National Academies Press: OpenBook
« Previous: Chapter 2 - Literature Review
Page 10
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 10
Page 11
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 11
Page 12
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 12
Page 13
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 13
Page 14
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 14
Page 15
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 15
Page 16
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 16
Page 17
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 17
Page 18
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 18
Page 19
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 19
Page 20
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 20
Page 21
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 21
Page 22
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 22
Page 23
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 23
Page 24
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 24
Page 25
Suggested Citation:"Chapter 3 - Study Approach." National Academies of Sciences, Engineering, and Medicine. 2010. Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes. Washington, DC: The National Academies Press. doi: 10.17226/14448.
×
Page 25

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.

10 3.1 General To accomplish the study objectives, the following major tasks were undertaken in this study: • Identify data needs • Evaluate data collection alternatives • Develop data collection protocol • Conduct supplemental data collection, manual review, and reconstruction • Create relational database • Incorporate data from previous studies into database Details of these tasks are presented in the following sections. The database was then analyzed to address the study objectives and the results are presented in Chapter 4. Finally, a proposed implementation plan for a long-term data collection effort was developed and outlined in Chapter 5. 3.2 Data Needs The primary goal to be achieved under the current study is to identify the distribution of impact conditions associated with serious injury and fatal ran-off-road accidents, includ- ing speed, angle, and vehicle orientation at impact. It is hoped that this information can then be used to select impact con- ditions to be used in full-scale crash testing of roadside hard- ware. In order to address this issue, the needed data elements were identified and are listed in Table 1. The data elements are categorized as available from: 1. Basic NASS CDS data. These data elements are already available as part of the basic CDS data. 2. Supplemental field data collection. These data elements will require field data collection. 3. Reconstruction. These data elements will require recon- struction of the crashes. The data collection plan presented in this chapter covers the data elements requiring supplemental field data collection and reconstruction. 3.3 Data Collection Alternatives Three basic alternatives were considered for the data col- lection effort in the current study: 1. New data collection system 2. Prospective special study under the NASS CDS program 3. Retrospective supplemental data collection for existing NASS CDS cases More detailed discussions of these alternatives are presented below. 3.3.1 New Data Collection The first alternative was to establish a totally new data col- lection system. The major activities required in the setup of a new data collection system at multiple sites include, but are not limited to, the following: • Establish data collection teams. This would require hiring of new personnel, establishing and furnishing the offices, purchasing the necessary equipment for conducting crash investigation, etc. • Train investigators in the basics of in-depth level crash investigation. The newly hired investigators would need to be trained extensively to acquire the required level of exper- tise, including both classroom and on-the-job training. This training would need to be extensive and comparable to what is used with the NASS CDS program. • Develop procedures for obtaining authorization to collect medical records. C H A P T E R 3 Study Approach

11 Variable Availability Case Screening Criteria Area type - PSU 1 Crash type - Single-vehicle, ran-off-road crashes 1 Vehicle type - Passenger vehicles only 1 Completeness of data on key variables 1 Injury severity - Serious and fatal injury 1 Variables of Primary Interest: Encroachment conditions at point of departure - Action prior to leaving travelway 1 - Speed 3 - Angle 3 Pre-impact vehicle trajectory - Vehicle path 3 - Maximum lateral extent of encroachment 3 - Total longitudinal distance 3 General impact data - Impact sequence 1 - Object struck 1 - Rollover occurrence 1 - Post-impact trajectory 3 Impact conditions – first harmful event - Impact speed 3 - Impact angle 3 - Vehicle orientation 3 Impact conditions – most harmful event - Impact speed 3 - Impact angle 3 - Vehicle orientation 3 Driver action - Evasive action 1 - Steering – vehicle path 3 - Braking 3 Controlling Variables: Highway type - Functional class 2 - Roadway type 1 - Speed limit 1 Travelway characteristics - Number of lanes 2 - Lane width 2 - Horizontal curvature - Point of departure and maximum 2 - Vertical grade - Point of departure and maximum 2 Roadside characteristics - Shoulder type and width 2 - Roadside slopes – widths and rates of slopes 2 - Median type, width, and slope 2 Traffic characteristics - ADT 2 - Percent truck 2 Struck object characteristics - Object type 2 - Impact performance 3 Vehicle characteristics - Type 1 - Make and model 1 - Curb weight 1 - Vehicle damage 1 - Occupant compartment deformation and intrusion 1 Highest occupant injury severity - Abbreviated Injury Scale (AIS) 1 - Police Injury Code (PIC) 1 EDR data 1 Table 1. Data needs for current study. (continued on next page)

12 • Establish cooperation with local agencies. This would include law enforcement agencies for the notification sys- tem, vehicle towing and repair facilities for access to the involved vehicles, hospitals and clinics for medical records/ information on injury severity, and transportation agen- cies for highway-related information. • Establish quality control procedures. To assure proper data collection in terms of validity and accuracy, appropriate quality control procedures would need to be established, similar to the Zone Centers in the NASS program. After the data collection system was established, additional activities would be required to establish the specific data col- lection effort, including: • Develop data collection protocol. The field forms and accompanying coding and instruction manuals, data col- lection procedures, data submission processes, and qual- ity control procedures would have to be developed for the specific data collection effort. • Train investigators in specific data collection effort. The investigators would have to be trained in the details of the specific data collection effort. This would be in addition to the basic training mentioned above. • Conduct pilot study. A pilot study would have to be con- ducted to work out any unforeseen problems in the data collection protocol. It is evident from the above discussion that the alternative of establishing a new data collection system was not a viable option for this study due to funding constraints. The startup costs would be prohibitive for such a short-term data collec- tion effort. However, this remains a viable alternative for a long-term data collection effort. 3.3.2 Prospective NASS CDS Special Study The second alternative was to establish a special study under the NASS CDS program. The special study would be prospec- tive in nature (i.e., data would be collected on new crashes) and could be within sample (i.e., only crashes that are already sampled under the NASS CDS program would be eligible) or outside of sample (i.e., all crashes are eligible). Again, this alternative is not viable for this study due to time and funding constraints. First, it will take a minimum of 12 to 18 months to set up a special study under the NASS CDS pro- gram. Second, this assumes that the NASS CDS program can accommodate a new special study on short notice, which is rarely the case. Because the CDS system itself requires a cer- tain number of crashes to be investigated and the researchers can handle only so many crashes (1 1⁄2 to 2 cases per week per researcher), the ability of the system to conduct special studies is limited. This limitation can be overcome by hiring new investigators specifically to handle the special study, such as in the case of the special study on large-truck crash causation. The addition of new investigators is not as time consuming or costly as establishing new data collection teams, but would still require more time and funding than available for the cur- rent study. However, this alternative remains viable for a long- term data collection effort. 3.3.3 Retrospective Supplemental Data Collection The third alternative was to conduct a retrospective study using previously investigated NASS CDS cases. This approach was similar to that successfully used in NCHRP Project 17-11 and the FHWA Rollover Study. In those studies, single-vehicle, ran-off-road crashes were selected from 1997 through 1999 NASS CDS cases. Since NASS CDS cases are oriented toward vehicle crashworthiness and occupant injury and lack details pertaining to the highway and roadside characteristics, sup- plemental field data collection and manual review and recon- struction of the cases were used to fill in the data gaps. A total of 559 cases were sampled under these studies. This approach can be implemented within a short period of time since it involves only existing NASS CDS cases. Sup- plemental field data collection protocol and manual review Variable Availability Variables of Secondary Interest: Time - Day of week 1 - Time of day 1 Environmental conditions - Light 1 - Weather 1 *Legends for Data Availability: 1. Existing NASS CDS data 2. Supplemental field data collection 3. Reconstruction Table 1. (Continued).

