Using machine learning to evaluate the impact of urban development on PM2.5 concentrations and health risks in Chinese cities

The rapid urbanization process has exacerbated PM_2.5 pollution and given rise to a series of health risks. Various studies have explored the link between urban development and PM_2.5 concentrations. However, few have translated these explanatory insights into scenario-based predictions of how future urban development affects PM_2.5 pollution and related health risks. Thus, this study aims to use machine learning (ML) algorithms to explore the effects of urban development on PM_2.5 concentrations and predict future PM_2.5 levels and associated health risks under various scenarios. Herein, data on ten key urban development indicators across four domains—land use, population and urbanization, economic growth, and environmental governance—were collected from 345 Chinese cities (2013−2023). A stacking model that integrates SVR, RF, DT, GBDT, and XGBoost was employed to simulate the relationship between these indicators and PM_2.5 concentrations and predict future PM_2.5 levels, while a meta-regression-Bayesian regularized trimmed (MR-BRT) model was used to assess PM_2.5-related health risks. Additionally, changes in PM_2.5 concentrations and health risks under six urban development scenarios were discussed. Finally, a spatiotemporal differentiated comprehensive urban development framework to suppress PM_2.5 and associated health risks was proposed. The results indicate that compared with single ML models, the stacking model has a better goodness-of-fit when fitting the complex relationship between urban development and PM_2.5 concentrations in urban-level tabular data. Single environmental governance measures have limited effects, while industrial structure transformation emerges as the core driver for short-term PM_2.5 mitigation in China. However, long-term reductions necessitate coordinated spatial and demographic planning. The findings provide empirically grounded strategies for cities to improve air quality and mitigate health risks, particularly applicable to developing economies experiencing rapid urbanization.