Development of Crash Modification Factors for Uncontrolled Pedestrian Crossing Treatments (2017)

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.

13 C H A P T E R 3 Site Identification The NCHRP Project 17-56 research team used multiple methods for identifying cities and states that had installed pedestrian crossing treatments at unsignalized intersections: • Identifying agencies using the team’s experience and knowledge from managing the Walk Friendly Communities program; • Contacting pedestrian coordinators at state and local agencies to request lists of specific sites or suggestions for jurisdictions to contact; • Soliciting agencies through the Association of Pedestrian and Bicycle Professionals (APBP) listserv; • Communicating with other research teams involved in ongoing projects related to pedestrian safety treatments; • Contacting vendors to obtain indications of cities with specific treatments; • Making personal telephone calls to selected agencies known or suspected to have treatments of interest; and • Conducting exploratory field work to identify specific treatment and comparison sites in mul- tiple cities across the United States, including Portland, Oregon; Tucson, Arizona; and Miami, Florida. The research team received approximately 80 emails in response to advertisements on various email lists. Additionally, the team developed and circulated a flyer to request information on locations where the pedestrian safety treatments of interest were installed. The descriptions were intended to help potential participating cities and state departments of transportation (DOTs) to better understand the types of treatments that were included in the evaluation. With this information, the project team identified agencies that looked promising in terms of number of sites and types of treatments installed. The team undertook follow-up interview calls with each promising agency. Approximately 35 agencies were further contacted by phone or email to gain their willingness to participate in the project, to get more information about the number and types of treatments available in their locality, and to determine the availability of their data. The research team developed a matrix of available sites by treatment type as the basis for deciding which agencies to pursue for detailed information. Cities were grouped for possible inclusion in this study into the following priority groups: • Priority 1—these cities were definitely eligible to be included in the study, based on large numbers of treatment sites or installations of hard-to-find treatments. • Priority 2—these cities could be included in the study if needed for sample size considerations. Data Collection

14 Development of Crash Modification Factors for Uncontrolled Pedestrian Crossing Treatments • Priority 3—these cities were unlikely to be included in the study since they have low numbers of potential treatment sites and would add little benefit (additional treatment installations) for the cost (logistics of incorporating another agency into the data collection process). Based on detailed information obtained from each city regarding available treatments of interest for this study, U.S. geographic distribution of cities, and other factors, the 14 cities selected for data collection for this study were Alexandria and Arlington, Virginia; Cambridge, Massachusetts; Chicago, Illinois; New York, New York; Miami and St. Petersburg, Florida; Tucson, Scottsdale, and Phoenix, Arizona; Portland and Eugene, Oregon; Charlotte, North Carolina; and Milwaukee, Wisconsin. The data collection activities began with finding the treatments of interest in the 14 selected cities and also finding suitable sites for comparison. Site selection focused on pursuing treat- ment (and comparison) sites on higher volume, multi-lane streets and/or two-lane roads at busy downtown crossings where pedestrian crash problems are more likely to be a problem and, therefore, where treatments are more likely to be needed. Compiling Site Data Given the number of potential issues when obtaining and combining data from multiple agencies, the project team identified treatment details, site characteristics, and crash data vari- ables that were needed for developing CMFs and CMFunctions (see Table 3-1). Site Characteristics Most of the study sites selected are intersections on urban, multi-lane streets, since these are sites with a high risk of pedestrian crashes and where countermeasures are typically most needed. Relevant information on geometric and volume characteristics was needed for each site. Google Earth aerial and street-view photographs were used to identify relevant site characteristics, as listed in Table 3-1. Dates were associated with the Google imagery, which allowed for recording of site characteristic changes over time (going back as far as 10 years). Treatment Characteristics In addition to site characteristics, it was crucial to collect data specifically on pedestrian safety treatments. Knowledge of where and when the treatment was installed was important. In many cases, treatments were found in combination with other treatments. For example, advanced YIELD or STOP markings and signs are often used in conjunction with the RRFB. Median islands or a refuge island are often used in combination with both the PHB and the RRFB at sites with a high AADT. These data have been gathered largely from agency installation records, from knowl- edge of agency contacts, and/or from Google Earth aerial and street-view photographs over a multi-year time frame. Crash Data The project team collected data on pedestrian crashes and other crashes for each treatment and comparison study site. All agencies had crash data available electronically, but not all agen- cies had hard copy police crash reports available for the full time frame of crash history. There- fore, the research team obtained electronic crash data files for use in this study. Information was also obtained on the distance from the intersection or midblock crossing where each crash occurred.

