Advancing Precipitation Monitoring and Forecasting Using AI

March 06, 2025

Simon Pfreundschuh

Hosted by Russ Schumacher

Download Video

Abstract

Monitoring and predicting precipitation is crucial for disaster prevention, agriculture, energy production, and water resource management. However, the high spatiotemporal variability of rainfall and the complex atmospheric processes governing its formation pose significant challenges to conventional observation and forecasting methods. With their ability to integrate various data sources and capture complex non-linear relationships, AI-driven approaches offer a transformative opportunity to improve precipitation monitoring and prediction, addressing both scientific and societal challenges.



In the first part of this seminar, I will present recent advances in applying machine learning to enhance global records of cloud properties and precipitation. This includes the development of the Chalmers Cloud Ice Climatology, the first continuous, high-resolution climate record of vertically integrated cloud ice concentrations, and the development of the next version of the Goddard Profiling Algorithm, which will improve global precipitation estimates for NASA's Global Precipitation Measurement Mission.

I will then outline a pathway toward the next generation of AI-based global precipitation monitoring and forecasting systems. This effort consists of two key steps: (1) Enhancing global precipitation datasets through integrated multi-sensor approaches that combine sparse observations from dedicated precipitation sensors with the continuous coverage of geostationary satellites, and (2) developing AI-based long-term precipitation forecasting models, work that I am currently undertaking in collaboration with NASA and IBM.

The proposed techniques will advance global precipitation monitoring and forecasting and provide a methodological basis for leveraging observations from current and future satellite sensors for the measurement and prediction of atmospheric variables. These methods will thus provide a basis for building next-generation AI-powered atmospheric monitoring and forecasting systems aiming to provide more accurate and timely insights to support decision-making in resource management and disaster response.