13 and reconstruction procedures had already been developed and field investigators at the Primary Sampling Units (PSUs) and the Zone Center personnel were already familiar with the protocol and procedures. Thus, this approach could be easily implemented for this study within the time and funding con- straints. Also, this alternative would allow cases from the pre- vious studies to be incorporated into the database with the new cases collected under this study. This third alternative of retrospective supplemental field data collection and manual review and reconstruction of exist- ing NASS CDS cases was, therefore, selected for this study. However, it should be noted that NHTSA had changed its pol- icy, starting with the 2003 data, to keep police accident reports in the file for only one year. This in effect eliminates the loca- tion information on existing NASS CDS cases. Thus, this alter- native of retrospective supplemental field data collection and manual review and reconstruction of existing NASS CDS cases is no longer a viable option. For the long-term data collection effort in the future, only the alternatives of a new data collec- tion effort or a special study under the NASS CDS system could be considered. 3.4 Data Collection Protocol As discussed previously, the plan for the current study was based on a retrospective supplemental data collection approach. This retrospective approach involved collecting supplemental field data and manual review and reconstruction of existing NASS CDS cases. The major components of the data collection protocol are summarized as follows: • Sampling plan • Supplemental field data collection • Manual review of sampled cases • Reconstruction of crashes to estimate impact speed Brief descriptions on activities pertaining to the supple- mental field data collection are presented in this section. 3.4.1 Sampling Plan As discussed previously, a similar retrospective supplemen- tal field data collection approach was used in two previous studies: NCHRP Project 17-11 and the FHWA Rollover Study. Supplemental field data were collected on NASS CDS cases from 1997 through 1999 in these two studies, as follows: • NCHRP Project 17-11 – 1997: 138 cases – 1998: 200 cases • FHWA Rollover Study – 1999: 221 cases The scope of the supplemental data collection effort for this study was, therefore, selected to include 2000 and 2001 NASS CDS cases. To maintain consistency among the three studies, the sampling criteria remained the same as the two previous studies. The sampling criteria included the follow- ing parameters: • Area type—rural and suburban. Urban PSUs were excluded from the sample because urban roadways tend to have lower speed limits and the roadsides are typically cluttered with fixed objects. More importantly, inspections at urban crash sites are generally less detailed with a higher percentage of incomplete data due to hazardous working conditions and traffic congestion. • Single-vehicle, ran-off-road crashes. Only single-vehicle, ran-off-road crashes were included in the sample. Single- vehicle crashes that occurred on the roadway, or involving parked vehicles, animals, or pedestrians, were excluded since the nature of the crashes is different from that of a ran-off- road crash. Similarly, multiple-vehicle crashes were excluded from the sample. • Passenger-type vehicles. Only passenger-type vehicles, i.e., passenger cars and light trucks with a gross vehicle weight (GVW) of less than 4,536 kg (10,000 lbs), were included in the NASS CDS sample. Heavy trucks, i.e., single-unit trucks with higher GVW and tractor-trailers, present very differ- ent problems than passenger vehicles. Also, reconstruction of crashes involving heavy trucks is much more difficult than those involving passenger-type vehicles. • Speed limit of 72 km/h (45 mph). Only crashes that occurred on highways with speed limits of 72 km/h (45 mph) or higher were included. Low-speed roadways tend to have lower design standards and have crash characteristics that are sig- nificantly different from those of high-speed highways. Thus, it is not desirable to mix crashes from both low-speed and high-speed highways. • Complete vehicle inspection, vehicle trajectory, and injury severity data. It would not be possible to reconstruct crashes without vehicle inspection and trajectory data, and those crashes would be of little interest to the proposed study. Thus, only crashes with complete vehicle inspection and trajectory data were included. Also, the emphasis of the study was on serious and fatal injury crashes, so the injury severity should, therefore, be known for the sampled cases. Table 2 shows a breakdown of the 2000 and 2001 CDS cases by the first four sampling criteria. In year 2000, there were a total of 4,307 cases, 2,929 (68.0%) of which occurred in the 16 rural and suburban PSUs, and 1,518 (51.8%) of which occurred on highways with speed limits above 72 km/h (45 mph). Of these crashes, 603 (39.7%) were single-vehicle, ran-off-road crashes. In year 2001, there were a total of

