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.
15 A systemic approach is inherently data driven, as data are used to identify pedestrian safety risks, locate sites with high risk features, and evaluate the cost-effectiveness of various treat- ment alternatives. The purpose of this step is to provide guidance on how to compile the data that will be needed to support all future steps in the process. The step also provides options for workarounds for some commonly missing data and gives examples of how other agencies approached this step. This step involves the following tasks, which are further described in the next sections. â¢ Compile roadway data, including traffic and pedestrian volumes (if available), for the relevant target facility types. â¢ Add land use and sociodemographic data using spatial methods to the specific locations for the relevant facility type. â¢ Count the focus crashes and add these data to the specific locations. Crash frequencies by location are the dependent (outcome) measures of safety. Completion of these tasks will result in a database, with a row of data or âobservationâ for each intersection or segment location, containing variables for all the key data in one place. Table 2 illustrates the basic structure of a database used in a systemic process (Seattle, in this instance). Each site ID is associated with a crash frequency, pedestrian and vehicle volumes, population and land use characteristics, and roadway features (note that only a partial list of sites and variables is shown). For more information on how these data will be used, read ahead to subsequent steps related to data analysis (Step 3) and site identification (Step 4), or see the Looking Ahead call-out boxes that flag important considerations made in this step, Step 2, that will have implications later in the systemic process. Compile Roadway Data for Focus Facility Type Based on the target crash types or locations (e.g., intersections, roadway segments, or both) identified in Step 1, build a database conducive to analyzing safety concerns at those locations. Ideally, the database will cover locations across the entire network or jurisdiction of interest. For many agencies, segments and intersections seem to provide the most workable options, but this may depend on how the agency compiles and stores roadway inventory data. Roadway characteristics are key factors to include in database development, as they are used as crash predictors and/or potential risk factors and can be used in the process to identify and prioritize sites for treatment. State DOTs and local agencies typically maintain many relevant features in roadway inventories. Specific roadway variables to consider are provided in Data for Intersections and Data for Segments. C H A P T E R 3 Step 2: Compile Data
16 Systemic Pedestrian Safety Analysis Traffic and Pedestrian Volume Data Along with the physical characteristics of roadway locations, in a truly systemic, risk-based process, it is important for best results to include two measures to help account for pedestrian exposure to crashes. These measures include â¢ Traffic volume data; and â¢ Pedestrian volume data for each site (collected or estimated). Definitions: Crash Predictors, Risk Factors, and Safety Performance Functions Crash predictors are any characteristic of the roadway, environment, vehicle, or population attribute that helps to predict future crashes based on a quantified association with prior crashes (such as in a safety performance function or SPF, see definition below). In other words, these are measured variables that are associated with crash frequencies and are used to estimate risk. Crash predictors themselves may not have a causal relation or theoretical basis for increasing risk, but they may serve as surrogate measures for certain risk factors. Risk factors are any attribute, characteristic, or exposure of an individual or roadway that increases the likelihood of a crash or increases the risk of a more severe injury outcome in the event of a crash. Not all risk factors can be measured from attributes associated with site characteristics (e.g., risk factors such as states of users at the time of a crash). The mere presence of a risk factor may not be sufficient to cause a crash, but a risk factor should have a plausible association with a contributing circumstance of a crash. SPFs are statistical models used to predict crashes based on site characteristics. These models always include traffic volume (average annual daily traffic or AADT) and, in the case of pedestrian crashes, SPFs should also include pedestrian volume (average annual daily pedestrian or AADP) or suitable surrogate measures. SPFs may also include other roadway features and, in the case of pedestrian SPFs, characteristics of the built and social environment around the site. Site ID No. of Pedestrian Crashes Motor Vehicle AADT Hourly Pedestrian Count % Seniors in Tract Distance to University Median Presence Crosswalk Presence â¦ 1 0 604 6 47 2.4 0 0 2 0 810 6 06 3.3 0 0 3 0 1,109 7 04 1.2 0 0 4 1 1,897 8 11 0.0 0 0 5 0 1,897 8 11 0.0 0 0 â¦ Table 2. Example database compiling key volume, land use, and roadway features for target sites.
