[A-HW22] River Channel Morphology, Water Resource Management, and Advanced Techniques

[AHW22-P15] Using a Bayesian network model for typhoon rainfall prediction

*I-TING KUNG¹, Tsunhua Yang¹ (1.National Yang Ming Chiao Tung University )

Keywords:Typhoon Rainfall Prediction, Bayesian Network Model, Disaster response

Typhoons in Taiwan primarily occur from July to October, often resulting in heavy rainfall that leads to loss of life and property damage. Accurate rainfall predictions are essential for effective disaster prevention and response, enabling predictions both the public and relevant authorities to take timely and appropriate precautionary measures. This study examined eight rainfall stations across various regions of Taiwan, utilizing historical typhoon data from 1987 to 2023 to identify events that meet Taiwan’s minimum wind speed standard of 17.2 m/s, categorizing them as tropical storms likely to impact the area. A Bayesian network model was developed to establish relationships among key factors influencing typhoon-induced rainfall, including wind speed, movement speed, and the distance of the typhoon's center to Taiwan. By integrating historical data for inference, this study predicts typhoon rainfall and evaluates the probabilities of different levels of rainfall intensity, with the aim of enhancing disaster preparedness and response capabilities.