Development of a Storm Surge Prediction Model Using Typhoon Characteristics and Multiple Linear Regression

Abstract

Storm surges pose a significant threat to coastal regions worldwide, particularly as sea levels continue to rise due to climate change. This study aims to develop a storm surge height prediction model for the southeastern coast of Korea using a multiple linear regression (MLR) approach. Typhoon characteristics, including location and intensity derived from best-track data, were used as independent variables, while observed storm surge heights served as the dependent variable. The model’s predictive performance was assessed using the Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Squared Error (MSE) and the coefficient of determination (R2). To enhance model accuracy and interpretability, a threshold-based model configuration strategy was implemented by categorizing data according to (1) the distance between the typhoon center and the observation point, and (2) the magnitude of the observed storm surge height. The results indicate that restricting typhoon events to within 900–1000 km of the observation site and segmenting surge heights into low and high ranges significantly improves predictive skill, especially for extreme surge events. For example, at Masan station, the model achieved an R2 of 0.82 for high storm surge height (>0.2 m), and Gwangyang station showed an R2 of 0.57 at a 500 km distance threshold, demonstrating substantial skill in predicting extreme surges. However, limitations remain in capturing the variability of lower-magnitude surges, suggesting the need for future research incorporating nonlinear and ensemble methods. This study provides a foundation for improving coastal hazard prediction and contributes to the development of more effective early warning systems and risk management strategies.