Performance of Longitudinal Barriers on Curved, Superelevated Roadway Sections (2019)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

31 Crash Data Analysis Crash data analysis is useful in understanding the fre- quency and severity of crash events and, when the data is adequate, often allows situational, behavioral, and impact influences to be discerned. This research was initiated on the premise that there is a potential safety problem associ- ated with typical longitudinal barriers when they are installed on curved road sections. The safety problem is believed to be exaggerated when traffic moves at high speeds on roadway sections that are superelevated, which allows drivers to easily negotiate the curves at high speeds. While anecdotal informa- tion suggests there is a problem, its magnitude and extent are not clear. This effort began with investigations of several avail- able crash data sources to determine the extent and magnitude of the problem and to gain insights on barrier performance so that effective standards and guidance could be generated. The sources of crash data included the following: • National Automotive Sampling System/Crashworthiness Data System (NASS/CDS) • National Automotive Sampling System/General Estimate System (NASS/GES) • Fatality Analysis Reporting System (FARS) These are all publicly available datasets maintained by the National Highway Traffic Safety Administration (NHTSA). The NASS/CDS dataset is the most detailed, but has the fewest cases. Its basic data has been supplemented by road features data in many cases through independent research. The NASS/ GES is the least detailed or comprehensive, but it reflects the full range of crashes nationwide and as such allows global metrics of specific safety issues to be derived. The FARS data- set provides more detail than GES, but it is only focused on fatal crashes, and its coverage of crash features lacks the detail to directly isolate crashes into longitudinal barriers on high- speed CSRS. Other data sources exist that could provide useful infor- mation, but they fall short of the needs of this research. For example, if an agency were to have a good highway features inventory, then it would be possible to identify all the loca- tions where CSRS exist. However, specific data on road curva- ture and the specific starting and ending points of the curves are rare. It is even rarer that features such as superelevation are available in a database. If these locations could be refer- enced, it might not be possible to accurately determine all the crashes that occurred in proximity of the feature. The process of determining crash locations is not based on GPS coordi- nates in many places, and often the location data is inexact or inaccurate. Thus, it is not likely that a sound estimate of a crash problem associated with longitudinal barriers on high- speed CSRS can be defined. Analysis of crash data was undertaken to understand the conditions that influence crash potential and barrier perfor- mance. The parameters examined included road curvature, vehicle type, number of road lanes, vertical elevation (i.e., road profile), lighting condition, surface condition, weather condi- tion, and speed limit. Cases where the longitudinal barrier was installed on a curved section were compared with cases where the barrier was installed on a straight section. A sum- mary of the analyses using these three datasets is presented. 3.1 NASS/CDS Data Analyses The NASS/CDS database was used to identify critical factors related to longitudinal barrier performance when installed on CSRS. Datasets from 1988 through 2009 were included in the analysis. The data included a total of 186,465 cases during these 22 years. The datasets were weighted to be representative of the total number of crashes. Both weighted and unweighted data are presented here for comparison. A summary of the results from the analysis is presented below. The first step in the analysis was to reduce the dataset to the cases involving a longitudinal barrier as a first harmful event. A total of 4,489 vehicles were found to have the first harmful event as collision with a traffic barrier. These cases included C H A P T E R 3

