Mamlouk, Vinayakamurthy, Underwood and Kaloush 1

1 EFFECTS OF THE INTERNATIONAL ROUGHNESS INDEX AND RUT DEPTH

2 ON CRASH RATES
3
4
5 by
6 1
Michael Mamlouk, Ph.D., P.E., F.ASCE*
7 Professor
8 Phone: (480) 965-2892
9 E-mail: mamlouk@asu.edu
10
11 1
Mounica Vinayakamurthy
12 Former Graduate Student
13 E-mail: mvinayak@asu.edu
14
15 2
B. Shane Underwood, Ph.D.
16 Assistant Professor
17 E-mail: Shane.Underwood@ncsu.edu
18
19 1
Kamil E. Kaloush, Ph.D., P.E.
20 Professor
21 E-mail: kaloush@asu.edu
22
23
24 1
Civil, Environmental and Sustainable Engineering Program
25 Arizona State University
26 P.O. Box 873005, Tempe, AZ 85287-3005
27
28 2
Deprtment of Civil, Construction and Environmental Engineering
29 North Carolina State University
30 Campus Box 7908, Raleigh NC 27695-7908
31
32 *Correspondence Author
33
34 Submitted for Presentation and Publication at the 97th Annual Meeting of the
35 Transportation Research Board
36
37

38 Abstract count: 197

39 Text and reference count: 4,187
40 Number of tables: 3
41 Number of figures: 9
42 Total word count: 197+4187+3,000= 7,384
43
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 2

44 ABSTRACT
45 Pavement distresses directly affect ride quality, and indirectly contribute to driver
46 distraction, vehicle operation, and accidents. In this study, analysis was performed on
47 highways in the states of Arizona, North Carolina, and Maryland to investigate the
48 relationship between accident rate and pavement ride quality (roughness) and rut depth.
49 Two main types of data were collected: crash data from the accident records and
50 International Roughness Index (IRI) and rut depth data from the pavement management
51 system database in each state. The ride quality and rutting values of crash and non-crash
52 segments were compared and showed close results. Crash rates were calculated using the
53 U.S. Department of Transportation method, which is the number of accidents per 100
54 million vehicle-miles of travel. Sigmoidal function regression analysis was performed to
55 study the relationship between crash rate and both IRI and rut depth. In all cases, the
56 crash rate did not show substantial increases until an IRI value of 210 inches/mile or a
57 critical rut depth of 0.4 inches. When the IRI or rut depth increased above these values
58 the crash rate increased. This is a key conclusion that provides empirically derived
59 thresholds for IRI and rut depth to reducing safety concerns.
60
61 INTRODUCTION
62 Several factors affect highway accident rates such as human factors, vehicular causes,
63 environment, roadway geometry, traffic volume, pavement condition, and their
64 combinations. Studies show that the majority of accidents are caused by human factors
65 such as distraction, alcohol, stress, physical deficiency and age (1). Although pavement
66 condition is not a major factor that affects accidents, the exact role of this factor is less
67 studied and therefore less well understood. On one hand, a good quality pavement may
68 reduce accident rates because of reducing driver distraction and providing good quality
69 vehicle operation conditions. Conversely, it may be argued that when the pavement
70 condition is poor, drivers tend to be more cautious and reduce speed, which in turn might
71 reduce the accident rate.
72 Pavement surface irregularities and rutting (Figure 1), directly affect ride quality, and
73 indirectly contribute to driver distraction, vehicle operation, and accidents (2). A
74 pavement that has a large number of surface defects (potholes, corrugations, very severe
75 fatigue cracking, etc.) can cause a vehicle to lose control when braking or turning,
76 especially under adverse environmental conditions (3). When pavement surface
77 irregularities increase, the contact area between vehicle tires and pavement decreases,
78 resulting in lower brake friction (4). Also, these irregularities can contribute to greater
79 vehicle instability since different friction forces may exist on the two sides of the vehicle.
80 Moreover, vehicles bouncing up and down on extremely rough pavements may result in
81 vehicle losing their loads causing accidents (5). Similarly, rutting may result in a driver
82 needing to exert extra effort to get out from the wheel path (if the rut depth is large), thus
83 leading to uncertain and in some cases uncontrolled lateral vehicle movement. In
84 addition, rutting is more hazardous in wet weather when water accumulates in the rut
85 path and leads to hydroplaning and loss of vehicle control. The problem can be further
86 exaggerated when human factors, such as distraction, alcohol, stress, physical deficiency
87 and age, are combined with pavement distresses.
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 3

