Professional Documents
Culture Documents
Volume Studies: By: Robertson, PH.D., P.E. Hummer, P.E
Volume Studies: By: Robertson, PH.D., P.E. Hummer, P.E
Volume Studies: By: Robertson, PH.D., P.E. Hummer, P.E
o o • t c • • • • • • o • o o o • o • • o • • • o • o • o • • • o o o t o • • • o o o o • o • o • o • o o • o o o o • a • o • o ~ 1 • 0 • 0 0 • 0 , 0 , , 0 • • o
0 0
Volume Studies
Original by:
H. Douglas Robertson, Ph.D., P.E.
joseph E. Hummer, Ph.D., P.E.
EJ;uJby:
&stian]. Schroeder, Ph.D.
1.0 INTRODUCTION 43
2.0 lYPES OF STUDIES 44
2.1 Intersection Counts 44
2.2 Area Counts 49
3.0 METHODS OF DATA COLLECTION 58
3.1 Manual Observation sa
3.2 Automatic Counti 62
4.0 DATA REDUCTION AND ANALYSIS 67
4.1 Manual Counts 67
4.2 Automatic Counts 68
4.3 Count Periods 69
4.4 Volume Data Presentations 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 :re-
mains 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 e£fi-
ciency 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.
Except for very light traffic conditions, intersection counts require multlplc observers. If many vehicle classes arc to
be examined at a busy intersection with several simultaneous movements, each observer muse be able to record data
for rwo or three lanes. Simplified methods of identifying vehicle classes arc sometimes desirable. For example, one
could classil}r ~ motor vehicles with two to four tires as automobiles and ~ motor vehicles with six or more ti~
as trucks. The classification scheme must be well understood by all observers before the beginning of the count.
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 yield-
controlled intersections, a second observer may be necessary, depending on me volume levels on the major approaches.
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, me counts should be rep-
resenta~C of the demand, despite the variations in timing (Roess, Prassas and McShane 2004).
Arrival volumes are not easy to observe, since the queu.es arc constantly changing. and may extend beyond tlte line of
vision of the obseiver. Additional observers arc normally needed to count queue lengths. while the primary observers
tount departure volumes. For greateSt accuracy. the queue can be counted every cycle or may be counted at the end
of each aggregation interval (for example, 15 min.). If only a single observer is available, a video camera can be used
to capture depanurc volumes, while the data collector manually records vehicle queues. [n this case, it is important to
find a synching point to be able to combine daca.sowces in the office.
Exhibit 4-1 illustrates how to estimate arrival volumes by observing departures and queue lengths. One can calculate
the arrival count for each interval by adding the net change in queue length to the observed departure count. Note
from the example that while the total departure and arrival volumes arc the same, the distribution of volumes across
counting intervals (or cycles) is different. This procedure es~ates arrival volume for the approach. Turning move-
~ents for these arrival volumes may be 'obtained by assuming that arrivals follow the same distribution of left, right
and.through movements as the.departure volumes.
Volume Studies • 45
2.1.3 Petknritm and Bicyck Counts
Pedestrian counts are usually taken at intersection crosswalks, ar midblock crossings, or along sidewalks or walking
paths. Pedestrian Aows are also common at modern roundabouts in urban or suburban settings. Similarly, bicycle
counts can be imporcant for planning applications and operational an2lyses, especially in areas where bicycle usage
for .recreational and commuting purposes arc common. Pedew-i~n md bicycle volume data are used for rraffic signal
and crossw2lk warrant studies, for capacity analysis, in collision studies, for site impact analysis and in other planning
applications.
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 intersec-
tions. 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 side-
walks for certain maneuvers. Chapter 12 describes procedures for performing pedestrian and bicycle volume counts.
Manual countS at roundabouts are possible if t:.raffic volumes are low or if multiple observers are used. The cognitive
task is more complicated from a regular intersection count, since the observer needs to remember the vehicle's origin
when he or she tallies the count as the vehicle exirs. Even if an analyst were co focus on only a single approach, this
method may result in multiple vehicles within the circle ac any given time. The potential for ector is grcaL To success-
fully perform a turning movement count, the observer(s) should be positioned at good vantage points.
