City-size-dependent interactive effects of building and green-space on PM_2.5 pollution: Evidence from 20...

Urban building and green space morphology jointly influence PM_2.5 concentrations, yet how their relative importance, nonlinear responses, and interactions vary across city sizes remains poorly understood. This study selected 20 representative Chinese cities across five city-size categories to investigate building and green space indicators on PM_2.5 concentrations using an XGBoost-SHAP and two-dimensional partial dependence plots. PM_2.5 concentrations varied significantly across city sizes, with higher levels in super (33.65 ± 10.01 μg/m³), large (33.29 ± 11.23 μg/m³), and small cities (32.79 ± 11.12 μg/m³), and the lowest level in super-large cities (30.12 ± 7.95 μg/m³). Building morphology showed greater explanatory importance than green space morphology, particularly in larger cities, whereas population density became more important toward smaller cities and dominated in small cities. The interaction between building average height and green total edge changed markedly when building average height exceeded approximately 20 m, while distinct interactions between building perimeter and green total edge, building coverage ratio and green total edge, and building perimeter and green patch fractal dimension were observed in super-large, large, and small cities, respectively. The identified thresholds provide quantitative evidence for differentiated building-green space pattern optimization and city-specific planning strategies for PM_2.5 mitigation.