Deep Learning Models for Tsunami Forecasting

Tsunami induced by seismic earthquake has generated significant disasters in the big cities around coastal regions. One of the most disastrous tsunami events, the 2004 Indian Ocean tsunami has killed 230.000 human casualties and destructed enormous infrastructures and economy in the coastal areas of several countries in the East and South Asia. The negative impacts of tsunami have initiated the vital need for an accurate and fast tsunami early warning system to anticipate infrastructure destruction and human casualties. This paper focuses on the development of a hybrid tsunami forecasting model by utilizing Delft3D-FLOW numerical models and data-driven models using computational intelligence to simulate and forecast tsunami wave amplitudes. Several deep machine learning models were built properly and optimized using Convolutional Neural Network and Long Short- Term Memory algorithms with single and multi-steps prediction methods. The proposed tsunami models provide reliable tsunami forecasts to be utilized for a real-time tsunami early warning system.