Data Collection 15 Category Data Elements Treatment Description of countermeasure Treatment Before condition (baseline) Treatment Midblock vs. intersection Treatment Other treatments installed at same time Treatment Date of installation (to the extent possible) Treatment Other changes during study period Treatment Photo of installation (to the extent possible) Site Characteristics Vehicle volume (AADT) Site Characteristics Pedestrian volume Site Characteristics Speed limit Site Characteristics Number of through lanes Site Characteristics Presence of median Site Characteristics Crosswalk presence Site Characteristics Urban vs. rural Site Characteristics One-way vs. two-way Site Characteristics Number of right turn lanes Site Characteristics Number of left turn lanes Site Characteristics Median width Site Characteristics Median type Site Characteristics Crosswalk type Site Characteristics Pedestrian signs Site Characteristics Presence of transit stop Crash Date Crash Time Crash Location Crash Injury severity Crash Weather Crash Lighting Crash Road/surface condition Crash Type of crash (e.g., rear-end, angle) Crash Pre -crash maneuvers (first or most harmful event) Table 3-1. Data elements collected for NCHRP Project 17-56. Summary of Site Characteristics Overview of Treatment and Comparison Sites A summary is given in Table 3-2 of the number of treatment and comparison sites that were found and coded into the database. Approximately 950 potential treatment and 800 potential comparison sites were originally identified. Using Google Earth aerial and street-view imagery, almost half of the original sites were eliminated for numerous reasons. Treatment sites were eliminated most commonly because they had unusual characteristics at the site, had the wrong type of traffic control signal, were too close to a signalized intersection, had unusual inter section or road geometry, were undergoing construction that precluded a pedestrian count, had a treat- ment that was installed too recently, or lacked good, available, crash or traffic AADT data. Com- parison sites were eliminated most commonly because they had recently installed traffic control signals or non-studied treatment types, were too close to a signalized intersection, had unusual intersection or road geometry, were undergoing construction that precluded a pedestrian count, or lacked good, available, crash or traffic AADT data. A total of 499 treatment sites and 476 comparison sites were included in the final database. The number of treatment sites includes 313 refuge islands, 292 advanced YIELD or STOP mark- ings and signs, 96 PHBs, and 50 RRFBs. Cities with the largest number of treatment sites include St. Petersburg (115 treatment sites), Tucson (85 treatment sites), and Portland (61 sites). Charlotte

16 Development of Crash Modification Factors for Uncontrolled Pedestrian Crossing Treatments had the most comparison sites (112), primarily because of the availability of a large sample of unsignalized crossings with existing 12-hour pedestrian counts. The number of treatment types is illustrated in Figure 3-1. Treatment Type Combinations Treatment sites often had combinations of two or more of the studied treatments. Table 3-3 describes the existing treatment combination possibilities and the number of sites with those combinations, ranging from one treatment per site to three treatments per site. Table 3-3 shows that 300 sites had only one treatment. A total of 146 sites had two treatments, including 57 sites with PHBs plus advanced YIELD or STOP markings and signs. There are 60 sites in the database City RRFB PHB Arlington & Alexandria, VA 2 22 0 0 23 22 Cambridge, MA 10 17 0 0 19 25 Charlotte, NC 2 34 0 2 36 112 Chicago, IL 3 33 3 0 36 37 Eugene, OR 3 28 6 0 29 27 Miami, FL 3 27 5 0 30 36 Milwaukee, WI 0 12 1 0 12 18 New York, NY 0 17 0 0 17 24 Phoenix, AZ 16 11 1 5 18 16 Portland, OR 53 40 2 2 61 33 Scottsdale, AZ 4 17 0 2 18 16 St. Petersburg, FL 113 19 32 3 115 45 Tucson, AZ 83 36 0 82 85 65 TOTAL 292 313 50 96 499 476 Advanced YIELD or STOP Markings and Signs Refuge Island Total Comparison Sites Total Treated Sites Table 3-2. Treatment and comparison site totals by city. Figure 3-1. Treatment type totals. 0 50 100 150 200 250 300 350 Advanced Stop Refuge Island RRFB PHB N um be r o f S it es Treatment Types