88 FIGURE 1 Surface irregularities and rutting may contribute to driver distraction,

89 substandard vehicle operation and accidents (right photo curtsy of Atkins
90 www.atkinsglobal.com).
91 The majority of the studies dealing with the effect of pavement condition on safety are
92 related to skid resistance (e.g., 6, 7). Only a limited number of studies focused on
93 exploring the relationship between accident frequency and roughness or rutting. These
94 studies showed that increasing road roughness, in general, increases the rate of accidents.
95 Very limited information is available to determine the pavement condition level the
96 agency needs to maintain in order to actively reduce accident risk. Transportation
97 agencies have been looking for the appropriate roughness and rut depth thresholds below
98 which pavement-related accidents can be reduced.
99 OBJECTIVES
100 The main objective of this study is to determine the international roughness index (IRI)
101 and rut depth thresholds that correlate to an increased accident rate. General trends that
102 relate accident rates to roughness and rut depth are developed. Accident data and
103 pavement conditions from three states in different geographic locations and climatic
104 conditions are collected. Accident severity levels are separated and related to roughness
105 or rut depth.
106 LITERATURE REVIEW
107 Pavement roughness can be defined as the deviations of a surface from a true planer
108 surface with characteristic dimensions that affect vehicle dynamics, ride quality, dynamic
109 loads, and drainage (8). The IRI is one of the most common parameter used to measure
110 ride quality. Note that IRI does not measure roughness per se, since roughness is related
111 to the micro/macro surface texture. IRI, therefore, measures the response to pavement
112 roughness or technically “ride quality.”
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 4

113 King (9) investigated the effect of road roughness on traffic speed and road safety in
114 Southern Queensland, Australia. The study found a strong relationship between higher
115 crash rates and increased pavement roughness. Crash rates involving light vehicles were
116 more affected by increasing roughness than crashes involving heavy freight vehicles.
117 Considering different crash severity levels, crashes resulting in hospitalizations and
118 property damage had the strongest increase in crashes over a small increase in roughness.
119 The study also found that speed is reduced when roughness increases. The study
120 recommended that traffic authorities managing rural roads need to reduce roughness to an
121 IRI value of 120 in./mile in order to provide a safer road environment.
122 Cairney and Bennet (10) performed a study to determine the relationship between
123 pavement surface characteristics and roadway crashes in Victoria, Australia. The authors
124 measured pavement properties with a multi–laser profilometer and linked them to crash
125 data with the help of the Global Positioning System (GPS). They found that there was
126 good correlation between roughness and crash rate. However, no clear relationship could
127 be found between rutting and crash rate.
128 In a study by Graves et al. (11), the authors found that a disproportionate number of
129 crashes were associated with certain pavement conditions, hence suggesting that they are
130 correlated. The analysis was performed in Alabama and further suggested data mining
131 could be a useful technique in the analysis process. Li et al. (12) examined the
132 relationship between crash severity and factors indicating the pavement condition in
133 Texas. Results indicated that crashes of higher severity occurred on roads with poor
134 pavement condition compared to roads with fair pavement condition. It was also noted
135 that relatively higher severity crashes occurred on roads with very good pavement
136 condition. Purposefully laying down rougher pavements on high speed roadways was
137 suggested as a potential solution to avoid high severity crashes. A more recent study by
138 Li (13) indicated that pavement with poor surface conditions are responsible for higher
139 crash rates.
140 Tehrani and Cowe Falls (14) explored the relationship between the IRI value and number
141 of collisions in the province of Alberta, Canada. The study indicated that the sections
142 with high IRI values have more crashes in comparison to those with low IRI values.
143 Also, the results indicated that there was a good correlation between rut depth and
144 number of crashes in 1 kilometer segments.
145 Cenek et al (15) performed a study to develop statistical models predicting the correlation
146 between rut depths and fatal and injury crashes on New Zealand’s state highway network.
147 The results indicated that there was an increase in crash rate where the rut depth is 10 mm
148 or higher. The study suggested that these accidents might have been caused by the
149 accumulation of water on the road surface. The authors concluded that the crash rate, for
150 dry crashes in particular has decreased slightly in sections where the rut depth is slightly
151 higher than the normal range.
152 Chan et al. (3) performed a study to understand the relationship between accident
153 frequency and pavement condition using IRI, rut depth and PSI as parameters for
154 pavement condition. The study used Accident History Database (AHD) and Tennessee
155 pavement management systems data focusing on four urban interstates with asphalt
156 pavement and a speed limit of 55 mph. The results show that IRI and PSI were
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 5

