Highway Safety Research Agenda: Infrastructure and Operations (2013)

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.

27 This chapter develops and presents a prioritized list of applied and fundamental research topics. The task was con- ducted as a two-step effort. The initial step involved the development of a “master list” of candidate research topics. The second step involved the implementation of the recom- mended methods to prioritize those topics. Each step is dis- cussed separately below. Defining the Method for Identifying Potential Research Topics This section defines the method to identify both applied and fundamental research topics and issues that could poten- tially be part of the National Agenda. The method identifies potential sources of the topics, describes the process of col- lecting topics from the sources, and defines the level of detail needed in the description of each potential topic. While the primary goal of this effort was to develop a list of potential research topics, an equally important goal was to develop this list in such a way that future potential users of the National Agenda will feel some ownership of that Agenda. This required widespread input from potential users (i.e., those who will choose research topics/projects either as research funders or researchers). Specifically, potential users should be involved throughout the process and shown where their input fed into the Agenda (e.g., any potential research topic listed in the master list will include the source of the idea). The list of potential topics must also include both applied and fundamen- tal research, and these two lists should be developed separately, because the sources of potential ideas, required level of detail, and user-reviewers may differ between these two categories. The research team developed an initial list of poten- tial research topics based on existing databases (e.g., TRB’s RNSs) and documents (e.g., knowledge gaps identified in the HSM). Sources considered for potential research topics are listed in Appendix J. In addition, several organizations and individuals were contacted for additional ideas. Appendix K contains a list of key stakeholder-reviewers. These appendices are not published herein, but are available on the TRB web- site at: http://apps.trb.org/cmsfeed/TRBNetProjectDisplay. asp?ProjectID=2727. Identification of Research Topics to be Prioritized The research team collected several RNSs covering a wide range of safety areas and issues. The research team utilized its extensive experience in identifying available research and research needs, based on previous work on the HSM and proj- ects for FHWA and NCHRP. RNSs were extracted from both national and local sources, including the sources provided by the NCHRP project panel members. The research team reviewed the following sources for potential RNSs: • AAA Foundation for Traffic Safety’s Traffic Safety Issues of the Future: A Long Range Research Agenda • AASHTO research needs, as defined in their plan: Toward Zero Deaths: a National Strategy on Highway Safety • AASHTO Safety Management Task Group • HSIS Research Topics • HSM Part D/CMF most wanted list/Knowledge gaps identified in the HSM • HSM webinars - topics of interest for CMF development discussed during HSM webinar series • Input from FHWA Office of Safety and Office of Safety Research and Technology • Knowledge gaps identified in the work plan for the 2nd edition of the HSM, NCHRP 20-7(279) • Pedestrian Safety Program Strategic Plan • Research Problem Statements from key TRB committees (including unsolicited research needs, which were submit- ted in September 2011 and are presented in Appendix L) • SHRP2 Safety Project C H A P T E R 3 Implementing the Prioritization Procedure

28 ideas. Another goal of the form was to have the proposer conduct some of the initial work such as searching for “cur- rent knowledge” before proposing a new research topic. This is the primary reason for including the treatment-related questions that must be answered by searching the HSM and the CMF Clearinghouse – the two primary sources of cur- rent knowledge of treatment effects. Prioritizing CMF Research Topics This section presents the results of the CMF-related applied research prioritization, including the method used to prescreen the initial list to a more manageable number of topics. Prescreening CMF Topics The combination of suggested research topics from all sources resulted in a list of 311 CMF research topics. Time constraints under the research project did not allow for conducting the VOR prioritization for all of these topics. Thus, the team determined that 50 topics would suffice to demonstrate the VOR method and prioritization process. Several steps were taken to narrow the list of RNSs to a smaller number. The steps are detailed below and include the following: 1. Eliminate statements not meeting scope of project. 2. Eliminate vague statements. 3. Eliminate or combine similar statements. 4. Eliminate statements related to highly rated CMFs in the CMF Clearinghouse. 5. Eliminate statements with low potential to affect fatal crashes. Eliminate Statements Not Meeting Scope of Project The first step of the prescreening was to eliminate RNSs that did not meet the scope of the project. This step elimi- nated statements that specified strategies that would not be implemented by a public agency for any reason based on past research, or professional input or knowledge on the topic (e.g., curb and gutter on freeway sections); were not related to operations or infrastructure modifications (e.g., roadside memorials); or were the topics of research already underway. Eliminate Vague Statements The list of RNSs was then examined to determine if each RNS had enough description and specificity to be used in the prioritization process. Statements were eliminated if they were too vague or did not have the information necessary • Topics identified by the National Highway Research and Technology Partnership • TRB RNSs Search Engine—search keywords “crash”/ “accident” • Unfunded high-priority NCHRP projects found in NCHRP Report 617 • Other sources identified by the project team, with input from the NCHRP panel A list of some of the major research topics that were con- sidered from these various sources is given in Appendix M, which is available on the TRB