Step 2: Compile Data 17 Both measures have been found to be important predictors of pedestrian crashes. If they are missing from the analysis, the other factors that appear to be correlated with crashes may be associated with crashes through their correlation with where people most often walk and/or drive. Failure to account for traffic volumes may lead to challenges in identifying the factors that contribute to increased crash risk or lead to incorrect conclusions about their relationship to crashes. Traffic volume data may be considered an element of roadway inventory, but some agencies may collect and store traffic and/or pedestrian volume data separately. However, traffic volume data should be linkable to roadway locations along with the other features. Traffic volume data tend to be widely available, with actual counts typically converted to an estimate of AADT. However, AADT data may be less available for non-arterial streets and local network segments in some jurisdictions. Since most agencies find that pedestrian crash and injury problems tend to concentrate on arterial and collector streets, the lack of volume data for local and/or residential streets may be less problematic or there may be some assumed lower volume categories for local streets. See the additional resources section for key references on traffic monitoring. More and more agencies are gathering pedestrian volume and trip data and working on meth- ods to generate appropriate measures of pedestrian activity to use for various purposes. How- ever, many agencies still lack networkwide estimates of pedestrian volumes, scaled to the site level for use in a systemic safety analysis. Many new tools and guides have been developed to support agencies in broadening and improving their pedestrian count data programs, and it is recommended that a pedestrian counting program be undertaken. In the meantime, there are other methods that can be used in a systemic analysis to help account for pedestrian volumes. See the troubleshooting call-out box on Page 22 and the additional resources section for more information on how to perform these tasks. Definitions: Pedestrian Exposure and Surrogate Measures Using the Safe States Alliance Consensus Recommendations for Pedestrian Injury Surveillance definition, pedestrian exposure is âan observable period or point during which a pedestrian experiences the possibility of suffering an injury related to the act of being a pedestrianâ (p. 20). See https://cdn.ymaws.com/www.safestates.org/resource/resmgr/ ISW8_Report_Final.pdf. Several constructs such as counts of pedestrians at crossings can be used to quantify pedestrian exposure to the risk of a crash or injury. In theory, not all pedestrian trips or activities result in exposure to a vehicle crash, but for the purposes of this guide, the terms pedestrian exposure, volumes, demand, and activity are used interchangeably. A surrogate measure is a characteristic or variable that may help predict crashes by approximating or capturing phenomena associated with a risk factor that may or may not be measured. A common example is the use of population and/or employment density (available data from the Census Bureau) to serve as a surrogate for pedestrian exposure when the latter is not directly measured.
18 Systemic Pedestrian Safety Analysis Other Roadway Data Needs Other features that are desirable but that may be lacking in roadway data for many agencies include pedestrian-focused facilities such as crosswalks and other crossing improvements. These and some other missing roadway features can sometimes be compiled from online aerial and street-view resources and field-checked for accuracy. Field data collection is also an option. These types of features have not often been available for system-wide analysis. It is worth noting that the presence of pedestrian facilities (e.g., marked crosswalks or median islands) may reflect pedestrian exposure or where demand is high. These features will be desirable to have during the screening and prioritization process, regardless of whether they are used in the risk analysis. Traffic speed monitoring data are also highly desirable but rarely available at a network level. Posted speed limit is typically used as a surrogate measure for traffic speed but may not accu- rately reflect travel speeds during different times of day (e.g., congested periods versus free-flow conditions). The following tables offer more suggestions on variables and measurement or ways the data have been aggregated to intersection or segment locations, if the data are not already compiled to the location types of interest. The case examples also describe risk variables that several juris- dictions have used. Data for Intersections Table 3 summarizes variables to consider compiling for intersection-focused analyses based on prior studies that have found these variables to be associated with pedestrian crash and injury Noteworthy Practice Ohio DOT reports that they use a hierarchical approach in assigning traffic volume to the intersection file. First, they use the volume information from the adjacent roadway sections. If that is not available for one or more intersection legs, they obtain any volume information that can be supplied by the metropolitan planning organization. If that is also not available, they assign a traffic volume value using default values for functional class by county. Seattle DOT had limited pedestrian count data when they began a systemic pedestrian and bicycle safety analysis project, but short-term counts were available for 50 locations across the city from several sources. Ballpark estimates of pedestrian volume were developed using the count data and characteristics of the 50 count locations. These procedures are explained in Sanders et al. (2017). The predictive model equations used variables such as nearby population density, household density, employment density, and presence of schools and university campuses to estimate pedestrian volumes at intersections across the city. The pedestrian intersection estimates were then parsed to adjacent segments. Due to the limited number of counts and other assumptions made in these estimation procedures, other measures of pedestrian activity were also included in the subsequent crash prediction models to help account for pedestrian exposure. See Case Example 1 for more information. The City also plans to collect more count data to improve future estimates.
