically becween 3 percem and 7 percem and can usually be obtained
from the local agency. The calculation is shown in Equation 17-16.
C,* R
Equation 17-16
State "B wanes to determine the benefit-cost ratios associated with the installation of RPMs along four similar 2-mile
long cwo-l.ane rural roads. Crash costs and crashes by cype are provided in Exhibit 17-14. Assume the following:
AMFRI'M.h.u.... ~
0.50
AMFmf.,.,-.ff.-J~
0.75
The discount rate obtained from your agency for RPMs is 5 percent with a service life of I0 years. The total installation cost is $100,000 per site. The maintenance cost for RPMs is $5,000 per site per year.
Solutiom Exhibit 17-15 shows the estimated benefits, annual costs and the benefit-costs for each AMF and segment.
_=
( N..,.. ( I- AMF))
N - . = (I (I - O.S)= 0.5]
N_..-_, =
[N_....., (1-AMF))
As currendy conducted by many practitioners, before-and-after srudies using collision data suffer from two serioUS
fu.ws and provide inconect and misleading results. The flaws are 1) che failure to control for the effects of changi.ng
conditions during the lengthy time periods required ro amass before-and-after collision statistics (seasonality, cime period, historical effects, etc.) and 2) the failure to correct for RTM. An example of seaso~cy and/or time period may
be evaluation of~ countermeasure where 5 complete years of data are available; however, only 5 months of data ~re
available after implementation. If this aftet period were during a particularly harsh winter, collisions would naturally
be higher because of the harsh winter and do not have anything to do with the coun~rme~ure. To account for ti-lls
seasonality or time period factor the analyst could usc a comparison group, where the comparison sites would natUrally fluctuate in a similar manner to treatment sites. These similar trends would negate most of the effects caused by
the harsh winter, allowing the analyst to pull only the effeets of the treatment.
Although co~parison groups are commonly used, the most rigoro~ way to overcome these two flaws is to tond~ct
an experiment using randomly selected control sires where the agency measures collisions but does n ot install rbe
improvements. However, rmdom sdection of conrrol sites is rardy done because agencies typically install treauneo.cs
at the sites in some predetermined manner. For this reason, it is more common to conduct observational before-afcer
studies such~ Comparison Group and EB methods (Hauer, 1997). Appendix A describes how to conduct a beforeand-after experiment with control sites and how to conduce other common countermeasure evaluation srudies. If" a
rigorous safecy study (such~ Comparison Group or EB) is needed to determine the effect of a particular counterme ::asure, the analyst mar, need to consider hiring an evaluation expert..
5.0 REFERENCES
American Association ofS12te Highway and Transportation Officials. Highway Safny MAnUAl. Washington, DC: AASHrO,
2010.
o/
Antonucci, N. D, K. K. Hanly, K. L. Slack, R. Pfcfer and T. R. Newnan. NCHRP Report 500: GWlanu for lmplnnmllllion
tiN AASHTO Snaugic Highway Safoty Plan, V6h.mu 12: A Guilk for Reducing CoOisions I# Sigruzljud lntn1tctions. Washington,
DC: Transporoi.tion Res=ch Board, National Research Council, 2004.
Balm, G., er al. Desktop &fmnu for Crash &duction Facton, FHWNSA-08/011. Washington DC: U.S. Deparunem of
Transportation, Federal Highway Adminismtion, 2008.
Box, P. C. and J. C. Oppcnlandcr. Manwd ofTnzffic Engillming Stu.ti.Us, 4th ed. Washington, DC: Institute ofTransportation
Engineers, 1976: p. 49.
Federal Highway Administration. Charat:UriJtia ofEmnging Road and Tnzil Usm and Thnr Safny. Washington, DC: U.S.
Department ofTransportation, 2004 .
Federal Highway Administration. Crttsh Cott Estimarn by Maximum Polia-&porud lnj11ry &vniiJ within &lrckd Cnt~h
~. FHWA-HRT-05-051. Washington, DC: Federal Highway Administration, 2005.
Federal Highway Administration. Highw.t] S4fitJ Enginming SluJin Proud~~ntl G!IUk. Washington, DC: U.S. Department of
Transportation, Federal Highway Administration, 1991.
Harl<ey, D: L. and R. Ruiz.. "HISAM: A Collision Data Base Manager." Tntn1porration Reuard! &cord: jo11rnal oftiN
Tntn1p0rtaJion Rrstard! Board 1238 (1989): 37-44.
Hauer, E. and A. S. Hillen. "Extenr and Some Implications of Incomplete Collision R.tporting. TntnJPonarjqn Reuarch
Rrcard:]MV7JIZ1 oftiN Tramportation Rrw:rch Board 1185 (1988): 1-10.
Hauer, E. and B. N. Persaud. "Problems Identifying Hazardous Locations Using CoUision Data. Trrmrpo~tm Rrstarch
Rrcorti:j011mal oftiN Transportation Reuarch Board975 (1984): 36-43.
Hauer, E. and B. N. Persaud. "How ro Estimate the Sakcy of Rail-Highway Grade Crossings and the Safety Effects of Warning
Dcvk.es." Tntn1p0rration Reuard! Rr(tmi: ]011rnal oftht TntnJPDrtation Reuanh Boani 1114 (1987): 131-140.
'.
Hauer, E., ct al. "Estimation ofSafccy ar Signalized Intersections." Tnznsportaiion Rntarch Rrcord:]o11rnal oftht T_n:msportation
Rntarch Board 1185 (1 988): 48-58.
Hauer, E. "Identification of SiteS of Promise." Tnmrporration Rmarrh Rr"'rd: journal ofw.Tnmrporration Rrstareh BoanJ 1542
(1996): 54-00.
Hauer, E. Obumttio1141 &fore-Afttr Studin ill RDdli 54fitJ. Amstudam, Netherlands: Elsevier, 1997.
Higle, J. L and M. B. Hecht. "A Comparison ofTechniqucs for the Identification of Hazardous Locations." Transporration
Rrcord: Journal oftht Transportlltion Rnearrh BoanJ 1238 (1989): 10-19.
Rrst~~rrh
Inscirutc ofTran.sportation Engineers. Sl4tittiuJ EvaJuation in Tnzffic SafotJ StudiD. Washington DCi ITE, 1989.
KnipUng. R. R., P. Walle:, R. C. Peck. R. Pfcfer, T. R Neuman, K. L Slack and K. K. Hardy. NCHRP Report 500, Volume 13:
A Glluu for kldrming CDIIiMns lnHiving Htllll] Trueks. Washington, DC: Transportation R.tsearch Boatd, National Research
Council, 2003.
Lacy, K., Srinivasan et.al. NCHRP R.tport 500, Volume 8: A O..Uk for .AJJJraring CDI/iMns /nll()/ving UliBIJ Polts. Washington,
DC: Transportation R.tsearch Board, National R.tsearch Council, 2004.
Mak. K. K. ct al. "Automated Analysis of High-Collision Location.s. TnlnJ!Drtaiion Rnt~~rch ~ }OIIrnal oftiN
TrrmspDnatiDn Rnearrh BoJUd 1068 (1986): 59--64.
McGee, H., S. Taori and B. Persaud. NCHRP R.cport 491: Crttsh Experimet Warrantfor Trttjfic SigMI.t. Washington, DC:
Transportacion Rtsea,rch Board, National Resca:ch Council, 2003.
Miller, T. ct al. Snuilivity ofRrstnlrrt A1IDauUm MH!tls 10 Dismmz RAu IfNi Un"JDrutl CDIJiril11U. FHWA./RD-35/092. McLean,
VA:. Federal Highway Admini.stracion, 1985.
Nacional Coopencive Highway Reseuch Program. NCHRP R.tport 128: Mttho.ls for Idmtif;ing fUwtrdow Highwa, Eltmmt.J.
Washington, DC: Transpo112tion Reseatch Board, National R.csearch Council, 1986.
National Highway Traffic Safety Adminisrncion. Tnzffic S4fitJ Factt 2008.Washin.gton, DC: NHTSA,, 2008.
National Safccy CounciL Estimllltd Com ofTraf!U CDI/iMns. 1990. Chicago: National Safety Council, 1991.
,.,.,.., .,
a A n l At
iMn f:nJTt(lN
Neuman, T. R. er al. NCHRP RA:pon )00, Volume 3: A Guide for Atid=ring ColliMns with Trtes in HUArdous lAcatiom.
Wa.shington, DC, National Cooperative Highway R=tch Program, Transponation RA:search Board, 2003.
Neuman, T. Ret al. NCHRP RA:pon 500, Volume 4: A Guide for Addrtning Head-On ColliJions. Washington, DC:
Transportation RA:seatcb Board, National R=tch Council, 2003.
Neuman, T. R et al. NCHRP RA:pon 500, Volume 5: A Guide for Addressing UnsignaliuJ lntnumon Co/Jisions. Wa.shingron,
DC: Transportation R.csearcb Boatd~. National R=rcb Council, 2003.
Neuman, T. R, ec al. NCHRP RA:pon 500, Volume 6: Guidtmct for lmplmrmtllrion ofthe MSHTO Stra~ Highway Saftty
Plan, A Guide for Atldrasing Run-Off-Road Co/Usions. Washington, DC: Transporwion R=rcb Board, National Research
Council, 200.3.
North Carolina Department ofTransportation. North Carolina Project Development Crash RA:duction Factor Information.
www.naloLorg/doblpreconsuucdua.ffidSafety/ RA:sources/ projcct_guidelrcgionalf:a.ctors.pdf. 2007.
Pons, I., J. Scuta, R. Pfefer, T. R Neuman, K. L. Slack, and: K. K. Hardy. NCHRP RA:pon 500:
Tran.~portacion
4 the AASHTO Smiugir HigbwiiJ SD.fny Pum. Vo/111111 9: A GuidLfor .&dudng Collisioru With 0/dn Driwr1. Washington, DC:
RA:search Board, National R=rcb Council, 2004.
Swb, T. TIN Poclttt T~ kn.lmt RtcwutruciUm Guide: A Cqmpflu TNJfic Alxidmt Rtftrmct Ha.ntihoolt. 6th cd. Tucson, AZ:
Lawyers and Judges Publishing. 2008.
Stok.es, R W. and M. I. Mutabci. Race Quality Control Method ofldencifying fuzacdous Road Locations. TrrwportaiWn
Rtsearrh &am/:]011rnai oftht Trruuportati4n &searrh &arri, Issue 1542 (1996).
:r.
Scum. J., R. Knipling, R. Pfefer, R. Neuman, K. L. Slade and K. K. Hardy. NCHRP RA:port 500, Volume 14: A G uidLfor
Rtt!udng Crashes Involving DroUISJ and D~t! Drivm. Wa.shington, DC: Transportation Rescacch Board, National Research
C!>uncil, 2005.
Tocbic, D. J. et al. NCHRP Report 500, Volume 7: A Guide for &Judng ColliMns on Horizontal CUTWS. Washington, DC:
Tran.~ponation RA:scacch Board, National Research Council, 2004.
U.S.. Department ofTcansportation. The 1989 Ann114i &pon on Highway Stzftty lmprovmu:nt Pro:rrzms. Wa.shington, DC: U.S.
Department ofTransportation, 1989.
Zegccr, C. V. et al. NCHRP RA:pon 500: Guidance for lmplmrmtalion oftht AAS/1(0 Strategic Highway Saftty Plan. Volumt
10: .A Guide for &Juring Co/Jisions Involving Pttfatrimu. Washington, DC: Transportation ReSearch Board, National Research
Council, 2004.
Chapter 18
. .. .... .. .. .. .......... ... ........ ... ......... .. ......... . .. ...... ..... ... ......... .... ...
383
384
2.1 Introduction
384
384
385
388
3.0
390
3.1 Introduction
390
390
394
405
406
4.1 Introduction
406
406
409
5.0 REFERENCES
412
1.0 INTRODUCTION
ypically, safety studies are conducted at sites wbe.rc large numbers of collisions take place and the need fo.r a
potential countermeasure may be necessary to reduce collisions. These studies, presented in Chapter 17, a.s:-e
reactive in nature because they require collisions take place in order to make a decision on what improveme.n.. -cs
are needed at the site. This chapter looks at three alternative safety s.tudies that are regularly conducted in lieu of co 1lision studies as p7ri4Ctive measures for identifying safety issues th:u may need remediation. These studies are typicallY
conducted as separate studies; however, many times they c:m be considered during ua.ffic impact studies.
This chapter first covers road safety audits (RSAs) in derail. RSAs have been conducted since the 1980s in the United
Kingdom and since the 1990s in New Zealand and Australia. Since the fitst road safety audits in the United States we&e
conducted in New York, Pennsylvania and Iowa in the lace 1990s, the RSA has become used increasingly as a proactiV'"C:
safety tool. This safety study can be conducted at any time during the life of a roadway and is highly recommcode
as a proactive measure during the construction and design process. Secondly, collision surrogate srudies, often aile .d
conHict studies, are described. Traffic conllicts are an alternative to coUision studies when linle-to-no collision data ar-c:
available (s~ch as a newly constructed or rural site), or a quick study is needed to determine safety problerps.la.stl~
the chapter discusses advisoty speed studies based on recent updates to guidelines provided by ITE. Advisoty speeif...s
"}
have been appHed in many different ways by different stare and local agencies. Recently the ManUAl on Uniform Traffic
ContrrJ! Dnticn (MUTCD) has recommended all advisory speeds be updated by 2020. The methods for posting
appropriate advisory speeds are updated in an attempt to bring some consensus among users.
RSAs can be a very useful tool for the engineer if used correctly. State and local agencies have many ~ompeting
interests indud.ing environmental, right-of-way, socioeconomic, roadway opacity and political interests. With all
these interests competing for limited funds, safety is sometimes overlooked. RSAs inject safety into the mix of ocher
interests by cxpUcidy identifying the safety impHcations of road management decisions. In addition, there arc other
advantages to conducting RSA.s.
Collision reports do not always identify many road safety rdated issues.
Roadway designs need to anticipate and accommodate common driver errors which can be uncovered
during a comprehensive safety audit.
It is much easier to design and build safer roads to begin with than to modify existing roads or entrenched
behaviors of many drivers.
Because RSAs arc newer in the United States, they arc typically conducted on existing roads; however, applying R.SAs
early in the design process offers the greatest opporrunity to influence the acrual design. Changes that improve safety
performance typically become more difficult, costly and time-consuming as a design progresses.
Internationally, RSAs are being mandated by the United Kingdom on all new roadways, while Australia, N~ Zealand
and other countries recognize the RSA as a valuahk tool in reducing collisions. International RSA sources are available onllne, including the United Kingdom's National Roads Authority Design Manualfor Roads and Bridga- ~lumt
5 (NRA, 2009), Ausualia's Austroads Road Safoy Audit GuUklirm (Au.suoads, 2002), and New Zealand's Tmnsfond
Road Saftty Awiit Promium for Projecr (Transfund New Zealand, 2004).
This process lays the basic framework for conducting a successful audit. The steps provide information on the work
to be done and by whom. A basic summary of the steps is provided below, with more detailed information provided
in the foUowing sections.
Sups 1-2: hkntifJ the ro/Uh to t:ndit-" tk mulit tum. The party responsible for assembling the audit team i.s
the road agency. Good ~didates for preconstruction audits are safety-oriented projects, high-profile projects usually
r~t U'" \J r r
"'UIII"'' t:"ITlf"U\1
.(
RSATeam
audit~ at the request of politicians or the public, and complex roadway designs. Good post-consaucrion candidate
projects include high-collision sires, high-profile projecrs, or sites where traffic characteristics have changed (or are
expected to change) due to long-term construction detour routes or new devdopments in the area.
St~s 3-6: OmJuct the -"it tmd report (11J firulinp. This is the backbone of the RSA. The aCCIW RSA procedt.~re
takes place during these steps and ~dudes the start-up meeting, site visit, audit analysis and presentation of findings.
The roadway agency is only present during the beginning and end of this series of steps ro exchange information.
These steps will be the primary focus in Section 2.3, Data Collection Procedures.
Steps 7-8: FoiJDw up (11J the mulit finJinp. The local agency follows up on the findings of the audit by issuing a
formal response co each of the safety issues and making improvements wherever feasible. These steps will be the focus
rR~<;.;c-~@@~~:~"1~
~~"'.?.~1-.~~~~~--~~
0
." ~~- !.. f:~..c;~~.- ~-. '~. . ., +
'.:_ . : . .. :~~.~.--~~-k~~
. ...:~~~~ .t-~ .,;t,~~!,'t>.~.>!il"5d1?::;n
..~- ~~~~.j ::'~!~
RSAStages
Master Promp!
Detailed Prompt
A.2 Quality,
Conditions, and
Obstructions
construction
post
construction
./
./
./
./
A.l.2
If no sidewalk is present, is
mere a walkablc shoulder (one
wide enough oo accommodate
cyclists/pedestrians) on the road
or other p:uhwayhnil nearby?
./
./
./
./
A.l.3
./
./
./
./
A.l.4
./
./
./
./
A.l.5
./
./
./
./
A.l.6
./
./
./
A. l.7
./
./
./
./
A.2.1
./
./
,;
./
A.2.2
./
./
./
./
A.2.3
./
./
./
./
A.2.4
Is me walking surface
adequate and well-maintained?
./
./
./
A.3.1
Arc sidewalks/walkable
shoulders continuow and on
both sides of me succt?
./
./
./
A.3.2
./
./
./
A.4.1
./
./
./
./
./
./
./
./
./
./
./
./
./
./
./
./
./
Connectivity
A.4 ugbting
A.4.2
A.5 Vtsibility
design
.>\.IJ
A. I Presence, Design,
and Placement
planning
A.5.1
A.6.1
A.6.2
A.6 Driveways
_,
-1
./
./
Source: Pdatritm &ad Saftry Amiit Guitklina tmd fumpt Lim. FHWA. 2007.
..
./
.
Alternative Safety Studies lS:;;;iii'
The field review is the primary focus ofSrep 4 and leads to the findings reported in Step 5. Prior to the site visit, auditors review any available crash data and collision diagrams if evaluating an existing site. This review will likely suggest
where, when, to whom, or during what events crashes are taking place and provide a good starting point for looking
at potential conflictS or problem areas in the field. In addition, it may be useful to take prompt lists along, especially
if the team is new or has not conducred an audit in some time.
It is a good idea for the audit team to travel together to and from the site to encourage discussion among team members.
As mentioned earlier, it is a good idea to designate a secretary and photographer to doaunem any observations during
the visit. If possible, the phocographer should sit in the front passenger seat to document problem areas from the driver's
point of view. This can be done using a still or video camera. It is a good idea to do a drive-through of the site along the
mainline and on aU the approaches and movements during both daytime and nighttime conditions, and even during
various peak traffic conditions. A nighttime visit is especially good for looking at street lighting and sign rcrroreflecrivity;
Conducting multiple visits gives the team an appreciation for the driving tasks and drivers' workloads under aU rypes
of conditions. An example could be a side-street permitted right-rum moVWtent with prohibited right-rum-<~n-red.
Driving this approach may reveal a signal compliance issue with the green ball indication, with drivers not yielding to
pedestrians during the permitted green phase. VISits to larger sites with multiple intersections may restrict the ability to
drive each corridor. In this scenario, a video and/or log of the points of interest should be taken while walking the site.
If the RSA is conducred on an existing site, it is wise to walk the site. There is no substitute for getting out of the caJ
and walking the site, especially if the site includes pedestrians. RDadside and pavement conditions are much more
apparent when the auditor is not in the vehicle. Push buttons, skid marks at the stop bar, roadside hazards such as
wood poles, multiple driveway enccances, broken curb cuts, protruding $idewalk tracks, pavement markings at the
crosswalks, signs and sight obstructions at driveway and side Street openings are all a:unples of things that ace much
easier to see out of the vehicle.
During the drive-through or walk-through, it is ideal to record observations direcrly on an aerial image or plan drawing
of the site. However, the drawings may be too large. Another good method for capturing information is to take pictures
in an orderly manner as the auditor walks or drives (at, for aamp!e, intersection signs). When taking pictures, auditors
should malcc sure to capture the intersection signs as they go. so they are aware of where the picture was t:aken. In addition, the photographer can provide a piaure number to the secretary to include in the notes during the drive or walk.
5 crashes/year
Occasional: I to 5 crashes/year
Infrequent:< 1 crash/year
'
~
'
Accident
Frequency
Category
Source: &ttd Saftty Audit Guilklinn. FHWA; 2006.
Once the analys.is is complete, a report of the findings is prepared fur the local agency. This report should include
background information related to the project being audited, the RsA team members and their expertise and materials provided to the team, all site visit dates and times, important notes (such as weather or unusual uaflic events),
prioriciucion method used, the overall site observations made during the visits and a list of the safety issues and suggested mitigation (usually one per issue).
2. 4.2 Presnumg FnuJn.11
Once the issues are prioritized (typically by risk and severity), a preliminary findings meeting is hdd with the local road agency. This is Step 6 of the RSA process. Unlike the preliminary audit meeting, this is more a roundtable
discussion. The idea is for the local road agency to have an opportunity to ask qtlescions and seek clarification of the
RSA findings. The meeting also provides the agency an opportunity co make additional suggestions for improvements
in conjunction with the RSA team. However, the meeting is not an opportunity for the local agency to persuade the
audit team to eliminate any concerns. Any and all safety concerns should be documented by the team in the RSA final
report; the local road agency may provide explanations in their formal response letter. The discussion items in this
meeting will be useful and can be included in the RSA report.
2.4.3 Prqlaring 11 FontJIIl Re4p0tue
The local agency is ~ponsible for fullowing up on the issues revealed in the final reporr. This is done in the form of a
formal response letter and is Step 7 in the RSA process. The response letter is relatively simple; fur each audit issue, the
agency should identify wlut action will (or will not) be taken, with a brief explanation. The response letter is provi-ded
to the RSA team and becomes part of the project record with the RSA report.
2.4.4 lru:orporllte Fnulinp
The final step in the RSA process is the responsibility of the local agency to incorporate the findings from the audit
team. The form21 response provided to the audit team is added to the RSA summary report and given to the local
agency for reference throughout the remainder of the construction project. The idea lx:hind an RSA is to be proactive
in reducing d\e potential for collisions; therefure, the findings should be conside.red seriously fur implementation as
soon as possible. Safety is the primary reason for conducting an RSA; agencies should rernemlx:r findings not imp lemented could be used against the agency ifan agency is later proven to be negligent. .
'
Alternative Safety Studies 3.89
3.1 Introduction
Traffic conflicts are interactions between rwo or more vehicles or road user.; when one or more vehicles or road users
take evasive action, such as braking or weaving, to avoid a collision (Parker and Zegeer, 1988). Engineers use traffic
conl!icts as a supplement to uaffic collision srudies in estimating the traffic collision potential at an intersection or
other location. Traffic conflicts arc useful because the srudy resulrs are often available much sooner than the resu lrs of
traffic collision srudies (when several years' data may be needed). Traffic conflict studies can also provide much more
derailed informacion than traffic collision srudies. However, conducting traffic conflict studies is nor simple. When
performed improperly, they may provide misleading information.
Traffic con8ict studies require a relatively small investment of time and other resources and require no special equip
ment. Trained observers watch traffic and note on a form when a conl!ict occurs. Observers usually require a week
or less of training. On a single intersection approach, one or two persons are usually needed for one-half to 3 days.
Besides training the observers, engineers establish srudy guidelines and analyze results. Research sponsored by FHWA
during the 1980s improved the state of the art of traffic conflict studies. Migletz. et al. (1985) demonstrated traffic
conllicts predict future traffic accidenrs about as weU as collision records. Manuals by Parker and Zegeer (1988, 1989)
provide excellent detail on how to conduct traffic conflict studies.
Traffic confucr studies supplement traffic collision studies in several ways. The magnitude of the traffic safety problem
at a particular location can be estimated from traffic conflicts. One possible result of a traffic confuct study at an inter
seCtion is a mean rate of traffic confucrs of a particular type per day. This rate may then be compared to a standard or
certain percentile rate from a sample of similar inrerseetions. Trearmenrs may be needed at the location ifthe ~bserved
mean rate is higher than the comparison rate.
The use of ua.ffic con8iets for estimating the magnitude of a safety problem is restricred due to the lack of good com
parison confliCt rates. It is time-consurciog to collect a database of traffic conflicr rates for comparison purposes. Glauz
and Migletz. (1980) and Migletz. et al. (1985) provide comparison rates for some common types of interseCtions. However, if the intersection types for the published confuct rates do not apply to the location under study, traffic confucts
may not be hdpful in estimating the magnitude of the safety problem.
Traffic con.l!ict studies are very useful in determining the types of safety problems that exist at a location. Once the
type of problem is known, possible countermeasures can be identified. There are, for instance, 14 basic types of traffic
conftiets at imersections (Exhibirs 18-4 through 18-17). The rd ative overabundance of one of these 14 types at an
intersection reveals a particular problem. Again, a database of typical rates must be available against which to compare
a location's rates. Lisrs of countermeasures that may reduce the occurrence of a type of conflict are available (Parker
and Zegecr, 1988). At many highway locations, it is impossible to obtain enough rdevant collision data to make such
a detailed diagnosis.
Traffic confucr data are often collected and analyzed to evaluate the effectiveness of a safety-related countermeasure.
Countermeasure evaluation using traffic conflicrs is attractive because traffic conflict data arc available to the analyst
before ttaffic collision data. In fact, a before-and-after srudy with traffic confucts may not need comparison or control
sites to overcome the history and maturation threats to experiment validity (see Appendix A for details on experiment
design). Using traffic conllicrs as the measure of effectiveness in a countermeasure evaluation may also eliminate the
threat to experimental v:ilidity posed by regression to the mean, since countermeasures are typically chosen on the
basis of collision data rather than conllicr data. To evaluate countermeasures, the conflict types being studied should
be closdy rdated to the countermeasures that have been implemented. If not, the true effeCtiveness of the counter
measures will remain. unknown (Parker and Zegeer, 1988).
; rear-end collision with a slow-moving vehicle during a green signal phase is involved in a traffic conAict. ObservtJS use
;,brake lights, squealing tires, or vehicle front ends chat dip or dive as indications that braking occurred and a conflict
'was possible. A collision or near miss during which no evasive actions were observed also counts as a traffic conflict.
Traffic conflicts can involve motor vehicles, pedestrians, bicycles and other road users. Rates of pedestrian and motor
vehicle conAiccs can be high ar intersections with appreciable pedestrian volumes.
Researchers have identified 14 basic types of con!Uccs at intersections, as shown in Exhibits I B-4 through 18-17. T)l.ese
conflict types apply at signalized intersections, unsignalized intersections and driveway openings. However, not all che
14 types apply at every interseccion. In mosr conflict studies, observers record only the conflict types that are related
to the study purpose rather than all 14 types. Traffic conAict types are not well defined for nonimersecrion locations
such as weaving sections, diverges, or merges. Preliminary observations or pilot tests are necessary to inform obserVers
which conBiccs to record at nonintersection locations.
III
~~
.
I I~L
--r1!!1 ~ ~ ~
- -~
---
---
I I
---.--J'l iL _j
' I IL.
f!--'
P
JI.
I'
I
---
t----
--.-
__. -
I~
t- - -
__ -
..
I I
I II
I
_ji
- .- -
!L
---
_j
---
!~
~---
I!
.I
I.IL
.
L I
-- -.;f-. _ _.,
l
.
lrl
1
I .
I I
_j' lI i'L
- - --- ---
--- --________
__
lr-
~CIJJ
~~
II
l .
I
"L
I.
~rn
_j
- .-
.-
'l ~
-
. I ~I
I
i
i
L
_j
___ . I#. __
. -
111
I
.1
. !
I
ccc-==-..-. .
~I
....;
---
--~
- .- -
~~~~~
I, I I . I ~~
.
I,
So=: G!auz., W. D. Uld D.]. Migle~ Application of
Tniffo unfticr Anaiym 41 lntmtiiJns. Natioml Cooperative
Highway Rese=h Program Report 219. TRB, 1980.
JilL .
rc
- ~~,'<'td'~
. :_
:;:~-e,
.
~:I:UllillMt!~~
.. 1 - _ -(:.l.lLI:
....'!;',~
...
.....
~ ;,_~
IL
g;IJ
t:l---
~- -~-.
~II
~~-
---
IL
I I
- --
!r- I !
.....,
fm-~-~~!_:;
-----
I*
I
I
be able w observe craflic con fliers and coum turning movements on an approach where three or fewer movements with
low or moderate volumes are made. If there are more than three movements or volumes are heavy, rwo or more observers will be needed. Observers should not have to look away from the location where conflicts are b eing watched
' to record turning movements. If turning movements are to be counted at an intersection roo far down stream for an
observer tO see both the approach conflict area and the intersection, a second observer counting only n uning movements is needed.
A traffic conflict observer should watch only one intersection approach or one end of a weaving a rea at a time.
Consequently, when it is desirable co study traffic conflicts at an entire intersection or weaving area, either larger
crews of observers must be used or particular observers must stay longer at the location. For example, if traffic conflicts per vehicle are desired on all four legs of a busy intersection and a full day of observation is needed on each
approach tO gather the appropriate sample, staffing options include:
1. a crew of six (four conflict observers and rwo turning movement coumers) is scheduled for 1 day;
2. a crew of three (two conflict observers and o~e turning movement counter) is scheduled for 2 days; and
3. a crew of rwo (one conflict observer and one turning movement counter) is scheduled for 4 days.
A traffic conflict study requires very little equipment. Observers will need forms, clipboards, pens, watches and a
place to sit (a vehicle or a folding chair). Electronic or manual turning movement recorders may easily be modified for traffic conflict studies with a template or by labeling keys so each key is associated with a conflict t)'l'e.
However, traffic conflict totals tend to be smaller than turning movement counts so electronic boards are usually
not required. Observers have used audiotape recorders co record data during traffic conflict studies. T he extra time
necessary to listen to the tape and code the data later makes this alternative less desirable. Finally, video can be used
during traffic conflict studies,.A digical video creates a permanent record so close calls can be re-evaluated. The
disadvantages of video, including "the extra labor to record and view media and the technical problems associated
with lighting and fields of view, usually outweigh the advantages. A well-designed data form remains the best choice
for most traffic conB.ict studies.
~~r-<-'""
.7.
~T
. -- !.~,.D,l-ll'/
..
:r:i...
..
...
0
""'
. . . ;..>;
. 0 .
Period
-::r.:.....~!~""!l"ff.--..- r-;:~
Topic
Introductory remarks
Oricnration to the training program
General background on tnffic safety
History of tnffic conflicu
Day2
Day3
Day 4
Question-and-answer session
Special con6ct types
Simulated Umited conflict counts
Day5
Discussion
lntcrsection.s Yt'ith unusual geometries
Use of other data forms
Day6
Day7
D~y8
Day9
rhylO
Source: .Parker. M. R. and C.'(. Zcgcer. Trrz/M Cowjlia Ttdmil:puJfor SafnJ tmd OpmaioN: Enginm's G..Uk, FHWAlP-8S-026. FHWA. 1988.
'lK MANI IAI (lf TRAN<;PORTATiflN
'
Observer consistency can be estimated by having two or more observers record conflicts independently at the same
location and time. If an observer at a particular type oflocation produces results that are consistent with other observ_ers, especially with experienced traffic conflict observers, his or her uaining may be considered complete for th at type
oflocation. Engineers setting up such a consistency test must make sure the observers see the same ponion of roadway
and are not influencing each other. An informal and usually sufficient method for judging consistency is simply to .
look over the completed data forms from different observers. A variety of statiscical testS are available for formally
estimating consistency. A correlation coefficient, a Z-test for proportions, or a paitcd comparison t-test can be used
to estimate the difference between two observers. A group standard deviation, a chi-square test, or an analysis of variance can be used to estimate consistency among a group of three or more observers. An F-rese can be used co compare
the variances between two groups of observers or between two trials with the same group of observers. Appendix: A
contains a discussion of experiment design (a formal comparison between observers is an experiment) and Appendix
C contains a discussion of common statistical tests.
There are few standards for observer consistency to indicate whether more uaining is needed. Based on total conflicts
at an intersection approach in 15-min. periods, Parker and Zegeer (1988) stare a correlation coefficient of 0 .95 be-
tween two observers is desirable. They also recommend observers whose conflict counts consistendy fall one or more
standard deviations above or below the group mean should be singled out for more training. However, if the gro.up
standard deviations are small, the group is ready for unsupervised data collection andsingling out particular trainees
for further training would waste time.
'fiL
Note: Basic ~terscctlon conflict types not shown had mean hourly rates less than 0.5. Statistics are based on sample
counts conducted in the Kansas City metropolitan area on all four approaches of signalized incccsccrions and on
the approaches with the right-of-way at unsignaliz.ed intersections. Counts were taken during the daylight, in dry
weather, and do not include secondary conflicts.
a. "All same direction~ includes left-tum same direction, slow vehicle, lane change, and righr-turn same direction conffiet types. "Through cross traffic~ includes cross traffic from left and cross traffic from right conflict types.
b Not available.
Source: Gbuz, W D. and D. J. Migletz. Appliemion ofTraffic Conflict Artal]sis at lntn1mions. Natio.W Cooperative Highway
Reseatch Program Report 219. TRB, 1980; ~liglea, ct al. Rrlationships bmutm Traffic Conflias andAiJmiJ. FHWARD-84-042, 1985.
398 MANUAL OF TRANSPORTATION ENGINEERING STUDIES. 2ND EDITION
:&<I
---"'
'
. ... . . . .
Consta.nt,
> ;'-'-:
,.. c.
~-'!.j'1.<<,~?,-~~.r-:-~~-i~~$.C:;:;
>:t
-- -
"-.~-.... _"!",..._~-----~:~"":"..;;-~:~~
1.28
80
1.5
86.6
1.64
90
1.96
95
95.5
2.5
98.8
2.58
99
.........~~
---
Once the rype of rare is chosen, the necessary sample size can be calculated. The equation for determining the salflple
size needed to estimate a mean conHict rare per unit of pme, if previous estimates of the variance of the mean and. the
mean arc available," is:
[(1oo -'-)
~
PC
mean
2
NT
Equation
18~1
where:
NT
PC
permirred error in the estim~te of the mean conflict rate, percent (if the mean hourly conflict rate is 6 and
PC = 50, the precision of the estimate is 6 t 50 percent of 6 or 3 to 9 conflicts per hoUr)
var
expected variance of the conflict rate, from previous studies or Exhibit 18-19
mtan expected mean of the conHict ate, from previous studies or Exhibit 18-19
Equation 18-1 can be used for conflict rates per any unit of time if prior estimates of the mean and variance of tPC
rate are available. If a previous estimate of the variance of the mean is available but not rhe mean itself, the equation
becomes:
1
NT
(...!._)
PQ
Equation 18.-2
var
where NT. t and varare as defined for Equation 18-1 and PQis the permirred error in the estimate of the mean coPBier rate, in confficrs per unit of time (if the levd of confidence corresponding ro tis 90 percent and PQ is 7 conffic-.:s
per hour, the acrual conflict rate will be within 7 conflicts per hour of the estimated rate 90 percent of the time).
EXAMPLE 18-1
For a.n approach to a signalized intersection with a total enuy volume of 35,000 vehicles per day, an engineer wan-es
an estimate of the mq.n number of same-direction confficts in an hour. The engineer would like to be 90 perceot su.t:"e
the estimate is within 10 percent of the actual rate. Using Equation 18-1 and Exhlbits 18-19 and 18-20,
NT=
[(too 1.64
f] .~
10
90
2
If conffict rates per vehicle arc desired, the sample size necessary to achieve a certain precision in the escimate of .::;;a
mean conftict rate is:
..:..'r
.,
;:
pq rz
NV = ~~
ppz
Equation 18-3
PP permitted error of the estimate of the proportion ofvehicles involved in conllia:s, in a proportion~ 0 and 1.
If the levd _of confidence corresponding tot is 95 percent and PP is 0.01, the actual conHict.ratc per vehicle will be
within 0.01 of the estimated rate 95 percent of the time. The sum ofp and q mwc be 1.0 in Equation 18-3.
A consetvative estimate of the sample size (that is, a larger sample will be gathered than probably is necessary to
achieve a given precision) can be provided &om Equation 18-3 without prior knowledge ofp and q. Ifp and q arc
unknown, they are assumed as 0.5 and Equation 18-3 reduces to:
11
Equation 18-4
Since conHict rates per vehicle are wually much closer to zero than to 0.5, use aJ). estimate of p and Equation 18-3
Hthe.r than Equation 18-4 to reduce the sample size estimate dramatically.
EXAMPLE 18-2
An engineer wants an estimate of the mean rate of right-turn same-direction conflictS per approach vehicle. The
engineer would like to be 95 percent sure the estimate is within 0.01 of the actual rare. Using Exhibit 18-20 and the
conservative Equation 18-4,
Nv = 0.25 1.9621
0.01
- 9,600.
Therefore, 9,600 approach vehicles would have to be obsctved to achieve the desired precision. If the engineer is certain that the con.ll.ict race is tess than 5 pcrcenr, Equation 18-3 can be wed and the needed sample size is much smaller.
1,800 vehicles
Inverse sampling is another sample size formulation for proportions chat may be useful for traffic conHict studies wing
races per vehicle. Inverse sampling applies when the proportion to be estimated is known co be small (for example,
less than 0.10) bur no rdiable estimate of the proportion is available. The sample size formulation given in Equations 18-3 and 18-4 may require a larger sample of vehicles than is necessary for a typical estimate. Inverse sampling
depends on the coefficient of variation, which is the ratio of the standard error of the proportion to the proportion
being estimated. The lower the coefficient of variation, the more precise the estimatt. Cochran (1977) bas shown:
CV <
\
\
\
.Jm
Equation 18-5
m-1
where CVis the coefficient of variation and m is the number of occurrences of the event to be obsetved (that is, the
number of conflictS). Exhibit 18-21 is based on this inequalicy (Equation 18-5). To achieve a certain coefficient of
variation, one keeps observing until m con.ll.ias arc counted. For instance, observing an approach until 27 coafficts of
a particular type arc recorded. guarantees a coefficient of variation ofless than 20 percent. Exhibit 18 21 illwuatcs how
difficult it is to gee a precis~ estimate of a very low proportion.