website at: http://www.trb.org/ Main/Blurbs/169240.aspx. RNSs were reviewed and a table was created with individual parameters to identify the related HSM chapter; focus area; research topic/title; research area (in terms of CMF, other applied research, or fundamental); user audience; research applicability; project description; source; year; status; and other comments. Research topics were added based on gaps identified from the CMF Clearinghouse, AASHTO and TRB Committees, and other sources. The project team reviewed and analyzed the quality of CMFs found in the HSM and discovered that 40 treatments were lacking information on applicability (e.g., exposure, target crash, site and roadway specification) even though they may have small standard errors. These were included in the RNS table and developed as research ideas. A total of 883 RNSs were developed follow- ing the activities described above. For each non-fundamental research topic, the title was modified to indicate the site type and roadway type, and reflect the decision faced by practitioners (e.g., what is the safety effect of providing one access point within the influ- ence area of a signalized intersection at a major arterial road with four or more lanes for both directions of travel). These parameters are needed when using the prioritization method for RNS screening. Typically, this process gener- ated two or more RNSs by breaking an original RNS into specific scenarios (e.g., site type, roadway type, project deci- sion, etc.). There was considerable variability in the level of detail provided for potential research topics, particularly among the various sources. In many cases, there was insufficient detail provided to allow for a meaningful analysis of the research need. It was determined that a template would help to (1) generate thoughts from the proposer, (2) improve con- sistency among proposed topics, and (3) provide the level of detail needed to prioritize research needs. As part of this proj- ect, a form was developed for use by those who propose new research issues for consideration (see Appendix N). The form is designed to capture enough information about the pro- posed research project to allow for prioritization without becoming so cumbersome that it discourages good research

29 then it would be less of a priority to conduct additional research on that topic. Eliminate Statements with Low Potential to Affect Fatal Crashes Finally, a query was conducted to determine how many fatal crashes would potentially be addressed by the counter- measure of interest for each RNS. To determine this number, the team performed queries of the NHTSA FARS for the year 2009. The team members performing the FARS query used parameters as indicated by the RNS title and description to make reasonable assumptions about what types of crashes would be addressed by the countermeasure. • Example: “Safety benefits of providing illumination at isolated pedestrian locations on 2-lane rural roads.” The FARS query parameters for this RNS were pedestrian crashes at nighttime on rural 2-lane roads. The result of this query was 485 fatal crashes. The results of the FARS query provided an estimate of the maximum number of fatal crashes that could be addressed by each RNS. Those with higher numbers of fatal crashes to address were considered to be a higher priority than those with lower numbers. A threshold of 500 potential fatal crashes was used to further eliminate statements from the list. The 500-crash threshold was selected because it was a threshold which eliminated a sufficient number of RNSs to bring the list down to a more manageable number for the next stage of the process. Additionally, the 500-crash thresh- old was low enough to allow for the possibility of including RNSs on bicycle or pedestrian topics, which had 630 and 4,092 fatal crashes in 2009, respectively. After these elimination steps were completed, the list con- tained 37 RNSs. These 37 RNSs were submitted through the prioritization process to calculate a VOR for each one. Estimating the Value of Conducting Research to Develop CMFs This section presents the results obtained after implement- ing the VOR method for estimating a dollar value associated with conducting research to develop CMFs. The overview of the VOR method was presented in Chapter 2. A more detailed description of the VOR method is available in Appendix C. Appendix D describes the method for estimating the mean and variance of CMFs: estimating the variance and conse- quently the standard deviation of a CMF is necessary in order to properly apply the VOR method. Appendices E, F, and G describe the Excel tool that was developed to implement the VOR method. The VOR method was applied to 37 CMF to perform other steps in the elimination and prioritization process (e.g., specification of the countermeasure of interest). • Example: “Operational and safety effects of pedestrian facil- ities at steep grades approaches.” This statement did not describe a specific countermeasure that would be studied. • Example: “Safety Effects of providing distance markers on freeways and expressways.” This statement did not provide enough description of what was meant by “distance mark- ers” so that the team could be sure they were prioritizing it correctly. Eliminate or Combine Similar Statements The list of RNSs was further examined to determine if any statements should be combined based on similarities in scope. Twenty-three statements were combined with other statements because they addressed the same countermeasure on different road classes or the countermeasures were similar and could be logically combined into one research project. • Example: “Safety Effects of Modifying Lane Width of 4- and 6-lane Urban Arterial Streets” and “Safety Effects of Modify- ing Lane Width of 4- and 6-lane Suburban Arterial Streets” were combined as “Safety Effects of Modifying Lane Width of 4- and 6-lane Urban and Suburban Arterial Streets.” • Example: “Safety Effects of Implementing Area-wide Traf- fic Calming in Suburban Areas” was eliminated based on similarity to another RNS entitled “Safety Effects of Traf- fic Calming Measures along Suburban Arterial Corridors (speed humps, chicanes and markings to be included).” Eliminate Statements Related to Highly Rated CMFs in the CMF Clearinghouse The next step was to check the list against the FHWA CMF Clearinghouse to determine if the proposed CMF research was similar to an existing entry in the Clearinghouse. Each statement was given a value of “yes” (statement matched exactly with a countermeasure in the CMF Clearinghouse), “related” (statement was similar to a countermeasure in the CMF Clearinghouse but not an exact match), or “no” (statement did not match any countermeasure in the CMF Clearinghouse). The team made use of the star quality ratings provided for each CMF entry in the Clearinghouse, denot- ing the reliability or quality of that CMF. For statements that had an exact or related match in the Clearinghouse, the team recorded the highest star rating of any CMF associated with the countermeasure, with the intention of using the star rating as a way of prioritizing the research statements. That is, if an RNS proposed research on a countermeasure that already had a highly rated CMF in the Clearinghouse,