32 both curved and straight roads. The variable “OBJCONT1” in the dataset was used to distinguish between “Concrete traffic barrier” (OBJCONT1 = 54), “Other traffic barrier including guardrails” (OBJCONT1 = 56), and “Bridge” (OBJCONT1 = 64). Table 3.1 shows the unweighted and weighted numbers of cases. Next, the cases were grouped by injury level—[Abbreviated Injury Scale (AIS)]—as shown in Table 3.2. Out of 4,110 cases involving longitudinal barriers (note that 379 cases have missing or unknown injury information), 993 (24%) resulted in serious injuries (AIS ≥ 3), 671 (16%) resulted in moderate injury (AIS = 2), 1,684 (41%) resulted in minor injury (AIS = 1), and 762 (19%) had no injury (AIS = 0). When using the weighting factor, the distribution showed 46% no injury, 44% minor injury, 6% moderate injury, and 4% serious injury. It should be noted that NASS/CDS data is biased toward more serious crashes. The data was then sorted based on curvature alignment (i.e., left or right curvature or straight). Table 3.3 data indicates that 27% of the vehicle crashes with barriers occurred on curved roads while the rest were on straight roads. For the weighted data, the portion of accidents that occurred on curved roads is one-third, while two-thirds occurred on straight roads. The number of cases for right and left curved roads is similar. The data for impacts with barriers was then used to compare curved versus straight road cases. Table 3.4 and Table 3.5 show the distribution of the vehicle class and the injury classification for curved and straight roads, respec- tively. The data indicates that 26% of the cases resulted in serious injuries for the curved roads compared with 23% for the straight roads for unweighted data. When the data is weighted, 3.7% resulted in serious injuries for curved roads, while the straight roads have 3.5% of the serious injuries. The data shows that the percentage of serious crashes on curved roads is similar to that of straight roads. It can be noted as well that the percentage of accidents by vehicle type is similar for curved and straight roads. Additional parameters examined in the NASS/CDS data included number of road lanes, vertical elevation, surface condition, lighting condition, weather condition, and speed limit. The data based on these parameters is listed in Table 3.6 through Table 3.11, respectively. The tables show the cases involving longitudinal barriers as the first harmful event for curved and straight roads on the left and only curved road cases on the right side for comparison. The following can be noted from the tables: • Crashes into longitudinal barriers on curved sections are more likely to occur on roads with fewer lanes (narrower roads). Table 3.6 shows that for one- and two-lane roads, the percentage of crashes on curved roads is higher than that of the combined (curved and straight) cases. The reverse is observed for roads with a higher number of lanes (wider roads). • Crashes into longitudinal barriers on curved roads are more likely to occur on uphill and downhill grades than on flat surfaces. For uphill and downhill roads, Table 3.7 shows that the percentage of crashes on curved roads is higher than that of the combined (curved and straight) road crashes. First Harmful Event Unweighted Weighted Number Percent Number Percent Concrete traffic barrier 2,066 46.02% 726,502 40.70% Other traffic barrier (includes guardrail) 1,773 39.50% 833,809 46.71% Bridge rail 650 14.48% 224,743 12.59% Total longitudinal barrier 4,489 100.00% 1,785,054 100.00% Table 3.1. First harmful event by barrier type (unweighted and weighted). Classified Abbreviated Injury Scale Unweighted Weighted Number Percent Number Percent AIS 2– 3,117 75.84% 1,586,921 96.39% AIS 3+ 993 24.16% 59,396 3.61% Total 4,110 100.00% 1,646,317 100.00% Missing and Unknown Cases (109 + 270) = 379 Note: AIS has six levels: 1: minor; 2: moderate; 3: serious; 4: severe; 5: critical; and 6: maximal. AIS 2– designates AIS 2 or less injury severity; AIS 3+ designates AIS 3 or higher injury severity. Table 3.2. Unweighted and weighted cases by AIS. Road Alignment Unweighted Weighted Number Percent Number Percent Curved Road 1,201 26.75% 576,478 32.29% Curved Road Right 601 13.39% 296,064 16.59% Curved Road Left 600 13.37% 280,414 15.71% Straight Road 3,288 73.25% 1,208,575 67.71% Total 4,489 100.00% 1,785,053 100.00% Table 3.3. Unweighted and weighted cases by road alignment.

