AI and the World's Hard Problems · Earth & climate science · arXiv · 2026-10-20
AirMPA: A Meteorology-to-Pollution Adapter for Global Air Quality Forecasting
Accurate real-time forecasts of air pollutants help reduce health risks, but traditional numerical systems are computationally costly and existing deep learning models often mix meteorological and pollutant variables. The authors propose AirMPA, a decoupled autoregressive model that predicts 13 pollutants at 0.4° and 12-hour resolution from meteorological fields, emission priors, and past concentrations. Over a 5-day horizon, it consistently outperforms Aurora and beats CAMS in about 90% of pollutant-lead-time cases.