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.
72 A P P E N D I X C The analysis approach for NCHRP Project 17-56 required the collection of pedestrian volumes at all study sites. Before beginning the counting effort, the team recognized that there were certain questions which needed to be answered: 1. What is the optimum time of day for conducting a short pedestrian count from which the entire daily pedestrian volume can be estimated? 2. How long should pedestrian counting be conducted at each site? The team was able to use existing pedestrian volume data from Charlotte, NC, to answer these questions. The analysis was based on a group of 204 unsignalized intersections in Charlotte. Charlotte DOT provided historical pedestrian crossing volumes for these intersections, collected during their regular turning movement counts. These counts ranged in length from 12 to 17 hours, but were typically 12-hour counts between the hours of 7:00 a.m. and 7:00 p.m. The data used in this analysis were only from the period of 7:00 a.m. to 7:00 p.m., even though more hours were available at some sites. None of these intersections were in the Central Business District (uptown Charlotte). Many of these intersections were fairly low in pedestrian volume. See Table C-1 for a frequency distribution of the intersection pedestrian volumes. Optimum Time of Day In order to answer the question regarding the optimum time of day for counting, the team analyzed the distribution of pedestrian volume throughout the day (Figure C-1). Even though some intersections had a range of 12- to 17-hour counts, only the data from 7:00 a.m. to 7:00 p.m. were used in Figure C-1. The data are presented in terms of the percentage of the daily volume represented for each 1-hour window. The percentages were calculated as a slid- ing 1-hour count divided by the total pedestrian volume for that day. For example, the point at 4:00 p.m. represents the number of pedestrians counted in the 4:00 p.m. to 5:00 p.m. time period. Also, the data were calculated on a âper-intersectionâ basis rather than by individual approach leg. Based on the trend in Figure C-1, either midday (11:30 a.m. to 1:30 p.m.) or late afternoon (4:00 p.m. to 6:00 p.m.) were the most fruitful times for pedestrian counting in Charlotte. Given that both time periods appeared fairly equivalent, the team made the decision to conduct the NCHRP Project 17-56 data collection counts during the late afternoon. Analysis of Charlotte Pedestrian Volumes to Determine Pedestrian Counting Procedure for NCHRP Project 17-56
Analysis of Charlotte Pedestrian Volumes to Determine Pedestrian Counting Procedure for NCHRP Project 17-56 73 1-Hour vs. 2-Hour Counts The second question pertained to the duration of pedestrian counting that would be needed. With a limited research budget, the team needed to obtain an estimate of pedestrian volume at an intersection using only the minimum necessary counting time. It was estimated that the budget would likely support counts only in the range of 1 to 2 hours per intersection. The team needed to determine whether 1 hour would be sufficient or whether a second hour of counting would provide proportionally more data for the time spent. The Charlotte pedestrian counts were provided for a full day (typically 12 hours) in 15-minute increments. This provided the ability to know the âground truthâ (the actual full 12-hour count) 12-hour volume Number of intersections 0-10 51 11-20 31 21-30 19 31-40 15 41-50 10 51-60 16 61-70 13 71-80 8 81-90 5 91-100 3 101+ 33 Table C-1. Distribution of pedestrian volumes at Charlotte sites. Figure C-1. Hourly pedestrian volume distribution at 204 unsignalized intersections in Charlotte.