33 Curved Roads Vehicle Class AIS 2– AIS 3+ Total AIS 2– AIS 3+ Total Unweighted Weighted Number Number Total Number Number Number Total Number Passenger Cars 562 199 761 (68.99%) 381,705 15,876 397,581 (74.56%) Pickups 97 38 135 (12.24%) 52,661 1,774 54,436 (10.21%) Utility Vehicles 117 40 157 (14.23%) 58,359 1,798 60,156 (11.28%) Vans 37 13 50 (4.52%) 20,692 379 21,072 (3.95%) Total 813 (73.71%) 290 (26.29%) 1,103 513,417 (96.28%) 19,828 (3.72%) 533,245 Unknown AIS and other vehicles cases = 399 unweighted for curved and straight roadways. Note: AIS has six levels: 1: minor; 2: moderate; 3: serious; 4: severe; 5: critical; and 6: maximal. AIS 2– designates AIS 2 or less injury severity; AIS 3+ designates AIS 3 or higher injury severity. Table 3.4. Unweighted and weighted cases by vehicle and AIS for curved roads. Straight Roads Vehicle Class AIS 2– AIS 3+ Total AIS 2– AIS 3+ Total Unweighted Weighted Number Number Total Number Number Number Total Number Passenger Cars 1,682 499 2,181 (73.02%) 776,569 29,147 805,716 (72.62%) Pickups 277 81 358 (11.99%) 156,789 4,732 161,521 (14.56%) Utility Vehicles 237 73 310 (10.38%) 98,547 3,425 101,972 (9.19%) Vans 93 45 138 (4.62%) 38,166 2,112 40,278 (3.63%) Total 2,289 (76.63%) 698 (23.37%) 2,987 1,070,072 (96.45) 39,416 (3.5%) 1,109,487 Unknown AIS and other vehicles cases = 399 unweighted for curved and straight roadways. Note: AIS has six levels: 1: minor; 2: moderate; 3: serious; 4: severe; 5: critical; and 6: maximal. AIS 2– designates AIS 2 or less injury severity; AIS 3+ designates AIS 3 or higher injury severity. Table 3.5. Unweighted and weighted cases by vehicle and AIS for straight roads. Table 3.6. Vehicle cases by number of lanes.

34 Table 3.7. Vehicle cases by vertical elevation. Table 3.8. Vehicle cases by road surface condition. Table 3.9. Vehicle cases by lighting condition. Table 3.10. Vehicle cases by weather condition.

35 • Crashes into longitudinal barriers on curved roads are more likely to occur on wet, snowy, and icy roads than on dry roads. In Table 3.8, the percentages of crashes on wet, snowy, and icy roads are higher on curved roads than that of the combined (curved and straight) roads. • Other parameters (lighting condition, weather condition, and posted speed) did not show significant effects on the crash distribution when comparing cases on curved roads with those on the combined (curved and straight) roads. Further analyses of this data for barriers on straight versus curved sections may be useful to isolate the differences. 3.2 NASS/GES Data Analyses Similar analysis was conducted using the NASS/GES data- base. Datasets from 1988 through 2009 were included in the analysis. The data included a total of 2,065,308 vehicle cases over these 22 years. The datasets were weighted to be rep- resentative of the total number of crashes. Both weighted and unweighted data are presented here for comparison. A summary of the results from the analysis is presented below. The first step in the analysis was to reduce the dataset to the cases involving a longitudinal barrier as a first harm- ful event. A total of 38,380 vehicles were found to have the first harmful event as a collision with a traffic barrier. These cases included both curved and straight roads. The variable “V_EVENT” in the dataset was used to distinguish between “Bridge structure” (V_EVENT = 34), “Guardrail” (V_EVENT = 35), and “Concrete traffic barrier or other longitudinal barrier” (V_EVENT = 36). Table 3.12 shows the unweighted and weighted number of cases. Next, the cases were divided into two injury groups using the maximum severity in the vehicle (MAX_VSEV) as shown in Table 3.13. The first group has no injury (O) (MAX_VSEV = 0), possible injury (C) (MAX_VSEV = 1), and non-inca- pacitating evident injury (B) (MAX_VSEV = 2). The second group has incapacitating injury (A) (MAX_VSEV = 3) and fatal injury (K) (MAX_VSEV = 4). Out of 38,380 cases involving longitudinal barriers, 5,581 (14.54%) resulted in incapacitating (A) and fatal injuries (K); 6,112 (15.92%) resulted in non-incapacitating evident injury (B); 5,854 (15.25%) resulted in possible injury (C); and 19,759 (51.48%) had no injury (O). When using the weighting factor, the distribution showed 163,418 (4.75%) resulted in incapacitating (A) and fatal injuries (K); 377,664 (10.98%) resulted in non-incapacitating evident injury (B); 497,291 (14.46%) resulted in possible injury (C); and 2,286,580 (66.55%) had no injury (O). Table 3.11. Vehicle cases by posted speed limit (km/h). First Harmful Event Unweighted Weighted Number Percent Number Percent Bridge structure 3,801 9.9% 340,126 9.89% Guardrail 19,771 51.51% 1,996,229 58.06% Concrete traffic barrier 14,808 38.58% 1,101,994 32.05% Total Longitudinal Barrier 38,380 100.00% 3,438,349 100.00% Table 3.12. First harmful event by barrier type (unweighted and weighted).