157 significant in all types of models, whereas the rut depth model performed well in
158 predicting the accidents that occurred during night time only.
159 Hu et al. (16, 17) developed mathematical relationships between IRI and driving comfort
160 and safety (driving workload). The authors developed threshold IRI values on road
161 segments at different risk levels for driving comfort and safety. They also concluded that
162 standard IRI values for pavement maintenance are beyond the comfort and safety
163 threshold for both car and truck drivers.
164 In summary, it can be suggested that roughness and rutting can be contributing factors for
165 traffic safety and crash occurrence. The literature suggests that pavement roughness has
166 good correlation with crash rate and affects crash severity. The contribution of rut depth
167 to traffic safety is not well defined. No guidelines are currently available in the literature
168 to assist highway maintenance authorities to maintain their pavement at a certain level in
169 order to minimize crash occurrences.
170 DATA COLLECTION AND PROCESSING
171 Data were collected from Interstate, U.S. and State roads in Arizona, North Carolina and
172 Maryland. The data were obtained for both flexible and rigid pavements without
173 distinction between the years 2013 and 2015. Two main types of data were collected:
174 crash data from the accident records and IRI and rut depth data from the pavement
175 management system (PMS) databases. Data were brought together from the national
176 Highway Performance Monitoring System (HPMS) public data release and open source
177 data available from the three states.
178 The specific data items that were used in the study were crash frequency and severity,
179 average annual daily traffic (AADT), IRI, and rut depth. As the duration of analysis is
180 one year for the study, the data were sorted and separated for each year in each state.
181 Crash data and PMS data were matched and merged together on the basis of location.
182 For Arizona and North Carolina, data matching was performed by taking road name and
183 milepost as common criteria. However, for Maryland, data were matched using latitude
184 and longitude or GIS coordinates as the matching criteria. The PMS data available in the
185 shape file format were extracted using ArcGIS and were converted into csv files for use
186 in further analysis. GIS coordinates along with the route ID which give information
187 about road name and milepost were obtained from the shape files. After obtaining the
188 filtered data, SQL queries were used to correlate the data and obtain the necessary results.
189 SQL queries were written for all the crashes and for each severity level separately. After
190 matching the data using SQL, Microsoft Excel was used to perform further analysis, and
191 grouping the data on the basis of IRI and rut depth.
192 During the initial screening of the data, the crash occurrences that state other factors, such
193 as weather condition, as the major cause of the crash were removed prior to the analysis.
194 However, in most cases the contributing factor for crash occurrence was not reported.
195 One mile segments are used for analysis and 1 year is taken as the period of analysis.
196 The PMS data in Arizona were provided for each mile post. However, in North Carolina
197 and Maryland, the PMS data were provided for every 0.1 miles. In order to maintain
198 uniformity throughout the analysis, the Maryland PMS data were averaged to every mile.
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 6

