Tsunami forecast using an adaptive inverse algorithm for the Chile-Peru source region

Abstract

The inverse method provides tsunami forecasts based on a regression analysis of near-field water-level data against a database of pre-computed tsunami waveforms corresponding unit slip of predefined sub faults. This thesis describes an improvement to this conventional inverse method by resolving the time sequence of tsunamigenic earthquake events. This adaptive approach provides additional degrees of freedom in the regression analysis to compensate for the errors in the user-specified earthquake time or the lag time in tsunami generation, and most importantly, to account for propagation of earthquake rupture. The algorithm also determines the initial subfaults for the regression analysis based on the arrival times of tsunami waves at water-level stations and those of the pre-computed waveforms. A hindcast analysis for the 1960 Valdivia-Chile, 1995 Antofagasta-Chile, and 2001 Atico-Peru tsunamis demonstrates the capability of the adaptive inverse method. The source parameters and subfault distribution for the Peru-Chile source region are based on seismotectonics and historical earthquakes. The present method produces better results in comparison to the conventional approach and indicates an apparent delay between the generation of the tsunamis and the reported earthquake times for all three events. The results suggest that the slip distribution and earthquake rupture propagation are very important for the near-field results while the bathymetry is the dominant factor which controls the shape of the waveforms in the far-field.