Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Early-Life, Long-Term, and Seasonal

Variations in Skid Resistance in


Flexible and Rigid Pavements
M. Alauddin Ahammed and Susan L. Tighe

Skidding contributes to up to 35% of wet weather accidents. Increased wet-weather accidents were shown to decrease by 35% with a net
temperature and surface wear and polishing may affect the available return of 540% after laying anti-skid surfacing at more than 2,000
friction and further increase skid-related accidents. Several studies have sites in London (7). An improvement of average friction by 0.1 could
attempted to examine and quantify these variations mostly with inade- reduce wet accidents by 13% (7, 8).
quate or inappropriate conclusions. The surface friction of both port- The available surface friction may be affected significantly by
land cement concrete (PCC) and asphalt concrete (AC) pavements was increased temperature. SN was shown to decrease by about 1.2 points
measured monthly to determine the influencing factors and quantify the for an increase in temperature of 10°C (9). The average difference
seasonal fluctuation. Skid number (SN) and pertinent data of the Long- of SN measured in winter and summer was shown to be 6 points (10).
Term Pavement Performance program were obtained for both PCC and Pavement surface friction also changes over time as a result of envi-
AC pavements, incorporating all geographic and climatic regions of the ronmental and traffic-related wear. Variation of surface friction with
United States and Canada, to determine the contributing factors and time is therefore regarded as an important measure of pavement
quantify the long-term and early-life variations of surface friction. Sur- deterioration (11). For safety and economy, the pavement should
face friction was shown to fluctuate as a result of ambient or pavement exhibit sufficient friction over the service life, withstanding the
temperature fluctuation at 0.35 British pendulum number per 1C change seasonal fluctuation and long-term wear and polishing. A small
in temperature. The effect of prior weather was shown to be insignifi- number of research studies were devoted to this area but ended
cant. Following the construction, AC and PCC surface friction was with inadequate or inappropriate results, conclusions, and models.
shown to increase by 5 SN in about 18 months and 4 SN in about This study is aimed at quantifying the seasonal and long-term vari-
21⁄2 years. Skid resistance was shown to decrease thereafter at 0.27 SN ation of asphalt concrete (AC) and portland cement concrete (PCC)
for AC and at 0.24 SN for PCC pavements per million vehicle passes. pavements’ surface friction, including the early-life increase, and
Cumulative traffic passes, pavement age, speed, and temperature during to determine the influencing factors to aid highway engineers in
the testing and PCC pavement surface texture types were found to be sta- selecting the appropriate pavement surface.
tistically significant for the prediction of long-term surface friction. AC
pavement long-term surface friction was shown to be more sensitive, as
compared with PCC, to predominant climatic condition. RELEVANT RESEARCH

In a Pennsylvania study, SN64 was shown to decrease by 1.7 points


In the year 2000, 42,643 fatalities and about 2.9 million injuries for a 7-day period without rain (dry spell). Models were developed
were reported from about 6.3 million traffic crashes on the U.S. for short-term changes in surface friction, correlating the dry spell
road network with an estimated $230.6 billion loss (1). In 2003, factor (DSF) and pavement temperature or a function of rain (9).
2,766 people died, and another 222,455 were injured on Canadian However, the developed models show an increase in surface friction
roads from a total of 156,904 police reported crashes, with an esti- in Pennsylvania but a decrease in North Carolina and Tennessee for
mated $25 billion loss (2). Skidding contributes to 15% to 35% of all the same DSF and temperature. Skid resistance was shown to be the
accidents and to 13.5% of fatal accidents during wet weather (3, 4). lowest in summer and the highest in spring regardless of asphalt mix
Several past studies have indicated that skid-related accidents can types (12). A significant short-term variation in SN was found, with
be reduced by an improvement in pavement surface friction (5). For low SN after a dry period that rejuvenated after rainfall. Models
example, the wet-to-dry pavement accident ratio was shown to were developed for predicting the low SN during fall from friction
increase sharply from 0.23 to about 0.7 as the skid number (SN), measurement at any time during the year, with the hypothesis that
measured at 64 km/h (SN64), drops below 41 (critical SN) (6). The seasonal variations are caused by polishing of microtexture and wear
of macrotexture (13). Others, however, have found that seasonal and
Department of Civil and Environmental Engineering, University of Waterloo, 200
short-term variations in SN occur as a result of temperature, rainfall,
University Avenue West, Waterloo, Ontario N2L 3G1, Canada. Corresponding and DSF.
Author: M. A. Ahammed, alauddinalam@hotmail.com. Models were developed for long-term skid resistance variation
using 3-year data from two highways in Ontario, Canada. The model
Transportation Research Record: Journal of the Transportation Research Board,
for Highway 401 (Equation 1) indicated that a mix with higher
No. 2094, Transportation Research Board of the National Academies, Washington,
D.C., 2009, pp. 112–120. stability will exhibit higher skid resistance because of the mix’s
DOI: 10.3141/2094-12 greater capability in resisting the coarse aggregate immersion in

112
Ahammed and Tighe 113

the matrix (14). In contrast, the positive sign associated with flow I1 through I5 = indicator variables for differences in mean SN
indicates that skid resistance will increase with increased flow. from one test pavement to another.

