Global agricultural adaptation case database and trend analysis based on large language models

The Paris Agreement mandates that countries report on their adaptation efforts to evaluate the adequacy and effectiveness of these measures. Agriculture, a critical sector in climate change adaptation, benefits significantly from global case studies that provide evidence, share experiences, and disseminate knowledge. However, the rapid expansion of these case studies presents challenges in extracting and analyzing relevant information effectively. To address this, this study developed a question‒answering information extraction framework that combines geographic analysis with ChatGPT. Guided by the Systematic Evidence Synthesis (ROSES) review protocol, we established a comprehensive global database of agricultural adaptation cases from 2000 to 2024. This database includes key information such as case distribution, climate stressors, adaptation measures, cost-effectiveness, and constraints, aimed at identifying major trends in agricultural adaptation. Our findings reveal the following: 1) Natural language processing technologies, particularly Large Language Models (LLMs), greatly enhance the efficiency and depth of extracting key information from adaptation cases. This advancement supports the frequent updating of the agricultural adaptation database. 2) There is a notable geographic imbalance in agricultural adaptation efforts globally. Adaptation cases are concentrated in central and southern Africa, southern Asia, Europe, and other regions. While there is diversity in responses to slow onset events, measures for extreme climate events are less common, indicating a gap in the sector's ability to address sudden and uncertain challenges. 3) Agricultural adaptation measures are evolving from individual technologies to more comprehensive approaches. The shift is from methods like crop improvement and irrigation adjustments to integrated measures such as climate-smart agriculture, conservation agriculture, and sustainable practices. These approaches collectively enhance adaptation capacity through technological, managerial, infrastructural, and biodiversity improvements, reflecting a deeper understanding and ongoing refinement of adaptation practices. This study highlights the significant potential of LLMs in improving the efficiency of information extraction and analysis for global adaptation research. It offers new methods for quickly summarizing adaptation cases in agriculture and potentially other fields, providing valuable insights and recommendations for global agricultural policymakers.