Investigating the multiscale associations between urban landscape patterns and PM1 pollution in China using a new combined framework

Urban landscape patterns serve as the spatial carriers of emission sources and influencing factors of air contaminants, which inevitably impact submicron particle (PM1) pollution. However, most studies ignore the spatial nonstationary and nonlinear effects of urban landscape patterns on the PM1 concentration. In this study, a novel framework was developed that integrates a multiscale geographically weighted regression (MGWR) model and Shapley Additive exPlanations (SHAP) machine learning method to explore the spatial heterogeneity effect, relative contribution and nonlinear impact of landscape patterns on the PM1 concentration at the national and urban agglomeration scales. The results indicated that urban landscape patterns were closely related to PM1 pollution and exhibited obvious spatial differences. The disorderly expansion of urban built-up areas and an irregular urban morphology could aggravate PM1 pollution, while urban fragmentation and connectivity could impose reducing effects. Dispersed urban landscape patterns could reduce the PM1 concentration in the Yangtze River Delta (YRD), Pearl River Delta (PRD), and Chengdu-Chongqing (CDCQ) regions, whereas compact and continuous landscape patterns positively affected PM1 pollution mitigation in the Beijing-Tianjin-Hebei (BTH) and Guanzhong Plain (GZH) urban agglomerations. The impact of urban landscape patterns on PM1 pollution was greater in urban agglomerations than elsewhere. The ENN_MN landscape index exhibited the highest feature importance and interpretability. The threshold effects between the urban shape indicators and PM1 concentration were more complex than those with the other landscape indices. This critical knowledge provides a scientific basis for further understanding the correlation mechanism between PM1 pollution and the landscape pattern, urban sustainable planning and air pollution control.