36 The data was then sorted based on curvature alignment. Table 3.14 indicates that 25.4% of the vehicle crashes occurred on curved roads while the rest occurred on straight roads. For the weighted data, the percentage of accidents occurring on curved roads is 26.71%. Approximately one-quarter of the crashes occurred on curved roads, while approximately three-quarters occurred on straight roads. The data was then used to compare curved versus straight road cases. Table 3.15 and Table 3.16 show the distribu- tion of the vehicle class and the injury classification for curved and straight roads, respectively. The data indicates that 15.74% of the cases resulted in serious injuries for the curved roads compared with 14.75% for the straight roads for the unweighted data. The data shows that the percentage of cases with serious injuries on curved roads is similar to that on straight roads. The percentage of acci- dents by vehicle type was found to be similar for curved and straight roads. Additional parameters examined included number of road lanes, vertical elevation, surface condition, lighting condi- tion, weather condition, and speed limit. The data based on these parameters is listed in Table 3.17 through Table 3.22. The tables show the cases involving longitudinal barriers as the first harmful event for curved and straight roads on the left and curved road cases on the right side for comparison. The following can be noted from the tables: • Crashes into longitudinal barriers on curved sections are more likely to occur on roads with fewer lanes (narrower roads). Table 3.17 shows that for one- and two-lane roads, the percentage of crashes on curved roads is higher than that for the combined (curved and straight) road cases. The reverse is observed for roads with a higher number of lanes (wider roads). • Crashes into longitudinal barriers on curved roads are more likely to occur on grades than on flat surfaces. For roads with grades, Table 3.18 shows that the percentage of crashes on curved roads is higher than that of the com- bined (curved and straight) road crashes. • Crashes into longitudinal barriers on curved roads are more likely to occur on wet, snowy, and icy roads than on dry roads. In Table 3.19, the percentages of crashes on wet, snowy, and icy roads are higher on curved roads than on the combined (curved and straight) roads. Classified Abbreviated Injury Scale Unweighted Weighted Number Percent Number Percent Non-incapacitating 31,725 82.66% 3,161,536 91.95% Incapacitating + K 5,581 14.54% 163,418 4.75% Missing and Unknown Cases 1,074 2.80% 113,395 3.3% Total 38,380 100.00% 3,438,349 100.00% Table 3.13. Unweighted and weighted cases by AIS. Road Alignment Unweighted Weighted Number Percent Number Percent Curved Road 9,750 25.40% 918,393 26.71% Straight Road 28,630 74.60% 2,519,956 73.29% Total 38,380 100.00% 3,438,349 100.00% Table 3.14. Unweighted and weighted cases by road alignment. Curved Roads Vehicle Class Nonincapacitating Nonincapacitating Incapacitating + K Incapacitating + K Total Total Unweighted Weighted Number Number Total Number Number Number Total Number Passenger Cars 5,114 826 5,940 (63.83%) 585,109 28,254 613,363 (70.04%) Pickups + Vans 1,033 168 1,201 (12.90%) 123,510 6,110 129,620 (14.8%) Utility Vehicles 644 158 802 (8.62%) 74,588 3,621 78,209 (8.93%) Buses 82 11 93 (1.00%) 9,763 1,125 10,888 (1.24%) Trucks 749 66 815 (8.75%) 24,196 1,243 25,439 (2.9%) Motorcycles 220 235 455 (4.90%) 10,520 7,655 18,175 (2.07%) Total 7,842 (84.26%) 1,464 (15.74%) 9,306 827,686 (94.52%) 48,008 (5.48%) 875,694 Other vehicle type, missing, and unknown injuries = 444 unweighted and 42,699 weighted for curved roadways. Table 3.15. Unweighted and weighted cases by vehicle and AIS for curved roads.

