Collecting population-representative bike-riding GPS data to understand bike-riding activity and patterns using smartphones and Bluetooth beacons

Abstract

Bike-riding GPS data offers detailed insights and individual-level mobility information which are critical for understanding bike-riding travel behaviour, enhancing transportation safety and equity, and developing models to estimate bike route choice and volumes at high spatio-temporal resolution. Yet, large-scale bicycling-specific GPS data collection studies are infrequent, with many existing studies lacking robust spatial and/or temporal coverage, or have been influenced by sampling biases leading to these data lacking representativeness. Additionally, accurately detecting bike-riding trips from continuously collected raw GPS data without human intervention remains a challenge. This study presents a novel GPS data collection approach by leveraging the combination of a smartphone application with a Bluetooth beacon attached to a participant’s bike. Aided by minimal heuristic post-processing, our method limits data collection to trips taken by bike without the need for participant intervention, carefully optimising between survey participation, privacy challenges, participant workload, and robust bike-riding trip detection. Our method is applied to collect 19,782 bike trips from 673 adults spanning eight months and three seasons in Greater Melbourne, Australia. The collected dataset is shown to represent the underlying adult bike-riding population in terms of demographics (sex, occupation and employment type), temporal and spatial patterns. The average trip length (median = 4.8 km), duration (median = 20.9 min), and frequency of bicycling trips (median = 2.7 trips/week) were greater among men, middle-aged and older adults. The ‘Interested but Concerned’ riders (classified using Geller typology) rode more frequently, while the ‘Strong and Fearless’ and ‘Enthused and Confident’ groups rode greater distances and for longer. Participants rode on roads/streets without bike infrastructure for more than half of their trips by distance, while spending 24% and 17% on off-road paths and bike lanes respectively. This population-representative dataset will be key in the context of urban planning and policymaking.