Global flood research patterns, disparities, and risk–attention mismatches revealed by large language models

This study provides a comprehensive global analysis of flood research output, themes, and geographic patterns from 1980 to 2025. Using 64,989 flood-related publications collected from OpenAlex and Web of Science databases, the authors applied rigorous screening procedures (PRISMA 2020 guidelines), semantic filtering with large language models (LLMs), and bibliometric analysis. They integrated disaster and socioeconomic datasets such as Emergency Events Database (EM-DAT) International Disaster Database, World Bank, and SCImago, and assessed the evolution of research, thematic structure, attention-risk mismatches, and disparities across regions, countries, and flood typologies.

Key findings reveal a marked surge in flood research since 2000, with developing countries surpassing developed ones in publication counts by 2023. Despite overall growth, there is a persistent mismatch between research attention and actual flood risk: over half of all countries, notably in Africa and South America, remain severely underrepresented relative to their risk. Research output is determined mainly by socioeconomic factors rather than risk exposure, leading to thematic concentration and limited multi-pathway governance in high-risk, resource-constrained nations. The authors recommend targeted investment, strengthened South-South cooperation, and systemic data sharing—especially in emerging high-risk regions—to close critical knowledge gaps and support more inclusive, risk-informed flood governance.