Evaluation of Pavement Safety Performance

CHAPTER 4. DATA COLLECTION

OVERVIEW OF DATA COLLECTION PROCESS

Because this was retrospective study, gathering data on existing low-cost pavement treatments was the most time-intensive aspect. The project relied heavily on the PFS volunteer States to provide the types and quantities of treatments available for analysis. The PFS volunteer States worked hard to provide information on the treatments, crash data, and roadway information. The following sections briefly outline the process used by the project team to gather the data necessary for this study.

Contact ELCSI-PFS Volunteer States to Identify Treatment Types and Availability of Crash and Roadway Data

The data collection process began by contacting the various ELCSI-PFS volunteer States to determine what treatment types would be available for evaluation under this study and to ascertain the availability of crash data for these sites. Because the treatments considered under this study were not necessarily installed as safety improvements, it was necessary for the State transportation department safety engineers to work closely with the pavements/materials/ maintenance engineers to identify what treatments are typically used and exact locations for those treatments.

Request Treatment Locations and Crash and Roadway Data

After the team had narrowed down the list of treatment types and the State transportation departments that were able to provide data, the team formally requested a list of treatment locations, crash data, and roadway data from the participating State transportation departments. This information was provided to the project team in various formats and further compiled into a consistent format by the project team for assessing the sample size.

Narrow List of Treatment Types and States Based on Sample Size

Based on the quantity (number and length) of treatment sites and availability and quality of crash data from the States, the team was able to further narrow the list of treatments and states. Having an adequate sample size for each treatment and for each State was important for evaluating the performance of each treatment, but also for being able to compare performance of the treatments between States. The only exception to this occurred with HFS, for which there are a relatively small number of installations nationwide, let alone in a given State.

Verify Treatment Locations, Installation Date, and Underlying Surface

Once a final list of treatment locations was provided, the project team worked with each State to verify the location and installation date for the purpose of mining crash data for each location. Pavement treatments considered in this study are typically installed over several months. For this reason, the installation year was effectively masked off when analyzing the crash data, leaving the years prior to the installation year as the "before" period, and the years after the installation year as the "after" period. The team worked with each State to verify that no other surface treatment or significant construction had been applied to the study locations during the after period. The team also worked with the States to verify the underlying pavement surface type. Most States were able to provide pavement or maintenance history records for the routes being studied for this purpose. These records were used to reconcile any discrepancies observed between the pavement surface type listed in the crash data records and the expected pavement type.

Identify Reference Sites and Collect Crash Data

A critical aspect of an EB before-after analysis is accounting for regression-to-the-mean, changes in traffic volumes, and time trends in crash reporting, by analyzing reference sites in addition to the actual treatment sites. This proved to be a difficult task as the treatments being considered are typically installed over long sections of a roadway that may have several variations in geometry, shoulders, median, roadway width, and traffic, among other factors. The project team used information in the roadway data file for each treatment site to identify reference site candidates that had characteristics as similar as possible to the treatment locations. One very important characteristic, because it relates to the pavement surface, is the pavement type for the reference sites. The team used a process similar to that described above to verify the pavement surface type at the reference sites. Once valid reference sites were identified, the project team collected crash data for the reference sites for use in the analysis.

Collect additional Data Collection: Climatic Data, Cost Information, and Materials Information

Additional data that were collected for each treatment site included historical climatic data, cost information for the various treatment types, and information on the materials used for each treatment. Climatic data was collected to assess any effects that climate (precipitation and temperature) may have on a particular treatment. Because the treatments were installed in a variety of climates, there is the potential that climatic conditions, precipitation in particular, could have affected crash frequencies and treatment effectiveness.

Cost information was collected for use in the BC analysis. Because of the large number of treatment sites, it was not possible to collect information on the exact cost of each particular treatment or project. Rather, the team looked at recent statewide bid averages to determine a rough unit cost for each treatment in each state.

Materials information was compiled from standard specifications and/or standard practices for each of the treatment types in each State. Of particular interest were any differences in the component materials (aggregates, binders) and tests used to measure various properties of these materials. As with cost information, it was not possible to collect materials information on each individual treatment site; rather the team simply looked at differences in standard specifications for each.

