Special Issue (Coastal Engineering)Paper
FUNDAMENTAL STUDY FOR THE DEVELOPMENT OF AN EFFICIENT STORM SURGE RISK ASSESSMENT METHODOLOGY
2023 Volume 79 Issue 17 Article ID: 23-17041
Details
2023 Volume 79 Issue 17 Article ID: 23-17041
We have investigated a probabilistic storm surge risk assessment methodology that simultaneously alters multiple typhoon parameters. This study focused on Tokyo Bay, implementing storm surge computations for a diverse set of typhoons generated via the Monte Carlo method. We confirmed that the hazardous typhoon tracks extracted through this method generally cover instances mentioned in previous research. Within the scope of this study, the exceedance probability distribution of storm surges essentially converged with the execution of 200 numerical simulations.
Subsequently, we conducted a preliminary investigation on the substitution of numerical simulations with machine learning to efficiently evaluate storm surge return periods. Specifically, a machine learning model was trained on the input-output relationships of numerical simulations. It was used to carry out pseudo storm surge simulations for approximately 5, 000 years’ worth of hypothetical typhoons, thereby demonstrating the feasibility of estimating the return periods of maximum class storm surges.
The future challenge lies in improving the variety of the storm surge database and the estimation accuracy for extreme storm surges where the data is relatively insufficient.
