Broadening Access to Climate AI Innovation

Abstract

Artificial intelligence (AI) offers significant potential for advancing climate science, particularly given its capacity to process vast datasets and improve the accuracy of weather forecasts and climate projections. As extreme weather events become more frequent and severe, AI systems are increasingly being explored to enhance localized prediction, preparedness, and response. At the same time, challenges pertaining to the use of AI in climate science remain, including methodological challenges, governance, ethical, and justice concerns, and limitations in explainable AI (XAI). To address these challenges, a workshop convened climate scientists, affected stakeholders, policymakers, and advocates to discuss and shape future directions for multidisciplinary climate AI innovation. Five key takeaways emerged: redefinition of weather extremes; AI uncertainty as an opportunity for engaging diverse climate knowledges; explainability as a foundational dimension of climate science AI; reconfiguration of climate expertise; and institutional and infrastructural conditions for responsible innovation. Building on these insights, a new knowledge agenda is proposed, calling for participatory data collection infrastructures, transparent communication of uncertainty in AI, and broader, more expansive definitions of explainability. This report is intended for all stakeholders including climate researchers, policymakers, and affected communities, offering pathways to advance the integration of AI in climate science in ways that are effective, equitable, practical, and socially meaningful.