Data Collection Items

Table 4 summarizes the list of items originally proposed for collection. It was not possible to collect this level of detail for all of the treatment sites because of the limited information maintained by each State for specific projects. However, those items that were actually collected are discussed in more detail later.

Table 4 . List of potential data collected for treatment sites.

¹Minimum of 3 years preceding application of the treatment and 3 years following application, as available.
PDO = Property damage only
SRV = Skid resistance value

SUMMARY OF DATA BY STATE

Table 5 and table 6 show the number of treatment sites and/or mi of treatments provided by each of the volunteer States for asphalt and concrete pavements, respectively. Note that these are overall treatment quantities after eliminating nonviable sites, not broken down by facility type.

Table 5 . Treatment strategies and quantities provided, by State, for asphalt pavements.

HMA = Hot mix asphalt
OGFC = Open graded friction course
HFS = High-friction surfacing
UTBWC = Ultrathin Bonded wearing course

Table 6 . Treatment strategies and quantities provided, by state, for concrete pavements.

OGFC = Open graded friction course
HFS = High-friction surfacing
HMA = Hot mix asphalt
UTBWC = Ultrathin bonded wearing course

FINAL TREATMENTS AND STATES

Although the team wanted to include all sites and States in the analysis, project timeline and resource limitations made it necessary to focus efforts on analyzing treatments in the States that were able to provide a variety of treatments, for both concrete and asphalt pavement, and a significant quantity of each. Therefore, treatment sites included in the final analysis for conventional (i.e., non-HFS) treatments were the sites provided by California, Minnesota, North Carolina, and Pennsylvania. For HFS treatments, the sites from roughly half of the States that provided them were included in the final analysis.

Conventional treatment sites provided by States that were not included in the final analysis were submitted to FHWA for in-house studies. These included treatment sites provided by Connecticut (UTBWC), Mississippi (cape seal), Montana (chip seal), Tennessee (various treatments), and Wisconsin (chip seal).

CONVENTIONAL TREATMENT DATA COLLECTION

This section describes the process followed to collect and query the data required for the study and provides summary statistics for these data for each of the four States included in the conventional treatment analysis.

Pennsylvania

Data for Pennsylvania were provided by the Pennsylvania Department of Transportation (PennDOT). These data included geometric, pavement, traffic, and crash data for the entire State. The data are linked together using the District Number, County Number, State Route Number, Segment Number, and Segment Offset variables.

Geometric Data

Geometric data were obtained from the PennDOT Roadway Management System and included the following variables:

• Access control.
• Divided/undivided.
• Divided width.
• Surface width.
• Speed limit.
• Number of lanes.
• Urban versus rural environment.
• Surface type.
• Shoulder type.
• Shoulder width.
• Shoulder paved width.

Note that for access-controlled roadways, the data are directional, that is, each record corresponds to only one direction of travel. For non-access-controlled roadways, the data encompass both directions of travel.

PennDOT also provided a list of segments where some change in the route/ segment/offset segmentation system had been made between 2003 and 2011. This is significant because data from different years are being queried and if the segmentation has changed, then the data cannot be matched correctly. This was evident for some locations that appeared to have very large changes in traffic volume from 1 year to the next. Segments with a change were not used for analysis.

Pavement Data

PennDOT provided a file with pavement history statewide. For each section of roadway, data on up to 10 layers of pavement is included. The data indicate the year of resurfacing and the surface type.

Traffic Data

Traffic data in the form of AADT were obtained from PennDOT from 2003 to 2010. The percent of trucks in the traffic mix was also provided. Because the study period extends to 2011 and these counts were not available, it was decided to apply the 2010 AADTs to 2011 as well.

