Climate AI is over-indexed on risk assessment

AI plays a vital and growing role in climate hazard and exposure assessment. However, a literature review covering more than 165 studies on AI and climate adaptation found a substantial knowledge gap when it comes to both the potential and limitations of using AI for two crucial tasks: evaluating vulnerability and resilience, and prioritization and implementation of adaptation measures. The review, supplmented by an online expert survey, highlighted the subjectivity, qualitative nature, and context-specific aspects of these tasks, which are challenging for current AI methods to address, as a central challenge. Addtional challenges for future research included: simulating complex long-term changes; improved data quality and access to computational resources; and the need for improved interpretability of results.

The over-indexing on risk assessment in current AI-enabled climate adaptation work highlights the need to invest in new research dealing with the more complex tasks of understanding vulnerability and resilience, and planning and implementing adaptation interventions. However, the factors that make these topics difficult for AI-driven innovation are not likely to fade soon, suggesting that stubborn gaps in AI capabilities for adaptation may persist.