Current State Of Practice And Measurement Error In Manual Community Bicycle And Pedestrian Count Programs.

Abstract

PURPOSE: To understand how and why communities implement manual bicycle and pedestrian count programs, and quantify the measurement error observed with common techniques. METHODS: An online questionnaire (Likert scale, open-ended, check box, multiple choice) was sent to Bicycle and Pedestrian Coordinators for every State Department of Transportation and various communities in the US (purposive “snowball” sample) asking about their manual bicycle and pedestrian count programs, such as purposes of the counts, characteristics of volunteers, use of automated count equipment, and application of adjustment factors. Simultaneously, bicycle and pedestrian count errors were assessed by 25 trained participants at 5 high volume intersections for 2 hours on 2 separate days in a small community (using 4- or 12-movement techniques). Participants were randomly assigned to tally information about direction and mode of travel, helmet usage, sidewalk bicycling, and age and sex of the traveler. Video recorders placed at the intersections served as the criterion measure. Absolute percent errors (APE) were calculated by taking the absolute difference between video and participant counts. Wilcoxon-Mann–Whitney tests were used to detect significant differences. RESULTS: The most common reasons for conducting counts reported by the 71 respondents were to “improve community-wide infrastructure, signage, and paint markings”, “analyze trends year-to-year” and “measure physical activity”. Many communities are not using automatic counters or making seasonal or day-of-week adjustments, and count infrequently and rarely for more than one day per year. The majority of community-volunteers are “first time” counters. APE for counting bicyclists and pedestrians with the 4-movement technique was 11% and 4%, respectively; whereas, the corresponding APEs for the 12-movement technique were 27% and 8%. Although differences between the techniques were not significant (p>0.05), large within-counter errors were observed (0-50%). CONCLUSIONS: The survey results suggest that data collected by many communities may be inappropriately utilized due to the infrequent nature of the counts. The ability to utilize the data are further complicated by the presence of profound within-subject counting errors. Funded by PacTrans, Region 10