Understanding how street environment affects e-scooter mode choice through travel experience

E-scooter sharing, emerging as a convenient and sustainable mode for short travels, has aroused riding comfort and safety concerns. How to intervene street environment to promote e-scooter sharing is worth investigation. While most studies explore the direct connection between street environment and e-scooter usage on aggregated level, less investigates how street environment affects e-scooter mode choice through travel experience on individual level. This study analyzes how Singapore e-scooter riders' experience of street environment influences their future mode choice by linking e-scooter trajectory data to a stated preference survey and applying the non-linear model of deep neural network with alternative specific utility functions (ASU-DNN). The results show that riders prefer to choose e-scooters when their average experience involve smoother pavement, denser street facilities, separate cycle lanes, fewer street pedestrians, medium naturalness and land use mix, and riding on secondary and tertiary roads. Enlarging the gap between street experience high peaks and low peaks boosts future probability of choosing e-scooters. Excessive riding duration in the past reverses its initial positive association with future e-scooter mode choice. Expanding dedicated cycle lanes, optimizing street design, integrating cycling infrastructure with various urban contexts, and providing rest amenities could be effective strategies to promote e-scooter sharing. • Average riding experience of pavement smoothness, cycle lane separation level, street junction and facility density positively affect future e-scooter probability. • Average riding experience of pedestrian volume negatively impacts future e-scooter probability. • Average riding experience of street naturalness, road hierarchy, and land use mix have an inversely U-shaped relationship with future e-scooter probability. • Enlarging the gap between experience high peaks and low peaks boosts future e-scooter probability. • Excessive total past e-scooter time reverses its initial positive impact on future e-scooter probability.