This project develops a new approach to climate modeling that brings community knowledge directly into climate artificial intelligence (AI) systems. While today’s AI models excel at large-scale forecasting, they often miss the local details that shape how people actually experience climate change. This research addresses this gap by integrating unstructured, place-based knowledge (such as local records, oral histories, and images) into advanced multimodal climate models.
By combining community data with traditional climate datasets, the project supports a more accurate, transparent, and ethically grounded form of climate innovation. It demonstrates how small, high-quality local data can improve regional and local forecasts, particularly for extreme events and climate risks that matter for planning and adaptation.
For society, this work has practical implications. More locally relevant climate information can improve public decision-making, support fairer allocation of adaptation resources, and help communities see their own knowledge reflected in the tools used to plan for climate change. In doing so, the project reimagines climate AI as an infrastructure that serves both scientific rigor and the public good.