199 Different states used different devices to collect IRI and rut depth data. In Arizona, the
200 profilometer was used to measure both IRI and rut depth. In North Carolina, the profiler
201 was used to measure IRI values and the profilometer to measure rut depth. In Maryland,
202 the Automatic Road Analyzer (ARAN) was used to measure both IRI and rut depth data.
203 Accident data were reported at 5 levels of severity. Although the severity levels are
204 similar in different states, they are named differently. Table 1 summarizes the severity
205 levels used in the three states.
206
207 TABLE 1 Severity Levels in Different States
Severity
Arizona North Carolina Maryland
Level
1 Damage without injury Damage without injury Property damage
2 Minor injury Injury level C
3 Non-incapacitating injury Injury level B Physical injury
4 Incapacitating injury Injury level A
5 Fatality Fatality Fatality
208

209 ANAYSIS OF RESULTS

210 Data Summary

211 The average IRI values were 72, 102 and 133 in./mile in Arizona, North Carolina and
212 Maryland, respectively. The average rut depth values were 0.06, 0.14 and 0.15 inches in
213 Arizona, North Carolina and Maryland, respectively. The differences in IRI and rut
214 depth values in different states could be because of the actual differences in pavement
215 conditions or because of other reasons such as type of data measured, type of measuring
216 equipment, data processing method, sampling method, unit length of pavement section,
217 and number of runs of measuring devices (18).
218 Table 2 presents the crash frequency data used for the analysis divided into different
219 severity levels. It can be noticed that the number of crashes used for analysis for
220 Maryland is considerably lower than the other states. This is due to the fact that crash
221 data obtained from Maryland did not cover the whole state since a portion of data was not
222 publicly available. Therefore, the Maryland crash rates were used to study their trend
223 with IRI and rut depth data, but not necessarily to show the actual crash rate values. Note
224 also that severity levels 2-4 (physical injury) are combined in the state of Maryland.
225 Also, no fatality crashes were reported in this Maryland sample of data.
226 TABLE 2 Summary Statistics of Crash Data

State (Year) All Severity Severity Severity Severity Severity

Severities Level 1 Level 2 Level 3 Level 4 Level 5
Arizona (2013) 31,514 21,748 4,473 4,149 838 306
Arizona (2014) 32,570 22,809 4,454 4,296 767 243
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 7

North Carolina 97,612 67,601 20,625 6,702 835 601

(2015)
Maryland* (2014) 807 607 204 -
227 *Partial crash data were obtained

228 Crash vs. Non-Crash Segments

229 The highways studied in the analysis can be divided into crash and non-crash 1-mile
230 segments. Crash segments are the road networks on which at least one accident has
231 occurred in the study period. On the other hand, non-crash segments can be defined as
232 the road networks on which no crashes have occurred during the study period.
233 The Arizona and North Carolina’s results show that the ride quality and rutting values of
234 crash and non-crash segments in each state-year are close to each other. This suggests
235 that ride quality and rutting are not the only factors affecting number of crashes but
236 possibly in combination with other factors such as traffic volume and others. Crash and
237 non-crash segments could not be separated in Maryland.

238 Crash Rate

239 For each state and each crash severity level, the IRI and rut depth values were broken
240 down to categories of 50 inches/mile and 0.1 inches, respectively, and the corresponding
241 number of crashes was obtained. The reason for separating the data into these bins was
242 to obtain a reasonable number of IRI and rut depth categories. For each category, the
243 crash rate was determined using the U.S. Department of Transportation method (19),
244 which can be calculated using the formula mentioned below.
C × 100,000,000
245 R= (1)
V ×365 × L× N
246 where,
247 R = Crash rate for the road segment per 100 million vehicle-miles of travel (VMT).
248 C = Average number of crashes in the study period
249 V = Average annual traffic volume entering the study area daily (AADT)
250 L = Length of the road segment in miles
251 N = Number of years of data
252 Since crash rates were calculated using 1-mile road segments and 1-year periods, both L
253 and N are equal to 1. In the rest of the analysis, only crash segments were used since
254 non-crash segments have crash rates of zero. If both crash and non-crash segments are
255 used, a large number of crash rates of zero will dominate the analysis and skew the
256 results.