74 Development of Crash Modification Factors for Uncontrolled Pedestrian Crossing Treatments for any intersection as well as the counts by specific times of the day (e.g., 4:00 to 5:00). The team used this information to simulate how much of the total day count was captured in a 1-hour block and a 2-hour block. Given that the time of day analysis resulted in a decision to collect pedestrian counts during the late afternoon, this counting duration analysis focused on the late afternoon time period (4:00 to 6:00 p.m.). If the 2-hour count (i.e., 4:00 to 6:00 p.m.) captured only twice as much of the daily percent- age as the 1-hour block (i.e., 5:00 to 6:00 p.m.), then there would be no proportional gain, and the 1-hour block would have sufficed. If the 2-hour count captured significantly more than double what was observed in the 1-hour block, this would represent a disproportional gain in data gathered and would provide a reason to conduct 2-hour counts at all study intersections. Table C-2 shows an example of the results of this analysis approach for six of the intersections. Even in this small example sample, it is obvious that results varied greatly by site. For example, Site 1 shows a significant gain from 2% to 10% when a second hour of counting is included and Site 2 shows that no pedestrians were observed during the first hour, but five were observed during the second hour. However, sometimes a second hour of counting was less than double the percentage seen in the first hour, such as with Site 4, representing an instance when the time invested in a second hour would not yield proportionally more data. Also, sometimes a second hour of counting did not yield any additional information, such as for Sites 5 and 6. The analysis began with data from all 204 intersections for which pedestrian volumes were available. However, the final analysis did not include intersections with a total daily count of zero pedestrians or intersections with an abnormally high number of pedestrians (elimination of outliers). The results showed that on average, a 1-hour count (5:00 to 6:00 p.m.) captured 8% of the total 12-hour pedestrian volume whereas a 2-hour count (4:00 to 6:00 p.m.) cap- tured 18% of the total volume. This result showed that the 2-hour count did yield slightly more than double the percentage compared to a 1-hour count. However, the team did not deem this a large enough gain to justify the substantial increase in costs to collect a second hour of counting at all study sites. A follow-up question to this result was, âAre there situations in which a second hour of counting would be more necessary?â The assumption was that intersections with low pedes- trian volumes would benefit more from a second hour of counting. This assumption was based on Figure C-2 and Figure C-3, where it can be seen that the percentage of the daily count cap- tured during a 1- or 2-hour count increased as the overall daily volume increased. That is to say, sites with higher overall pedestrian volumes provided a better estimate of the daily volume from a short count. However, the analysis of sites with low daily pedestrian volumes showed similar results to the results from all sites. As shown in Table C-3, the percentage increases when adding a second hour Table C-2. Example analysis results for six sites. Intersection ID 12-Hr Total Pedestrians (âGround Truthâ) 1-Hr Total (5:00-6:00) 2-Hr Total (4:00-6:00) Percent Captured by 1-Hr Count Percent Captured by 2-Hr Count 1 135 3 14 2% 10% 2 88 0 5 0% 6% 3 157 16 38 10% 24% 4 53 13 22 25% 42% 5 20 0 0 0% 0% 6 24 3 3 13% 13%
Analysis of Charlotte Pedestrian Volumes to Determine Pedestrian Counting Procedure for NCHRP Project 17-56 75 Percentage of Daily Count Observed from 1-Hr Count Percentage of Daily Count Observed from 2-Hr Count All sites 8% 18% Low volume sites (1-20 peds/day) 6% 13% Med volume sites (21-50 peds/day) 9% 21% Table C-3. Analysis of low volume sites. Figure C-2. Percent of daily pedestrian volume captured in a 1-hr count from 5:00 to 6:00 p.m. Figure C-3. Percent of daily pedestrian volume captured in a 2-hr count from 4:00 to 6:00 p.m.
76 Development of Crash Modification Factors for Uncontrolled Pedestrian Crossing Treatments of counting for low volume and medium volume sites were still only slightly more than double the 1-hour yield (e.g., 6% vs. 13%, 9% vs. 21%). Again, this was deemed to be an insufficient gain of data for the added cost of collecting a second hour. Given the results shown in Table C-3, the protocol used for the NCHRP Project 17-56 data collection was to count 1 hour of pedestrians at each treatment and comparison site. Although all data were collected in the late afternoon ârush hourâ period, the actual time of collection varied throughout the study depending on factors specific to the location and time of year. For example, pedestrian counting at the Arizona sites was done at an earlier time period (4:00 to 5:00) since those counts were taken in the fall and the daylight hours were shorter.