An Efficient Method for Far-field Tsunami Forecasting

Abstract:

We have developed a hybrid method to forecast far-field tsunamis by combining traditional, least-squares inversion for initial sea surface displacement (LSQ) and time reverse imaging (TRI). This method has the same source representation as LSQ, which involves dividing the source region into a grid of “point” sources. For each of these, a tsunami Green’s function (GF) is computed using a basis function for sea surface displacement whose support is concentrated near the grid point. Instead of solving the linear inverse problem for initial sea surface displacement using regularized least-squares, we apply the TRI method to estimate initial sea surface displacement at each source grid point by convolving GFs with time-reversed observed waveforms recorded near the source region. This tsunami-source estimate is then used to forecast tsunami waveforms at greater distance. We apply this method to the 2011 Tohoku, Japan tsunami because of the availability of an extensive set of high-quality tsunami waveform recordings. The results show that the method can predict tsunami waveforms having good agreement with observed waveforms at near-field stations not part of the source estimation, and excellent agreement with far-field waveforms. The spatial distribution of cumulative sea surface displacement agrees well with other models obtained in more sophisticated inversions, but the temporal resolution of this method does not resolve source kinematics. The method has potential for application in tsunami warning systems, as it is computationally efficient and can be applied to estimate the initial source model by applying precomputed Green's functions in order to provide more accurate and realistic tsunami predictions.