257 Ride Quality Analysis

258 The relationships between crash rates and IRI values were investigated. Several curve
259 fitting functions were tried such as exponential, power, etc. The sigmoidal function
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 8

260 provided the best fit among other functions. Sigmoidal function models were developed
261 between the average crash rates and the average IRI values for each crash severity level.
262 During the analysis, data points that are obviously outside the typical range were
263 removed from the analysis. Equation 2 shows the sigmoidal function used in each
264 category.
α
265 log R=δ+ β +γ ¿¿
¿ (2)
1+e ¿
266 where,

267 R = Crash rate for the road segment per 100 VMT
268 D = Average IRI (in./mile)
269  = minimum logarithmic value of R
270   = maximum logarithmic value of R
271  = parameters describing the shape of the sigmoidal function
272
273 Figures 2-5 show the relationship between crash rate and IRI values for all severity levels
274 combined and for individual severity levels for Arizona, North Carolina and Maryland,
275 respectively. The graphs show an average crash rate value in each category of IRI of 50.
276 In all cases, the crash rate does not basically increase up to a certain IRI value, above
277 which crash rate starts to increase. This trend was noticed for various crash severity
278 levels. This phenomenon suggests that if the IRI value is kept below a certain value,
279 crash rate can be reduced.
280
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 9

281
282
283 FIGURE 2 Relationship between crash rate and IRI values for different severity
284 levels (Arizona 2013).
285
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 10

286
287
288 FIGURE 3 Relationship between crash rate and IRI values for different severity
289 levels (Arizona 2014).
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 11

290
291 FIGURE 4 Relationship between crash rate and IRI values for different severity
292 levels (North Carolina 2015).
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 12

293
294 FIGURE 5 Relationship between crash rate and IRI values for different severity
295 levels (Maryland 2014) (partial crash data were used).
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 13

296 Rutting Analysis

297 Similar to IRI data, the rut depth values were broken down to categories of 0.1 inches for
298 each state and each crash severity level. The relationships between crash rates and rut
299 depth values were investigated. Sigmoidal function models (similar to Equation 2, except
300 that D is the average rut depth in inches) were developed between the average crash rates
301 and the average rut depth values for each crash severity level.
302 Figures 6-8 show the relationship between crash rate and rut depth values for all severity
303 levels combined and for individual severity levels for Arizona, North Carolina and
304 Maryland, respectively. Similar to the IRI graphs above, these graphs show an average
305 crash rate value in each category of rut depth of 0.1 inches. In all cases, the crash rate
306 does not basically increase up to a certain rut depth value, above which crash rate starts to
307 increase. This trend was noticed for various crash severity levels. This phenomenon
308 suggests that if the rut depth is kept below a certain value, crash rate can be reduced.
309
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 14

310
311
312 FIGURE 6 Relationship between crash rate and rut depth for different severity
313 levels (Arizona 2014).
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 15

314

315
316 FIGURE 7 Relationship between crash rate and rut depth for different severity
317 levels (North Carolina 2015).
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 16

318
319 FIGURE 8 Relationship between crash rate and rut depth for different severity
320 levels (Maryland 2014) (partial crash data were used).
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 17

321 Critical Pavement Condition

322 The aforementioned results show that the crash rate is not greatly affected by either IRI
323 or rut depth at low values, but at some critical value the crash rate increases. To
324 objectively define this critical threshold the sigmoidal function was used to estimate the
325 IRI or rut depth values when crash rate becomes non-zero. The second derivative (R”) of
326 the sigmoidal function, Equation 3, was first used to identify the inflection point of the
327 function.

