Tsunami travel time prediction using neural networks

Abstract

The present work reports the development of a nonlinear technique based on artificial neural network (ANN) for prediction of tsunami travel time in the Indian Ocean. The expected times of arrival (ETA) computation involved 250 representative coastal stations encompassing 35 countries. A travel time model is developed using ANN approach. The ANN model uses non-linear regression where a Multi-layer Perceptron (MLP) is used to tackle the non-linearity in the computed ETA. The back-propagation feed forward type network is used for training the system using the resilient back-propagation algorithm. The model demonstrates a high degree of correlation, proving its robustness in development of a real-time tsunami warning system for Indian Ocean.