Response of PM2.5 concentration to climate variability and climate change prediction in China

Air pollution, particularly fine particulate matter (PM_2.5), poses severe environmental and public health challenges. Understanding the complex interactions between PM_2.5 concentrations and meteorological factors is crucial for effective air quality management and policy development. However, existing predictive models often struggle to capture the nonlinear and spatiotemporal dependencies of PM_2.5 variations, limiting their interpretability and accuracy. To address these gaps, this study developed a machine learning-based model using extensive historical environmental and meteorological data to analyze the nonlinear response of near-surface PM_2.5 concentrations to multiple spatiotemporal drivers. A monthly-scale prediction model was established, integrating CMIP6 climate projections under different emission scenarios, projected emission inventories, and multi-source auxiliary data. The findings reveal that PM_2.5 concentrations are highest in winter and lowest in summer, with significant seasonal and regional variations from 2015 to 2023. The model demonstrated strong predictive performance, particularly over the North China Plain and Northeast urban agglomerations (R = 0.825), though performance was weaker over the Tibetan Plateau. Key meteorological factors influencing PM_2.5 concentrations include specific humidity, 500 hPa wind, and short-wave radiation. Under future climate scenarios, PM_2.5 concentrations in South and East China are projected to decline during 2025–2030, while Northern China may experience seasonal increases under low-emission scenarios. Regional climate changes, such as increased precipitation and wind speeds in certain areas, further influence PM_2.5 concentration patterns. This study provides a novel, data-driven approach to quantifying the impact of meteorological fluctuations on PM_2.5 variations, offering valuable insights for air quality forecasting and policy formulation under different climate scenarios.