Crash Data

The PennDOT Crash Database is maintained by the Bureau of Highway Safety and Traffic Engineering’s Crash Information Systems & Analysis Division. Data from 2003 to 2011 were provided. The compiled crash data contain many variables related to the location, time, and characteristics of each crash. The following notes relate to the exclusion or inclusion of crash data and definition of crash types:

• Included only non-intersection-related mainline crashes (if intersect_type = "mid-block").
• Excluded crashes where harm_event = "hit deer" or "hit other animal".
• Excluded crashes where road condition = "snow", "slush", "ice", or "ice patches".
• Defined injury crashes as those where fatal_count or tot_inj_count is greater than 0.
• Defined ROR crashes as those where relation_to_road = "Outside trafficway".
• Defined wet crashes as those where road_condition = "wet".

Construction Data

PennDOT does not maintain a central file with construction information. For the treatment sites used the PennDOT staff did not indicate any other construction had taken place within the study period. It is possible some other works may have occurred; however, with the large mileage of treatment and reference sites available, if some works did occur at some locations, the impact on the results of the analysis would be negligible.

Treatment Sites

Sites on the initial list of treatment sites eligible for the study were subsequently removed for two reasons.

• Sites with a missing volume between 2003 and 2011 were removed.
• Sites with a change in the segmentation between 2003 and 2011 were removed.

Reference Sites

Reference sites were identified by first identifying road segments in the same county and with the same route number as the treated sites. Then the various geometric and traffic variables were compared to ensure that the range of values for these variables were consistent. The pavement data were linked to these data, and any segments that were resurfaced between 2003 and 2011 were removed.

Similar to the treated sites, potential reference sites were excluded if any year had a missing traffic volume count or the segmentation had changed between 2003 and 2011.

After querying the data, it was found that some sites have highly variable AADTs between years. After consultation with PennDOT, no satisfactory explanation was found for such systemic variation. To avoid biases in the SPFs to be developed due to incorrect traffic volume data it was decided to eliminate any segment where the AADT for any single year deviated from the segment average for the entire study period by 25 percent or more.

Table 7 through table 9 provide summary statistics for the treatment and reference site data in Pennsylvania.

Table 7 . Summary statistics for Pennsylvania treatment site geometry.

Avgshldwid = Average of left and right shoulder width
AC = Asphalt
CS = Chip Seal
Micro = Microsurfacing
SS = Slurry Seal
Max = Maximum
Min = Minimum
PCC = Portland cement concrete
HMA = Hot mix asphalt
UTBWC = Ultrathin bonded wearing course

Table 8. Summary statistics for Pennsylvania treatment site AADT and crashes.

ELCSI = Evaluation of Low-Cost Safety Improvements
AADT = Average Annual Daily Traffic
totrateb = total crash rate per mi-yr in before period
totratea = total crash rate per mi-yr in after period
injrate, wetrate, rorrate, wetrorrate = injury, wet-road, ROR, and wet-road ROR crash rates per mi-year, respectively
Max = maximum
Min = minimum
PCC = Portland cement concrete
HMA = Hot mix asphalt
UTBWC = Ultrathin bonded wearing course

Table 9 . Summary statistics for Pennsylvania reference sites.

Avgshldwid = Average of left and right shoulder width
AADT = Average Annual Daily Traffic
totrate = total crash rate per mi-year
injrate = injury crash rate per mi-year
wetrate = wet-road crash rate per mi-year;
rorrate = ROR crash rate per mi-year
wetrorrate = wet-road ROR crash rate per mi-year)
AC = Asphalt concrete
PCC = Portland cement concrete
HMA = Hot mix asphalt

North Carolina

Data for North Carolina were provided by the North Carolina Department of Transportation (NCDOT) and the Highway Safety Information System (HSIS). The data provided by NCDOT included a list of potential treatment sites and construction data, including paving projects. The data provided by HSIS included geometric, traffic, and crash data. The data are linked together using the County Number, Route Number, and Milepost variables.

Geometric Data

Geometric data were obtained from HSIS and included the following variables:

• Urban versus rural environment.
• Median width.
• Shoulder type.
• Surface width.
• Terrain.
• Median type.
• Speed limit.
• Surface type.
• Number of lanes.
• Roadway class.
• Shoulder width.
• Paved shoulder width.