To enhance the accuracy of a manu21 roundabout count, video observations can be wed. Given an appropriate van-
tage point that caprures the entire roundabout, an analyst can count different approaches to the roundabout by watch-
ing the video repeatedly. A national study (R.odegc.rts et al., 2007) on roundabout operations has wed this approach,
aided by the we of multiple cameras mounted on a telescoping pole in the center island of the circle. Others have
estimated roundabout turning movementS wing multiple time-synchronized data collection units (list et 21., 2006).
To further facilitare the manu21 count of roundabout volumes, manufacturers of data collection equipment are begin-
ning to include roundabout analysis features UAMAR 2008). But even with erthanced data collection equipment and
video observations, a manual roundabout count remains a challenge.
Given the challenge of counting roundabout flows, recent technologies and research have increasingly focused on au-
tomated roundabout data collection using video-image processing techniques. This approach requires a good vantage
point for video observations and further that the image is calibrated. Video-based roundabout counts are already be-
ing offered by some private companies and are expected to grow as roundabout insrallations become more frequent.
The median U-mrn allows through movements along the major and minor streets to operate within the normal in-
tersection zone; however, ALL left turn movementS are required to utilize one of the U-turn openings to complete che
movement. Many of these inte.rsections arc located in the state of Michigan, which has led to many people calling rh is
intersection configuration a "Michigan Left."Thesuperstreet divertS minor left AND through movements by maki ng
drive.rs access one of the U-turn openings. Typically, the main street movements are allowed to operate in a normal
fashion, because the additional left turn signal phase does not hinder the opposing through movemenrs. This is be-
cause the two main street through movements act as one-way pairs on each side of the l"{ledian, allowin g ne:u perfect
progression that is nor possible with two-way progression at a rypical intersection. However, at times it is not possible
to allow left turn movements due to issues that may arise such as right-of-way restrictions.
Counrs at these intersections are even more chall.enging than at a roundabout, because of the larger spatial extent of
the intersection. Another key difference to the roundabout, is that not all counts are path-based. Luckily, the major
movements at a superstreet (typically) operate like they would at a regular intersection. They can thus be counted
using any of the manual or automated study methods discussed in this chapter. The challenge lies with the mipor
street movements. Depending.on the volumes of these movements, one or multiple observers in the field may be able
to perform a manual couot to capture t,hese Hows. At greater volumes, video observations may be used to allow for
repeated viewing and detailed count of the different movementS. Researchers bave found that video cameras posi-
tioned at the locations shown in Exhibit 4-2 can be sufficient, although additional cameras may be necessary if visual
occlusion through landscaping and signage is a concern. An overhead van rage point i.s generally helpful and preferable
if available.
To develop an 0 -D matrix of turning percentages, the observer(s) samples the puning movement percentages for' '3.
short period and then applies these percentages to a general appro~ch volume couot. Depending on the BuauacioOS
in traffic, a 15-min. sample per approach may be sufficient, supplemented by a 2-hour approach count. To impro-ve
the accuracy of the 0-D matrix, an average of multiple sampling periods can be used. If traffic volumes are high, a
sampling rate can be used to record the path of every fifth vehicle (or greater). If rate sampling is used, the sampling
period should be extended longer than five min.
Because of differences in rime-of-day crave! patterns, different 0-D matrices should be estimated for the Ml, noo.J'l,
and PM peak hour, or any other time period used in the analysis. Exhibit 4-3 illuscrates the computation. ln rhe tP'--
ample, the average 15-min. 0-D macrix was obtained by manually counting approach turning movementS for.each
approach. which were then used to estimate turning percentages. The approach demand was obtained by simply tally-
ing the vehicles on the approach at a location upstream of the intersection. Using the oudined method, a roundaboi.J. t,
for example, could be couoted by a simple observer, who samples the 0-D matrix at different times of the day. For
increased efficiency, the approach volumes are then obtained by automatic data collection equipment as discussed lac-=:r
in this chapter. It is important that while the 0-D macrix in Exhibit 4-3 was sampled for only 15 min., the approach
volumes represent a count performed over the full analysis period.
a)
~ ~
····i·············)oo
0
~-···········t····
C:*
~ .............!..··· ·····%···· .........)-
... +
b)
~ ...... 1
l••••t··············>-
C>>e '
at::)
......)o
The .sampling method is appliable to all intersections that require path-based counr:s. The 6dd observer lim esti-
mates 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 path-
based count applications in the future.
