Probabilistic prediction Of Extreme weather events based on ai/physics SYnergy – POESY

The POESY project aims at improving the probabilistic prediction of high-impact weather (HIW) events with an innovative combination of standard physical modelling approaches and computationally-efficient Artificial Intelligence (AI) methods. Probabilistic prediction currently takes the form of small ensembles of perturbed forecasts (O(10)) with a kilometre-scale resolution. In this project, AI and in particular deep generative models will be applied to increase both the probabilistic and spatial resolutions of ensemble forecasts. The ultimate goal is to provide ensembles of several hundreds of weather forecasts at sub-kilometre scales, which are practically unfeasible using standard modelling with the available computational resources. As a consequence, this project could lead to a strong methodological break in the design of ensemble forecasts. It shall also lift scientific locks for the real-time monitoring of HIW and in many domains where HIW predictive information is critical.