Intersection-Related Roadway Variables Measurement Methods Traffic volume Typically, average daily traffic (ADT) or AADT is available for state road networks. Subtypes may include â¢ Major and minor road volumes (for intersection legs) â¢ Volume assignment by functional classes (surrogate measure) â¢ Turning movement counts â¢ Heavy vehicles percentage May need to collect additional data and develop estimation procedures to generate estimates for network locations not covered by regular traffic monitoring. Pedestrian volume â¢ Counts of pedestrians crossing any leg of intersection â¢ Average AADPcrossing at intersection (estimates) based on modeling of a sample of actual counts Agencies may need to develop a sampling and estimation strategy, coordinate with agencies that have count data, and/or collect additional data to improve estimation accuracy. See the forthcoming FHWA resource, Guide for Scalable Risk Assessment Methods for Pedestrians and Bicyclists (Turner et al., in press), and other resources mentioned in Additional Resources in Chapter 3. Transit stops Presence of transit stops. Note the other transit activity measures listed in Table 5. Transit measures have been found to be associated with pedestrian crash risk in both intersection and segment-based analyses. Number of traffic lanes â¢ Total number of traffic lanes (all types, all legs) â¢ Entering through lanes â¢ Number of lanes on main/largest approach â¢ Maximum number of lanes pedestrian must cross in one maneuver The data listed above are all ways traffic lanes have been counted at intersections . All generally have been positively associated with increasing crash risk. Number of intersection legs Count the total number of legs entering an intersection. (Short distance offset legs may be included.) Crosswalk length â¢ Maximum crosswalk length â¢ Major/minor road crosswalk lengths Traffic control type â¢ Signalized â¢ Four-way stop control â¢ Two-way stop control â¢ No traffic control, yield control, other On-street parking â¢ Presence of parking on one or more legs â¢ Proportion of all legs/sides with parking Commercial driveways Presence or number of commercial driveways within X distance Leading pedestrian interval Presence (or amount of time) of leading interval Pedestrian signals and detection â¢ Presence of pedestrian countdown signal heads on all legs â¢ Type of activation (active, passive, or Puffin) Unrestricted/restricted turn phasing â¢ Presence of protected pedestrian crossing phase (no left turns during pedestrian walk phase) â¢ Presence of all red during walk phase Turning lanes Presence of one or more lanes dedicated to right or left turning movements. Speed limit â¢ Highest entering speed limit of any leg â¢ Major and minor road speed limits Actual traffic speed monitoring data may be preferable, but no prior studies have been identified that included actual measured traffic speeds. Intersection skew (angle > 90Â°) Presence of one or more angles with angle > 90Â° No identified pedestrian studies included this measure, but it has been found to be associated with motor vehicle crash types; may affect sight lines and turning speeds. Crosswalk markings and type (high visibility or standard) Presence or proportion of legs with crosswalk markings Sidewalk coverage Proportion of all legs/sides of intersection with sidewalks ADA-accessible curb ramps Proportion of landing areas with ramps that meet accessibility guidance Other Other facility/roadway or relevant environmental variables as locally determined (e.g., walk signal timing per pedestrian walking speed) Table 3. Potential pedestrian crash risk variables for intersection analysis.
20 Systemic Pedestrian Safety Analysis risk. These variables were identified from the literature on crash risks at intersections, from effec- tive treatments, and from risk or conflict principles. Many more variables could be included, and local knowledge of the network and consideration of risk principles should inform the selection of priority variables. As discussed earlier, some pedestrian features (such as crosswalk markings at an intersection and ADA-accessible curb ramps) have not necessarily been analyzed with respect to pedestrian crash risk but could prove helpful to include in the database. This information will allow track- ing of features present and may aid prioritization of improvements during the screening and ranking process. Data for Segments If focusing on non-intersection or segment crash issues, consider compiling the variables in Table 4. These variables were also identified from the literature on crash risks, effective treat- ments at uncontrolled locations, and risk principles described in Step 3. There are potentially many more variables that may be important, including risks that have not been thought about or included in prior studies. However, it is also important to be realistic in developing the database. Consider basic risk principles of volumes of users, speed, distance, conflict points, and other features of the trafficway and roadside that could potentially affect pedestrian crash risk. (See the risk concepts at the beginning of Step 3.) Varied networks may also have different data needs, depending on design and operational factors. Therefore, it is important to think through the data that are important for the jurisdiction. With those cautions in mind, consider including many of the factors mentioned above that were identified from prior studies. Add Other Pedestrian Crash Exposure Measures to Facility Data For a systemic process that aims to identify locations most at risk of future collisions and treat the factors that increase risk, it is important to account for both traffic volume and pedes- trian volume in risk analysis and prioritization. In addition, other built and social environment measures have been found to be important crash predictors with respect to where crashes have tended to occur. These data types are described in this section. Looking Ahead Whether or not pedestrian exposure measures are used to identify risk factors, they will almost certainly be needed to evaluate the suitability of candidate sites for treatment (in Step 4), to help with selecting appropriate countermeasures (Step 5), and, ultimately, for evaluating cost-effectiveness and justifying the treatment plan to be implemented later (in Step 6). For example, a location that has certain roadway-related risk factors for pedestrians (i.e., multiple lanes, higher speed limits, or no median islands) may lack the land uses to generate pedestrian trips and may not warrant treatment given other potential priorities. Thus, investment in gathering and including roadway, land use, and travel volume data earlier in the process will pay off in later steps of the process.
