Nonlinear and spatial non-stationary effects of land finance on urban expansion at the county level in China: Insights from explainable spatial machine learning

While previous studies have highlighted the complex role of land finance in urban expansion in China, few have explored its nonlinear and spatial non-stationary effects together. This study uses an explainable spatial machine learning model, integrating Geographically Weighted-Random Forest (GW-RF) with SHapley Additive exPlanation (SHAP), to investigate these effects at the county level. We identified nonlinear relationships between Land Conveyance Fees (LCFs) and urban expansion, with thresholds increasing from 1 billion to 9 billion yuan over time. Initially, land finance had limited effects, but its impact grew rapidly within certain ranges, stabilizing in more developed counties. Sectoral LCFs exhibited similar patterns, with industrial LCFs having a steeper accelerating influence, particularly in later periods. Spatial SHAP maps revealed disparities, with stronger impacts in developed urban areas and lower effects in underdeveloped regions. Over time, land finance's influence expanded in western and central China, especially in industrial-driven areas like the North China Plain. The influence of LCFs decreased with distance from regional centers and smaller county populations. These findings provide valuable insights for transitioning land finance policies and managing urban growth more effectively.