Professional Documents
Culture Documents
To cite this article: Shahriar Najafi, Gerardo W. Flintsch & Alejandra Medina (2017) Linking
roadway crashes and tire–pavement friction: a case study, International Journal of Pavement
Engineering, 18:2, 119-127, DOI: 10.1080/10298436.2015.1039005
TECHNICAL NOTE
Linking roadway crashes and tire –pavement friction: a case study
Shahriar Najafia*, Gerardo W. Flintschb and Alejandra Medinac
a
Virginia Department of Transportation, Salem District, 731 Harrison Ave., Salem 24153, VA, USA; bDepartment of Civil Engineering,
Virginia Tech, 200 Patton Hall, Blacksburg 24060, VA, USA; cVirginia Tech Transportation Institute, Virginia Tech, 3500
Transportation Research Plaza, Blacksburg 24061, VA, USA
(Received 14 September 2014; accepted 29 March 2015)
Tire – pavement friction is a factor that can affect the rate of vehicle crashes. Several studies have suggested that reduced
friction during wet weather conditions, due to water on the pavement surface reducing the contact area between the tire and
the pavement, increases vehicle crashes. This study evaluates the effect of friction on both wet- and dry-condition crashes.
The data for the study were provided by the New Jersey Department of Transportation. Regression analysis was performed
to verify the effect of friction on the rate of wet- and dry-condition vehicle crashes for various types of urban roads. It was
found that friction is not only associated with the rate of wet-condition vehicle crashes, but it also impacts the rate of dry-
condition vehicle crashes. The analysis also suggested that the developed regression models could be used to define the
friction demand for different road categories.
Keywords: friction; road safety; locked-wheel; skid resistance; crashes
study were provided by New Jersey Department of on areas with a high rate of vehicle crashes. This means
Transportation. that data collection is mostly being done at the project
level on specific areas. Statistics show that 55.4% of
agencies collect friction measurements at the project level.
Background Only 33.9% of agencies collect network-level friction data
Friction is an important performance parameter of the road (Flintsch and McGhee 2009). To be able to collect
surface. It is defined as the force that provides resistance to network-level friction data, agencies need to implement
slipping when the tires are prevented from rolling (Shahin equipment with high daily output. In the USA, locked-
2005). The friction between tire and pavement is a critical wheel trailers are the most predominant friction tester used
factor in reducing crashes (Ivey et al. 1992, Henry 2000, by state DOTs. They measure the steady state friction –
Najafi et al. 2011). Most skidding problems occur when also called Skid Number (SN) – on a fully locked wheel at
the road surface has friction deficiencies due to wetness a certain speed (40 mph based on the ASTM standard).
(Flintsch et al. 2012). According to some sources, one- The locked wheels use two types of test tires; Smooth
third of all wet-condition crashes are related to surface (ASTM E-524) and Ribbed (ASTM E-501) (Flintsch et al.
characteristics of roadways (Fwa 2005). Traditionally, it 2010, Najafi 2010). Ribbed-tire locked-wheel trailer was
has been found that pavement friction is high during dry used in this study. Figure 1 shows a locked-wheel skid
conditions. Data collected with a Sideway-force Coeffi- trailer. The SCRIM is the most commonly used friction
cient Routine Investigation Machine (SCRIM) device has tester in Europe. SCRIM devices can measure friction
proven that dry friction is very similar for various surfaces continuously. The device has a bigger water tank
(Fwa 2005). Once the pavement gets wet, the surface compared to locked-wheel trailers, so it can test a longer
friction drops significantly. Pavement surface friction length of roadway network, on average, 125 miles
gradually drops as the water film thickness increases; (200 km) of roadway every day (Fwa 2005).
however, the changes in pavement surface friction are not Friction data can be used in developing Pavement
substantial (Fwa 2005). Several studies have quantified the Friction Management (PFM) programmes. PFM includes
effect of water film thickness on surface friction engineering practices and periodic data collection to
(Kulakowski and Harwood 1990, Rose and Gallaway ensure an adequate friction level for the riding surface.