Data Collection 17 that have advanced YIELD or STOP markings and signs plus refuge islands. There are 53 sites that have three treatment types: either RRFBs or PHBs in combination with advanced YIELD or STOP markings and signs and refuge islands. As discussed in Chapter 4, CMF development was not possible for all of these countermeasure combinations because of small sample sizes that would not have produced reliable CMFs. Database After valid treatment and comparison sites were selected, the locations (streets and latitude/ longitude) and the site and treatment characteristics for each site were entered into the final database. Pedestrian volume, vehicular traffic volume, and crash data were then compiled and added for each site. As relevant data were collected and received from each city, they were also entered into the existing database, as appropriate. In addition to the pedestrian volume counts that were available from a few cities, pedestrian volume was also collected in the field by Kittelson & Associates, Inc., and Quality Counts (QC) for 915 of the 975 total sites (discussed in the next section). These pedestrian volume counts were accompanied by site photographs. These photographs, in combination with Google Earth current and historical overhead imagery and the most current Google Earth street-view imagery, allowed for a historical analysis of site characteristics over time. Often, sites had changes in crosswalk types or numbers of lanes over the last 10 years. Additionally, sites with multiple treatments some- times had those treatments installed in different years. For example, sites sometimes had refuge islands installed and then, several years later, advanced YIELD or STOP markings and signs plus a new crosswalk marking type installed. These types of changes over time were recorded in the final database using the historical and date-stamped imagery tools in Google Earth. For all sites, important characteristics recorded in the final database included those described in Appendix G of this research report (“NCHRP Project 17-56 Database Creation and Data Entry Methodology Notes”), intersection and midblock locales, road features, signage, area type (urban or suburban), bike lanes, trails, intersection characteristics, crosswalk types, crossing distances, and other pertinent attributes. Single Treatment Advanced YIELD or STOP Markings and Signs (AS) Refuge Island (RI) Rectangular Rapid Flashing Beacon (RRFB) Pedestrian Hybrid Beacon (PHB) 300 97 196 5 2 Two Combined Treatments AS+RI AS + RRFB AS+PHB RI + RRFB RI + PHB RRFB+ PHB 146 60 25 57 4 0 0 Three Combined Treatments AS + RI + RRFB AS + RRFB + PHB AS + RI + PHB RI + RRFB + PHB 53 16 0 37 0 Four Combined Treatments AS + RI + RRFB + PHB 0 0 499 Table 3-3. Treatment site combinations.