30 5. Countermeasure information. This included estimates of the following: a. The CMF for Total Crashes – This was obtained from the CMF Clearinghouse, the HSM, and in some cases, review of studies that were not in the CMF Clearinghouse or the HSM. In some cases, CMFs were not available in pub- lished studies for the particular treatment under consid- eration. This required the project team to use their judg- ment in determining the appropriate CMF estimate. In some cases, the project team used the information about the CMF from a closely related treatment. (It is impor- tant to note that treatments with highly rated CMFs in the Clearinghouse had already been eliminated based as part of the prescreening procedure discussed earlier.) b. The Standard Deviation of the CMF before Proposed Research is Undertaken – If CMFs from multiple stud- ies were available, then the standard deviation of the CMF before research was obtained based on the proce- dure described in Appendix D. The CMF worksheet in the Excel tool implements this procedure. If only one CMF was available, then the standard deviation was assumed to be 0.2*CMF. This assumption was based on the results from two recent studies presented at the 2012 Annual Meeting of the TRB that estimated the standard deviation of the CMF in addition to the stan- dard error of the estimate of the CMF: i. Paper 12-1658 entitled Safety Effectiveness of Convert- ing Signalized Intersections to Roundabouts, and ii. Paper 12-2521 entitled Crash Modification Factors for Changing Left Turn Phasing. c. The Standard Deviation of the CMF after Proposed Research is Undertaken – The procedure for estimat- ing this parameter is described in Appendix D. Esti- mating this parameter requires an estimate of the standard error of the CMF for the new research. Here, we assumed that the new research will have a data col- lection plan and statistical methods that are at least as rigorous as those used in the previous projects such that the standard error from the new project will match that of the previous projects. If previous research did not provide standard errors for the CMFs, the project team’s judgment was used in estimating the standard error. For treatments mainly aimed at reducing pedes- trian crashes, the standard error of a minimum of 0.1 was used unless previous research results showed stan- dard errors less than 0.1. In other cases, the standard error corresponding to the CMF being statistically dif- ferent from 1.0 at the 0.05 significance level was used. Estimated VORs Table 5 shows the VOR (rounded to the nearest ten thou- sand dollars) for each problem statement sorted in decreasing research need statements that were obtained after prescreen- ing the 311 CMF research need statements to remove those that the research team felt were of low priority (the pre- screening process is discussed in the previous section). Note: The appendices for this report are available online via the TRB website at: http://apps.trb.org/cmsfeed/TRBNetProject Display.asp?ProjectID=2727. Inputs for Implementing the VOR Method The following is a brief overview of the inputs that were needed in order to use the Excel tool for determining the VOR for each problem statement. Further details about the inputs for each problem statement, along with the assumptions, are provided in Appendix P. 1. The number of miles by functional class (for treatments on roadway segments) and the number of intersections (by type and functional class of the major road) that could be treated. To estimate this parameter, the project team used information from multiple sources: a. HPMS – this database provides an estimate of the num- ber of miles by functional class in the United States. b. HSIS – this information system has data from seven states in the nation with detailed information about the roadway inventory for state maintained roads. c. Percentage of miles or intersections that could be treated. This was estimated based on a survey of project team members and states represented in the NCHRP panel. 2. Countermeasure implementation cost information. This included an estimate of the initial cost of the project, the annual maintenance cost, the countermeasure service life, and the discount rate (assumed to be 3.0 percent). This information was estimated for each countermeasure by conducting a search of state DOT contracts and letting lists, public agency reports on the costs of safety strategies, federal and national reports on costs of countermeasures, news reports on safety strategies implemented in cities or towns, and Internet searches targeted at specific coun- termeasures. The full database of countermeasure costs compiled for this effort is presented in an accompanying spreadsheet and described in Appendix R, which is avail- able on the TRB website at: http://apps.trb.org/cmsfeed/ TRBNetProjectDisplay.asp?ProjectID=2727. 3. Target crash cost information. This was estimated based on the severity distribution of the target crash. The infor- mation from Council et al. (2005) (FHWA-HRT-05-051) was used for estimating the cost for each severity level. 4. Limiting benefit-cost ratio. The limiting benefit-cost ratio defines the minimum ratio above which a project is deter- mined to be worthy of funding. Different states may use a dif- ferent value for this parameter. For this exercise, we assumed the limiting benefit-cost ratio to be 1.0 for all projects.