1997, Najafi et al. 2013). This can be achieved by using appropriate pavement
Several factors can affect pavement friction. Accord- design, construction and maintenance techniques. Guide-
ing to Hall et al. (2009), these factors can be classified into lines and techniques for providing adequate surface
four groups: (1) pavement surface characteristics (i.e. texture and friction levels can be found in FHWA
microtexture, macrotexture, material properties and Technical Advisory TA 5040.36, ‘Surface Texture for
temperature), (2) vehicle operating parameters (i.e. slip Asphalt and Concrete Pavements’ (FHWA 2013).
speed and driving maneuver), (3) tire properties (i.e. Locations with high rates of wet-condition crashes need
footprint, tread design, inflation pressure, rubber compo- to be identified and further investigated. This can be done
sition and temperature) and (4) environment (i.e. climate, by computing the ratio between wet-condition crashes and
temperature, water, snow and ice, and contamination) total crashes (wet- and dry-condition) and then using one
(Hall et al. 2009). Seasonal variations can have a of the following procedures (FHWA 2013):
significant effect on pavement friction (Fwa 2005). This
effect can be explained by changes in temperature and
precipitation (Jayawickrama and Thomas 1998). Overall
friction is lower in summer months compared to other
times of the year. This is mainly due to accumulation of
polished particles from pavement or contamination from
vehicles (Flintsch et al. 2012). Since both hot-mix asphalt
surface and tires are viscoelastic materials, temperature
can affect their properties. Several researchers have
studied the effect of temperature on pavement friction
measurements (Bazlamit and Reza 2005, Bianchini et al.
2011). Although several attempts have been made to
quantify the effect of seasonal variation on friction
measurement (Hill and Henry 1981, Wambold et al. 1989,
Jayawickrama and Thomas 1998, Gonzales 2009), there is
no practical model currently available for this purpose.
Friction measurements are typically collected as part
of the state’s wet accident reduction programme (WARP) Figure 1. Locked-wheel trailer.
International Journal of Pavement Engineering 123
3
121
. Set a specific threshold for the wet-condition crash Table 1. Fatal and injury-causing accident counts.
ratio, above which a location will be identified as an
Number of sites with accidents
elevated wet-condition crash location. This ratio can
vary between 0.25 and 0.5, depending on geo- Surface
graphical and climate conditions. condition Speed limit
. Compare the wet-condition crash ratio with the Functional class Dry Wet Min Max
average crash ratio for a specific highway functional
class in that area. The location will be identified as Urban Principal Arterial 20308 5010 25 55
Urban Interstate 3963 1251 35 65
an elevated wet-condition crash location if the
Urban Minor Arterial 2979 714 25 55
computed ratio is above the average by a specified Urban Freeway Expressway 2409 665 25 65
percentage. Urban Collector 20 11 25 50
. Set the threshold at the minimum of the total number Urban Local 12 4 45 50
of crashes within a segment. Rural Principal Arterial 758 166 35 65
Rural Interstate 510 170 50 65
The American Association of State Highway and Rural Major Collector 687 158 30 55
Transportation Officials Guide for Pavement Friction
Management outlines several methods for highway
agencies to establish Investigatory Level (desirable) and Friction measurements were divided into intervals
Intervention Level (minimum) thresholds for pavement (bins) with increments of two friction units (20: SN
friction and texture (Hall et al. 2009). Most of these below 20, 22: 20 , SN # 22, etc.) and the numbers of
methods require historical data for pavement friction, crashes for each friction interval were determined for
which are not readily available for most states. different road classes. This number was then used to
find the rate of vehicle crashes in million vehicles per
mile (MVM) using the following formula (Gan et al.
Data collection 2012):
The friction database includes the network-level friction
Number of crashes £ 1; 000; 000
measurements collected by a ribbed-tire, locked-wheel Crash rate ¼
skid trailer. The friction database also includes traffic Exposure
ð1Þ
(annual average daily traffic (AADT)), speed limit, road Number of crashes £ 1; 000; 000
¼ ;
functional class (Urban Interstate, Urban Principal AADT £ 365 £ Y £ L
Arterial, etc.), and the location (route name and the
where AADT is total annual average daily traffic; Y is
milepost) of the collected data.
the study duration in years and L is the length of the
The crash database includes the location information
roadway segments.
of the crashes (route and milepost), type of accident (fatal,
The duration of study was 1 year. Average AADT was
injury-causing, etc.), road surface condition at the time of
used for each friction bin.
accident (wet, dry, etc.), and the date of the accident. The
analysis of this project considered only the crashes that
had at least one injury or fatality. For practicality, only the
Regression analysis
data for one year (2007) were analysed.