14 4,090 cases, 2,833 (49.3%) of which occurred in the 16 rural and suburban PSUs, and 1,500 (52.9%) of which occurred on highways with speed limits above 72 km/h (45 mph). Of these crashes, 593 (39.5%) were single-vehicle, ran-off-road crashes. Combining data from the two years, there were a total of 1,196 eligible cases that occurred in rural and suburban PSUs on highways with speed limits above 72 km/h (45 mph), and involving single-vehicle, ran-off-road crashes. As shown in Table 3, of the 1,083 eligible cases with known injury severity, 348 (32.13%) resulted in serious to fatal injuries [Abbreviated Injury Scale (AIS) ≥ 3], 229 (21.14%) resulted in moderate injury (AIS = 2), 385 (35.55%) resulted in minor injury (AIS = 1), and 121 (11.17%) incurred no injury (AIS = 0). However, it should be noted that the sam- pling scheme for NASS CDS is biased toward the more seri- ous crashes. When the cases are weighted according to the sampling scheme, the distribution of injury severity is very different: 43.64% no injury, 40.15% minor, 8.32% moderate; and 7.90% serious to fatal injury. Thus, all analyses shown herein show both unweighted and weighted frequencies and percentages. Table 4 shows the distribution of the eligible cases by the number of lanes. The vast majority of the cases, 998 (83.44%), occurred on highways with two or three lanes. Another 38 (3.18%) occurred on one-lane roadways (i.e., ramps). The remaining 160 cases (13.38%) occurred on highways with four or more lanes. The weighted distributions are similar, 3.45% for one lane, 82.02% for two or three lanes, and 14.53% for four or more lanes. The similarity between the unweighted and weighted percentages suggests that the severity of crashes is similar for different highway types, though slightly higher for highways with two or three lanes. Table 5 shows the distribution of the eligible cases by vehicle type. Passenger cars accounted for the majority, 696 (58.19%), of the eligible cases, followed by pickup trucks, 247 (20.65%), and sport utility vehicles, 198 (16.56%). The weighted distri- butions show a higher percentage for passenger cars (64.60%) and lower percentages for the other vehicle types. This sug- Year Total No. of Cases 16 Rural and Suburban PSUs Speed Limit 45 mph Passenger Vehicle/ Single-Vehicle Ran- Off-Road Crashes 2000 4307 2929 1518 603 2001 4090 2833 1500 593 Total 8397 5762 3018 1196 Table 2. Breakdown of 2000 and 2001 NASS CDS cases by screening criteria. Unweighted Weighted Abbreviated Injury Scale Number Percentage Number Percentage No Injury (0) 121 11.17 280,985 43.64 Minor Injury (1) 385 35.55 258,559 40.15 Moderate Injury (2) 229 21.14 53,554 8.32 Serious Injury (3) 175 16.16 23,074 3.58 Severe Injury (4) 80 7.39 20,846 3.24 Critical Injury (5) 68 6.28 5,190 0.81 Maximum Injury (6) 25 2.31 1,712 0.27 Total 1,083 100.00 643,920 100.00 * Missing Cases = 113 unweighted (45,970 weighted) Table 3. Eligible cases by maximum abbreviated injury scale.

15 gests that a higher proportion of crashes involving passenger cars had lower injury severity. The final screening criteria include documentation of vehi- cle trajectory, complete vehicle inspection, and known injury severity data. Of the 1,196 eligible cases, only 437 (36.54%) met all three criteria. Table 6 shows the distribution of these 437 cases by PSU. Note that three of the PSUs (4, 73, and 81) do not have any complete cases. Two other PSUs (5 and 43) have only two and four complete cases, respectively. Also, three other PSUs (8, 9, and 75) have less than 20 complete cases. Since the targeted sample size was only 400 cases, it was decided to eliminate seven PSUs (4, 5, 8, 9, 43, 73, and 81) from the sampling due to overly small number of cases, which ren- ders the data collection effort inefficient. The number of sam- ple cases was thus reduced from 437 to 404 cases. Distribution of the 404 sampled cases by PSU is also shown in Table 6. In order to make sure that the sampled cases are reasonably representative of the NASS CDS cases, and thus the overall crash population nationwide, a check was conducted on a few key variables, including highest injury severity, number of lanes, and vehicle type. As shown in Table 7, of the 404 sampled cases, 139 (34.41%) resulted in serious to fatal injuries (AIS ≥ 3), 94 (23.27%) in moderate injury (AIS = 2), 142 (35.15%) in minor injury (AIS = 1), and 29 (7.18%) with no injury (AIS = 0). The dis- tribution of the sampled cases was quite similar to that of the eligible cases shown previously in Table 3 with a slight decrease in the percentage of crashes with no injury. The same is true for the weighted distributions. Table 8 shows the distribution of the eligible cases by num- ber of lanes. The dominance of highways with two or three lanes is even more pronounced for the sampled cases with the weighted percentages, increasing from the 82.02% for the eli- gible cases (see Table 4) to 90.36% for the sampled cases. The proportion of crashes on one-lane roadways also increased slightly. Correspondingly, the weighted percentages of crashes on highways with four or more lanes dropped from 14.53% to only 5.69%. This drop in the proportion of cases occurring on highways with four or more lanes is not surprising given that only three of the sampled PSUs are in suburban areas, where multi-lane facilities are more common. As shown in Table 9, the distributions of the sampled cases by vehicle type are similar to those of the eligible cases, shown previously in Table 5. Passenger cars accounted for about 65% for both the eligible and sampled cases. The proportions of sport utility vehicles and vans/minivans decreased some- what for the sampled cases while the percentage of pickup trucks increased. Overall, the distributions of these key variables for the sampled cases were reasonably similar to those of the eligible Unweighted Weighted Number of Lanes No. Percentage No. Percentage 1 38 3.18 23,809 3.45 2 & 3 998 83.44 565,855 82.02 4 160 13.38 100,227 14.53 Total 1,196 100.00 689,891 100.00 Table 4. Eligible cases by number of lanes. Unweighted Weighted Vehicle Type No. Percentage No. Percentage Passenger Car 696 58.19 445,651 64.60 Sport Utility Vehicle 198 16.56 103,434 14.99 Van/Minivan 55 4.60 26,138 3.79 Pickup Truck 247 20.65 114,668 16.62 Total 1,196 100.00 689,891 100.00 Table 5. Eligible cases by vehicle type.

16 Eligible Cases Complete Cases Sampled Cases Area Type PSU No. Percentage No. Percentage No. Percentage 2 59 4.93 31 7.09 31 7.67 4 35 2.93 0 0.00 0 0.00 11 145 12.12 59 13.50 59 14.60 13 130 10.87 86 19.68 86 21.29 43 100 8.36 4 0.92 0 0.00 48 114 9.53 40 9.15 40 9.90 76 109 9.11 41 9.38 41 10.15 78 85 7.11 43 9.84 43 10.64 Rural Subtotal 777 64.97 304 69.57 300 74.26 5 16 1.34 2 0.46 0 0.00 8 28 2.34 15 3.43 0 0.00 9 64 5.35 12 2.75 0 0.00 12 94 7.86 47 10.76 47 11.63 45 60 5.02 38 8.70 38 9.41 73 48 4.01 0 0.00 0 0.00 75 57 4.77 19 4.35 19 4.70 81 52 4.35 0 0.00 0 0.00 Suburban Subtotal 419 35.03 133 30.43 104 25.74 Total 1,196 100.00 437 100.00 404 100.00 Table 6. Eligible, complete, and sampled cases by primary sampling unit. Unweighted Weighted Abbreviated Injury Scale Number Percentage Number Percentage N o Injury (0) 29 7.18 88,968 41.15 Minor Injury (1) 142 35.15 87,723 40.58 Moderate Injury (2) 94 23.27 16,063 7.43 Serious Injury (3) 69 17.08 11,387 5.27 Severe Injury (4) 30 7.43 9,966 3.68 Critical Injury (5) 32 7.92 3,056 1.41 Maximum Injury (6) 8 1.98 1,024 0.47 Total 404 100.00 218,187 100.00 Table 7. Sampled cases by highest injury severity.