SN100 = 0.714 ( MS) + 0.356 ( flow ) + 1.048 ( voids )


A yearly decrease of 0.22 SN for rural roads and 0.26 SN for urban
roads was observed in a Maryland study. Surface friction was shown
40.904 to decrease by 1 SN for 0.56°C (1°F) increase in temperature (16).
+ − 17.323 (1)
[ EQT ( F )] 0.081 With this trend, the SN will drop from 50 to 0 (no friction) for an
increase in temperature from 5°C to 33°C (28°C change), which
seems to be unreasonable. A relationship between the surface fric-
where
tion [British pendulum number (BPN)] and temperature on an inter-
SN100 = SN at 100 km/h, mediately polished surface was developed as given by Equation 4
MS = Marshall stability, (17 ). The developed correlation indicates that skid resistance will
flow = Marshall flow (0.25 mm), decrease by 0.232 BPN for 1°K increase in temperature. However,
voids = voids in mix, and no correlation was found when combining all the unpolished with
EQT (F) = equivalent traffic (F = commercial vehicle equivalence completely laboratory-polished asphalt pavement surfaces, leaving
factor). the data and model in question.

The state of Wisconsin uses Equation 2 for the prediction of trans- BPN T = 125.2508 − 0.232 T (4)
versely tined PCC surface friction deterioration (15). The positive
sign-associated HV indicates that friction will increase with an
where T is temperature (°K) and BPNT equals BPN at temperature T.
increase in the percentages of heavy vehicles on the design lane.
A few other related studies are available in the published liter-
This may not be justified, because heavy vehicles with more and
ature but with similar deficiencies. Lack of logical interpretation
larger tires than those of passenger cars will cause greater surface
might have resulted in such deficient or conflicting findings and
wear and thereby further reduce the available surface friction.
models (18).
ln ( FN ) = 3.99 − 0.0419 ln ( LAVP ) − 0.00129 DOL
+ 0.004774 HV (2) OBJECTIVE AND SCOPE

where It has been seen that research on seasonal, early-life, and long-term
variations of pavement surface friction are scarce and mostly defi-
FN = friction number at 60 km/h, cient, with inadequate or inappropriate results, conclusions, or mod-
LAVP = summation of all vehicles expected to pass over during els. This research has attempted to closely evaluate these three aspects
the design life of the pavements (millions), to aid highway engineers in selecting the pavement surface with
DOL = limestone, dolomite, or ankerite content of coarse appropriate skid resistance. More specifically, the objectives of this
aggregate materials (percentage by weight), and paper are (a) to examine and quantify the effect of environmental fac-
HV = percentage of heavy vehicles in design lane (% lane tors, such as ambient and pavement temperatures, rainfall and DSF,
per average daily traffic). pavement type and mix properties, on wet pavement skid resistance
A model (Equation 3) was developed for the prediction of SN fluctuation month to month; (b) to quantity the early-life increase of
with normalization of seasonal effects (10). The justification of the skid resistance; (c) to examine and quantify the effect of surface age,
variable traffic passes, asphalt and concrete mix properties, annual average
temperature, wet days, vehicle speed, climatic regions, and so forth,
⎡ ⎛ 2π ⎞ ⎤ on long-term performance; and (d) to develop performance models
⎢sin ⎝⎜ 365 ⎠⎟ ⎥
⎣ ⎦ for seasonal and long-term surface friction variations.

is not clear. Furthermore, it is completely illogical that the SN of a


particular pavement will vary with the SN of other pavements (vari-
DATA COLLECTION
ables I1 through I5).
The seasonal variation of pavement surface friction was measured
⎛ 2π ⎞
SN 64 = 32.28 − 0.14 ( temp5 ) + 0.031( RF5 ) − 0.66 sin ⎜ from February to October 2007 for six PCC and five AC pavement
⎝ 365 ⎟⎠
JD
surfaces. The PCC surfaces include smooth surface, burlap, broom,
and Astroturf dragged surfaces and exposed aggregate texture,
+ 13.53 I1 − 3.12 I 2 − 2.78 I 3 − 9.52 I 4 + 7.43 I 5 (3)
and longitudinal tined surfaces. Each surface texture configuration
consists of three specimens; measurements were taken on a total
where
of 18 concrete specimens. All specimens were prepared in the Cen-
SN64 = SN at 64 km/h, tre for Pavement and Transportation Technology (CPATT) labora-
temp5 = average of daily temperatures for the 5 days before tory at the University of Waterloo, Ontario, from standard 30 MPa
friction measurement, ready mix concrete used for PCC pavements. Fifteen AC surface
RF5 = cumulative rainfall over the 5-day period preceding course specimens were obtained from five sections, which con-
the measurement, tained aggregates from the same source and were constructed by
JD = Julian calendar day corresponding to the day of a single contractor in one season (same-age pavements). They
measurement, and included HL3 (two sections), Superpave®, stone mastic asphalt,
114 Transportation Research Record 2094