Step 2: Compile Data 21 Segment-Related Roadway Variables Measurements Traffic volume Typically, ADT or AADT is available for state road networks. Subtypes may include â¢ Major and minor road volumes (for intersections) â¢ Volume assignment by functional class (surrogate measure) â¢ Heavy vehicle percentages Data may be less available for non-arterial streets and local networks. Agencies may need to develop a sampling strategy to cover all street/ area types and follow standard practices to generate AADT volume estimates for the entire network. Pedestrian volume It is challenging to account for pedestrian volumes crossing a length or segment of roadway. Ideally, counts of pedestrians walking along the roadway and of pedestrians crossing anywhere along a segment could be included. It may be feasible to collect counts of pedestrians crossing at non-intersection-marked crosswalk locations. AADPs for pedestrians walking along segments were included in the analyses of segment-related data for this report and found to help predict where pedestrian crashes occurred. In addition to âwalking alongâ measures of pedestrian volume, many of the measures in Table 5 were included to capture risks associated with potential pedestrian attractors or areas of pedestrian activity that could contribute to midblock crossings. Transit â¢ Presence of stops within X distance of segment midpoint or endpoints â¢ Number of stops along segment Note the other transit activity measures listed in Table 5. Transit measures have been found to be associated with pedestrian crash risk in both intersection and segment-based analyses. Total through lanes Number of through lanes (average, either end of segment; midpoint number of through lanes; or number proportionally weighted) Median with/without crossing facilities Presence of a continuous raised (not painted or two-way left-turn lane or TWLTL) median Median islands with pedestrian crossing Count of raised median islands with pedestrian pass through refuge along segment. Could consider median island presence at intersection. Two-way left-turn lane Presence of TWLTL Midblock crosswalks Presence or count of marked crosswalks with unsignalized approaches along a segment On-street parking Presence (any, one, or both sides) or proportion of segment covered by striped parking Pedestrian hybrid beacon or PHB Presence or count of the facility type along a segment Rectangular rapid flashing beacon Presence or count of the facility type along a segment High visibility crosswalk markings Presence or count of the facility type along a segment Advance stop/yield markings and signs Presence or count of the facility type along a segment Speed limit Posted speed limit or weighted average speed limit along segment Segment length Length of segment; may be estimated from spatial data Sidewalk coverage Presence of sidewalks along zero, one, or both sides, or proportional coverage from front frontage data Distance to nearest signalized crossing As described or activated beacon along same road Right- or left-turn lanes at adjacent intersections Presence or counts of different lane types at adjacent intersection Signals at adjacent intersections Proportion or number of adjacent intersections with traffic signals Table 4. Potential pedestrian crash risk variables for segment analysis.