-- , ........... _,..__ , ,. .. ,. , .. ,,.,.nu.t,.- I""TI H''IICC "''MI"\ c:nmnM
L
INTERSECTION TRAFFIC
Location _ __ _ _ __ _ _ _ _ _ _~-:-:-:-:-~::\---Date
Observer(s) _ __ __
C"' Conflict
Day
Ql
Ql
F--.
"0
.... ~
"' "'
c..
...
c ::z:
......
~ ~
Vl=
:::1
t:
Ql
Ql
E
:::1
0
>
s:.
0
"'e
Q.
Q.
Left-Turn
Same
Direction
sc
Right-Turn
Same
Direction
sc
Slow
Vehicle
Lane
Change
c sc
c sc
Opposing
Left-Turn
Right-Turn
From-Right
sc
c sc
<t:
Total
c +sc
Daily Count
Rate Per 1,000 Veh
Source: Parker, M. R. and C. V. Zegeer. Traffic Conflict uchmqun for SaftlJ and OpmuioiU: Enginm's Guitk, FHWAIP-88-026. FHWA. 1988.
CONFLICTS SUMMARY
_ _ __ L.eg
Numbl";~~h
<$)
of Reconliog "'riod
c sc
c sc
c sc
c sc c sc
c sc
All Same
Direction
sc
All Through
CrossTraffic
...<II
.s::
....
0
sc
..
;
-
------
i.: .
.
~.
--
..,
..
;.
' lilt
'-4-
..
' r -.;
.....
ACfORCODES
t~
. . ........ .. .~~:/i'T:<to
~~~~r~
.
~~
.').;
ACTION CODES
. .~.
' ...~.J~~~~~~
Name:
Date:
Tune Peria<i:
Intersection:
Direction Oeg with actor 1}:
Weather:
Tune
Actor 1
Action
Actor 2
Action
Comments
'
Source: Hwnmer ec al. An &alsumon ofuJUiing W11VS lAgging J4i Tt.m Signm Pluuing. Fi711ll &port, FHWA-IN
JHRP-89-17. Purdue Univenicy,Joim Highway Research Project, 1989.
Exhibits 18-22 and 18-23 conu.in forms for recording mffic con8ias (the forms are reproduced in Appen<lix E). The
form in Exhibit 18-22 includes 12 of the 14 basic types of con8ias expected on intersection approa~es. Only the
I:WO types of basic pedestrian conflictS are omitted. Observers enter one line on the form in Exhibit 18-22 after each
obscrva.tion period (that is, each 20- or 25-min. period bei:WCCn breaks). A mechanical or electronic coumer may be
helpful in keeping a running total of con.llias by type during an observa.tion period. Inappropriate columns of the
form in Exhibit 18-22 could be voided in advance to prevent the mistaken entry of data or an alternative data sheet
could be made by the analyse prior to collecting data.
Exhibit i8-23 shows a form that requires one line be entered for each conflict. For each conflict, observers record
the rime, the position of each involved vehicle (actor), the movement of each involved vehicle (action) and comments that help describe the con.llict or event. Analysts should develop codes for actors and actions to meet the
individual needs of each study. Exhibit 18-24 shows a sample form with actor and action codes that was developed
for a conflict srudy concentrating on left rurns. With a form such as that given in Exhibit 18-23, no counter is
needed, obse.rvcrs have much more flexibility in coding unusual occurrences and analysts have very detailed data
to work with. Also, recording the rimes of conflias makes it easier to chedc observer consistency during training.
However, the form requires more data reduction and may distract observers (while writing. they may miss some
conflias) at sites with high numbers of conflias.
\
\
r.
'J
. ... ,._.
~~<;;"""
'
,...
~--.
=t -~~JJ. .
........ ~-
. .
!~
D
~
,1
lo~
s
1l
~-
,.,.. ._
~:.-
.~ .. ...-....
-~:;-~~~A
ACfOR CODES
f.
,.
ACTION CODES
Name:BMS
Dare: 12119/09
!:\
S Srop
D ~ Decderate
i~
A Accelerare
B BackUp
w .. weave
Ttme
Actor 1
Action
5:03
5:05
5:05
Actor 2
.Action
l1
s
s
"
3
11
11
5:07
5:12
5:13
5:16
5<17
5:18
5:19
Comments
--
Several key points apply equally to traffic conflict dara forms with one line per observation period (Exhlbir 18-22) and one line
per conflict (Exhibit 18-23). Fusdy, c:xrensive resting of and practice with a. form is CS$ential. Secondly, observers must complete
all header" information or the dara on a form are useless. A critical piece of header information that is often forgotten is the
direction of the approach leg being nudied. Thirdly, every form should bave a place for comments, which are important during
dara analysis.
JllliL
uemely large, this is best accomplished by manually scanning the completed data forms instead of by a computer program.
A manual scan allows flexible scoring of conflicts by type and allows the analyst ro obtain a good feel for the data. During
data reducrion it may also be appropriate to sum counts from individual approaches into a coral for an intersection.
If conflicts per unit of rime are of interest, analysts need to adjust for unobserved rime periods. For example, suppose
a particular study calls for an hourly conflict rate. The observers used 20-min. data collection blocks with 10-min.
breaks becween blocks so 40 min. of data were collected per hour. Therefore, multiplying the number of conflicts in
rwo adjacent blocks by 60/40 provides the needed rare per hour. Analysts should adjust for unobserved time periods
with data from similar time periods that were observed, not by assuming constant conflicr rates throughout a long
period. A total of 30 conflictS recorded from noon to 3:00 p.m. should not be expanded to a standard 11-hour day
(7:00 a.m. to 6:00 p.m.) coral of 110, because the 3 hours observed did not covet either the a.m. or p.m. peak hours:
It usually would be permissible to use a 3:00 to 4:00 p.m. count of x conflicts and a 5:00 O 6:00 p.m. count of y
confHas to esrimarc the 4:00 to 5:00 p.m. count would have been (x + y)l2 conflicts.
Conflict rates per vehicle observed are produced by combining the conflict sum and the appropriate rurning movement
count. An intersection or approach count should be done at the same time the conflict counts are recorded.
Once the data are reduced, the purpose of the srudy should guide analysts. For many studies, mean rates of particular
types of conflictS and the accompanying standard deviations are necessary. Appendix C provides descriptions of and
formulas for conunon statistical procedures. If the srudy purpose is to determine the magnirude of the safety problem
or to generate countermeasures, comparison conflict rates are needed. For agencies without their own comparison
rates, Exhibit 18-19 provides typical rates for some types of conflicts at intersections.
4.1 Introduction
Most U.S. hlghway agencies post advisory speeds for some changes in horizontal alignment. The first known record
for setting advisory speeds goes back to 1940 (Moyer). Eight years later, the MUTCD adopted the curve advisory
speeds for implementation throughout the United States.
Recent research sug,aesrs that methodologies commonly used for the determination of advisoty speeds for changes in horizon raJ aligrunent can be improved upon with the use of updated design criteria that are more reflective of current driving
conditions. In 2010, ITE approved a new informational reporr, Methodologies for the Iktmninmion ofAdvisory Speedr describing three methodologies that may be used to determine advisory speeds including: design speed equation, traditional
ball-bank indicator and the newer accelerometer method. These three techniques are very closely related and should provide
the same or similar advisory speeds. It is expected many of the procedures for setting advisory speeds used by agencies will
now have a standard criteria to compare agaipst, rhus eliminating many of the very conservative speeds currendy posted.
Three acceptable methods bave been developed for determining advisory spds along horizontal curves: design speed
equation, traditional ball-bank indicator and the newer accelerometer method. The three techniques are very closely
related ar1d should provide the same or similar advisory speeds. It is expected many of the procedures for setting advisory speeds used by agencies will now have a standard criteria to compare against, thus eliminating many of the very
conservative speeds currently posted.
..___....,,
"==. . . . --.
-- ------.-
.::~
------
The accelerometer is similar to a ball-bank indicator but measures laceral acceleration only: Some accelerometers, such as
the one shown in Exhibit 18-27, will automatically correlate the laceral acceleration to a ball-bank reading. Either thdace
acceleration or the correlated ball-bank reading can be used to determine the po-sted advisory speed.
Both ball-bank indicators and accelerometers arc very easy to inscall. They can both be mounted permanently; however, they are cypically mounted semipermancntly with suction cups or Vdcro strips. Both unitS have methods for
zeroing. or "kvding." when placed on a flar surface. Further information on how to use the equipment to collect the
various data elements is discussed in Section 4.2.4.
second person be available to manually record data while the driver drives a constant speed and pays attention to !he
road. The accdc:rometer method only requires one person because the device will store data, allowing one to download
the data in the office.
Ball-bank and accderometer me!hods require !hac the respective devices be calibrated prior to collecting any data. The
easiest method for calibration of either device is to use a low-volume undivided roulway and center the vehicle over
the centerline of !he roui along a lengthy tangent section. The driver and passenger should not change places once !he
device is calibrated as a shift in body roll and/or suspension may .cause a change in de8eccion.
The speedometer of the vehicle should also be calibrated prior to conducting any studies. This should be done every
5 mi.lcs per hour (mph) (8 kilometers/hour [km/h)) on a periodic basis, ideally with a laser speed gun. The calibrated
speed should be used when conducting the acrual ball-bank or accelerometer study. Tires should be in8ated to the
proper pressure so body roll is not signiJi.cancly affected.
4.2.3 s-pk Siu R.tqui-mttnu
For each of the three study methods described, great precision of the cscimate is not critical and a reasonable cstimace of
the average will yield an appropriate wvisory speed. It is a good idea tO take three to five measurements for each method
of analysis. If the design equation method is used and radius is needed, a minimum of three measurements along different parts of the curve will suffice. The superclcvation can be dcrennined using a levd ar different locations along the
curve. For the ball-bank indicator and accelerometer methods, a minimum of three runs should be completed ac each 5
mph (8 kmlh) increment in both directions of cravd, for a coral of six runs per 5 mph (8 kmlh) increment.
4.24 Mu.tfll'ing RMiiru s~
The design equation method requires the curve radius and superclevation. The easiest way co coUcct curve radii data
is to overlay circular templates onto an aecial image. This can be done using hand-drawn or computer-generated templates that are scaled to the referenced accial image. An alternative is co use the chord and middle ordinate" method
which is slighdy more cime-ronsuming. Using this method, the equation for dcrermining the radius of a curve is:
l2
R=-+Bh 2
where:
R ~ curve radius. (ft.)
I chord length.(ft.)
h
1.
Equation 18-6
ca
Superelevation is another input into the design equation. Superelevadon (l") can be determined from plan drawings, but in many cases it will need to be determined in the field. If a field calculation is needed, a carpenter's
level can be used by taking one end of the level, laying it on the pavement and raising the other end until th.e
bubble indicator reads true. The superelevation is measured as the vertical distance divided by the horizontal
distance (the length of the level) and expressed as a percent. As indicated earlier, measurements should be taken
in several locations, ideally in the center of each lane. The minimum superelevation in the curve should be used
for determining the advisory speed.
Source: Seyfried, I<. and J. Pline. "Guidelines for the Determinacion of Advisory Speeds." rrE}tnmutl, January 2009.
Equation 18-7
where:
Radiw (ft..)
Sour= Seyfried, K. and J. Pline. "Guidelines for the Determination of Advisory Speeds." ffE]ournal, JanU2rf 2009.
Exhibit 18-30 shows the calculated values for standard radius and superelevacion. The data sht provided in Appendix E-46 is useful for recording field-collected clara and documenting the work done at thar site.
Once the driver and data collector are ready, an assumed starting speed should be determined. Two simple methods
for assuming a starting speed are 1) arbitrarily choosing a speed 10 mph (16 kmlh) below the posted s peed limit or
2) drive 5 mph (8 km/h) below the driver's comfortable speed. As mentioned earlier, the driver should make sure h e
or she is driving the "calibrated speed" and not the speed noted on the speedometer. A minimum of three passes in
each direction is necessary to obca.in a fair reading of the ball-bank indicator. This will yield a minim ttm of six tot al
passes through the curve at each speed until the desired advisory speed is determined. The speed should be inccea.sed
or decreased by 5 mph (8 km/h) until the average ball-bank reading is not exceeded for either direction o f travel based
on Exhibit 18-29. Data collection forms are available in Appendix E-47.
Direction of
Travd
North
13.4
South
11.4
North
ll.S
South
10.0
Solution: The ball-bank results indicate an advisory speed should be posted. Based on the results, the northbou~ d
direction of travel q>nsistently yields higher ball-bank readings (the superelevation is most likely lower in th::3-t
lane); therefore, this 'will be the direction used to set the advisory speed. The first set of 35 mph (56 km/h) runs i..Jl
the northbound direction yielded an average ball-bank reading of 13.4 degrees, which is higher than the recoif.1
mended reading based on Exhibit 18-29. The speed was reduced by 5 mph to 30 mph (8 km/h to 48 km/h) and tb-e
crew made three more runs in each direction. As expected, the northbottnd direction again yielded higher reading-S
The average ball-bank reading was 11.8 degrees, which falls below the maximum required reading of 12 degrees.
Therefore, the agency should post an advisory speed of 30 mph (48 km/h) along this curve.
4.3.3 AccekrtmJeter
The accelerometer method is becoming a more popular method for determining advisory speeds. The accelerometer is very similar to the ball-bank indicator, with only two major differences. Firstly, this method only requires
one person for the study since the accelerometer will store data that can be downloaded at a later time and date.
Secondly, an accelerometer measures lateral acceleration only (measwed in "g~s), and does not take into accounr
body roU or superelevation. However, each of the three methods has been calibrated and should give roughly the
same advisory speeds.
The srudy is conducted in exactly the same manner as the ball-bank indicator method. The lateral acceleration is
read from Exhibit 18-29 in lieu of ball-bank readings; however, some accelerometers will convert lateral acceleration to b<!ll-bank readings. The data collection forms associated with the ball-bank indica~or method in Appendix
E can easily be manipulated to account for lateral acceleration.
5.0 REFERENCES
American Association ofState Highway and Transportation Officials. A Policy on QotMtric lksign ofHighways and Struts
("Ac.cderometer and Ball-Bank.") Washington, DC: AASHTO, 2004.
Austroads. Road Saftty Audit Guuulines. Sydney, New South Wales, Australia: Austroads, 2002.
Cocliiail, W. G. Smltpling TtthniqUIJ, 3rd cd. Ne'iv York: Wiley, I9n: pp. 76-n.
Federal Highway Administration. Manrilll on Uniform Traffic umrrol Dtvices for Strem and Highways. Washington, DC:
FHWA, 2009.
Federal Highway Administration. Ptdestrian Road Safety Audit Guwlines and Prompt
Administration, 2007.
Federal Highway Admini.stration. &ad Saftty Audit Guuulirus. Mclean, VA: Federal Highway Administration, 2006.
Glauz, W. D. and D.]. Miglett. Application ofTra.ffic Conflict Analysi.s at Intersections. National Cooperative Highway
~rch Program Report 219. Washington, DC:Transportacion Research Board, 1980.
Hummer,]. E., R. E. Montgomery and K. C. Sinha. An Evalrilltion ofuadint~JmUS lAgging Lift Tum SignAl Phasing, Final
Report, FHWAIIN/]HRP-89/17. West Lafayette, IN: Pusdue University, Joint Highway Research Project, 1989.
Miglea, D. ]., W. D. Glauz and K. M. Bauer. &/4tionships bnwn Traffic Conflim and Accidmtr, FHWA/RD-84/042.
Washington, DC: Federal Highway Administration, 1985.
Moyer, R. A and D.S. Berry, "Marking Highway Curves with Safe Speed Indications," Proceedings of the High~y Research
Board, Vol. 20, 1940.
National Roads Authority. Design Manualfor Roatfs and BriJgn-VoiUtM 5. Dublin, Ireland: National Roads Authority,
2009.
Parker, M. R and<; V. Zt:geer. Traffic Conjlia Techniques for Safoy and Optrasions: Enginm's Guiik, FHWA-IP-88-026.
McLean, VA; Federal Highway Administration, 1988.
Parker, M. R. and C, V. Zegecr. TrtlfJic Conjlia Ttclmiqutt for Safny and Optrasions: Obsnwri Manual, FHWA-IP-88-027.
McLean, VA:. Federal Highway Administration, 1989.
Public Road Administration. Manrilll on Uniform Traffic Control Dtvim for Smm and Highways. Washington, DC: Public
Road Administration, 1948: pp. 39, 53.
Seyfried, K. ar~d]. Pline. "Guidelines for the Determinacion of Advisocy Speeds. ITE ]Du11141. Jar~uary 2009.
Trarisfund. Road Saftty Asdit Promiurts for Projtm. Wellington, New Zealand: Transfund, 2004.
Chapter 19
\ '
Roadway Lighting
Origitud By.L FJJis IGnr. D. Eng., P.E.
EdiudBy:
413
414
414
415
415
415
415
4.0
419
3.5 Collisions
420
423
424
424
424
424
425
5.0 REFERENCES
426
5. 1 Literature References
426
427
427
1.0 INTRODUcTION
he ability to dearly sec the roadway is essential for the safe and efficient Bow of ttaflic on our bighwaf3. However, in many instances, limi12tions of the hwnan eye prevent vehicle headlights alone from Cl)lllplctdy satisfying visual nighttime driving requirements. Fixed roadway lighting supplements vehicle headlights by extending
the visibility range both longitudina.lly (distance along the roadway) and latcrally (distance across the roadway), thus
aiding the driver by providing earlier warning of hazards on or near the roadway. Research shows the nighttime collision rare can be reduced by the provision of ad~uate roadway lighting (Box, 1971, 1972a, 1972b, 1989; Walker
and Robe~, 1976). lighting defines the roadway geometries, such as the edge of pavement, curves and dead ends,
and illwnirutcs obsttuctions in or near the roadway, including channelization islands, bridge pictS and parked cars.
lighting allows the driver to sec a pedestrian in the road?ny beyond the headlight beam and even before 'the pedeStrian enters the road. It also aids pedestrians by illuminating obstacles on the: sidewalk and roadway in their vicinity.
Rr"'~that:il\t l
inktinn
41~
Lighting raises the surrounding brightness level to which the driver's eye.s adapt and increases the driver's concrast
sensitivity, resulting in an overall improvement in the driver's ability co see. Fixed roadWlly lighring also conrributes ro
a more pleasant and comfortable night d.riving environment, which in turn reduces driver fatigue and improves driver
efficiency. Lighting is an aid to police surveillance, and a reduction in stree.t crimes may be experienced foUowing
installation of improved srreet iUuminacion. Auto theft, assault and vandalism are three of the types of night crimes
most frequently cited as being reduced. While the reported impaccs of Lighting on crime are statistically inconclusive,
there are srrong indications the fear of crime is reduced following increases in street lighting and chat feelings of safety
are higher (lien, 1979, Sherman et al., 1997). The negative aspects of lighting include glare, collisions with light
poles, initial insrallation costs and continuing maintenance and energy costs for the lighting system. Ught pollution
is a concern with roadway lighting and should be considered to maximize the benefits to roadway users and minimize
the unWllnted effects of sky glow, light trespass and glare (NLPIP, 2007).
Light sources include (Szary, Maher, Srriz.ki and Moini, 2005):
mercury vapor (MV)
metal slide
low pressure sodium
high pressure sodium (HPS)
HPS recro white
HPS resrrike
induction lighting (Icetron, quam light [QL])
fluorescent
comp:act fluorescent
light-emitting diode {LED)
solar
- -_
..
-- . .,._.,r ~ .. : ~
:f-:c
'"_._
~ --~,
...
---'
-.;;.
--~~-
Road aod.A.ma
Average Luminance,
Veiliog Luminance
Ratio (muimum)
Luminance Uniformity
to L
L. to L
Freeway class A
0.6
3.5 to 1
6 to 1
0.3:1
Freeway class B
0.4
3.5 to l
6 to 1
0.3:1
Commercial
3 to 1
5 to 1
Intermediate
0.8
3 tO 1
5 10 1
R=idential
0.6
3.5 ro 1
6 to I
Corrunercial
1.2
3 to l
5 to 1
Intermediate
0.9
3 to 1
R=idenrial
0.6
3.5 to 1
6 co I
5 to 1
Oasaification
L to L"""
Expressway
0.3:1
Major
10
0.3:1
Collector
Commercial
0.8
3 co 1
Intermediate
0.6
3.5 co 1
6 to 1
Residential
0.4
4 to I
8 co 1
0.4:1
Loc:al
Commercial
0.6
6 roI
I 0 ro 1
Intermediate
0.5
6 to l
10 tO 1
-R=idencial
0 .3
6 to 1
10 to 1
0.4:1
Wwninaocc
Uniformity Ratio,
R1
Pnanent Cla.ssi6c:ation
R2 and R3
R4
E_ toE_..
Freeway class A
3:1
Freeway class B
Commercial
10
14
13
Intermediate
12
10
Residential
12
17
15
lntermedb.te
13
ll
R=idential
Commercial
12
10
lntcrmedb.tc
Residential
CLusi.6cation
Exp_,_,3:1
Major
Commercial
3:1
Collector
o ,.
~"""'
t"f"'\r'Tf,....l
4:1
Notes:
I. L. veiling luminance
2. These tables do not apply to high mast interc:hangc lighting S)'ltcm.s (e.g., mounting heighu over 20 meters).
3. The relationship between individual and respective lwninance and illuminance values is derived from general conditions for
dsy paving and straight road sections. This rclacionship does not apply to averages.
4. For divided highways, where the lighting on one roadway may differ from that on the other, calculations should be made
on cac:h roadway independently.
5. For freeways, the recommended values apply to both mainline ;and ramp roadways.
For approximate values in candelas per square foot, multiply by 0.1.
\For approximate values in foorcandles, multiply by 0.1.
Source: Reprinted by permission from the IESNA Handhoolt, 9th Edition. By the 1lluminating Engineering Society of
North America.
Crosswalla rnversing roadways in the middle oflong blocks and at suect inreneaions should be provided with additional
illumination.
For approximate values in footcandles, multiple by 0.1.
'For pedestrian identification at a distance. Values u 1.8 mea:rs (6 feer) above
Source: Reprinted by pc.rmission from the IESNA Handbook, 9th Edition. By the lUwninacing Engineering Society of
North America.
Rl
0.10
Mostly diffuse
R2
Mixed
(diffuse and specular)
R3
Slightly specular
o representative mean luminance coefficient. (llluminat.ing Engineering Society of Noeth America, 1983, RP-8-83, Chap tee
16, Table 16-3, page 7.)
Sou=: Reprinted by permission from the IESNA HanJhqo!, 9th &JitiDn. By the Illuminating Engineering Society of
North America.
ITel:ll'
C,LAJJ,tCAtiON
'
--- -
M.Ao~o
ovT
COI.I. I Cto .
t.OCAt. . , ,
LANO
Ust
~
~
e::z ---,
~
co<~>'
...........,
..NttCM I OIA.fl
. .
EXAMPLE OF:
STREET CLASSIFICATION
AND
LAND USE
Source: Ammcm National Prtu:tice for &adway Lighting, illuminating Engineering Society of North America, 1983, RP-8-83,
FJgUre 1, page 6.
Fruway: divided major roadway with full control of access and with no crossings at grade. This definition applies
to coli and noncoll roads.
Expressway: divided major ro~dway for through traffic with partial control of access and generalJy with inter
changes at major crossroads. Expressways for noncommercial traffic within parks and parklike areas are gene...UY
known as parkways.
Major: that part of the roadway system which serves as the principal nerwork for through-rra.ffic Bow. The roo res
connect areas of principal uaflic generation and important rural highways entering the ciry; also known as arceri:Js.
Colkctor: distributor and collector roadways servicing traffic becv.een major and local roadways. Th ese are road
ways used mainly for traffic movements within residential, commercial and industrial areas.
Local: roadways used primarily for direcr access co residential, commercial, indUStrial, or other abutting pro~. LO
cal roadways do not carry through traffic. Long local roadways will generally be divided into shorter sections by c0l
lector roadway systems. Additionally. local streetS within residential areas should be grouped into rwo genero t)'J'es:
1.
Alley: narrow public wayS within a b~ock, generally used for vehicular access co the rear of abutting p ropertieS
. Foorway: paved or otherwise improved areas for pedestrian use, located wirhin public sueec rights o f way, wW ch
also conain roadways for vehicuJar traffic.
Ptdmrian walkway: public walk for pedestrian traffic not necessarily within the right of way for a vehicularcr:J'lic
roadway. Included are skywal.ks (pedestrian overpasses), subwalks (pedestrian runnels), walkways giving access 1:0
parks or block interiors and midblock street crossings.
Bilttway: any road, street, path, or way that is specifically designated as being open co bicycle travel, regardless of
whether such facilities are designed for the exclusive use of bicycles or are co be shared with other transponaciO.n
modes. There are rwo general rypes of bikeways.
1.
Typt A~signaud bicyck laM: portion of roadway or spoulder that has been designated for use bJ
bicyclists. It is distinguished &om the portion of the roadway for motor vehicle traffic by a paint Sl!ip e"
curb, or other similar device.
2.
Typt B-bicyck trail separate trail or path from which motor vehicles are prohibited and which is br
the exclusive use of bicyclists or the shared use of bicyclists and pedestrians. Where such a trail or puP
forms a part of a highway, iris separated from the roadways for motor vehicle traffic by an open sp~ee
or barrier.
Commm:Uzl: business or industrial area of a municipality where ordinarily there are many pedestrians duringch~
night hours. This definition applies to densely devdoped business areas outside, as well as within, the centralp.r-.ot:
of a munio;ipality. The area contains land use that attracts a relatively heavy volume of nighttime vehicular aci/.
or pedestrian traffic on a frequent basis. A commercial area without night pedesttian activity may .l>e classificl ~
a special land use.
Roadway light ing , ll ~
lntmnediizu: those areas of a municipality often characterized by moderate nighttime pedestrian activity such as
in blocks having libraties, community recreation centers, large apartment buildin~, commercial buildin~, or
neighborhood retail stores.
&sidmrial: residential development, or a mixture of residential and small commercial-establishments, characterized by few pedestrians at night. This definition includes areas with single-family homes, condominiums and/or
small apartment buildings.
. Probkm at?as: certain land wes, such as industrial areas, office packs, commercial parks and public parksi may be
located in any of the foregoing area classifications. The classification selected should be consistent with o:pected
nightt.ime pedestrian, vehicular and.Qther rdared activity. Within any given municipality there should be consistent application of the definitions.
3.5 Collisions
When undertaking collision analyses as parr of a roadway lighting srudy, four general data items arc needed for each
collision that occurs during the study period. Refer to Chapter 17 for more information on traffic collision studies.
These data items arc the location, severity. type and time of day (day or night). The location element includes the
street name and whether it occurred at a specific interscaion or at midblock. In the case of &ccWay studies, the six
desirable location categories include:
1. on mainline between interchanges
2. on mainline within interchanges
4.
at racnp
entrance to mainline
5. in racnp proper
6. at ramp inrcrseaion with crossroad
Collision severity classes include fatal, injury (overall or A, Band C type), or property damage only (PDO).ln the case
of fatal or injury collisions, the number of collisions involving a fatality or an injury should be wed, not the number
of persons killed or injured. Even when a cost analysis is being performed, it i.s preferable to assign average unit costs
to collisions by class rather than by numbers of persons involved. Due to their low frcquericy of occurrc~:~ce, fatal and
injury collisions are sometimes combined for analysis PUJPOSCS
The primary collision types that arc weful in lighting studies ace:
l. vehicle/pedestrian and/or bicycle
2. vehicle/vehicle
3. vehicle/fixed object
\In spccialiud studies, other distinctions may be useful, such as sepuating vehicle collisions into two categories: !hose
~nvolving oilier moving vehicles and those involving parked vehicles. Fixed-obj~t collisions may. be categorized by
!~e of obstacle struck such as light pole, sign post, guardrail, bridge pier and so on. Pedestrian collisions may also be
of particular interest as a special study. The basic tabulation is by day or night, from the light condition indica ted by
rhe investigating officer. This item is generally coded on collision summary reports and shown on collision diagrams.
If not, it will be n~essary to determine light conditions directly from the fully detailed collision reports.
Up to 5 percent of collisions can be expected to be noted on the collision report as occwring at dusk or dawn. These
collisions should not be classified arbitrarily. One option is to omit tabulation of this group. The disadvantages
include reduction of sample size and incompleteness in total collision data ror comparison with other srudies, crosschecking, or usc in nonlighting-related studies. If rhis group is included, there are two methods of placing the dusk
or dawn collisions into a day or night category. The simpler merhod is to assume all dawn collisions occurred during
darkness and all dusk collisions during daylight. The preferred method of grouping is to assume the time of occurrence is reasonably accurate as shown on the collision repon. The local sunrise or sunset time is determined from
readily available tables (see, for example, Exhibit 19-5) and corrected for daylight saving time if it is in effect on the
collision date (AsUQnomical Applications Depanment of the U .$. Naval Observatory, 2008). Ifthe collision occurrod
w,ithin the period &om 15 minutes (min.) after sunset to 15 min. before sunrise, it is classed as a night occurren~e,
and all collisions during the remaining period are daylight events.
Collision records obtained from computer files may pose some problems. The printouts may code the time of collision occurrence only to the closest hour and some tabulations do not identify exact location, object scruck, or other
pi~ of information that may be viral ro an accwate analysis of collisions as related to lighting. Pasr experience has
shown errors of 20-60 percent when comparing automated data versus manual tabulation from original collision
repon forms (Box, 1971). If there are serious doubts regarding the validity of the computer printout data, the original
collision repon forms should ~.used. These are commonly available as electronic files and are accessible to the analyst
for most studies. Suggestions for lo<:atin~and possibly correcting erroneous collision data are given in Chapte r 17.
The effecr of lighting on roadway collisions is tenuow and easily masked by other variables. Controls in v:~riability
can be achieved by careful selection of the analysis method, provided an adequate collision sample size is available.
In an e:xisting<.nndltion study, it is generally desirable to ~ 3 to 5 years of daca. In a before-and-after srudy, 2 to 3
years' data are typically wed for each srudy period. Research should be conducted to determine ifchanges other than
lighting levels have occurred in the area which may inHuence nighttime coUisions. To account for any time effects, ic
is advisable to include a set of unimproved comparison sites, having essentially similar characteristics as the improved
sites, in the after period. More 9-etails regarding the design of before-and-after experimentS are given in Appendix A.
The decision criteria discussed in Exhibit 3 of Appendix A (Exhibit A-3) may be used to determine whether a change
in the number of coUisions recorded during a before-and-after srudy is significant. These criteria assume collisions
are Poisson-distributed and also recogoiu the before count is a random variable. To illustrate the use of Exhibit A-3,
consider an wban roadway that received a lighting improvement. A total of 12 nighttime collisions were recorded
422 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
during the 2-year before period prior co che improvement, and three occurred during che 2-year after period following
.the improvement. This represents a decrease of (12 - 3)/12 = 75 percent, which is greater than the figure value of 67
:percent for a 95 percent confidence level. Therefore. it may be concluded with 95 percent cc:rtaincy char che reduction
'in collisions was not due to chance alone. If rhe number of collisioo.s in the before period had been considerably larger,
say 26 occurrences, che same conclusion would be reached if che number of collisions in the after period were rcdueed.
to 13 or fewer, since che 26/13 = 50 percent change is greater than the figure value of 46 percent.
EXAMPLE 19-1
A typical computation of night traffic volume percentage at a single location is made as follows:
. Sunset: 19:05 hows + 15 min. = 19:20 hows
Sunrise: 05:48 hours - 15 min. = 05:33 hours
Night traffic (19:20 to 05:33) = 4,200 vehicles
24-hour count = 17,000 vehicles
Night percentage .. 4,200 vehicles I 17,000 vehicles =25 per~t
The night percentages are computed separately for each seasonal count at the location. These percentages are theO.
averaged to provide the typical annual percentage of night traffic by type of route. This seep is repeated for each route:~
Night traffic percentage by type of route may be applied individually when conducting lighting studies. However, the:
type of abutting land use typically changes from point to point along major traffic routes. Portions of a given rouce:
may carry predominandy industrial-cype craflic; ocher portions m~y carry shopper craffic; and scill others, socialrecreational traffic. To simplify the collision studies, a cicywide major route night volume percentage should be used.A separate residential night volume percentage may easily be computed by averaging four seasonal counts for one:
typical screet in each of several different neighborhoods.
1. Single-family-low cost
2. Single-family-medium cost
3. Singl~-farni1y-high cost
4. Multiple-family- low rent
5. Multiple-family-medium rent
Roadway lighting 423
Past experience indicates a consistent range of 23-27 percent of total vehicle miles are traveled at night and the use
of an average value of2S percent for the night portion of the 24-hourvolume appears warranted (Box, 197l).Ifit is
desired to determine the actual percentage of night traffic ar a given location, hourly tabulations of volume are needed
for all36S days of the year and should include the volumes in both directions of travel. These hourly data are generally
available only from a limited number of automatic recording stations along freeways or other high-volume facilities.
%N= lOOAlf
AN+AD
Equation 19-1
where AN is the number of night collisions and AD is the number of day collisions.
A condition ofN that is equal to the night traffic volume percentage would typically represent a collision rate at night
equal to the day rate. A night collision percentage in excess of the night traffic percentage may be evidence of inadequate illumination.
For example, a major study of lighted urban freeways found an average of 32 percent of collisions occurring at night,
while the unlighted freeway average was 44 percent (Box, 1971). Similar findings luve been reached in a before-andafter study of major routes (Box, 1989). Based on such research, a rule of thumb is sometimes applied, with night
proportions up to 30 or 35 percent representing reasonable safc:ty, while greater percentages may indicate a need for
lighting improvemenL
R AN(l-P)
Aoi'
Equation 19-2
iwhere:
iJ?
AN
AD
R=3AN
AD
Equation 19-3
E=RAtf'
N .. 1- P
Equation 19-4
EXAMPLE 19-2
Exhibit 19-6 shows how cwo alternative lighting projects could be compared. Project A is evidendy more effective,
although a simple ranking using night percentage would rank project B as first.
Based on assumed local cxpcric=c for routeS provided with recommended illwnination lcvcl.
b From assumed lighting design calculations of amortized capical cO-tt, plus annual maintenance and eoccgychacges, minus
existing lighting cosu (if any).
The ENvalue is subuacted from the existing number of night coll isions, and the ;,nnual cost per reduced collision is
calculated as in the preceding example. A separate calculation is made for each project, and tbe rankings then follow
in simple order, starring with the lowest annual cost per reduced collision.
Since pedestrian collisions arc especially susceptible to reduction by improved lighting, rhis class of collision can be
singled out for separate analysis. Priorities can be established along major routes by directly calculating the number of
collisions per mile and total number at each intersection. Sound engineering judgment must be used to prevent one
or rwo unusual collisions from creating unrealistic programming. Priority schedules for residential lighting may be
established by comparing crime percentages (night percentage of total daily) for the various areas or by using the total
number of night crimes. The nacwe of the crimes may also be a consideration.
After installation of lighting improvements, an evaluation can be conducted to determine rhe actual effectiveness of
the improvements. The actual effectiveness of the lighting improvements can increase the reliability of funu:e cost
COfllpariso~ and analysis, as shown in Exhibit 19-6.
5.0 REFERENCES
5.1 Literature References
American Association of Srate Highway and Transportation Officials. Roadway Lighting Drsign GuitU. Washington, DC:
AASHTO, 2005.
Asaonomical Applications Dep=mem of che U.S. Naval Observatory. "Sun or Moon Rise/Set Table for One Year." February,27,
2008. Washington, DC: U.S. Naval Observatory. www.usno.navy.mil/USNO/asaonomical-applications/data-services/rs-oneyearus.
Retrieved December 4, 2008.
Box, P. C. "Rdationship Between Illumination and Freeway Accidents: llluminaring Enginuring, May-June, 1971.
Box, P. C. "Comparison of Accidents and lUumination." HighwiZJ &uarth Rmrd 416 (1972~): I 0.
Box, P. C. "Freeway AccidentS and Illumination." HighwiZJ &uarch &cord416 (1972b).
Box, P. C. "Major Road Accident Reduction by lUumination." Transportation Rnta.rch Rtcord:journAl oftht Trttnsportalion
ReJtarch Board 1247 (1989).
lES Design Practice Committee. "Available Lighting' Computer Programs: A Compendium and a Survey. LD+A. (March 1981): 35.
Illuminating Engineering Society of North America. Amm(ll!l National Standard Pracriu for Roadway Lighting. ANSI/IES RP8-1983. New York: Illuminating Engineering Society of North America, 1983.
Uluminating Engineering Sociery of North America. "Section 14, Roadway Lighting." lES Lighting Handbook, 1987 Appli~ation
VqJu~m. New York lllwninacing Engineering Society of North Amc:rica, 1987.
illuminating Engineering Society of North America. Tht /ESNA Lighting Handbook: Rtftrma & Applicarum. New York:
Illuminati.ng Engineering Society of Nonh America. 2000.
Instituce ofTransponatio_n Engineers. Traffic Enginm-ing Handbook. 6th ed. Washington, DC: ITE, 2009.
National Lighting Product Information Program. LightingAnswm. Troy, NY: National Lighting Product Informacion Program,
Lighting Research Center, Rensselaer Polyrecbn.ic Lnsititute, 2007.
Sherm;,n, L W., eta!. Prwmting Crimt: Wha: WOrkr. What Doem'~ Whals Promising. A Report to the United Stam Congrw.
Washington, DC: U.S. Deparancnt of Jwtice, Office ofJwtice Programs, National Institute of Justice R.cscarch, 1998.
1ie~, J.. M.
U.S. Department ofTransponarion. RoadwiZJ Lighting H4ndboole. lmplemenation Package 78-15. Washington, DC: Federal
Highway Admlnistttci.on, 1978.
Walk.:r, F. W. and S. E. Robera. Evaluation of Lighting On Accident Frequency at Highway Intersections." Traruportation
&starch &cord: journal ofthe Trrrn1portation Rntarrh Board 562 (1976).
426 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
Chapter 20
EJiudby:
Rohert S. Foyk, P.E.
