A Copula-XAI framework for assessing compound typhoon disaster-chain risks and driving mechanisms in coastal mountainous cities: Evidence from Fujian, China

Fujian Province, China. To address the limited quantitative understanding of compound disaster chain risks in highly urbanized mountainous coastal regions, this study develops an integrated framework combining a Copula-based joint probability model with explainable machine learning (XGBoost–SHAP). Using Fujian Province as a case study, we identify high-risk areas, quantify exposure inequality, and analyze key driving factors, nonlinear thresholds, and transition mechanisms across typical typhoon disaster chains. High-risk areas of typhoon disaster chains in Fujian Province show a clear spatial contrast, with single disaster chains being widely distributed and compound disaster chains strongly clustered. Although compound-chain high-risk areas account for only 0.8 % of the provincial area, they concentrate 12.6 % of the population and 14.1 % of economic activity. Correspondingly, population and GDP exposure lift values reach 16.3 and 18.2, respectively, which are substantially higher than those of single disaster chains, indicating pronounced exposure inequality. Overall, typhoon disaster chain risks follow a “natural triggering–social amplification” pathway and exhibit nonlinear threshold behavior. Transitions from single to compound disaster chains are governed by two dominant pathways: socioeconomic-driven and naturally driven transitions. These findings support fine-scale identification and differentiated management of compound disaster chain risks in mountainous coastal cities. • An integrated Copula–XAI framework is developed to assess typhoon disaster-chain risks. • Eight disaster-chain types are identified, spanning single and compound chain patterns. • Compound chains occur 13.6 times more frequently and show a “small-area, high-intensity” exposure pattern. • Disaster-chain risks follow a “natural triggering–social amplification” mechanism with clear nonlinear thresholds. • Transitions from single to compound chains are governed by socioeconomic- and natural-factor-driven pathways.