How can AI and ML methods, especially fundamental models, help to improve the prediction of extreme weather events such as flooding and overcome existing deficits in the representation of atmospheric dynamics in current models?
What approaches are being developed to optimize the RAINA model for extreme weather events?
Can the AtmoRep model be extended to enable a high-resolution and statistically robust prediction of extreme weather events with a spatial resolution of 1 km?
What should the application for a precise short-term forecast for parameters such as temperature, wind and precipitation look like?
Extreme weather events, like the Ahr Valley flood in 2021, have severe impacts on people and nature. Reliable forecasting of such events is therefore a key research goal. Artificial intelligence (AI) and machine learning (ML) are becoming increasingly important in this field, but current models still show weaknesses, such as underestimating wind and precipitation peaks. The RAINA project aims to develop a foundational model for weather forecasting that addresses these deficits. Based on the AtmoRep model, the first model for atmospheric dynamics, RAINA aims to achieve a globally unprecedented spatial resolution of 1 km. The project will use various data sources and the technical expertise of the German Weather Service and the Jülich Research Center. Optimization approaches, such as diffusion models, will be integrated with the support of the University of Bonn. The prototype application aims for a precise short-term forecast of temperature, wind, and precipitation, intended to surpass operational models.
Project Partners