1.0 INTRODUCTION
2.0
430
TYPES OF STUDIES
430
431
2.2 Inventories
432
437
438
438
438
446
5.0
429
448
448
448
450
REFERENCES
450
1.0 INTRODUCTION
ransportation planning is~ complex process that involves the evaluation and selection of highway or transit hcilities to serve present and future travel demand. The decision-making process involves state DOTs, mettopolitan
plwrung organizations (MPOs), CUJ'al planning organizations, larger municipalities and sometimes regional effota as wdl (Mc:yer and Miller, 200 1). Studies ue conducted to gather information for transportation madding dfons
or to evaluate the potential impactS of specific programs and projects. In almost all cases, softwue programs ue wed c:xtensivdy in the model,ing and evaluation process. Tcavd demand forecasting models, in particular, require extenSive dam
for creating mulrimod;U tranSportation netwooo that ue interconnected with land use and socioeconomic ~ta foruscrs.
~ta needed for these models and studies usually include inform~tion about population and
rr~vd characteristics and tranSportation facilities. The ~ta ue often organized and analyt.ed to:
The
economic utiviry,
n .,.t"i\
4?Q
The data needed for a planning study vary according co the scope and nature of the project, past planning activities,
cost of data collection and so on. Generally, rhe transportation planning process requires that data from a wide variety
of sources be complied, organized and analrzed. Data collection efforts can include most of the major operational
surveys described in this book, including volume, speed, travd time, transit, goods movement studies and invemories
ofland use, traffic control, geometries and other characteristics. Refer to those chapters for derailed discussions of the
design and implementation of these studies. ln addition, Appendix B describes the process of designing, testing and
implementing surveys. Derailed explanations of transportation modeling processes are beyond the scope of this book;
refer ro the lTE Tranrportn.tion Planning Handbook for further information on this topic (ITE, 2009).
management planning and corridor srudies. ProjectS commonly categorized as medium-term extend beyond the time
span and narrow focus of short-term plans but are more special.ized and immediate than the long-term planning efforrs described below. TSM d ements identify traffic engineering, public rransportation, rcgularory, pricing. m~e
ment and operational-type improvements to the transportation system.
430 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
Long-range plans project transportacion needs of an area and identifY the projectS to be constructed o ver a 20-year
(or 25-year} period to meet these needs. Long-range transportation plans (LRTP) rend to be capital imensivc with
relatively long implementation cimes. Although the problem-solving process is similar for the long- and shon-cerm
planning processes, the LRTP planning process requires more sophisticated methods for forecasting future transportation demand than are required for shore-term plaMing projects.
2.2 Inventories
Once the scope and scale of the planning effort have been determined and the boundaries of the srudy area established, the usual next nep is to assemble the data required. The information most commonly needed for planning
purposes includes information on land use and popularion, the chanaetisrics of the transport:trion system and information on travel demand. Inventories, surveys and studies are made to determine traffic volumes, land uses, origins
and destinations of aavdetS, population, employment and economic activity. Collecting completely new data for a
study may not be necessary. Before 6dd collection ofdata begins, existing sources of data should be identi1ied and the
data aamined to see if they are useful for the study in question. The information gathered is summarized for each
traffic zone and for the aisting highway and tnnsit systetn.
2.2.11Arul iJ,e mul. PopulAtion Data
Land use inventories' provide important data on land characteristics and activities including current land use and
vacant land. ln general, land use data are often colkcted to provide:
n~TIIAM<;~:>f'IFITAT1nM
EDITION
\A wide range of information on land use is useful in the tr.10Sportation planning process. Thi.s includes:
I
Several types of land use surveys can be conducted to provide new data, or to update and supplement available information. The decision a.s to the type of survey must be based on its purpose(s) and the resources available. Land
use data may be recorded in the fidd using several different techniques. One of the simplest methods is to recor~
the d:~ta di.reetly on maps or aerial photognphs. This m4p-r<rtl' m>C ofland use inventory may be appropriate in
small towns oi: cities where the only da.ca required are simple use cla.s:&ilications. Typical black line prints (or plots) a.t a.
scale of 1 inch (in.) =400 feet (ft.) or larger may be used. Complementary maps and land data are available dectronically and are used co check cxi.sti.ng data and maps in helping to identify gaps or errors for update effons. A second
technique, fold li.rting, is the most common method for transportation planning. Traditionally, this method has been
carried out using a fidd form similar to that shown in Exhibit 20..2.
0-o,.~....:don
1I
Oogr..,hlolotlon
X
I I
"'-;r'"~ w-
";:::'
ITT
TTl
chi\
I I
,.,....,.
Na-=;4:!' ....
I ll
,.,..UI
rn
~~
_y_____
-~~-
,.loor ':"::!,
PlVi'...;:::=:,::.,u~~-
II IIIIIIIIIIIII
....._
~-"'"
,._,., ...!:_i=::!.'='=~::..
"'~ctACtvt141
......... ~=.:.~-
._.._.t\'1"
"-oall.-='~~:.~.._
::::-
F~r
:,.liM,
A~
flro"
Wl4hh
HMOf'~GIAO N-::.~-tlt_.
"::" .
~ IT r I r I OJ
Wl4ft'l
~ ::: ~-;.;:.:.
,.~
lcrth
H~Ot'b~ltlg Non::,.~.WeAll~-
-~~
,._w-e ...,.
..y.....w.,J....L,~-1--+..LJ..,..w_..L+.~I I I I I
...
n._, ;::=,
OJ OJ 0
TTl [0
Blodt.
G . .DO'..,hlc:''"-tlol'l
....!:!::~no
I I I I I
8 14do
Oogr...,oto..-t!OI'I
~:~no
I I I I I
_..
f I
iJ5
0::::: ~~-:. I I I I I I I I I I I I I I I
A sepame form is usually reserved for each block, with a line for each pucd or land we and dwelling unk Observations may be made &om the vehicle, or on foot if parking or traffic congestion presents a problem. Data from the
forms arc input directly imo a computer database. Surveys of existing land use can be classified as "inspection" or
"interview surveys, depending on whether the dwellings and other places must be entered. Inspection surveys (often
called "windshield surveys") are accomplished without entering the building. Combined inspcction~intcrvicw studies
arc needed when exterior inspection docs not yield enough informacion. Windshield surveys arc frequendy sufficient
for transportation and traffic needs.
The field listing technique may be applied easily wing a laptop computer and either database or spreadsheet ~oftwuc
(depending upon the size and complexity of the :~.pplicacion). Data collected in the field can, if necessary, be ported
to another computer system in the office. [f the land we informacion is co be incegl7.ted with a number of other datab= and will be retained and rewed, :1. geographical information system (GIS) may be appropriate. GISs integrate the
map {or geographically oriented system) with the database, providing a very powerful tool for cracking and analyzing
spatially oriented information.
Land use dat:~. arc traditionally presented on a map showing the land we by general ategory of we (residential,
commercial, industrial, institutional, parking and recreation, transportation, utiliries, agriculture and water). Each
ategory is given a different color tO provide visual differentiation. Statistical summuics may be prepared to show
the cotalland area devoted to each category of use, and may be broken down into subareas, or traffic zones. Land usc
data collected for one study are valuable for later studies. Inconsistencies in categories of land we activities and dat:~.
formats may make it difficult to compare results with subsequent studies. To minimize these problems in the United
Scates, a standard system for identifying wd coding land we activities was devdoped (Urban Renewal Administration
and Bure:~.u of Public Roads, 1965). The coding system provides four levels of detail on land use activity. Each Levd is
subdivided into one-, two-, three- and four-digit categories.
In the hierarchy of highway facilities shown in the c:xarople in Exhibit 20-4, freeways, expressways and major arterials
constitute the major (principal arteria!) highway system, while collector and local streeu compose the local street system.
The number ofjurisdictions invoJ..:ed in a study area varies depending on the siz.e and complexity ofthe area under srudY
Functional and jurisdictional classificatio~ are done in conjunction with the geometric inventory of che streets.
.,.
-~-.:~~
~~
'""~~:--
STREET CLASSIFICATION
AND . IIHERSfCTION
TRAFFIC CONTROL DEVICES
~'
t..ICIND
eto cu.aac,
~-
-. -
......."".
nueu*'
....._,..c
...-
._ .....
-- ~---- ~ il
Facility characteristics m3.y be obWr!cd from existing m.ffic conuol device (TCD) wd geometries inventories. Speed
wd tnvel time ma.y also be needed for selected ro3.dw3.y wd tr.lllSit links. These d3.ta ue collected 3.5 described in
Chapters 5 and 9. Existing opacity and kvds ofservice can be determined using the methods described in the Highway Capacity. Manwzl (HCM) (TRB, 2000). The d3.ta required for signaliz.ed intcncctions include basic geometries,
signal timing informlu.ion, existing volumes, 3.re3. type and driver populacion (commuter or oilier).
2. External trips, funher classified as imernalexternal or external-imernal trips, have either their origin or
destination outside rhe study area.
3. lmernal-incernal rrips have both their origin and destination within the study area.
roadside interviews
postcard/mail-back surv~
rag-on-vehicle method
lights-on study
The study area may be a central business district (CBD) or other major activity center. Dara collected usually pertain
to auto and truck travel, although there is an increasing need for pedestrian and bicycle trips as well. Separate surveys
of rail, bus, raxi and air travel are made as desired to obtain additional travel information. Practitioners can have
success with conducting surv~ over the Internet. In some contextS, an electronic survey actually may have a higher
response rate than whar could be expected through tnditional tecliniques.
TYPICAL
METROPOLITAN
i
I
1NTAV.~WI!ft6
&
Reo f..IANTE.RfitS
f'"LARit
RCO Ft...AG
e
..
i-
Exhibit 20-5 illustrates a typical layout of a roadside interview station for a moderate-volume, two-lane, two-way road.
The location should be fairly level and have a sight distance of over 800ft. (244 m). Where only part of the driv~ can
be interviewed, one or more bypass lanes are needed co avoid congestion. A paved shoulder of a rural route or a section
of curb lane (cleared of parked CaJ's).may ~e used for the interview, leaving the regular traffic lanes for use as a brpass.
M'ost stations are operated for 16 hours a day (6:00a.m to 10:00 p.m.), Some stations are operated 24 hours on major
routes. Additionally, 24-hour volume counts must be taken concurrently with roadside interviews. Data from the z4hour count are used to expand survey data and to account for the sample of vehicles interviewed. Since freeways :iJ-Ce
normally too congested to permit stoppages for interviews, other techniques must be used to obtain travel inforro.a tion. Several alternatives are discussed bdow.
Stopping drivers usually requires the assistance of a uniformed police officer for traffic conuol, slows traffic considerably and may antagoniz..e the public unless skillfully handled. Generally, a large ponable sign explaining the projeCt
prepares the motorist for the dday and often enlists cooperation in answering the questions more readily. (Check wi -ch
appropriate authorities on signs necessary to meet traffic control laws.) It can be useful to involve the news media in
explaining the need for the study. Every effort should be made to avoid congestion, not only for safecy and for mai.Ptaining good public rdations, but also because congestion may cause local drive..S to detour around the interview sr::ation and thus distort the traffic flow patterns. Not every vehicle can'be stopped on a high-volume route; usually, ofl.lY
a sample of drivers is interviewed. The required sample size iS determined using the methods described in Appendix :13.
Samples usually consist of20-50 percent of the traffic. For a 50 percent sample, it is satisfactocy to interviewdri~rs
of three vehicles and pass the next three vehicles.
A typical crew at a location carrying 3,000 to 5,000 vehicles per day in both directions includes a survey chief, [\IV'O
recorders, six interviewers and one or two uniformed police officers (and other individuals as required by local regui:;ations). The number Qf inrerviewers should equal the number of vehicles stopped in each group. Recorders count a.r:t-d
record all traffic (by type and direction). A police officer may be used to hand each driver a card explaining the purpo~e
of the survey as each car is stopped. Exhibit 20-6 shows a form for recording 0-D informacion. Column 5 of tb-JS
form may be used to record the purpose of the trip. The form should be modified to meet individual study needs. TPe
sample data are then expanded to represent the total traffic volume and to obtain total estimated origins and desrin~
tions. This method has several advantages and disadvantages. The most complete and accurate informacion is usual.l.Y
obtained when personal contact is made between respondent and interviewer. The response rate is greater (rdative co
the voluntary return technique), thereby minimizing the survey bias, and samples can easily be chosen from a craff.i.c
stream to satisfy planned statistical standards. The roadside interview technique is more expensive than several other
techniques b~use more personnel are required. On high-volume facilities, there may be some traffic delays durir.Jt..g
the survey, especially during peak travel periods. This technique is often dangerous, especially on higl)-volwne Facil..i.ties, because survey personnel must operate on the highway and interfere with the regular flow of traffic.
Transportation Planning Data 43 ~
8laeiOn ...,..,...
Time: e.o~n:
and;
""""---:.
OOOiiiiOiJOO
---
AOUiiOMd
l5iiiOi"
~--------------------------
A vmation of the roadside interview is often used on comprehensive downtown parking studies to interview drivers
who have just pa.rlced their cars. An interviewer is responsible for interviewing a sample of the parkc;n in a certain
&cilicy. Origin, destination, loation parla:d, time of day, purpose of trip and parking duration information are most
often obtained. Details on this procedure are given in Ch2pter 16.
er caa. e--c., et
c: ....._.... '
ox ...
J11.S..a t.
NO SfAA4111' tUllllt6D
aao.
J"lM-
~t..r'
w..,.
.... ,...,. a
T
a-...
;: ::::... 8
Clt.y
... ..-r
,...........
z...ile
:.
Dr.
co. ................~
:tor\. . . . , hp . . _.l'
0Al'M't.
-:r
C:b.i.cac &a.l'..-.:r
b..
8Cl
.,,.., ~ ~.11 r .. r r
.,....
,...... o r r ....
~T
a:...-.;1 .......
-rak
t ... r ""'" ~ or
The postcard questionnaire technique is relatively inexpensive. Traffic delay is less than for direct interview, and untrained personnel on be used for handing out cards. A major disadvantage is possible sample bw due co better cooperation by some drivers. Care is required in location of disuibution points to intercept a representative cross-section of
uips. Through uucks and passenger vehicles will not provide a high percentage rerum. It may be difficult co include
all important vehicle movements, especially in largt cities, and, like the interview method, it requires stopping traffic.
Observation points must be located carefully. If only informacion on through traffic is being collected, at least rwo stations are needed: one representing the origin (for one direction) and the other the destination. To uace vehicle movementS through an area, every reasonably heavy bypass point must be surveyed. In addition, intermediate Stations may
be used to establish routes or to give more: detailed information on origins and destinations. The reduction of data
requires considetable labor co match the license plate numbers listed on the field sheets of each 0 -D station against
nearby 0 -D stations to uace the route: origin and destination of each vehicle. Simple computer programs and spreadsheets on be used in this process. Often, not more chan 60 percent of the license numbers on be traced through
this study. Trips may bCgin and end within the study area and may never encounter a station. Errors in tranSCribing
license numbers also reduce the number of matches made. The time between obscrvacions of a vehicle indicates fiairly
accUrately whether stops are made in the business disuict, but knowledge of a time gap docs not make it possible to
ascenain the purpose of the scop.
license plate obsecw.tions must be matched at a minimum of twO locations. If SO percent of all license plates were
recorded at one location and SO percent at another, it could generally be expected that only 2S percent of all license
places were observed at both locations. Sampling rates at various locations will differ widely depending on vantage
poincs, volume _and skill of observers. Video cameras can be used to get a 100 percent sample at all locations, bu.t
manual data cxuaction ftom video can be time-<:ansum.ing.
n .... t ,
AA 1
.~ - - ~-~~~
'
<D
./~ ~'\a>
W,/
\,_~ __ ,
,,_
<3)
..
- .....
~ ~
Destln.otlon
Senti o n
2
3
4
CoiSum T1
Volume Vi
Ori&ln Station
2
so
10
IS
13
88
210
6S
20
21
38
18
17
12
14
97
99
__2~ - - -32S
---
10
Row S unlT_.
VoiVJ
95
106
~0
3 10
200
37S
80
87
IS
42
84
400
-~-
----
-~~
Rocss, Prassas and McShane (2004) suggest a method for expanding license plare counts to estimate total volume
between poincs in this type of study. For this method, the cocal volume at all observation pointS must be recorded for
the period of interest. Sample data :ue presented in Exhibit 20-8.
The variables for count expaiiSion :ue defined as follows:
r.N=
TU(N-1)
(5!!L)
Equation 20-1
11
f; - 7;
\.')
F1 - 1j
Where
TiiH
Tlj(N, IJ
number of trips from Station ito Station j afi:er the (N - I)"' iteration of the data trips
r;
V,
\j
F;
The 0 -D volumes of Exhibit 20-8 :ue ~ded using these factors. Results :ue rounded to the nearest vehicle. The
results of several iterations :ue shown in Exhibit 20-9.
f.~
.t$~~J_fu,'@i-t?J~~~?(P-it;. ,.i.c;i~~~~~ilc~~~"~'l';i~"-~;,"_j6,.l',l~i','i'\if.,
~ "-~
~v, . -~ ,:-~,-:_ ~~-.- ~
' ....~_t; ;"\-t; --,;.:; -~~1! -~-1-. _-:' ~ - . .- _...:..-.,
__ . _
Destination
Station
2
3
4
T.
,..
...
"' . - ! ... _ _
' -..
1
2
3
4
1
50
10
IS
13
88
210
2
8
65
3
20
21
4
17
10
12
38
18
97
325
3.35
15
42
84
400
4.76
14
---
95
106
80
87
368
250
310
200
375
F. _
2.63 .--
2.92--2.50 4.31-
----
1135
Origin Station
2
19
161
3
60
66
27
Ill
44
69
306
345
325
1.06
400
t.l6
125
27
37
44
234
210
251
200
0.90
0.80
.:!.
..
Origin Station
99
v,
200
2.02
F
2.39
(a) Field Dar.a and Factors for Iteration 0
Destination
Station
- " '- -. ,, -: 1
4
63
38
54
190
T
267
292
229
347
1135
250
310
200
375
0.93
1
2
1135
3
4
T
v
F
Origin Station
1
115
26
33
44
218
210
0.96
2
16
150
23
41
230
200
0.87
3
60
70
107
74
311
325
1.05
4
66
42
55
213
376 ;
400
1.06
T
257
. 288
218
3n
1135
----
---
v.
F.
250
0.97
1.08
0.92
1.0!
310
200
375
1135
Origin Station
Ill
2
3
4
T
27
31
43
212
210
0.99
v
F
2
15
146
21
39
221
200
0.90
3
61
75
105
76
317
325
1.03
4
67
45
54
220
386
400
1.04
v.
254
293
211
378
1136
250
310
200
375
0.98
1.60
0.91
0.99
--
1135
2 39 2 63
lJ_ .
~-:_7*~-
Time: Begin:
End:
UCense
Time
Number
TNc:k
or
Sus
OUt-
Ucense
of
Number
~-------------------
Time
Truc:k
or
o.Aof
Bus?
~---------------------------
The method is especially adaptable to locations where ttaffic is too .hea.vy to be stopped for driver interview. Licerue
plate studies a.re also very useful in sma.ll-sca.le, limited 0-D studies, or where the destination is known (CBDs. shopping centers, airportS) (Rocs.s, Pcassa.s and McShane, 2004). Tracing license plate numbers over ala.rge number of exits
and entrances is difficult. The method has the adV1111tage of a.llowing the observers to obtain daa without depending
on the cooperation of individual drivers, as is required by interview methods. The IW:lihood of a bWed sample because of poor driver cooperation is less with this method than with the methods described ea.rlier. This method does
not work wdl for large a.rea.s because of the enensive labor required, but is pa.nicula.rly adaptable to srudies ofsingle
routes or fa.cilities. Each srudy must be completed in one day, and it must be continuous. This method does not produce any information on the purpose of the trips, nor does it produce information on tranSit trips or vehicle pa.rking.
445
comprehensive metropolitan area 0-D study. Manuals of detailed instructions and guiddines on these s tudies should
be consulted for additional details beyond the scope of this manual. The basic concepts are described below.
3.2.1 Dwelling-Unit Interviews
The study area is first divided into analysis zones that are as homogeneous as possible. Every dwdling uni t in each zone
is identified. Sample sizes are selected for each zone based on housing density, and the dwelling units to be surveyed
are systematically selected within the zones. Each dwelling unit selected is initially contacted by mail to alert its residents to the impending studies and advise them an interviewer will be contacting them in person. Th e mailing also
indicates what information will be req uested and approximately when the interview will be conducted . A uip dia ry
(or travel diary) is included so respondentS can record information themselves. The interviewer can then confirm and
supplemenc the self-recorded data.
Interviewers are hired, t.raincd and assigned to conracr personally the residentS of the selected dwelling units. They are
inscrucced to interview oQly the sdecred unitS (that is, they cannot substitute another dwelling unit if no one is at the
selecred unit or it is vacant at the time of the interview). A minimum of three repeat visits are made to residents who
are not home duriqg th~ initial contact attempt.lmervi~ ace usually conducred during the early evening hours (6:00
to 10:00 p.m.), when there is greater likelihood of people being at home.
The information collecred includes social and economic data on all of the household residents as well as da1:1 on all
trips made by each resident age 5 or older during the preceding weekday. Dwelling-unit data consist o f the number
of people living at the unit, the number of vehicles owned and/or operated by unjt residents and derails (e.g. gender,
age, etc.) on each person. Travel information is collecred for each person living in the unit for trips made during the
preceding weekday. Information includes. time of trip, origin and destination, mode of travel, purpose of trip and
parking information for each link in each trip by each person. Interviews are not generally conducted on Mond ays
because preceding weekday (F~iday) trips are easily forgotten over the weekend.
Once the interviews have been complet~, the data on travel trips are coded by origin and destination. Each trip c:nd
(origin or destination) is coded ro the zone in which it occurred. The data arc then expanded to the full sample us ing
a zonal faeroe that is calculated by dividing the total number of dwelling unitS in the zone by the actual number of
interviews completed suocessfully. Other facrors are used to convert the data to an average weekday. The survey rcsul ts
are compared to the screenline counts (see Chapter 4) and correction facrors are developed as described lat.er in che
chapter.
3.2.2 Vehicle Owner Mail Q!ustionnaires
This method involves mailing questionnaires or return-addressed, stamped postcards co owners of motor vehicles f.iving within the survey area. This survey may be combined with driver interviews or postcard questionnaire surveys of
traffic entering and leaving the study areas. The elements not obtained include taxi and ttansit patterns. T he ~pie~t
is asked to .record all trips made by motor vehicles on the day after the card is received, normally a weekday. A poceJ'lcial source of bias comes &om the possibility of getting lower response from vehicle owners who don't appreciate t:.he
benefits of such a survey or who don't have good reading and writing skills. ~ a result, the travel patterns of th e:se
populations are under-represented in the survey results. Different-colored cards may be used for private passcngc:r
vehlcles and trucks.
Names and addresses of vehicle owners are obtained from State motor vehicle registration 6.les. Cards are marked :;~.C
cord.ing to the address zone before mailing, ~d tabulations are made of the number of cards for each vehicle classsc:..,.t
co each zone. This enables proper estimation of the total ttips based on the completed cards returned, as compared
with the total vehicles listed for each zone. Truck Beet owners (three or more crudes) may be contacted petsonally, co
ensure better accuracy for this relatively small but important group. Some state highway departments have developed
d~tailed procedural manuals for this type of survey. An inexpensive variation of this method that has been used s~C
cessfully is tO Send the questionnaire tO the motorist along with the annual vehicle license certificate.
3.2.3 Intentiews at WOrkplaces and Spedal Gnum#ors
Travd questionnaires may be distributed co all employees of an acrivity center, such as a large industrial plam or a
group of office buildings, or to people visiting an acrivity center such as an airport, shopping center, transit rennin~
or other special generator. For employees, completed forms are picked up the same day they are disuibutecl Data 0' .61
where employees live, how they get ro work. times of arrival and departure, parking information and trip cosu ca-:J:l
all be obcained for auto drivers, auto passengers and transit and taXi passengers. It is important to encotirige agw ..d
response rate by keeping the questionnaire short. Preparation of questionnaires is reviewed in Appendix B.
Transportation Planning Data 44:::;:7
This method is most dfeaive when thc:ce :ue a few largtt employ= involved; however, it will work under other cin:umstances if the employers fed the survey is important. After obtaining the cooperation of mmagement, individual firms will
often distribute and pick up the questionnaires within their own organizations. This type of study often yidds information
of direct value to the employers and thus hdps .!CCUI'e their cooperation. It is important to =rd the total number offorms
distributed to c:adt fum as wdl as the number of employees in c:adt 6rm so nip dan for each mmpany can be expanded
properly. Similar surveys can be conducted at other special generators, such as aiipom and shopping ccncers. Depending on
the circumsances, parrons can be intervicwal or be given cards to complete and mail back.
3.24 Trtmsil Rouu P~ Quutitmtl4iru
This study (:also called an on-board transit survey) is confined to the ultimate origins and destinations of passengers
using a particular transit route. It is used primarily for planning improvements to the routes or schedules of transit
units. One or rwo survey persons ride each transit vehicle and distribute a questionnaire card (and pencil) to each passenger who boards the vehicle. Passengers may be askd to complete the card and return ic to ndd survey personnd
when leaving the vehicle, or it may be mailed back. This study is best suited to lighdy travdedlines where passengers
are all seated. It is difficult for sanding passengers co .6.ll in the questionnaire cards. The results are expanded to represent 100 percent of passengers, as based on the ratio o( the total riders to the number of cards .6.lled in.
Rerum-addressed postcards may be used when transit loa.ds make it impossible to conduct this study properly on
moving vehicles. The cards are handed co passengers as they board or alight, to be complered and returned u a later
time. The return rate may be low, as with other postcard studies. Considerable care muse be exercised in the analysis
to make certain the results are not biased. There may be a tendency for only the regular commuters or only those
subject to most crowding. or other discomfort, to show interest in returning cards. Transit surveys are discussed in
more derail in Clupter 13.
2. A natural burier such as a rivet ot a railroad track i.! selected as a screenline thac divides the incemal area
into cwo parts. The estimated 100 percent count of traffic crossing the saeenline as derived from the internal and atemal surveys is compared with actual field countS.
3. A cordon line comparison involves deriving traffic ar a cordon line station from the internal survey and
comparing this with similar trips of area residents as derived from the c:xternal survey.
4. Public uansit riding derived from the internal survey can be compared with total riding observed in the
fidd or obtained from transit company records.
5. The accuracy of trip reponing may be estimated by sdecting a wne in an area of high employment for
which employment 6~ are available and comparing the number of persons employed in that wne with
the number of work trips into the woes, as determined from the expanded interview data. In this comparison, work trips by all modes of travd must be included. Proper allowance must be made for those persons
walking and bicycling to work, average absenteeism and the likelihood of employees making midday shopping or business trips to and from the area.
If these comparisons reveal siuble discrepancies, adjustments should be made in the survey daca to correct them.
Exhibit 20-11 shows a comparison of screenline and 0-D data.
I
I
5.0 REFERENCES
American Society of Civil Engineers. American Socirty ofCivil Engineers Urban Planning Guide, ASCE ~uals and Repot!S on
Engineering Practice No. 49. New York: ASCE, 1986.
Kraft, W H., W S. Homburger and J. L. Pline. Tmffic Engineering Handbook, 6th ed. Washington, DC: Institute of
Transportation Engineers, 2009.
Institute ofTransportation Engineers. Transportation Planning Handbook, 3rd ed. Washington, DC: ITE, 2009.
McGrath, W. and C. Guinn. simulated Home ln~rvicw by Tdevi.sion: Origins and Destination Techniques and Evalu;.rion."
Transponati~tn &search &cord:}rnmuzl oftiN TMnsponation &search Board41 (1970): n~3.
Meyer, M.D. and E. J. Miller. Urban Transportation Planning, A Derision-OrimtedApproach. New York: McGraw-Hill, 2001.
Nelson, D . C. Manual ofTransponation Engineering Srudier. Washington, DC: Institute ofTransporration Engineers, 1994.
Roess, R. P.., E. S. Prassas and W. R. McShane. Traffic Engineering. Upper Saddle River, NJ: Pearson Prentice Hall, 2004.
Stopher, P. and A. Meyburg, Urban Transponatiqn M~tfkling and Planning. Lexington, MA: Lexington Books, 1975.
Transportadon Research Board. Conftrmu Procttdinf1 42, lnnowz.tions in Traut/ Dmurrul Mofkling, Volume 2: Papm.
Washington, DC: TRB, 2008.
Transportation Research Board. Highway Capacity ManuaL' Washington, DC: TRB, 2000.
Transportation Research Board Special Report 288: Mnropolitllll Travel Forrcasring, Currmr Practiu and Furun Direction.
Washington, DC: TRB, 2007.
Urban Renewal Administnrion and Bureau of Public Roads. Statu/ani UmJ Ute Coding Mamuzl. Washington, DC: U.S.
Government Printing Office, 1965.
Chapt er21
..
"'
Environmental Impacts of
Transportation Projects
Original By:
Donna C Nelson, Ph.D., P.E.
EJiutl By:
Dani~IJ Fitu!ley, P.E.
1.0
INTRODUCTION
451
2.0
452
453
454
455
456
457
458
459
459
459
459
460
4.0 REFERENCES
461
4. 1 Literature References
461
461
462
1.0 INTRODUCTION
ransportation systems affect the environment in many ways. The environmental impactS of ttanspornuion proj ccts are coming unde close scrutiny because of priorities across the United States and around the world. Alon.g
with safety and. efficiency, environmental responsibility is a primary goal of transportation agencies. The impact
transportation and oilier industries exen on the environment can result in climate change. Transportation accoun t::S
for approxiinarely one-third of greenhouse gas emissions in the United States, with 72 peccent of ttansponation.'s
impactS coming from road use (AASHTO, 2008). Climate change has potential impacts on transportation througl'l
increases in weather/climate extremes and Boeding of coastal transportation infrasrructUie (TRB, 2008). For inst:wcc=: ,
more frequent intense precipitation could lead to increased disruptions to operations at transportation facilities, e~
sion of road base foundations and scouring of bridge supporrs and pipes.
In the United Stares, governmental regulations require that an environmental impact Statement (EIS) be prepared fo .X
any transportation project that will affect me human and physical environment. For fedeml-aid transportation projecn, th.~
EIS muse indu<le srudies of the social, economic and environmental impacts of the proposed system (US DOT. 1975). ~
comprehensive EIS on a large project may include the following:
natural resources, including prime and unique farmlands, wetlands, threatened and endangered species,
natural land forms, groundwater resources and energy requirements;
rdocation of individuals and families, including the number of households displaced, neighborhood
disruption, avUlable housing, number of businesses displaced or affected, documentation of public
pmicipacion and any unusU21 circumstances;
air quality studies, including microscale impacts, mesoscale and mobiie stationary impacts, analysis
methodology and a description of how consistent the project is with the state implementation plan (SIP);
highway noise studies, including the identification of sensitive receptors (such as schools and hospitals);
comparison of future noise levels with Federal Highway Administration (FHWA) criteria and existing noise
levels; noise abatement measures; and noise problems with no reasonable solution;
wetlands and coastal zones' studies to document the analyses, practical measures to minimize lwm and that
there are no pn.ctical alternatives;
social and economic impact studies of impacts on lifestyle, uavd patterns, school districts, churches,
recreation, businesses, minorities and ethnic groups, urban quality and secondary impacts;
water quality issues, such as erosion, sedimentation., use of deicing and weed control products, chemical
spills, groundwater contamination, stream modifications, impoundment, as wdl as impacts on fish and
wildlife;
flood hazard studies including impacts on beneficial floodplain values, incompatible devdopmem, measures
to minimize flood risks and the evalU2tion of alrernatives; and
construction effects including impacts of the construction process on air quality, noise, water, traffic
detours; and the impacr of spoil and borrow (the need to dispose of or find soil and other materials).
A detailed discussion of each of these studies is beyond the soope of this chapter. The <tWO types of impacts most commonly studied by =ponation engineers ace noise and air quality impacts. Many components of EIS scudies fall
beyond the usU2I practice of tn.nSportation and traffic engineers and should be conducted by a specialist in the specific
area. The traffic impacts of new projectS ace d.iscussed in Chapter 22. The net impact of a specific project is typically of
interest in a study, either compared to the baseline conditions or the net difference of a comparable alternative. Many
of the studies needed to supply basic data for the EIS (including volumes, speeds, traffic projections and other basic
studies) are discussed in other chapters in this book. The reader is referred to those chapters for detailed information
on roUection of that data. A Blsuprint for NEPA Docummt Contmt, conducted as pact of NCHRP Research Project
25-25, can aid an analyst with environmental reporting (Crosett et al., 2005). Simulation tools, as discussed in Chapter 11, can also be used for determining environmental impacts of aansponation projecrs; refer to Chapter 11 for
simulation study informacion.
cl=aaeristics of environmental noise that ace of particular concern are (Cohn and McVoy, 1982):
magnitude of the sound;
frequency of the sound;
temporal distribution of~ sound; and
time variance of the sound.
The 1114gnitude of sound is perceived by the human ear as a shon-duration Huccuation in atmospheric pressure. The
levd o~ sound pressure or ~e magnitude of a specific sound (or ambient sounds) is expressed in decibds (dB). The
!decibel scale is logaridunic; therefore, an increase of I dB reflectS a tenfold increase in the sound pressure level. Sound
jpressure levels are generally adjusted to one of three scales: A. B, or C. The A-weighted sound-level scale is used co
1measure the magnitude of traffic noise because it most closely reflects the response of the human ear to transportation
hoise. The A scale (referred to as dBA) de-emphasizes lower-frequency sounds.
The foqumcy if> of a sound is determined by the number of times per second the sound pressure fluctuates between
posirive and negative values on a sinusoidal wave. Noise &om transportation sources are not usually pure tones but are
broadbanded sounds with a wide frequency range. Typically, the range of human hearing is 20 hem to 20 kilohertz.
Tnnporal distribution of noise is important because the time of day, day of week and month or season affect the perception and effect of noise on the receiver. Noise levds are usually higher in the daytime than at night. However, nighttime noises may be perceived as more offensive. Noise levels acceptable at one time of day may not be acceprable at
another time.
TirM varittnc~ of sound refers to the fact that enviionmencal noise is rarely stationary; that is, the magnitude often
varies over very shore periods of time. The passage of an automobile or airplane might cause the noise level in a quiet
neighborhood to rise by 10 to 20 dBA for a short period of time.
Several noise descriptors can be used to account for the above characteristics.
L50 is the sound level exceeded 50 percent of the time (that is, the median sou.nd level).
Ll 0 is the sound level exceeded 10 percent of the time.
L90 is the sound level c:xcecded 90 percent of the time.
n.
'
'~
.. ~~~~'J.lo~li~~~~t~~~~~r~'t;~;;t~t~~!;.)l~tt.~t~~~T,(~'!,Jit(~?.L'"''-'
- '0
-~
-JJ
-~ . :!'... 0: ._..,..,. "'. - ... ~ .r:.~ :....:,,.,.,. -~
...:>~~~ ......, . :c.~ .,.{1:~. ":.--. ~-\t~t~.~-:~t:~t:tl>,-~
Design Level,
Design Level,
Land Usc Category
L (h)
L 10(h)
Description of Land u~ Ottegory
,<;
57 (Exterior)
GO (Exterioc)
67 (Exterior)
70 (Exterior)
72 (Exterior)
75 (Exterior)
None
None
52 (Interior)
55 (Interior)
Traffic noise analysis includes the following steps for each alternative of a highway project:
1. identification of existing activities, developed lands and undeveloped lands for which development is
planned, designed and programmed, which may be affected by traffic noise from the highway;
2. determination of existing noise levels;
3. prediction of traffic noise levels;
4. determination of traffic noise impactS; and
5. examination and evaluation of alternative noise abatement measures for reducing or eliminating the traffic
noise impaets. (FHWA, 2006)
traffic; a second group located toward rhe interior of rhe residential area, where the arterial traffic is still a major factor
'.in establishing the noise bur where orher noises of the community are beginning to make significant conrribucions;
.~nd a third group deep enough imo the community that the only noise measured is from the community itsdf.
The noise level measun:mem analysis should involve the foUowing steps (FHWA, 1996}:
1. Adjust mea.~ured levels for calibration drift.
modeling of five standard vehicle types, including automobiles, medium trucks, heavy trucks, buses and
motorcycles, as well a.1 user-defined vehicles
modeling of both constant-How and interrupted-flow uaffic using a 1994/1995 field-measured database
modeling of the effects of different pavement types, a.1 wdl as the dfccts of graded roadways
sound level computations based on a one-third ocave-band dambase and algorithms
graphically interactive noise barrier design and optimization
attenuation over/through rows of buildings and dense vegetation
multiple d.ifiTaction analysis
parallel barrier analysis
contour analysis, including sound level contours, barrier insertion loss contours and sound-level difference
contours
For manual calculations of noise levels, NCHRP Report 174 (TRB, 1976) includes a comprehensive design guide for
highway noise computations. An c:xam.ple of noise contours from a study is presented in Exhibit 21-3.
20ft
o.,.,........
Source: Kugler, B. A. et al. NCHRP Report 174. Highwtty Noist: A Dmgn GuUU for Prtdictitm 111111 CtmtroL Copyright,
Nacional Academy of Sciences, Washington, DC, 1976. Reproduced with permission of the Truuportacion Research Board.
Ozone (0 )
Particulate maner (PM)
Carbon monoxide (CO)
Nitrogen dioxide (NO)
Sulfur dioxide (SO)
Each region is measured for the five criteria pollutants to determine if the area is classified as attainment o r "nonattainment" (Forkcnbrock, 2004). Maintenance" status is used to reference specific pollutantS when an area moves
from an attainment to nonattainment level. Exhibit 21-4 shows the national air quality standards that must be met
to achieve attainment statuS. Responsible state agencies and metropolitan planning organizations (MPOs) within
nonanainment rcgiops mUSt prepare state implementation plans (SIPs) that ensure no mnsponation projectS or policies will inpease regional emissions or cause a pollutant violation. When mnsportation improvement programs arc
consistent with the corresponding SIP, transportation conformity is achieved. Nonanainment or maintenance status
can result in rules that a transportation project must not cause an increase in a specific pollurani:, or the implementation of more stringent analY'i.s procedures. The impact of transportation projeas on greenhouse gas (GHG) emissions
is also an imponant consideration. In 2003, tranSportation accounted for 27 percent of the GHG emissions in the
United StattS (EPA, 2006). The prim.ary human generated componentS of GHG emissions include carbon dioxide
(C01), chloroBuorocarbons (CFCs), hy.d!oBuorocarbons (HFCs), perfluorocarbons (PFCs) and sulfur hc:xafluoride
(SFJ. GHG emissions are typically reponed in terms of C01 equivalent unics.
