Abstract
Tsunamis are distinguished from ordinary waves and currents owing to their characteristic longer wavelengths. Although the occurrence frequency of tsunamis is low, it can contribute to the loss of a large number of human lives as well as property damage. To date, tsunami research has concentrated on developing numerical models to predict tsunami heights and run-up heights with improved accuracy because hydraulic experiments are associated with high costs for laboratory installation and maintenance. Recently, artificial intelligence has been developed and has revealed outstanding performance in science and engineering fields. In this study, we estimated the maximum tsunami heights for virtual tsunamis. Tsunami numerical simulation was performed to obtain tsunami height profiles for historical tsunamis and virtual tsunamis. Subsequently, Bayesian neural networks were employed to predict maximum tsunami heights for virtual tsunamis.
Highlights
Tsunamis triggered by undersea earthquakes around the Pacific Ocean have frequently caused loss of human life and property damage in coastal areas
This study explores the estimation of maximum tsunami heights with Bayesian neural networks (BNNs)
Tsunami height profiles at the epicenter and gage in port were used as fundamental datasets, and normalization for tsunami height and time was performed in advance
Summary
Tsunamis triggered by undersea earthquakes around the Pacific Ocean have frequently caused loss of human life and property damage in coastal areas. In the field of tsunami research, data analysis and hydraulic experiments for tsunami events are challenging because major tsunamis are rare and laboratory installations for hydraulic experiments are exceedingly expensive For these reasons, many researchers have focused on developing numerical models to predict accurately tsunami heights defined as vertical height from the troughs to crests of tsunamis and run-up heights defined as vertical heights from the datum levels to the highest inundated areas for several decades [2,3,4,5,6]. BNNs were employed, and the pairs of tsunami height profiles at the site of the tsunami epicenter and Jumunjin port (37.89 N, 128.83 E) considering two historical tsunami events: tsunami events in 1983 (40.54 N, 139.02 E) and 1993 (42.34 N, 139.25 E), were used as the input variables for training and testing. BNNs were used to predict the maximum tsunami heights for the virtual tsunamis
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
