Naturalistic Driving Study: Development of the Roadway Information Database (2014)

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.

46 A p p e n d i x A The user needs assessment effort for the project included four components: • Identify user/stakeholder groups who will use the roadway data and geographic information system (GIS) tools when accessing the Roadway Information Database (RID). • Identify roadway data elements necessary to answer research questions using the naturalistic driving study (NDS) data. • Prioritize roadway data elements to be either acquired through existing data sets or collected under the mobile mapping data collection project (S04B). • Solicit user/stakeholder feedback on development of GIS tools to use the RID. A.1 develop Candidate List of Relevant Roadway data elements After identifying the user/stakeholder groups, as described in Chapter 2 of the report, the next step in this task was to iden- tify roadway data elements necessary to answer research ques- tions using the SHRP 2 NDS data. This included a broad range of safety-related research questions that may be asked by users identified in the previous subtask, including those required for the SHRP 2 analysis projects (S08). The result was an exhaustive list of roadway data elements that were candidates for inclusion in the RID. A.2 Refine Candidate List The team used this candidate list as a starting point in the process to define what data elements were to be included in the RID. An exhaustive literature review was conducted to determine the roadway data elements that have been identi- fied in safety analyses as being related to intersection or lane departure crashes—the high-priority areas for SHRP 2. The literature identified roadway data elements that have been shown to have a correlation with crash frequency or severity and determined which elements have already been identified by researchers as being relevant. Data elements may contribute positively or negatively to crash risk. Although a large number of sources were consulted, information from 55 journal papers or reports contributed to the process of refining the candidate list of data elements. The literature review was also used to prioritize data elements, as described in another section in this document. The team next considered data necessary to answer the research questions posed in completed SHRP 2 projects. Researchers for the SHRP 2 S02 project, Integration of Analy- sis Methods and Development of Analysis Plan, assembled research questions from SHRP 2 projects S01 and S05 researchers, resulting in a set of 448 research questions. Ques- tions were synthesized and prioritized into eight global research topics. The S04A team reviewed the 448 research questions with a focus on the eight topic areas. Relevant data elements identi- fied in the questions were added to the candidate list. The following is an unranked list of the eight topic areas identified by S02 researchers (Boyle et al. 2012): • How do driver interactions with roadway features influ- ence the likelihood of lane departure crashes? • How do driver interactions with intersection features (con- figuration and operations) influence crash likelihood? • What is the influence of driver impairment (e.g., alcohol) on crash likelihood? • How does driver distraction influence crash likelihood? • How does driver fatigue affect crash likelihood? • How do aggressive driving behaviors influence crash like- lihood? • How do advanced driver support systems influence crash likelihood? User Needs Assessment

