Identification of High Pedestrian Crash Locations

CHAPTER 5. DEVELOPMENT OF THE GUIDEBOOK AND WEBINAR WORKSHOP

OVERVIEW

The findings from the literature and interviews were incorporated into the Guidebook on Identification of High Pedestrian Crash Locations and a 90-min webinar workshop.⁽¹⁾

DEVELOPMENT OF GUIDEBOOK ON IDENTIFICATION OF HIGH PEDESTRIAN CRASH LOCATIONS

Background

One of USDOT’s top priorities is the improvement of pedestrian and bicyclist safety. FHWA promotes safe, comfortable, and convenient walking for people of all ages and abilities. Part of this effort has been to encourage a data-driven approach to identifying and mitigating safety problems. An initial step in reducing the frequency of pedestrian crashes is identifying where they occur or where there is a concern they are likely to occur. As part of an FHWA project, the Guidebook on Identification of High Pedestrian Crash Locations was developed to assist State and local agencies in identifying high pedestrian crash locations such as intersections (points), segments, facilities, and areas.⁽¹⁾ The process of identifying high pedestrian crash locations results in a prioritized list of potential locations on the roadway system that could benefit from safety improvement projects.

Process to Identify High Pedestrian Crash Locations

Several agencies were contacted to gather information on how they identify high pedestrian crash locations. This information coupled with findings from a review of the literature generated the process shown in figure 3 . The steps are as follows:

1. Select approach. The Guidebook focuses on the traditional (also known as “reactive”) approach. If a proactive approach is preferred, the Guidebook gives suggestions on other reference documents.
   
   
2. Gather data. The typical data needed consist of crash data (including severity, crash type, contributing factors, and, importantly, the location of the crash preferably as latitude and longitude coordinates) and roadway characteristics (e.g., the number of lanes or traffic control devices present). Exposure data in the form of vehicle counts, pedestrian counts, and/or turning and crossing movement counts for specific locations may also be desired.
   
   
3. Plan assessment. The substeps are to select scale (e.g., intersections, segments, or area), performance measures (e.g., crash frequency or crash rate), and screening method (e.g., simple ranking or a more complex approach that requires software).
   
   
4. Conduct assessment. Several tools are available to assist in conducting the assessment, with most having data requirements in addition to crash data. Some of the tools can require additional training or a skill set in a GIS before an assessment can be conducted.
   
   
5. Prioritize locations. After the selected performance measure(s) and screening method(s) are applied to the study network, the resulting list of sites can be arranged using a simple ranking or by considering adjustments or community priorities.

Details about completing each of these steps are discussed in the Guidebook. The Guidebook concludes with supporting materials grouped within the following sections:

• Supplemental Material A—Example of Safety Index: provides additional details about the index the City of Los Angeles uses.
  
  
• Supplemental Material B—Screening Method Examples: presents examples of several screening methods.
  
  
• Supplemental Material C—Online Maps: presents several examples of online maps being produced by numerous city and State DOTs to show pedestrian crash data.
  
  
• Supplemental Material D—Advice From Previous Studies: discusses several previous studies that have documented analyses to identify HCLs along with different approaches to identify and rank locations.
  
  
• Supplemental Material E—Glossary: provides a list of definitions relevant to the document.

Lessons Learned During Development of the Guidebook

Most agencies now have available the geographic coordinates of crashes, which results in the ability to quickly illustrate where crashes occur. All the interviewed agencies use a GIS to identify HCLs. The agencies generally start with identifying high crash intersections and then group the sites into facilities and/or areas. GIS tools are used to aid in the grouping; however, several agencies noted that visually confirming the grouping is how they set the limits for their corridors and areas.

Agencies have considered surrogates to identify locations of concern, such as activity centers, walk scores, or citizens’ comments. Pedestrian exposure data are typically not used to identify sites because of the lack of good data for significant portions of their network. The analysis period ranges from 1 to 3 yr. The agencies noted that pedestrian and bicycle crashes are different from motor vehicle crashes and require unique efforts.

The current skill sets needed to work with crash data include familiarity with a GIS, the ability to work with attribute tables, and programming skills. Key lessons learned include the importance of the following:

• Analysts with the needed skill set are necessary.
  
  
• Partner agencies that share data are better informed and can develop their program more efficiently and effectively.
  
  
• Agencies with access to reliable geographic coordinates for crashes can more accurately locate their crashes.
  