18 Development of Crash Modification Factors for Uncontrolled Pedestrian Crossing Treatments Appendix A provides descriptions and examples of treatment installations. Samples of pedes- trian counts and site photographs are given in Appendix B. Treatment and Comparison Site Characteristics Most of the sample sites are on roads with four or more lanes (see Figure 3-2). However, 136 of the 499 treatment sites (27%) and 87 of the 476 comparison sites (18%) are on two-lane roads (see Table 3-4). The research team decided to focus more on sites with four or more lanes and pedestrian crossings on arterial streets and at transit stops, since these are the types of sites where pedestrian crashes are more likely to occur (compared to local, two-lane streets) and where pedestrian safety treatments are more likely needed. Figure 3-3 provides a summary of crosswalk types. The most common types are crosswalks with no markings, continental-style crosswalks, and ladder crosswalks. Standard (parallel-line) Figure 3-2. Sample site distribution by number of lanes. 1 2 3 4 5 6 7 8 Treatments 0 136 71 156 96 29 10 1 Comparisons 0 87 53 150 165 13 8 0 0 20 40 60 80 100 120 140 160 180 To ta l S it es Number of Lanes Table 3-4. Study site distribution by number of lanes, city, and treatment or comparison site. Total Number of Sites Total Number of Sites Alexandria VA 3 0 0% 3 100% 2 0 0% 2 100% Arlington VA 20 12 60% 8 40% 20 15 75% 5 25% Cambridge MA 19 11 58% 8 42% 25 19 76% 6 24% Charloe NC 36 5 14% 31 86% 112 2 2% 110 98% Chicago IL 36 16 44% 20 56% 37 8 22% 29 78% Eugene OR 29 9 31% 20 69% 27 7 26% 20 74% Miami FL 30 3 10% 27 90% 36 2 6% 34 94% Milwaukee WI 12 1 8% 11 92% 18 9 50% 9 50% New York NY 17 4 24% 13 76% 24 9 38% 15 63% Phoenix AZ 18 0 0% 18 100% 16 0 0% 16 100% Portland OR 61 12 20% 49 80% 33 3 9% 30 91% Scosdale AZ 18 1 6% 17 94% 16 3 19% 13 81% St. Petersburg FL 115 58 50% 57 50% 45 7 16% 38 84% Tucson AZ 85 4 5% 81 95% 65 3 5% 62 95% Grand Total 499 136 27% 363 73% 476 87 18% 389 82% Comparison Sites ≤ 2 Lanes ≥ 3 LanesCity State ≤ 2 Lanes ≥ 3 Lanes Treatment Sites

Data Collection 19 crosswalks were found at about 100 crossing sites. Multiple crosswalk types (80 sites) are those that have two or more descriptive codes (combined): for example, a yellow crosswalk with stan- dard markings, a staggered crosswalk with continental markings, or a diagonal crosswalk with ladder markings. The number of treatment sites and comparison sites is shown in Figure 3-4 and Table 3-5 by intersection and midblock configuration for each city. Most of the treatment (75%) and comparison (76%) sites are at intersections. Furthermore, a great majority of treatment and comparison sites are in suburban areas (see Figure 3-5). In terms of transit use, 41% of treatment sites are at transit stops, and approximately half (50%) of the comparison sites are at transit stops (see Figure 3-6). Collection of Pedestrian Count Data The previous section summarized the site characteristics of all of the treatment and compari- son sites currently in the database. This section describes the process used to collect pedestrian count data at the majority (915) of those sites where pedestrian counts were not available from city data files. Figure 3-3. Sample site distribution by crosswalk type. A B C D E F G H I J K L M Mu lp le Treatment 61 51 30 18 142 0 1 215 0 6 32 6 0 63 Comparison 322 44 10 0 50 0 0 48 0 3 15 1 0 17 0 50 100 150 200 250 300 350 To ta l S it es Crosswalk Type Type Idenfier No Markings A White Std B Yellow C Staggered D Ladder E Zebra F Piano G Connental H Dbl. Cont I Diagonal J Brick/Stp/Blk K Unknown L Raised M Combined Mulple Figure 3-4. Sample site distribution by intersection (INT) and midblock (MB) (all sites). INT 75% MB 25% Treatment Sites INT 76% MB 24% Comparison Sites