and polymer modified asphalt on the CPATT test track at the sion were used in the analysis, except that PCC surfaces were
Waterloo landfill site. ranked by SN exhibited by each texture configuration. Finally, data
All specimens were exposed to the outside natural environ- from 238 PCC pavement sections in 38 states and provinces of the
ment, and surface friction was measured monthly from February United States and Canada were used in the analysis. For AC pave-
to October 2007 using the British pendulum with the equipment ments, data from 256 sections in 33 states and provinces were used
at roadside. For a few months, BPN was measured directly on the in the analysis.
AC road surfaces at midlane. The road is located at landfill and Average SN of multiple measurements on each section was calcu-
was mostly closed for traffic during the testing period. The surface lated and used as the available surface friction of each section at each
texture was measured (with the sand patch method) time to time to age level. The time (age) between successive friction measurements
determine possible changes in the surface macro-texture level. and cumulative age for each section was calculated. The temperature
Daily low and high temperatures and rainfall data were obtained during the testing, vehicle speed, and all other data were converted
from the University of Waterloo weather station, in addition to pave- to metric units. Mean annual average daily traffic (AADT) and per-
ment and ambient temperatures during the testing. This enabled centage of trucks on LTPP lanes were calculated from traffic data dur-
determination of the true effect of the environment and develop- ing the years of friction measurements. Cumulative traffic on LTPP
ment of a model for seasonal variation that correlated the statistically lanes was then calculated. Mean annual wet days (percentage of num-
significant variables. ber of days in each year) and annual average mean temperatures were
For long-term friction performance, field data of the Long-Term calculated from weather data of 1986 through 2005 (20 years). For
Pavement Performance (LTPP) Release 21 were obtained for both PCC surfaces, average 28-day compressive strength was calculated
PCC and AC pavements, incorporating all geographic and climatic for each section, with appropriate correction for test at different ages.
regions of Canada and the United States. The PCC pavements data For AC surfaces, mean Marshall stability, flow, voids, and VMA were
included all sections under Global Positioning System (GPS)-3, calculated. The coarse aggregate percentage and maximum aggre-
GPS-4, GPS-5, and GPS-9, consisting of 1,692 SN measurements, gate size in the mix were also obtained for each surface layer.
while AC pavements data included all sections under GPS-1, GPS-2, Table 1 shows the summary statistics of the data that were used in the
GPS-6, and GPS-7, consisting of 2,742 measurements. For each analysis. The data cover four climatic regions: dry freeze, wet freeze,
section the information and data obtained included friction (skid) dry no freeze, and wet no freeze.
number, traffic use, age, annual wet days, annual average temper-
ature, climatic region, speed and temperature during the testing,
asphalt mix gradation, air void (AV), voids in mineral aggregates ANALYSIS AND RESULT
(VMA), stability, and PCC surface texture type and compressive
strength. Seasonal Variation of Pavement Surface Friction

The PCC and AC wet pavements’ uncorrected surface friction was


DATA PROCESSING shown to vary by up to 10 BPN and 13 BPN, respectively, depend-
ing on pavement or ambient temperatures during the testing and
The monthly skid data were routinely checked during the actual test- other factors that need to be established. Figure 1 shows the month-
ing for accuracy and consistency relative to that of preceding months. to-month fluctuation of corrected (normalized to 20°C) wet surface
Measurements were repeated if any doubtful situation occurred. Sur- friction for both AC and PCC pavements, which is likely to be related
face friction measurements were taken on a wet surface after thor- to seasonal variation of pavement surface characteristics. As shown
ough cleaning to remove any dust or other debris, so as to determine in the figure, the variation of skid resistance for both AC and PCC
the true effect of rainfall, temperature, and DSF. surfaces is almost identical (slight deviations are probably related to
The surface friction data in the LTPP database consist of mea- test repeatability). The difference between the lowest and highest wet
surements taken by several types of friction tester. Therefore, mea- surface friction was shown to be 8 BPN for both AC and PCC sur-
surements taken by a single equipment type with the maximum faces. This variation agrees well with that indicated by Jayawickrama
number of measurement points were selected for consistency in and Thomas (10) for the seasonal variation of 6 SN between winter
the analysis and findings. The selected equipment is the ASTM and summer, taking the SN value as 77% of the BPN (19). Statisti-
locked-wheel skid trailer (ASTM E 274). Among the locked- cal tests between AC and PCC surfaces and among various AC sur-
wheel friction tests, a smooth tire was used for a couple of sections. faces have shown that neither pavement types nor pavement-mix
These sections were also filtered out, for consistency with the method types are statistically significant at 5% significance level for monthly
of skid measurement and equipment used. Only the locked-wheel variations in wet surface friction.
skid test data obtained with the ribbed tire were used in the analy- Figure 2 shows the effect of pavement surface and ambient tem-
sis of PCC surface friction variation. For AC pavements, friction peratures on surface friction variation. As shown in the figure,
data were further sorted, and measurements were taken with the trends for the air and surface temperatures are identical and over-
ASTM locked-wheel skid trailer (ASTM E 274) from a single lap. This indicates that the seasonal variation of surface friction
manufacturer, to further reduce the data volume and to be more can be estimated from either of these, although ambient tempera-
consistent. The friction data of all sections were then individually ture has shown a slightly better correlation, with a correlation
checked for accuracy, practicality, and consistency, such as for coefficient (r) of 0.88. The developed correlation between surface
unusual increases or decreases. All the doubtful data were filtered friction and temperature is given by Equation 5. The model is sta-
out to obtain meaningful and useful results and models. As the main tistically significant at 5% level of significance. The developed
objective was to examine the skid resistance performance over the correlation shows that for 1°C increase in temperature, the wet sur-
long term, sections with at least two friction data in time succes- face friction will decrease by 0.35 BPN. Taking the SN64 as 77%
Ahammed and Tighe 115