31 Research Topic Title/Project VOR/Year Safety effects of installing post-mounted delineators on rural two-lane roads - tangents and curves $ 165,160,000 Safety effects of installing transverse rumble strips prior to a horizontal curve requiring speed reduction on a two-lane rural road $ 151,010,000 Safety effects of providing a sidewalk at urban and suburban arterials in place of a shoulder, including park and transit areas $ 98,090,000 Safety effects of in-lane advance curve warning marking on rural two-lane roads $ 58,800,000 Safety effects of installing raised pedestrian crosswalks at intersections on urban and suburban arterials $ 57,720,000 Effects of multiple parallel turn lanes on pedestrian-vehicle conflicts and safety at urban signalized intersections $ 30,970,000 Safety effects of placing converging chevron pattern markings on rural two-lane roads - curves $ 22,020,000 Safety prediction models for rural two-lane roads with various offsets of trees and other vegetation $ 19,660,000 Safety effects of narrowing roadway at pedestrian crossings $ 19,540,000 Safety effects of changing speed limits on urban and suburban arterials (both system-wide and speed zoning) $ 17,330,000 Safety effects of LED lighting in reducing pedestrian nighttime crashes $ 14,980,000 Safety effects of installing continuous shoulder rumble strips on rural multilane divided roads $ 10,210,000 Safety effects of in-lane advance curve warning marking on rural multilane highway $ 9,170,000 Safety effects of installing post-mounted delineators on rural multilane roads $ 6,790,000 Safety effects of installing centerline rumble strips at multilane roads with median barriers $ 4,980,000 Safety effects of installing “Botts’ dots” on rural two-lane roads $ 4,910,000 Safety effects of placing converging chevron pattern markings on rural multilane highways - curves $ 2,510,000 Safety effects of exclusive pedestrian signal phases at urban intersections $ 1,480,000 Safety effects of installing continuous shoulder rumble strips on rural multilane undivided roads $ 1,120,000 Develop a CMF for installing wide edgelines (8 in.) on rural two-lane roads versus the standard (4–6 in.) $ 1,090,000 Safety effects of retiming signal change intervals to ITE standards (four-leg urban intersections) $ 840,000 Safety effects of reversible flow lanes along five-lane arterial street corridors $ 790,000 Safety effects of widening external paved shoulder of two-lane rural roads $ 700,000 Safety prediction models for rural multilane highways with various offsets of trees and other vegetation $ 570,000 Safety effects of removing roadside obstacles along urban arterial streets $ 480,000 Safety effects of adding lanes by narrowing existing lanes and shoulders of four- and six-lane Urban and Suburban Arterial Streets $ 400,000 Safety effects of installing “Botts’ dots” on rural multilane roads $ 270,000 Safety effects of retiming signal change intervals to ITE standards (three-leg urban intersections) $ 200,000 Safety effects of increasing the forgiving area from current standards for freeways and expressways (four and more lanes) $ 150,000 Safety effects of widening external paved shoulder of four- and six-lane multilane highways $ 130,000 Develop a CMF for installing wide edgelines (8 in.) on urban and suburban arterials versus the standard $ 110,000 Safety effects of improving superelevation (to meet standards) of horizontal curve on rural multilane highways $ 100,000 Develop a CMF for installing wide edgelines (8 in.) on freeways and expressways versus the standard (4–6 in.) $ 50,000 Safety effects of increasing the forgiving area from current standards for rural multilane (four and more lanes) highways $ 30,000 Develop a CMF for installing wide edgelines (8 in.) on rural multilane highways versus the standard (4–6 in.) $ 20,000 Safety effects of changing speed limits on rural two-lane roads (both system-wide and speed zoning) $ 0 Safety effects of changing speed limits on rural multilane roads (both system-wide and speed zoning) $ 0 Table 5. VOR results for CMF research topics.

32 order of the VOR. In general, higher values of VOR indicate that there is more value in conducting research on that topic. Hence, the VOR can be used for prioritizing the projects. In order to determine the VOR for a group of projects, the user can add the VOR for the individual projects within that group, e.g., if an agency is interested in conducting research for developing CMFs for the “Botts’ dots” treatment for rural 2 lane and multilane roads, the VOR for such a research study would be $4,910,000 (VOR for rural 2 lane roads) + $270,000 (VOR for rural multilane roads) = $5,180,000 (combined VOR for rural 2 lane and multilane roads). Details of the inputs and results for each of the CMF- related RNSs are given in Appendix P. Sensitivity Analysis and Concluding Thoughts As discussed in Chapter 2 and Appendix C, there are many factors that influence the VOR of a particular research study. One such factor is the number of target crashes for a particular treatment, which is a function of the number of miles/sites that could be treated. If the number of target crashes is high, the VOR will be high as well. As discussed earlier, the number of miles/sites was estimated using data from HPMS, HSIS, and an estimate of the percentage of target miles that could be treated (based on the judgment of the project team and a couple of the states). Since this is a critical input to the implementation of the VOR method, future applications of the VOR method to rank potential CMF research projects for a particular jurisdiction should strive to estimate the number of target miles/sites for a particular treatment by making use of the jurisdiction’s roadway inventory files in addition to the expertise of local engineers. Another factor that influences the number of target crashes for a particular treatment is the distribution of crashes in the target sites. For applying the VOR method, SPFs using data from an HSIS state were estimated and calibrated to represent the United States using the procedures discussed in Appendices E and F, which are available on the TRB website at: http://apps. trb.org/cmsfeed/TRBNetProjectDisplay.asp?ProjectID=2727. For future applications of the VOR method to a particular jurisdiction, SPFs calibrated using data from that jurisdiction may provide more accurate results. The VOR is also dependent on the implementation cost and the annual maintenance costs of a particular treatment. In this study, we compiled data from various jurisdictions and used a number that the project team considered appro- priate. For future applications of the VOR method to a par- ticular jurisdiction, local estimates of implementation and maintenance costs should be used. Finally, the VOR is dependent on the estimated CMF and the difference between the standard deviation of the CMF after the proposed research and the standard deviation of the CMF before the proposed research. If the difference between the two standard deviations is high, the VOR will be high as well. As discussed earlier, estimating this difference between the two standard deviations requires an estimate of the stan- dard error of the CMF for the new research. In this study, we assumed that the new research will have a data collection plan and statistical methods that are at least as rigorous as those used in the previous projects such that the standard error from the new project will match that of the previous projects. If previous research did not provide standard errors for the CMFs, the project team’s judgment was used in estimating the standard error. For treatments mainly aimed at reducing pedestrian crashes, the standard error of a minimum of 0.1 was used unless previous research results showed standard errors less than 0.1. Since the assumption of 0.1 for the stan- dard error was somewhat arbitrary, the research team felt the need to conduct some sensitivity analysis for four research problem statements to assess the impact of using different standard errors (i.e., standard errors of 0.05 and 0.15 in addi- tion to 0.10). Table 6 shows the results of this exercise. The sensitivity analysis illustrates the importance of tak- ing care in estimating the standard error. While the relative rank remains the same among these four cases for the different Research Topic Title/Project S.E. = 0.05 S.E. = 0.10 S.E. = 0.15 Safety effects of exclusive pedestrian signal phases at urban intersections $ 360,654 $ 1,481,461 $ 3,503,610 Safety effects of providing a sidewalk at urban and suburban arterials in place of a shoulder, including park and transit areas $ 23,931,321 $ 98,093,131 $ 230,880,661 Safety effects of narrowing roadway at pedestrian crossings $ 4,771,049 $ 19,541,285 $ 45,945,208 Safety effects of LED lighting in reducing pedestrian nighttime crashes $ 3,711,613 $ 14,980,095 $ 34,224,253 Note: S.E. denotes standard error. Table 6. VOR per year for various values of assumed standard error.