U-Turn 9 6% . 37
Left 27 19% I ll
North Through 75 53% 578 307
Right 30 21% 123
Total 141 100% 578
U-Tum 15 6% 59
Left 33 14% 130
East . Through !56 64% 958 615
'
Right 39 16% 154
Total 243 100% 958
U-Tum 0 0% 0
Left 21 24% 80
South Through 42 48% 331 • 160
Riglu 24 28% 91
Total .. ff7 100% 331
U-Turn 6 4% 24
Left 24 14% 95
West Through 60 35% 678 238
Right 81 47% 321
Total 171 100% 678
Volume Studies • 49
2.2.1 Cordon Counts
Agencies make :1. cordon coum by encircling m area such as a central business district (CBD) or other major activity
center with an imaginary boundary md counting vehicles and pedestrians at all of the points where srreets cross the
cordon. Observers classify each vehicle by type, direction of travel and occupmcy and typically use 15- ro 60-min.
intervals. The counts show the amount of traffic entering or leaving and ehable an estimation of the vehicle and per-
son accumulations within the area. ·
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 trav-
eled 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 accumula-
tion. 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.
Numbers of
Passenger Vehicles Other Vehicle
1/4 Hour Starting Including Taxicabs Trucks Excl. Bus &; Str. Can
---
---
Date Recorder
Pedestrians
Traffic Side of Street
Weather Hours From To
- --
- --
---
Da.te Recorder
i Street
Compiled by Street
St St St St
i/4Hour Side Sick 'IOoal Side Side Tooal Side Sick Tooal Side Side Total
Sw1lng Ia 0... lo 0... Ia o.. lo o.. lo 0... lo o.. Ia Oal Ia o•• Ia o.. lo Oat lo o.,, lo Oac
: .
Total
A--ae
..... }loot
Total
Average
Peak Hoa.r
'
All Penon SWIUIW'f Sheet
Date Traffic on S=r
Weather Compiled by
J/4
-..lahoaadv. ,.......o..m.....dV"oa
Hoa.r
SanloJ
..........
v.blda Tracb
Tamil
WUda
1\Wc
VCiolda
Toal
Wo.IJdac lahoood
PuocDp
Ve!Uda Tta<b
Ttu.oia
VCiolda
Mloc
WUd. Wolkios
Teal TocaiiD
Owl>oa...t PluO..I
y-a
~
.....
Hoat
Volume Studies • 53
2.2.2 Screen Line Counts
Screen line counts are made to record travel from one area to another. While a cordon count is used to count all traf-
fic entering and leaving an area (such as a central business district), a screen liQe count is used to capture traffic flows
from one area to another area. The screen line is some form of natural or human-made barrier with a limited number
of crossing points where volumes are counted. Examples include rivers, railroad lines, or urban freeways with a limited
number of crossing points. Analysts use screen line counts to check and adjust the results of 0-D studies or to validate
traffic distribution results of a rransponation planning study (see Chapter 20). They may also be used tO detect trends
or long-term changes in land use, commercial activity and travel patterns.
For screen line counts, hourly intervals for a 12- co 24-hour period are usually used on a weekday. Using several count-
ing periods that arc I week or more apace will preclude any unusual conditions occurring on a given day from causing
a bias in the data. Classification counts may also be desirable when conducting a screen line count. Upon completion
of the count, the screen line crossings (hourly or total) are compared co the crossings predicted by uansporcation plan·
ning srudies discussed in Chapter 20. The result of this comparison is then used to adjust the ~pottarion planning
model chat predicts 0-D Bows. In other words, the screen line study results are used to calibrau chc planning modd.
Exhibit 4-7 illustrates the layout of screen lines in a regional screen line count study. Another screen line example and
display of traffic accumulation over rime is given in Chapter 20.
A
LAgtnd
_ .......,.~
c:J.__
~ .......... ......
~
pr-..q AUMu.nt~•el•""""ar,an
eut 111 ,, a
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 supple-
ment the data obtained from permanenc-count stations to obtain estimates of seasonal and monthly volume variations
at addicionallocarions in the transportation network.
