Carbon-efficient ride-pooling at intercity railway hubs: Nonlinear topological mechanisms and algorithmic insights

Source abstract

Ride-pooling (RP) can reduce transport emissions, but its carbon reduction potential across spatial contexts remains unclear, especially at intercity transfer hubs. Using citywide regulatory ride-hailing data from Suzhou, China, this study examines how meso-level path topology features are associated with carbon reductions in railway station pooling. Railway station pooled trips show 62.6% higher carbon reductions per trip than the overall pooled-trip sample average, influenced by nonlinear mechanisms and interaction patterns rather than higher pooling intensity. An explainable machine learning framework shows that carbon efficiency is primarily associated with route quality, as reflected in path topology, operational scenarios, and built environment factors. In particular, detour constraints, common drop-off configurations, and their interactions with distance act as critical thresholds. Building on these insights, we develop a dual-layer classification framework that can identify pooling matches with high carbon reduction potential prior to dispatch (accuracy ≥ 96%). These findings suggest that intercity railway stations may serve as favorable contexts for low-carbon RP when demand concentration and route quality align. This insight can help guide matching strategies that are enabled by AI and informed by carbon efficient in shared mobility systems.