Dynamic assessment and prediction of typhoon disaster risk in Beibu Gulf: From the analytic hierarchy process and projection pursuit clustering perspective

Abstract

The Beibu Gulf, located over the South China Sea, has frequently experienced tropical cyclones and typhoons over recent decades, significantly impacting marine ranching activities. This study introduces the Typhoon Disaster Prediction Algorithm (TDPA), a high-resolution disaster assessment algorithm specifically designed for marine ranching risk evaluation. By integrating meteorological and disaster data from 59 typhoons (1984–2019), the TDPA identifies key influencing factors through diagnostic analysis and quantifies their contributions using the Analytic Hierarchy Process (AHP). A high-resolution disaster prediction grid is then generated via Projection Pursuit Clustering (PPC) to classify potential typhoon damage levels in real-time. The results demonstrate: (1) Maximum wind speed is the most critical meteorological factor, exhibiting a strong positive correlation with all components of marine ranching disaster losses. (2) Marine aquaculture economic losses account for 96.51 % of total disaster impacts, with a significant correlation to typhoon center distance, amplifying indirect damages.; (3) The TDPA achieves high prediction accuracy, with a Pearson correlation of 0.68, Spearman coefficient of 0.57, and an overall accuracy of 83 %, closely aligning with historical disaster data. By leveraging high-resolution meteorological data, TDPA provides fine-grained spatial and temporal disaster estimates, offering an interpretable and adaptable algorithm for emergency management. This algorithm enhances disaster prevention and mitigation strategies, supporting more effective risk-informed decision-making in the Beibu Gulf region.