22 Systemic Pedestrian Safety Analysis Potentially Important Pedestrian Crash Exposure Measures Each jurisdiction and each network are different, and therefore the specific measures that may be compiled and used to understand pedestrian crash risk will likely vary. However, it may be helpful to consider types of data and variables that have been used and found to help predict where pedestrianâmotor vehicle crashes tend to occur. For example, varied measures of transit activity, commercial land uses, and population measures such as average income and younger and older ages have been found to be associated with increased pedestrian crash risk in some analyses, even when traffic and pedestrian volumes were also accounted for in the analysis. Such measures should be considered for inclusion in the database, in addition to traffic and pedestrian volumes. Other built environment and demographic measures have been identified as being associated with pedestrian crashes primarily through their linkage to pedestrian activity. Measures such as population and employment density measures, household density, mode share, and others could serve as potential surrogates for pedestrian volume data if these data are not yet available. However, note that these surrogate measures may not serve as adequate substitutes for actual pedestrian volume data if volume estimates are not included at all. Troubleshooting Both traffic and pedestrian volumes are needed for systemic safety analysis. If pedestrian volume data are not available for the entire network, consider one or more of these alternatives: â¢ If count data are available for some locations, use modeling to estimate volumes for the network based on existing long- and short-term or annualized counts. â¢ Collect count data at a representative sample of locations, then use the data to estimate volumes at other locations across the network. â¢ Use alternative/additional surrogate measures of the roadway, built, and social environment (see Table 5). However, note that multiple measures may be needed, and surrogate measures of pedestrian activity (e.g., populations, numbers of households, employment density, or land uses) may not fully or accurately represent pedestrian volumes. This could potentially lead to incorrect interpretations of analyses. Table 5 summarizes variables that may be useful, along with traffic and pedestrian volume, to account for potential pedestrian crash exposure associated with the built and social environments. Other Considerations The scale or measurement of a variable may affect its association with crashes. In analyses performed for this project, the versions of variables scaled at closer buffer distances to the facility location provided a stronger association to pedestrian crashes than the greater distance versions. For example, the number of buses stopping within 150 feet of an intersection or midblock loca- tion was a stronger predictor than the number of buses stopping within 500 feet. It is reasonable
Step 2: Compile Data 23 to collect such measures at several scales for testing with local data. The exact way these variables are scaled or measured may be less important than the fact that they are included. Count Target Crash Types by Location and Add to Database The final task in compiling data will be to count the target crash types (identified in Step 1) by location and add these counts to the analysis database. These crash frequencies are the dependent variables in the analysis (described further in Step 3) and will also likely be use- ful at least for reference during the treatment site identification process. Besides the crash types selected for systemic focus, it may be desirable to count total pedestrian crashes, other crash subsets, and bicycle and auto-oriented crashes and add them to the database. The crash database developed in Step 1 can be used to determine the frequencies of focus crash types by location. Exposure-Related Variables Measurements Roadway functional class Functional classification is usually available in roadway inventory. This measure is included in this table, as it may help to account for traffic volume if networkwide traffic volume data are not available. Transit activity measures â¢ Numbers of buses stopping within X distance of facility location (intersection/segment) or along a segment â¢ Potentially other measures of transit activity (e.g., boarding/alighting data) Commercial land uses; mixed, residential land use â¢ Number or square footage of commercial properties aggregated in spatial reference to facility location â¢ Similar measures of other land use types (residential, mixed, or institutional) Area population â¢ Total population (average of census blocks) within X distance of facility location Employment density â¢ Numbers of employed persons working within distance of facility Household density â¢ Numbers of households within X distance of facility Area population income â¢ Mean/average income of residents within X distance of facility â¢ Percentage of residents below the poverty line within distance of facility Area population age groups â¢ Proportion of population 65+ years â¢ Proportion of population < 18 years Area population vehicle ownership and/or mode share percentages Mean/average percentages of residents without access to a motor vehicle or who walk to work, within X distance of facility location Other urban density measures May take the form of local planning variables; building volumes; commercial building volumes within distance of facility Alcohol vending establishments Number or density of alcohol establishments within distance of facility location Universities, schools â¢ Density of (within X distance) or distance to nearest institutions from facility location Slope/grade â¢ â¢ â¢ â¢ Other pedestrian/traffic generators â¢ Examples include shopping centers, stadiums, theme parks, recreational facilities, hospitals, or large parking lots Change in slope or grade on segments or intersection approaches may be associated with pedestrian crash risk. If elements capturing slope or change in grade are not present in roadway data, data can be obtained from the National Elevation Data Set from the U.S. Geological Survey (https://pubs.er.usgs.gov/publication/70156331). Table 5. Other variables to account for pedestrian crash exposure at any location type.
24 Systemic Pedestrian Safety Analysis The database should now be complete and will form the basis for risk analysis in Step 3 and network screening and ranking to identify and prioritize treatment sites in Steps 4 and 6. See the next section for additional resources on developing the necessary data types to account for pedestrian exposure to potential crashes. Additional Resources Chapters 2 and 3 of the technical report provide more details on data sources and risk-related variables from prior research that agencies may consider compiling. Below are additional resources on data collection and volume estimation. FHWAâs Synthesis of Methods for Estimating Pedestrian and Bicyclist Exposure to Risk at Areawide Levels and on Specific Transportation Facilities https://safety.fhwa.dot.gov/ped_bike/tools_solve/fhwasa17041/ index.cfm Turner et al., in press FHWAâs Traffic Monitoring Guide https://www.fhwa.dot.gov/policyinformation/tmguide/ NCHRP Report 797: Guidebook on Pedestrian and Bicycle Volume Data Collection (Ryus et al. 2014) http://www.trb.org/Main/Blurbs/171973.aspx Resource Link FHWAâs pending Guide for Scalable Risk Assessment Methods for Pedestrians and Bicyclists