17 cases, given that the sampled cases are not truly a represen- tative sample of the eligible cases. Rather, it is a sample of convenience to make sure that the sampled cases have com- plete documentation of the vehicle trajectory, vehicle inspec- tion, and information on injury severity. 3.4.2 Supplemental Field Data Collection Data elements requiring supplemental field collection are shown in Table 10. The protocol for the supplemental field data collection effort was developed, including the field forms and the accompanying coding and instruction manuals. The field forms were used by the PSU investigators during the actual data collection while the manual provided definitions of the data elements, field data collection procedures, and coding instructions. Note that given the retrospective nature of the data col- lection approach, there was an implicit assumption that the data elements would not change significantly with time. This is a reasonable assumption for most of the supplemental data elements, such as roadway, traffic, and roadside character- istics. As for the struck-object characteristics, there was an additional assumption that any damaged objects would be replaced in kind, i.e., the replaced object or feature would have the same characteristics as the original that was damaged. The investigators would compare the site and struck-object characteristics at the time of supplemental data collection to those at the time of the crash, using photographs from the case files to make sure that these assumptions were accurate. Cases in which the site and/or struck-object/feature charac- teristics had been changed significantly would be deleted from the sample. Unweighted Weighted Number of Lanes No. Percentage No. Percentage 1 14 3.47 8,531 3.95 2 & 3 356 88.12 195,360 90.36 4 34 8.42 12,296 5.69 Total 404 100.00 216,187 100.00 Table 8. Sampled cases by number of lanes. Unweighted Weighted Vehicle Type No. Percentage No. Percentage Passenger Car 212 52.48 140,692 65.08 Sport Utility Vehicle 64 15.84 67,169 11.25 Van/Minivan 23 5.69 3,502 1.62 Pickup Truck 105 25.99 45,511 21.05 Total 404 100.00 256,874 100.00 Table 9. Sampled cases by vehicle type. Highway type - Functional class Highway characteristics - Number of lanes - Lane width - Horizontal curvature - Point of departure and maximum - Vertical grade - Point of departure and maximum Roadside characteristics - Shoulder type and width - Roadside slopes – widths and rates of slopes - Median type, width, and slope Traffic characteristics - ADT - Percent truck Struck-object characteristics - Object type - Impact performance Table 10. Data elements requiring supplemental field data collection.

18 There were two sets of field data collection forms: • Supplemental Highway Data Collection Form • Object Struck Data Collection Form A complete copy of the field forms and the accompanying coding and instruction manuals are included as Appendix C and will not be repeated here. The Supplemental Highway Data Collection Form was completed for each sampled case. The form contains 20 data elements under four general headings: • Case Identification: 1. Year 2. Primary Sampling Unit 3. Case Number-Stratum • General Highway Data: 4. Land Use 5. Class Trafficway 6. Access Control 7. Average Lane Width 8. Roadway Alignment at Point of Departure 9. Radius of Curve 10. Roadway Profile at Point of Departure 11. Vertical Grade • Roadside Data: 12. Curb Presence 13. Curb Height 14. Shoulder Type 15. Shoulder Width • Slope Data: 16. Roadside Cross Section at Point of Departure 17. Number of Slopes 18. Lateral Offset to Beginning of Slope 19. Rate of Slope 20. Width of Slope An Object Struck Data Collection Form was completed for each object involved in the crash. The form contains seven data elements under four general headings: • Case Identification: 1. Year 2. Primary Sampling Unit 3. Case Number-Stratum • General Struck Object Data: 4. Impact Number 5. Object Type 6. Material • Dimensions of Struck Object—annotation • Photography: 7. Photographs Taken? Due to the large number of potential roadside objects and features, the variables are necessarily very general without spe- cific details. Instead, investigators were asked to provide anno- tations or descriptions and photographs of the struck object. Since the data collection protocol was similar to that of NCHRP Project 17-11 and the FHWA Rollover Study, the Zone Center staff and PSU investigators were already famil- iar with the data collection protocol. Thus, the data collection experienced little problem or difficulty. The actual field data collection was conducted by PSU investigators under the direction of the Zone Centers: Veridian Corporation for Zone Center 1 and KLD Associates for Zone Center 2. After a qual- ity check was conducted by Zone Center personnel for accu- racy, the completed data were forwarded to KLD Associates, which was a subcontractor for this study. The supplemental field data were then combined with the regular NASS data in the manual review of the cases. 3.4.3 Manual Review of Sampled Cases Additional data elements not available from the computer- ized data file or supplemental field data collection were gleaned from manual review of hard copies (in electronic form) and reconstruction of the sampled cases. The data elements coded from this manual review are shown in Table 11. Part of the review included verification of data elements that were already coded under existing NASS CDS or supplemental data collec- tion, such as: • Highway data—highway type, number of lanes, divided/ undivided, presence/absence of shoulder, and impact sequence • Roadside feature impacted—guardrail, tree, ditch, etc. • Driver input—steering and/or braking Encroachment conditions at point of departure - Speed - Angle Pre-impact vehicle trajectory - Vehicle path - Maximum lateral extent of encroachment - Total longitudinal distance General impact data - Post-impact trajectory Impact conditions – first harmful event - Impact speed - Impact angle - Vehicle orientation Impact conditions – most harmful event - Impact speed - Impact angle - Vehicle orientation Driver action - Steering – vehicle path - Braking Table 11. Data elements requiring reconstruction.

