Comparison of Machine Learning Approaches for Tsunami Forecasting from Sparse Observations

Abstract

We have explored various different machine learning (ML) approaches for forecasting tsunami amplitudes at a set of forecast points, based on hypothetical short-time observations at one or more observation points. As a case study, we chose an observation point near the entrance of the Strait of Juan de Fuca, and two forecast points in the Salish Sea, one in Discovery Bay and the other in Admiralty Inlet, the waterway leading to southern Puget Sound. One ML approach considered is to train a support vector machine (SVM) to predict the maximum amplitude at the forecast points. We also explored the use of two deep convolutional neural networks, a denoising autoencoder (DAE) and a variational autoencoder (VAE) to predict the full time se-

Comparison of Machine Learning Approaches for Tsunami Forecasting from Sparse Observations

“…Due to space constraints, we have only shown the full time series predictions for two sample realizations from the test data, #1127 from the fakequakes data set and the L1 event. The website Liu et al (2021a) includes plots of the time series and predictions for all of the other test problems.…”

“…Availability of data and code. All of the data and code used for this work are available through the website Liu et al (2021a), along with additional plots of time series predictions for other test problems not shown here. The code is also available on Github, and the version used to produce the figures in this paper is permanently archived at Liu et al (2021b).…”

Tsunami Early Warning From Global Navigation Satellite System Data Using Convolutional Neural Networks

“…(2016) and archived at (Melgar, 2016). The tsunami waveforms for each realizations were generated using the GeoClaw Software (Clawpack Development Team, 2020), and available at (Liu et al., 2021b).…”