To determine the significance of the effect of friction
(regressor) on the rate of vehicle crashes (independent
Data analysis variable), an analysis of variance (ANOVA) was
performed using SAS software. The summary of
The first step for analysing the data was to match up the
ANOVA statistics for Urban Principal Arterial roads is
friction measurements with corresponding crash location.
provided in Table 2. According to the P-values provided in
This was done by comparing the route and milepost
Table 2, friction (SN) is a significant factor in the model,
information. If the difference between the milepost of the
with a confidence level greater than 95% (P , 0.05).
crash location and the friction measurement that were on
To verify the adequacy of the linear model, residual
the same route was less than 0.1 miles, they were
analysis was performed. Residual analysis is a diagnostic
matched. The summary of the crashes that were
method for examining the adequacy of the fit of a
successfully matched with friction locations, as well as
regression model. There are several assumptions that we
the range of speed limit for each roadway classification is
make in any regression analysis. According to Montgom-
provided in Table 1. Since urban collector, urban local,
ery et al. (2001), these assumptions are as follows:
and all rural roads had very few matched locations
compared to other types of roads, they were not included . The relationship between response and regressors is
in further analysis. linear.
4
122
124 S. Najafi et al.
3.62155
20.02871
variance.
2.90102
20.04507
Graphical analysis of the residuals is a common way of
,.0001
Pr . F
,.0001
,.0001
Adj. R 2
F value
0.8073
18.83
2 9.38
t Value
Parameter estimates
Standard error
Coeff. var.
14.61751
0.17323
0.00393
3.26129
2 0.03689
21
22
1
Intercept
Variable
Table 2.
0.25024
(a) Normal probability plot (NPP) before transformation (a′) Normal probability plot (NPP) after transformation
1.0 1.0
0.8 0.8
Normal Cumulative Distribution
0.4 0.4
0.2 0.2
0.0
0.0 0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
CDF of RSTUDENT CDF of RSTUDENT
(b) Residual plot for predicted values (b′) Residual plot for predicted values after transformation (x Ln(x))
4 4
3 3
2 2
1 1
0 0
–1 –1
–2 –2
0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8
Predicted Value Predicted Value
(c) Residual plot for regressor (c′) Residual plot for regressor after transformation (x Ln(x))
4 4
Studentized Residual without Current Obs
3 3
2 2
1 1
Potential outliers
0 0
–1 –1
–2 –2
20 25 30 35 40 45 50 55 60 65 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2
SN logSN
summarises the model’s parameters. According to P- line is negative for all cases except for Urban Freeway
values, friction is a significant factor in all the models Expressway dry-condition crashes. This supports the
except for Urban Freeway Expressway wet-condition hypothesis that increasing the friction level decreases the
crashes (in which P . 0.05). The slope of the regression rate of both dry- and wet-condition vehicle crashes.
6
124
126
Analysis of variance
Source DF Sum of squares Mean square F value Pr . F
Model 1 5.77753 5.77753 115.97 , .0001
Error 21 1.04624 0.04982
Corrected total 22 6.82377
Root MSE Dependent mean Coeff. var. R2 Adj. R 2
0.22321 1.71191 13.0384 0.8467 0.8394
Parameter estimates
Variable DF Parameter estimate Standard error t Value Pr . jtj 95% Confidence limits
Road functional class Surface condition at time of crash Regression model .R 2 P-value
Urban Principal Arterial Dry Y ¼ 7.11 2 1.47 £ Ln (X) 0.84 , 0.0001
Wet Y ¼ 1.79 2 0.37 £ Ln (X) 0.73 , 0.0001
Urban Interstate Dry Y ¼ 2.33 2 0.47 £ Ln (X) 0.36 0.0144
Wet Y ¼ 0.96 2 0.21 £ Ln (X) 0.73 , 0.0001
Urban Minor Arterial Dry Y ¼ 11.78 2 2.58 £ Ln (X) 0.65 , 0.0001
Wet Y ¼ 3.09 2 0.69 £ Ln (X) 0.35 0.0009
Urban Freeway Expressway Dry Y ¼ 20.88 þ 0.43 £ Ln (X) 0.24 0.01
Wet Y ¼ 0.49 2 0.08 £ Ln (X) 0.10 0.15
Note: Y ¼ rate of fatal and injury-causing crashes (MVM), and X ¼ ribbed tire skid number (SNr).