19 The main function of the manual review was to conduct detailed reconstruction of the crashes to estimate parameters such as: • Vehicle encroachment conditions—angle and orientation • Vehicle trajectory after encroachment—vehicle path • Impact conditions—angle and orientation • Impact performance of struck roadside safety feature With the exception of the reconstruction of impact speed, which was performed by the project staff, the manual review and reconstruction were conducted by Zone Center personnel from KLD Associates. Two reconstruction coding forms were designed specifically for coding of these manual review and reconstruction data elements: one for the first event or impact, and one for subsequent events or impacts. Copies of the recon- struction coding forms and the accompanying coding and instruction manual are shown in Appendix C and will not be repeated here. Zone Center personnel were trained on the manual review procedure and the coding of the data elements. Under the reconstruction coding form for the first event, there are 20 data elements under six general categories: • Case Identification: 1. Year 2. Primary Sampling Unit 3. Case Number-Stratum • Encroachment Data: 4. Departure Angle 5. Vehicle Heading Angle • Vehicle Trajectory Data: 6. Driver Action 7. Longitudinal Distance of Travel 8. Number of Trajectory Profile Points 9. Lateral Offset of Trajectory Profile Points 10. Maximum Lateral Offset • Impact Conditions—First Event: 11. Location of Impact 12. NASS CDS Data 13. Impact Angle 14. Vehicle Heading Angle at Impact • Separation Conditions—First Event: 15. Location of Separation 16. Separation Angle 17. Vehicle Heading Angle at Separation • Subsequent Event/Final Rest 18. Subsequent Event 19. Location of Final Rest 20. Vehicle Heading Angle at Final Rest Under the reconstruction coding form for subsequent events, there are also 20 data elements under six general categories: • Case Identification: 1. Year 2. Primary Sampling Unit 3. Case Number-Stratum • Current Event Identification: 4. Current Event Number 5. Current Event Location • Vehicle Trajectory Data: 6. Driver Action 7. Longitudinal Distance of Travel 8. Number of Trajectory Profile Points 9. Lateral Offset of Trajectory Profile Points 10. Maximum Lateral Offset • Impact Conditions—Current Event: 11. Location of Impact 12. NASS CDS Data 13. Impact Angle 14. Vehicle Heading Angle at Impact • Separation Conditions—Current Event: 15. Location of Separation 16. Separation Angle 17. Vehicle Heading Angle at Separation • Subsequent Event/Final Rest 18. Subsequent Event 19. Location of Final Rest 20. Vehicle Heading Angle at Final Rest The completed case, including data from the regular NASS CDS data collection, the supplemental field data collection, and the manual review and reconstruction, was then sent to the project staff for final quality control and reconstruction to estimate the impact speeds. 3.4.4 Reconstruction of Impact Speed As mentioned above, the completed cases from KLD Asso- ciates went through one final quality check by the project staff to assure completeness and accuracy. The cases were then reconstructed to estimate the impact speeds. Reconstruction of single-vehicle, ran-off-road crashes is greatly complicated by the wide variety of roadside objects. For example, Table 12 shows a list of first harmful events caused by objects struck from the 1999 Fatality Analysis Reporting System (FARS) data. It is obvious from the list that the object struck varies widely, from impacts with roadside hazards (e.g., trees and utility poles) to roadside safety devices (e.g., guardrails and crash cushions) to terrain features (e.g., embankments and ditches). In order to accurately identify impact conditions associated with these accidents, it is critical to implement crash recon- struction procedures appropriate for each of the hazards listed. In general, reconstructions of single-vehicle, ran-off-road crashes primarily involve calculating energy losses and gains

20 after leaving the roadway. Energy changes during ran-off-road crashes can generally be attributed to one or more of these seven categories: • Vehicle crush • Damage to roadside feature • Tire braking • Tire side slip • Vehicle rollover • Change in vehicle elevation • Friction between vehicle and roadside feature Key data elements needed to accurately estimate these energy changes include, but are not limited to: • Impact sequence • Vehicle crush profile • Impact angle/principal direction of force during crash • Vehicle trajectory, including tire mark measurement and description • Driver action, i.e., steering/braking • Roll distance and number of quarter roll • Changes in elevation along the vehicle path • Extent of damage to roadside feature It should be noted that these data elements pertain to per- ishable evidence that have to be collected at the time of the crash investigation. For a prospective study in which data are collected on crashes as they occur, the study can be designed to properly document the required data elements. However, in the case of a retrospective study like the current project, the data availability and quality is limited by what was actually collected and could be lacking for some of the data elements. The availability and quality of the data elements can be divided into the following general categories: • Data elements that are well documented and coded in the NASS CDS cases, such as impact sequence, vehicle crush profile, principal direction of force, and number of quar- ter rolls. The quality of these data elements is typically high and no further work is needed. • Data elements that are documented and coded in the CDS cases, but the quality of the data may be somewhat question- able, e.g., driver action. These data elements would need to be checked against other available evidence, such as the scaled diagram, annotated remarks, and photographic doc- umentation, to verify the accuracy of the coded data. • Data elements are documented, but not coded, and the quality of the data may vary greatly from case to case, e.g., vehicle trajectory, tire marks, impact angle, and roll distance. These data elements would have to be gleaned from the scaled diagram, annotated remarks, and photographic documentation. • Data elements that are not documented. The two areas where existing NASS CDS cases may not contain sufficient information are elevation changes along the vehicle path and the characteristics and sustained damage of the impacted roadside feature(s). These data elements would have to be gleaned from the photographic documentation to the extent possible or the information collected in the supplemental data collection effort. It should be noted, however, that the implicit assumption was that the data from the supple- mental data collection were the same as at the time of the crash, which may or may not be true. Although deformation of roadside features is an important source of energy dissipation for some crashes, many ran-off- road crashes would not involve deformable fixed objects. For the limited number of cases where this energy dissipation factor is important, it may be necessary to make estimates of deformation from case photographs and supplemental site investigations. Change in elevation during a crash is generally not an important source of energy change unless the vehicle has traversed a very deep roadside embankment. Elevation changes along the vehicle path can be estimated by recording the dimensions of the various side slopes. While the general principle of identifying the energy loss parameters during the collision and summing the total to determine the change in velocity from the point of impact to the final resting position is rather straightforward, the actual Object Frequency Percentage Tree 2,997 26.09 Embankment 1,213 10.56 Guardrail 1,078 9.39 Utility Pole 1,018 8.86 Ditch 887 7.72 Curb 681 5.93 Culvert 592 5.15 Fence 490 4.27 Sign Support 368 3.20 Other Post/Support 308 2.68 Concrete Barrier 275 2.39 Bridge Rail 158 1.38 Bridge Pier/Abutment 155 1.35 Wall 119 1.04 Luminaire Support 103 0.90 Boulder 79 0.69 Building 79 0.69 Shrubbery 56 0.49 Bridge Parapet 36 0.31 Equipment 26 0.23 Fire Hydrant 25 0.22 Other Longitudinal Barrier 23 0.20 Snow Bank 23 0.20 Traffic Signal Support 22 0.19 Unknown 22 0.19 Impact Attenuator 11 0.10 Other Fixed Object 506 4.41 Other Object (not fixed) 135 1.18 Total 11,485 100.00 Table 12. Object struck as first harmful event from 1999 FARS data.