Parciculate (PM,.)"'
Particulate (PMu)..
So=: Forkcnbrock, David J. and Jason Sheeley. NqiRP, Report 532, Ejfoaive Mnhods for Environmmtll/justiu Assessmmt.
Copyright, National .Academy of Sciences, Washington, DC, 2004. Reproduced with permission of the Transportation
Research Board.
The microscale CO analysis is the most common air quality assessment. This analysis focuses on ground-levd or "hot. spot" impacts (EPA, 2006). This process includes combining the microscale effects and demographics of affected ~reas.
: Microscale analyses generally employ a dispersion model to predict concentrations at critical recepcor sites. Jncersec. tions are studie<l because vehicles produce gr~cer CO emissions while idling or traveling slowly. Therefore, a common
assumption can be made thar ifCO limits are not exceeded at worse-case intersections, the other criteria poilu rantS will
also nor be exceeded. Regional air quality assessment is more difficult to conduce and requires more extensive data
collection. Both cypes are ar~ of interest with respect m federal and state agency standards. In additio n to the commonly studied pollutants, regional studies examine volatile organic compounds as well. The state air pollution control
agency and ocher reviewing agencies should be contacted before analysis is begun. Mitigation m~ures are required if
violations oflocal standards result &om a planned transportation project. Some project alternatives migh t result ill less
impact chan ocher project options, for instance, various choices ofland use options or public transit ver sus highways.
tion can be wed to estimate the magnitude of the dust problem and to recommend mitigation techniques, including
minimization of the amount ofland cleared, replanting, use of dust suppression methods and others. These cypes of
controls are part of the standard specifications for most transportation projects.
Air qualicy impaets due to traffic congestion occur when access to existing highways must be restricted during construction. The impaCts of these diversions can be assessed in much the same manner as any other highway project;
however, due to their temporacy nature they are usually determined to be insignificant. ImpaCts associated with construction also involve emissions from construction equipment. Emission rates for construction equipment arc published by EPA and impaets can be analyzed using conventional modeling techniques. In most states, constmccion
impaCts are considered temporacy and/or insignificant. Their discussion in air qualicy impact reports is usually a
commitment to undertake all appropriate mitigation measures.
;"'t,:o:A'..~.-..o~~~
tt<,_
Source: Forkcnbrock, David J. and Jason Sheeley. NCH.RP, Report 532, Effiaiw
Mahods for Environmml41f~ Auasmnu. Copyright, National Academy
of Sciences, Washington, DC, 2004. ReprodllCCd with pamission of the
Transportation Research Board.
460 MANIIAf nF TRAN<;Pr'\RTIITinM cw:,.ICCDIIr ,.,., ' " ' ' '
.4.0 REFERENCES
i4.1 Literature References
American Association of State Highway and Transportation Officials. Primu on Transportation and Clitn4te Changt:.
Washington, DC. AASHTO, 2008.
Code of Federal Regulations. Proudures for Abatnnmt ofHighway Traffic Noise and Constructwn Noise. Title 23 : Highways, Part 772. Washington, DC: U.S. Government Printing Office, 2009.
Cohn, L. F. and G. R. McVoy. EnvironmmtalA114lysis ofTransport4tion Systmu. New York: Wiley, 1982.
Crossett, J. et al. Synthais ofDatil Ntttis for EA and EJS Docummtation- A Blueprint for NEPA Docummt Con1mt.
National Coc!perative Highway Racardt Program Projca 25-25, Task 1. Washington, DC: Transportation Research
Board, 2005.
Federal Highway A4rninistration. Highway TrajJic Noise Analyris andAbatnnt:nt Policy and GuidAnce. Washington, DC:
F~deral Highway Adminisuation, 1995.
Federal Highway Administration. Measuremmt ofHighway-&Iaud Noise. Washington, DC: Federal Highway Administration, 1996.
Federal Highway Administration. Traffic Noist: Modt:l Vmion 2.5. Gainesville, FL: University of Florida for the Federal
Highway Administration, 2004.
Federal Highway Administration. Highway Traffic Noist: in tht: United St4tn-Problnn and Raponst:. Washington, DC:
Federal Highway Administration, 2006.
Fork,enbroclc, D.}. and J. Sheeley. National Coc!pcrative Highway Research Program Report 532: Effictiw Mt:thodsfor
Environmmtai justice Assmmmt. Washington, DC: Transportation Research Board, 2004.
Frey, C. H., N. M. Roupluil, and Z. Haibo. speed- and Facility-Specific Emission Estimates for On-Road LightDuty Vehicles on the Basis of Real-World Speed Profiles. Tramport4tion &search &cord: jour1141 ofthe Trllnsportlltion
&search Board 1987 (2006): 128-137.
Soil S~ey Staff. W(b Soil Survey, Washington, DC: Narural Resowces Conservation Service, U.S. Department of
Agriculture. http://websoilsurvey.nrcs.usda.gov/app/HomePage.htm.
Transportation Racardt Board. National Coope.rarive Highway Research Program Report 174: Highway Noise: .A
Dtrign Guide for Prediction and Contrr~L Washington, DC: TRB, 19!6.
Transportation Research Board. Potenti4llmpam ofClimate Change on U.S. Transportlllion. Special Report 290. Washington, DC: TRB. 2008.
U.S. Department of Transportation. Environmmtill A$1emnmt Notebook Snits. Washington, DC: U.S. DOT, 1975.
U.S. Environmental Protection Agency. Grrt:nhouse Gas EmissWns ftom the U.S. Transportiltion
Washington, DC: O~ce ofTransportation and Air Quality. U.S. EPA, 2006.
Sfqr, 19!}()-2003.
Zegras, P., D. Guruswamy and H. Rojas. *Transportation Modeling for Energy and Environment: U.S. Experience
and Relevance to the Developing World." TrrznsportiJtion &sea~h & cord: ]ou1714i ofthe TransportiJtion Research Board
1487 (1995): 41-48.
Chap ter 22
................... ........ .. ............... ....... ................................
464
464
464
465
466
3.0
4.0
463
473
473
477
478
479
3.2 Capacity
479
480
480
481
481
SUMMARY
482
4.1 Presentation
482
5.0 REFERENCES
483
1.0 INTRODUCTION
ypically, traffic~= and impact studies are conducted to assess the transportation impacts of proposed dtvc::J.opments and other land use changes. Studies might also be required for transportation infrastructure cbange5
such a.s a new roadway or a road widening. The proposed development could be a new office building, subdiv.i.sion, &ccory, or shopping center. Proposed changes in land use might include the redevelopment of an existing are: ::a
into an area that includes a mix of uses. Traffic impact studies project future transportation demands, assess the impac;;: 1:
of changes in demand and suggest ways for mitigalling the adverse effects of land use changes in defined geognphi.:.
areas. For these studies, mmportation dnnand is defined a.s the need for movement of people and goods by all fro.. .::S
of transportation, including autos, car pools, transit, ta:ri, trucks and bicycles; and movement of pedesuians in th. &=
vicinity of a proposed development-
The design and implementation of meaningful impact studies are complex processes. The results and recommendations of the studies are heavily dependent on the experience and knowledge of the persons conducting them, as well
as those reviewing them. This chapter provides an oucline of the analyses required with an emphasis on the collection
and organization of materials and data for traffic access and impact analyses. A complete study draws upon data collection and analysis procedures described in a number of chapters in this book as well as several other references. These
include the most recent issue of Transportation Impact Analyse. for Site Dl!llewpmnit (ITE). Trip Gm"ation (ITE) and
the Highway Capacity Manual (HCM) (TRB). Local and/or state guidelines must also be followed when conducting
a traffic access and impact study.
Computer-based methods are used for many of the analysis steps required for traffic access and impact analysis. For
example, computer-based methods are typically used for level ofservice (LOS) and capacity analysis, as well as for most
steps in trip distribution, traffic assignment and modal choice modeling. Available software packages will also estimate
trip generation and parking generation for various land uses.
when the devdopment contains more than a specified number of dwelling units or square foo!'a&e
when financial assessments arc required and the extent of impact must be determined
when the development will require a significant amount of transportation improvements
when a previous c.ransportation impact analysis for a site has been deemed out of dace
at the judgment or discretion of public agency staff
when the devdopmcnc is in a sensitive area
Transportaritm lmptUt Analyses for Sit~ Dnt~~pmmt (ITE, 201 O) recommends that in lieu of other locally establish~
thresholds, a traffic access/impact study should be conducted whenever a proposed devdopment will generate 100 or
more added (new) peak direction trips to or from the site during the adjacent roadway's peak hours or the devdopment's peak hours. The ra'tionale supporting this recommendation is that:
100 vehicles per hour are of a magnitude that C:an change the LOS of an c:xistin~ intersection approach.
Left- or right-tum lanes may be needed co accommodate site traffic satisfactorily without adversdy affecting
through (nonsitc) traffic.
Virtually any m2jor cra.ffic generator (which may include approved or anticipated developments) must be considered
as a potential candidate for traffic impact analysis. Examples include high-density residential areas, offices, retail/commercial hotels, business parks, hospitals/medical offices, schools, industrial facilities and scadiumsfcoliseums.
1A Study Timing
Transportation needs should be a major consideration for new or expanding developments throughout the planning
stages, including site sdcction. Detailed formal studies, however, may only be required ac specific development planning stages. Under normal circumstances there are several stages in the development process where traffic access/
impact studies are potentially appropriate.
Zoning and rezoning applications
Redevdopment
Land subdivision applications
Erivironmental assessment
Site plan approval
Building permit application
Formation of a special-purpose district
Development agreements
Amendmcn~ to comprehensive plans
Pennits for major driveways
Annexations
Signal warrants
All-way scop warrants
Consiruction staging
Separate srudies are not needed at each development stage. However, studies completed very early in the development
process may ne.ed to be updated to include additional detail as the sire plans become specific. In some cases, the planning process will result in a substantial reformulation of che development program and plan, resulting in a need for
re-analysis and re-examination of study findings and conclusions. The initial study should be reviewed at each phase of
the development co ensure consistency with the current development plan or to indicate the need for additional srudy
because of substantial changes in impact over those predicted initially. For scaged developments and projects, me original studies should be checked ar each stage and updated as appropriate.
An 2pproved study methodology should be reached with the approving agency on the scope of study and appropriate
assumptions for the analysis prior to the commencement of the study. Examples of issues that may be discussed with me
reviewing agency include (ITE, 2010):
What components of a full site transportation impact srudy are needed to address issues associated with rhe
site, proposed development and nearby transportation system?
How detailed. a.n analysis is needed for the trip generation forecast? Should standard equations and rates be
used, or is a special study needed? Should modal split be considered? Should pass-by and/or captured traffic
be analyzed? Is an internal!external analysis warranted and if so, how detailed should it be?
How large should the study area be? What is the area of influence of the project?
Aif: traffic counts needed? Which da}'$ and hours should be countedl What growth rate is appropriate for
background traffic?
How should adjacent developments be considered in the study? How should areawide growth estimates
and future traffic assignments should be used? Should existing zoning be used to generate future traffic for
undeveloped parcels?
466 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDmON
peak-period turning movements for site and stcecc, and for surrounding major intersections and
interchanges if pare of the project sco~;
vehicle classification counts;
adjustment faaors co rdatc count dara to design pe.riod;
machine counts to verify peaking characteristics;
I\
Traffic volumes
Land use
Dcmogr:ophics
Transportation S}'ltem
I Crash history (3 years ifa:vaila.ble) adjacent to site and at nearby major intersection.!
if hazardous condition has been identified
Source: TTrliUJortafiqn lmpactAnlliyusforSi~ Ikw/4pmmt: an ITE Rmlmmnuled Practiu. ITE. Table 3-2, page 17, 2010.
The location and siu: of all approved projects in the vicinity of the project site should be shown and can be presented
as in Exhibit 22-4.
~::~~~:~~-:~~~:~~~J~.~-:z.=~~~
ROUTE
~ -
NOTTO.c.A.L...8
.,..., c:,.._,.
eo
fWt
Pt:tl\.EY
....V. LeMA
8'1'1.CJT-~--
maq
~__) + : ~
r---------------~
SMD,
~.....-.-~
----
ttt Q K$F
(APP. P"AO.JECTG!
- - P-~
l ~I"AA..JecT 8
. . -
1B tCSF .. Cl.'1WNG
and
Ls;oetP
17 tCSr
f'RO'CCT
[Af'P:-~01
CAM'OiltJit PRo.I:CT
1IH
(APP;..~e!
7!
.503 IC$F
maq
~1
WAN.FAC1\IItlloiG. KSF
$l\JOY IOTIRSI'CliONI
:)
~~
.II!
I
I
I I~
.
olfmt
cma-
-~~'f., 2 .. KS"
.....
N<71'TO...,......
Motllha.O..-- ..
....,.
loUR.;. \.OMA
-+ ..............
~....
maq
$1UO"f' watrCTQNI
- - ......cllt01Uilu
. , _ . . . . 14 .,aw OIJUCI (IIOlH
Pmi...EY
in Exhibit 22-7, traffic demand peaks b~cween 5:00 and 7:00 p.m. for both directions. Turning
should be conducted during this period.
movem~nt
counts
Fifteen-minute rum.ing mov~mrot counts should be obtained for intersections and driveways during the period. The
peak hour can be identified more closely by selecting the four co~cutive 15-minute (min.) periods that contain
the highest total counts by direction. Existing directiona.l volumes can be presented as shown in Exhibit 22-8. Orher
methods for presenting turning movement volumes include tables, graphic intersection summary diagrams and intersection flow diagrams. Examples of these are given in Chapter 2. These data will be used for the LOS analysis for
these locations.
Residential
Office
_____ .. ...
.,.._
300
ccunM11:t
.i
j
250
200
1!!10
100
!50
--
12 t
rnc:aq
II
10 H
12 1
lJ
<4
II
Ill
..
HOUR aiDING
SO IS
-~/W
--~
.....................
.
-i..
CCUCT CM'IE:
~
-S1UOY
lf~S
~-1
COUNTS, 'IIOOCUS
1t~ GA.~ MMsoAY
31\.$ lQ 15 ,......
I1'1(Jq
Source: Mohlc, Grover & ASsociates.
determining trip generation rates for similar developments in the area for the same time of day and
multiplying the race per unit area on a proraca basis (FHWA, 1985);
using trip generation rates &om a similar area and land use;
obtaining trip generation rates and equations from ITE's publication Trip Gmmztion (TfE, 2008) or ocher
accepted pubLiShed sources;
using the tc:chniques available on computer sofrware packages; and
using information coUccted from household tcavd surveys.
Published nati-onal rates oc local rates acceptable in the cc:sponsible jurisdiction may be used. Many state, ~onal and local
agencies have: established their own aip generation database for sites within their boundaries. ContaCt the appropriate agencies to determine ifsuch a database exists and ifit is apprQpriau for the eutmlt stUdy. Narional sources can be used as starring
points in estimating the amount of traffic that may be generated by a specific building or land use. Whenever possible, these
national rates should be adjusted to rdiea local or forecasted conditions. These national sources should not be used without
the application of sound judgment and a dear understanding of the stati.stical significance of the proposed rates.
Traffic Access and Impact Studies 473
If existing local data samples are limited, collect additional local data to provide a credible sample size on which to base
the trip generation estimate. Local trip generation data should be collected at sites that exhibit similar characteristics
to the development being s!Udied and that are self-contained, with adequate parking not shared by other activities.
Developments with Wlique demand or parking management schemes should be considered carefully for inclusion in
the study. The following guidelines are suggested for collecting data on a similar site (ITE, 2005).
To obtain daily machine counts, select a generator where automatic counts can be made wirhour doublecounting turning vehicles and without counting through traffic. Directional counts should be in 15-min.
periods. Daily counts should be made for a full24-hour period at the minimum, although a 48-hour
period would be preferable. If feasible, a full7-day period should be counted.
If peak traffic hours of the development are unknown, conduce automatic counts during a typical week of
the year to provide data concerning the weekday and weekend peak hours.
For uses that do nor demonstrate substantial weekly or seasonal variations, select average days for the
analysis.
For developments chat exhibit
major seasonal variations, tksign days
(approximating the 30~ highest hour)
should be selected.
Location
Independent variables
Other data
If there is reason to expect travel characteristics for the proposed development will
be unique, manual counts or controlled
interviews are needed. to determine average weekday person trip ends by mode, the
number of trips actually generated by the
site and the number of trips acrcacted to
the site from traffic passing the site on the
adjacent street. Information on sire/development characteristics of the survey generator may be obtained through an interview
with the site owner or manager, telephone
conversations, mail-back questionnaires
and/or measurements as necessary. Obtain
informacion on as many variables as pos
sible to determine which is the most closely correlated to trip generation. The data
used to apply ITE trip generation rates are
listed in Exhibit 22-9.
sruov "'TERstenONs
IIH80UNO
our aOUNO
.....,_
..
..
__
..
--our..,._
..._,_
McJq
Source: Mohlc, Grover & Associates.
Trip distribution models can also be applied to determine appropriate trip distributions. A common trip distribution
modd is a gravity model which biases che number of trips on the distance from the source. Another type of model
that can be employed for large-scale projectS is an areawide travel demand mode,!. Use of cravd demand modds is
most appropriate when substantial changes in land use, transportation f.tcilities, or both are anticipated during the
analysis period. Most modeling effortS are based in part on origin-<lestination (0-D) surveys. The results of such surveys, stratified by trip purpose and translated into a matrix of trip exchanges among transportation analysis zones, can
provide a geographic distribution of uavd demand to and from the project zone. Many computer-based uavd dem:a.nd
modeling packages permit che user to extract the traffic assignment for a given traffic zone.
The surrogate data method employs socioeconomic and demographic data. Common data dements include population and employment. These data, combined with trip length, arc useful to determine trip distribution.
2.1.3MoJAJ Split
Motlal split refers to the distribution of all person trips, generated by development, among the various transportation
modes available: automobile {drive alone or shared ride), transit (subway, regional rail and bus), motorcycle, bicycle
and walking. In subwban areas of the United States and Canada, the automobile is usually the dominant mode of
access; however, other modes should not be ignored. The importance of transit increases in more intensely developed
areas; in some large CBDs, transit is the domina{lt mode during peak periods. Most published trip generation <hta
are based on surveys of vehicle trips. Such surveys are generally conducted in suburban areas where sites with separate
parking lots make it relatively easy to track vehicles entering and leaving a given site. Thus, if a traffic impact study is
being performed in an urban area where che use of nonautomobile modes is significant, adjustments for modal split
will be reqUired.
One method of estimating modal split is by analyzing existing modal splits for similar developments in similarly located
sites with similar transit service levds. Local transit agencies can often provide historical data on transit use in the
project area. Assuming a constant level of transit service into the future, these modal split data can chen be applied to
the prospective trips generated by new development.
Areawide models can be useful for projecting modal split for planned transit services. They can also project the effectS
of transit's modal share on major changes in the overall transportation and land use patterns of a community. Only
limited cia~ are available for estimating the percentage of trips made using modes ocher chan auto and transit. MPOs
occasionally have survey.data on prevailing use of walking and cycling for some trip purposes. A survey should be
\
\
.47 e
lA~MIIAI (')!=
f'TIIr'ltr'r
'"'"~ "'"''~""'"
I ''
'
'
. considered if it appears such modes account for more chan 5 percent of all trips generated. In the absence of informal cion, mathematical models are used co forecast the probability that a trip will include uansit. These models are often
. described in terms of a disutility function that accounts for the generalized costs associared wirh modal choice.
Build-up
Transportation plan
Growth races
Most metropolitan areas maintain forecastS of fucure travel demand. These forecasts are based on computer-assisted
models which reflect all officially anticipated land use and tranSportation network changes. Travel' demand model
forecasu cypically have only ~ few defined horizon years, which might not match up wi th the forecast years of the
proposed development. Therefore, imerpolation techniques must be employed to determine the relevant volumes for
the study. These forecasts can be used in a cumulative analysis, with the followi ng precautions:
1. The projecrion may be out of date and not reflect changes in local land usc or highway plans subsequent
to the uaflic forccasr.
2. Regional traffic forecasts may only be available on a daily basis. If so, there is no straightforward way to
derive the peak-hour volumes required for a detailed uaflic impact analysis.
3. The land use and transportation network components of travel demand models are frequendy too coarse in
scale for the level of detail needed in traffic impact studies.
2.2..3 Trends or Grtnvth RAu Method
Average growth rates can be used to project nonsite traffic. The growth r~te method is one of the simplest methods to
use; however, it often resulu in inaccurate projections, as it assumes the patterns of growth rates in ua.ffic volumes will
continue tluough the study target year, or will change predictably. Moreover, use of such unconstrained growth could
easily result in unrealistic projected future uaflic levels. Growth in through traffic may be estimated by recent growth
trends in traffic volumes in or near the study area, or from traffic volume projections used in che area transportation
plan. If area transportation plan volumes are used, all site volumes associated with che development chat have already
been projected must be subtracted. If the study year is not che same as the transportation plan projection year, Straightline or variable growth rates should be used to interpolate between eurtem volumes and transportation plan forecast
year volumes. The growth rate selected should be justified in che study report. Tlte growth rates applied should be
appropriate for the situations. For example, a 2-3 percent annual growth rate may be applied co the existing traffic in
well-developed urban areas to reflect possible changes in land use. For a newly developing area, a 5-7 percent annual
growth rare may be applied to reflect the pace and type of development raking place.
This method requires that recent development trends and population growth rates will continue ar approximately the
same rare or at a rate char is predicable. If the study horizon is greater than I 0 years, or growth rates are expected to
change, another method should be used. This method is the least accurate because growth is never U!liform tluoughout a community and its transportation necwork. Past traffic growth may not be indicative of future growth.
safery;
circulation patterns;
traffic conuol needs;
transit needs or impaets;
transportation demand management;
neighborhood impacts;
on-site adequacy and off-site parking facilities (if any are to be used for site generated parking);
pedestrian and bicycle movements;
service and delivery vehicle access;
driveway location and operation; and
air quality and no_jse impact.
478 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
The impact analysis and the development of improvement plans are conducted in an iterative process for each rime
horiwn and key location. The intent is to show the relationsltip becween operations and geometry and to assess deli
. ciencies, as well as to identify alternatives for further consideration. Analyses should lx conducted for existing condi"tions, predcvelopment conditions and for conditions with the proposed project to gauge the incremental impactS of
the project and the incremental needs it generates. k. stated in the introduction to this chapter, a comple te description
of all the analyses required for a complete, comprehensive study is beyond the chapter's scope. Import ant topics are
discussed briefly below. Air quality and noise impacts are covered in Chapter 21.
-f..
S"l\JOY
Oflr:RSI:CTlONS
TUAN~Ci WOW:..OfT
\t'Q..lNES , \t:HK:ILS
PER HOUR
GlltOWTH FAC'I'OR
APPitOWO
~CTS
~C1
nv.rnc.
......... a - - .............
maq
TRAFFIC VOLUMES
HOUf'
3.2 Capacity
Several methods, including the HCM (TRB, 2000), intersection capacity utilization and critical lane technique, c~
be wed ro assess the adequacy of key intersections to handle additional demand generated by the new developme$lt
(lTE, 2008). Apply the same techniques for the existing situation and to the projected alternatives. For exampLe,
projected volumes at signalized intersections should be analyzed wing the same capacity analysis techniques as tho.se
used to characterize existing conditions. All unsignalized intersections should be analyzed wing the same unsignali~ d
analysis technique, unless signaliz.ation is warranted and planned. Traffic signal warrants, such as those contained j.n
the M1Z11wzi on Uniform Traffic Control Dmces (MtiTCD) (FHWA. 2009), may lx used to determine whether si~
nalization is desirable, particularly for the peak period. If the analyses indicate mitigation measures needed i.nduc3e
improvementS co the geometries at the srudy intersections, improvements in LOS and intersection capacity utilizarico- n
can be represented as shown in Exhibit 22-14.
'
Traffic Access and Impact Studies 47' 9
OECUitntcs
C\IWUIILAJJW TRATF'IC
---'
GALE NA ST.
AND
FELSPARST.
"'
o.,. loo.Av os -
~LIE_
~[llF.
GALENA ST.
AND
BAOOKHOLLOW C1R.
GALENA ST.
AND
RlJTLE ST.
~
LfCENO
_ _ _ _ , _ _ UIC
---
,M::r.
- -- ~ tllWUIC ('t.).
_.......,.._ _ _. . . . , . ....,. ('8).
maq
No.ccr + ,.,.,.,
_LI:If- PMW~Cr
(I)..
If signalization is proposed for access to a development along existing anerial ro~dways, it is important to assess the
impact of the new traffic demand on operation of the through street especiilly if a new signal will be pan of the sueet's
progression plan. The HCM defines anerial roadway levels of service based on operating speeds and type of faciliry.
The required analysis m:ty consist of :1. microsimul:~.cion :tnalysis to show that a proposed signal will, or will not, disrupt the platoon Bows operating along the :I.CtCrial.
D.rivcway placement is critical to the overill function of the street necwork operations. Driveways of adjacent parcels
should be combined when possible or spaced as far from each other to minim.i.z.c the traffic conBict points. Sjghc
distance is also a key component when conducting driveway design. Study recommendations md conclusions should
provide safe and efficient movement of traffic to and from, within md past, the proposed development, while minimizing the impact co nonsite trips. Sice access objectives arc co serve abutting properties, preserve roadway capacity,
maintain efficient traffic Bow md maintain safety. Among the factors co be considered ue the following:
480 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
site access for service vehicles should be examined based on size and operating characteristics;
adequacy of site driveways and the internal site circulation scheme must be studied; and
the design and location of driveways for the amount and type of traffic that will be using both the adjacent
street and the driveway must be analyzed.
Estimated driveway volumes, by turning movement, can be shown as in Exhibit 22-15 and used in the analys is of access points (driveways) and inrernaJ traffic circulation. ITE (2010) deals with topics related to site access and off-site
improvements.
_B!_
'I
~
...... .............
Ill
-.ttiT--
S'R.OV~S
..
.._
~
Source: Mohle, Grover & Associates.
al., 1989) and Livable Streer.s (Appleyard, 1981) suggest guidelines for measuring such impaccs, and concain strategies
and techniques for controlling the intrusion of nonlocal traffic into neighborhood streets.
4.0SUMMARY
The purpose of a sire traffic access and impact study is to assess the effects a particular development will have on the
surrounding transportation network, to determine what provisions are needed for safe and efficient site access and
traffic Sow and to address related mobility issues. Other transportation objectives could include jurisdiction-specific
goals such as the promotion of transit. The study report documen ts the purpose, procedures, assumptions, findings,
conclusions and recommendations of the study. Common uses for these reporu include:
providing developers or designers with recommendations on site selection, sire transportation planning and
traffic impacts;
assisting public agencies in reviewing the artribuces of proposed developmencs in conjunction with requescs
for annexation, land subdivision, zoning changes, building permits, or other development review;
establishing or negotiating mitigation requirements where offsite impacts require improvemenrs beyond
those otherwise needed; and
as a basis for levying impact fees or assessing developer contributions to roadway facility improvements.
4.1 Presentation
The sample table of contentS below provides a framework for organizing the report in a maightforward and logical
sequence (ITE, 2010). Some studies are easily documented using this oudine; however, additional sections may be
warranted because of speci..Gc issues to be addressed, local study requirements and the results of the study. Inapplicable
sections may also be deleted from the report. Chapter 3 and Appendix D provide useful information on communicar
ing data to the public and presentation techniques, respectively. The documentation for a traffic access and impact
study should include at a minimum:
1. Study purpose and objectives;
2.
5.
Trip generation;
6. Trip distribution;
7. Modal split;
8. Traffic assignment resulting from the development;
9. Projected future traffic volumes;
10. Assessment of the change in roadway operating conditions resulting from the devdopmem traf6c; and
11. Recommendations for site access and cra.nspon:ation improvements needed to maintain traffic ll.ow to,
from, within and past the site at an acceptable and safe LOS.
The report should lead the reader step by seep through the various stages of the process'and to the resulting conclusions
and recommendations. The report may have several different audiences. Sufficient technical detail must be included
to allow the scalf of the reviewing agency to follow the path and methodology of the analysis. The report must also be
understandable to nomechnical decisionmakers a.n d interested citizens. The following suggestions are o ffered by ITE
(2010).
I.
Clearly document assumptions. Assumptions based on published sources should be specifically referenced.
If less-available sources are used, a more derailed explanation may be necessary.
3. Whenever possible, pre5enc data in cables, graphs, maps and diagrams rather than narrative tex t. This
enhances clarity and ease of review.
4. Discuss findings and recommendations with the reviewer prior to submitral of the final repor t .
5. Do not include political views or statements in the report; it should be an objective, technical analysis.
6.
Inadequate reports should be returned co the preparer by the reviewing agency for completion or modification as needed.
5.0 REFERENCES
Appleyard, D. Livabk Struts. Berkeley, CA: University of California Press, 1981.
Deakin, E., E C. BosselmiliUI, D. T. Smith Jr., W. S. Hom burger and B. Beukers. &sidmtia/Sm,r D'sign and Traffic Control.
Englewood Cliffs, NJ: Institute ofTransportation Engineers and Prentice Hall, 1989.
Federal Highw:~.y Administracion. D~eiiJpmmt andApplication ofTrip Gmeration Rates, HHP-22. W:tShington, DC: U.S.
Department ofTransponation, Federal Highway Adminisuacion, 1985.
Federal Highway Administration. Manual on Uniform Tmffic Control Devim. Washington, DC: Federal Highway
Administration, 2009.
h.stituce ofTranspottation Engineers. Gui&lints for Driveway Location and Design. Washington, DC: ITE, 1986.
Institute ofTransportation Engineers. Trip Generation, 8th ed. Washington, DC: ITE, 2008.
Institute ofTransportacion Engineers. Tmnsportation Impact Analymfor Site DeveiiJpmmf: An ITE Proposed Recommentied
Practice. W:tShington, DC: ITE, 2010.
Transportation Research Board. Highway Capacity Manual. W:tShingron, DC: TRB, 2000.
Appendix A
'
Experiment Design
OriginAl By:
joseph E. Ihmmn, Ph.D., P.E.
UptiAuJBy:
&stilm]. Schroetln; Ph.J?.
1.0 INTRODUCTIO.N
485
486
2.1 Definitions
486
2.2 Objectives
486
2.3 Statisticallnference
486
487
..
5.0
487
487
488
489
489
490
492
492
494
FACTORIAl DESIGNS
6.0 REFERENCES
494
' 498
1.0 INTRODUCTION
xpe.riments are comparisons between two or more conditions chat are manipulated by. or arc under the control
of. the experimenter. Experimena are conducted in a synematic and scientific manner so infuena::s can be
drawn about g~cral populations from measuremena taken from samples of the population. For example, one
could consider a seria of twO spot-speed studies to determine the effects of a change in the speed limit on the 85th
percencile speed to be an experiment. However, a spot speed study conducted to dctennine the 85th pcrcencile vehicle
speed on a highway would not be an experiment since the analyst makes no comparison and docs not manipulate the
conditions during the study.
Expcrimena are one of the major means by which transportation engineers gain an understanding of the transportation sptcms they design and operate. Experiments arc also one of the major uses of the data collection methods
described in this !IWlual. However, a poorly designed experiment can provide invalid rcsula (that is, results from
which no inferences ourside the entities acrually measured can be made), or may mili ine.fficient use of data collection resources. This appendix provides a brief overview of the fa.cers of experiment design that are wed _m~t often by
A ............ ..a: .. A
AO~
transportation engineers, so these undesirable outcomes may be avoided. General concepts and terms are presented
first, followed by discussions of several types of experiments.
Experiment design reaches far beyond the relatively simple concepts provided in this appendix. Readers seeking details
on more advanced topics are encouraged to consult textS such as Cochran :u~d Cox (1957), Anderson and Mclean
(1974), Hicks (1982) and Montgomery ( 1984). The advice of professional statisticians is essential for the design of
advanced experimentS with significant costs.
2.2 Objectives
The process of experiment design begins with recognition of a problem. Next, the engineer muse determine whether an
experiment is needed ro solve the problem. This determinacion is often made with a vety rough estimare in mind of the
resources required to conduct an experiment, and the potential benefitS to be gained. Better estimates of the resource
requirements of an experiment wiU be available after the design is complete, which will guide the decision of whether
to proceed. In some cases, modifications of the experimental design may be considered at this time to condense the data
collection effon co the most pertinent data d ements. Oftentimes, there are a number of variables of potential inrerest,
that may be rank-ordered based on priority and need of the particular element for analysis purposes. The central tasks
of experiment design are to define the MOE(s), list the units to be studied and determine a treatment for each unit.
The experiment design is complete when an analysis plan is developed.
. construct confidence intervals based on a low probability of type I errors, because those are usually more critical. Most
'; r::ypc If errors merely preserve the status quo. Sample-size formulas are available for experiment design t hac mini{l'li ze
:borh rype I and rype 11 errors. More discussion is offered in Appendix C, and specific sample size examples were given
'in Chapter ).
Appendix A 48-;;?'
An example of m effective paired comparison design is in testing the a.ccuracy of traffic conflict data collectors after
training. A newly trained collector md a teacher together spend one-hill hour at each of 15 different intersection
approaches, and indcpcndendy estimate the number of confficrs observed. Each intersection approach represents a
block generating a pair of measurements: one for the student and one for the teacher. The hypothesis tested is that
there is no difference between the srudent's and the teacher's estimated numbers of conflicts at approaches. Exhibit
A-1 provides sample data and calculations that show the hypothesis is accepted. The experiment design is effective
because the variation from different intersections is removed. By contrast, it would be possible with random matching
of treatments and units foe the student to obsc.rve more higher-volume intersections than the teacher.
A.vg(d)
StdDev
rn
= 0.133 = 0.08
6.413
.Jl5
For 14 degrees off=dom,I(0.025levd) 2.145. Since 1 computed i.s ~than c(0.0251evcl) ac.ccpt null hypothesis ar 95%
con6dence levd.
5
t I l
,..#"
..,.~
~t.lf"'oo
l"'t"\ITII'''\~1
typ~ of paired comparison. In a before-and-after experiment, measurements are taken at the same location r:wice: on ce
before a change and once afterward. Before-and-after conditions are sometimes also referred to as "pre" and "post"
conditions. The treatments in a before-and-after experiment are "not changed" and "changed," the units are points in
time and space (that is, a location at a particular rime), and the blocks are the locations. Before-and-after experiments
are attractive for scatistical and practical reasons. In some cases, che unh can also be a human subject who participates
in the experiment during boch treatment conditions.
Statistically, a before-and-after tXperiment alto~ a paired comparison to be performed, removing from consideration
variation between locations. Practically, before-and-after experiments can be condu!=ted during improvement programs and require measurements at fewer locations than ocher experiments. Before-and-after experiments are easily
understood by engineers and nontechnical readers and make intuitive sense.
.
... ::...
...,,,, ~-~~.~--
-~,!:~\1~::~ _J'i~---..~u_::# -:i~~?;;~~~~i?.!!:r#f.~~~~~t
:~:~,~~ll''~l>"''?jf'1Pf!~1f),'-;;-~,-,..,~~:&l'f;l
I~~
~~~--~; __ :.. ~-
- ~
Number of Sections
in Group
Fiest Year
12,859
4,457
1,884
791
_,.o..... o. .-. .
Second Year
Change {o/o)
OA04
Increase
0,832
-16.8
1.301
-35.0
L841
-38.6
374
2.361
-41.0
160
3.206
-35.9
95
3.695
-38.4
62
4.968
-29.0
33
4.818
39.8
14
6.930
-23.0
33
~~~
10.390
-22.0
'9venl.l first. and second-yeu average collisions per section is approximately 0.7.
\Average =13.33.
Source: l{auer and Persaud, 1982.
for most intersections may lead the analyse to claim a reduction in collisions across the network, when in fact the
overall mean number of collisions per inte.rsection stayed largely unchanged.
Regression to the mean affectS before-and-after aperimencs whenever experimental units are chosen on the basis of a
high or low MOE. For instance, treating the highesc-coUision locations in a jurisdiction with a program ofintense traffic law enforcement and evaluating the results on collision rates with a before-and-after experiment is a classic candidate
for regression to the mean bias. The high-collision locations would probably have experienced a decline in collisions in
the after period anyway, regardless of the treaanem.
The third drawback co before-and-after designs is of concern when adaptation co the trearmenc by rhe unit.s is nor
instantaneous. In chis situation, experimenters use wnnn-up periods to allow the unit to adapt. For eJCa.mple, expcr~
menters should wait for one to several months after installing a TCD to coll~cr after data on motorist reaction. Etpe n. menrers should also take care not to unwittingly use data from the period when the treatmen t was being constructed.
Experimenters interested in the "novelty" effect of a rrc::aunent do not use warm-up periods.
Designers overcome the fourth drawback, inmbility, by using a sample of sufficient size and by using stariscical af\alysis methods to draw conclusions from the data. The main point co remember regarding instability is a count of before
events is a random variable, rather than a constant, and must therefore be treated with statistical techniques.
Designers can deal with history and maruration, the fili:h and sixth drawbacks to a before-and-after experiment. by
shortening the rime span of the srudy. Designers changing from 3-year before and after periods co !-month periods
greatly reduce the chance of large history or maturation biases. Again, it is helpful to use a surrogate measure that can
be collected quickly. Experimenters also artempt to overcome history and maturation by literature revi ews that identify and quantify biases. Manual adjustments for possible history and maturation biases rely heavily on judgment and
are often open to ~eba~e, however.
.