  
• Agencies with strong support from others within the city can develop and implement stronger programs.

Some of the cities suggested that the list of sites and plans should be shared with the public so residents know where the city is performing work and how those decisions were made.

Examples of approaches used and lessons learned from previous studies include the following:

• Several methods should be used to rank sites to ensure consensus.
  
  
• Not considering pedestrian exposure in some manner could significantly affect results.
  
  
• Other measures (e.g., walkability score) should be considered along with crash data, since crashes are rare events.
  
  
• If evaluating segments, a minimum segment length is needed (suggested as at least 0.2 mi).
  
  
• During evaluation of pedestrian crashes, a longer period may be needed due to the relatively small number of reported pedestrian crashes (e.g., 5 yr rather than 3 yr).

The methods used to identify and evaluate sites with a high crash frequency have evolved, in the recent decades, in the following ways:

• The availability of geographic coordinates (latitude and longitude) for crashes has resulted in the ubiquitous use of GIS platforms for displaying the locations and density of crashes on maps.
  
  
• Certain map displays are more likely to successfully convey crash density. The use of larger symbols or color-coded symbols (e.g., green–yellow–red scale) seems to be most appropriate for quickly identifying HCLs. In most cases, showing symbols for individual crashes does not display well because the crash symbols overlap on the map.
  
  
• Map displays with zoom capability can be used to quickly identify high crash areas at a citywide scale yet still provide the option to view individual crashes at a specific intersection or street. Maps that display additional information once clicked are ideal for exploring crash patterns or attributes at a detailed level.
  
  
• Several different approaches are used to identify and display HCLs, and often, these approaches are not well documented on the mapping application. In some cases, the mapping software application automatically determines how to group nearby crashes based on view level. In other cases, agencies predetermine how nearby crashes will be grouped and displayed in the mapping software.

When considering approaches other than crashes, recent advances in statistical techniques have provided several methods and tools that can be used to identify locations with concerns for pedestrians. These techniques include SPFs, the HSM,and systemic analyses.⁽⁵⁾ The techniques provide the opportunity to allow comparisons between a city’s data and national trends. The growth of better statistical techniques also permits the profession to better handle RTM and low-sample challenges.

WEBINAR WORKSHOP

FHWA hosted a webinar workshop in May 2017 for FHWA staff and panel members. The webinar workshop had the following three objectives:

• Provide lessons learned about the identification of high pedestrian crash locations.
  
  
• Present the features of the developed Guidebook.
  
  
• Provide examples of the screening process using a real dataset to overview the Guidebook and demonstrate its use.

The webinar workshop covered the following:

• Overview and objectives.
  
  
• Advice and lessons learned.
  
  
• Process of identifying locations of interest.
  
  
• Overview of available GIS technologies.
  
  
• Demonstrations.
  
  
• Discussion.
  
  
• Closure.

Previous sections of this chapter summarize the information presented about the initial three bullets covered during the workshop. The portions of the workshop that focused on a GIS and the demonstration illustrated the steps of the process to identify locations of interest (i.e., locations with high pedestrian crashes). Emphasis was placed on the use of a GIS in identifying locations of interest, given the wide availability of appropriate data and the growing interest and use of a GIS for such data analysis platforms.

The diagram in figure 4 summarizes the basic structure of a GIS. The main components for a GIS are a core to host, manipulate, and analyze data (shown in red or as rectangles or a cylinder); inputs consisting of data and processing packages (shown in green or as cards or a barrel); and outputs consisting of data and information (shown in yellow or as rounded cards or a rounded parallelogram).

During the webinar workshop, prerecorded steps were shown to demonstrate the use of a GIS package to implement one of the screening methods to identify locations of interest. The scenario for demonstration was the application of the sliding window method to identify locations of interest along a corridor in Austin, TX. The inputs for the analysis were two layers of data: one with corridor crashes and one with the segment representing the analysis corridor, as shown in figure 5 .

The corridor main segment was split into subsegments, and buffers were created around each subsegment, each buffer representing a sliding window for analysis (shown in figure 6 ).

The data in the crash layer were then merged into each sliding window to collect the crash frequency per window. The crash frequency per window displays how crash frequency varies along the corridor using a color scale (as shown in figure 7 ).

Finally, the layer with the results of the sliding window analysis was extracted and exported to a spreadsheet for further analysis and graphics as necessary (a plot from the exported data is shown in figure 8 ). The webinar workshop was closed by taking questions and comments from participants.