21 reconstruction is greatly complicated by the wide variety of roadside features. There is not a single procedure that can be used to reconstruct all ran-off-road crashes. Instead, differ- ent reconstruction procedures are needed to accommodate the wide variety of roadside features and types of impact. There are a number of existing procedures that have been developed for reconstructing special types of ran-off-road, fixed-object crashes, including: • Pole support structure (25) • Rigid barrier (15) • Semi-rigid and flexible barrier (14) These roadside features accounted for about 55% of all ran-off-road, fixed-object fatal crashes, as shown in Table 12. For the remaining 45% of crashes, the vast majority can be grouped into one of the following five categories: • Roadside terrain • Rigid hazards • Drainage structures • Buildings and walls • Fences and shrubbery New reconstruction procedures were developed for these five categories of roadside features. Brief discussions on recon- struction procedures for the various roadside features are pre- sented in the following sections. 3.4.4.1 Pole Support Structures A computerized reconstruction procedure was developed for ran-off-road crashes involving pole support structure, including breakaway and nonbreakaway utility poles, lumi- naire supports, and sign supports (25). Energy loss is grouped into three major categories: • Vehicle crush. The CRASH3 (27) reconstruction program was utilized to estimate vehicle crush energy based on vehi- cle crush measurements. • Fracture of pole. Energy associated with breaking or frac- ture of the pole was estimated based on empirical test data. • Post-impact vehicle trajectory. The CRASH3 reconstruction program was also utilized, to the extent possible, for estimat- ing the energy or speed loss associated with the post-impact vehicle trajectory. Otherwise, manual calculations were per- formed for the reconstruction. This procedure was utilized whenever possible for recon- struction of crashes involving pole support structures, e.g., utility poles; luminaire, sign, and traffic signal supports; other post/supports; and fire hydrants. 3.4.4.2 Rigid Barrier Another procedure was developed for reconstructing rigid barrier impacts during a study to assess rollovers on concrete barriers (15). This study found that vehicle/barrier friction was a major source of energy dissipation during a crash. Again, energy loss is grouped into three major categories: • Vehicle crush. The CRASH3 (27) reconstruction program was utilized to estimate vehicle crush energy based on vehi- cle crush measurements. • Friction. Energy loss associated with vehicle/barrier friction was estimated as a function of the length of barrier contact. • Post-impact vehicle trajectory. The CRASH3 reconstruction program was also utilized, to the extent possible, for estimat- ing the energy or speed loss associated with the post-impact vehicle trajectory. Otherwise, manual calculations were used for the reconstruction. The vehicle crush energy was then matched to the energy associated with the lateral velocity of the impacting vehicle. If both energy estimates are comparable, the procedure was believed to be reasonably accurate. If not, the vehicle crush energy would be adjusted appropriately and a new estimate of the impact speed was generated. This iterative procedure has been found to give reasonably good estimates of impact speed when used to evaluate findings from full-scale crash tests. 3.4.4.3 Semi-Rigid and Flexible Barrier A reconstruction procedure for semi-rigid and flexible bar- riers was developed in a study of ran-off-road crashes (14). This procedure utilized similar techniques for estimating vehicle crush and trajectory energy losses. Energy loss associated with the deformation of semi-rigid barriers was estimated from a series of computer simulations that correlated impact severity to maximum barrier deflection. The impact severity is calcu- lated using the following equation: where: IS = Impact Severity M = Vehicle mass V = Vehicle velocity θ = Impact angle The IS value has been shown to be a good indicator of the degree of loading and maximum deflection of a barrier during an impact. Unfortunately, the maximum barrier deflection after a crash is seldom measured during a NASS CDS investi- gation. Thus, the permanent barrier deflection was estimated from available photographic documentation. The measured or IS M V= ( )1 2 2   sinθ

22 estimated permanent barrier deflection was then related to the maximum dynamic deflection, which in turn was used to estimate the IS value from the impact. The impact speed could be estimated from IS value along with the impact angle or by traditional energy loss calcula- tions, including vehicle crush, barrier deformation, and post- impact trajectory. An iterative procedure similar to that used to reconstruct rigid barrier crashes was developed for this application. The procedure from Erinle et al. (14) was refined and updated for use in the current study. The revised procedure also included techniques for reconstructing impacts with guardrail terminals and crash cushions. 3.4.4.4 Roadside Terrain Impacts involving embankments and ditches could be reconstructed if detailed information is available on the ter- rain and any associated gouges in the terrain along with the vehicle crush. Efforts to model vehicles traversing hazardous roadside terrains have established reasonable measures of the forces and energy associated with vehicle undercarriage com- ponents gouging into the terrain (28). Furthermore, for crashes involving vehicles plowing into steep embankments virtually head on, vehicle crush measurements would produce a good estimate of the total force generated between the embankment and the vehicle. Finally, energy losses associated with rollover accidents have been investigated through computer simulation for a variety of passenger vehicles (29). Hence, impact speeds for crashes involving roadside terrain could be estimated by combining conventional trajectory analyses, such as that used in the CRASH3 reconstruction program, and incorporating procedures for estimating the effects of terrain gouging and vehicle rollover. 3.4.4.5 Rigid Hazards For rigid obstacles, such as bridge piers and parapets, boul- ders, and heavy construction equipment, there is little energy dissipated by the rigid hazards themselves. Thus, reconstruc- tions could be based almost entirely on vehicle crush energy and post-impact trajectories. These procedures would be sim- ilar to those used by Mak and Labra (25) to reconstruct pole crashes in which the poles remained intact. 3.4.4.6 Drainage Structures Drainage structures, such as culverts and curbs, are often traversed during a ran-off-road accident without a significant speed reduction. Full-scale crash testing and computer sim- ulation have shown that speed losses during curb impacts are very low (30). These simulation and test findings were used to obtain gross estimates of the total speed loss associated with curb impacts. Thereafter, other reconstruction techniques could be used to estimate the total energy lost during the post- impact trajectory of the vehicle. Culverts offer significantly greater challenges. Cross-drainage culverts with high headwalls can act as a rigid hazard and could be reconstructed based largely on vehicle crush as described in the previous section. Crash tests of cross-drainage culverts that have been cut to match the slope and/or grated to reduce the severity of crashes have shown that these hazards provide very little energy dissipation (31). This low level of energy dissipation would allow crashes involving these hazards to be reconstructed based on the post-impact trajectory alone. Unfortunately, reconstruction of crashes involving parallel drainage structures were somewhat more difficult. Crash test- ing has indicated that vehicles striking culverts under drive- ways or intersecting streets are frequently subjected to violent rollovers. Where possible, procedures for estimating energy losses during vehicle rollover formed the basis for reconstruct- ing rollover crashes associated with culvert accidents. Con- ventional trajectory analyses were used whenever the vehicles remained upright after striking the culvert. 3.4.4.7 Buildings and Walls When buildings and walls are struck in a more or less head- on configuration, conventional reconstruction techniques are applicable only if the building or wall is relatively rigid. No procedure has been developed that can effectively estimate the energy required to break through a building or wall. How- ever, if the structures remain intact, the building or wall was treated as either a rigid hazard or a rigid longitudinal barrier, depending on the nature of the impact. 3.4.4.8 Fences and Shrubbery Most fences, including chain link and wooden privacy fences, provide relatively little energy dissipation when struck by an automobile traveling at a high rate of speed. Similarly, small shrubs do not offer significant resistance to an impact- ing vehicle. Therefore, crashes involving these hazards were reconstructed using conventional procedures unless the fence had an unusual construction or the shrubs were large enough to pose a major obstacle to a vehicle. In summary, by utilizing and refining available reconstruc- tion techniques, it was possible to produce accurate estimates of the impact conditions for most ran-off-road crashes. The reconstruction procedures discussed above should account for almost 90% of the serious injury and fatal ran-off-road crashes.