328 (3)
329 Where D is ether the average IRI value in inches/mile or rut depth in inches.
330 Next, the slope of the sigmoidal function at this point was taken and extrapolated until it
331 intercepted the horizontal axis, as shown in Figure 9. The slope of the line at this point
332 was determined by taking the first derivative (R’) of the sigmoidal function, Equation 4.
333 The intersection of the tangent line with the x-axis was selected as the critical IRI or rut
334 depth.

335 (4)
336 In a few cases, the curves ended slightly below the inflection point because of the low
337 accident data available. In such cases, the procedure for determining the critical IRI and
338 rut depth values was kept the same for consistency. The amounts of error involved in
339 these cases were within the rounding errors of IRI and rut depth used in the study.
340

341
342 FIGURE 9 Determination of critical IRI or rut depth to minimize crash rate.
343
344 The critical IRI and rut depth values were determined for the cases of all severity levels
345 combined for each state as shown in Table 3. The table shows some variations in critical
346 IRI and rut depth values for the different states. These differences could be because of
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 18

347 differences in the type of measuring equipment, data processing methods, sampling
348 method, or number of runs of measuring devices. The table also shows a difference in
349 the critical IRI values of Arizona in 2013 and 2014, which could be within the
350 measurement errors. Considering all values in the table, the critical IRI value can be
351 taken as the average of all values for all cases. In such a case, the critical IRI value is 210
352 inches/mile, whereas the critical rut depth value is 0.4 inches. This is a key conclusion
353 that provides empirically derived thresholds for safety concerns.
354 Since crashes are typically caused by several combined factors, one of the limitations of
355 this study is the fact that the effects of IRI and rut depth on crash rates were investigated
356 separately and without considering other potential contributing factors. Future studies
357 need to combine the effect of IRI and rut depth with other factors. Other contributing
358 factors include human factors, vehicular malfunction, environmental factors, and
359 roadway geometry.
360

361 TABLE 3 Critical IRI and Rut Depth Values in Different Cases to Minimize Crash
362 Rate
State Year Critical IRI (in./mile) Critical Rut Depth (in.)
2013 192 ---
Arizona
2014 152 0.35
North Carolina 2015 268 0.35
Maryland 2014 208 0.4
Average (rounded) 210 0.4
363

364 CONCLUSIONS
365 The following conclusions can be drawn from the results of this study.
366 1. IRI and rut depth values of crash and non-crash segments in each state-year
367 combination were close to each other. This suggests that ride quality and rutting
368 are not the only factors affecting number of crashes, but possibly in combination
369 with other factors such as traffic volume and others.
370 2. There is a unique relationship between IRI and crash rate in all cases, indicating
371 that crash rate does not basically increase up to a certain IRI value, above which
372 crash rate starts to increase. This phenomenon occurred for individual crash
373 severity levels as well as for all crash severity levels combined.
374 3. Similar to ride quality, crash rate does not essentially increase up to a certain rut
375 depth value, above which crash rate starts to increase. This phenomenon occurred
376 for individual crash severity levels and for all crash severity levels combined.
377 4. The average critical IRI value above which crash rate starts in increase is 210
378 inches/mile, whereas the average critical rut depth value above which crash rate
379 starts in increase is 0.4 inches.
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 19

