Abstract
Tsunami inversion analysis uses well-known quantities such as the propagation speed of tsunami waves in a homogeneous medium (i.e. sea water) and the bathymetry. Because of its simplicity, this method is expected to be an alternative tool to the conventional tsunami forecasting system that uses seismic waves for which propagation is affected by heterogeneity in the crust and upper mantle. This study proposes novel approaches for estimating initial tsunami sources using computational intelligence. A genetic algorithm combined with a pattern search method is introduced to optimize a set of unknown parameters. The method has been tested using the actual tsunami resulting from the 2011 Tohoku event. The proposed method generates random and scattered unit sources inside the region being inverted. This leads to a better approximation of the initial profile of a tsunami, reduces the number of parameters required, and suppresses the negative effects of regularization. Furthermore, a multiple time window analysis is introduced to estimate the spatial and temporal distribution of tsunami-genic slip on the earthquake fault. In general, the proposed method has improved the ability to reveal the underlying physics associated with tsunami generating processes.