23 3.4.5 Conduct of Data Collection The work on supplemental field data collection, quality control, and manual review and reconstruction of the sam- pled cases was conducted over a period of approximately 12 months. Of the 404 sampled cases, 15 were found to have major construction/reconstruction at the crash sites and thus were eliminated from the sample. One additional case was eliminated because it involved two vehicles. Thus, the final sample size was reduced from 404 to 388. 3.5 Data from Previous Studies NCHRP Project 17-11 and FHWA’s Rollover Study incor- porated the same data collection procedures as used in the cur- rent study and included at total of 485 cases from NASS CDS for the years 1997 through 1999. These studies were conducted by the Texas Transportation Institute (TTI) and therefore the data from the two studies will be referred to collectively as “TTI data.” Because the TTI data was collected and processed using the same protocol as the data collected in this (17-22) study, it was believed to be appropriate to combine the two data sets into a single file. Unfortunately, upon comparison of basic crash data, such as departure velocity and angle, it became apparent that the two data sets were not sufficiently similar to be combined. The biggest differences were found in departure and impact angles. For example, the average departure angle for the TTI data was found to be 19.9 degrees, compared to 17.2 degrees for the 17-22 data. This 15% difference in average departure angle was considered to be excessive. When a simple T-test was applied to compare the two data sets, differences in departure angle were found to be significant at the p = 0.001 level. These findings prompted a more careful examination of the differences between the TTI data set and the NCHRP 17-22 data set. It was discovered that the TTI cases were recon- structed from scene diagrams downloaded from the NASS CDS website. These scene diagrams had been converted to PDF format before being posted on the website. Unfortu- nately, the process of converting the scene diagrams to PDF changed the scaling of the drawings. The compression in the longitudinal direction was found to be greater than the com- pression in the lateral direction. As a result, all angle measure- ments were corrupted. 3.5.1 Manual Review and Crash Reconstruction of Prior Cases In order to salvage the 485 cases included in the TTI data set, it was necessary to obtain the original scene diagrams and repeat the reconstruction process for all of the cases. Unfor- tunately, supplemental data forms for 35 of the TTI cases were lost in transit from College Station, Texas, to Lincoln, Nebraska. Although reconstructions were possible for these 35 cases, much of the supplemental information such as road- side topography, land use, highway classification, and highway alignment could not be determined. 3.5.2 Incorporation of Prior Data into Database After the reconstructions and manual reviews were repeated for the TTI cases, the TTI and 17-22 data sets were subjected to a comprehensive evaluation to determine the appropriate- ness of combining them into a single data set. Each impor- tant variable was tested to determine the significance of differences between the two data sets. Whenever a variable was found to be significantly different at the p = 0.05 level, all 877 cases were re-examined to identify the source of the error. In some cases, the errors were found to be related to the way a specific parameter was measured. For example, the heading angle at departure was measured from –180 to 180 degrees in the 17-22 data and from 0 to 360 degrees in the TTI data. These errors were easily corrected. Other data elements were found to have been poorly recorded on the supplemental data forms. For example, in some cases, the roadside slope was recorded as the highway grade. In this sit- uation, the research team was forced to re-examine every case to compare photographs at the scene with the recorded high- way grade. Whenever there was reasonable evidence of an error, the entire file was examined for evidence of the highway grade. In some cases, the highway grade was found in investi- gator notes on the supplemental data forms. In other situa- tions, the elevation changes along the roadway were recorded between the point of departure and at a point where the vehi- cle re-entered the roadway. These elevation changes were then used to estimate highway grade at the crash site. Unfortunately, there were many cases where the highway grade could not be identified and the variable had to be labeled as unknown. This type of examination was undertaken for a large number of data elements that were found to be significantly different in the two data sets. As shown in Table 13, most variables with significant dif- ferences between the two data sets were corrected and the two data sets could be considered to be relatively similar. Unfor- tunately, significant differences remained for some variables, including speed limit, vehicle weight, height and width of object struck, rollover, and vehicle class. Differences in speed limit and vehicle weight are believed to be appropriate. The national speed limit law was repealed in late 1995 and was not implemented immediately in many states. In fact, 18 states had not implemented any change in speed limit before the end of 1997. Many of these states eventually raised speed limits.