33 standard errors, it is clear that the standard error has a sig- nificant impact on the absolute VOR. If the absolute VOR is erroneously inflated, then this could result in the selection and funding of projects that are not cost effective. Further work is recommended to improve the manner in which the inputs are estimated when data are not available. Prioritizing Fundamental Research and Non-CMF Applied Research Topics This section presents the results of the Fundamental and Non-CMF Applied Research prioritization. Estimated Relative Utility Index (RUI) The RUI was computed for all applicable RNSs that were categorized as fundamental or non-CMF applied research. There were a total of 21 fundamental research topics and 29 non-CMF applied research topics. Table 7 and Table 8 provide the prioritized lists of fundamental and applied non- CMF research, respectively, including the research topic, focus area, project description, and RUI. The research topics are sorted from highest to lowest priority within each table based on the RUI. The complete prioritization results are presented in Appendix Q, including the individual scores for the factors included in the RUI. Table Q.1 (in Appendix Q) provides the Research Topic Focus Area Project Description RUI Updating Databases of In-depth Crash Attributes - Expansion to F- SHRP 2 Data Management Enhance knowledge about roadway crashes for improved understanding of crash causes. Continue maintenance and digitization of the library, including database updates and online accessibility improvements. 83 Develop/Test Procedures to Identify Locations for Cost-Effective Programs of System-Wide Improvements Network Safety The purpose of this project is to develop and test procedures to identify locations for cost-effective programs of system-wide/region-wide/area-wide improvements. 81 Develop a Tool for Assessing the Extent to which a New Statistical Method or Modeling Approach Succeeds in Identifying The Cause- Effect Relationship in the Data Evaluation Methods With highway safety data, cause-and-effect conclusions obtained through the use of existing statistical methods are unreliable. The objective of this research project is to develop a tool for assessing the extent to which a new statistical method or modeling approach succeeds in identifying the cause-effect relationship in the data. 81 Development of Driver Behavior Models for Structural Modeling Evaluation Methods The objective of this project is to develop realistic representations of road-user behavior for use in structural models, with consideration given to all road users, including drivers, pedestrians, and motorcyclists. The focus should be on driver behaviors that are associated with infrastructure design, traffic control, and vehicle operation (e.g., reaction time, visual acuity, speed choice, etc.). 81 Establish Criteria to Assess the Validity of Surrogate Measures and Safety Relationships. Evaluate and Validate Candidate Surrogate Measures (Continuation of Research Idea #355) Evaluation Methods The objective of this project is to conduct a state-of-the-art review of knowledge in the area of surrogate safety measures and synthesize the findings. The synthesis should identify and define candidate surrogate measures. It should also establish criteria to assess the validity of alternative surrogate measures. The criteria should consider the use and usefulness of each measure in road safety evaluation. Thereafter, research is needed to identify potential roles for each candidate measure (e.g., for countermeasure evaluation, or as an independent variable in a safety prediction model). This research component would consist of a series of separate research projects. Each project would evaluate and validate one candidate surrogate safety measure or one specific class of related measures (e.g., surrogates from simulation modeling, surrogates from field studies). 81 Refine Geometric Design Models and Process Including the Science of Safety Alignment The objective of this project is to perform a critical review of the format, structure, and basic assumptions included in the AASHTO Policies governing geometric design of highways and streets. AASHTO presents basic geometric guidance information inconsistently. A critique of all models is needed as to their adequacy and applicability across the range of location, traffic conditions and functional classification. This research effort should directly involve the AASHTO Geometric Design Task Force and Subcommittee on Design. The inclusion of the Science of Safety and the HSM are key components of this research project. 80 Develop a National Data Warehouse and Archived Data Data Management The purpose of this project is to develop a data warehousing and archiving system. As a result, data previously gathered and analyzed could be available for future analyses. 80 Developing Methodologies to Determine Cost–Benefit of Data Investment Data Management Encourage the use of information on the basis of quality data and analytical processes to make better safety decisions. Develop methodologies to determine cost-benefit of data investment. 73 Understanding the Influence of Road Features on Crashes - Expansion to F-SHRP 2 Data Management Enhance knowledge about roadway crashes for improved understanding of crash causes. Pursue second phase of rollover study to link empirical results to simulation analysis to establish a more comprehensive understanding of the influence of road features on rollover crashes. 71 Table 7. Prioritization results for fundamental research topics. (continued on next page)

