Real-Time Prediction of Alongshore Near-Field Tsunami Runup Distribution From Heterogeneous Earthquake Slip Distribution

Files
TR Number
Date
2023-01-05
Journal Title
Journal ISSN
Volume Title
Publisher
American Geophysical Union
Abstract

Real-time tsunami prediction is necessary for tsunami forecasting. Although tsunami forecasting based on a precomputed tsunami simulation database is fast, it is difficult to respond to earthquakes that are not in the database. As the computation speed increases, various alternatives based on physics-based models have been proposed. However, physics-based models still require several minutes to simulate tsunamis and can have numerical stability issues that potentially make them unreliable for use in forecasting—particularly in the case of near-field tsunamis. This paper presents a data-driven model called the tsunami runup response function for finite faults (TRRF-FF) model that can predict alongshore near-field tsunami runup distribution from heterogeneous earthquake slip distribution in less than a second. Once the TRRF-FF model is trained and calibrated based on a discrete set of tsunami simulations, the TRRF-FF model can predict alongshore tsunami runup distribution from any combination of finite fault parameters. The TRRF-FF model treats the leading-order contribution and the residual part of the alongshore tsunami runup distribution separately. The interaction between finite faults is modeled based on the leading-order alongshore tsunami runup distribution. We validated the TRRF-FF modeling approach with more than 200 synthetic tsunami scenarios in eastern Japan. We further explored the performance of the TRRF-FF model by applying it to the 2011 Tohoku (Japan) tsunami event. The results show that the TRRF-FF model is more flexible, occupies much less storage space than a precomputed tsunami simulation database, and is more rapid and reliable than real-time physics-based numerical simulation.

Description
Keywords
tsunami, earthquake, forecasting, runup, data-driven modeling
Citation