24 17-22 Data TTI Data Variable Units Mean StdDev. SEM Mean Std Dev. SEM P Value Dep. velocity km/h 80.00 26.00 1.32 78.70 25.30 1.15 0.48 Dep. angle deg. 17.20 11.90 0.60 16.90 10.20 0.47 0.70 IS value kJ 41.70 59.60 3.02 36.90 74.90 3.41 0.31 Degree of curvature deg. 2.27 7.50 2.65 2.65 6.72 0.32 0.45 Driver action 3.92 3.03 0.15 4.16 3.09 0.15 0.27 Month 6.68 3.45 0.17 6.39 3.00 0.14 0.20 Access control 2.27 0.92 0.05 2.28 0.94 0.04 0.82 Accident time 0.48 0.30 0.02 0.52 0.40 0.02 0.10 Alignment 1.53 0.79 0.04 1.61 0.80 0.04 0.17 Curb height mm 5.59 29.23 1.48 8.24 39.31 1.86 0.27 Curbs 0.09 0.40 0.02 0.08 0.36 0.02 0.64 Departure side 1.49 0.50 0.03 1.43 0.50 0.02 0.07 Divided/undivided 1.43 0.50 0.03 1.38 0.49 0.02 0.14 Grade % 1.50 1.67 0.08 1.39 1.49 0.07 0.10 Highway speed limit mph 57.45 9.24 0.47 55.68 9.44 0.43 0.006 Land use 1.76 0.43 0.02 1.70 0.47 0.02 0.06 Lane width m 3.69 0.55 0.03 3.64 0.52 0.02 0.18 Lat distance from departure to rest m 0.07 13.45 0.68 1.24 13.22 0.60 0.20 Lateral travel m 0.37 12.22 0.62 1.02 12.93 0.59 0.45 Heading angle at point of rest deg. 166.15 111.15 5.63 165.27 111.47 5.21 0.91 Long. distance from dep. to rest m 46.40 37.83 1.91 44.84 40.05 1.82 0.56 Long. travel, 1st encroachment Material of object struck No. of slopes Object diameter Object height Object length Object width Road class Road condition Road profile Road surface Rollover Shoulder type Shoulder width Sideslip angle Vehicle weight Weather X-section at departure m cm cm cm m m deg. lb 39.14 5.02 4.11 33.54 475.84 2937 68.91 2.77 1.36 0.52 1.21 0.59 1.27 1.77 -1.02 3348.32 1.24 5.43 30.89 2.60 1.81 26.55 699.70 6326 292.38 2.86 0.82 0.82 0.65 0.49 0.74 1.31 38.61 861.96 0.69 2.66 1.56 0.13 0.09 2.81 62.33 922.8 19.15 1.29 0.04 0.04 0.03 0.02 0.04 0.07 1.38 43.59 0.03 0.13 39.79 4.70 3.94 29.44 215.86 1145 292.38 2.86 1.31 0.53 1.25 0.50 1.30 1.86 0.63 3154.16 1.20 5.45 34.71 2.29 1.59 38.90 240.01 2229 433.98 1.43 0.72 0.89 0.75 0.50 0.84 1.40 38.59 738.30 0.57 2.67 1.58 0.11 0.08 2.66 17.60 388.1 38.97 0.07 0.03 0.04 0.03 0.02 0.04 0.07 1.76 33.52 0.03 0.13 0.78 0.06 0.15 0.36 0.0001 0.12 0.0003 0.33 0.32 0.90 0.46 0.008 0.55 0.37 0.46 0.0003 0.32 0.89 Table 13. Comparison of 17-22 and TTI data.

25 Recall that the TTI data included crashes from 1997 through 1999 while the 17-22 study included data from 2000–2001. Thus, it is not surprising that speed limits were found to increase between the time of data collection for the TTI and 17-22 data sets. Similarly, the average weight of the vehicle fleet increased dramatically during the 1990s. In the early 1990s, the 5th and 95th percentile passenger vehicle weights were 1,800 and 4,400 lb respectively. By 2002, the 5th and 95th percentile weights had increased to 2,500 and 5,200 lb, respectively. This dramatic increase in vehicle weight would be expected to cause the average weight of crash vehicles to be higher in 2000 and 2001 than during the 1997 through 1999 period. Hence, the nearly 200 lb increase in average weight between the TTI and 17-22 data sets is not unexpected. Careful examination of the two data sets revealed that the differences in the width and height of the object struck between the two data sets could be attributed to overrepresen- tations in the number of tall trees impacted in the 17-22 data and of wide ditches in the TTI data. Note that the increase in the number of trees or the number of ditches was not sufficient to produce statistically significant differences in the object- struck category. However, the number of very tall trees (15 m or more) in the 17-22 data was sufficient to produce significant differences in the height of the object struck. Further, a rela- tively small number of wide ditches in the TTI data produced significant differences in the width of the object struck. The number of rollovers in the 17-22 data was found to be significantly greater than in the TTI data. As shown in Table 13, 59% of the cases from 17-22 involved vehicle rollover com- pared to only 50% for the TTI data. A careful evaluation of each case in both data sets could not provide any explanation for the magnitude of the difference in rollover frequency. The only possible explanations for the high rollover rate is that the 17-22 data also had 47% light-truck involvement compared to 38% for TTI data. Although light-truck sales were growing during the 1997 through 2001 time frame, the 9% increase in light-truck involvement is unexpectedly high. Further, even though light trucks are known to have a higher risk of roll- over, the overrepresentation of light trucks is insufficient to explain the full magnitude of the difference in rollover rate. The rollover rates for both cars and light trucks were found to be significantly higher in the 17-22 data than in the TTI data. The 17-22 data had 50% and 69% rollover rates for cars and light trucks, respectively, while the comparable numbers from the TTI data were 44% and 59%. Unfortunately, the fundamental differences in rollover rate could neither be eliminated nor explained. In spite of the differences found in the six variables described above, differences between the two data sets were not statis- tically significant for the vast majority of data elements. Based upon this finding, combining the two data sets was deemed acceptable. Note that finding differences not to be statistically significant does not necessarily imply that the data sets are similar. Users should use caution whenever using the com- bined database to examine highway or crash characteristics that are close to the threshold of statistical significance. 3.6 Relational Database The design of a relational database for the purpose of storage and retrieval of crash data was developed and imple- mented. In addition to the data collected under this study, the crash database also stored data from NCHRP 17-11 and the FHWA Rollover Study. The crash database design revolved around the Oracle server, which is an object-relational database management system providing an open, comprehensive, and integrated approach to data management. The crash database was composed of a data file containing different types of elements (e.g., CASE_NUM, CASE_ID, DEPARTURE ANGLE, etc.). A user process (or a client process) and a server process were used for successful communication between users and the crash database. Together these two processes enabled users to run various queries on the database. Access to the crash database could be obtained by directly issuing SQL commands or through the use of an applica- tion that contains SQL statements. The Oracle crash database processes the commands and returns results to the users. It is physically located on a server residing at the Nebraska Trans- portation Center of the University of Nebraska–Lincoln. Cur- rently, logging in directly on the host computer is supported, i.e., the computer running the Oracle crash database server is used for database access. The communication pathway is established using the inter-process communication mecha- nisms available on the host computer. Logging in via a two- tiered (client-server) connection, where the machine on which the user is logged in is connected directly to the machine run- ning the Oracle crash database server, and via a three-tiered connection, where users will connect to the Oracle crash data- base server via network server(s) by using a customized appli- cation, are possible but have not been implemented. However, remote access to the database is available using Windows® Remote Desktop Connection (password protected). Data ele- ment names and definitions are presented in Appendix D.

Next: Chapter 4 - Results »
Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes Get This Book
×
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

TRB’s National Cooperative Highway Research Program (NCHRP) Report 665: Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes quantifies the characteristics of ran-off-road crashes and identifies appropriate impact conditions for use in full-scale crash testing.

Appendices A through F of NCHRP Report 665, which are as follows, are available online:

Appendix A: Annotated Bibliography

Appendix B: 1997–2001 NASS CDS Cases

Appendix C: Supplemental Data Collection Protocol

Appendix D: Database Content

Appendix E: Additional Tables, Plots, and Analysis Results

Appendix F: Proposed Data Collection Forms Continuous Sampling Subsystem

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!