Robust prediction of hourly PM_2.5 from meteorological data using LightGBM

Retrieving historical fine particulate matter (PM_2.5) data is key for evaluating the long-term impacts of PM_2.5 on the environment, human health and climate change. Satellite-based aerosol optical depth has been used to estimate PM_2.5, but estimations have largely been undermined by massive missing values, low sampling frequency and weak predictive capability. Here, using a novel feature engineering approach to incorporate spatial effects from meteorological data, we developed a robust LightGBM model that predicts PM_2.5 at an unprecedented predictive capacity on hourly (R² = 0.75), daily (R² = 0.84), monthly (R² = 0.88) and annual (R² = 0.87) timescales. By taking advantage of spatial features, our model can also construct hourly gridded networks of PM_2.5. This capability would be further enhanced if meteorological observations from regional stations were incorporated. Our results show that this model has great potential in reconstructing historical PM_2.5 datasets and real-time gridded networks at high spatial-temporal resolutions. The resulting datasets can be assimilated into models to produce long-term re-analysis that incorporates interactions between aerosols and physical processes.