Pavement Data

NCDOT indicated which of the treated site segments were asphalt or concrete pavements and provided pavement project history files for helping to verify the underlying pavement type. For the reference sites, the surface type variable was relied on as the existing pavement type.

Traffic Data

Traffic data in the form of AADT were obtained from HSIS from 2000 to 2010. The percent of trucks in the traffic mix was also provided.

After querying the data, it was found that some sites experience very high variability in traffic counts from year to year. After discussing with HSIS staff, it was concluded that this may be owing in large part to re-mileposting of the roadway segments from year to year. By comparing the roadway characteristics of a sample of these sites, it was found that the functional class often changed from year to year. In an attempt to eliminate as many of these re-mileposted locations as possible, the team removed any segment from the data for which the functional class or number of lanes changed from 2000 to 2010.

There was an additional concern with the 2009 AADTs in that many of these counts seemed out of line with the preceding and following years’ AADT count. In particular, on divided roadways, the 2009 AADT appears to be roughly half of that reported in either the preceding or following year. The suspicion is that in 2009, the AADT for one direction may be what is reported. No resolution was found for the data, and it was decided to substitute the 2010 AADTs for 2009.

Crash Data

The crash data were provided by HSIS for 2000 to 2010. The compiled crash data contain many variables related to the location, time, and characteristics of each crash. The following notes relate to the exclusion or inclusion of crash data and definition of crash types:

• Exclude intersection-related mainline crashes (if loctype is any type of intersection crash).
• Excluded crashes where acctype = "animal".
• Excluded crashes where rdsurf = "snow", "slush", or "ice".
• Defined injury crashes as those where severity = fatal, A class, B class, or C class injury.
• Defined ROR crashes as those where acctype = "ran-off-road right", "ran-off-road left", or "ran-off-road straight".
• Defined wet crashes as those where rdsurf="wet".

Construction Data

Construction history data was provided by NCDOT for 2000 to 2010. This list includes resurfacing and other projects, as well as concrete pavement ratings for verifying which sites were concrete.

Treatment Sites

Sites on the initial list of treatment sites eligible for the study were removed if they had an AADT under 500. As was discussed under Traffic Data, some segments are believed to have been re-mileposted but no record is available indicating the old and new mileposts for the same section of road. In an attempt to eliminate as many of these re-mileposted locations as possible, any segment for which the functional class or number of lanes changed from 2000 to 2010 was removed from the data.

Reference Sites

Reference sites were identified by first identifying road segments in the same county and with the same route number as the treated sites. Then the various geometric and traffic variables were compared to ensure that the range of values for these variables were consistent. The pavement data and construction data were linked to these data, and any segments that were resurfaced or had other construction works between 2000 and 2010 were removed.

Similar to the treated sites, potential reference sites were excluded if the functional class or number of lanes changed from 2000 to 2010.

Similar to the Pennsylvania data, it was found that some sites have highly variable AADTs between years. To avoid biases in the safety performance functions (SPF) to be developed due to incorrect traffic volume data, it was decided to eliminate any segment for which the AADT for any single year deviated from the segment average for the entire study period by 25 percent or more.

Table 10 through table 12 provide summary statistics for the treatment and reference site data in North Carolina.

Table 10 . Summary statistics for North Carolina treatment site geometry.

Avgshldwid = Average of left and right shoulder width
AC = Asphalt
PCC = Portland cement concrete
Max = Maximum
Min = Minimum
OGFC = Open graded friction course
HMA = Hot mix asphalt
UTBWC = Ultrathin bonded wearing course

Table 11. Summary statistics for North Carolina treatment site AADT and crashes.

