Tiny sky dust, suddenly readable — forecast errors cut in half

The air is full of specks you can't see — wildfire smoke, desert dust, sea salt, exhaust — and for years they've been the blur in weather and air-quality forecasts. A research team led by Chongzhao Zhang found a shortcut: feed raw satellite light measurements straight into aerosol simulations, letting a trained AI sort out what's particle and what's reflected sunshine. Tested over China, prediction errors fell by about half. Better forecasts of haze and storms start with seeing the sky more clearly — and the work is in the Journal of Advances in Modeling Earth Systems.