4.2.1 Overcoming &gression to the Mean
Regression to the mean is the last of the serious drawbacks to before-and-after experiments. Experimente.rs can avoid it
while selecting units for ueaunent or can overcome it during anal~is. During unit selection, experimenters can avoid
regression to the mean by randomly choosing uni.ts for treaunenr from the entire list of units. For instance, in eva}t.lating the effectS of a program of school woe speed enforcement, every school zone in the srudy area must be a candidate
for enforcement and data collection. T he major problem with this approach is the units most in need of treaun.em
(that is, school wnes with higher speeds) are not necessarily chosen for treaunent. The experiment results do noc sO. ow.
whether enforcement in higher-speed w oes is effective. Similarly, a municipality wanting to test a new intersection
safery treatment would have to randpmly select treatment and control sites, which may leave the worst inrersecrion
U!lueated while the experiment is ongoing.
Experimenters can overcome regression to the mean bias during anal~i3 if the experimental data are Poisson-disccibuted. D ata are Poisson-distributed if they meet certain conditions (Steen, 1982).
L
They are suucrured as a number of events that happen during a particular interval of time (or space or
other dimension).
2. The underlying race of event occurrence does not change through the time studied.
3. The occurrence or nonoccurrence of an event in one time segment is unrelated to the probability of eveP t
occurrence in subsequent time segments.
4. More than one occurrence in a very short time is unlikelf:
Traffic collision and trafli.c conHicr are examples of data that are usually Poisson-distributed.
For data that are Poisson-distributed, Hauer and Persaud (1982) provide a useful method to adjust the nwnber of
events experienced in the before period for regression to the mean. The expected number of after events at a locac.i on
that had It events before is estimated by the number of events that occucred in the before period at locations with k +
1 events divided by the number of locations with It events in the before period. For example, consider in Exhibit A-- 2
the 1,884 roadway ~ons with two collisions each in the before period. An estimate of collisions in the after period
adjusted for regression to the mean is (3 x 791)/1,884 c 1.3 collisions per section. If an experimental treatment h~d
been applied to these sections and records had indicated an average of 1.5 collisions per section in the after period.. a
naive anal~t would have concluded the treatment had been effective in reducing collisions when actually the trearme:t:'lt
had probably been harmfuL
The major problems with this adjustment for regression to the ~ean bias are that it only applies co Poisson-distribure::d
data and that it requires a large database. Hauer and Persaud (1982) analyzed a collision data set from 82 sices 3.0- d
reponed estimation using the adjusunent method above was very poor. However, when the data were examined ~
mulacively (~flat is, the anal~ts c:xamined sections with four or more collisions together instead of analyzing scccio~
with four collisions separately from sections with five collisions, six collisions, etc.), the adjustment m~thod produce::' a
more accurate results than did unadjusted before-and-after comparisons.
Appendix A 49>
Hauer and Lovell (1986) have demonstrated another procedure for removing regression co che mean bias during che
an~ysis of before-and-afte.r experiments. The proccdu.re involves rdativdy advanced scarislical techniques chac are beyond the scope of chis appendix but should be less data intensive chan the adj ustment described above.
-__
-
10
... ..
-. . ........,.,__
I
~ ,..... .,.,.
..
._.
...,....,...,
_prillo_lho_llftll.. _
10
,. \
...... _ _ bf!lw
f--
-- ,,\r\'
Solution: Exhibit A-3 shows that for 14 before events an analyst needs a 60 percent reduction to reject the null hypothesis. Since
a 50 percent reduction was experien ced
(chat is, (7/14) x 100 percent) the an~ysc
cannot reject the null hypothesis at the 95
percent levd.
10
d ltnot
..
._.
10I
40I
/'; ' - -
'r-
i\
~~
I' ~
Ill
~ t-..
o . m
10
--
10
* *
l,llea5Ur~
control units during th~ b~forc and after periods like ueatment units, but do not alter control units dwing
t~e experiment.
Exhibit A-4 illustrates the control unit concept. Mean values from a before-and-after with control experiment could
be tested with a statistical Hest. A typical null hypothesis is that the mean difference berwecn the experimental units
b~fore and after treatment is the same as the mean difference betW:een the control units in the before and after periods.
Con..ructl
edjustm
. on a nd
ntperloci
Legend
Control ehe(8)
-- ....
Control atteta
~I
....-
_ .J _
.... -----
l"rojectalte
..,.,,.
MOE -
Mea~uro
A
0
E>cp~ed MOE
--. e"~~:J
Actual
"11m
Ab<
of Etfectlvenesa
Before-and-after with control c:xperim~nts are strong but require more resources than do other e:xperimem types. Like
a simple b~fore-and-after experiment, the before-and-after with control experiment may require a rclativdy long time
for data collection, and the randomly chose.n units may be widely dispersed. However, With control units the amount
of data collection will increase ovc.r a simple before-and-after c:xpcriment. In addition, it is essential before-and-after
with control experiment be set up early, since control units must be chosen randomly from all possible unics.
Some peopl~ believe a before-and-after with control experiment is nqt feasible wh~ the treatment is to be implemented across an entire jurisdiction or agency. However, Council ct al. (1980) suggest an experimenter can take advantage
of the fact that implementation of a treatment is not instantaneous, due ro budgeting and personnd constraints, to
establish control units. Suppose guardrails arc to be improved over a 2-year period on all freeways in a jurisdiction. An.
experimenter may be able to select at random some freeway sections to be improved in the first year and some to be
improved in the second year. Then the sections selected for second-year improvement can serve as control units during
the first year. The key to this plan is random control and treatment unit sdection. If the project engineer does not
permit random selecti;on, no true control" is established and a bias IIUY affect the results.
A ~fore-and-after with control experiment may be politically unpopular and open to liability questions if high-collision or high-priority unics arc untreated. If an. experimenter is faced with this criticism, Council eta!. (1980) suggc.u
that the treatment is still unproven (that is why an experiment is needed), so withholding it is not necessarily harmful.
In addition, the experimenter may promote the idea of an experiment with a promise chat high-priority units not
treated during the experiment will be treated immediatdy after the final set of experiment measwements, regacdfcss
of the experiment outcome.
Locating enough high-qualiry comparison units is a challenging task in many experiments. The number of comparison
units needed is estimated from standard stadstical sample-size formulas. A common method for choosing comparison
units is to choose :t unit for treatment and then use the unit next to it for comparison. For example, in evaluating a
newTCD at minor inrcrsectiow, an experimenter randomlysdeas the intersection of Main Street and First Street for
treatment. He would then use the next minor intersection along Main Street (that is, Main Street and Second Street) as
a comparison unit. He should not use rhis method if operatiow at the sites are dependent, however. If drivers can sec
the TCD at Main and First &om Main and Second, for instance, the larter is probably not an appropri:tre comparison
unit. Comparison units are often much easier to locate when experiments are designed prior to or during the before
period rather than during or after the after period. Reassembling data &om several prior years to locate adequate comparison units is extremely difficult, even in agencies with outstanding recordkecping systems.
',ANOVA resul(S from an experiment with four factors (L, T, P and S), where the model included only main effects and
:two-way interactions. Analysrs use means testS to find rhe particular levels of significant factors that deviate from other
'\~vcls. Exhibit A-6 illustrates the application of a means tesr ro the main effects of the four-factor experime nt.
The notation L''T for example means the interaction between the lefi-turn volume factor (L) and the through volume factor (1').
Mean delay
Number of
Factor
Level
Observations
(seconds per
vehicle)
Appendix A 49~
Complerdy randomized fa.ctorial designs have one or more observations in every cell of the matrix of factors and levels.
Exhibit A-7 illustrates such a design with rwo facton, each at four levels, and two replications. The units, numbered 1
through 32, were assigned ar random to a cell in the matrix. Completdy randomi-zed factorial designs must be at least
pa.rtially replicated (that is, at least a few cdls muse have more than one observation) to allow the significance of the
main effects and all the interactions to be estimated without additional assurnptio?s.
Unit numbers
I__
Levels
of factor
traffic control"
None
Signs only
Markrngs only
None
Two
patrols
per hour
Four
patrols
per hour
Contlnuous
patrols
31
19
2
28
18
32
. 21
22
26
27
I .
3
7
10
12
15
13
30
14
29
24
17
25
23
'1 1
16
20
9
Many types of advanced factorial designs arc available. Bltxlt designs utilize groups of similar units, as described in the
section Paired Comparisons" above, co remove extraneous sources of variation efficiently. Block designs are useful
when.treacing some units one day and other units another day, for example. The possible effect of the day created can
then be scpararcd from the effect of the treatment. Within each block, experimenters assign units to a treatment at
random. Experimenters use nmetltlesigns when one or more factors are embedded within other factors so it is impossible to obtain an interaction berween the factors. For example, nested designs are convenient when an experimenter
identifies several jurisdictions and then sdecrs for treatment several households within each of those jurisdictions.
The c:xperimenter can separate the effect of the jurisdiction from the effect of the household (Anderson and Mclean,
1~~.
'
Other advanced factorial designs called ftacticnaJ factorials allow the experimenter to gain full information on main
effects and partial information on interutions while running only a portion of a replication. Fractional factorials arc
very difficult to design bur can be very efficient, especially when treatments are expensive. Experimenters oftm use
fractional factorials in exploratory work to determine which variables should be examined in depth. Exhibit A-8 shows
the design of an experiment with eight factors, each at two levds, that would have required 256 units for a full replication. The c:xperi.menters designed a one-<[uarter fractional factorial experiment that required only 64 units (labeled run
number in Exhibit A-8) and provided complete informacion on all eight main effects and all interactions between
two factors. LAtin SifU4rt and Grruco-LAtin sqtunT designs are very special..iz.ed fractional factorial designs that have been
applied successfully in transportation engineering experiments. Designers interested in block, nested, fractional, or
other advanced fa.ctori21 designs are encouraged to consult a professional statistician or a text on statistical c:xperiment
design such as by Cochran and Cox (1957), Anderson and Mclean (1974), Hicks (1982), or Montgomery (1984).
'\
Run
Nu:mb#lr
1
2
3
4
5
8
7
8
16
16
17
18
19
20
21
22
23
24
31
32
33
34
..
...
Variable
Combination
BCOEF
ACOEF
BEF
315.
36
37
36
'39
40
41
42
CE
'14
28
29
30
I
AB
ABCO
GH
ABGH
COGH
ABCOGH
DE
ABOE
21
Run
Numbe~
co
10
11
12
13
25
26
Variable
Combination
AEF
BCDEFGH
ACDEfGH
BEFGH
AEFGH
BCF
ACF
BOF
AOF
BCFGH
ACFGH
BOFGH
AOFGH
BCOEH
ACDEH
BEH
AEH
BCOEG
ACOEG
BEG
AEG
BCH
ACH
43
ABCE
OEGH
ABOEGH
CEGH
ABCEGH
FH
ABFH
COFH
ABCOFH
FG
ABFG
COFG
ABCOFG
OEFH
ABOEFH
. CEFH
ABCEFH
OEFG
ABDEFG
. CEFG
ABCEFG
44
46
46
47
46
49
50
51
52
53
54
55
56
57.
!58
59
60 .
BOH
AOH
BCG
81
e2
. ACG
BOG
ADG
63
54
'--rKey
Level
With No
U.tter
A
B
0
E
F
G
H
I
V:attable
Th~ghvolurne
Opposing Volume
Lefttum volume
Alght-turJ'> volume
Mean vehicle speed
Upstream signal distance
~ signal dlatance
.,..._nt
Units
t.stter
VPH
VPH
VPH
VPH
MPH
Miles
Mil-
300
Bypass Jane
Variables A through H e ll at "no letter'.' ' - '
300
20
20
35
0 .5
0.5
No
Csvel
With A
LAtiM
700
700
50
30
30
1.5
1.5
Yes
Sowcc: Bruce, E. Land J. E. Hummer, "Dday Alleviated by Left-Tum Bn>ass Lanes." TransportatitJn &uardl &rord 1299.
Transportation Research Board, National Resea.rch Council, Washington, DC, 1991.
6.0 REFERENCES
Anderson, V L. and R. A. Mcl ean. Design of Experiments, A Realistic Approach. New York: Marcel Dekker, 1974.
Bhattacharyya, G. W. and R. A. Johnson. Statistical Conctpts and Methods. New York; Wiley, 1977.
Bruce, E. L. and J. E. Hummer. "Delay Alleviated by Left-Turn Bypass Lanes." Trauportation Research &cord: journal ofrhe
Transporuztion &search Board 1299 (1991).
Cochran, W. G. and G. M. Cox. Experimental Designs. New York: Wiley, 19)7.
Council, F. M., ec a!. Accidmt Research Manual FHWNRD-80/016. Wasb.ington, DC: U.S. Department ofTcansportation,
Federal Highway Administration, 1980.
Federal Highway Administration. Highway Saftty Evaluation Procedural Guide. Washington, DC: U.S. Deparcmenc of
Transportation, Federal Highway Administration, 1981.
Hauer, E. and J. Lovell. "New Directions for teaming about the Safety Effect of Measures." Transporuztion &search &cord:
Jountal ofthe Transportation &uarch Board 1068 (1986}.
Hauer, E. and B. Persaud. "Common Bias in Befo~d-After Accidents Comparisons and Irs Elimination." Transportation
Rmarrh &cord: jountal ofthe Transportation &search Board 905 ( 1982).
Hicks, C. It Fundammtal Concepts in the Design ofExpmmenr:s. New York: Holt, Rinehart and Winston, 1982.
Hummer, J. E., It E. Montgomery and K C. Sinha. An Evaluation ofLeading venus LaggingLtft Tum Signal Phasing. FHWN
INIJHRP-89/17.1ndianapolis, IN: Federal Highway Admini.mation and Indiana Deparonent ofTcansportation, 1989.
Montgomery, D. C. The Design andAnalytis ofExpmmmts, 2nd ed. New York: Wtley, 1984.
Steen, F. H. Elemmts ofProbability and Mathematical Suztistics. Boston, MA: Duxbury Press, 1992.
Washington, S. P., M.G. Kar!aftis and F. L. Mannering. Statistical and Econometric M~thods for Transportation DaUI Analysis.
Boca Raton, FL: Chapman & Hall, CRC Press LLC, 2003.
Weed, It M. "Revised Decision Criteria for Before-and-After Analyses. Transportation &s~arrh &cord: ]ountal ofthe
Transporuztion &search Board I068 (1986) .
Appendix B
0. 0
..
..
Survey Design
Original by:
joseph E. Hummer, Ph.D., P.E.
Edited by:
2.0
INTRODUCTION
499
500
500
METHODS
500
501
504
505
505
505
506
507
501
PROTECTING RESPONDENTS
508
509
510
8.0 PRETESTS
: 510
. 510
512
11.0 REFERENCES
514
1.0 INTROOUCT,ION
ransportation engineers have used surveys for many years, particularly in tranSportation planning. H owever, man~
engineers assigned to design and conduct surveys have not had training or experience in survey preparation. Thos..::::.
engineers are often swprised to learn even small surveys are very difficult to design and can fail ifdesigned poorlrThis appendix serves as a swting point for beginners and provides a reference for those already experienced in designin5
surveys. There is no substitute for experience in conducting surveys. AD. engineer inaperienced in surveys who is conducting a survey that is c:xpcnsive or important should seek help from an experienced professional.
Surveys ace essentially systematic, structured conversations. Surveys range from written questionnaires where eacJ::::1
respondent answers the same questions with the same set of possible answers to rambling interviews held cogeche.)IIC.
by a central theme. Most survey results ace coded and analyzed using standard statistical techniques (diseussed jr.:::a.
Appendix C).
Appendix B 49~
This appendix will cover methods for conducting surveys, selection of a relevant and significant sample, how to compose appropriate questions to get at the stated objective(s), preparing the interviewers and administering the survey.
Whether new or experienced with conducting surveys, this appendix should serve as a good resource for all who wish
to learn more about possible surveying options.
2.0 METHODS
Survey data are collected from the verbal or wrin:en responses of a sample of people. Common methods of collecting
written responses include a m.ail~ut and mail-back form, a hand~ut and mail-back form and a hand~tit and handback form. Common methods of collecting verbal responses include telephone interviews a,nd personal interviews.
Personal interviews are conducted at the reSpondent's home, at the respondent's workplace, at the survey analyse's office, at some public place, at an event, or almost anywhere. There are also hybrid methods that combine the foregoing
techniques. Each method has advantages and disadvantages.
Exhibit B-1 shows some of the relative advantages of three common survey methods: personal interviews, telephone
interviews and mail~ut and mail-back questionnaires. The major advantage of the personal interview method is the
ability to probe responses in depth and the high probability of obtaining responses from the desired sample. By contraSt, m.ail~ut and mail-back questionnaires offer lower costs and minimize interviewer bias; however, response rates
decline. Telephone interviews usually take less time than the other methods shown in Exhibit B-1 but do not allow the
presentation of visual' stimuli. Also, with the advent of caller ID, telephones are not answered as often as in previous
decades. Tdephone interviews generally have uaits that are between the acremes of the personal interview and the
mail~ut and mail-back questionnaire, making them an attractive choice for many surveys. Interview surveys have
improved in recent
with advances in computers. Interviewers lising handheld computers in the 6eld or personal
computers in the office can immediately code and score responses. Computers also help interviewers follow complex
skips, where some respondents are not asked some questions.
years
Hand~ut and hand-back or hand-out and mail-back questionnaires are also common in transportation srudies. They
have many of the advantages and disadvantages of mail-out and mail-back questionnaires but can be Jess cosdy. Costs
are especially low if workers_ (bus drivers, for example) distribute and/or collect questionnaires while performing their
Aft
l.AAMII A. f t"'\1:::
TDI\~IC:DI"\OTAT11"\M
Ct..lt":lt..ICI:Dit..lt':. C:Titntr:C
")Mn tnrTII"'\t..l
AJrcc.k, R A. and R. B. Scttk. TM SID'W) Rnurrh H~. Homewood, ll.: R.iclwd D. Irwin, Inc., 1985; Sudman,
S. and N. M. Bradburn_ Asking QJusti4N. Sm franci=, CA:. Jo=y-Bass Publications, 1982.
regular duties. Hand-<lut and hand-back or hand-<lut and lll2il-back surveys also have higher response rates than mallout and mail-back surveys.
The choice of a survey method depends on the type of questions and the sample anticipated. Typically, a designer
tentatively chooses a method in the early stages ofsurvey development and then re-evaluates his choice while drafting
questions and selecting a sample.
Appendix B 501
and complete frame is nece:ss:uy for mosc random sampling techniques. Compiling a frame is often very difficult because
directories, tax rolls and ocher common framing materials are often our of dare or error-prone.
3.1.1 Simple !Un1dom Sampling
Random Sampling is the purest form of probability sampling. With chis method, each sampling unit in the population
has an equal chance of being chosen. The analyse chooses units to be sampled by assigning a number co each unit in
the population and then drawing random numbers. Simple random sampling can be accomplished with "f'lacemmt,
when sampling unirs arc d igible co be drawn more than once, or without replacemmt, when units cannot be drawn a
second time. Some formulas for analyring samples with replacement are simpler, but for most surveys there is no difference in the analysis. Therefore, to remove the possibility that a person is asked the same questions twice, sampling
without replacement is standard.
Simple random sampling has a few advantages over ocher sampling methods, bur for most surveys ir is inferior. Simple
random sampling is easy to understand, is commonly used (making ir easier to compare results wich ocher surveys)
and ic allows a designer co select a sample with no population data besides a &arne. However, simple random sampling
is usually inconvenient, since survey p~onnel must track down widely dispersed units. Worse still, simple random
sampling is almost always less precise chan sampling methods that utilize more infor1112tion about the population. With
the same size sample, ocher sampling schemes will usually produce bcuer estimates of means and proportions than will
simple random sampling.
Exhibit B-2 provides formulas for estimating mean values, variances and needed sample sizes for a simple random
sample. The sample-size formula foe continuous (nor proportional) data requires an estimate of the variance of the
mean for the entire population or a prior estimate of the coefficient of variation, which is the variance divided, by the
f!lean. Analysts can obcain chis prior estimate from an educated guess, the results of a prior survey, or a pilot study of
the cucrent population. The sample-size formulas are based on the desire to estimate the mean value of some characteristic within a specified tolerance from the true mean value for the entire population. For example, an analyst may
want to estimate mean household income within $1,000 of the true mean.
This random sampling scheme is best explained duough an e:xample. Instead of choosing individll21 people as the sampling units for a rravel survey of the residents of a town, the analyst chooses neighborhoods as the sampling units. Then
the individual people are known as the sampling subunits, and each subunit in a sampled neighborhood would be asked to
respond to the 5U1'Ve)': This sampling plan is known as cluster urmpling beau.se the sampling unit (a neigbbochood) consists
of a cluster of subunits. Ouster sampling is a very popular way to conduct surveys in rransportarion and many ocher 6dds.
I.cwer cost is the lll2in advantage of duster sampling c;yver other sampling methods. The rravd survey mentioned above
saves monc:y by sending survey ~nne! to a relatively small nwnber of neighbochoods rather than to dwdl.i~ scattered
across town. In addition, duster sampling does not require a &arne of all the subunits in the population. The rravd survey
mentioned above only needed liscs of neighborhoods in the town and of residents of the sampled neighborhoods. The major
502 MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDffiON
~'51-"?'!'
t..~~
'Yif!?"
~~
Quanricy to be .r:.
Mean
. y=
Vaciahle npfi,
Formula
IY;
Individual observation
i: Observation number
n: To cal number of observations
l _l_
n
Variance of Mean
L:(yi- y)
s2
_.:.::i
=,_,_l~-
(n-1)
Sample Siz.e
n=(~J
1mean
n=(~~J
a
n
--
Proportional Data
Proportion
p=Variance of Proportion
var(p) = pq
n-1
Sample Size
n= t pq
d2
or
t2q
n =-2--
---
rp
a) Formulas for sampling with replacement or for sampling without replaoement when less than 5% of the units in the
population sampled
Appendix B s~~
disadvantage of duster sampling is convenient units and subunits must be available. Ouster sampling is impossible if reasonable dusters cannot be defined. Another disadvantage of duster sampling. cited by Fmk and Kosa:off (1985), is a sa of
relatively cnmplc:x formulas for estimating sample sizes, means and variances. Those formulas differ depending on the exact
duster sampling technique used and are not teproduc.ed in this appendix.
3.1.4 Systematic Sampling
Systematic sampling is often used instead of nndom sampling. lt is also called an N name selection technique. Aher
the required sample size has been calculated, every Noh record is selected from a list of population members. As long
as the list dOC$ not conuin any hidden order, this sampling method is as good as the nndom sampling method, Its
only advantage over the nndom sampling technique is simplicir:y. Systematic sampling is frequently used to select a
specified number of records from a computer file.
3.1.5 Multistage Sampling
The four methods covered so IU-simplc, stratified, duster and systematic-are the simplest random sampling stratc
gies. In most real applied social research, the analyst would use sampling methods that arc considerably more complc:x than these simple variations. The most important principle here is the survey designer can combine the simple
methods described earlier in a vuiety of useful ways that help us addteSS the sampling needs in the most efficient and
effective manner possible. This is called multistage sampling (CteSWcll, 2009).
For example, consider the idea ofsampling Oregon state residents for face-to-face interviews. Clearly. the survey designer would want to do some type of cluster sampling as the first stage of the process. He rni.ght sample townships or
census tracts throughout the stue. But in cluster sampling he would then go on to measuce ~Min the dusters he
selects. Even if the analyst is sampling census tracts he may not be able to measure everyone who is in the census tract.
So, he might set up a stratified sampling process withln the dusters. In this case, he would have a rwo-stage'~ampling
process with stratified samples within duster samples. By c.ombining different sampling methods the analyst is able co
achieve a rich variety of probabilistic sampling methods that can be u.sod in a wide range of social research contexts.
STUOIF~
INn
Fnmnr..~
pnc "represemadvc~ city, even though the population includes all cities. When using this method, the researcher mlist
~c confident the chosen sample is truly representative of the entire population.
In addition, designers should move questions likely to bias responses to later questions. For example, do not follow
a question on traffic congestion wim a question asking for "the worst suburban problems of our time" with "traffic
congestion" as one possible response.
Some SUJ:Yeys contain skips, where certain respondents are directed co skip certain questions. If a survey contains
skips, designers muse ensure questions pertaining co all respondents are not skipped. Provide clear directions fo r skips,
especially for mail-back surveys when respondents cannot ask a survey worker for help.
Checklist qrstions provide respondents with a list of possible responses, from which they may choose one or more
items. The keys to constructing unbiased checklist questions include (Alreck and Settle, 1985):
The lise must include all possible answers.
The items must be mutually exclusive.
There should be more variation between items man within individual items.
"No opinio~" or "don't know" options should be offered on most checklists (and with most "yes or no" questions) to
include all possible answers and to avoid bias (Converse and Presser, 1986).
Closed survey quesci9ns may provide scales to me respondents so they can choose positions on a continuous spectrum
of possibilities. Scales are easier to construct than checklists since designers must simply fix the extremes and establish
an increment on the scale. In comparison to other forms of closed questions, scales may save space on a written survey
or save time in an interview. Once a scale has been presented to or learned by the respondent, it may not have to be
repeated while different questions are asked. Scales also lead to more possibilities for numerical analysis than "yes or
no" and checklist questions which are usually analyzed only as proportions (Aireck and Senle, 1985).
Among the most common types of scales are Likert, ordinal and ranking. Likert scales measure the deg ree ro which
respondents agree with a scacemenr, and usually run from I = "strongly agree" rhrough 3 = "neutral" or "no opinion"
w 5 ="strongly disagree." Likert scales are very common but there are several good reasons for caution about cheir
use. Respondents generally wane co agree with the analyst more often rhan disagree, so biased responses are more likely
with Likert scales (Converse and Presser, 1986). Also, Likert scales sometimes confound extreme and intense vieWs
on an issue. For example, a person "strongly agreeing" with a statement char "More highways must be buik to solve
traffic congestion" may be expressing a strong desire to hllve two highways built (an intense view) or a w~er desire co
have 100 highways built (an extreme view). Rewording questions can help overcome this problem with Likert scales.
In an ordinal scalt, items ace in some logical sequence. For example, for a question on when people prefer co perform
errands, a convenient (and oversimplified) ordinal scale may be 1 ="before work," 2 ="during lunch" and 3 = after
work." Ordinal scales provide good data on where items are relative co other items, but provide less information for
the analysis rhan do checklist questions.
Ranking st:Alts require the respondent to put a lise of items in order using some criterion. Ranking scales are useful for simulating real-life situations when choices must be made. However, data on rhe absolute standing of eaCh item on the lisr are
unavailable from a ranking scale. Care should be taken in carefully writing instrUctions for ranking scales. For inscancc;, a
survey may ask a respondent to rank five possible retail locations in order from 1 co 5, with 1 being the sire the respondent
is most likely to visit and 5 the least likely. Confusing wording in ranking questions such as this may lead the respondent
to rank each retail site on a scale of 1 to 5, instead of putting them in acrual rank order. Consult Creswell (2009), Rea a.nd
Parker (2005), or other references if further details on the abovementioned scales and descriptions arc needed.
Regardless of question format, in every effort to compose questions designers should consider some general tips. The
main rule on composing questions is to keep the study objectives in mind. Every question should help satisfy those
objeCtives. Effective questions should also:
Be as shon as possible. Use standard demographic questions and terms whenever possible (Sud man and
Bradburn, 1982). For example, srandard U.S. Census Bureau occupational category definitions are available
(Exhibit B-3) that will allow easy comparisons of results to other sucveys, among other advantages. In
fact, borrowing successful questions from other analysts (after obtaining permission and promising to give
proper credit to the author) is an acceptable, inexpensive way to constrUct many surveys quickly.
Avoid technical jargon and slang. Use words simple enough to be undersrood by the leasr sophisticated
respondent. A survey of the U.S. general public should not exceed a grade 7 to 9 reading levd. For comparison,
this chapter is written for a college-educated audience at a grade 14 co 16 reading level. Many word processing
and giammar checking computer programs judge the reading level required to understand text.
Define terms carefully. Respondents sometimes misinterpret very simple terms. The word "you, for
example, can mean the respondent or the respondent's family, depending on the content of previous
questions and other factors (Sud.man and Bradburn, 1982).
Avoid controversial and other words that evoke strong responses. Respondents associate such words as rnm,
liberal and traffic congtstion with strong images that can bias the response to an otherwise mundane question.
Avoid racial, eth.n.ic, sexist, or otherwise biased language. Look for the presence of biased language during
pretests (discussed later).
Avoid leading questions. For instance, "How clear is it to you that mass transit saves fuel . .. ?"
Avoid loaded questions. Loaded quescion.s usually associate one aspect of the issue under question wirh
some desirable trait. Alreck and Settle (1985) provide the following example of a loaded question: "Do you
advocate a lower speed limit to save human ~ves?"
Avoid double negatives. For instance, oo you agree or disagree that highways should not. .. ?
Avoid "double-barreled" questions. In double-barrded questions, respondents may answer one part one wa:Y"
Appendix B 50~
and another pan another way. Double-barreled questions usually result from uying to do too much in one
question and are tteaced by creating separate questions from the pans.
Be very careful about prompting respondencs' recall. Mentioning an example of a possible answer during
the question may bias the response in favor of the item mentioned.
IWiponses to threatening questions may be biased. The number of respondents claiming to use seat bela,
ride public transportation and perform other socially desirable actions will probably be overestimated unless
techniques (see Sud.man and Bradburn, 1982, for example) are employed that account for these biases.
For instance, this could include a comparison group survey of a similar population which does not have a
treatment installed.
Avoid hypothetical questions. Ifhypothetical questions cannot be avoided, append them onto one or more
questions on real-life cxpcrieno:s. Then use additional questions to find out what respondents were thinking
while answering the hypothetical question (Converse and Pl=er, 1986). For aample. a designer may want to
ask whether a respondent would use an clearonic map that shows traffic congestion in her automobile. The
designer might lead the hypothetical question by a question on whether the respondent listens to traffic reportS
on the radio and might follow the hypothetical question with a checklist of the desirable fi:atures ofan electronic
map.
Ask respondents to recall events in the roost immediate time frame feasible. The choice of an appropriate
time frame avoids ukscoping, the tendency to underestimate the passage of time since an event occurred.
Longer time frames are possible for events that were very important to respondents.
Construaing high-quality swvey questions
JOB CATEGORIES
A. I!AIIUFACTVR- Of 'f'RAMIPOffllmOM EQUif'tiiEifl
I . OTHER MAHUMCT\Jft.
C. AGRICUL1URE, FO~ AH0 RSHIERY
D. MlfClNO
E. BUSINESS SERVlCESAHO REPAIR SERVICES
\
l
.til. Ml I A I 1"\1: TO A
~IC:Df'\DTAT11"'\~I
OCCUMnON TYPES
A. I'ROfUaiUI.CIIl ~
a. FMMlR OR IMAM MMM1!1R
c. I'MIIII L.IIIICIIIeR OR M11M I'Oia!IIM
ll. . , . , . LAIIORI!JI
E. MAICA8I!II. OPI'ICIAL. CNMift 01' A BUSINESS
F.
CURICALMD - . v i WORDRS
H.
a. uua
I.
~f\11"\
rr'\ITII"\111
Notice the skips in questions 2, 3 and 5. The usc of special graphics, such as bold arrows and dash-dot boxes, helps
guide respondents through those difficult areas. Other features to note include:
other responses arc provided;
the .language used is easy to understand for the general population and assistance is available fo.r those who
may be illiterate;
precise directions are given with several questions;
demographic questions arc placed at the end of the form; and
respondent comments arc encouraged.
The designers could have improved the survey form by eliminating the coding spaces. Also, many on-board surveys
suffer from biased samples, since those who ride more often will receive more forms. To adjust for that bias, a. question
on how often the respondent rides the bus is usually i.qcludcd in all on-board $urvcys (Doxscy, 1983).
8.0 PRETESTS
During pretesting, survey conditions are simulated to scrut.inize the survey method and questions. If the survey is
found lacking during the pretest, as most are, the designer can revise it before data collection. Pretests are necessary
for all surveys. Some authors insist that an analyst without enough resources to pretest should not conduct a survey
(Sudman wd Bradburn, 1982).
While every facet of survey design can benefit from pretesting. pretests are especially useful during cwo particular stages in the process of developing a survey. Frrstly, pretests are useful while beginning to compose questions. The results
of pretests during this time may prompt wholesale changes, such as revising the format, dropping an entire sequence
of questions, or adopting different cypes of questions. Secondly, pretests are useful after questions have been fully
developed, in a "dress rehearsal" The designer can reword questions, add closed question choices, reform scales and
delete individual questions. However, larger changes are rarely warranted at this later stage. Pretesting data recording
forms, data reduction methods and data analysis techniques (with made-up data, if necessary) is strongly encouraged
during the second stage of pretesting. Many surveys of all sizes have &Jeered after data collection when analysts found
themselves with sets of inappropriate or unusable responses.
Pretest a respondent sample rhat is similar to the actUal sample, especially during the "dress rehearsal." Although drafts
of questions should be circulated for comment among colleagues, pretesting on colleagues, employees, srudenrs, relatives, or other groups unlike the sample of respondents may not be helpful. Informing respondents they are part of a
pretest is useful for pretests during early stages of survey development if decailed probes of the responses to questions
ace then performed. When respondents ace nor informed a pretest is being conducted, the pretest is more realistic, but
it will not gen,erate as much information that could be useful for diagnosing problems.
The interviewers who will conduct the survey should also conduct the pretest. The survey designer should inrensdy debrief
the interviewers about each respondent during :t pretest. The licerarure recommends pretest sample sizes &om 20 to 75.
mend designers avoid bright-colored (pink, blue, or yellow) paper. White, off-white, beige and many pastel colors are
fine for mailing. Mill-back furveys achieve highest response races with first-class scamps and lowest response rates with
510 MANUAl OF TRANSPORTATION ENGINEERING STUDIES. 2ND EDITION
1bulk mail permits, althou~h the latter is less expensive per response received. Assemble mail-out and mail-b acbur;eys
'with a cover lerter explaining the purpose of the survey, asking for parricipacion, derailing how complered sur:veys are
:~o be mailed back and providing other information.
The issue of respondent compensation musr be addressed before conducting the survey. Most rransportation surveys do
not compensate respondents. Compensation introduces an additional expense char may restrict sample si-ze and JTial
bias results. However, designers should consider compensation for longer surveys or when respondents perform ditficult tasks. Compensarion shows appreciation, indicates goodwill, gains attention, or creates an obligation to respond
(AI reck and Setde, 1985).
Respondents do not have co be compensated with cash. Designers can reduce the paperwork burden of providing cash
(receipts, tax statements, etc.) by providing small gifts chat respondems will value. The on-board bus survey discussed
earlier in the appendix (Exhibit B-4) promised each respondent a pass for free rides and is a good example of such a
small, valued, noncash gift.
------.....-------0___,
..,,?1=:.7_-:-------41!
------------__:____
-.
a ____
:: :::=- - g
=------------- ...... .---..------~=-.::..~.:=':'.:.-
---=~==
=--_ .. .....,._
mL;.,-
o0
.......
- --
~-------------
""'
--
-.:":""
: .:u
. ..__
_.. _g. -
g ::..8
- 0
--
- - -
en
-
,~....
11------------......
------------o
.F ~Jt---m=-;;:11''''
1
~------........
I
.e
--c
:.::=-:.::~:..:..- - -
n0
::.-=......
g
Cll
-a
WI===--:.~
---- -- ~---
=.:.:~:::-
;;;;;;
..~ w
.,.., ..
ft-.r'
_____..
nooltn\t
"
._.,
t::::::;.:=--.1
...,_,,... ___._....
......... ._.. .........
=:..----
:.:::=-~~::
i2
~
-...
. .. . .........,.
._
......,_
:::!::...
==.::;;-.:=~.-:::.:.
c..:
&:!
1'1
c:
__:_ I
--~ ..-.
il:'='a"
,.. _.... ..
..., _.._
a -~:,:.,;::;-.o;r:':l;'.~-n:-~~=.-..'l:'- . - =:..
:.~:..,.--:_.
H
--::r-
D=:.-s-:::_... ~
"""'
.....
---_
--------
...... ___ _
"*'
_,..
-.
...
<.... - .. -----~
UL~ f :..::::::.::
,.....
--- g
t ........
.....
...
Appendix B 51 --'
.:_....__l!t;
...,~:~;...
,~,..__
~i?'''
0 _.......,.
[!]
.
0
[!]
Terminate caU
m
Sole male
Sote female
Voun~r
maa.
Sole maJe
8 or more
specify _ __
III
male
OJder
male
Younger
female.
2nd ?OLinge..
----
m...
male
femele
Older mete
4U1 oldeat
male
Srd oldeet
mel
2nd olde.t
nu~1e
0 1 - m l
e or more
_ pecll'y! _ _ _ ~..:::::.__ __;;::::.......,=--~::...J!..c:::::...._ _..:::;....a..:::::;;..._ _..:::;:o..L..:::::;;..._ _
..::::::,.__===--'
Tell the peraon on ihe phone which membet' e>f the houMhold
Ja the sel-=ed repondent and attempt an Jntervlew.
Source: Grovc:.s, R. M. and R. L. Kahn, Surveys by TtkphoN!, Academic Press, Inc., 1979.
Sampling error. Sampling nror is due to the sdection of a sample from t4e subject population. Most
samples from a population will provide results that difftt frqm other samples. Ana.lysts can control the
amount ofsampling error when they use accepted proa:dwes for random sampling.
Coding and reduction errors. Analyses can detect and correct most cot/jng tmd miw:tWn before analysis.
Thorough training of coders, close contact between analyses and coders and frequent double-checks of coded
and reduced data are needed.
Biased Responses. Biar~d responses are very di.fficult to detect in surveys. However, once detected, there
are matherna,tical procedures to estimate the effects of biases on survey results (Cochran, 1977}. In
addition, Brog et aL (1982) and others have explored typical biases in transportation surveys and have
made suggestions for overcoming some of them. A thorough pretest is one of the best ways to avoid biased
responses.
