Urban form, terrain, and spatial inequality in electric vehicle charging demand: Evidence from a mountainous megacity

This study examines how urban form, terrain, and accessibility shape spatial variation in fast-charging demand for electric vehicles. We develop an interpretable two-stage framework that combines an attention-based spatiotemporal graph convolutional network for short-term demand prediction with a static interpretation module linking predicted demand to built-environment, infrastructure, and terrain-related attributes. The framework is applied to Chongqing, a mountainous megacity where steep topography and fragmented road networks reshape effective accessibility. The results suggest that charging demand is not explained by population density or charging-station supply alone. Instead, demand patterns are associated with the joint effects of road connectivity, land-use mix, spatial dependence, and terrain-related accessibility constraints. Several densely developed areas show persistently low predicted demand, indicating that topographic barriers and functional mismatches may limit the use of charging infrastructure. By contrast, central districts with dense road networks and diverse urban functions show concentrated demand. The findings highlight the need for accessibility-oriented and terrain-sensitive approaches to charging-infrastructure planning in heterogeneous urban environments.