Spatio-temporal characteristics of taxis in Bangkok, Thailand, across multiple pandemic waves

• Tucker decomposition is utilized to unravel trip patterns of taxis in Bangkok. • COVID-19 has a significant impact on both taxi’s travel time and distance. • Trip patterns of taxis differ greatly across different pandemic waves. • Taxis have played a vital role in accessing essential services during the pandemic. This research investigates the impact of COVID-19 on taxi trends in Bangkok, Thailand, using data from 7 million conventional taxi trips over a four-year period—encompassing pre-pandemic, pandemic, and post-pandemic periods—by a non-negative Tucker decomposition approach. According to the decomposition results, the fourth-order taxi tensor of size [1527 × 1527 × 24 × 48] could be divided into a core tensor and four factor matrices, with the optimal ranks of 30, 30, 3, and 3 for H3 index of origins, H3 index of destinations, operating hours (time of day), and operating months, respectively. In terms of temporal dimensions, trip patterns of taxis dynamically change as COVID-19 situation develops, but there is no distinction between taxi trips on weekdays and weekends. Regarding spatial dimensions, the majority of taxi trips are short-distance, as most of them originate and destine within the same areas. Tucker decomposition also highlights a marked increase in trips between transportation hubs and healthcare facilities during morning hours, underscoring the role of taxis as a vital transportation mode for accessing essential services during the pandemic. We expect that, with more data availability, a deeper understanding of urban mobility patterns would be better comprehended; and, this piece of information would, in turn, allow policy makers to make informed decisions concerning the development of more sustainable urban transportation considering both the roles of transportation modes and the changes in ridership’s behavior concurrently.