Nonresponse. Nonresponst is a problem in surveys when the portion of the desired sample that does not
respond differs from the portion of the desired sample that does respond and a bias of unknown si.z.c: is
introduced. Correcting for nonresponse requires time and dfon, so smaller surveys may suffer more from
noruesponse errors. Response rates from 70.percent are typical for tranSportation surveys. However, there
is no staneWd response rate that should cause concern for analyses. A 20 percent response race where
the r-eSpondents are similar to the nonrespondencs i.s superior to a 70 percent response rate when; the
respondentS differ &om the nonrespoodents.
'
The best way to treat nonresponse is through prevention: Designers should consider response rare during every stage
ofsurvey development. A classification scheme for non response developed by Cochran (I 977) is helpful for designers.
The classes include:
non,om-agc people in the sample are not given the opportunity to respond to the survey.
unable to answu: people are reached but do not have sufficient information ro answer the question.
unwilling 10 answ~r: people are reached but refuse to respond.
not-at-home: survey workers try but cannot reach some people in the sample.
Designers can prevent many noncoverage problems by using an up-to-dace frame (list of units in the population). Most
"unable to answer" problems should be detected during pretesting. If respondents are frequently unable co answer
a question,. the designer can change the format of. reword, or discard the question. Skillful introductions, proper
inducements and polished questions will reduce the number of people unwiUing co answer. Finally, survey workers
should repeatedly attempt to contact not-at-home nonrespondents. A procedure for determining the optimum number of callback attempts is available from Cochran (1977).
Stopher and Sheskin (1982) describe ways tO detect and treat nonresponse bias in transportation surveys. Usually,
analysrs can judge the c:nenc of nonresponse bias after a special effort to call or visit a small number of nonrespondencs. Unbiased nonresponse can be coneaed by distributing the survey to more people without violating the original sampling plan. Analyses can correct biased nonresponse mathematically using the different characteristics of the
respondent and nonrespondenc subsamples.
11.0 REFERENCES
Alreck. P. A. and R. B. Settle. The Surory 1/nMT'(h Handbook. Homewood, IL: Richard D. Irwin, 1985.
Brog, W., E. Ed., A. H. Mcyburg and M. 1. Wermuth. "Problems ofNonreporced Trips in Surveys ofNonhome Acrivicy
Patterns. Tri11Up0rt4tion llntarrh &am:/: fou17141 ofrhe TTlliiSJ>Drtation 1/nearch Board 891 (1982).
Cochran, W. G. Sampling 'ftthniquu, 3rd ed. New York: Wt!ey, 1977.
Converse, 1. M. and S. Presser. Survty QJmtWm: Handcrafting the Suzndardized Questitm1111irr. Newbucy Park. CA: Sage
Publications, 1986.
Creswell, ]. W. llntanh Design: Qua/iuuive, Qu=riUllive, an4 Mixed Methods Approaches. Newbury Park, CA: Sage
Publications, 2009.
Oox.sey, L B. "Respondent Trip F~uency Bias in On-Board Surveys. Transportation ResraTth &cord: fou17141 ofthe
T~rtllrion &uarr:h &an/944 (1983).
Fink, A. and 1 Kosecoff. How to CAndua Survryt: A Step-by-Step Guide. Beverly Hills, CA: Sage Publications, 1985.
Groves, R. M. and R. L Kahn. Survtys by ulephone. New York: Academic Press, 1979.
Luman, S. R. and C. F. Man.ski. "Alternative Sampling Procedures for Calibrating Oi.saggregate Choice Models." Transpor14rion
Research Rrcord:Jou1714l.i7fthe Trr:nsportation Res=ch &ard592 (1976).
Rea, L M. and R. A. P~r. Dttigning an4 Condstding Survty ReseiZT(h: A Comprrhmsive GuU/e- Third Edition. San Francisco,
CA; Jo.ssq-Bass Publications, 2005.
Shesk.in, I. M. and P. R. Stopher. "Pilot Testing of Alternative Adminiruative Procedures and Survey InsmunentS. TramportaiWn
Resrarrh Rrcord:foumall1j'the T~rt4rion llntarrh Boarr/886 (1982).
Stopher, P. R. and I. M. Sbeskin. "Method for Determining and Reducing Nonresponse Bias." Transportation &search Record:
jollmAI ofthe Tramportarion Reuarrh Board886 (1982).
Sudman, S. and N. M. Bradburn. .AJJting Questions. San Francisco, CA; Jossey-Bass Publications, 1982.
5 14 MAfo:lUAl OF TRANSPORTATION ENGINEERING STUDIES, lND EDITION
AppendiX C
:
... .......... .. ... ............ ........... . .. ... ......... ........... ........... ....... ... .
Statistical Analyses
Original By:
L Ellis King, D. Eng., P.E.
UpJAudBy:
Ba.rtUm]. Scbrouler, Ph~D.
1.0 INTRODUCT19N
515
516
516
517
520
521
522
3. 1 Central Tendency
3.2 Variability
4.0 STATISTICAL INFERENCE
522
526
528
4.1 Estimation
528
529
532
533
536
6.0
REPORTING RESULTS
536
7.0
REFERENCES
538
1.0 JNTRODUCTtON
fr:er field data have been collected in a uaffic engineering study, the informacion is arranged or tabul2c~d
for visual inspection and analysis. Then the application of appropriate statistical techniques aids J.n
malci.ng the proper and most effective evaluation of the study results. To ensure complete and accllf~-ce
knowledge of existing traffic conditions, the field study must be designed and performed properly wich due coosideradon for the following statistical analysis: If engineers collect and analyz.e data with appropria te statisti C:::~!
p rocedures, they can avoid inaccurate and improper interpretations of the traffic situation. Success in determif-1 ing traffic improvements depends greatly on the reliability and correct interpretation of the informacion rb-:;;~.t
describes the traffic problem.
This chapter describes the arrangement of data into a convenient form in the section "Data Reduction, and discusses
the subject of statistical analyses in the sections "Descriptive Statistics" and "Statistical Inference. "Descriptiv( Sc~:
tiscics" is concerned with summarUing the data collected in a traffic engineering study, while "Stacisticallnkreoce:
describes the procedures for the devdopment of statistical estimates and the testing of statistical hypotheses. The 6n. ,:;;;;a)
section of this chapter describes some of the calculation aids that are available.
Appendix C 51 !!!5
This ap~ndix describes very basic statistical analyses used by transportation engineers. Detailed guidance on statistical
analysis principles for transportation applications can be found in Washington et a!. (2003) or other general statistics
references (Meier and Brudney, 2002, Moore and McCabe, 2006, Grecnshidds and Weida, 1978, Taylor and Young,
1988). When anticipating complex analyses, engineers should seck the advice of a professional statistician prior to data
coUection. Costly and frustrating analysis mistakes can be avoided with thorough planning and good professional advice
beforehand. Similarly, while spreadsheet software allows for basic statistical analysis, more sophisticated tools may be
necessary for spccialittd applications.
.
49
43
46
40
44
47
49
....
~Q~~ .. ~
.. .
48
52
45
45
47
41
45
53
40
45
46
45
41
44
38
45
46
;~tt.,7{'1;
,.. -
37
35
46
. -
44
50
48
43
42
4()
41
48
48
39
47
48
45
41
55
48
51
46
51
54
42
-. . ,..
46
41
50
35
44
52
53
52
49
47
42
42
46
44
49
41
49
45
44
35
41
44
45
47
49
52'
35
41
44
37
38
41
44
42
45
45
46
47
4.7
47
44
46
48
49
49
49
49
50
52
'53
53
54
55
44
39
42
TRAN~Pt'IRTATIIiN Fl\lt;IMFFRI~Ir.
40
40
42
45
42
40
43
45
46
45
46
43
45
46
48
4.8
48
41
44
45
46
48
50
51
51
52
46
48
~r.oTor'"'
~-2
Frequency Distribution
Analysts can condense and summariu: the large number of daca poinu shown in Exhibit C I by developing a frequency table. The consuuction of a frequency table first requires the sdection of the group or class siu:. If roo few or
too many groups are sdecred, detail is lost in the data reduction. The appropriate number of classes generally ranges
from 10 to 20. After the field data have been collected and tabulated, the range in measurements is de~erm ined by
subtracting the lowest from the highest values. The range is then divided by 10 and 20 to estimate, respectively, the
maximum and minimum class si~ thar are reasonable for the observed data. A convenient class siu: is selected within
these minimum and maximum values. After the analyst determines the class siu:, he or she selects class intervals to
completely define the acrual sample values that are contained in each class. These limiu arc written to the same precision as the original data. Exhibit C2 shows a frequency table for the speed data in.Exhibit Cl.
~
..
~.~
..-
Class
Boundaries
..
. . . ._,.. .<
...
....
au.
i fi!D
: :- ~~ ~+t&}~>~w1::~:!'~~~5
Cumulative Frequency
Rd&live
Midpoillt
Class
Fl't<}IWICf
Frequet~cy
33-34
33.5
35-36
35.5
37.38
Number
Relative
0.030
0.030
37.5
0.030
0.060
39-40
39.5
0.060
0.121
41-42
41.5
0.136
17
0.257
43-44
43.5
0.121
25
0.378
45-46
45.5
15
0.227
40
. 0.605
47-48
47.5
10
0.152
50
0.757
49-50
49.5
0.106
57
0.863
51-52
51.5
0.076
62
0.939
~3- 54
53.5
0.046
65
0.985
55-56
55.5 .
O.Gl5
66
1.000
57-58
57.5
66
1.000
66
1.000
CJ.a.;s
Inu;:val
32.5
34.5
36.5
38.5
..
..
40.5
42.5
44.5
..
46.5
48.5
50.5
52.5
54.5
56.5
58.5
Total
A _ _ ...... ....,~ ..
1::'4"9
Afcer che class intervals have been recorded on che frequency cable, each field-recorded vehicle speed is placed in che
appropriate class interval. Summing the number of vehicles in each class interval gives the frequency of occurrences
for each of che intervals selected in che spoc-speed study. The resulting cable of occurrence in che various class intervals
is known as a.frtqumcy distribution, and the sum of che class frequencies is equal co che sample size or total number
of field observations.
A reLuive.frequency distribution is obtained by dividing che cotal number of vehicles in each class by che total number
of vehicles in che sample; che relative frequencies muse total 1.0. Relative frequency distributions provide a more
convenient format for data summaries because the user does nor need co refer to che sample size. The use of relative
frequencies, expressed cicher as proportions or percentages, also permi!S direct comparison of che resul!S obtained from
different studies wich varying sample sizes. Exhibit C 2 shows boch frequency and relative frequency.
The cumuiJltivt foqumcy dim-ibutum provides a listing of che total number of observations chat are less chan or greater
chan a specilied value. A cumula.tive frequency distribution is developed beginning with the class at eicher che bo.t.tom
or che .top of the frequency cable. The class frequencies, ei.ther acrual or relative, are summed in the selected direc.tion
wich respect .to the class boundaries. The class boundaries are the most extreme values included in a given class and
are calculated to one-half unit of greater precision chan che original data. Exhibit C-2 shows che cumulative frequency
distribution developed from che frequency distribution discussed previously. If the frequency summation is from .the
small to the large values of che study variable, as in Exhibit C2, .the analyst matches che higher class boundary wich the
corresponding cumulative frequency. The analyst seleru che lower class boundary for this matching when he develops
che cumulative frequency distribution from che large co che small measurements.
The informacion contained in a frequency distribution cable may also be presented graphically as a histogram, frequency diagram, or cumulative frequency diagram. When plotted for che complete range of observations, chese diagrams
provide an opportunity co evaluate the data shape wich regard to conformance to a recognized distribution, degree of
symmetry and presence of irregularities.
Analysrs construct a histogram by plorcing a diagram with class boundaries on che horizontal axis and che corresponding frequency or relative frequency of occurrence on the vertical axis. The resultant rec.tangle represen!S che number
of observations wi.thin each class interval. Exhibit C3 is a histogram for che classes of spot speeds shown in Exhibit
C2. Al.though actu2.1 or relative frequencies may be sdeaed, the use of rdacivc frequencies or percentages provides a
diagram chac has general application and facilitates comparisons since it eliminates the in.8uence ofsample size. Nevertheless, the sample size should always be reported to fully document the study.
25%
::c:
2 20%
Gl
E;
{;' 15%
c
Cll
:I
tr
! 10%
Ll.
Cll
"'
'i
5%
0::
0%
+---..3334
3536
37 38
3940
4142
4344
45 -
47 -
49 -
51 -
53 -
55-
57-
46
48
50
52
54
56
58
25%
C'
...~
.e:
20%
Cll
(';' 15%
c
Cll
:I
cr
~ 10%
LL.
CD
>
;;
~
iii
5%
0:::
0%
30
35
40
45
50
55
60
The.fr(qutncy dUlgram is cons eructed by plotting a graph with the midpoim of each class on the horizonral axis and che
corresponding frequency or relative frequency of me class on the vertical axis. Exhibit C-4 shows a frequency diagra.fYl
for the classes of spot speeds from Exhibit C2. The points a.re plotted and then connected by straight lines. The diagram is closed by connecting the extreme points with the next class midpoints that have a frequency of zero. k wich
me hiscogra.rn, actual or relative frequencies may be selected for ploning.
The cumultztive frequency diagram is constructed by plotting a graph with a class boundary of each class on the hoc~
zontal axis and the corresponding cumulative frequency of the class on the vertical axis. The higher class boundary lS
matched with the corresponding cumulative frequency when me frequency summation is from the small to the!a.r~e
values of the study variable, and the lower class boundary is selected for chis matching if the cumulative frequency is
sununed from the la.rge to me small observations. A smooch curve is used to connect the plotted points, and exccnde:G
to the rwo extreme class boundaries with cumulative frequencies Q( 0 and 1. On cumulative frequency diagram.S,
analysts generitly express class frequency on a relative basis. Exhibit C-5 shows a cumulative frequency diagra.rn of ci::J.e
spot speed data from Exhibit C-2.
Appendix C 51~
100%
90%
~
/.
/.
80%
41
70%
~
c
60%
41
!'
:J
.50%
II..
Cl
>
40%
;:1
"5
E
:s
(,)
30%
20%
/
v
...,.. ~
0%
30
35
I
I
I
I
I
I
I
.i
I
I
I
I
I
I
I
'
I
I
I
I
I
I
/o
10%
I
I
I
I
I
I
I
40
45
so
55
60
Cumulative frequency <llilgrams are useful to visually determine certain percentiles of the distribution. Vehicle speed
srudies arc often interested in the 85th percentile speed, while pedestrian walking speed srudies typically emphasize
the 15th percentile speed. These fearures are highlighted in Exhibit C5.
Us~g percentiles of speed elimioates the imp:a.ct of outliers, while encomp:a.ssing a grc:a.tcr portion of the wnple than
the mean observation. N :m c:x:m1ple, pedestri:m wallcing speed affects sign:a.l riming (duration of the B:a.shing "don't
walk" pluse). If the phase duration were designed for the mean walking speed, 50 percent of pedestrians would be
w:Uking too slowly to make it :a.cross the entire speed. On the other h:md, designing for the 15th percentile speed ensu= 85 percent of pedestrians luve sufficient time, while being more efficient th:m a design for the slowest observed
walking speed.
I ....
....
I:
1110
~
~
t!s
::
110
100
PO
.I
1111_1
110
B
it
=:
120
120
..
. .....
140
. ...
I ...
~
....
100
II:
PO
Oi
eo
f--
til
IIII
- '
~~~~ ~I!~
MONTHOf'YCA._
DAYOP"W&EM
OAILVtAAII'I'IC UANA.....,.
z
~
..:
r-
"
1
1111
_.
~ :;:~
f1
l ' I'l'l
.,.,.
...... .
~
...
=~:~==
~
..:.
:- =
,~,c,~~~
t
i ~ a ~ ~ ~
HOUIIIP2':MOD
HOUIJIILY1AAP'ft0 WII.NAT'IQNS
_,..ST.
ELM ST.
o-
I~ DD
~1)~ .~=:::_
CJ~c._ rcr
J=:ktJ o o.cr o.
_Lc::tC1J __ ~~ao--l:J-o-o
~ ~"~ ~:___ s
o
.
9tJ D
D D d~~ =-<--h>>
,.A.TA\.
OfMC.R
.._
~: oo~
<: -
Spatial distributions are analyzed using geographical information system (GIS) software. For example, the analyst
may be interested in all records at or within a certain "buffer" of an intersection. Details on geospatial analysiscan be
Speed
i = -
EquationC-1
where:
arithmetic mean
If the measurements have been placed into classes as shown in Exhibit C-2, analysts use the following relationship -c:o
compute the arithmetic mean.
Appendix C SZ 3
x=
J:,ftul
J:.ft
Equation C-2
where:
= arithmetic mean
L/; u
Lfi
Most procedtires for determining the minimum sample siu in various transportation engineering studies are based on
the mean as the desired measure of central tendency. In scacistical inference, the mean is generally the most efficient
estimator of the average value for the population characteristics being studied. However, in some cases the mean can
be skewed by oudiers that increase or decrease the estimate.
3.1.2 Metlilln
The median represents the middle value in a series of measurements that have been ranked in order of magnitude and
divides the measurements into two equal parts. When the number of observations is odd, the median is the middle
value in the list of ranked measurements. For an even number of ranked measurements, the median is the arithmetic
mean of the two middle values. The 50th percentile value is equal to the median. The median is a useful average measure because it is less affected by extreme values than is the arithmetic mean.
3.1.3Motle
The mode is that value or class midpoint that occurs with the greatest &equency in the distribution of data. The mode
is not a reliable "average" for small samples because sevecal values with the same frequency can occur by chance. As
the sample siu increases, both the median and the mode become more meaningful as estimates of central tendency.
Depending on the data set, it is possible to observe multiple modes or peaks in the distribution. An aample would be
a gap distribution of a population that includes two different driver types (aggressive and conservative).
3.1.4 E.xamplu
The following aamples of computations for the various measures of central tendency are based on the observations
in Exhibit C-1 for unclassed data and Exhibit C-9, which is an extension of Exhibit C-2, for classed data. When speed
measurements are grouped into classes, Equation C-2 is used to calculate the mean speed. From Exhibit C-9, the summation over all classes is obtained for the product of the class midpoint and the corresponding class frequency. This
value is then divided by the sum of &equencies for all classes (that is, the total number of observations) to provide the
arithmetic mean, which is 3,006/66 or 45.5 mph (73.2 km/h). This procedure is similar for unclassed data, except
that in accordance with Equation C-1, the individual measurements are summed in the numerator, which is then
divided by the total number of observations, again giving a mean speed of 45.5 mph (73.2 kmlh).
'
-,..;-
~~~ ~--
Class Bounduies
. .... .
-~~
.. . .. ....
.
.........
~.,.
.-.,.1>.. . . 0 .
..
:..~---... -!
~~11~~,
~-
. .
. .-
;...-l_-trf~~
Class Frequency
Oass Interval
Clau Midpoint
u
fu
fu
33-34
33.5
35-36
35.5
71
2.521
37-38
37.5
75
2,813
39-40
39.5
158
6,241
41-42
41.5
374'
15,500
43-44
43.5
348
15,138
45-46
45.5
15
683
31,054
47-48
47.5
10
475
22,563
32.5
34.5
36.5
38.5
-40.5
..
42.5
44.5
46.5
I
48.5
49-50
49.5
347
17,152
51-52
. 51.5
258
13,26 1
53-54
53.5
161
8,587
55-56
55.5
56
3,080
57-58
57.5
Total
66
3,006
137,910
50.5
52.5
54.5
56.5
58.5
The median or 50th percentile value is found for classed data by a linear interpolation across the class in which the middle
value occurs. In the sppt-speed example, this value is the mc:an of the 33rd and 34th frequencies and lies in the class with
boundaries of 44.5 and 46.5. A linear interpolation for the location of33.5 in this class is accomplished by calculating:
.
(
33.5 -25)
median= 4-4.5 + (46.5 - 44.5) x
_
= 44.5 + 1.1 = 45.6 mph
40 25
The values of25 and 40 are, respectively, the cumulative frequencies at the class boundaries of 44.5 mph and 46.5
mph (71.61unfh and 74.8 kmlh). For the unclassed data of Exhibit C-1, the median is the arithmetic mean of the
twO middle values and, as mph, is calculated as:
45+46
2
median= -
= 45.5 mph
Annonrliv
li:,t;:
The mode is the class midpoint for that class with the largest frequency of occurrence and is equal to 45 mph (72
km/h) from Exhibit C-9. For the unclassed data of Exhibit C-1, the most frequently occurring value or mode is 45
mph (72 km/h).
3.2 Variability
A statistic describing the variability, dispersion, or spread of sample data is also valuable. The following three measures
of variability are widdy applied in the summary of transportation engineering data:
Range
Standard deviation
Standard error
Although the range is easier to compute, the standard deviation is a more reliable measure of data variability. The
standard error scales the standard deviation as a function of sample si1.e.
3.2.1/Umge
The range is the interval berween the smallest and the largest observations. The range is markedly dependent on the
size of the sample and is greatly infiuenced by outliers or erroneous measurements. This summary statistic is readily
inrerprera.ble and is often used with small samples of 10 or fewer observations. As the sample size increases, the variability within the sample may increase while the range remains essentially unchanged. Due to this phenomenon, the
range should not be used in any comparative evaluation, such as before-and-afrer 'analysis, that involves saniples of
different sizes. However, the range is useful to give readers an idea of outlier observations that may be hidden by the
average. For example, the mean estimate in a delay srudy may mask the fact that isolated drivers experience much
longer wait times (while most have lower delays).
.ECxt -i)2
s=
n- 1
Equation C-3
where:
r
= standard deviation
= arichmetic mean
x,
individual observation
$>::
Equation C-4
: where:
J; ,.
s,-
If the shape of the data is approximately a bellshaped curve or the normal probability distribution curve, multiples of the standard deviation can be expressed on either side of the mean
and represent limits that con~ various percenrages of the total values in a selected sample.
Limits oft I. 2 and 3 standard deviations about
the mean contain, respectively, 68.3, 95.5 and
99.7 percent of the coral observations. These relationships arc shown in Figure C-11.
3.2.3 StandArd Error
The standard error of the estimate is calculated by dividing the standard deviation by the
square root of the sample size.
SE
=..[Ti
Equation C-5
.................. _......,._:a..
... ..........
..
x-a.-
where:
SE
_.,.... ~ ,
standard error
.. standard deviation
sample size
3.2.4 ExAmples
The following illustrative calculations for range, standard deviation and standard error are based on the spot-lpCecl
data shown in Exhibit C-1 and C-9, for unclassed and classed data, respectively. The range for the classified daca i .S
the difference between the upper cl2ss interval for the highest speed group and the lower class interval for the lowes -c
speed group and is equal to (56 - 35) or 21 mph (34 km/h). For unclassed data, the difference between the maxin:u~
and minimum observation is equal to (55 - 35) or 20 mph (32 km/h), the exact value. The application of Eqwrio .C-1
C-4 permits the computation of the standard deviation for classed data. The numerator contains the followingclas.6
summations, which are tabulated in the "total row of Exhibit C-9.
Appendix C
SJ.71"
3006
137,910-bb
66 - 1
md the squm root of this expression gives a value of3.9 mph (6.3 kmlh) as the standard deviation.
The standard deviation for the unclassed data in Exhibit C-1, calcularcd according to Equation C-3, using actual measurements rather than class midpoints and class frequencies, gives a value of 4.4 mph (7.1 krnlh). The lack of agreement between the two calculated values is an indication of the loss in accuracy that may occur when data arc grouped.
The standard error is calculated by dividing the standard deviation by the positive square root of che sample size. For the
sample in Exhibit C-1, the sample size is 66 observations and the resulting standard error is (~) or 0.54 mph (.87 kmlh).
4.1 ~stimation
When sample measurements of vehicle speeds, volumes, occupancy and other characteristics are made, the tabulated
resultS coo.sist of a number of different values and a single value is often selected from the array co represent the data.
This representative value, or average, is genccally computed as the mean, the median, or the mode. This average value
is then generally taken not only to represent 'the sample of traffic observations, but also to represent the entire population of values from which the sample was taken. The resultS of many srudies are reported in the form of single-value
or point estimates.
The accuracy of a sample: depends on two factors. The first factor is chanc~. which is the probabk amtlUnt ofnror due:
co change in the average of any sample and which can be estimated. This section describes how to calculate this estimate. The second factor affecting the accuracy of the sample is the mahod ofsampling. The sample represents only the
population &om which it was drawn. In a speed srudy, if only the fastest speeds were recorded, the average: would be:
representative of only the fastest dement of traffic on that street. Similarly, a speed study would commonly focus on
free-Bowing vehicles that arc not inhibited by (slower) preceding vehicles. To obtain a truly unbiased or representative
sample of speeds in a spot-speed study, vehicles must be selected at random in such a way that each vehicle has an equal
chance of being included in the sample. The requirements for a representative sample are:
the sample must be selected without bias;
the componentS of the: sample must be complctdy independent of one another;
there should be no underlying d.i.ffcrenc;($ b~en areas from which the data are selected; and
.conditions must be .the same for all items constirucing the sample.
t
ttt<l!
I 'T'll"\tl
,.,.,..111,.,.~111 ,..
1" ...11~1,.,-
~"""'~
r",..,rTII'"\11.1
; For example, jfa laser or radar gun is used, a random sample could be obtained by measuring every nm vehicle, regard\ less of its spee<l. The value of n depends on traffic inrensicy and how quickJy measurements can be taken and recorded
1,{which may be automatic). A sampling frequency of n=l is typically unrealistic, except for very low-volume lo cations.
Values of n=S or n=10 are more reasonable, while S[ill assuring efficiency in data collection.
Assuming the sample was chosen in an unbiased manner, it is possible to calculate the degree of error that may be due
to chance by alculacing a confo/mct inttrVal {CI) estimau:. A CI of the population mean is usdUl in reportin g res~ts
of various transporution srudies and is expressed by the following inequalicy:
tas
taS
..fit
..Jn
x--<f,l<x+Equation C-6
where:
t~
= statistic of the 1 disuibution for (n-.1) degrees of freedom and the probabilicy defined by the subscript
Appendix C 529
t~:m;~~4t~~~j,.~~~~[~~it~:.~~\}1,~)~~~~~~~1'\$it;o~~~~~$~~~~
~. ~ J. ~ '. :.. ~~~'~iJII'~~ _";:."',;! - ~ ,!-;.~ :i'~-t\~.:;;~.:.. ' "}~'1:~.~~~\t~' .. ' .>-~f~..t"::~, .J~... }~~~;...~~~-tt~r.~~.~~~~-.~~
\:, .a , '"11'":; ,,
t''
lA
Confidence Level
90%
95o/o
99o/o
Degrees of Freedom
(alpha O.OS)
(alpha 0.025)
(alpha = 0.001)
6.314
12.706
63.657
2.920
4.303
9.925
2.353
3.182
5.841
2.132
2.776
4.604
2.0 15
2.571
4.032
1.943
2.447
3.707
1.895
2.365
3.499
1.860
2.306
3.355
1.833
2.262
3.250
10
1.812
2.228
~. 169
11
1.796
2.201
3.106
12
1.782
2.179
3.0S5
13
14
1.n1
2.160
3.012
1.761
2.145
2.9n
15
1.753
2.131
2.947
16
1.746
2.120
2.921
17
1.740
2.110
2.898
18
1.734
2.101
2.878
19
20
1.729
2.093
2.861
1.725
2.086
2.845
21
1.721
2.080
2.831
22
1.717
2.074
2.819
23
1.714
2.069
2.807
24
1.711
2.064
2.797
25
1.708
2.060
2.787
26
1.706
2.056
27
1.703
2.052
z.n9
z.n1
28
1.701
2.048
2.763
29
1.699
2.045
2.756
30
1.697
2.042
2.750
40
1.684
2.021
2.704
60
1.671
2.000
2.660
120
1.658
1.980
2.617
co
1.645
1.960
2.576
:The following example illustrates the development of an interval estimate of the mean speed for the classed data that
!u-epresented in Exhibit C-9. Sample values of mean. and standard deviation were previously calculated as 45.) mph
taS
~nd 3.9 mph (73.2 km/h and 6.3 km/h), respectively. The term Vn is now calculated as
2.000 3.9
tas
{n
calculated as 2.000
(.f66)
or 1.1. This results in lower and upper limits of 44.4 and 46.6,
This interval is the measure of precision of the sample, assuming the sample selection is random and unbiased. lf
the sample error is too large, more observations should be obtained, making certain, of course, conditio ns have not
changed since the original observacions were made.
4.2.2 Proportions
When a study is made in which the result is expressed as a proponion or a percentage, the precision of the estimate is
indicated by rhe following inequality:
Equation C-7
p-~~~<<p+~~~
where:
=1.0- p
t = statistic of the t distribution for (n-1) degrees of freedom and the probabiliry defined by the subscript
n = number of observations
IPii
The term ..J-;;- is defined as the standard error of the estimate fo~ the proportion. Exhibit Cl2 provides rypic:J.
values of the t statistic for the various degrees of freedom and for selected levels of a. The determination of the sample:
precision for proponions or percentages is very similar in procedure and interpretation to the situation for the mea~
of a sample discussed previously.
& an example of a percentage estimate, 400 drivers were randomly sampled in regard co their location of em lo en..::=
IPii
0.3. 0.7
and 120, or 30 percerit, reported they worked dowmown. The term' -In is calculated as L.960 times 400 a5
the precision of the percentage estimate for a confidence coefficient of 95 percent. That is, the estimate of 30 pero:n C::
is subject to a possible sampling error of :t4.5 percent, and the upper and lower limits of the percentage escimatear~
expressed as 25.5 I>< 34.5 with a confidence of95 percent.
Appendix C 531
If rwo samples of data are taken from the same population, there will probably be a difference between the averages
of these two samples. This d.iffercnce would be due to chance alone. A greater number of observations probably provides a smaller chance error, and hence a smaller difference berween the study averages, because the same population
is being sampled. However, if rwo studies are made at different locations where conditions ace not identical, the two
populations will have different mean values. This difference, added to the chana variation between the two studies,
equals the to~ observed difference between the means of the two studies.
Because differences between pairs of sample means from a given population occur only due to chana error, these differences are subject to the laws of probability and follow' a normal rurve. Any difference that is large enough to &II
at an extreme point on this curve is not within the realm of chana error and represents a significant difference. The
signifoanu urt for equality or inequality of the means of rwo populations with unequal variances is based on samples
of 30 or more observations for each population. For smaller samples, nonpararneuic tests arc more appropriate as
discussed in section 4.4.
r==
z
2
!1.+!1.
t=
nl
Equation C-8
Dz
where
= statistic of the t distribution
X1.
s1
11
The computed value oft is then compared with the aitical tvaluc (t), as obained from Exhibit Cl2, to determine the
significance of the difference between the two sample means. The value oft, is selected in accordance with the specified level of significance (a). The value of 0.05, corresponding to 95 percent confidence, is often chosen as the lc:vd
of significance, although an a of 0.0 1 to 0.10 is within the proper range for most evaluations of uansportation data.
If ~c computed value (either positive or negative) oft is greater than t? the d.iffercnce between the two means is considered significant and not due just to chana variations alone. When the calculated t value (either positive or negative)
is less than the critical t value, the difference berween the two means is defined as nonsignificant and due to chance
variation alone.
,4.3.2 Proportions
~alysts p~rform significance t~ting for the difference b~tween two proportions or percentages in a similar manner
bccept chat they us~ the following ~quation to compur~ the t statistic.
'
Pt-P>
t=
1)
Equatio n C-9
-/ Poqo {f. + n2
where:
t
p1
n1
4.3,3 &mnples
The following aample is presented to illusttate significance testing. Under old parking regulations, a srudy s howed
185, or 28.5 percent, of 648 vehicles were parked overtime. After new parking regulations were adopted, a similar
srudy revealed that 119 of 512 vehid~ or 23.2 percent were parked overtime. Th~ weighted average, pO' of the cwo
28.5 X 648 + 23.2 X 512
percentag~ is first calculated as
648 + 512
or 26.2 percent, md 'lois equal to (100.0- 26.2) or 73.8
percent. The various valu~ are now inserted in Equation C-8 to determine the calculated value oft as (28.5 - 23.2)
divided by the square root of 262 x 73 8 (~ +
or 2.038. The critical t value is obtained from Exhibit C 12 as
1.960 for a significance levd of 5 percent. Because t is larger chan t1 the difference between the two percentag~ is
significant, and the new parking regulations appear to be effective in reducing overtime parking. An example for nonproportion data is given in section 6.
s!V
Appendix C 533
1$ = 1- 6'J;df
n(n2- l)
Equation C-10
where:
r, = Spearman rank-order correlation coefficient
~ =
Critical values for r are given in Exhibit C- 14. Any calculated value greater than the critical value in absolute value is
a statistically significant indication of independence.
Freeway 1
Freeway2
Freeway 1,
X
Freeway2,
Difference
d =X-Y
dl
0.100
0.050
0.010
0.900
1.000
1.00 0
0.8 29
0.886
1.00 0
0.714
0.786
0.929
0.643
0.738
0.88 1
o.600
0.700
0.83 3
10
0.564
0.648
0.794
11
0.536
0.6 18
0.818
12
0.497
0.591
0.780
13
0.475
0.566
0.74 5
14
0.457
0.545
0.716
15
0.441
0.525
0.689
16
0.425
0.507
0.666
17
0.412
0.490
0.64 5
18
0.399
0.476
0.62 5
19
0.388
0.462
0.60 8
20
-o.3n
0.450
0.59 1
21
0.368
0.438
0.576
22
0.359
0.428
0.562
23
0.351
0.418
0.54 9
24
0.343
0.409
0.537
25
0.336
0.400
0.526
26
0.329
0.392
0.515
27
0.323
0.385
0.505
28
0.317
0.377
0.496
29
Q.311
' 0.370
0.487
30
0.305
0.364
0.478
,--!
--
--
--
---
--
Substituting the example data in Exhi>r C-13 in Equation C9 gives a coefficient of.
r,
=1 -
6x20
( _ )
7 71 1
=1 -
0.36 = 0.64
The crirical r, value is obtained from Exhlbit Cl4 as 0.786 for a significance level of 5 ptrcmt and n = 7. Since c]::JL<:
calculated r, is less chan the critical value, it may be concluded chat speeds on the rwo freeways are dependent (that i.5
are related to each other).
Numerous other nonpararneuic rests are available for use when uansponacion data are nor suitable for the cbssic~
Statistical tests and only one has been presented here. Complete explanations and illustrative examples of the proc~
dures may be found in Taylor and Young (1998) and Washington ec al. (2003).
Appendix c
53~
Histogram of SPEED
Mldpolat
Count
33.~
0
2
2
4
9
8
15
35.5
~7.5
39.5
41.5
43.5
45.5
47.5
49.5
51.5
53.5
55.5
~1.5
SPEED
SPEED
10
7
~
3
1
0
N
66
MIN
35
N=a"
......
..........
...........
MEAN
45.5
MAX
MEDIAN
~.s
STDEV
4.43
~s
Measures of central tendency and variability are also readily calculated in spreadsheet software using built-in functions
in Microsoft Excel such as MEAN(a-z), MEDIAN(a-1.), MODE(a-1.), MIN(a-z), MAX(a-1.) and STDEV(a-z), where
(a..z) corresponds co the range of cells containing the sorted or unsorted data. Microsoft Excel also allows the analyst
to conduct basic Hests and other tests of signifi=:tce as pare of its analysis tool pack. The reader should consult the
sofrware documentation to become familiar with these procedures.
In anorher application, Exhibit C-17 shows resultS of multiple simulation scenarios. For each (hyporhetical) scenario,
borh rhe mean and standard error of travel time are shown, calculated from multiple iterations or runs. This form
of display allows rhe user to visually compare rhe performance of two scenarios, while acknowledging the variability
in rhe estimate. The c:xh.ibit shoWs ~ both scenarios 3 and 4 result in a low uavd time along the tested corridor.
However, rhe error bars show the variabiliiy across runs is less for scenario 4. So, while scenario 3 results in the lowest
ove~ mean, soenario 4 may be preferred since it also enhances uavd time reliability.
45
~====--~:_~
======~--
40+---------------------------------------~
3P
30
25
20
15
10
5
0
Scenario1
Scenario2
Scenario3
Scenario4
ScenarioS
Appendix C 537
7.0 REFERENCES
Greenshields, B. D. and F. M. Weida, revised by D. L Gerlough and M. J. Huber. Statiuits with Applicariomto Highway Tra.ffit
Annf];es, 2nd ed. \X'estpon, CT: no Foundation for Transportation, Inc., 1978.
Meier, K. J. and J. L Brudncy. Appli~d St4Jittia for Puh/ic AdministT1JrWn, 5th ed. Belmont, CA: Thomson ~Wadsworth, 2002.
Moore, D. S. and G. 0 . McCabe. Introduction to the Practiu ofSratistia, 5ch ed. New York: W.H. Freeman and Company,
2006.
Taylor, M. A. P. and W. Young. Tra.ffit Analy1is: Ntw Tt:chnol<Jgy and N= Soluriom. North Melbourne, Ausrralia: Hargreen
Publishing Company, 1988.
Washingron, S. P., M. G. Karlaftis and F. L. Mannering. St.ztisriazl and Econommic Mtthods for Tramporration Dal4 Analjsis.
Boca Raton, FL: Chapman & Hall, CRC Press LLC, 2003.
Appen d ix D
:
..
.. ..
..
... . . . . . . . .
.. ..
... .
......
... 0
...
. . . . . . . ..
.. .
..
.. ..
c .... .
. . . .. .
.. .
. ...
Supplemental Material on
Communicating Data
Origimd By:
Dtnma C Nelson, Ph.D., P.E.
H. Douglas Robertson, Ph.D., P.E.
Edited By:
Bastian]. Schroeder, Ph.D.
RobertS. Foyle, P.E.
1.0 INTRODUOION
540
2.0
DESIGN OF GRAPHICS
540
2. 1 Section Overview
540
540
2.3 Tables
542
543
549
549
2. 7 Graphics Checklist
550
550
550
5.0
550
552
554
4. 1 Section Overview
554
555
555
557
557
558
REFERENCES
559
Appendix D 53!JIIII"
1.0 INTRODUCTION
his a.ppendix presents supplementary material to Chapter 3, ~Communicating Data to the Public." While
Chapter 3 contains key concepts and principles of data display and report writing. this appendix provides more
detail on additional graphic types, sections of a written reporc and presentation techniques. The three ways co
ddivet content co an audience are presented sequentially, consistent with Chapter 3.