20 Development of Crash Modification Factors for Uncontrolled Pedestrian Crossing Treatments City State Treatment Sites Comparison Sites Total Number of Sites Intersection Midblock Total Number of Sites Intersection Midblock Alexandria VA 3 3 100% 0 0% 2 1 50% 1 50% Arlington VA 20 17 85% 3 15% 20 16 80% 4 20% Cambridge MA 19 17 89% 2 11% 25 21 84% 4 16% Charlotte NC 36 20 56% 16 44% 112 82 73% 30 27% Chicago IL 36 28 78% 8 22% 37 30 81% 7 19% Eugene OR 29 16 55% 13 45% 27 22 81% 5 19% Miami FL 30 19 63% 11 37% 36 27 75% 9 25% Milwaukee WI 12 10 83% 2 17% 18 16 89% 2 11% New York NY 17 11 65% 6 35% 24 20 83% 4 17% Phoenix AZ 18 12 67% 6 33% 16 11 69% 5 31% Portland OR 61 53 87% 8 13% 33 21 64% 12 36% Scottsdale AZ 18 11 61% 7 39% 16 8 50% 8 50% St. Petersburg FL 115 78 68% 37 32% 45 29 64% 16 36% Tucson AZ 85 81 95% 4 5% 65 60 92% 5 8% Total 499 376 75% 123 25% 476 364 76% 112 24% Table 3-5. Study site distribution data including city, treatment or comparison site, and intersection or midblock site. Urban 19% Suburban 81% Treatment Sites Area Type Urban 11% Suburban 89% Comparison Sites Area Type Figure 3-5. Sample site distribution by area type (all sites). Transit Stop 41% No Transit Stop 59% Treatment Sites Transit Stop 50% No Transit Stop 50% Comparison Sites Figure 3-6. Sample site distribution at locations with and without transit stops (all sites).

Data Collection 21 Approach The data collection followed a consistent and pre-established process for collecting pedestrian counts. Several key steps are the following: 1. Confirmed list of treatment sites and comparison sites. a. Confirmed identified treatment(s) at treatment site. b. Eliminated sites with a traffic signal. 2. Due to limitations in the project budget, pedestrian volumes for 24-hour periods could not be compiled. The project team investigated the option of collecting pedestrian volumes for 1 hour versus 2 hours. Collecting data for 1 hour would allow for data collection at more sites, but the team needed to determine whether a 1-hour data collection period would provide sufficient accuracy. To make this determination, the team a. Filmed/recorded pedestrians crossing at the site for a 2-hour period generally between noon and 6:00 pm for first city (Portland, Oregon). b. Viewed the 2-hour period from 3:00 to 5:00 pm by intersection leg in 15-minute incre- ments (this time period was chosen because this was usually the 2-hour period with the most number of pedestrian crossings). c. Compared the results of the Portland 2-hour counts to the 12-hour counts from Charlotte, North Carolina (from city records). This comparison of 1-hour to 2-hour counts showed that 1-hour counts were essentially as good as 2-hour counts for developing estimates of daily pedestrian volumes. Typically, 1-hour counts were conducted between 4:00 and 5:00 pm. Details of this analysis are given in Appendix C. Appendix D describes the method for adjusting these short counts into day-long pedestrian volumes. 3. Filmed or recorded pedestrian counts for a 1-hour period at most sites in subsequent cities studied for estimating pedestrian AADT. The team provided QC with specific location information (e.g., intersection or midblock street names, and latitude and longitude of each site), a link to a Google map location, and an indication of which treatments were expected to be present at the location. QC used this infor- mation to locate the sites, confirm the expected conditions were present, document differences between what the team expected and the actual conditions, and document other circumstances that seemed potentially inconsistent with the needs of NCHRP Project 17-56. A summary of the treatment and comparison sites and the status of the pedestrian volume data collection, by city, follows. Number of Treatment and Comparison Sites The initial list of sites in need of pedestrian counts was reduced for several reasons. For some sites in Charlotte, North Carolina, and St. Petersburg, Florida, pedestrian counts were already available from city records. Other sites were eliminated for reasons such as ongoing construction during data collection periods and the presence of a newly added traffic signal. As data collection commenced in the field, additional sites were removed because the locations were controlled by a signal and no longer qualified as uncontrolled crossing locations; there was construction at or adjacent to the site; and/or the treatments that were expected to be installed were not present. Table 3-6 summarizes the number of sites by city at which data collection was attempted (sites listed), and successfully completed (sites counted). The pedestrian volume data collection commenced in July 2014 in Portland, Oregon, and concluded in November 2014 with the three Arizona cities (Phoenix, Scottsdale, and Tucson).