TABLE 1 Descriptive Statistics of the LTPP Data

Statistics Count Min Max Mean SD

Rigid Pavements
SN 743 22.0 65.5 46.6 6.86
Age, years 743 0 15.95 2.85 3.22
AADT 277 462 13,637 4,757 3,153
Cumulative traffic, millions 277 0 42.02 5.85 7.82
Truck, % 277 2.9 68.5 18.1 12.6
Speed, mph 743 41.8 88.5 65.3 4.64
Air temperature, °C 655 0 41.7 22.3 8.25
Compressive strength, MPa 677 20.64 56.67 37.2 7.60
Annual avg. temperature, °C 743 2.8 22.9 13.4 4.25
Annual avg. wet days, % 743 9.2 51.8 34.6 8.16
Flexible Pavements
SN 1,229 23.0 68.5 44.8 8.01
Age, years 1,229 0 12.05 2.12 2.26
AADT 474 407 16,700 3,723 3,387
Cumulative traffic, millions 474 0 37.3 3.6 4.98
Truck, % 474 3.8 44.9 15.5 9.85
Speed, mph 1,229 57.9 88.5 67.5 7.86
Air temperature, °C 1,164 0 42.8 22.5 8.22
Marshall stability, N 218 3,526 19,907 8,336 2,927
Annual avg. temperature, °C 1,079 2.7 25.1 12.7 4.23
Annual avg. wet days, % 1,079 6.5 56.6 35.0 9.59
Max. aggregate size, mm 484 9.5 38 16.9 4.1
Coarse aggregate, % 515 2 69 39.4 11.33
VMA, % 266 11.8 29.2 17.5 3.11
Voids, % 432 1.7 17 5.8 2.57

of the BPN value, as found by Ahammed and Tighe (19), a BPN Attempts had been made to correlate seasonal surface friction
of 0.35 would mean a SN64 of 0.27, that is, for each 1°C increase in variation with the mean texture depth, mean 1-day, 3-day, 5-day,
temperature, the SN will decrease by 0.27 SN64. This estimate agrees and 7-day low as well as high temperatures and total rainfalls, and
closely with the finding by Hill and Henry (9) for temperature effect, number of dry days preceding the testing day. None of these vari-
as follows: ables were statistically significant at 5% significance level. All
these observations indicate that the seasonal variation of surface
friction cannot be predicted in advance, unless the temperature of
BPN T = 75.181 − 0.35  T (5)
that particular time and date is known. The prior environmental
conditions are also insignificant. Short-term variation resulting
where T is temperature at °C. from oil spillage, debris and dust on the surface, and the washing

75
73

72

PCC AC
71

70
69
68

68
BPN

67
66

66

65

65
66

69

66

64

65

65

65

64

71

70

60
ne

r
ly
ry

st

er
be
-9

ril

ay
3

Ju

gu
-2
ua

Ju
Ap

ob
ch

em
ch

Au
br

ct
ar

pt
ar
Fe

O
M

Se
M

Months

FIGURE 1 Month-to-month variation of pavement surface friction.


116 Transportation Research Record 2094

80
Surf_Temp Amb_Temp Amb_Temp Surf_Temp

75
Amb_Temp, r = 0.89
Surf_Temp, r = 0.84

BPN
70

65

60
0 5 10 15 20 25 30 35 40
Temperature, °C

FIGURE 2 Effect of pavement surface and ambient temperatures on monthly friction


variation.

out of contaminants can be significant depending on the extent of pavements. The trends in the figure show that PCC surface friction
contamination. increases by about 4 points (SN64) on average from the initial SN,
whereas the increase is about 5 points (SN64) for AC pavements.
Pavement surface friction was shown to start decreasing after these
Early-Life Increase of Pavement Surface Friction initial periods of increase.

