Evaluating the Potential of ChatGPT to Support Climate Risk and Adaptation Assessment

Abstract

Adaptation to climate change is increasingly urgent, as efforts to curb greenhouse gas emissions falter. Scaling up adaptation finance is essential to address climate risks, but no adaptation inventory covers all sectors and regions globally, especially for vulnerable, information-scarce communities. Large language models (LLMs) like ChatGPT could help bridge these gaps through rapid scoping of climate risks, adaptation options, programme costs and potential maladaptation. This paper uses structured conversations with ChatGPT to explore adaptations to climate hazards in the United Kingdom (for a national perspective), Bangladesh (for an education sector) and Ghana (for vulnerable communities). Queries were run multiple times to test consistency of outputs and contextual awareness. Early results are promising when compared with published information and expert insight. Nonetheless, practical steps can be taken for more effective use of LLMs, and these are captured in a checklist for users. Further research is needed to compare ChatGPT with other LLMs in giving reliable, domain-specific information about climate risks and priority adaptations.