380 Future studies need to combine the effect of IRI and rut depth with other contributing
381 factors such as human behavior, vehicular malfunction, environmental factors, and
382 roadway geometry.
383
384 ACKNOWLEDGEMENTS
385 This project was funded by the National Transportation Center @ Maryland
386 (NTC@Maryland), one of the five National Centers that were selected in this nationwide
387 competition, by the Office of the Assistant Secretary for Research and Technology (OST-
388 R), U.S. Department of Transportation (US DOT). The cost share contributed by
389 Arizona State University is greatly appreciated. Appreciation is also given to
390 transportation engineers in Arizona, Maryland and North Carolina for answering queries
391 and providing data used in this study.
392
393 REFERENCES
394 1. Garber, N.J. and L.A. Hoel, “Traffic and Highway Engineering,” 5th Edition,
395 Cengage Learning, Stamford, CT, 2015.
396 2. TRB, “Influence of Roadway Surface Discontinuities on Safety: State of the Art,
397 Transportation Research Circular No. E-C134, Transportation Research Board,
398 National Research Council, Washington D.C., 2009.
399 3. Chan, C. Y, B. Huang, X. Yan, and S. Richards, "Investigating Effects of Asphalt
400 Pavement Conditions on Traffic Accidents in Tennessee Based on the Pavement
401 Management System (PMS)." Journal of Advanced Transportation 44(3), 2010,
402 150-161.
403 4. Wambold, J.C., et al., Effect of Road Roughness on Vehicle Braking, Highway
404 Research Record 471, 1973, pp. 76-82.
405 5. Burns, J.C., Roughness and Roadway Safety, Transportation Research Record
406 836, 1981, pp.8-14.
407 6. Hall, J. W. et al., “Guide for pavement friction,” NCHRP Project: 1-43,
408 Washington, DC, 2009.
409 7. Noyce, D. A. et al., “Incorporating Road Safety into Pavement Management:
410 Maximizing Asphalt Pavement Surface Friction for Road Safety Improvements.”
411 Draft Literature Review and State Surveys, Midwest Regional University
412 Transportation Center (UMTRI), Madison, WI, 2005.
413 8. Sayers, M., "Development, implementation, and application of the reference
414 quarter-car simulation." In measuring road roughness and its effects on user cost
415 and comfort. ASTM International, 1985.
416 9. King, B., “The Effect of Road Roughness on Traffic Speed and Road Safety,”
417 Ph.D. dissertation, University of Southern Queensland, Australia, Nov. 2014.
418 10. Cairney, P., and P. Bennet, "Relationship between Surface Characteristics on
419 Selected Rural Roads in Victoria Rural Roads." ARRB Conference. ARRB Group,
 Mamlouk, Vinayakamurthy, Underwood and Kaloush 20

420 2008.
421 11. Graves, S.J., R. Daniel, and A.D. Michael, “Mining and Analysis of Traffic
422 Safety and Roadway Condition Data.” 46th Annual Transportation Research
423 Forum. Transportation Research Forum, 2005, pp. 618-626.
424 12. Li, Y., L. Chunxiao, and D. Liang, “Impact of Pavement Conditions on Crash
425 Severity.” Accident Analysis & Prevention, 2013, pp. 399-406.
426 13. Li, Y., “Safety Impact of Pavement Conditions.” Transportation Research Board
427 93rd Annual Meeting, 2014, pp.14-5373.
428 14. Tehrani, S., and L. Cowe Falls, "A Comparison of IRI and Collision Rates in the
429 Province of Alberta." Canadian Society for Civil Engineering. Canadian Society
430 for Civil Engineering. 1024, 1, 2012.
431 15. Cenek, P. D., et al., The Relationship between Crash Rates and Rutting, New
432 Zealand Transport Agency Research Report No. 545, 2014.
433 16. Hu, S., and X. Qiao-Jun, “Safety Evaluation Method of Freeway Pavement
434 Condition.” Journal of Transportation Systems Engineering and Information
435 Technology, 2013, pp. 130-135.
436 17. Hu, J. et al., “Research on Comfort and Safety Threshold of Pavement
437 Roughness,” Presented at the Annual Transportation Research Board Meeting,
438 Washington, D.C., 2017.
439 18. Mamlouk, M. and C. Zapata, “The Need to Carefully Assess the Use of State
440 PMS Data for Use in the MEPDG Calibration Process,” Presented at the Annual
441 Transportation Research Board Meeting, Washington, DC, 2010.
442 19. Federal Highway Administration, U.S. Department of Transportation,
443 https://safety.fhwa.dot.gov/local_rural/training/fhwasaxx1210/s3.cfm.