Pavement surface friction usually increases at an early age until the


construction debris and loose materials on the surface are cleaned Effect of Rigid Pavement Texturization Methods
up by traffic movement and environmental actions such as rain. The on Surface Friction Variation
full benefit of aggregate microtexture is available once the bitumen
or cement coating from the surface aggregate is completely removed. For determining the contribution of different texturization methods
on skid resistance, the SN at a single speed of 64 km/h (40 mph)—
The surface friction then starts to decrease and continues to decrease
SN64—was chosen. The variation of average SN64 and the ranks of dif-
until pavement surface distresses (e.g., raveling and cracks) cause
ferent surface texture configurations are shown in Figure 4. Rank 1
the friction to increase, if any friction increase actually occurs. A
represents the surface with the highest skid resistance, with a descend-
data set comprising two opposite trends likely to compensate for
ing SN value as the rank decreases. Among the PCC surfaces, cross-
each other in all analysis and in the model is aimed at determining
textured surface with transverse groove after burlap drag was shown
the effect of contributing factors to surface friction variation. There-
to exhibit the highest skid resistance (average SN64 = 50), while
fore, the processed LTPP data were divided into two groups: the first
the Astroturf dragged surface was shown to exhibit the lowest skid
group consisting of increases in surface friction during the early age
resistance (average SN64 = 41).
and the second group consisting of decreases in surface friction after
the early-age increase. This allowed for quantifying both early-age
and long-term skid resistance variations and for developing models. Skid Resistance Changes with Surface Age
The trends for AC and PCC pavements’ early-age surface friction
are shown in Figure 3. Full surface friction was attained in about The trends for change in pavement surface friction after an initial
30 months, on average, for PCC pavements and in 18 months for AC (early-life) increase are shown in Figure 5. The trends in the figure

65

55
AC
PCC
SN64

45
PCC
AC
35

25
0 1 2 3 4 5 6 7
Surface age, year

FIGURE 3 Early-life skid resistance changes for AC and PCC pavement surfaces.
Ahammed and Tighe 117

60

49 50
50 46 46
45

SN64
43 44
41
40

30

)
)

)
3)
)

(3

(2
(7

(4

(1
(6

(5

t(
ne

ne
rf

ve
p

oa
on
rla
tu

oo

o
Ti

Ti
Fl
tro

ro
m
Bu

Br

&
ia

G
ed
As

rf
D

s.
ov
Tu

an
ro

Tr
G

&
p
rla
Bu
Texture type

FIGURE 4 Average SN and ranks for different texturization methods of PCC.

show a consistent decrease in skid resistance with the age of the testing, annual average temperature, annual average wet days, and
pavement surfaces after the early-life increase. Figure 5 also sug- codes for climatic regions (dry versus wet and freeze versus no
gests that AC pavements’ surface friction decreases at a higher rate, freeze). For AC pavements, additional predictor variables were
than that of the PCC pavement surfaces, and the loss mainly occurs maximum aggregate size, coarse aggregate percentage, AV (%),
for about 8 years after the early-life friction increase. VMA (%), Marshall stability, and flow. For PCC pavements, addi-
tional variables included concrete compressive strength and a code
that accounts for the difference in surface texture type.
LONG-TERM FRICTION PERFORMANCE MODELS The best correlation of each independent variable (IV) with the
DV was selected by means of the curve estimation module in SPSS.
Multiple regression analysis in statistical analysis software Statisti- The attempted trends include linear, quadratic, inverse, logarithm,
cal Package for the Social Sciences (SPSS) 15.0 was used for devel- exponential, and power. The trend that showed greater correlation
oping the long-term friction performance models. The available coefficient (r) and made logical sense was selected for transforma-
variables were entered in the model, and statistically significant tion of each IV, if not for the linear trend. In all modeling attempts,
variables were selected through proper examination of the models. the normality was checked on the basis of the distribution of stan-
The predictor variables that were statistically significant at a 5% dardized residual. Observation points with standardized residual
significance level make practical sense and improve the prediction absolute values exceeding 2.0 were filtered out as outliers on the
power of the models and therefore were selected for building the basis of guidance in statistical analysis text (20). The filtering
final models. The dependent variable for these models is the aver- process was repeated until the normality condition was met. The
age SN. The predictor variables that were included for both PCC multicollinearity was checked on the basis of the variance infla-
and AC pavements are surface age, cumulative vehicle passes, per- tion factor (VIF), where a VIF value of less than 4 to 5 indicates no
centage of trucks, vehicle speed, ambient temperature during the multicollinearity problem (20).

70
PCC AC PCC AC
60

50
SN64

40

30

20
0 5 10 15 20
Surface age after initial period of friction increase, year

FIGURE 5 Changes in skid resistance with age after initial increase.


