Global and local impacts of multi-dimensional urban form on PM2.5 concentrations: A case study of the downstream urban agglomeration in the Yangtze River Delta

Abstract

The acceleration of urbanization and the expansion of population scale have led to increasingly prominent PM2.5 pollution. Taking the urban agglomeration in the lower reaches of the Yangtze River within the Yangtze River Delta as the study area, this research innovatively constructs a 1D-2D-3D multi-dimensional urban form indicator system. By integrating spatial autocorrelation analysis, hot spot analysis, the Optimal parameters-based geographical detector (OPGD) model, and the Geographically Weighted eXtreme gradient boosting (GW-XGBoost) model, this study systematically reveals the spatiotemporal evolution characteristics of PM2.5 concentration from 2014 to 2022 and the global and local driving mechanisms of multi-dimensional urban form on PM2.5 concentration. This study aims to fill existing research gaps and provide scientific support for the precise prevention and control of PM2.5 pollution in urban agglomerations. The results indicate that: Multi-dimensional urban form exhibits significant spatial differentiation, with high values of 1D and 2D forms concentrated east of Nanjing, while high values of 3D forms are distributed west of Nanjing. The correlation coefficients of Building density (BD) with Mean building height (MBH) and Floor area ratio (FAR) reach 0.78 and 0.85, respectively, indicating a distinct characteristic of urban vertical expansion. The regional annual average PM2.5 concentration shows a continuous downward trend (decreasing from 60.77 μg/m 3 in 2014 to 29.52 μg/m 3 in 2022), with reductions ranging from 29.55% to 40.61% across individual cities. Spatially, a hot spot area at the 99% confidence level, centered around Nanjing and Ma’anshan, is formed, presenting a stable pattern of “high in the middle and low on both sides.” Regarding the global driving mechanisms, River density (RD), Digital elevation model (DEM), and Road network density (RND) are the core factors influencing PM2.5 concentration (with q -values of 0.3232, 0.2604, and 0.1852, respectively). Pollution risk is highest when DEM is in the 19–34 m elevation zone (concentration reaching 30.06 μg/m 3 ), and all factor interactions exhibit nonlinear enhancement effects. Significant spatial non-stationarity exists in the local driving mechanisms. The regulatory effects of urban form are stronger in core cities (Nanjing, Shanghai), with local R 2 values ranging from 0.30 to 0.35. Specifically, RD exhibits a significant positive driving effect in the central region of Nanjing–Ma’anshan–Wuhu (coefficient 0.80–1.00). In key transportation areas such as northern Shanghai, the coefficient of RND reaches 0.50–0.70. The positive effect of Proportion of transportation land (PTL) is prominent along expressways and around logistics hubs (coefficient 0.30–0.50). In contrast, in peripheral cities (Anqing, Chizhou), the local coefficients of determination (R 2 ) are only 0.16–0.20. The mitigating negative effect of the Proportion of water body (PWB) exhibits a “water-adjacent attenuation” characteristic. This study effectively compensates for the shortcomings of traditional research in the systematic integration and methodological applications of characterizing nonlinear relationships, accounting for spatial heterogeneity, and analyzing multi – dimensional urban form systems. It provides scientific support and specific pathway references for the precise prevention and control of PM2.5 pollution and urban form optimization at the urban agglomeration scale. The findings carry important practical value for air quality improvement and sustainable development in similar high – density urban agglomerations.