4-year Postdoc position on AI and S2S prediction

The position to be filled is embedded in the project ASPIRE - Advancing Subseasonal Predlctions at Reduced Computational Effort, which is funded by the European Research Council (ERC) through a Starting Grant.

AI-based weather prediction models are revolutionizing the way weather forecasts are made. Compared to numerical weather prediction models, these models allow for high-quality forecasts at substantially reduced computational costs. Current data-driven AI models for weather forecasting focus on short to medium range forecasts of 1 to 10 days. For weather forecasts on the so-called sub-seasonal time scale (2 weeks to 2 months), forecast skill can be obtained based on slowly varying modes in the tropics. It is currently unknown if AI-based weather prediction models are capable of mapping to such sub-seasonal time scales. Therefore, the aim of this position is to investigate the feasibility and adaptation of data-driven AI weather forecasting models for sub-seasonal time scales. Within the project ASPIRE you will implement state-of-the-art AI-based models and train them with a local emphasis on the tropics. You will then work on fine-tuning these models for specific applications, which will be enabled through high-resolution numerical simulations generated in the project. Your task will be further to develop and implement probabilistic approaches for these models, e.g. through ensembling techniques, to account for uncertainties in the forecast.

With this opening, we are looking for an early career researcher working on:

• the implementation of probabilistic AI-based weather forecasting systems in collaboration with the Junior Research Group "Robust and Efficient AI" at KIT-SCC
• the fine-tuning of AI-based weather forecasting models with a special focus on the source region of forecast errors
• the fine-tuning of AI-based weather forecasting models and implementation of locally higher resolution using simulation data generated with ICON, the NWP model of German Weather Service
• outreach activities and public relations for ASPIRE, e.g through the project's website and representation at conferences and workshops
• development of a user-infrastructure for model deployment available to the public
• preparation and publication of scientific papers

We offer an exciting and dynamic work environment in a newly established Young Investigator Group “Meteorological Data Science” at KIT, one of the largest institutions of research and higher education in natural sciences and engineering in Europe. Interactions with the KIT-based research groups Robust and Efficient AI (Dr. Charlotte Debus), AI for probablistic weather forecasting (Dr. Sebastian Lerch), Atmospheric Dynamics (Prof. Peter Knippertz) and Tropical Meteorology (Prof. Andreas Fink) are part of the position. Furthermore, we collaborate closely with German Weather Service and researchers at research institutions in Europe and North America. Networking and training opportunities for early career researchers are offered at KIT.