118 Transportation Research Record 2094

Models for Rigid Pavements’ Long-Term Friction Y = pavement age in years after an early-age increase in surface
friction (age since construction minus 21⁄2 years),
The outputs of the modeling attempts for long-term surface friction V = cumulative traffic passes in million after an early-age
had shown that annual average temperature, annual average wet days, increase in surface friction (total traffic since construction
and different climatic regions that distinguish between dry (dry–freeze minus traffic passes in 21⁄2 years),
and dry–no-freeze) and wet (wet–freeze and wet–no-freeze) weathers, T = temperature during the testing in °C, and
as well as between freeze (dry–freeze and wet–freeze) and no–freeze R = rank for different texture configurations in available skid
(dry–no-freeze and wet–no-freeze) are statistically insignificant or resistance, relative to the average friction number exhibited
meaningless. This indicates that the PCC pavement long-term surface by all surface textures.
friction is less sensitive to environmental condition. The percentage
The surface texture that had exhibited above-average skid resis-
of trucks was also shown to be statistically insignificant at 5% signif-
tance based on the network ranked above 1.0 and vice versa. The
icance level, but the number of cumulative traffic passes was shown
ranks for various surface textures are: Astroturf drag = 0.87, burlap
to be significant. Concrete compressive strength was also shown to
drag = 0.92, broom drag = 0.93, diamond ground = 0.96, Astroturf
be statistically insignificant. Pavement age and cumulative traffic
drag and tining = 0.98, grooved float = 0.99, tining = 1.04, and
passes were shown to be statistically significant in separate models.
burlap drag and transverse groove = 1.08.
Two sets of the model were, therefore, developed. The summary of
As shown in Table 2, all the predictor variables are statistically
these two models is presented in Table 2 and given by Equation 6
significant at 5% significance level ( p-values are less than 0.05) for
(Model A) and Equation 7 (Model B).
both models. The coefficient of determination (R2) is 0.59 for Model
A and 0.70 for Model B. The low R2 values do not indicate any prob-
SN S = 21.767 − 0.717 Y + 40.345 R − 0.198 S (6)
lem, if each regression coefficient is statistically significant at a
preselected significance level and makes logical sense and if the
SN S = 35.840 − 0.240 V + 35.486 R − 0.308 S − 0.131 T (7)
regression diagnostic does not show any problem. The regression
coefficients associated with age and traffic indicate that PCC sur-
where
face friction reduces at 0.7 SN per year or 0.24 SN per million vehi-
SNS = SN at speed S, cle movements. Surface friction increases with improved texture
S = vehicle speed in km/h, rank and decreases with increase in speed and with increase in

TABLE 2 Summaries of Long-Term Friction Performance Models

Parameter N Coefficient t-Value p-Value VIF R2

Rigid Pavement Long-Term Skid Resistance: Model A (Equation 6)


Intercept 153 21.767 3.69 0.000 0.592
Age, year 153 −0.717 −10.80 0.000 1.07
Relative rank 153 40.345 7.02 0.000 1.04
Speed, km/h 153 −0.198 −6.28 0.000 1.08
Rigid Pavement Long-Term Skid Resistance: Model B (Equation 7)
Intercept 127 35.840 6.70 0.000 0.701
Cum. traffic, millions 127 −0.240 −9.61 0.000 1.02
Relative rank 127 35.486 6.84 0.000 1.03
Speed, km/h 127 −0.308 −11.30 0.000 1.03
Test temperature, °C 127 −0.131 −4.89 0.000 1.01
Flexible Pavement Long-Term Skid Resistance: Model C (Equation 8)
Intercept 468 63.079 29.394 0.000 0.484
Age, year 468 −1.208 −10.00 0.000 1.03
Dry–wet code (DW) 468 5.321 10.17 0.000 1.04
Freeze code (FNF) 468 2.697 5.34 0.000 1.13
Speed, km/h 468 −0.179 −5.72 0.000 1.14
Test temperature, °C 468 −0.242 −8.74 0.000 1.08
Flexible Pavement Long-Term Skid Resistance: Model D (Equation 9)
Intercept 467 59.644 26.20 0.000 0.412
Cum. traffic, millions 467 −0.265 −3.57 0.000 1.04
Dry–wet code (DW) 467 5.901 10.78 0.000 1.03
Freeze code (FNF) 467 3.691 6.96 0.000 1.12
Speed, km/h 467 −0.133 −4.03 0.000 1.16
Test temperature, °C 467 −0.293 −10.16 0.00 1.09

NOTE: N = number of observation points.