ELCSI Treatment Type	Site Type		AADT Before	AADT After	Years Before	Years After	totrateb	totratea	injrateb	injratea	rorrateb	rorratea	wetrateb	wetratea	wetrorrateb	wetrorratea
Chip Seal	Multilane	Min	10,995	11,500	6.00	3.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Chip Seal	Multilane	Max	35,173	33,333	7.00	4.00	127.41	125.00	46.33	41.67	3.57	9.26	19.31	18.02	0.00	9.26
Chip Seal	Multilane	Mean	26,689	25,797	6.93	3.07	20.41	32.02	6.91	10.18	0.54	0.81	3.20	4.31	0.00	0.66
Chip Seal	Two-Lane	Min	500	500	2.00	2.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Chip Seal	Two-Lane	Max	24,368	32,000	8.00	8.00	41.85	100.00	31.25	50.00	31.25	7.58	17.54	22.73	9.62	2.40
Chip Seal	Two-Lane	Mean	1,980	1,947	5.74	4.26	0.83	0.97	0.41	0.40	0.10	0.04	0.17	0.15	0.02	0.00
Diamond Grinding on PCC	Freeway	Min	31,889	33,000	9.00	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Diamond Grinding on PCC	Freeway	Max	53,889	74,000	9.00	1.00	411.11	100.00	77.78	19.23	11.11	19.23	55.56	19.23	8.55	19.23
Diamond Grinding on PCC	Freeway	Mean	42,701	44,581	9.00	1.00	17.19	9.17	4.31	1.79	0.63	0.56	2.89	1.35	0.27	0.47
Microsurfacing	Freeway	Min	20,206	23,676	7.00	2.00	1.14	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Microsurfacing	Freeway	Max	30,314	32,000	8.00	3.00	95.24	15.15	12.99	6.33	10.42	4.44	10.42	15.15	3.00	0.00
Microsurfacing	Freeway	Mean	26,779	27,168	7.40	2.60	21.71	5.78	3.04	2.00	2.91	0.44	2.88	2.05	0.32	0.00
Microsurfacing	Multilane	Min	4,242	4,500	8.00	2.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Microsurfacing	Multilane	Max	8,032	8,200	8.00	2.00	1.48	1.69	0.62	0.93	0.00	0.00	0.45	0.74	0.00	0.00
Microsurfacing	Multilane	Mean	6,708	6,967	8.00	2.00	0.76	0.60	0.20	0.28	0.00	0.00	0.23	0.12	0.00	0.00
Microsurfacing	Two-Lane	Min	628	620	7.00	2.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Microsurfacing	Two-Lane	Max	9,539	12,667	8.00	3.00	8.24	250.00	4.46	10.10	4.76	3.70	6.49	5.56	2.86	3.70
Microsurfacing	Two-Lane	Mean	4,912	4,751	7.30	2.70	1.09	3.18	0.53	0.44	0.20	0.05	0.27	0.14	0.06	0.04
OGFC	Freeway	Min	17,563	17,000	2.00	2.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
OGFC	Freeway	Max	158,656	160,333	8.00	8.00	465.61	345.68	116.40	98.77	25.97	30.30	125.00	66.67	13.51	30.30
OGFC	Freeway	Mean	66,006	69,228	5.95	4.05	39.74	40.91	12.16	10.34	2.21	1.20	9.97	7.29	1.03	0.79
OGFC	Multilane	Min	14,961	13,313	2.00	3.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
OGFC	Multilane	Max	31,371	29,000	7.00	8.00	7.94	11.11	4.76	3.79	0.00	0.28	1.13	0.00	0.00	0.00
OGFC	Multilane	Mean	17,696	16,094	2.83	7.17	1.70	3.32	0.98	1.14	0.00	0.05	0.19	0.00	0.00	0.00