Control-count sracions are distributed across the cransporcation network and placed at strategic locations. They can
be divided into major and minor concrol·couar 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
'.i the year. At minor control-count stations, a 5-<iay weekday continuous count is made every other month. There are
;i
:j S4 • MANUAl OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDmON
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.
where
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).
where
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 di-
rectly 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:
y
PHF=~ Equation 4-4
4*V, •.•m
where:
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) .
Volume Studies • 57
3.0 METHODS OF DATA COLLECTION
The rwo basic methods of counting traffic are manual observation and automatic counts. For the purpose of this chap·
ter, manual observations will be defined as any count where individual vehicles are tallied by an analyse either during
field observations or from video recordings. By conuast, automatic counts use automated technology to ~rform the
count, reducing the analysts' taSks to set-up, calibration and analysis. It is important to note both rypes of counts can
be performed in the field or during pose-processing in che office.
Many types of counts require classifications that are obtained more easily and accurately with trained observers in a
manual count. Examples include vehicle occupancy, pedesuians, intersection turning movements and vehicle clas-
sifications. While most of these measures can also be obtained by automatic means, the additional set-up effon is
typically not time-efficient for shon-duration counts.
The main reasons for conduCting manual counts are time and resources. Practical applications often require less than
10 hours of data at any given location, with most counts focusing on peak-hour conditions. Thus the'effort and
expense co sec up and remove automated equipment is not justified. While there will always be a need for manual
counting, advances in technology and the ability to use existing equipment for counts (for Cll:ample, loops or cameras
at signalized intersections) are shifting the balance away from manual and towards automatic methods.
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 bi-
cycle 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 of-
fice. 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.
p p
~·
p T T IT
~
I I
I I
.,
I
I
I
I
I
I
I
I
I
I
I ,.
I
I I
--
I
p . IT I ..1. ........ jp IT
I
r
p IT - - -I T I
I I I
I I I
I I I
I I I
-· . I
I I I
p I I p IT
I I
I
I
I
1
I II
I
I I I
I
+
I
I
lp_ ;-.:
,
I
p I p T
r
!T
I I
I
I
;' I
.I
~
I I
I I
I I
I I I
I
Electronic count boards are compact, lightweight, handheld computers with di.fferem buttons allocated to differel-1 •
movements at an intersection. They are much simpler in design and visual clisplay chan laptop computer.;, and ftaru.c-e.
rugged casing and long battery life. Electronic count boards contain an internal clock that separates tho data by whac-
ever interval is chosen, so no field forms are needed. Electronic count boards have an advantage over Wly sheeG and
Volume Studies • ~
mechanical counting boards, since paper forms are
more sensitive to weather (wind and rain), and are
sometimes hard to keep organized for long studies.
Mosc importantly, they preclude the need for manu-
al data reduction and summary. Data may be traru-
feued directly from the field to a computer in the
office via modem or USB connection upon recurn
from the field. In the analysis software the data are
summarized and processed, and the results displayed
in a sdecred presentation format. This eliminates the
data reduction step required with tally sheets and
mechanical count boards.
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
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.
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
3.1.4.2 Obrerver Location
Observers must position themselves where they can most clearly view the t~affic they are counting. Observers must
avoid vanrage points blocked regularly by trucks, buses, parked cars, or other fearure.s. They should be located well
away from the edge of the travel way, both as a personal safety precaution and to avoid di.macting drivers. A posi-
tion above the level of the street and clc:ar of obstructions usually illOrds the best vanuge point If several observers
are counting at the same site, it is helpful for them to maintain visual contllct with one another, and to be able ro
communicate so as ro coordinate their activities. Given char observers are likely positioned on opposite corners of an
intersection, rwo-way radios or cell phones are helpful.
Protection from the elements is also an important consideration for rhe observer. Proper clothing ro suit prevailing
weather conditions is critical. Safety vests should be worn if the observer is near traffic at any time. Observers may
co unt from inside vehicles as long as their view is unobstructed. Sitting in an automobile is safer and more comfort·
able during inclement weather than sitting outside. Observers should park the vehicle in a legal space that is close
enough to _the intersection so they can sec the farthest lane of their assigned approaches. Parking on priVllte property
is sometimes convenient, but observers musr first obtain the property owner's permission. While sitting outSide,
observers may use chairs to prevent fatigue and umbrellas for protection from the sun, as long as these devices are
not distracting to drivers. A sign indicating that a traffic count is under way usually satislies drivers' curiosity about
observers they can view from the roadway.