37 Straight Roads Vehicle Class Total Total Unweighted Weighted Number Number Total Number Number Number Total Number Passenger Cars 14,917 2,600 17,517 (63.66%) 1,564,498 74,033 1,638,531 (68.01%) Pickups + Vans 3,267 510 3,777 (13.73%) 367,179 16,891 384,070 (15.94%) Utility Vehicles 2,529 603 3,132 (11.38%) 268,516 12,482 280,998 (11.66%) Single Unit Truck 42 0 42 (0.15%) 1,091 0 1,091 (0.05%) Trucks 2,359 168 2,527 (9.18%) 61,236 2,391 63,627 (2.64%) Buses 201 21 222 (0.81%) 25,934 1,887 27,821 (1.16%) Motorcycles 143 156 299 (1.09%) 8,000 4,970 12970 (0.54%) Total 23,458 (85.25%) 4,058 (14.75%) 27,516 2,296,454 ( 95.32%) 112,654 (4.6%) 2,409,108 Missing and unknown injuries =1,114 unweighted and 110,848 weighted for curved roadways. Nonincapacitating Incapacitating + K Nonincapacitating Incapacitating + K Table 3.16. Unweighted and weighted cases by vehicle and AIS for straight roads. Table 3.17. Vehicle cases by number of lanes. Table 3.18. Vehicle cases by vertical elevation. Table 3.19. Vehicle cases by road surface condition.

38 • Other parameters (lighting condition, weather condition, and posted speed) did not show significant effects on the crash distribution when comparing cases on curved roads to the combined (curved and straight) road cases. 3.3 FARS Data Analysis Datasets from the FARS for the years 1982 through 2010 were used in the analysis. These datasets include only cases where one or more fatalities occurred. Years prior to 1982 were not included in the analyses because the variables in these datasets were less descriptive. The 29-year dataset con- sidered in the analysis included a total of 905,289 cases with at least one fatality. First, the data was truncated to include only the cases where a longitudinal barrier was the first harmful event. A total of 41,634 (4.60%) cases involved a longitudinal barrier as the first harmful event. A variable “HARM_EV” in the dataset was used to distinguish between “Bridge rail” (HARM_EV = 23), “Guardrail face” (HARM_EV = 24), and “Concrete barrier” Table 3.20. Vehicle cases by lighting condition. Table 3.21. Vehicle cases by weather condition. Table 3.22. Vehicle cases by posted speed limit (km/h).

39 (HARM_EV = 25). One additional variable, “Cable barrier” (HARM_EV = 57), was introduced after the year 2008, which was included in the “Guardrail face” (HARM_EV = 24) cate- gory in prior years. Cases with this variable (HARM_EV = 57) were added to the “Guardrail face” cases to be consistent with prior years. The data was sorted based on roadway alignment and listed in Table 3.23. There were 30,181 (72.49%) fatal crashes involving guardrail barriers, 6,591 (15.8%) involving concrete barriers, and 4,862 (11.7%) involving bridge rails. A total of 16,738 (40.2%) fatal crashes involving longitudi- nal barriers were on curved roads and the remaining 24,896 (59.8%) cases occurred on straight roads. Although crashes on curved roads account for only one-quarter of the total number crashes (based on the NASS/GES dataset), crashes on curved roads are more severe than crashes on straight roads. Table 3.24 and Table 3.25 show the distribution of fatal crashes based on the vehicle class and barrier type for curved and straight roads, respectively. Fatal crashes on curved roads are about half (50%) the number of fatal crashes on straight roads for passenger cars, pickups and vans, and utility vehicles. This figure of occurrence increases to 63% and 76% for Single Unit Trucks (SUT) and large and heavy trucks, respectively, when comparing fatal crashes on curved roads with those on straight roads. About 69% of fatal crashes involving longitudinal barriers on straight roads occur with guardrails, while the remaining crashes occur with bridge rails or concrete barriers. For curved roads, 77% of fatal crashes occur with guardrails and the remaining crashes occur with bridge rails or concrete barriers. The motorcycle data shows that the fatality numbers on curved roads are twice as high as those on straight roads. This observation is true for bridge rails and guardrails while the concrete barriers have similar values on curved and straight roads. Additional parameters examined included number of road lanes, vertical elevation, surface condition, lighting condi- tion, weather condition, and speed limit. The data based on these parameters is listed in Table 3.26 through Table 3.31, respectively. The tables show the cases involving longitudi- nal barriers as the first harmful event for curved and straight Road Alignment Curved Roads Straight Roads Straight and Curved Roads Number Percent Number Percent Number Percent Bridge Rail 1,508 3.62% 3,354 8.06% 4,862 11.68% Guardrail and Cable Barrier 13,008 31.24% 17,173 41.25% 30,181 72.49% Concrete Barrier 2,222 5.34% 4,369 10.49% 6,591 15.83% Total 16,738 40.02% 24,896 59.79% 41,634 100% Table 3.23. Vehicle crashes by road curvature and barrier type. Table 3.24. Vehicle crashes by vehicle and barrier type for curved roads.