Slurry Seal	Two-Lane	Min	1,396	1,500	8.00	2.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Slurry Seal	Two-Lane	Max	3,771	3,400	8.00	2.00	9.62	12.50	9.62	0.89	0.00	0.00	0.27	0.00	0.00	0.00
Slurry Seal	Two-Lane	Mean	2,370	2,536	8.00	2.00	1.04	0.81	0.82	0.08	0.00	0.00	0.01	0.00	0.00	0.00
Thin HMA	Freeway	Min	6,270	6,200	2.00	2.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Thin HMA	Freeway	Max	117,500	124,500	8.00	8.00	1083.33	958.33	333.33	208.33	100.00	68.63	421.88	187.50	50.00	41.67
Thin HMA	Freeway	Mean	52,143	55,330	5.44	4.56	27.75	29.25	9.09	8.19	2.56	2.80	6.56	7.12	0.85	0.86
Thin HMA	Multilane	Min	984	2,200	2.00	2.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Thin HMA	Multilane	Max	61,962	63,000	8.00	8.00	1657.14	1500.00	557.14	333.33	333.33	74.07	328.57	333.33	14.29	40.00
Thin HMA	Multilane	Mean	19,323	19,588	5.04	4.96	18.15	20.70	6.26	6.03	0.92	0.96	3.20	3.37	0.15	0.29
Thin HMA	Two-Lane	Min	500	520	2.00	2.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Thin HMA	Two-Lane	Max	27,767	31,354	8.00	8.00	500.00	388.89	100.00	125.00	41.67	83.33	66.67	100.00	8.33	41.67
Thin HMA	Two-Lane	Mean	4,927	5,058	5.14	4.86	2.25	3.03	0.90	1.08	0.15	0.22	0.42	0.53	0.03	0.04
UTBWC	Freeway	Min	10,325	11,000	2.00	2.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
UTBWC	Freeway	Max	153,167	162,000	8.00	8.00	1033.33	2150.00	333.33	400.00	83.33	100.00	150.00	400.00	17.54	100.00
UTBWC	Freeway	Mean	69,054	74,410	6.65	3.35	43.51	63.15	13.13	13.06	2.46	3.53	7.70	13.53	0.84	1.85
UTBWC	Multilane	Min	4,932	12,000	4.00	3.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
UTBWC	Multilane	Max	41,833	51,667	7.00	6.00	155.88	94.55	61.76	35.26	10.00	7.49	26.47	22.99	1.61	2.58
UTBWC	Multilane	Mean	22,717	23,799	6.16	3.84	38.07	25.72	14.94	9.72	1.63	0.66	5.51	3.96	0.14	0.10
UTBWC	Two-Lane	Min	2,704	2,000	6.00	2.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
UTBWC	Two-Lane	Max	11,216	7,873	8.00	4.00	47.62	142.86	23.81	71.43	1.85	1.23	23.81	7.25	0.54	0.00
UTBWC	Two-Lane	Mean	5,730	4,536	7.05	2.95	3.10	3.47	1.80	1.80	0.06	0.02	0.75	0.39	0.01	0.00
UTBWC on PCC	Freeway	Min	22,625	23,000	8.00	1.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
UTBWC on PCC	Freeway	Max	107,250	113,000	9.00	2.00	701.09	2150.00	138.59	400.00	25.00	100.00	111.41	400.00	12.50	100.00
UTBWC on PCC	Freeway	Mean	66,031	67,988	8.27	1.73	37.67	72.15	9.19	12.58	2.87	5.54	6.92	12.11	1.09	2.60
UTBWC on PCC	Two-Lane	Min	6,420	5,325	6.00	4.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
UTBWC on PCC	Two-Lane	Max	6,938	8,000	6.00	4.00	5.85	41.67	2.92	41.67	0.00	0.00	2.78	0.54	0.00	0.00
UTBWC on PCC	Two-Lane	Mean	6,530	6,582	6.00	4.00	1.19	4.08	0.31	3.63	0.00	0.00	0.28	0.04	0.00	0.00

