Cross-validation between GPS-derived trajectories and activity-travel diaries for transport geography studies

Abstract

Transport and health studies need elaborated contextual information to establish causally relevant associations between built environment factors, mobility characteristics, and health outcomes. However, current approaches face various challenges in reliably obtaining contextual attributes. Therefore, investigating the capability of combining contextual attributes collected from different time-geographic approaches is essential for promoting data quality of contextual attributes but missing in previous studies. To address this issue, this study employed data collected in a cross-sectional survey in Hong Kong to cross-validate activity-travel diaries and GPS-derived trajectories, two primary data collection approaches in transport geography studies. We recruited 782 participants in four representative communities through a stratified sampling scheme. Participants were asked to fill out activity-travel diaries and carry a GPS-equipped smartphone on a working day and a non-working day, respectively. Staying/moving events were identified from the GPS-derived trajectories using a spatiotemporal clustering algorithm while activity-travel records were manually aggregated as activity-travel events, which were then matched with the corresponding staying/moving events. Rigorous preprocessing revealed that about 90 % of the 8500 detected events matched well across the two approaches, and about 76 % of them are one-to-one matched events. Yet, the matching rate shows significant disparities between different socio-demographic groups and geographic and activity contexts. The findings indicate that GPS-derived events can mitigate recall biases in the temporality of activity/travel, while diary-based events provide enriched contextual attributes unavailable from GPS data alone. Our study systematically articulates the (mis)match between different approaches for collecting contextual attributes, and it provides essential insights and protocols for a broad scope of environmental and transport studies that need elaborate and comprehensive contextual information.