47 • What variables or pre-event factors are the most effective crash surrogate measures, what explanatory factors are associated with crashes or crash surrogates, and what ana- lytical models can be developed to predict crashes or crash surrogates? The first two questions deal specifically with roadway fea- tures. The first focuses on lane departures, which is typically a rural issue. The second deals with the relationship between intersection features, driver behavior, and crash likelihood. Intersections can be either rural or urban. Roadway features related to lane departures and intersections were the highest priority. Five of the remaining six research questions focused on driver behavior. Although roadway factors can exacerbate neg- ative driver behavior and increase crash likelihood, no specific roadway features were identified in the research questions. The last question relates to developing crash surrogates. No road- way features were specifically identified for this question. A.2.1 Solicit Additional Feedback and Finalize the List of Data Elements After the candidate list was developed, the team obtained additional information from the relevant user groups identi- fied previously. This information was obtained in several ways. First, the team reviewed roadway data elements that had been included in the Highway Safety Information System (HSIS), Model Inventory of Roadway Elements (MIRE), and the Model Minimum Uniform Crash Criteria (MMUCC). Data elements included in those inventories not already iden- tified were added to the list. A webinar was held in September 2010 that also solicited feedback from potential users. A.2.2 Data Elements and Priorities from MIRE, MMUCC, and HSIS Several user groups have already contributed significantly to identifying important roadway data elements for safety analyses. Many users were involved in development of MIRE, MMUCC, and HSIS. MIRE was developed to define critical safety data inventory elements for state and local agencies in order for them to conduct their own safety analyses, as well as those necessary for them to take advantage of FHWA’s Interactive Highway Safety Design Model (IHSDM) and SafetyAnalyst (Council et al. 2007; MIRE 2010). The MIRE effort identified the first set of elements by reviewing data elements from IHSDM and SafetyAnalyst, the Highway Performance Monitoring System (HPMS), the Traf- fic Safety Information Management System (TSIMS) of the American Association of State Highway and Transportation Officials (AASHTO), and MMUCC, as well as the team’s knowledge of state inventory databases, the team’s knowledge of efforts related to development of the Highway Safety Man- ual, and the MIRE project team’s knowledge of data needed for safety analyses not traditionally included, such as pedes- trian, bicycle, and roundabout safety. The MIRE effort also provided a priority for the data ele- ments identified (priorities in MIRE are listed as “1st,” “2nd,” and “do not recommend”) based on data requirements for the safety tools described above, requirements for MMUCC, consideration of whether states were already collecting the element, and the team’s knowledge of what may be included in existing and future safety analyses and tools. When rele- vant, the difficulty of collecting the data was considered. The MMUCC is a set of voluntary guidelines to help states collect the consistent, reliable crash data necessary to improve highway safety; establish goals and performance measures; monitor the progress of programs; and allocate resources for enforcement, engineering, and education. The criteria include the minimum variables that should be used to describe a motor vehicle crash to make informed decisions leading to highway safety improvements. Elements were included if they were deemed necessary for highway safety decision making. MMUCC was developed by some of the leading traffic safety experts in the United States. These included representatives from safety, engineering, emergency medical services, law enforcement, and state and federal agencies (MMUCC 2010; MMUCC Guideline 2008). Because HSIS, MIRE, and MMUCC include data elements that have been exhaustively reviewed by national, state, and local transportation agencies; traffic safety experts; law enforce- ment; and emergency medical services around the country, the team felt that this provides sufficient representation from these groups regarding which data elements are important in safety analyses. A.2.3 Solicitation of Additional Feedback The team presented the methodology to develop the candi- date list of roadway data elements discussed in the previous step at the July 2010 SHRP 2 Safety Research Symposium in Washington, D.C. The team received a few comments about additional data elements, which were added to the list. The next step was to conduct a survey to solicit feedback. A list of questions was developed in Survey Monkey, which asked respondents to review the initial list of roadway data elements and then add any additional items that they felt were necessary. A list of safety professionals who had attended one annual SHRP 2 Safety Research Symposium from 2008 to 2010 was obtained, and attendees were invited to partici- pate in the survey. Invitations were sent to 177 people in July 2010. The survey was available for approximately two weeks. Eleven people responded to the survey.

