Unravelling the spatial dependence and nonlinearity of driving forces shaping China’s inter-city transportation mode structure

The rapid expansion of transportation infrastructure in China underscores the importance of understanding the inter-city transportation mode structure. However, current research predominantly focuses on individual travel choices, lacking comprehensive analysis of city-level inter-city travel dynamics and the spatial dependencies and nonlinearities influencing these patterns. This study addresses this gap by utilising comprehensive data of inter-city flow records between 296 cities from Amap and Tencent, calibrated with official data, to reveal the inter-city transportation mode structure in China. We develop a Spatial Durbin eXtreme Gradient Boosting (SDXGBoost) model to analyse the driving factors of this structure. The model offers an innovative method for incorporating spatial lag effects into machine learning. Our findings highlight the significant role of spatial lag effects and nonlinearities, showing that increased rail travel in one city promotes rail usage in neighbouring cities, whereas air travel exhibits limited spill-over effects. Cities with advanced urbanisation, significant economic development, and high administrative status tend to have high proportions of rail and air travel, attracting demands from surrounding areas with backwash effects. Geographical location and terrain also exhibit nonlinear effects. The findings provide valuable insights for tailored transportation infrastructure investments and policy-making, ensuring efficient and equitable transportation development across diverse regions.