AeroChem improves long-term air quality forecasting with physics and AI
AeroChem: Closed-loop Physics-Informed State Space Modeling for Long-term Chemically-Reactive Air Quality Forecasting--Extended Version
Machine LearningArtificial IntelligenceComputational Engineering, Finance, and Science
Summary
Forecasting air pollution over long periods is hard because pollutants move far, weather changes, and chemical reactions are complex. The authors propose AeroChem, a new approach that combines physical laws and artificial intelligence in a feedback loop to better predict air quality over time. Their method mixes a physics-based model and a neural network that helps correct and improve forecasts step-by-step. They also created a new dataset from Hanoi to test their approach, which showed more stable and accurate predictions, especially when pollution changes suddenly.
What this means in practice
- •For environmental monitoring teams: Deliver longer and more reliable forecasts of urban air pollution using combined physical and AI models with feedback correction.
- •For urban planning departments: Provide improved forecasts of sudden pollutant changes to aid in city health and development planning.
Authors
Kien Le Trung, Tung Kieu, Dinh Duc Nha Nguyen, Cuong Do, Bin Yang, Phi Long Nguyen
Abstract
Accurate long-term air quality forecasting is challenging due to long-range transport, meteorological variability, and nonlinear chemical reactions underlying secondary pollutant formation. Purely data-driven models often suffer from long-horizon drift and lack physical consistency, while existing physics-informed approaches commonly follow open-loop designs that fuse neural and physical components only at the output level. We propose AeroChem, a closed-loop physics-informed state space framework for chemically reactive air quality forecasting. AeroChem adopts a two-branch co-correction design: a UDE-based process branch provides a structured but imperfect physical prior, while a Mamba-based virtual measurement branch captures long-range temporal context and produces probabilistic virtual observations. Neural Kalman Fusion couples the two branches by correcting the latent physical state at each forecasting step and feeding the posterior state back into the simulator, reducing long-horizon drift. We further introduce HanoiAir, a real-world dataset integrating pollutant concentrations, meteorology, and emission inventories. Experiments on HanoiAir and a public benchmark show that AeroChem improves long-horizon stability and sudden-change forecasting.