48 The first question asked respondents to identify which user category best represented them. Responses are shown in Table A.1. Respondents were asked to enter additional data elements that they felt should be included for roadway segments (non- intersection), which resulted in 19 data elements. Respondents were also asked what additional intersection data elements should be included, which resulted in a list of 25 elements. A number of the data elements were already included in the list of roadway data elements. Several items were traffic operation data, such as annual average daily traffic. The roadway data ele- ments suggested by survey respondents that were not already included were added to the list. Respondents were also asked for contact information and were included in a later survey to prioritize the data elements. Once the survey was completed, the list of roadway data elements was finalized. The list was maintained in a spread- sheet. Roadway elements for all roadway segments were kept in one spreadsheet, and those specific to intersections were maintained in another. Information about other sources that had suggested that the data element was important was also recorded. For instance, one column was used to indicate whether the data element was listed in MMUCC, other col- umns were used for HSIS and MIRE, and another column indicated whether the data element was listed in a research question from SHRP 2 projects S01, S02, or S05. A.3 prioritization of Research Questions The candidate list of roadway data elements was quite extensive, and, given limited resources in both S04A to acquire existing data sets and S04B to collect mobile data, it was necessary to prioritize roadway data elements. The team assessed priorities and created a prioritized list, which was used to guide mobile data collection. Elements ranked the highest were a top priority for mobile data collection if they could feasibly be collected and if the data element was expected to remain consistent through- out the data collection period. For example, edge drop-off, while a desired data element for run-off-road analysis, has a high temporal change rate and consequently did not make it to the final list of data elements collected with the mobile vans. Considerations in prioritization of the data elements included the following: • Stakeholder priorities; • Importance to SHRP 2 analysis projects (S08); • Focus on lane departure and intersection crashes; • Importance to future safety research questions; • Importance to future nonsafety research questions (e.g., asset management, travel demand modeling, transportation air quality analysis); • Priority from MIRE; • Importance from surveyed literature with focus on crash severity; • Curve properties (e.g., radius); • Resources to collect a data item in relationship to its importance; • Whether the feature can be collected with few additional resources, even if it is lower ranked; • Whether the feature will be available with reasonable accu- racy from an existing database; • Whether the feature could be extracted from forward NDS data acquisition system roadway video; and • Team to review sample data set. A.3.1 Literature Review Data elements were prioritized using several sources of infor- mation. First, the priority assigned to each data element in MIRE was recorded. Next, a comprehensive literature review was conducted for lane departure and intersection crashes. These two crash types were selected because they were the focus of SHRP 2. The number of times a roadway data ele- ment was reported in the literature as having either a positive or negative correlation to crashes was reported. A.3.2 Survey Another survey was conducted to ask user groups about the data elements they felt were the most important. Originally, the team had planned a webinar where attendees could inter- actively respond to questions. However, the SHRP 2 webinar services could only post six interactive questions during the webinar. Since this was not sufficient to obtain feedback, the team used a hybrid webinar/survey approach. Attendees from the 2008, 2009, and 2010 SHRP 2 Safety Research Symposia were sent an email inviting them to the webinar/survey. In Table A.1. User Groups Responding to Survey User Group Respondents National safety or roadway agency (i.e., FHWA, NHTSA) 2 State highway agency 1 Traffic safety researcher (university or consultant) 2 Human factors researcher 1 Asset management expert 2 Insurance group 1 Other • Pedestrian advocate • Automotive ITS researcher 2 1 1

49 addition to those individuals, the team also identified five human factors researchers who were familiar with the SHRP 2 NDS. Human factors researchers would have a different per- spective on which roadway factors may present a safety risk than traditional roadway safety professionals. Human factors researchers had not been included in any of the groups sur- veyed to identify roadway data elements. Two state DOT traffic safety engineers were also invited to participate. The webinar was hosted by SHRP 2 and held on Septem- ber 30, 2010. Ninety-five sites registered for the webinar. Fifty- eight logged into the webinar, and 31 to 37 responded to the questions. The webinar consisted of the following: • Introduction and summary of SHRP 2; • General background on the S04 project; • Description of data collection from state and local agencies; • Method used to identify and prioritize roadway data ele- ments; and • Description of what was included in the survey. Respondents were asked to respond to several questions during the webinar. First, they were asked about their level of expertise in using GIS for safety analyses. Of those who responded to this question (31), 26% considered themselves novice users, 29% indicated they were occasional users, and 19% and 10% were experienced or expert users, respectively (Figure A.1). The second question asked respondents (37 answered) about their level of expertise in integrating disparate data sets, such as roadway inventory data and crash data. Fig- ure A.2 shows that 30% felt that they were novice users, and 24% were occasional users; 19% indicated they were experi- enced users, and 14% indicated they were expert users. The third question asked respondents how frequently they used federal roadway data, particularly for safety analyses. Out of the 31 responses, 42% indicated that they never or seldom use federal roadway data, as shown in Figure A.3. Another 23% indicated that they occasionally used the data, and 29% were frequent users. Attendees were asked how regularly they used county or municipal roadway data, particularly with respect to safety analyses. Out of the 34 responses received, the majority, 65%, never or seldom use county or municipal data, while 21% indicated they occasionally use it and 6% indicated they use it frequently, as shown in Figure A.4. The final in-webinar poll question asked attendees how fre- quently they used state roadway data, particularly with respect Figure A.1. Level of expertise using GIS for safety analyses. Figure A.2. Level of expertise in integrating disparate data sets. Figure A.3. Use of federal roadway data. Figure A.4. Use of county or municipal data.

