Energy-efficient AI for Extreme Weather Events Forecasting
Funded by ZUG
Project Leader Prof. Dr. Christoph Meinel and PD Dr. Haojin Yang
Cooperation Partners TUM (Prof. Dr. Xiaoxiang Zhu, Dr. Zhitong Xiong), GFZ (Dr. Zhiguo Deng) and Deutscher Wetterdienst (Dr. Michael Bender)
Runtime 2023 – 2026
The goal of the project is to develop AI-based precipitation forecasting models for Germany with a special focus on extreme weather events. For this purpose, the most efficient and powerful AI algorithms (label-efficient AI models) will be developed, which at the same time aiming at a significant saving in resource consumption when using AI (binary neural networks). In addition, novel datasets (Integrated Water Vapor and Slant Integrated Water Vapor) from GNSS observations will be explored and used to increase the forecasting performance. Through a freely accessible platform to be developed in the project, the project will contribute to a significant improvement of climate forecasting performance and a great saving of computational costs.