22 Development of Crash Modification Factors for Uncontrolled Pedestrian Crossing Treatments Summary In summary, pedestrian volumes were collected at a total of 491 treatment sites and 424 com- parison sites in 14 cities from July through November 2014. The pedestrian counts were used to estimate the pedestrian AADT for each of these sites. Please note that the number of sites presented in Table 3-6 only represents sites where new pedestrian counts needed to be collected. It does not include the approximately 100 sites (from Charlotte and St. Petersburg) where pedes- trian counts were already available from the city files and provided to the project team. Collection of Crash and Average Annual Daily Traffic Data Crash Data To develop CMFs, it is necessary to collect crash data for each treatment and comparison site. It is also important to understand how the crash data for each agency are collected so that the appropriate crashes are included in the query for analysis. Agencies in each of the 14 cities were contacted to obtain the necessary information to understand and subsequently request crash data for each study site. Information gathered included but was not limited to: • Number of years of available crash data; • Any major changes that occurred to the crash databases or recording systems in the last 10 years; • An explanation of how crashes are assigned to intersection and midblock locations; and • Whether hard copies of the crash reports were available. Once an appropriate contact was identified, steps were made to obtain crash data. Table 3-7 gives an indication of which agency provided data for each city involved in the study, the range of years of available data, and whether hard copies of the crash reports were available. It is important to note that the research team was not able to obtain any hard copy reports from any agency, primarily due to privacy laws. This means that crash diagrams and City Sites Listed Sites Counted Treatment Sites Comparison Sites Treatment Sites Comparison Sites Arlington & Alexandria, VA 32 33 30 30 Cambridge, MA 23 28 19 26 Charlotte, NC 36 35 36 35 Chicago, IL 40 40 37 37 Eugene, OR 30 30 30 28 Miami, FL 34 38 29 39 Milwaukee, WI 18 25 12 18 New York, NY 20 25 17 24 Phoenix, AZ 22 20 20 17 Portland, OR 86 39 66 36 Scottsdale, AZ 19 18 19 18 St. Petersburg, FL 90 54 90 47 Tucson, AZ 86 69 86 69 TOTAL 536 454 491 424 Table 3-6. Summary of treatment and comparison sites for pedestrian volume data collection.

Data Collection 23 narratives of each crash could not be used to locate the crashes relative to the treatment(s) of interest. The majority of agencies had crash data available electronically; however, much work was needed to identify crashes at particular sites. For instance, some cities maintained crash data in a spatial format, so queries in a GIS environment were used to associate crashes to sites. In most cases, the research team was given the full crash record for the entire desired city. In a few cases, the agency pulled the full crash record for applicable crashes only for the sites provided by the research team. Average Annual Daily Traffic Data It was also necessary to obtain traffic volume data for the road segments surrounding each leg of the treatment or comparison site. In addition to obtaining crash data from each agency, the research team also sought contacts for AADT data. In most cases, AADT data were obtained from local agencies, unlike the crash data, which came primarily from state DOTs. Efforts were made to obtain multiple years of AADT records where possible. Develop Crash Modification Factors The original objectives of this research were to (1) quantify the relationships between pedestrian safety and crossing treatments at uncontrolled locations and (2) develop CMFs or CMFunctions by crash type and severity for four treatments. Table 3-7. Crash data availability and summary for study sites. City Agency to Provide Crash Data Years of Data Available Total Years Hard Copies Available Alexandria, VA Virginia DOT 2004–2013 10 No Arlington, VA Virginia DOT 2004–2013 10 No Cambridge, MA Cambridge DOT 2004–2013 10 No Charlotte, NC HSIS* 2004–2013 10 No Chicago, IL Chicago DOT 2008–2012 5 No Eugene, OR Oregon DOT 2004–2013 10 No Miami, FL Florida DOT 2006–2012 7 No Milwaukee, WI Wisconsin DOT 2004–2013 10 No New York, NY New York DOT 2008–2012 5 No Phoenix, AZ Arizona DOT 2004–2013 10 No Portland, OR Oregon DOT 2004–2013 10 No St. Petersburg, FL Florida DOT 2006–2012 7 No Scottsdale, AZ Arizona DOT 2004–2013 10 No Tucson, AZ Arizona DOT 2004–2013 10 No *HSIS = Highway Safety Information System