The analyst should identify count locations ahead of dalll collection or plan sufficient rime to scope' the site just before
the count. In general, it is goo<i practice to arrive ar the sire at leasr 15 min. before the scheduled counr start rime to
allow for set-up and familiarization. It also helps to watch traffic for the equivalent of a few signal cycles.
The observer must concentrate his or her attention on accurately recording each count in the proper place or with the
proper button. Special care must be taken with deccronic counting boards to ensure they are properly oriented to the
geographic and geometric layout of the intersection. When rwo or more observers arc working together, time intervals
musr be maintained and coordinated accurately. Observers should also look for and note on their forms, or in a log,
any temporary traffic events, such as collisions or maintenance activities, that may lead to unusual traffic countS. For
this reason, it is good practice to always carry a field notebook, even if using automated count equipment. It is also
good practice to have the data colleccor llllly each data collection sheer before handing it to the analyst ..This practice
improves accuracy and decreases missed countS.
Modem technologies for automated counts can principally be divided into on-road technology and roadside technol-
ogy. 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 com-
municate with roadside readers.
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 col-
lected 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.
Agencies establish petmanenc-count stations whe,re they desire long-term, continuous counts (for example, 24 hours
a day, 365 days a year). The volumes collected at chese stations are usually part of an areawide program co monitor
traffic charactetistics and trends over time. Automatic count technology is also available in portable format for use in
temporary applications. ·
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 technol-
ogy) 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 tech-
nology 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
c) On-Road Magnetic Inductance 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.
:•
~:.
3.2.2.2 Roadsitk Count Ttchrwlogy
Many modem automatic count technologies can reliably detect vehicles from a roadside or overhead location. Ad-
V211CCS in radar, laser, microwave, infrared and video-image processing technologies allow for automated and non-
destrUctive volume measurements at intersections and in freeway applications. While some reports claim reliability
:: concerns with some of these technologies (Federal Highway Administration (FHWA) 2006), it is expected these
~~ concerns will decrease over time. In fact, others cite good reliability in research applications (Banks 2008), suggesting
:t! the bigger concern with these technologies may lie with long-term applications and maintenance.
.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 intersec-
tions. 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 equip-
mentis available, but used less frequently than in-road equipment. Exhibit 4-13 shows a selection of modern
roadside data collection equipment.
:~
'.
q
::1
i
\
l
\
\ ,
( 64 • MANUAL OF TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
3.2.2.3 V'ut'eo !mag~ Processing
Video-image processing systems. can automatically collect volume and other data from video. The analysis process
typically involves computerized meas\u'em.ent of lighting changes in pixels on the video; however. the exact algorithms
are proprietary and vary among different manufacturers of video-image processing technology and software.
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 dif-
-I fen;nt movements on one approach of the intersection. Because of sight-angle restrictions, multiple c:a.meras :a.re used
J
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 re-
corded. 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.
3.2.4.1 Preparation
Field work should 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 ofequipment to be used and the deployment
procedures to be followed. All equipment should be checked to see that it is limctioning properly. An ample supply ofacces-
sories (such as nails, clamps, tape, adhesive, chains, locks, batteries) and all necessary tools should be on hand.
• 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 vol-
ume 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
4.0 DATA REDUCTION AND ANALYSIS
Following collection, raw data must be placed in a form suitable for analysis. This reduction usually c onsists of con-
. verting rally marks to numbers, summarizing the data by calculating subtotals and totals, and arranging the dara in
·an appropriate format. The analysis may range &om a simple extraction of descriptive information ro a sophisticated
statistical treatment, depending on the type of study being conducted.
~ LU 12:11
PM 8:00
110.:1.:..:.:15;:.___ __
110..:"'=._ __
WAIIMrCoDtlllloM
AN~
w~
PM~
c-ar<•> - - -- -·
0
l
:l'
tt City of Raleigll
Pullllc Worb
:;• This computer software accomplishes the reductions and produces summaries of the data and desired calculated
...
;i":
.,,
values (such as ADT, peak-hour factors and percent turns). These tools save analysrs time and often eliminate errors.
·;; 4.2.1 Qnnerlmg Axle Counts to Wlbick Counts
·ll
lit Older automatic cotintcrs used to be driven by single-point sensors such as pneumatic road tubes, tape Switches, or
.:.