34 Research Topic Focus Area Project Description RUI Investigate the Accuracy of Automated Pedestrian/Vehicle Conflict Video Data Collection in Comparison with Human Observations Pedestrians Research on the use of video data collection to detect, measure, and evaluate pedestrian/vehicle conflicts and the accuracy compared to human observations. 60 Motorcycle Crash Causation Study Motorcyclists Devising effective crash and injury countermeasures requires comprehensive researching into current causes of motorcycle crashes and defining the population at risk. The objective of this research is to conduct on-scene, in-depth research on motorcycle crashes and user population at risk, and to harmonize such research with previous studies. 59 Developing Valid Estimates of Motorcycle Exposure for Safety Analysis Data Management The objective of this research is to identify a reliable means for estimating motorcyclist exposure for motorcyclist safety analysis. 58 Evaluation of Existing and New Commercial Vehicle Exposure Measures to Generate Meaningful Safety Estimates for Freeways and Multilane Highways Using these Surrogate Measures Commercial Vehicles The objective of this research is to evaluate the validity and usefulness of existing and potential exposure measures corresponding to key crash data variables for the purpose of generating more valid and meaningful relative risk estimates. This may include surrogate exposure measures, but should include actual measures of normal driving, including those based on: 1. Highway/traffic monitoring. 2. Naturalistic driving baseline (random epoch) data. 3. Fleet-based real-time documentation of random trip “exposure points.” 4. Government or carrier records (e.g., Hours-of-Service [HOS] logs). 5. Surveys. 6. Non-crash controls (e.g., non- crash trucks and drivers). 57 Empirical Testing of W-Beam Guardrail to Expected Safety Performance Roadside Furniture Evaluation of W-beam guardrail failures to the expected performance. 56 Development of a Safer Concrete Barrier Roadside Furniture The objective of this research project is to develop a new concrete barrier design that can be as economical and durable as safety shape and single slope barriers without producing the same rollover potential. The new barrier shape should essentially eliminate vehicle climb, minimize the risk of head-slap against the barrier, and be slip-formable. 54 Evaluation of the Applicability of and the Effect of Detection Technologies Related to Pedestrian Real-Time Data (Walking Speeds, Crossing Times, Etc.) and Vehicular Operations at Intersections Toward Enhancing the Safety and Operations of Signalized Intersections at Urban Collector and Arterial Roadways Advanced Technology and ITS The purpose of this project is to test and apply existing detection technologies to capture real-time pedestrian walking speed/location data and behavioral needs in conjunction with vehicular traffic operations to accommodate the data input in a new intersection signal system. 48 Investigate Measures to Enhance Interactions Between Pedestrians and Large Commercial Vehicles (Trucks and Buses) in Urban Areas Pedestrians Research identifying pedestrian safety improvements with regard to large commercial vehicles, especially in urban areas. 47 Develop Methods and Tools for Design and Evaluation of Roadway Departure Programs and Treatments. Develop Finite Element Models for Representative Vehicles to Support Simulation of Various Types of Crashes Evaluation Methods Develop methods and tools for design and evaluation of roadway departure programs and treatments. Develop finite element models for representative vehicles to support simulation of various types of crashes. 70 Data Collection and Analysis of Vehicle Paths and the Safety Effects of Alternative Curve Design Elements (Compound Circular Curves, Spiral Transition, and Tangent-To-Curve Transitions) Alignment Develop an experimental and/or field study protocol to collect data to study the nature of vehicle paths and operating speeds on transition curves, focused primarily on freeway exit ramps. Determine if, and how much, excessive lateral vehicle shifts are caused by compound circular curves, spiral transition, and tangent-to-curve transitions; compare the results with collision and speed reduction data. 69 Assessment of Compatibility of Modern Vehicles and Roadside Safety Hardware Roadside Furniture Assessment of compatibility of modern vehicles and roadside safety hardware. There is concern that the newer model vehicles may be incompatible with current roadside hardware that was developed using older models in crash tests, and that this problem will become worse in the future. 63 Develop an In-depth Understanding of Crashes with Utility Poles Roadside Furniture The objectives of this project are to (1) Identify the human, site, and traffic conditions associated with crashes involving poles and (2) Combine those characteristics with those already available related to vehicle type size, speed, and trajectory to develop better empirical/analytical methods of separating the predictable collisions, subject to cost-effective treatments from the purely random collision, that which is not reasonably predictable. 61 Table 7. (Continued).