50 covered a category. For instance, the first category listed items included in horizontal and vertical curvature, as shown in Figure A.6. Features under that category (e.g., horizontal curve length, maximum superelevation) were listed individu- ally, and respondents could select from the following: • Most important; • Not as important, collect if cost-effective; • Possibly useful; and • Not useful. Since there was not a logical way to get respondents to pri- oritize the list, respondents could select a response as many times as they wished. For instance, a respondent could select “most important” for every data item. Respondents were asked to only indicate “most important” for a few top-priority items, but there was not a method to ensure that this happened. Categories included the following: • Horizontal and vertical curvature; • Cross section; • Regulatory and warning signs; • Pavement markings; • Road surface condition; • Countermeasures; • Other, which included guide signs, service signs, and bridge features (location, type, bridge deck width); • Roundabout characteristics that differ from regular inter- sections; Figure A.5. Use of state data. Figure A.6. Example webinar survey question. to safety analyses. Out of the 32 responses, 41% had never or seldom used state data, while 25% and 22% used it occasionally or frequently, respectively. Responses are shown in Figure A.5. After initial information was provided about the S04A proj- ect and the method to identify data elements was presented, the S04A team reviewed the components of the online survey. Webinar attendees were then directed to the online survey, where they were able to indicate which data elements they thought were a priority. Respondents were to immediately fill out the survey, but a grace period of two weeks was provided. The survey was conducted using Survey Monkey. In order to better display the information in the survey, roadway ele- ments were divided into logical categories, and each question

51 • Intersection control; and • Interchange ramp characteristics. The next question asked respondents to rate the roadway types they felt data collection should focus on. For instance, is it more important to collect urban Interstates or rural minor arterials? This question helped prioritize data collec- tion for the mobile data collection effort in S04B. An example of the question is provided in Figure A.7. Only six people responded to the online prioritization survey. Of those who responded, 67% indicated that they have some experience with driver/user-focused traffic safety research, 50% indicated that they have some experience with vehicle-focused traffic safety research, and 100% indicated that they have some experience with vehicle-focused traffic safety research. The list of roadway data elements was updated with a col- umn that indicated whether any of the survey respondents had indicated the data element as being “most important” by one or more researchers. A.3.3 Discussion with Human Factors Researchers To get input and feedback from human factors researchers, a group was identified and targeted for the webinar/survey. However, a human factors conference was scheduled for the same time, so none were in attendance at the webinar or filled out the survey. The S04A team contacted five human factors researchers who were familiar with the SHRP 2 Safety program and asked them individually for input. Their input was recorded in an additional column of the list of roadway data elements as being ranked “high importance” by one or more human factors researches. A.3.4 Creation of Initial Prioritized List Information from the previous steps was used to create an initial priority of data items. A point system was used to score all the roadway data items in the list. Points were assigned if the data element was the following: • Included in the research questions from S01, S02, or S05; • Included as a priority 1 in MIRE; • Included as a data element in MMUCC; • Included as a data element in HSIS; • Identified in the literature review; • Ranked “high importance” by one or more survey respon- dents; or • Ranked “high importance” by one or more human factors researchers. Points were added, and an initial list of priorities was devel- oped. The initial ranked list with data elements having a sum of three or higher is shown in Table A.2 for tangent sections/ Figure A.7. Example questions about roadway types.

52 Table A.2. Roadway Tangent Data Elements Ranking Three or Higher Data Element Features Sum Access Access control 3 Driveway density 5 Road classification Route/street name 3 Functional class/type 4 Clear zone Curb presence 3 Type of objects within clear zone 4 Cross section Type (i.e., regular, two-way left-turn lane [TWLTL], parking, acceleration, high-occupancy vehicle [HOV], reversible) 3 Median width 3 Number of lanes 4 Median type 5 Lane width 6 Horizontal curvature Location, including PC and PT 3 Tangent length between adjacent curves 3 Direction of curve 3 Horizontal curve deflection angle 3 Length 4 Presence and type of spirals 4 Presence and amount of superelevation 4 Radius or degree of curve 7 Illumination Overhead lighting type 5 Other countermeasures Type edgeline or shoulder rumble strips 3 Pavement markings Centerline (i.e., dashed, solid) 3 Edgeline 4 Regulatory signs Pass/no pass zones 3 School area 3 Speed limit (location) 5 Road surface Surface type (i.e., gravel, asphalt, portland cement concrete [PCC]) 3 Surface friction 3 Roadway defects Surface irregularities 3 Pavement edge drop-off 4 Shoulder Right shoulder paved width 3 Left shoulder type 3 Left shoulder paved width 3 Right shoulder type 5 Right total shoulder width 6 Vertical curvature Vertical curve length 4 Grade (percent) 7 Warning signs Horizontal alignment signs and location (i.e., chevron, curve advisory speed) 3