40 Table 3.25. Vehicle crashes by vehicle and barrier type for straight roads. Table 3.26. Vehicle cases by number of lanes. Table 3.27. Vehicle cases by vertical elevation.

41 Table 3.28. Vehicle cases by road surface condition. Table 3.29. Vehicle cases by lighting condition. Table 3.30. Vehicle cases by weather condition. roads on the left and curved road cases on the right side for comparison. The following can be noted from the tables: • Fatal crashes into longitudinal barriers on curved sec- tions are more likely to occur on roads with fewer lanes (narrower roads). Table 3.26 shows that for one- and two- lane roads, the percentage of crashes on curved roads is higher than that for the combined (curved and straight) road cases. The reverse is observed for roads with a higher number of lanes (wider roads). • Fatal crashes into longitudinal barriers on curved roads are less likely to occur on grades than on flat surfaces. For roads with grades, Table 3.27 shows that the percentage of crashes on curved roads is lower than that of the combined (curved and straight) road crashes. This is opposite to what was found in the NASS/CDS and NASS/GES datasets. • Other parameters (surface condition, lighting condition, weather condition, and posted speed limit) did not show significant effects on the crash distribution when compar- ing cases on curved roads with the combined (curved and straight) road cases. 3.4 NCHRP Project 17-22 Data Analysis The dataset from NCHRP Project 17-22, “Identification of Vehicular Impact Conditions Associated with Serious Ran-Off-Road Crashes,” was also used to investigate barrier

42 performance when installed on curved road sections. This dataset supplements existing NASS/CDS data with additional information pertaining to the roadside such as side slope, roadway alignment, curvature, grade, profile, and roadside barrier characteristics (including post-crash measurements). The NCHRP 17-22 dataset was combined with the NASS/ CDS data and the cases that involved impacts into longitu- dinal barriers on curved road sections were identified. Forty crashes were found where the vehicle impacted a longitudinal barrier on a curved road section. These cases were summa- rized and information containing a description of the crash, a crash diagram, crash scene, barrier, and vehicle pictures, and road characteristics were extracted. Due to the small number of cases found in the database, no significant conclusions could be extracted from the anal- ysis. The cases were analyzed and grouped into three main categories: (1) barrier redirected the vehicle successfully; (2) barrier (including end terminal) caused rollover; and (3) other special cases. Out of the 40 cases, 32 were included in group 1, where the barrier redirected the vehicle back to the roadway. In 10 of these cases, the driver and occupants had no injuries (only property damage). In 16 of the remaining cases, the crash had no fatalities, but had injuries and prop- erty damage. In most of these cases, the vehicle was redi- rected by the barrier, crossed the travel lane, and remained upright. The remaining six cases had one or more fatali- ties. In these cases, the vehicle was redirected after the first impact, but impacted another barrier or an obstacle on the opposite side. The second category (other special cases) had seven cases where the barrier did not safely redirect the vehicle: • In two of the seven cases, the vehicle hit a concrete barrier and rolled after impact. One case had low injury and the other was fatal (occupant was unbelted). • In one case, the vehicle hit a W-beam bullnose and rolled over. • In one case, the vehicle hit a turned-down end terminal, which caused the vehicle to vault and roll over. • In one case, the vehicle broke through the end terminal and hit a tree. An AIS 3 injury was recorded. • In one case, the vehicle impacted a Thrie-beam barrier and rolled over. An AIS 2 injury was recorded. • In one case, the vehicle hit a W-beam barrier and rolled over. The occupant was unbelted and the crash was fatal. The last category had only one case where the vehicle hit the back of a Thrie beam barrier. The vehicle vaulted the Thrie- beam, continued into the opposite traffic lanes, impacted another Thrie-beam barrier, and came back into the traffic lanes. Two occupants died in the crash. 3.5 Data Analysis Summary Various datasets were analyzed to isolate a specific safety problem. The findings are summarized by dataset as follows: • NASS/CDS: This represented the most detailed set of data, albeit most of the data items focus on the impact and injury Table 3.31. Vehicle cases by posted speed limit (km/h).