Ahammed and Tighe 119

temperature during the testing (driving). All these variations make susceptibility of the asphaltic concrete to environmental changes
practical sense. probably justifies such variation.
The regression diagnostic has shown that VIFs are 1.01 to 1.08, The developed models show that AC pavements’ surface fric-
lower than the acceptable maximum value of 4 to 5 that leaves no tion reduces at 1.2 SN per year or 0.27 SN per million vehicle
doubt about the model. The errors were shown to be normally dis- passes, which is relatively higher than that of the PCC pavements’
tributed about the mean, and the scatter plot of errors has shown surfaces. AC pavements will exhibit 5 to 6 greater SN if the weather
no particular pattern, further proving the adequacy of the devel- is predominantly dry (dry–no-freeze and dry freeze) as compared
oped models. The models, therefore, may be used to predict the with wet (wet-freeze and wet–no-freeze). If the weather is pre-
PCC pavements’ surface friction after about 21⁄2 years of construc- dominantly freezing (dry–freeze and wet–freeze), the AC pave-
tion. For example, if surface friction after 10 years is to be pre- ments will exhibit 3 to 4 lower as compared with no freeze weather
dicted, the Y variable in Equation 6 would be 71⁄2 years. For (dry–no-freeze and wet no freeze).
pavement management, it is recommended to apply both models
and use the lowest friction value obtained from these two models
for preventive action. CONCLUSIONS

This study has examined the seasonal, early-life, and long-term vari-
Models for Flexible Pavements’ Long-Term Friction ations of both AC and PCC pavements’ surface friction on the basis
of the field data. The seasonal variation of both AC and PCC pave-
In AC pavement long-term surface friction models, the annual aver- ment wet surface friction was shown to be similar. The wet surface
age temperature and annual average wet days were shown to be sta- friction was shown to vary mainly as a result of ambient or surface
tistically insignificant, as with the PCC pavements. However, the temperature variation, which is probably associated with changes
codes defining different climatic regions were shown to be statisti- in tire hardness. The surface friction variation was shown to be
cally significant. The correlations of surface friction with asphalt 0.35 BPN for 1°C variation in temperature. Prior weather, such as
mix properties such as voids and VMA were shown to be impracti- dry-spell, rainfall, or temperature conditions, was not shown to be
cal, with a very high correlation coefficient, making all other IVs statistically significant for seasonal skid resistance variation. The
insignificant. A close study by the authors also showed that neither seasonal variation examined in this study, however, did not account
AV nor VMA has statistically significant or meaningful correlation for the short-term variation resulting from surface contamination
with the SN. This is probably because both AV and VMA are inter- such as dust, debris, oil, or rubber deposit.
nal properties of the mix, while the skid resistance is an external The LTPP data have shown that full surface friction is attained
property of the pavement surface. The coarse aggregate percentage in about 11⁄2 (AC) to 21⁄2 (PCC) years, on average, following the
and maximum aggregate size were also shown to be statistically construction. The early-life surface friction increase was shown
insignificant for long-term surface friction variation. Pavement age to be about 4 SN for PCC and 5 SN for AC pavements. The sur-
and cumulative traffic passes were shown to be statistically signif- face friction was shown to decrease thereafter with age, mainly
icant in separate models resembling the PCC pavement surfaces, because of traffic-related wear and polishing. PCC and AC sur-
and therefore, two models were developed. The summaries of these face friction was shown to decrease at 0.24 SN and 0.27 SN,
two models are also presented in Table 2 and given by Equation 8 respectively, per million vehicle passes, or 0.7 SN and 1.2 SN,
(Model C) and Equation 9 (Model D). respectively, per year. AC pavement surface friction was shown
to be affected more by predominant weather conditions as com-
SN S = 63.079 − 1.208 Y + 5.321 DW pared with that of PCC pavement surfaces. Four alternative mod-
els were successfully developed for the prediction of long-term
+ 2.697 FNF − 0.179 S − 0.242 T (8) surface friction. In practice, both models should be utilized, and
lower friction value should be used for pavement management or
SN S = 59.644 − 0.265 V + 5.901 DW investigation purposes.

+ 3.691 FNF − 0.133 S − 0.293 T (9)


ACKNOWLEDGMENTS
where DW is dry versus wet weather code (dry weather = 1, and
wet weather = 0), and FNF is freeze versus no freeze weather code The authors thank the Natural Science and Engineering Research
(no-freeze = 1, and freeze = 0). Council and Cement Association of Canada for funding the complete
The R2 values for Models C and D (Equation 8 and 9) are 0.48 and research on pavement surface characteristics.
0.41, respectively. However, as shown in Table 2, all the predictor
variables for both models are statistically significant at 5% signifi-
cance level. Equation 8 and Equation 9 also show that the coeffi-
REFERENCES
cients associated with the IVs are logical. Regression diagnostic has
also shown that errors are normally distributed, whereas the scatter 1. Noyce, D. A., H. U. Bahia, J. M. Yambo, and G. Kim. Incorporating
plot of errors has shown that errors are randomly distributed about Road Safety into Pavement Management: Maximizing Asphalt Pave-
a line with mean equal to zero with no particular trend. The VIFs ment Surface Friction for Road Safety Improvements. Midwest Regional
are from 1.03 to 1.16, indicating no multicollinearity problem. The University Transportation Center Traffic Operations and Safety (TOPS)
Laboratory, University of Wisconsin–Madison, April 29, 2005.
significance of DW and FNF indicates that the performance of the 2. Canadian Motor Vehicle Traffic Collision Statistics: 2004 and Road
AC pavements is more affected by the environmental condition as Safety in Canada: An Overview. Transport Canada, http://www.tc.gc.ca/.
compared with the performance of the PCC surfaces. The greater Accessed Nov. 07, 2006.
120 Transportation Research Record 2094