International Journal of Pavement Engineering 127
1257
(a) Urban Principal Arterial - dry (a′) Urban Principal Arterial - wet
3.0 0.8
Crash rate = 7.11 -1.47 îLn (SNr) 0.7 Crash rate = 1.79 -0.37 îLn (SNr)
2.5 R-Square = 0.84 P-value < 0.0001 R-Square = 0.73 P-value < 0.0001
0.6
Crash Rate (MVM)
1.5 0.4
0.3
1.0
0.2
0.5
0.1
0.0 0.0
0 20 40 60 80 100 0 20 40 60 80 100
Friction (Skid Number Ribbed) Friction (Skid Number Ribbed)
1.40 Crash rate = 2.33 -0.47îLn(SNr) 0.35 Crash rate = 0.96 -0.21îLn(SNr)
R-Square = 0.37 P-value = 0.0144 R-Square = 0.73 P-value < 0.0001
1.20 0.30
Crash Rate (MVM)
0.80 0.20
0.60 0.15
0.40 0.10
0.20 0.05
0.00 0.00
0 20 40 60 80 100 0 20 40 60 80 100
Friction (Skid Number Ribbed) Friction (Skid Number Ribbed)
(c) Urban Minor Arterial - dry (c′) Urban Minor Arterial - wet
6.000 1.8
1.6
5.000
Crash rate = 11.78 -2.58 îLn (SNr) 1.4 Crash rate = 3.09 -0.69îLn(SNr)
R-Square = 0.65 P-value < 0.0001 R-Square = 0.35 P-value = 0.0009
Crash Rate (MVM)
4.000 1.2
1.0
3.000
0.8
2.000 0.6
0.4
1.000
0.2
0.000 0.0
0 20 40 60 80 100 0 20 40 60 80 100
Friction (Skid Number Ribbed) Friction (Skid Number Ribbed)
The regression models can be used to define the been shown in Figure 3(b). Similar friction threshold
minimum allowable friction level for various types of can be defined for other road classes. Overall, the dry
roads. The relationship between crash rate and friction crash rate is higher than wet crash rate. This is due to
is illustrated in Figure 3. Highway agencies should first wet time exposure and it can be corrected if wet
define the minimum allowable crash rate for various weather information is available. Consideration needs
road networks. Suppose that their goal is to decrease to be given on the available funding when one tries to
the rate of wet-condition vehicle crashes for Urban define a minimum allowable crash rate. Defining the
Principal Arterials to 0.5. This requires that the friction minimum allowable crash rate is beyond the scope of
be at least 33 [1.79 2 0.37 £ Ln (33) < 0.5]. This has this paper.
126
128
8 S. Najafi et al.
Annual Meeting Paper No. 14-5617. Compendium of Papers Journal of the Transportation Research Board No. 2209.
DVD, Washington, DC. Washington, DC: Transportation Research Board of the
Najafi, S., Flintsch, G.W., and McGhee, K.K., 2013. Assessment National Academics, 18 – 25.
of operational characteristics of continuous friction measur- Shahin, M.Y., 2005. Pavement management for airports,
ing equipment (CFME). International Journal of Pavement roads, and parking lots. New York, NY: Springer Science
Engineering, 14 (8), 706– 714. doi: 10.1080/10298436.2012. Business Media, LLC.
667097. Viner, H., Sinhal, R., and Parry, A., 2004. Review of UK skid
Roa, J.A., 2008. Evaluation of international friction index and resistance policy. In: Symposium on pavement surface
high friction surfaces. Virginia Polytechnic Institute and characteristics (of roads and airports), 5th International
State University [online]. Available from: http://scholar.lib. Symposium on Pavement Surface Characteristics-Road and
vt.edu/theses/available/etd-12182008-225503/unrestricted/ Airports, Toronto, Ontario, Canada.
Thesis.pdf [Accessed 30 Jul 2013]. Viner, H., Sinhal, R., and Parry, A., 2005. Linking road traffic
Rose, J.G. and Gallaway, B.M., 1997. Water depth influence on accidents with skid resistance-recent UK developments. TRL
pavement friction. Transportation Engineering Journal, 103, Paper Reference PA/INF4520/05.
491– 506. Wambold, J.C., et al., 1989. Pavement friction measurement
Schram, S., 2011. Specifications for aggregate frictional qualities normalized for operational, seasonal, and weather effects.
in flexible pavements. In: Transportation research record: McLean, VA: Federal Highway Administration.