Nonlinear relationships and spatial heterogeneity between geographical environment and mental health among middle-aged and older adults in China

The nonlinear dynamics and spatial heterogeneity between the "geographical environment and mental health" represent pivotal elements in ongoing theoretical debates and empirical discrepancies. However, existing analyses are often fragmented and constrained by issues such as the predominance of linear models, multicollinearity, and omitted variable bias. This study bridges these gaps by employing the eXtreme Gradient Boosting (XGBoost) model and SHapley Additive exPlanations (SHAP) to examine the non-linear effects, spatial variability, and interactions of the natural, built, and social environments on the mental health of middle-aged and older adults in China. The findings reveal that nearly all geographical environmental factors exhibit non-linear relationships with mental health, with maximum temperature, NDVI, population density, and per capita park green space area showing the most pronounced effects. Their directional impact and marginal effects follow divergent trends. Geographical environmental factors not only exert independent effects on mental health but also interact in complex ways. Moreover, the relationship between geographical environmental variables and mental health displays significant spatial heterogeneity. K-means clustering analysis further identifies four distinct regions where geographical environmental factors differentially influence mental health. These results offer valuable insights for resolving theoretical disputes, understanding empirical variations, and informing urban planning, landscape architecture, and environmental management strategies.