3. Hoerner, T. E., and K. D. Smith. High Performance Concrete Pavement: 13. Saito, K., and J. J. Henry. Mechanistic Model for Predicting Seasonal
Pavement Texturing and Tire-Pavement Noise. FHWA-DTFH61-01-P- Variations in Skid Resistance. In Transportation Research Record 946,
00290. FHWA, U.S. Department of Transportation, Washington, D.C., TRB, National Research Council, Washington, D.C., 1983, pp. 29–37.
2002. 14. Emery, J. J., M. A. Lee, and N. Kamel. Skid Resistance Predictive
4. What’s all the Noise About? Separating the Truth from Myths About Models. Pavement Surface Characteristics and Materials. ASTM STP
Tire-Pavement Noise. Better Roads, Vol. 73, No. 8, 2003, pp. 8–14. 763. ASTM, West Conshohocken, Pa., 1982, pp. 61–72.
5. Skid-Resistant Aggregate in Ontario. MERO-005. Materials Engineering 15. Drakopoulos, A., T. H. Wenzel, S. F. Shober, and R. B. Schmeidlin.
and Research Office, Ministry of Transportation, Ottawa, Ontario, 2003. Crash Experience on Tined and Continuously Ground Portland Cement
6. Rizenbergs, R. L., J. L. Burchet, and L. A. Warren. Relation of Accidents Concrete Pavements. In Transportation Research Record 1639, TRB,
and Pavement Friction on Rural, Two-Lane Roads. In Transportation National Research Council, Washington, D.C., 1998, pp. 140–146.
Research Record 633, TRB, National Research Council, Washington, 16. Song, W., X. Chen, T. Smith, and A. Hedfi. Investigation of Skid Resis-
D.C., 1976, pp. 21–27. tance of Hot-Mix-Asphalt-Surfaced Pavements in Maryland State High-
7. Kennedy, C. K., A. E. Young, and I. C. Butler. Measurement of Skidding way Network System. Presented at 85th Meeting of the Transportation
Resistance and Surface Texture and the Use of Results in the United Research Board, Washington, D.C., 2006.
Kingdom. Surface Characteristics of Roadways: International Research 17. Bazlamit, S. M., and F. Reza. Changes in Asphalt Pavement Friction
and Technologies. ASTM STP 1031, ASTM, West Conshohocken, Pa., Components and Adjustment of Skid Number for Temperature. ASCE-
1990, pp. 87–102.
Journal of Transportation Engineering, Vol. 131, No. 6, ASCE, Reston,
8. Hosking, J. R. Relationship Between Skidding Resistance and Accident
Va., 2005, pp. 470–476.
Frequency: Estimates Based on Seasonal Variation. TRRL Report
18. Ahammed, M. A., and S. L. Tighe. Statistical Modeling in Pavement Man-
No. RR 76. Department of Transport, Crowthorne, Berkshire, UK, 1987.
agement: Do the Models Make Sense? In Transportation Research
9. Hill, B. J., and J. J. Henry. Short-Term, Weather-Related Skid Resis-
tance Variation. In Transportation Research Record 836, TRB, National Record: Journal of the Transportation Research Board, No. 2084, Trans-
Research Council, Washington, D.C., 1981, pp. 76–81. portation Research Board of the National Academies, Washington, D.C.,
10. Jayawickrama, P. W., and B. Thomas. Correction of Field Skid Mea- 2008, pp. 3–10.
surements for Seasonal Variation in Texas. In Transportation Research 19. Ahammed, M. A., and S. L. Tighe. Pavement Surface Mixture, Texture
Record 836, TRB, National Research Council, Washington, D.C., 1981, and Skid Resistance: A Factorial Analysis. In Proc., 2008 Airfield and
pp. 82–86. Highway Pavements Conference. Transportation and Development
11. Pavement Design and Management Guide. Transportation Association Institute of ASCE, Seattle, Wash., 2008, pp. 370–384.
of Canada, Ottawa, Ontario, 1997. 20. Montgomery, D. C., and G. C. Runger. Applied Statistics and Probability
12. Elkin, B. L., K. J. Kercher, and S. Gulen. Seasonal Variation of Skid for Engineers, 3rd ed. John Wiley & Sons, Inc., New York, 2006.
Resistance of Bituminous Surfaces in Indiana. In Transportation Research
Record 777, TRB, National Research Council, Washington, D.C., 1980, The Surface Properties–Vehicle Interaction Committee sponsored the publication
pp. 50–58. of this paper.

You might also like