Real-time Attribution of Extreme Precipitation Events on Subseasonal-to-Seasonal Timescales using AI Climate Agents

This project advances how scientists understand and attribute extreme weather events—specifically heavy rainfall—occurring weeks to months in advance. At the effort's core is the development of an AI-powered tool called ClimateAgent. Unlike conventional climate models, which rely primarily on static datasets and deterministic forecasting, ClimateAgent uses advanced multimodal artificial intelligence to analyze and reason across diverse sources of data, including satellite imagery, meteorological reports, and real-time environmental observations.

What sets this initiative apart is its use of causal inference to generate "what-if" and "even-if" scenarios, allowing scientists to simulate alternate versions of past events. For example, it can help estimate how much more intense a storm became due to climate change, or how different the outcome might have been under cooler global temperatures. These insights will be made accessible through a user-friendly, interactive AI interface designed for researchers to query, refine, and challenge the model's conclusions in real time.

This work enables more accurate and transparent attribution of extreme precipitation events on subseasonal-to-seasonal timescales. ClimateAgent may improve early warning systems, guide climate adaptation planning, and inform climate policy debates with rigorous, real-time evidence. This approach also lays the groundwork for applying similar AI agents to other climate extremes, such as droughts or hurricanes, enhancing society’s ability to anticipate and adapt to a changing climate.