35 Research Topic Focus Area Project Description RUI AASHTO Design Criteria and Design Model Research including the Science of Safety (Relates to Project #69 and #536) Alignment Based on the findings from the previous project, a series of research studies to fill the gaps in design policy formulation would occur. The following issues (which would need confirmation) are believed to represent core needs. Task 1. AASHTO Horizontal Curve Design Model. Task 2. Roadside Design Criteria for the Urban Environment. Task 3. Cross Section Design Criteria for the Urban Environment. Task 4. Relationship of Level of Service to Substantive Safety. Task 5. Influence of Geometric Design Dimensions on Highway Maintenance. Task 6. Discretionary Decision-making, Tort Law, and Risk Management State Practices. Task 7. AASHTO Sight Distance Design Models. 83 Demonstrate Application of Surrogates and an Understanding of Safety Issues (Continuation of Research Idea # 356) Evaluation Methods This project would follow the previous one (356). It would consist of a series of separate research projects, each project focusing on one of the surrogate safety measures evaluated in the previous research project and identified as having the most promise. One objective will be to demonstrate the application of the surrogate safety measure (or class of related surrogate measures) for safety evaluation. A second objective will be to evaluate the ability of each safety surrogate (or class of related surrogate measures) to describe facility safety, or to quantify the safety effect of a treatment. 82 Updating the Cost of Crashes Data Management Develop and deploy support that facilitates an evidence-based decision-making approach to improve safety. Update Cost of Crashes Report and Guidance. 81 Evaluate and Improve Criteria for Collecting Speeding-Related Crash Data Speed Better define the relationship between speed and safety. Evaluate and improve criteria for collecting speeding-related crash data. 76 Develop Safety Prediction Models for Corridors with Different Access Point Density on Multilane Highways and Expressways Access Management Develop safety prediction models for corridors with different access point density on multilane highways and expressways. 75 Develop a Program for New Safety Prediction Models for Part C of HSM Safety Tools Research is needed to expand the range of intersection types addressed in predictive methods in Chapters 10, 11, and 12 in HSM Part C. 72 Safety Effects of Dynamic Speed Feedback Signs on Curves along Two-lane Rural Roads Speed Better define the relationship between speed and safety. Evaluate dynamic speed feedback signs on curves. 71 Development of National Service Level Criteria for the Interstate and National Highway System for Rest Area Facilities With Explicit Safety Impacts Rest Area The objectives of this research effort are to develop national service level criteria for safe and secure rest area facilities development and maintenance along interstate highways. 64 Safety Effects of Increasing Drivers' Response to the First Signalized Intersection When Entering Urban Areas (Following Rural Travel) Intersection Traffic Control Investigate methods to increase the safety of the "first signal" in the urbanized area such as real-time activated vs. static flashers. 63 Barrier System Maintenance Procedures Based on Damage and Potential Safety Impacts Roadside Furniture The final product expected from this research is an objective set of guidelines for determining the limits of damage to a longitudinal barrier that will not cause the barrier to have unacceptable safety performance. The guidelines will also address which safety feature components require replacement or refurbishment when repair is deemed to be necessary. 62 Best Practices in the Identification and Prioritization of High Pedestrian Crash Locations/Areas Pedestrians Research to generate best practices in pedestrian problem area identification (including the use of GIS, crash data, and land use data) and prioritization to assist practitioners in accurately and systematically identifying pedestrian risk areas that could be proactively treated. 62 Develop Safety Prediction Models for Corridors with Different Access Point Density on Two-Lane Rural Roads Access Management Develop safety prediction models for corridors with different access point density on two-lane rural roads. 61 Safety Performance of Large Trucks by Volume and Time-of-Day for Planning and Operations of Freeway and Multilane Highway Networks Commercial Vehicles The recommended study would review the literature and available data sources on large truck crash rates and risks by time-of-day. This would include crash distributions; crash harm distributions (derived from severity statistics); mileage exposure distributions; time-of-driving distributions (i.e., from naturalistic driving studies); and any other data metrics representing either a numerator or denominator in the overall time-of-day risk assessment. A successful study would derive time-of- day crash risk functions to inform carrier dispatching and other operational decisions. Potential approaches to how to incorporate heavy trucks into freeway and multilane highway network operations with safety explicitly quantified. 60 Development of Incident Duration and Secondary Crash Prediction Models for Freeways and Multilane Highways Advanced Technology and ITS Development of an incident duration and secondary crash occurrence prediction models on freeways and multilane highways for use by Transportation Management Centers, so users can make more informed travel choices. 58 Table 8. Prioritization results for non-CMF applied research topics. (continued on next page)

36 Table 8. (Continued). Research Topic Focus Area Project Description RUI Safety Predictive Models for Multi-Vehicle Collisions Based On Tunnel Characteristics and Service Areas, and Response Needs Tunnels Determine probabilities for multi-vehicle collisions based on tunnel characteristics and service areas, and develop predictive models for crashes in tunnels and response requirements. 58 Development of a Bicycle Safety Prediction Methodology for Road Segments and Signalized and Unsignalized Intersections Bicyclists The proposed research will explore exposure measures (effects of distance, time, traffic volume, number of lanes, etc.) and develop a predictive method for geometric treatments for bicyclists: 1. Along road segments (marked bicycle lanes, separated bicycle lanes, and raised bicycle lanes). 2. At signalized intersections. 3. At unsignalized intersections. 57 Development of Safety Prediction Models for Pedestrian Target Crashes at Midblock Locations with Marked Crosswalks Along Urban Multilane Roadways Pedestrians Development of safety prediction models for pedestrian target crashes at given design characteristics of midblock locations with marked crosswalks - urban multilane roadways. 56 Development of Safety Prediction Models for a Set of Two Three-Leg Intersections Intersection Geometry Development of safety prediction models for a set of two three-leg intersections on rural two-lane roads, rural multilane highways, and urban and suburban arterials. 56 Operational and Safety Effects of Marking a New Crosswalk at Midblock Locations Pedestrians Research to evaluate how new pedestrian facilities affect pedestrian exposure data and to determine the increased facility use by conducting before and after case studies of pedestrian facility projects. 55 Develop Safety Predictive Models for Roundabouts in Rural and Urban Conditions with Four Approaches and One- or Two-Lane Entries/Approaches Intersection Geometry Develop Predictive Methods for Roundabouts for Part C of the HSM. 53 Safety Impacts to Pedestrian at Different Levels of Visibility and Sightlines at an Urban Intersection Pedestrians A study of the safety impacts of lighting and other roadside elements such as newspaper boxes, shrubs, and other obstructions on driver-pedestrian visibility at urban intersections. 