'~
.}
loops. They therefore recorded axles, not vehicles, and the raw count had to be converted in the analysis step. While
ri methods exist to convert axle counrs to vehicle counts, modern data collection equipment typically makes this step
,:P. unnecessary. Modem pneumatic tubes arc delivered and inscallcd in pairs, and built-in logic automatically determines
~ a v_chicle and classification count. Other equipment based on magnetic induction technology also gives a vehicle and
c:Wsification count automatically. For guidance on how to convert axle to vehicle counrs the reader is referred to
j R.ocss, Prassas and McShane (2004) or Robertson cr al. (1994).
l
j
\J
'I
\l 68 • MANUAL Of TRANSPORTATION ENGINEERING STUDIES, 2ND EDITION
I
: 4.3 Count Periods
! The counting period sdected 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
Automatic counts are usually c:dcen· for a~mum of 24 hours. They may extend for 7 days, a month, or even a year.
The interval most commonly used is 1 hour, al!hough smaller intervals may be desired for certain purposes. Smaller
intervals r~uire greater computer storage space and reduction time, and the analyst should consider an appropriate
file management structure to organize count data. Modern database management systems allow the analyse to store
counts from multiple locations over long periods of time. The database can be queried for a particular study period.
Many providers of automated count technologies also offer services for data management, including databases and
query tools.
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 sea-
cion 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 quan-
tified by calculating, for the control-count data, me proportion of the tocal sampling period volume occurring dur-
ing 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
•f§l' .. ---.... - ;'~W~
~ . . ' '!. ~ ~· -~ ! , • . .- :,.,....~~
~ ..• ~ • .• ~- ·j./."
-~~!:f'~lt~1f~~~~~t'"~':~1~-·~t"" .... ~,~~tt~1!e~~,~~~
~-- ,,::~ .~_:.-,..~~- ...... ·~. ~~tl-•:: ~~- s .,... : .. 'f ~~;~~!·/ .... ~,.~-!.~·t-v ... -•.
Control Station Data
Tune Cou.nt (vehicles) Proportion of Period Total
7:00-7:1SAM 720 72016,864 .. 0.105
7:15-7:30AM nG 776/6,864 a 0.113 !
·~-
Dr.
~'r!Z. . ~.·--
j
l lO
140 t 120 r--
! 130 ~0 JIO r-
'o
~
120
110
1
l.
100 r-- t--
": 100 • 90 - f- f- r-- - t-
I • i
90
I ~ r----- ,_
:>' 80 I • ~
80 f- f- f- t-
.s • • I I I I ~ 70
fl
~ ~F.l~f~~"'g:lJr.>!:l
~ ~ ~ ~' ~ ~ .,~
?0
- ~ ~ < ~ - ~ < ~ ~ ~ 5
MH.IholYeoc l>o.IolWotk
Mool.bl11'rallk Varildou Dalli1nllic Variatloos
9.0 - DMainSt
-
f
~
8.0 -
7.0
6.0
_ • Railroad St
- s.o
I ro 4.0
~~
":' 0 3.0 ..
·~a ..
l! 2.0
!!i 1.0
~
...
E
0.0
n.
~ - ...
~ ~
.... ~ ~ ~
'T ,.:. :t 0 ... ;;;"' . .. ..
"'~ "';;; ~... ,.:. "' d.~ ;:; ~ ... "
~
~
-6 - - ....
- - - - - li! ;:; ~
... ~
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 the•sn: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.
~~·
I
;\.
'Ii .
,,
·;
:r1:seleCt riinet . • ·
F!o.ni ,.' _·.-106:
-:-·oo~M."----::J--,_.
. · To . .j 05:45 PM
.
. .,.
. ~
::J
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 GJil«tion T«hno!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.
Hummer, J. S. "Recenr Supersrreet lmplemenracion and R=arch." Confirmu Procudings ofthe 3rd Urban Smn Sympotium.
Seactle, WA. Washington, DC: TRB, 2007.
JAMAR Technologies. Data Col.kction Equipmmt. Retrieved December 29, 2008, from JAMAR Technologies: ja manech.com/
index.htrnl. ·
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, ID-
2004. 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 Sympo•ium 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.