~AAMII.At
r-----+--~
I
1
r-- - - ....
I -'---I
I
-+
-+
1/3
1/3
_.J..
[Q][Q][Q][Q][Q)[Q][Q]
IQ] 131EJIQ] [Q] [QJ IQ]
[QJ IKI [Q] [Q] [QI [Q]
[Q][QJ[QJ(Q][Q][Q][Q]
IQ][Q][Q]IQ][Q][QJIQ]
[Q][Q]IQ]JQI[Q]{Q)[Q}
IQ]IQ][Q](Q)[Q)[Q)[Q]
[QJ[Q](Q]IQ][Q][Q)[Q]
[Q][Q)[Q][Q][Q)[Q}IQ]
[Q] [Q] IQl [Q] [Q] [QJ [Q) .
~
--I>
---
most likely to be remembered. The lower right intersection is the second strongest visual position and, as the last item
viewed, is also likdy co be remembered. Items placed away from these locations tend to serve as background informacion. They suppon irems in areas ofstrength but are less likdy to be remembered. These points apply to photographs
as well. For example, a photo of a landscape with one-third sky an4 two-thirds landscape is more visually appealing
than one with one-half sky and one-half landscape.
'
The visual weight of items or areas on a graphic is also imponant. V'uual weight is an elusive combination of position,
shape. color, contrast and meaning. If there is more than one object on the page, the eye tends to focus on the visually
heaviest object first. The eye either stays focused on this point or continually rerunJS to it from other items {Exhibit
D-2a). If all the items on a page have the same visual weight, there is little for the eye to focus on. With no point to
focus on, the viewer may become bored and the eye may move off the graphic (Exhibit D-2b). Items placed within
the focus areas (as shown in Exhibit D-1} draw the viewer's attention.
The shape of the items can also direct the viewer's focus around the page (Exhibit D-2c). Items shaped like arrows
that point to the lower right corner tend to direct the interest of the observer off the page. Placing a shape that either
stops the flow or points back onto the page helps the viewer focus on the page and on the message (Exhibit D-2d}.
2.2.2 Onnposilion
The strucnue of a graphic can be compared to that of a written paragraph. In a paragraph, the first sentence informs
the reader what the paragraph is about: It td.ls the reader what to expect. This is analogous to the first itern of focus
on the graphi~ The main body of the paragraph develops the idea presented in the first sentence. Similarly, the main
body of the graphic serves as the backgtound, supponing the main message of the visual. Points of inserest are estab~
lished by the objects of heavier visual weight. The final sentence reinforces, concludes and tells the reader what the
paragraph was about. For a graphic, this usually corresponds to the item at the lower right area of focus. This is the
last thing the viewer sees before leaving the graphic.
2.3 Tables
Formal tables serve rwo basic purposes: to summarize information or data discussed in the rext and ro compile reference data. The purpose of the table guides how it is designed and where it is placed in a report. Summary tables save
space and enhance comprehension by providing condensed information in meaningful form. A short, weU-designed
tabulation can replace a lengthy section of explanatory text loaded with statistics. Tables designed for oral presentations are usually most effective when they are simple. The design of tables for oral presentations is discussed in section
4. Tables included in written reports may be more complex, as the reader has the time to study and undemand them.
Reference tables generally provide material to support the text but are nor needed in the text for clarity. These tables
are generally placed in the appendices.
2.3.1 Structural Elements ofa Table
Some general guidelines for the devdopment of effective tables are described in this section. For reference, Exhibit
D-3 shows the strucrural dements of a table. Formal tables are titled and numbered for reference. Tables are normally
numbered in consecutive order, starting from the beginning ofeach article, chapter, or book. Arabic numerals are used
to identify tables occurring in the main body of a work. Double numbering (1-1, 1-2, l-3) can be used ro locate tables
in their respective sections of a report. Double numbering is appropriate when sections contain numerous tables or
graphics. Position tables as close to the texrural reference as possible but after the initial reference to them. The ririe
should briefly explain the content of the table and should be pertinent to the text. If the data are not original, ieference
the source. Data items can also be referenced in the body of the table, as footnotes to specific data items, in columns,
or in the title of the table. A subtitle contains any additional explanatory or descriptive information required to make
the main title clear. Use the same type face as the title, but use either a smaller size or lighter weight.
Column and row headings should be brief. Vertical column headings and subheadings usually contain the dependent
variables. Horizontal row headings (shown as the srub in Exhibit D-3) generally contain the independent variables.
Define any abbreviations used in the headings as footnotes to the table.
The field (or body) contains the tabulation of numerical or verbal data referred to by the stub and headings. Each unit
of the fidd where a horizontal row intersectS a vertical column is called a cell Whenever possible, data items in columns should have a uniform degree of acc:;uracy. Fractions ace usually converted to decimals and numbers are aligned
by decimal point. Plus or minus signs ace placed immediately to the left of numbers; missing entries are identified
by a dash. Shore answer entries that require multiple lines may be right-justified unless the spacing berween words is
noticeably large. Lines (rules) or grids are used to separate certain columns or lines, or to frame the table in a box. This
may help to distinguish the headings, separating heading area from the data in the fidd, and.improve the readability
of the table. Rules should be thin lines; heavy lines may draw attention away from the table and make it difficult to
read. Rule.5 work best for very small or very large tables.
/Col~~
~
'---.. Yeac
12
Otber
Arterials
6
11
. 1981
13
10
.lL
~978
10
1915
10
10
11
fM'
1987
Stub
1983
Interstates
Interstates
27
24
26
23
24
18
.--
~Decked Heads
+-Column heads
Acterials
38
39
44
43
49
45
~Cells in
body of table
! Each
table should stand alone (be understandable without referenct to text). All references and notes are handled
!separately, without reference to the text or ocher tables. Reference numbering stares at the upper left:, with a new serit:s
Jor each table (for example, a, 1, or ). Number tabular elemenrs from left co right across the table. Foornoresshould
be the same width as the table. If possible, place each note on a separate line. Footnotes titled "Note" qualify, explain,
or provide information relating to the table as a whole. They are alwa~ placed first in the sequence offoomotcs. Footnotes titled "Source" cite the sou ret of the information and are always last in the sequence of foomotes.
Readers are accustomed to reading information from lefr to right. However, by definicion, rabies are struct ured ~round
venical columns. Data in vertical columns appear co be easier to read than horizontal comparisons. The c ricial problem is to make the left-to-right relationship clear with respect to dara in venical columns (Whice, 1988) _
Organize the data to make imerrelacionships as visible as possible. Presem data so comparisons are possibk both
within a table and among cables. Organize columns of data to suppon the purpose of the table (for instan ce, to show
similarities or differences, statistical trends, or interactions). For example, a table presenting data on high accident
locations may be otdered from high to low based on the accident experience.
Exhibit D-3 shows the repair perctncages grouped by the need for repair for "interstates; and "ocher arterials." If the
designer wanted to com pate repair needs by faciliry rype, the second levd decked headings would be *In cerstate" llfld
"Ocher Arterials." This would pur the data for interstates in the rwo left columns and for ocher arterials in the r;WO
right columns. Thus, a different purpose for the cable corresponds to a different ordering of rhe data.
If users must c:xtract data from a table based on a name, place the enuies in alphabetical order based on the mllj or
sueet. When designing reference cables, consider the ways in which readers may want to use these rabies.
Avoid large gaps berween cohunns. Gaps. tend to confuse the reader because the eye must move across empry artaS t:O
locate the next item in a row. Using generous spacing berween the lines can help improve horizontal tracking, whlch
makes it easier to read across a cable. If a cable is large or complc:x, place it on a separate sheet. Small, simple tables
should be placed directly in rhe tc:xt. Make sure the cable supplementS and complementS, but does not duplicllce,
informacion in the text. All tables should be referred to in the tc:xt.
Appendix 0 ~
-- -- --- .
.
-----+-----~~====--~-----
280
280
~- -----
240
220
200~
180
------b-----~---------~-.,---------HIO
140
- - -- - - -- - .., _ _ _ _ _ _ _ _ _ _ .. _ _ _.._ _ _ _ _ , __
120
...:___
.'---
__
11me period
.Not!': HI Lo graphs can be used to display expected end observed values.
45
40
ic:
35
30
25
15
,E
z
20
10
s
0
1980
i:
CAA
1982
..
1988
Veer
1~
\ 2.4.3 Pictogr11nu
.
!A pictogram integrates graphics, symbols, or icons into various types of graphs, including bar graphs, line graphs and
! pie charts. Exhibit D-6 is an example of the use of pictorial symbols to illustrate sign legibility distances. Pictograrns
\hat are too complex may be difficult to interpret.
Di s tances shown in
~e e t
r- 800
800-
.....f
w~0~
~~- ~~~
300.....f~
~~
""" 700
1-
600
t-
500
t-
400
I-
300
t-
200
' - 100
Source: J.F. Paniaci. Legibility and Comprtthmsion ofTraffic Sign Symbols, Research Report. Federal Highway Mminisuation,
McLean, VA.
~~17.:
Accidental Death
r--=--
Rates
by States, 1971
---,
,o
.......
==BELOW 50.0 13 STATES
lllll 50.0 TO 58.9 11 STATES
8,_ 80.0 TO 6$.9 13 STATES
70.0 & OVER 13 STATES
to..tu: filM".,,~-- o,. ~o;sc-.- 0t1 ~~
..-e... . ..,.
~v.a, c
:::=====.
---------------------,:
!-Coordination
Mth od\ttr
OfPJlfzati:lns !
--------~Departments,
l~~~~~!~~~~!:~~~-j
Appendix D 54::;;;o?
NO
~ ~ 8IQHT
DISTANCE
YES
'\'I!
YES
EXJa'TNi
DOES
OR l'liJ1..IAE T'I'IAtPIC SA'nSFV
ANY OF THE 8IQNAL ~AU.AT10111 WARRANTS?
ACCC)ERT
RECORD
WORSE WrTH
NO IJICFFCANT CHANG
1'1 ACCIDENT AECORO
OR' NODATA
YES
8TAGE I
Source: J. L Kay, L. G: Neudorff, and F. Wagner. Critnia for Removing TMjfic Signals aru/ Usm' Guide. (FHWA-IP-80-12).
Alexandria, VA; JHK Associac~ Lnc. and Wagner-McGee AssociaEeS, 1980.
IltiDJ;iiLI''
..
..
~ .r~
lint line
v.. t
TASK
I* I
Proje<t M-.ting
A
&I
..
~ole.;,qs.
B OMiopMm~l.tat.rill
~ IIIIP..,...RMeo~Malbllng\1\doo
12
Year2
11 21 21
l
i
t5
ij_~
~~
i
.i
:
!
!I
i
I
I
I
I
I
I
I
I
I
I
I
!!
'
I'
_l
'
iI
"
fkl
i
I
I
'
I
I
iI
l t
:I
i
i !I !
t .
:
I
I
I
I
ln
i
:i .
i;
'
I
I
YearS
54 57
51
'
i
I
I ~
iI
J.
.l
!
!
I
!
I
I
' !
I
!;
I
at
*!
_!r : l
!
II
v....
Yur3
~
i
II
I
, .~
The scale (or ~pect ratio) of the graph also affects the way data are perceived. A ratio of height to width of 3:4 is common. Designers may need to experiment, changing the scale of :1. graph ro determine the most effective aspect ra.rio
for its usc. Too great a range of values causes the graph to appear compressed. Some spreadsheet packag-es will allow
the usc of more than one vertical axis scale. Label them clearly co :~ovoid confusion. Type fonts (such as Helvetica and
Ttmes Roman) should be consistent with those used on tables and in the ten ifpossible. The we of different sizes and
weights of the same font type works weU.
develop associations;
build rerenrion; and
create an aesthetically pleasing product.
There arc some general rules for using color in pleasing ways. A color whed (which shows the 12 basic hues or colors)
is a very useful roo! in choosing colors. h can be purchased in any art supply store. The primary colors are red, yellow
and blue. The other (secondary) colors are produced from these three. Cool colors (blue and predominately blue) are
relaxing and appear to recede on the page. Warm colors (red, or red and yellow) are stimulating and advance to the
foreground. Green, brown and red-purple lie berwcen rhe warms and cools and arc therefore relatively neurral.
An important use of color is to aruact attention, primarily through conrrasr. Color should be applied to the elements
of greatest significance. A bright color used with black is the most effective. Complementary warm and cool colors can
also be used sparingly. Blue appears to be the most popular color. The colors used in a graphic should fir the overall
mood. Too much color can detract from rather than improve the readers understanding of rhc graphic.
Tide page
Copyright notice
Disclaimers
Forward or preface
Acknowledgmenu
Table of comenrs
List of tables
List of figures
Summary or executive summary
Body of the reporr
List of references or bibliography
Appendix
Glossary
Index
The sequence presented follows normal conventions but may be altered to fir a given situation. Each component is
discussed briefly below.
subject m:mer further. Rq>orts describing cransporution engineering studies do noc usually contain bibliographies
unless the author wants to encourage further study or aplanation.
3.2.13 Appnulix
Detailed material that supports but is not essential to the body of the report should be placed in an appendix.
Appropri:a.te appendix materials include supporting d:a.ta, detailed aplanations of methodologies or procedures,
derivations of formul:a.s, conversion f.u:tors, lists of symbols, d:a.ta collection formaa, d:a.t.l. collection P.rotocols and
checklists. Appendices are effective means of fully documenting :a.nd supporting the results of a study without
cluttering the body of the report.
3.2. U Glosury
If the report is written for nontechnical re:a.ders or if new and unf.uniliu terms are introduced, a gloss:a.ry .or list of
defin.itio~;~s should be included in the report. If few new terms are wed, footnote definitions may suffice.
3.2.15lrulet
Transport:a.tion engineering reportS seldom contain indices unless the report is voluminous or intended for frequent
reference. Th~ inda is more aact and detailed than the table of contents. The index liru all major subjects alphabetically along with the .,ages where each subject is addressed in the report.
C"Tlll"''~c:c
-,,.,u"\ s:nrTtf'\M
Order of.Andie:nce lntuest or Familiarity. Treating the most populu subject first, rcgardiess of its imponanc;e, may
be a way to gain the readers' interest and lead them into less familiu subjects.
Grouping Similar Subjecu T~ Similar subjects should be grouped together and presented in the same section/chapter to ease reading and understanding. For e:nmple, planning chapters may be followed by design chapters,
which in turn ue followed by operations and safety chapters.
methods used;
data collected (for example, type, amount, when, where, etc.);
analyses performed;
findings;
conclusions; and
rccomiiiendations.
Appendix 0 553
Examples of subjects chat would be appropriate for a transportation engineering study report include:
alternatives developed or examined:
selection of alternatives, traffic control devices (TCDs), or routes; and
evaluation results, such as:
cost analysis or financial impact;
environmental impact;
traffic impact; and
implementation plans or recommendations (for example, preferred alternative).
Use Simple, Clear and Concise Language. Remember, a transportation engineering study report is a presentation of
technical faccs and their implicarions, not Literary art. On me other hand, write the report to hold the reader's attention.
Do not use complicated wording in an effort to sound important." Most readers prefer simple, common language.
- Use a Good Style Manual. The elements of style in wriring rechnical reports vary &om one manual to the next. Pick a good
manual, adopt a set of conventions and maintain uniformity in spelling, grammar, punctuarion and formar throughout the
report. Sevetal good technical wriring style manuals are included in the references at the end of this chapter. Agencies such
as U.S. DOT and the Transportarion R=rch Board (TRB) also maintain technical wriring style manuals.
Page Numbering. Number every page except the tide page and the letter of transmittal. The summary or the ~st page
of text can be "Page 1." Advance pages are nwnbered with lowercase Roman numerals. Appendices may continue the
numbering of the main body or can be numbered A-1, A-2 and so on. Place page numbers at the top or bottom center
of each page.
Break Up Lengthy Narratives of Pure Text. Headings and subheadings lead the reader through the author's train of
thought and presentation of &as. Make ample use of summary tables and figures. Photographs may also be useful in
describing certain situations or conditions.
&plain the Methodology Used. The credibility of the results of a uansportacion engineering study often hinges on the
type and amount of data collected and the methods used to analyze those data. Therefore, the report should clearly, yet
concisely, provide the reader with sufficient infon:nation about the data and methodologies employed t.o establish an acceptable level of confidence in the findings. Explain the application of standard methods by referencing a text or manual
on the subject. Special methods require more explanation that may be placed in an appendix to avoid cluttering the body
of the report. The sources of data and informacion used in the study should be clearly identified. Authors should also
specify the amount, place, time and conditions under which field data were collected.
Rdationsb.ip of Findings to Conclusions and Recommendations. This relationship is similar to the construction of
a ho~ The findings are the foundation on which the conclusions rest. The conclusions, in rurn, support the recommendations.
Reviews and Editing. A good technical writer, a panel of reviewers, or an e:xperienc.ed editor should review the report,
if possible. This simple practice often turns a dull, Lisdess recitation of facts and figures into an interesting, understandable and perhaps even enterta.ining piece of writing. The goal for any report is that it be read and understood correctly.
A report that readers ill?ore or misunderstand has no value and the work and dfort to produce it will have been wasted.
4.0 PRESENTATIONS
4. 1 Section Overview
This section discusses some additional aspects on making oral presentations that adds to the discussion in Chapter
3. Sometimes the only information on a transportation project the public sees or hears about is from a presentation.
Effective presentations help explain important results of a srudy, can inform and educate the public about the issues
554 MANUAL OF TRANSPORTATION ENGINEERING STUDIES. 2ND EDmON
. involved in a project and can provide feedback on which of several alternative designs are preferred and why. There~ fore, the presemer needs to understand these addidonal concepts in preparing for and presenting information.
There may be facsimile or call-in questions. These should be filtered by someone in the studio and then shared with
the presenter{s) by the session moderator. See Section 4.5 on how to answer questions &om the audience.
Appendix D 55~
to the audience. It is better to present less and cover it thoroughly than-to discuss too much and have the listener feel
you did a superficial job.
4.3.3 Presentation
The medium of the oral report is the speaker. Therefore, the speaker's adivery style is critically important. Posrure,
gestures, eye contact and &.cia! expression, voice projection, enunciation, pronunciation and the degree co which
the speaking style seems relaxed, conversational and expressive all affe~ the attitude and receptivity of the audience.
Four styles of presentation delivery include extemporaneous;-impromptu, manuscript and memorized techniques
(Michaelson, 1987). The o:tnnporaMOus talk is usually preferred for. the oral presentation of a technical report. In
this method, the oudine of the speech is carefully prepared but not committed to memory. The impromptu (offthe-cuff) method is obviously inappropriate for the presentation cif.lil\e report; however, it is an appropriate style
for the question-and-answer session. Prepared and read manuscripts dr'Jm.f"'orized methods tend to be inflexible and
sound artificial Practice, however, will improve the quality of an~tation.
'""'-'
Oral presentations frequendy include questions from the audi~ec. ~uestions may be asked during the presenta
cion or be deferred to a question-and-answer period that follo'IIJ t}tt formal presentation. While each method has
its advan cages, the mOS[ satisfactory arrangement is to defer q~t0tfonittg until the formal presentation has been
completed. Potential questions are often answered during the remainae'r of the talk.
The presentation should be made in an appropriate location. ldeallr.,the space needs to be comfortable and sized
for the number attending. It should contain all the necessary equipmerX for the presentation, usually a lectern with
a light and computer projection equipment. Sometimes the hos~ presenters to bring their own laptop for
use with the projection equipment, but that is quickly changin~tlttrs are presenting during one session, put
your presentation on a memory stick for use in the host computer.
'
4.3.4lnstant Urukrstaruling
: .. ~
~ ~,..
If some portion of a written report is not immediately clear, th~ay reread it, look back or ahead, or even
consult other sources. However, the spoken word only lases during
, ment of presentation. The speaker must
be exceedingly careful co be as clear as possible at all times, payin . . . ar attention to audience reaction. Instant
understanding is &.cilitated through voice, language and the usc:.,
s~tions and summary statements.
.
~:...,~
Proper use of voice includes adequate projection and distinct enli:nt:t.rtion to permit even those persons Sealed
farthest away to hear and understand the material presented. The speaker should use pauses freely to break up the
flow of ideas into meaningful thought unirs. Also, speaking wilh sufficient forcefulness and using a variety of inflections avoids monotony and gives life and meaning to words. A speaker must be sensitive to the rate of delivery
and must pace the various remarks for understanding, variety 1-nd emphasis. Practice is essential to master all of
these techniques.
...The audience must understand the vocabulary used. If the readers~a. written report encounter an unfamiliar
word, they can refer to a dictionary. In an oral repon, the sp~ mmnrezine unfamiliar terms foe the listeners. Obviously, the problem varies from one situation to another, but the speaker should be sure the listeners
understand the terminology. A broad vocabulary permits an expression of thoughts with clarity and precision in a
vigorous ~d colorful whion. Transitions and summaries help guide the listener through the developmenc of the
presentation topic. After a section of the repon is developed, summarize the main points briefly before moving on
to the nat section. Also, show the relationships among various sections.
4.3.5TOrganize and 'rehearse the presentation for the established time limit. The average rate of speaking is 100 to 150
worqs per min. (with listener comprehension at up to 600 words per minute). Tuning can be checked by practicing
with audiotape, videotape, or with a live audience. If available, videotape is the most effective. Ic gives the presenter
the opportunity to see. how he or she is coming across and how 10 impc~ his or her presentation skill.s. Simply
speaking aloud hdps check the organization of the calk; however, whc:P-Practicing in an empty room there is a
ten~ency to speak much faster than on tape or to an audience. Rehearsal will help the speaker achieve a relaxed,
at-case posture, and a smooth, confident delivery.
14.4
)A well-conceived talk, carefully tailored to fit ilie audience, will fail if it is difficult to follow. A confused reader of a
report can regress as needed to wade through an obscure passage; a confused listener is likely to be lost forever. P resentations are organized around three major divisions: the introduction, the body and the conclusion.
4.4.Ilntroduc1Um
The introduction prepares the audience for th'e body of the presentation. The introduction should motivate the listener, catch the interest and inform about the'concent of the report. Motivation is accomplished by dwelling briefly
on the importance of the subject and by establishing a distinct impression that something worthwhile is to be offered.
The speaker can inform listeners about the tof'ic of the presentation by:
4.4.3 eo~lusion
'
The conclusion normally fu.lfills the following; three main functions:
1. summary of the various subcon'clusiyns presented during the presentation
2. Do not use an opening line of nonsense patter to gain time to think of your answer. For c:xample, do not
start with, "Let me just think ouc loud for a minute" or "That is a very difficult question. Pause briefly
before starting to speak if you need a moment to organize your ideas. Repeating the question clearly will
allow you to gain some time and confirm that you undeiStand the question. Do not assume the questioner
will deduce the answer correctly froin a long discussion of pros and cons. Be explicit. If your answer must
be long or compia, make a brief SUflllllary statement.
3. Do not bluff. If you do nor know rhe answer to a quelition, the following choices are available: "I am sorry,
I cannot answer your question;" or "If you please, I would like to consult my notes (references, partners)
before I answer that question;" or "! can nor answer char now. T wUI have the answer chis afternoon (tomorrow, next week) ."
4.
Don't be evasive. Answer the question you arc asked, or do nor answer. Do not answer a different question. The questioner will not be impressed or fooled. Avoid hedging. Sometimes the answer is not simple;
ir must be qualified or have limits set. In these cases, tty to make che limirs quantitative and make the
qualifications precise.
5. Be specific. Use precise words. Favor citing a number over using a descriptive phrase. Never assume you
and the listeners have the same scale of reference for imprecise words (words such as dup, high, wann,
6. Do not try co answer a question you do not understand or one chat is imprecise. If the question is unclear,
your answer will be unclear. Even worse, it may be wrong. Do not guess at what the quelitioner has in
mind, ask for clarification. You may rephrase the question slighdy, but do not change it. Ask for confirmation that your interpretation is correct. If you are answering questions from a group and receive a com
ment far off the point, or far wrong,_decide if it is best to assume others also recogni-ze this and move on
as politely as possible. Be polite but firm in your posicion; however, never be argumentative or insulting.
Above all, remain calm. If you arc: asked a hostile question, restate it using positive phrasing. Avoid single
or double negativeli. It may be helpful to restate a negatively phrased question or one that uses angry words
or emotionally loaded terminology into one that is clear, precise and contains less inflammatory language.
7. Never use slang and avoid jargon. Slang is imprecise and it can give the impression the speaker is poorly
educated. Jargon is vocabulary known to a group of specialistS but generally unfamiliar to outsiders. Often,
jargon can be replaced by a few more common words.
4.6.1 Organization
Well qeforc: the prelicntacion date, consider the following:
Does the introduction aplain why the subject is significant?
Arc: the major points covered withour excessive detail?
Will the audience understand all the terminology? Is the talk simply arranged in a logical sequence? Are the
visl12ls ~mple and visible?
4.6.2 Rehetn'SII!
Rehearse the presentation until you feel comfortable with the material.
Rehearse with the visuals you intend to usc:.
Identify and practice fielding possible questions.
Is the room equipped with the necessary projection equipment; is it in working order?
Is the temperature of the room neither too warm nor coo cold?
5.0 REFERENCES
Berthouex, P.M. and D. Hindle. "Abour rhe Presentation of Your Paper: Oral Communication: So me Guidelines on AnsWering
Questions." Engineering Educarion (December 1989}: 243-244.
Bel(, P. C. Manual ofTraffic Enginuring StuditJ, 4rh ed. Washington, DC: Institute ofTransporralion Engineers, 1976.
Editorial Scaff of the University of Chicago Press. Tht Chicago Manual ofStyk. Chicago, IL: Uni~ersity of Chicago Press, 1982.
Hodges, J. C., M. E. Wh.inen, W. B. Homer, S. S. Webb and R. K. Miller. Harbrau Colkgt HaMboolt., lith ed.: San DiegO
CA: Harcourt Brace Jovanovich, 1990.
Hodges, J. C, M. E. Wh.inen, W. B. Horner, S. S. Webb and R. K. Miller. Harbrt Colkgt Handbook, 13th ed. Chicago:
University of Chicago Press, 1998.
.
Insurance lnstirute for Highway Safety. Fan:s. Washington, DC: Insurance Institute for Highway Safety, 1991.
Kay, J. L. L. E. Neudorff and F. Wagner. Crittri4 for &moving TraffitSignaiJ and U1m Guitk, FHWA-IP-80 12. Alexandria,
VA:. JHK and Associates Inc. and Wagner McGee Associates, 1980.
Ksanier, M.G., J. W. Presley and D. C. Rigg. Prmtiu-HaU WOrkbook for Writen, 4th ed. Englewood Cliffi, NJ: Prentice Hall
1985.
Lala!li, N. and D. G. Gerard. "Communicating Eff=ivelywith the Public." ITE]D-urnai(Occober 1995): 51}.-52.
Lalani, N. and S. B. Colman. "Making Effective Technical Presentations." ITEjournal (January 1995}: 29-38.
Mason, J. "Communicating with Elected Officials. ITE]ournai (December 2005): 18-22.
Michaelson, H. H()IQ UJ wn~ and Publish Enginming Papm and Repom, 2nd ed. Philadelphia, PA:. lSI Press, 1987.
National Safety CouncU. kcidmt Fan:s. Chicago, IL: National Safety Cou~cil. 1986.
Paniati, j. F. Ugibility and Comp"hmsion ofTTYJjfic Sign Symbou. Research Report. McLean, VA: Federal Highway
Administration, 1988.
TurnbuU, A. and R.N. Baird.
Rindurt and WU!Ston, 1975.
White, J. Graphic De1ign for tht Ekctronic Agt. New York: Watson-Gupcill f?ublicarions, 1988.
Appendix 0 55~
.......
Append ix E
!,
io
561
1.1 Overview
561
1.2 Coding
562
564
This appendix conu.ins useful ~esources for transporracion studies. The first part discusses how analysts can create a
time-stamp macro in Microsoft Excel that can be adapted to various data collection applications. The second pan
presents commonly used data collection forms. Each form may be removed or photocopied !Or use. The fOrms are ident:i.fied as to source in the manual by chapter number. The details of use are described in the text near the page on which the
form appears or is first mentioned.
The method can be applied in real-time field studies or from video observations. In all cases, the analysts predefines
ce.ru.in keys to correspond to specific events of interest. It is important to practice the actual data collection befOre
going live,~ especially if the macro was previously used for orher sru.dies. Coding a large number of events can create
a high cognitive codirtg load, and human factors' principles should be considered when laying out the key structure.
Foe example, the aforementioned signal compliance srudy might use keys "A," s and o to denote vehicle arrivals in three lanes. These keys would be operat~ with the lefr hand. The right hand may then be used for example to
and "K. The c:nct configuration depends on
record signa! phases "green; "yellow" and red; using keys H;
the specific study and usee preference.
If the coding load becomes too high, the analyst may decide to supplement additional data at a later time (from video)
using a common uro-time reference point. Events ~ then be ordered by sorting by time stamp. Alternatively, two
coders can be used co collect different data, either on the same or separate keyboards. Note that MS Excel will not
recognize an external number pad for this macro, but will support a full external USB keyboard.
1.2 Coding
The time-stamp macro relies on a builtrin MS Excel function, NOWQ, that r~pons the instantaneous computer system time. The function is automatically updated every rime the Excel worksheet is updated in any form (for example,
by entering a value). During data coUecrion, every rime the analyst presses one of the preconfigured burrons, the
macro (very quickly) performs three taSks:
1. It enters the key code in the accive column (A) and row (l) and moves the input cursor inco the next column (B).
2. It copies and pastes the cell containing the NOW() function into the active cell.
3. It moves the input cursor back into column A, buc down one row (2), where it waits for the next key input.
The actual macro code is wriucn in VB, a programming language included in Microsoft Office produces. The macro
is activated by hitting a uSrarc Recording~ burton that is in turn associated with a function in.VB. The macro continues to respond to keys until the analyst hits "Stop Recording," at which cime Excel is returned co its normal srate.
Exhibit E-l shows a possible layout of che input screen. The Excel worksheet contains a title bar, the "scan:" and "stop"
buuoru, the system time stamp with the NOWO function and a data table.
-
'
LLO
!
23
24... i
.........
_. ..,..
rrrrt '": . . , . , . .
251
2'
_,._
lfltlfll ............
.........
V I
2S
IT L IIITJIJNfllf)ll ..........
lfllffiiUJIIIIII Ill ........
.!
'
- .. ..... ..
i----------i:~:~~~~~=-~:~~t~~: _::.~::_
~- .......
-
- -- ~
\The macro only records data in the first two columns. Column C contains equations that automatically deduct the
:active row time sump (see cdl Bll in Exhibit E-1 ) from the first time sramp in the study (cell $B$10) using the
.function [=Bll-SB$10]. Since the NOWQ funCtion works with the system time in che format DD:HH:,MM:S$.00,
column D multiplies the values of column C by the factor 86,400, which is the number ofseconds in a day. The values
in column D are easily added or subtracted in the analysis.
The analyst may use the remaining columns in the worksheet to code other functions that can be used as real-time
coding aids. For example, in a co.mpliance study, a function could indicate the active signal phase recorded in the
macro through the function {=lF(Allr"H","RED","")]. This function will return the statement "RED" if cell All
cont:~.ins the letter "H" and returns a blank cell
otherwise. During a busy li~ld study, this em help the malyst
double-check visually whether a key was pressed correctly.
r)
The "scan recording" and "stap recording" buttons are associated with subroutines in the VB Macro. Exhibit E-2
shom the components of the VB.
'!'~"...
~--
:::r
::np,:~
Sub StartRecoll"'l ()
.q
.Lppl1cae1on.OnXey "a", PrlntTiJDe.l"
.lp1l1cacion.onJCey "b", "Ptio.tTiiDeR"
..,. pE.:::!ld::...::Sc=ub"----------- -- -- - --
::~
- - - --
- - --
- - -'.
Sub StopReeorcliDq ()
Prln~Timei!O
...
:
Sub Start &cording, which is associated with the "St:lrt Recording~ control button in the MS Excel
worksheet. It assigns the print1imeA" subroutines to a specific button pressed by the user (in this case the::
letter":~."
or "A").
Sub SUJp &cording, which is associ:~.ted with the "Stop Recotding" control button in the MS Excel
worksheer. It disassociates the "PrintTime" assignmentS from the button press.
Sub Print1imd, which is the copy and paste procedure discussed above (in this case for the letter "a" or
"A"). It prints the selected letter, moves into the next column to the right, copies and pastes the time stamp
and moves back to the initial column and down one row.
All three subroutines need to be uscr-cusromiud to reflect wharever keys the analyst wishes to use in the study. In rhc oeample shown in Exhibit E-2 only keys "A" and "H" are defined and all remaining keys will simply rerum the key vall.l(. The
macro oould easily be enended to include mOSt keys on an ASQI keyboml, including all letters, numbezs and symbols.
Appendix E S631J
Symbol
In
tin
,ft
yd
ml
....
"'ml'
yd'
...
..
..,.._
"'
011
"'
yd'
211.4
"2.114 '
G.3
0-$
1.1.
--mbo
NIEA
.....,..._
tqt,l. . wrd.
--
..
I GilloN
""'"" ....
oobloywdo
....,
om
m
'"
.....
1m'
h
...
211.4
.....
......
....
VOLUME
JU
.....
0.~
o.oa
._,
.
ldlo-
... 1100.-'1
--1......
--....
rniiU..
......
..............
.--.
Symbol~
T.oFtnd
1Symbol!
""
mna-
0.00
..
....
G.4.
lnoh..
"'m
.......
foot
!em
....
""
"""
hi
.............
m...,.
1.1
110
to
NIEA
0.18
1U
1.2
0.4 . .
f\lfUit' I'MM'*
I _ . ..._._
-(100-..,
~-
o.m
N-
I
I
I
.......
...... ................
1--..1:=1--- jc
"C
vd
--
........
,;
"''
ftll
d'
mil
..........
...
poundo
lb
lb
~d-
ftoz
...........
VOLUME
mlllllilon
ml
.......
mR
MASS'"""'.........
2.2
ml
,,..
"'
2.1
1.1
In-
-..
---'"'
- - -.... -
-
--o.e
1......
to
...""
"''"
om
ltNI'EM'MIE 1 -
''f
I"-:ply I
lENGTH
2.1
MAss~h11
fq-
Symbol
....
Q.t
I poundo
RuWounoM
........
1-...
""
To Find
.......
.... ~--u
~~-
-- ....
---
1~1
LENGTH
"""'
.... "
.. ,._
ft01
..,pn>...,.,.C<>-..........10MolrioMoliour"
_..,.. _
0.03
2.1
.-
PI
.., .......
............
ue
o.a
q.....
. . . .et
,~
qt
lll l
ftll
yd' .
TEMPBIA'IUM 1-a
Jc.o..-
..--1 ..~
1~1-
.b.nn,mrfiY J:
1;1:,~
Bound on
Strtet
Weather
From
To
Numbers of
Passenger Vehicles
Including Taxicabs
Other Vehicle
Trucks
- - I
- -Dace
Recorder
Pedestrians
Traffic
Side of
Scrcct
Weather
Hours From
1/4 Hour Starting
To
IDbound
Outbound
---- -Due
Recorder
Dac~
on
Str~CI
Weather
Street
Compiled by
Street
Screec
Sc
l/4 .Hour
Stutins
Side
Ia
O.c
Side
In
Our
Sc
Total
lo
Ouc
Side
In
Out
Side
lo.
Ow
St
Sc
Total
"'
O.c
Side
"'
Out
Side
lo.
O.t
Total
lo.
Ouc
Side
Ia
Out
Side
Ia
Ow
Toed
ho
--
ou<
-To<al
A--a<
Pealt Hocu:
Appendix E 56-::;;ii'
Tttfficon --------------~-------
114 Hoou
Compil~bY------~--------------
I ..._,..,.
Vcbldar
Tracb
TtoAiit
Srreet
Nioc
Teool
= I
Total
Tracb
Tn..uit
MiK
Teal
lloo'::"
Traffic on
Weather
Compiled by
U4
s..ru..,
"""' ..........
Vdides
TNdoo
1iuoit
Mloc:
VcWda v.b;da
'WoiDoc
Toal
w.o-1
.._
Tooal
A--ac
r..k
"""'
Source: Box and Oppcnlandcr, 1976. ~- 40. Refer to Exhibit 4-4.
VdUdes
S trtet
PcooaaOuobowociV~a
T....D
Tn.uit
Vdoideo
Mite
Vd.ida
loal
Wall<la&
Toea! Ia
o...t.o...cl PluO.t
1:~ '
fjj)'r,;;r,c;,~, "")'z.,.J~""~n!' '
~ ,,, .. .,
.~Jii.!J.Jk" t;rJ!,i.f?u.;{,l\lu(.!:~jij;~~;,~-o/rofo'-!AJ.lr'~~ 1.1Sll1>'9$,r~.:; ;..~>~.
,.,._ ,;.~Th~,l-
, , ... ,..
~~ -~
,. -. - .
tO------
Oay- - - - - -
Date
Ob~rvt<------------------
:T IP
I
I
I
I
I
I
J
I
I
I
IP
I
I
I
II
I
I
IT
I
I
IT
II
I
I
'II
I
I
I
I
I
I
I
JT
II
I
I
I
I
I
I
IT
I
I
I
.I
I
I
I
I
I
-I
'0
t1>
"'a.
)(.
.
m
'
Route
MilepOsts
County
Weather--------------
Day
Date
Start Time
xrractlon Date - -- - - - - - - - -
End Time
TSR Unit
Diskette I
Length State
Speed Blns
Mat configuration:
TSR State
Director(-- - - -- - - - - -
Speed Stats - - - - - - -
Stream Stata - - - - - - - -
Counts
I_
I
I_
Lane _ _ _
Direction _ _
State ;..- - - - - -
Expanded M i l e p o s t s - - - - - - - - - - - - -
.Lane--- Direction - - - -
,_
lane---Olredlon - - - -
lane--- Diredlon - - -
Noted Problems - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
It Width
Bridges - -- - Unslgnalized - - - - - -
Slg.