52 Best Practices and Pedestrian Safety Concerns Related to Transit Access in Urban Areas Pedestrians Research on (1) pedestrian safety concerns around transit (including bus stops, light and heavy rail, and streetcars), around at-grade rail crossings, and along railways and on (2) best practices related to transit access and increasing transit ridership through pedestrian facility improvement. 52 Development of National Service Level Criteria for Freeways and Multilane Highways with Four Lanes or More with Explicit Safety Impacts of Different Drainage Features Roadside Furniture The objectives of this research effort are to develop national service level criteria for functional drainage features (culverts, inlets, rutted pavement lanes that pond water, ditches, under drains, etc.) for freeways and multilane highways of four lanes or more. 51 Development of a Pedestrian Safety Prediction Methodology for Unsignalized Intersections Pedestrians The proposed research will develop a predictive method for geometric treatments for pedestrians at unsignalized intersections. 49 Systematic Data Management of Crashes on Shared Use Pathways Shared Pathways The purpose of this project is to develop recommendations for a national and standardized data management for shared use pathway crash and injury data: collection, analysis, and publication system. 49 Accessible Pedestrian Signals Pedestrians Research on APS devices, specifically the impacts and benefits for non-disability users, guidance on maintenance audits and protocol, as well as guidance on where APS devices are most beneficial and should be prioritized, or where fixed-time operation should be used. 49 Safety Impacts to Pedestrian at Different Levels of Visibility and Sightlines at a Roundabout Pedestrians A study of the safety impacts of lighting and other roadside elements such as newspaper boxes, shrubs, and other obstructions on driver-pedestrian visibility at roundabouts. 44 Animal-Vehicle Crash Severity Reduction Using Advanced Technologies along Freeways and Multilane Highways Advanced Technology and ITS This study will examine innovative ways to mitigate the damage from animal- vehicle crashes using traditional measures as well as advanced technologies. Since preventing such crashes has proven to be very difficult, this study will investigate ways to reduce the severity of the crashes. 43 Costs of Vehicle-Train Collisions At Urban And Rural Crossings Highway- Railway The objective of this research is to develop a cost model that takes into account direct costs from multiple perspectives that accrue as the result of a vehicle-train collision at urban and rural railroad grade crossings. 42

37 combined list of prioritized research, including both fun- damental research and non-CMF applied research. While Table Q.1 appears to show a good distribution of the two types of research (i.e., not top-heavy with one or the other), it may be more appropriate to consider them separately. Reasons for separating the two types of research are discussed in the Lessons Learned section following this table. Tables Q.2 and Q.3 pre- sent the results separately for fundamental research and non- CMF applied research, respectively. Stakeholders may use the prioritized lists to help identify projects for funding. This is similar to the current method used to select NCHRP projects for funding, but the prioritized list will help to support the decision-making process. Lessons Learned This project is the first attempt to develop a formal method for prioritizing fundamental research and non-CMF applied research related to highway infrastructure safety. The priori- tization method evolved through the course of the project. As such, there were several lessons learned and opportunities for further refinements as discussed below. General Stakeholder Input It would likely become too cumbersome to incorporate general stakeholder input in the prioritization process for fun- damental research. It is also the goal of this project to develop a more quantitative method for prioritizing fundamental research, moving away from the traditional “voting” meth- ods. With that said, stakeholders will play a vital role in the process, including the identifying research needs to be priori- tized and helping to select the final list of topics for funding. The latter will be conducted once the list of research needs has been prioritized using the method described above. Other Applications of Prioritization Process The prioritization of CMF-based research is based on a highly rigorous and repeatable method. An attempt was made to develop a similar process for prioritizing funda- mental research, but the method has proven to be impracti- cal for a repeatable process. As such, the alternative method proposed in this project was developed for the prioritization of fundamental research. The team further determined that the prioritization method for CMF-based research would not be appropriate for non-CMF applied research. As such, the method proposed for fundamental research was adjusted slightly and used for the non-CMF applied research. There are several factors and detailed data included in the prioritization of research. It would be useful to identify high-priority research (conduct initial screening to elimi- nate low priority topics) before collecting further details and completing the more rigorous prioritization process. This would reduce the effort in collecting detailed data for each proposed research project and limit the detailed data collection to those that pass the initial screening pro- cess. The proposed prioritization process for fundamental research includes eight factors. Several of these factors could be included in a modified weighting scheme to conduct an initial screen of CMF-based research. For example, target crashes (frequency and severity) would be a quick and easy method to screen potential research topics. However, tar- get crashes alone could overlook important topics such as pedestrian and bicycle research. As such, the number, quality, and average effectiveness of existing CMFs could be included in the modified weighting scheme. These data are readily available from the CMF Clearinghouse. Stakeholder input would be another potential factor to include in the weighting scheme, but this would include a higher level of effort to solicit and summarize feedback. Program Monitoring and Evaluation For both fundamental and applied research programs, it is important to monitor and regularly (e.g., annually) evaluate the programs to ensure that they are providing the intended products. In addition, any benefits derived from funda- mental research through subsequent use in applied research projects should be documented for use in future research prioritizations. Separation of Fundamental and Non-CMF Applied Research While the same method was applied to prioritize both the fundamental research and non-CMF applied research, it may not be appropriate to compare both types of projects side-by- side. Instead, it may be more appropriate to consider funda- mental research and non-CMF applied research separately. The reason for this is that some of the factors used to determine the RUI are inherently different. For example, the factor related to the “Extent of Impacts on the Science of Safety” will likely favor fundamental research as these topics tend to have more far-reaching benefits compared to a specific non-CMF applied research project. Conversely, the probability of success is likely to favor non-CMF applied research projects as they tend to be less complex and shorter duration than fundamental topics.