Using Strava Data for Active Transportation Planning (Part 2)

Part 2: Placement of Bicycle Counters

Trisalyn Nelson, Founder BikeMaps.org, Professor & Dangermond Chair of Geography, University California Santa Barbara

Meghan Winters, Associate Professor, Faculty of Health Sciences, Simon Fraser University

Welcome to Part 2 in our series on how the BikeMaps.org team is using Strava data for active transportation research and planning! In our first post, we focused on methods our team has developed for street level mapping all bicycling ridership using Strava data. Using statistical models, we integrate Strava, GIS, and official bicycling count data and predict all bicycling. A key to accurate mapping of ridership is the availability of representative official bicycle counts; the more representative the bicycle counts the more accurate the model. But, how can we get representative bike counts?

Several planners and engineers have contacted the BikeMaps.org team over the years asking for help locating counters optimally. To be honest, at first, we didn’t really know how to help. Sometimes our team had local expertise, but we needed a data driven approach to determining where counters might be effective.

Read Part 1 of Using Strava Data for Active Transportation Planning

Enter Strava data… again. Strava provides a continuous sample of ridership on almost every street segment, 24 hours a day and 365 days a year. The temporal patterns in Strava data can be used to characterize the type of bicycling on a street. A lot of bicycling during commute periods likely indicates the street is being used for commuting. High bicycling volumes on the weekend mornings may the street is used for recreational bicycling. A data driven approach to locating bicycle counters is to characterize and group street segments by similar Strava bicycling behavior, and then get some counts from each group of streets. We used machine learning and time series analysis to group street segments into groups based on temporal patterns of Strava ridership (Brum-Bastos et al. 2019). We found six groups that differed based on: level of use (low to high), commute vs non-commute, and type of street (major road vs connecting streets) (Figure 1).

Most often, official bicycle counts focus on high ridership areas, and while important data from low ridership streets are needed for city wide monitoring and modelling all levels of bicycling. To collect data that represents the range of bicycling patterns in a city, we recommend at least some official counters be placed within each bicycling behavior group.

• Note on the above: Each street has a 24 hour ridership pattern (available 365 days a year) (a). Using methods from time series analysis and machine learning we characterizing 24 hour ridership patterns on each street segment and create groups of streets with similar bicyclign behavior. (b) A map of 6 bicycling behavior groups. We recommend distributing official bicycle counts to ensure at least some counts in each groups.

Reference

Brum-Bastos, V., Ferster, C., Nelson, T., and Winters, M. (2019). Where to put bike counters? Stratifying bicycling patterns in the city using crowdsourced data. Transport Findings. November. https://doi.org/10.32866/10828