Unsig. - -- -- - - -
Sig.
Unslg. - - - - - - - -
Left Side - - - -
HorlzontaiCUrvwture - - - - - - -- - - -- - -- - - - - - - - - - -
Terrain - - - - - - - - - - - - - - - - - - - - - - - - - _ . : . - - - - ,
OevelopmentType - - - - - - - - . . , . . . -- - - -- - -- -- -- --
om-Cha~ ---------~---------------
'
.-.
.- .
.
'
.
Data Collection Log
Removal
Deployment
i
I
Site ID
Route
Coonty
State
Milepost
Date
Collected
Driving
Time
----
Setup
Driving
Time
Retrl~a1
.:r~~~~~~~~:~::~:~~~~~!~:.:;:~~~~~~~~;~.;Ar;~~;~~~r
Darn CoUection Summary
Site ID
lAtta Typl
Speed
Limit
I I I
No. of No. of
Miltpo.t lues Ow.ys
Horitonlal
Tma.in
Cuswtun:
I Other
..
"d~ifil~~an.m
..
~.!~ :.,,_-
~-
-- ;.,lf_~'C'#-~!it.u.
- ~-- ~-~.c:, ., - ' _.. - ,-,-- ----.
il'i -. ~1@'k
:::'.~ . :. >;;. ,:;,<!<_'":;~~- ::. '\'
.r..,l}il~
_@l)jG,@I.~~tK:~
<-r
- "~-
........
*-
.. . - -
. ........ . . , ... .. .
-- ... .. .
Site Information
Genem//nformllllon
.Ii
......
#Oo
l n~<r.;-cJi on
Anitr'it
A;J!:ncy or t/Jmpaflj
Oi!e P<rf~ri!E~
Analysis Time P<ricd
a esc
Ae<a Typo-
~ C!h:r
Juri>dicti~
FM
_______
/npvlln/1/a/ Param111rs
193.9
Analy:isYm
_______________
-~
.,
Humberul h~E~ . Il
fr!.flow Sll!'d, FFS (milh)
SIJn;sycount in~nial, ~ (s)
;crri.in3. v.,
T<JEI ;<.hicl~s
40
15
C~le
::'-0
22~
lerg:h. C (.;)
...
Input FJeld Data
Hu~r of V:hidesln
Court n:er.sl
Cbck
C-;ele
Tim:
tlumber
,:)
4:1)~
4<42
+.47
.. .. .
Ou":u=
1l
1:3
:2
:2
16
,:)
It
...
14
-~
:0
:;,
:;,
:0
:2
..,.
13
0
0
12
i2
-~
r.
u;
37
Total
- Computations
--Totah-.!tldES In queue, lV01 =
64
Cf (fx. A1 S-2)
8C
'"-- - --G:
--l--..-
t4
10
= v
lll.fllh.:roft;tlessur\'E)cd. N.:
3iY
9,5
=ci.q ci.4
7.o
0.42
!.7
. :1.2
Appendix E 57S
I
I
.J
L~~--------------------~----~--------------------~-----------------
_______ BoundT~
~:
Unk--------------------------------~----------------------
G,.do
~:
~------------~-------------
w..tl>v. - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Oboe,._
I~
Oilier.--------------------Obo
8thv.h.
8111 ..11.
21 '
22
22
24
21
21
27
:za
21
10
30
11
31
12
22
13
sa
14
34
16
,.
17
t<ilwt..
10!1> wh.
37
38
1t
38
20
40
ColumnS
---lv!lhl
to Exhibit 6-4.
8111wh.
35
,.
~fer
411>-""'lllo wh.
lOCI! - .
1e
Source:
Tomo t - c l l l -n
No.
Cal
011
If
'
+ \' + '
lcl
lAS)
. ... .
..
~-
.....
'
;o.- .
.:- ..
.f
,.,.~
'
..
-.... '
""
"
..
..
Survey Dare
' Location
Crosswalk across
Computations
9
10
R + [W/3.5] + 2(N- l)
G c _ _ _.sec
11
l2
13
14
15
16
17
18
19
20
21
22
23
T=
total survey
rime x60
T=
_ _ _ sec
D=
t<T-l)rn x too
Ds
_ _ _%
24
25
26
r=
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
'
Totals
"'
~~.
:&OO'MJ1l.&>'~b
rJ~:i".JW'l;tfu1:~~~f.'i[~;furiij)~f'5
:Z," ~~">'::-\";,::.>;:;~~;;;."t.":..'t-?t:rf.;'l1J;:
..~'::.::?.:
.~~'J)
; .. . -
,_ .-
- .
.
,
'" _- __ ' .,., "
- --- -.
- .<> :;......
. . .:J
: ~ .
"'--~''
~-~
~-,-.
~--
Survey d:tte
Crosswalk surveyed
Crosswalk across
Divided Ro3dway?
Yes
No
5 or fewer
Number of rows
Tally
Total
Cumulative
Computations
6 [0 10
11 ro 15
16 to20
21 to25
26 tO 30
31 to 35
36 to 40
7
8
41 to 45
46 ro 50
10
loca~on
__________________________________________________________
Time
to
Stopped by Traffic
c:
0
ul
vi
z
Right__ _ .:....__
Straight
---
Straight
Voluntary Full Stop
Left
4
Right
Stopped by Traffic
Non-Stopping
c::
0
....
vi
z
----- - - - - - - - -
Date
Recorder _ _ _ _ _ _ __ _ _ _ _ _ _ __
Appendix E >1~
Area Type._ _ _ _ __ _ _ _ _ _ _ _ _ __ _
to _ _ _ VVeamer ____________ ______________ _
Observer
c
A
R
T
R
u
c
Source: Adapted from MokJrist Compliance with Srandltrr/ Traffic. Refer to Exhibit 8-5.
l~tion
_______________________________________________________________________
ilme
to
Weather --------------------------------------------
N.S.E.
,., ..... W
.. un
il!!
!"
!"
1eui!S padwnr
"
;:
&'
pa)j
"'
"
."en
60
!!.
.,..t
~
C'l
;;;
...
"
~
UllaJ!)
___.,._____
,
-- .
~
N. S.E.W
- - - - - - .. Ufl
~'""''-
Grl!i!n
j
,..~
;:
...:!.3
.."'
Q,
~!!.
,...f
~
C>
["'
"
"'
~
,1
Date _______________
:oaie"
Red
Jumped Signal
I
5
;:
...
.,;
'
~~------------------------------
,;~~1;1,~~iWiiJ~~~t.~ift.~~@Wi~&nur11i1~~::j;~~;;,'t.w!;-.~r~~i:"~;;
... n r c-
. ,.
.
..
,~-
Locarion
, Direction ofTravel
Time
Date
I
--~
, . ,.
- -
. . ., , .. ,,_.
- .
.. .... .
. . . . . . . .~
Area Type
ro
Indication
Wearher
Olcrver
Acci.o n
Cars
Trucks
Total Cars
Total Trucks
Turned on GorY
'
to rum on R, turned
onG
Full Stop
'
On~No Queue
No Stop
Full Smp
On Red Queue
Sropp~ by Cross Traffic
Toals
Notes: _____________________________________________________________________
Souoce: Adapr~ from Mowrisr Ormp/Uma with Sumtlttrri Traffic Control Drnca. FHWA-RD-89-103. ~fer to Exhibit 8-7.
~~}
; ,, ....,~~ -,l'.
:[: . . 1.,.
-: r
-e- 1 :,
....... .. - . . ,.
.::r..">-'~.. ~-........\
,."'
,._.
T~
Pedotllcr=in&
to
of
St. in
Oir~cdon
_ _ _ _ _ _ _ __ _ _ _ _
~============~~~----------~~~~
M;~s~~~hl,-----------lii____TI;o;,,.~,------
lcrosswalk}
"
~ I~
~
~ 1:p .
o!!
,S!
J ~~
"'
~jiiJ
I.
,!(
t;
-z l~
Croued
O~conlly
::::::::
"' 3:
hl~ j~t
Q
"
Total
O>ll _ _ _ _ _ _ __ __
Reoorder _ __ _ _ _ __ _ _ _ _ _
Appendix E ~ 3
Weacher
Route
Direction
Trip Started at
At
Trip Ended at
Trip #
(l..ocauon)
(Mileag<)
(l..ocation)
(Mllea&<l
At
Location
Tunc
Location
\l
1
C RA a
~t. IC"'\
Cr'IITinM
Type
Delay
(accond.s)
l:r.l;
.... . .
..
- ..
..
Wearhec
Route
Sw-rPoinr
E.nd Point
Ru.n
StartTtme
FuilibTame
Travel Tame
Vehides Met
Vehides
Ovuuking
Vehicles
hssed
V~des
Start Tame
FuilibTimc
TcavdT..me
VehidesMct
Ovutaldng
Vehicles
Passed
Bound
1
2
3
5
6
7
8
9
10
Total
Average
Ru.o
_ _ Bound
1
..
4
5
6
7
8
9
10
'
Total
Avenge
Comments
O~cvec
Location ______________________________________________________________________
Time
Weather _ _ _ _ _ _ __ _ _ _ _ _ _
tQ
Date _ _ _ _ __ _ _ _ _ _ _ __ _ _
Adults
Observer _ __ __ _ _ __ _ _ _ __ _
Children
tN
1:
c:
::2
::2
:E
11:I
:;
Street Name
....
E
z"'
~
VI
Adults
Children
1ime
Date
Walking Speed
Crossing Dimnce
Critical Gap
Tally
10
13
14
--'
--
17
19
20
21
-~I
15
16
18
Total
--
12
---
II
:..--
22
23
24
--
25
26
27
28
29
30
. 31
32
..-
33
34
35
36
37
38
--
'
39
40
Tow Adequau Gaps
Appendix E 547'
ceck.&
O S".&~..,.~-""""*'pol'"~ .ap..
.....)l.ery. . . ., .... . c.
[J No .~.s-m.. or ibo.Sdtu
0 Too JnOCh tnlic
'Ia;;;=====
c_
Loc:uior.ofptoblcmc
. , . ....
Rodac; (clodt on<)
t2S4S6
Ov..
ONo
c v...
ONo
ov..
ONo
ov..
ONo
0 Yo
IHMdbe"'""bJ41n.tn:~
0~.!'-ot~
-------
=~-:.~=:~:O:o~
Cro. 'Witb ctM ape?
OScvy~
0Non...UIIJII>oo4
Dlny, kltt crnn.te or ltrllllh
CJ Dlny olr 4., ., n Knabilt ftlUo!l
0 SoiiWthlna _ _ __ .
Loaodonootp-
0 Scnv ~ Dtlwn...
c--ot.-,....-~
C OWDOt",.W.to~~chl:~
.....
------ -
l . __
s.__
--
oo......-r..
c~,:..:-:-,
--
:t. _ _
T otal _ _
u .. u
, ,.. ,o
~~"""-->
I 2 ) .. I 6
lAhf\111/\f ()C
- ----
1 2a s'
0 Some~l..-.,rt~\ngl.:
0 NH4fd ,.oot SCU&. Oowtn. ot tNft
RAdacol<iodoOOM)
a...._ (clo<\t_,
CtlR
I ..
~bdi::ftc~JMO!
OS<uy ....
occpw~-.coei'Ot'
OY..
C'ftlaa~wwbr ~ ~oo~~lj
0 Somt pa>~~~.....
O~WIJKIO WW.
2 ) ..
'
R.dll&l (<irO.OM}
.,)456
_~
0 Sonw pfObl.ttb:
0
........
...
---
TO/\M('Qf'\QTATint.l
rl~lrltrrnltr
""'"' ,.....,,_,_. ,
""
0.
-. :
... . .
. '
..
'..
.....~~""#.'
''.t"':
...
~~
~Xt~~Jii~~~gn?fjj
\ Route
Block Number
Day
Date
Weather
Observer
Route Segment
Boardiog Passengers
Reduced
From
To
Full Fare
Reduced
Fare
Full -+
Transfer
Transfer
..
Transfer
Alll'assq
Dar~ - - -- - - - - -
a Arriving U>ad
Route
Number
I Direction
Wearher - - - -- - - - - -- - - - Observer
Block
Number
Vehicle
Capacity
Arriving Tune
Scheduled
Actual
Passengers
, ~~~t.~~(>~iji.~~;/ft:_',;;:;:i~J,ir,t),..- ,,p, .. v
.~,-;_,,'1"
.;1
.~,
''#o'
8l0 tK HUMBER
ROUTE NUMBER
OATE
DAY
WEATHER
08 SERV Eft
Lou:lo11
Oo
Pnsan4ers
Oil lu4
Ti..- Chtc11:
A :marlr.s
13-10.
Appendix E 591
o-:
M elhod:
TllpNumber.
Tllp lkan llrM:
TrlpEnd11""':
Weather:
"oute:
Olrection:
VehldoT~:
111
121
locellon
llmet
131
Time
corwol
point
-d
elo than
..
141
(51
(Ill
-d
O.loy
,nmo,
Stop
lime
than
welldng .
COUM
welldng
.w~c
olgnol,
Renwrb:
m
151-131
Deloy
lime
(MeJ
;~:5'1i11 ~ .
..
..
..
~.
~-
...
.,
..... ..
"
~~j.i:w.
o:,.. ?",
. ~'~~~
~.
,_ ....
\Dare
Ciry/Area
Stop
Number
Scop
Location
Arrival
Departute
!we
ll.IDC
Type of
Basiness
or Activity
Load
Factor for
Truck or
Trailer
Quantity
Type
After Scop
Notes
Jsr
2nd
~ 3rd
: 4th
5th
6th
7th
8th
9th
lOth
ll"th
12th
13th
14th
15th
16th
17th
18th
19th
20th
21st
22nd
23td
24th
25th
Comments
Driver
""''
..
~~~
;...-~!110 -~~~w~t-~~~r~ruful&?~~~:v~;-.;;xJ:~~~lffit~~JJ21J;f~~~~
Suee! Parking Tlltnover Su.mmazy Sheet
Dace
S~t
Weather_ __ _ __ _ _ _ __ _ _ _ _ _ __ _ _ _ _ _
1imc
Location
Facility
Block
Facility
Type
Number
of Stalls
Total
Tumover
Vehicles
Ho11tly
Turnover Rate
(vehid es/stallf
bour)
~mmcn~--------------------------------------------------------------------~----Observcr
Source: RdU 10 Exhibit 16-1I.
~~~~~a~
_' .!.,.
Parking Survey
The
\
Department ofTr:mspomuion is conducting an analysis of the patlcing conditions in chis area. We
are asking each driver to complete the following questionnaire. Your answers will be kept anonymous and data will be
summarized so that the analysis and report will in no way disclose data on individual!. We thank you in advonce for filling
out this brief survey. Your time is valuable, and the answers you give will help us improve parking in the area.
I. What is the primary purpose of your trip? (check only one)
OWork
Tourism/~creation
O School
0 Shopping
2. What was the specific location you visited immediately after parking your car here?
Street address or bwinw narne:
3. How long were you parked at chis locaLion?
hours
minules
4. Where did your trip start (where did you come from just prior to parking here)?
Nearest street intersection:
5. Please share any additional comments you have about parking in chis area;
Source:: ~er to Exhibit 16-12.
Appendix E 601
~iB;~~-.~ .:~~~~~~~~r
Parking lotervicw Sheet
Weather
Date
EndTtme
Start lime
Location
Trip Purpose
Tune
(Work, School,
Shopping.
Start (Parlred)
End {Left
Stall)
Trip
Destination
Personal,
Street adress
or Bo.siness
Other)
name
Hours
Miou~
Walking
DisWICC
'j
I
(
I
Comments
O~rvet
i'~~!)ibit E;_39. Traffic Conflict F,orm Wlth Or~ line Per,Tun~ P~rl~~ :;~t:~: ;.:. : ..
..
'!
.:
:i
;~,,-:.~} ~~~
. .. .
-rO
~-
Location
Leg Number(s)
Date
..
..
~---
i .i 1
"".,
Cl3 :::
'e
~
~
1
<
Observer(s)
C=Conflict
I;el\~1\111:
Same
Directio~
Right-
RightSlow
Lane Opposing Thm
Same Vehicle Change Left-Tum From
Right
Direction
Thm
RightThrough
Tum
FromFromFrom
R!ght
Right
Left
LeftThm
c sc c sc c sc c sc c sc c sc c sc
Total
C + SC
\
SC - Secondary Conflict
Daily Count
Rate Per 1,000 Veh
LeftTum
il
A ll
Through RightAll Sam~ Through
Thm
On
From- From
Direction CrossLeft
Red
Left
Traffic
sc c sc c sc c sc c sc c sc c sc
.'i';:;.::,titW~~
~~tr.v....;...r~~~;..-~'.i;!'!la~'#"'"'''- !'!!~~,,,,, "~'''"''"""*~;,!~<;,.~":r"':>-,;P,"-:-:.,~;,~1".f''
ACTOR CODES
ACTION CODES
Name:
Date:
Time Period:
Intersection:
Direction (leg
with actor I ):
Weather:
Time
Actor I
Action I Actor 2
Action
Comments
rnrn
rnrn
rnrn
rnrn
IIJQOO'
G.oQII'liPI'I0c1oclioon
Ftoo'
FJ~
Ft~r
"-'CS.nc tvP
,.,....,..
0 ~: :: ;;~
OMtp(Oc)f\ or .....,..,
ll.tldonc type
"'u' fl
""H
R~t'fiM
o~:: ~~:::r
Retklonc tviM
"'"'"
0!::::
~~:;.:,
us.
1TrnTrrn. rrcr1
O.aetipllo" o t
""d V t-$
lr iTT"itnTr TiTJ
IT.iTrrrnlrriTl
-""-'*""-... ,.,.. .
I I I I I I I I I I I I Ill
O..CtfpciOn ol U nd UM
tt~i>
o . .c ripUon ofLaMU c
~ Olhtt
,..
'
...
" ..~~~
- . ~
Origin-Destination S tudy
Field Sheet
Location
Station Number
Time: Begin:
\nbound
End:
Origin
Outbound
Oestinatfon
Route Used
Parking
Othe r
Date~
S"'""'-ZIOnltt,.li.laJI,..y
Loc:uiOOMIIII1)'Jie
Observez
Appendix E 6CP 5
8
~
i
d
:I
(I
,.
: !
.,'
..... .,.....
..----..------, ...------
!
~-
SoutCC:
:!,
"
-~ ~--u,nT <tTt..-. t
f"t
""~"'"
f"r"IITM"\fd
~or
'f-
Coo,.rau ...
Location
nme: Segln:
End:
Ucens
Number
Statlon Number
- - - - - - - - - - - - - - - Weather
Time
Truc::l<
or
8WJ
Out
ot
State
Dater._ _ _ _ _ _ _ _ _ __
ueense
Number
Time
Truck
or
Bus
OUt
or
State
?
~n~'---------------------
Aoot>ndix E 607
...
"'
. , ..t;
)>
z
c
)>
r-
.....
::0
)>
"'""'0
;D
E
5
z
5
6
9
10
22
23
24
25
26
27
28
29
0
m
0
=l
5z
IS
~SUM(G6:07)
06/08
ll6/E8
F6/F8
I
0
1
l
',.
. ,
1'01'.\L
~sUM(l>t.:li~J
s
..
06108
-SUM(07:117)
0
SUM(H6:H7)
-sv~
H6/H8
Tobit A
13
-"'
N
z
;D
Y'ES
12
.....
"'
c
zm
m
zGl
4S
SUM('E6:E1}
IS
30
SUM(F6:F7)
. ,_
11
;;;
Gl
0
60
-sUM(06:07)
..
7
2S
''
lncr<8llng Proponions?
'-'~ ~
Auumc lhat 1ho followi~_i_l*_E_etCCC1 tancc dam set wilh 1 2as:ecc>nd interval was eol1eeted:
, I
Gop/Loa Slu ( soc)
J
s
14
15
16
17
18
19
20
21
"'
z
30
31
32
33
34
Colnmn N6.:
Criricol Cop (scc):
Atplcd Gop (se<)
0
0
-DI71(100-SDSI6)'100 0
018/(IOI)..SD$16)"100 ..0 I !!I(100-SD$17-SOS I6}" 100
-DI91{100-$DSI6)"100 -1>19/(1 OO-SDS17-SOSI6)"1 00
-G81S1S8' 100
HS/SI$8' 100
D2Difl 00-SD$1 6)0 I00 cOlO/( I OO-S0$17-$0SI6)'1 00
-sUM(D 16:0 20) SUM(E16:1l10)
~SUM(fi6:F20)
08/SISI'IIJO
c8/SI$8'1 00
F8/S1S8"100
3
5
7
9
TOTAL(''!.)
0
0
0
..
s
8
.,
..
0
0
0
DI9/(IOO-SUM(SDSI6:SDSI8))"100
0201(100-SUM(SDS 16:SDSISll'l 00
Sll~ I(G 16:020)
0
0201(1 00-SUM(SDS16:$DS 19))'100
StJM(HI6:H?O)
Table U
Colmh Number
Crlrlul Gop (Itt)
Atuprod Gop (sro\
3
5
7
'TOTAl..
Cri<l<ol Gap v,
M .. n Crillcol Gop (,.c)
1
2
E6
EIS/100'0532
B19/IOO"OSJ2
E2lYIOO"DS32
028" 100/17
Dll/SRSJ2
0
Hl9-029
0
0
f00-SUM(D30:.EJO)
-(OlVV 00)' F532
E6
0
0
F6
-<l6
-fi31-SUM(D31 :Fll)
H6
-GJI'IOO/HlO
-G321SHSJ2
SUI'-t(lll8:HJ Q
"'(FI9/IOP)'E$32
i"'(FlD/lOO)'E$32
l!1t"100/F18
IJ.l/SI!$32
-(Dll'P26+E32'
F30'100/GI9
Total
lf.\ l/(11$32
'
.
'\
'" '"
In dEx
.. . . .
"
......................
A
MDT. S,c average annual daily
.................
traffic
.....
National Standard
Practice for Roadway Lighting, 41 5
analogy, 475
.~
.I
.,i
~;
..
.. ;
:;
296f
urival volumes, inrersecrion counts
and,45. 45f
Index 609
.
'
!ij.!
j!
~i
u
'
it
. I
IJ
~!
r:
'
ii
..
i
i.
!
''
automatic coums
for bicycles, 242-245
data reduction for, 68
for Aow, 183
freeways and, I ~Sf
for pedesuians, 242-245
periods for, 69
for public transportation, 281-283
fortesr vehicles, 165
for TID, 161, 165
for volume data collection, 62-66
automatic passenger counting
(APC), 272, 281
ITS and, 281
automatic traffic recorder (ATR),
192
automatic vehicle identification
(AVI), 173
automatic vehicle location (AVL),
281-282
GPSand, 281
ITS and, 281
uavel time variability from, 282f
visualization with, 282f
pavement markings. s~~ also
cemerline markings
compliance with,_144
at crosswalks, 388
reaordlecrivity of, 139-140, 139t
reuorellecrometer for, 139f
average, 522.
mobility, 197
for spot speeds, 523f
average annual daily .traffic
(AAD1},49,57, 192
.collisions and, 355
ddinition of, 9
Avcrage-CarTcchnique, 161
8
ball-bank indicator, 407-408, 407f
advisory speed and, 409t, 411-412,
411!
centerline markings and, 139
definition of, 9
bar graphs, 27t
grouped, 28, 28f
stacked, 29, 29f
for volume counts, 71f
beacon, 10
before-and-after test, 92, 93t
analysis for, 492
for bicycles, 255
comparison in, 494
with conuols, 492-494, 493f
drawbacks to, 489-490
in experiments, 489-494
hismry in, 489, 491
for lighting, 415
maturation in, 489,491
overcoming drawbacks, 490-492
for pedesuians, 255
forTCD, I36
units in, 489, 493
warm up period in, 491
benefit-cost ratio
for councermeasures, 374t, 377379, 378t
definition of, 10
i:
Bureau ofU:nsus, 450, 508t
CF. &(car-following
CFCs. &( chlorolluorocarbons
CFS. S(( Commodity Flow Survey
c
CAD. s~e computer-aided design
calibration, 205
ddiniuon of, 10
of inputs, 223-227
for lasers, 85
for radar, 85
from screen-line ~unts, 54
forSLM, 455
for spot speed, 88
with video-base counts, 65
capacity, 479-480
definicion of, 10
charts, 27c
area, 544, 544f
Row diagrams, 548, 548f
high-low graphs, 543, 544f
organization, 547, 547f
picrograms, 545, 545f
project progress, 549, 549f
scatistical maps, 546, 546f
cypcs of. 543-549
checkers
data collection and, 264, 272
definition of, 10
lice~-placc marching by, 334
for public uanspomtion, 280
surveys by, 284
chcckiiscs
bikeability, 252
for graphics, 550
for prescntacions, 558-559
quescions, 506
for transportation studies, 6t
walkability, 252
TTDand, 161
chord
for curve radii, 408-410
definition of, I 0
CL S~t conlidence interval
classificaciorts
a.rea. 419-420
for areas, 419-420
for bicycles, 419
counts, 54
ddinicion of, 10
of inventories, 316
for mobility and aecess, 43 5t
for pedesuians, 419
for road sumces, 41St
of streets, 312f. 418,419,435,
436f
of vehicles, 185-186, 185f
r.
;.
,;
..
clipart, 27t
t/
(.
'
I
'
~
,.
unreported, 356
volume and, 354
':~ .
'
;:-,
:;I
-.
...I
- .. i
collision races, 11 9
~I
: i
I
;I
~~I
. :[
. '
:i
:l
:_..:
!
?
~i
-..!i
:.~
!l
:i
;.
j
;
'
'
..L
j,
i!
'
.;:
comparison
in before-and-after rest, 494
in experiments; 487-438
history and, 494
maturation and, 494
regression to the mean and, 494
units and, 494
compatibilicy, principle of, 27
....
for, 131
copyright norice, on written
reports, 551
cordon count, 50-53
accumulation computations for, 53r
ADTand,50
definition of. 11
elW!lple of, 53f
OD and, 50
summary sh~t for, 51f-52f
cordon line, 431,438,449
cosine error
crosswalks
complil'l~e
wirh, 252
definition of. 12
pavemera marking at, 388
pedestrians and, 239f
QOSand, 253
unsignalized inrersecrions and, 252
volume and, 238
definitioH or, 11
countermeasures
benefic-cost rario for, 374t,
377- 379, 37Bt
for collisions, 369- 3 79
definition of, 11
e\aluarion of, 379
cycle length
definition of, 12
for signali:z.ed intersections, 44
simulation and, 202
TID and, 99
D
daily factors, in control councs,
55- 56, 56t
daca collection
for bicycles, 253-256
checkers and, 264, 272
for collisions, 348- 357
crime and, 8
forms, 564
for freeways, 191-195
individual vehicle selection method
for, 79-84
for inventories, 318-320
for lighting, 415-424
for multi-use paths, 241 f
for parking, 332-339
for pedestrians, 253-256
pitfalls o( 8 point data, 221-222
for public transportation, 272-285,
273f
i :
:l
.I
'!
'
j.
I
'
I
~
;!
ft
!!
l
..,,
;~
:1
~
u
p
l!
.f:r.
definition of, 12
demand
diagrams, 27c
:rr.-~:1>-;-
i:'Zi:.:~~
design
of experiments, 485-496
factorial, 494-496
of ~phics, 25-34, 540-550
for interchanges, 203
principles of. 26
of surveys, 499-514
of tables, 542-543
~i
:i
il,.
ll
ll
I~
I
I
i.
design spttd
advisory speed and, 406
definicion of. 12
design equation method and, 410
on freeways, 178
stopping sight <llitances and, 114t
YIELD signs and, 114
ti
!'
I
dilemma zone
I.
definicion of, 12
laser and, 79-82
radar and, 79-82
dererminisric, 205
definition of, 12
rraffic, 219
transportation, 463, 481
volumes, 223-224
aerial surveys for, 187
definition of. 12
detector occupancy and, 187
on freeways, 187, I 88f
LOS and, 187, 188f
occupancy and, 187. . .,...
UAVs for, 187 ! \ ~~~
Doppler effect
definition of. 1i
lighting and, 370
traffic
disuibucion
~ngineering judgment
Evaluating lntnucticn
improvnnmt.t: An Enginuring Study
Guidt(NCHRP), 208
Excel,536
expressways, 419
experiments
before-and-after rest in, 489-494
comparison in, 487-488
design of, 485-496
inferential statistics wim, 486-487
paired comparison in, 488, 488t
random assignment in, 487
unpllied comparisons in, 487
172
s~~ evacuation
response curve
Floating-urTc:chnique, 161
flow
automatic couna for, 183
definition of. 13
freeways and, ~ 82-183
HCMand, 182
speed and, 183
vph and, 182
FABCO, 351-352
ERC.
facton, 486
l
]
I
I
I
;I
I
.I
of parking.
330-332
definition of, 12
field procedures for, l 02- I 03
graphics
-~
ch: ckliscs for, 550
communicalion with, 2S-34
com posilion of. 541-542
design of, 25- 34, 540-550
foc:t.l poinrs i.n. 540- 542, 541
visual weight in, 541f
..--
sysr~ m
Ad ministration
H
handheld counc boards, 59-60
for bicycles, 239
for gaps, 110, 123
for pedestri:ws, 239
for TCD compliance, I46
hi~ wry
incidents
definition of. 14
freeways and, 190
index, in wr;tten reports, 552
indirect measurements, of speed, 79
individual vehicle selection method,
for d:ua collection, 79-84
inferential statistics, 92, 528-535
estimation in, 528-529
wirh experiments, 486-487
nonparameuic tests in, 533-535
proportions in, 531
run pies and, 528-53 I
significance resting in, S32- 533
informative changes, principle of. 27
inputs, 205
calibration of. 223-227
simulation and, 228r
:r.
:'ji
l; l
impact, 463-483
"
;~
I;
~ j~:
interchanges
definicion of, 14
design for, 203
diverging diamond, 49
expressways and, 419
overpasses and, 194
single-point urban, 49
weaving segments and, 192
-e'
j:r.,.
;~tl
' ,J.
:i~
,:
.,,
"
!' '' 1
; ,~
:li
:/r-1
-!r1
,;,
~ .
~ H~i
-! il j
i:J'j
lid
:- ;f. ~
;,,i(iI ~-<!
.]f'i
~ ~. ~
ifl
I
:n:i
:~1,,.:
H~~1
;I~
i iI ~
'f
.
-: fi
~
'
!ri
; i t-1
,ii.J
: ,,1
;fj
li
I
?'-;,i
.1,
' l I
.j
t!
;:
'0:
i>
~.
i':
:
i
L
1-
f:
f:.
i'
'
~r~
h_:
::-3
r.
:l :
-1
for superscreeu, 47
at unsignaliz.c:d intenections, 44
intersection sight distance (ISO),
112- 11 5
MSHTO and, 115
measurementS for, 113f
STOP signsand, 114-115
TCD and, 135
time gaps and, 114-115
YIELD signs and, 114
interviews
dwelling-unit, 447
for parking. 336--339, 338f
su~and, SIO
K
KABCO, 351-352
collision severity and, 15
key counts, 55
L
lag. 15, 110
land use, inventories for, 432-434
la.ne-changing, 225-226
algorithms for, 15, 225
definicion of, 15
traffic c.onftica and, 394
laptop computers, 60
for bicycles, 239
for gaps, 110
intersections and, 98
for pedeStrians, 239
for saturation How, 105
forTCD compliance, 146
for test vehicles, 162
la.sers
calibration for, 85
cosine error for, 80, 8 II
definition of, 15
Doppler effect and, 79-82
lSD and, 112
round off error for, 80
for spot speed, 84-86
Luin square, 496
learning elfect, 210
level, 15
Engineers
iteration, 13, 229
ITS. Stt intelligent transportation
systems
J
jay-W21ki.ng, 13, 155, 253
judgfnentsampung, 504-505
license-place matching
by cbcckers, 334
for parking. 332-336, 333f
for path-based councs, 48
M
macroscopic models, 202-203
definicion of, 15
for queue length, 105
magnitude, of noise, 452-453
R~trort!foctivir:y,
Maximum-CarTcchniquc, 161
mean, 91
arithmetic, 523--524
ccnrraltcndency and, 522
definition of. 15
MOE and, 492
public tr:tnsporcation and, 288
str:tci.fied r:tndom sampling and, 502
ML.
s~~ managed
lanes
middle ordinate
for
radii, 408-410
definition of, 15
midlock, 50
marwacion
in before-and-after test, 489, 491
comparison and, 494
MOE 211d, 489
minor sneet
approaches of, 125
definicion of, 15
peak bour wunnt and., 127
:.
: :~
'~~
..
'
:,
1:
;
:.
I
:i
:i
:!
'
';i
"
:!
cum
mobility
averages, 197
classification for, 435r
median, 91,524
median U-curn. Su Michigan
U-rum intersection
MOBILE6, 459
,,I
r:
:
j:
.'
!l
..; !~
I
'.
!i
;! i
..
I I
, I'
'i
N
NAICS. See North American
Industry Classification System
:i
i
.i
~ .
'
..,'
.,
,.
~~
~~tiona!
:I
~.
I!
~
~.
~
'~
..
ji
l
!Iil
!i
'I
...Ji1,
. r.
ji
I!
networks, 205
bicycles wd, 252-253
definicion of. 16
mosuresof,223
pedestrians and, 252-253
$Creening of, 359-369
$imulacion and, 220f
warnuus for, 132
ll
'
. j.'I
,.J!
,:
.,
il
11
l,;
'I!
!! ~
1:
!l
t:.
ti
:i
'Ji
~
,,'i
I
noncoverage, 514
not-at-home, 514
nonattainment area
air quality and, 457
definicion of. 16
i'
0 -D. Su origin-destination
,,
'I'
no stop, 149
0
occupancy
definicion of, 16
operating speed
algorithm for, 227
definicion of. 16
$Unulation and, 227
validation and, 227
opcimiz.ation
algorithm, 16
delinition of, 16
simuh.cion and, 208
steady-stare, 210
oral presentations, 555
ordinal scale, 507
organiz.ation charts, 547, 547f
origin-destination (0-D)
cordon count and, 50
da.t2 c:xpa.ruion for, 443t
definicion of. 16
field $heet for, 440f
parking and, 326
path-based counts and, 47-48
peak hour and, 47
screen-line councr and, 54, 449f
simuh.cion and, 212
survcys,437-438
TAZs and, 432f
transportation planning and, 437,
448-449
ortho-recrilication
dcfirtition of. 16
with video-base counts, 65
outputs, 205
simuLa!ion and, 228t
372t
p
pace speed, 16, 91
paired cowparison, in experiments,
488, 488c
paratransit, 16, 264
parking, 323-343
accumulation ~cudic:s of, 330-332,
332f
block and curb face numbering
system for, 326f
forCBD, 325
congestion and. 325, 331
curb, 326-327
dara collection for, 332- 339
data reduction and analysis for, .
339-340
data tabulation form for, 341f-343f
duration summary shecc for, 335f
generation srudies of, 330-332
interviews for, 336-339, 338f
invenrories, 325-332
license-plate matching for, 332336,333
on major street, 331
occupangand, 330, 332,340f
0-D and, 326
online m<tpping cools for, 326
pom:uds for, 336-337
space counting for, 339
srrat~gies for, 344
summary form for, 328f
surveys for, 337- 338, 337f
transportation planning and,
436-437
for trUCks, 292 '
tutnover summary sheer for, 336f
usage of, 328r, 329-330, 329r
passenger-car equivalent (PCE)
definition of, 16
HCMand, 185
passenger load. Stt load
path-based couna, 46
de6nirion of, 16
pedometers, 242
percentile
definicion of, 17
gaps and, 110
of speed, 520
for spot spds, 523f
percentile speed, 85 th, 84, 84t,
91-92, 520
definition of, 17
peak hour warrant and, 127
YIELD signs and, 133
percent of congested travel, 197
percent variation, 197
Perception, Identification,
Emotion, and Volition (PIEV), 134
perceptual organization, principle
of, 26
perflurocarbons (PFCs), 457
perfocmance measure, 17. Se~ also
specific pcrformanu mtas11m
permissive turn
definicion of, 17
at signalized intersections, 45
permicted error, 83
pcrson-houcs per year, 197
person injury, in collisions, 351352
person occupancy, 184
ofHOV,187
MOEand, 187
PFCs. Se~ pcrflurocarbons
Index 621
i
~
i;
i
!!
I
i
j
1
l
;
g
'~
,.
~
~
~
I)
trucking, 292
CDD and, 300
loading and, 299-302
routes for, 298
surveys for, 448
unloading and, 299-302
weight and dimension !imics for,
297t, 302-304
WlM and, 303-304, 303t
vehicle.~
user perception
definition of, 21
<{05,252,253,25~256
video, 60-61, 65
at approachc, 65
for imenections, 98-99
for saturation Bow, I 05
signal phase and, 255
for speed decemiliutioo, 82
for speed craps, 85-86
with test vchiclc, 161
;~
'I
TID.
'!
).
r-eese, 488
'I
u
UAVs. S" unmanned aerial vehicles
i
l
l
i
1
1
}
I
~
r,
l.Ii
!i
!I
v
validation, 205
definition of, 21
operating speed and, 217
for outputs, 217-228
variabilicy, 526-528
v/c. s(( volwne-to-capacicy ratio
vehicles. Su also probe vehicle; test
vehicles
classification of. 18~186, 185f
conflict and, 251-252
definition of, 21
emissions from, 217-218, 217f
on freeways, 185-186, 185f
heavy, 18~186
pedestrians and, 251-252
routes by, 219
types of. 219
vehicle identification number
(VlN), 352
vehicle intercept method, 446
TID, 172
w
walkability checklists, 252, 257f
walking speed, by pedestrians, 246
walkways, for pedestrians, 252, 419
warm up pedod, in before-andafter test, 491
warning signs, 118
definition of. 21
placemenr of, 134
warrants
. for collision severity, 131
for coordinated signal system, 131
for crash, 131
definition of. 21
eight-hour vehicular volume,
124-125, 125t
engineering judgment and, 21, 123
y
{
<