ELCSI = Evaluation of Low-Cost Safety Improvements
AADT = Average annual daily traffic
totrate = total crash rate per mi-year
injrate = injury crash rate per mi-year
wetrate = wet-road crash rate per mi-year;
rorrate = ROR crash rate per mi-year
wetrorrate = wet-road ROR crash rate per mi-year)
totrateb = total crash rate per mi-yr in before period
totratea = total crash rate per mi-yr in after period
injrate, wetrate, rorrate, wetrorrate = injury, wet-road, ROR, and wet-road ROR crash rates per mi-year, respectively
MAX = maximum
MIN = minimum
PCC = Portland cement concrete
OGFC = Open graded friction course
HMA = Hot mix asphalt
UTBWC = Ultrathin bonded wearing course

Table 12 . Summary statistics for North Carolina reference sites.

Avgshldwid = Average of left and right shoulder width
AADT = Average annual daily traffic
totrate = total crash rate per mi-year
injrate = injury crash rate per mi-year
wetrate = wet-road crash rate per mi-year;
rorrate = ROR crash rate per mi-year
wetrorrate = wet-road ROR crash rate per mi-year)
AC = Asphalt concrete
PCC = Portland cement concrete
Max = maximum
Min = minimum

California

Data for California were provided by the California Department of Transportation (Caltrans) and HSIS. The data provided by Caltrans included a list of potential treatment sites based on pavement maintenance and preservation projects from 2004 through 2007. The data provided by HSIS included geometric, traffic, and crash data. The data are linked together using the District Number, County Number, Route Number, and milepost variables.

Geometric Data

Geometric data were obtained from HSIS and included the following variables:

• Number of lanes.
• Median type.
• Median width.
• Access control.
• Terrain.
• Design speed.
• Rural versus urban environment.
• Surface type.
• Shoulder width.
• Paved width.
• Surface width.
• Lane width.
• Divided versus undivided.
• Roadway class.

Traffic Data

Traffic data in the form of AADT were obtained from HSIS from 2000 to 2009.

Crash Data

The crash data were provided by HSIS for 2000 to 2009. The compiled crash data contain many variables related to the location, time, and characteristics of each crash. The following notes relate to the exclusion or inclusion of crash data and definition of crash types:

• Exclude intersection-related mainline crashes (if int_rmp = "ramp intersection", "mid-ramp", "ramp entry", "ramp area/intersection street", "in intersection", or "outside intersection (non-state route) within 250 ft.").
• Excluded crashes where veh_invl = "animal".
• Excluded crashes where rdsurf = "snowy, icy".
• Injury crashes defined as those where severity = fatal, A class, B class, or C class injury.
• ROR crashes defined as those where miscact1 for any involved vehicle = "ran-off-road".
• Wet crashes defined as those where rdsurf="wet".

Construction Data

Construction history data were provided by Caltrans for 2000 to 2009 as a statewide list of capital improvement projects. This list includes resurfacing and other projects to verify that no other construction had been completed at the treatment site locations.

Treatment Sites

Sites on the initial list of treatment sites eligible for the study were removed if they underwent other significant construction during the study period. Other sites were moved when the roadway was a divided road but the treatment was only applied in one direction. In an attempt to eliminate any re-mileposted locations, those segments for which the post mile prefix variable "psmilprf" changed during the study period were removed. The "psmilprf" variable indicates that a segment has been reposted, realigned, or overlaps with another segment.

Reference Sites

Reference sites were identified by first identifying road segments in the same district and with the same route number as the treated sites. Then the various geometric and traffic variables were compared to ensure that the range of values for these variables were consistent. The construction data were linked to these data, and any segments that were resurfaced or had other construction works between 2000 and 2009 were removed. Similar to the treated sites, potential reference sites were excluded if the post mile prefix variable "psmilprf" changed during the study period.

Table 13 through table 15 provide summary statistics for the treatment and reference site data in California.

Table 13 . Summary statistics for California treatment site geometry.

Avgshldwid = Average of left and right shoulder width
AC = Asphalt
PCC = Portland cement concrete
Max = Maximum
Min = Minimum
OGFC = Open graded friction course
HMA = Hot mix asphalt
UTBWC = Ultrathin bonded wearing course

Table 14. Summary statistics for California treatment site AADT and crashes.

ELCSI = Evaluation of Low-Cost Safety Improvements
AADT = Average annual daily traffic
totrateb = total crash rate per mi-yr in before period
totratea = total crash rate per mi-yr in after period
injrate = injury rate
wetrate = wet-road rate
rorrate = ROR rate
wetrorrate = wet-road ROR crash rates per mi-yr
MAX = maximum
MIN = minimum
PCC = Portland cement concrete
OGFC = Open graded friction course
HMA = Hot mix asphalt
UTBWC = Ultrathin bonded wearing course

Table 15 . Summary statistics for California reference sites.