53 intersection approaches (38 elements out of a total of 153), and Table A.3 shows the list for data elements specific to inter- sections (nine elements out of a total of 113). A.3.5 Develop Final List of Prioritized Data Elements The initial prioritized listed was developed in the previous step. A final list, which was provided in the S04B proposal, was developed in this step. The final step was prioritized using several pieces of information. The assessment at this point was subjective. A preliminary test run of the SHRP 2 data acquisition sys- tem (DAS) was evaluated to determine the roadway data ele- ments that could feasibly be extracted in the event that they are not available in roadway data sets from SHRP 2 S04A or S04B. The team reviewed the list of roadway data elements and the DAS data set and commented on which roadway fac- tors can be extracted from the DAS data set, which may be useful to S08 and other researchers. A summary is provided as Appendix C to the S01E report (Hallmark et al. 2011). The types of crashes that each data element was likely to address were also determined. For instance, type of median is most related to cross-centerline crashes. The team also determined whether it was feasible to collect the data element with mobile data collection. The cost for collection of each individual item was also considered. Using the information about whether the data element could be obtained from the forward view of the DAS, consideration of type of crash addressed, and feasibility and cost were used to develop a final prioritized list, which was presented to SHRP 2. The final list consisted of two tiers of critical items (Tier 1 and Tier 2), which were to be collected as part of the SHRP 2 mobile data collection project S04B. The two tiers of data are as follows: • Tier 1 44 Horizontal curvature: Radius, length, PC, PT, and direc- tion of curve (left or right based on driving direction). 44 Grade. 44 Cross slope/superelevation. 44 Lane information: Number, width, and type (e.g., through, turn, passing, acceleration, carpool). 44 Shoulder type/curb (and paved width, if it exists). 44 All MUTCD signs (location and message). 44 Intersection location. • Tier 2 44 Intersections: Number of approaches and control (un- controlled, all-way stop, two-way stop, yield, signalized, roundabout). Ramp termini are considered intersections. 44 Median presence: Type (depressed, raised, flush, barrier). 44 Rumble strip presence: Location (centerline, edgeline, shoulder). 44 Lighting presence. 44 Guardrail: Begin and end, type, and material. A.4 References Boyle, L. N., S. Hallmark, J. D. Lee, D. V. McGehee, D. M. Neyens, and N. J. Ward. 2012. SHRP 2 Report S2-S02-RW-1: Integration of Analy- sis Methods and Development of Analysis Plan. Transportation Research Board of the National Academies, Washington, D.C. Council, F. M., D. L. Harkey, D. L. Carter, and B. White. 2007. Model Mini- mum Inventory of Roadway Elements—MMIRE. FHWA-HRT-07-046. Federal Highway Administration, Office of Safety Research and Development. Hallmark, S., Y.-Y. Hsu, L. Boyle, A. Carriquiry, Y. Tian, A. Mudgal. 2011. SHRP 2 Report S2-S01E-RW-1: Evaluation of Data Needs, Crash Sur- rogates, and Analysis Methods to Address Lane Departure Research Questions Using Naturalistic Driving Study Data. Transportation Research Board of the National Academies, Washington, D.C. MMUCC Guideline: Model Minimum Uniform Crash Criteria, 3rd ed. 2008. Model Inventory of Roadway Elements (MIRE). http://www.mireinfo .org/about.html. Accessed June 2010. Model Minimum Uniform Crash Criteria (MMUCC). http://www .mmucc.us/about-mmucc. Accessed June 2010. Table A.3. Roadway Intersection Data Elements Ranking Three or Higher Data Element Features Sum Cross section Intersection/interchange type 5 Number of lanes by approach 4 Turn lanes (number, type) 4 Median type 3 Pavement markings Crosswalks 3 Intersection control Type of intersection traffic control 5 Presence of left-turn arrow (indication of left-turn phasing) 3 Illumination Lighting type 4 Other Sight distance 3