Nonlinear driving mechanism of PM2.5 and spatial governance considering time series information entropy

PM2.5 pollution significantly impedes sustainable development in large urban agglomerations. Understanding its driving mechanisms allows the development of targeted management strategies. However, focusing on static factors alone limits the ability to regulate the pollution reduction pace. Achieving stable and efficient PM2.5 reduction in the Guangdong–Hong Kong–Macao Greater Bay Area (GBA) is challenging. This study employed time series information entropy, random forest, partial dependency plot, and multiscale geographically weighted regression (MGWR) to explore the nonlinearity and spatial heterogeneity of static and dynamic influences on PM2.5. We then applied K-means clustering to design region-specific control strategies. Results revealed that temperature (46.39%) and wind speed (44.12%) primarily drove static PM2.5 intensity, while changes in gross domestic product (GDP; 47.50%) influenced its evolution. Enhancing road networks and building density fosters compact cities that steadily reduce PM2.5. Minor fluctuations in population density and nighttime lighting significantly reduced PM2.5, indicating that stable demographic and economic transitions support sustained reductions. At the subregional level, cooling strategies should be prioritized in the central GBA, while regulating population growth, GDP, and road construction is key in peripheral areas. This study clarifies PM2.5,s dynamic driving mechanisms and offers actionable insights for managing both the level and timing of reductions.