Avgshldwid = Average of left and right shoulder width
AADT = Average annual daily traffic
totrateb = total crash rate per mi-yr in before period
totratea = total crash rate per mi-yr in after period
injrate = injury rate
wetrate = wet-road rate
rorrate = ROR rate
wetrorrate = wet-road ROR crash rates per mi-yr
AC = Asphalt concrete
PCC = Portland cement concrete
Min = Minimum
Max = Maximum

Minnesota

Data for Minnesota were provided by the Minnesota Department of Transportation (MnDOT) and HSIS. The data provided by MnDOT included a list of potential treatment sites and data on recent paving projects. The data provided by HSIS included geometric, traffic, and crash data. The data are linked together using the District Number, Route Number, and milepost variables.

Geometric Data

Geometric data were obtained from HSIS and included the following variables:

• Shoulder width.
• Shoulder type.
• Surface width.
• Surface type.
• Median width.
• Median type.
• Surface type.
• Surface width.
• Number of lanes.
• Lane width.
• Roadway class.
• Rural versus urban environment.

Pavement Data

MnDOT provided resurfacing information during the period 2000 to 2010.

Traffic Data

Traffic data in the form of AADT and commercial vehicle AADT were obtained from HSIS for 2000 to 2010.

Crash Data

The crash data were provided by HSIS for 2000 to 2010. The compiled crash data contain many variables related to the location, time, and characteristics of each crash. The following notes relate to the exclusion or inclusion of crash data and definition of crash types:

• Include only nonintersection-related mainline crashes (if loc_type = 0, 1, 8, 9, 10, 11, 25, 90, 99 (various non-intersection categories)).
• Excluded crashes where acctype = "collision with deer" or "collision with other animal".
• Excluded crashes where rdsurf = "snow", "slush", "ice/packed snow", or "snow/slush".
• Injury crashes defined as those where severity = fatal, A class, B class, or C class injury.
• ROR crashes defined as those where accdigm = "ran off road left side" or "ran off road right side".
• Wet crashes defined as those where rdsurf="wet".

Construction Data

MnDOT used the highway pavement management system to verify construction history for the treatment sites and reference sites and ensure that no other pavement treatment had occurred at these locations during the study period.

Treatment Sites

The list of treatment sites was provided by MnDOT.

Reference Sites

Reference sites were identified by first identifying road segments in the same county and with the same route number as the treated sites. Then the various geometric and traffic variables were compared to ensure that the range of values for these variables were consistent. The pavement data were linked to these data, and any segments that were resurfaced or had other construction works between 2000 and 2010 were removed.

Table 16 through table 18 provide summary statistics for the treatment and reference site data in Minnesota.

Table 16. Summary statistics for Minnesota treatment site geometry.

Avgshldwid = Average of left and right shoulder width
AC = Asphalt concrete
PCC = Portland cement concrete
Min = minimum
Max = maximum
HMA = Hot mix asphalt

Table 17. Summary statistics for Minnesota treatment site AADT and crashes.

ELCSI = Evaluation of Low-Cost Safety Improvements
AADT = Average annual daily traffic
totrateb = total crash rate per mi-yr in before period
totratea = total crash rate per mi-yr in after period
injrate = injury rate
wetrate = wet-road rate
rorrate = ROR rate
wetrorrate = wet-road ROR crash rates per mi-yr
MAX = maximum
MIN = minimum
PCC = Portland cement concrete
HMA = Hot mix asphalt

Table 18. Summary statistics for Minnesota reference sites.

Avgshldwid = Average of left and right shoulder width
AADT =Average annual daily traffic
totrateb = total crash rate per mi-yr in before period
totratea = total crash rate per mi-yr in after period
injrate = injury rate
wetrate = wet-road rate
rorrate = ROR rate
wetrorrate = wet-road ROR crash rates per mi-yr
AC = Asphalt concrete
PCC = Portland cement concrete