Scalable validation of automated bicycle and pedestrian counters: evidence from Washington, D.C.

Abstract

Data from automated pedestrian and bicycle counters is important in planning for active transportation. Counter validation approaches vary, from spot checks at initial installation of individual counters to rigorous comparison with manual count data. Since best practices differ, the data validity from counters is often unknown. This study identifies (1) the minimum number of manual count hours needed to reliably validate automated counters and (2) key factors affecting the accuracy of correction functions. Findings suggest that using 12–15 hours of count data to validate automated counters is the point of diminishing returns to develop reliable correction functions across site volumes. Counters that passed our validation criteria were associated with a 46% decrease in mean RMSE. For bicycle counters, greater accuracy was associated with validation on warmer, non-summer days, during peak hours, and with no precipitation. For pedestrian counters, higher accuracy was associated with validation during summer, spring, or early winter with no precipitation.