43 severities. Data for 1988 through 2009 was analyzed for crashes into concrete barriers, other guardrail, and bridge rails. Of these, 46.7% hit an “other barrier” guardrail and 41% hit a concrete barrier. The analysis involved 4,489 U.S. barrier impact cases, which when weighted reflected 1,785,054 crashes. The weighted results indicated the following: – Serious injuries (or worse) occurred in 4% of the crashes, moderate injuries in 6%, minor injuries in 44%, and no injuries in 46% of the crashes based on an AIS scale. – Crashes involved passenger cars 74.6% of the time. – Crashes occurred on curved roads 32.3% of the time. – Other analyses isolated frequency of crashes by light- ing conditions, road surface condition, number of lanes, and grade. It was noted that crashes into barriers were more likely to occur for narrow roads (two or fewer lanes), on wet or icy pavements, and on uphill or down- hill grades. – Because the data did not include measures of the radii, superelevation, or shoulder features for any of the crash locations, it was not possible to isolate any specific safety problems for barriers on CSRS. • NASS/GES: This more general dataset provided less detail on crashes and included some different data items. The dataset included 2,065,308 vehicle cases over the same 22-year period: – The data for first harmful event includes 38,380 cases of collision with a traffic barrier. Of these, 58% hit a guardrail and 32% a concrete barrier. The differences in percentages can be attributed to variations in data definitions. – Incapacitating injuries or fatalities occurred in 4.75% of the crashes, and non-incapacitating injuries in 91.9% of the crashes. – Crashes occurred on curved roads 26.7% of the time. – The analyses of other conditions led to similar insights related to crashes with barriers as above. – The data was also insufficient to isolate any safety prob- lems for barriers on CSRS. • FARS Data: This dataset was compiled for all fatal crashes and the dataset contained 905,289 cases for the years 1982 through 2010. Fatal crashes with the first harmful event being hitting a longitudinal barrier were isolated. The bar- riers types included bridge rail, guardrail face, concrete, guardrail end, and cable barriers (after 2008): – Crashes with guardrail face were represented in 76.3% of the cases and concrete barrier in 15.7% of the cases. – Crashes occurred on curved roads 40.3% of the time. – The analyses of other conditions led to similar insights related to crashes with barriers as above, but it was noted that fatalities were more likely to occur on grades. – The data was also insufficient to isolate any safety prob- lems for barriers on CSRS. • NCHRP 17-22 Data: This project created a data of CDS cases of longitudinal barrier impacts from three studies. Supplemental data reflecting roadway conditions was added to the 700+ cases. It was hoped that this supple- mental data would provide some relevant roadside crash cases. Forty usable cases were isolated, but only seven were related to CSRS conditions. The Team decided that this was too few to derive any meaningful insights. The analyses of crashes revealed what is generally known, that is, that crashes occur more frequently on curves than on tangent sections. The available data does not, however, allow much mining into the effects of the various design features associated with basic curves, much less with super- elevated curves. Given that vehicles are known to leave the road on curves more frequently (e.g., due to loss of side fric- tion, visibility issues), it is appropriate to consider whether the barriers deployed for these situations are providing compara- ble safety. The available sources of crash data do not typically include sufficient details about the roadway curvature or the barrier type, dimensions, or placement relative to the shoulder to allow safety performance to be analyzed. Further, many state DOTs cannot link their roadway geometry and barrier inven- tories to crash data. Various sources of data were explored, but none were found to offer useful insights on any variations on the safety of longitudinal barriers installed on CSRS. It was therefore concluded that available crash data would not be able to provide specific insights on whether typical longitudinal barriers function similarly on CSRS as they do for tangent sections of roadway. The absence of specific data in police crash reports on the types of barrier impacted, the nature of the curve, or details about the shoulder configu- ration made it necessary to use other means to analyze the safety performance of barriers used in CSRS situations.