Systematic Comparison of Tsunami Simulations on the Chilean Coast Based on Different Numerical Approaches

Abstract

Tsunami inundation estimates are of crucial importance to hazard and risk assessments. In the context of tsunami forecast, numerical simulations are becoming more feasible with the growth of computational power. Uncertainties regarding source determination within the first minutes after a tsunami generation might be a major concern in the issuing of an appropriate warning on the coast. However, it is also crucial to investigate differences emerging from the chosen algorithms for the tsunami simulations due to a dependency of the outcomes on the suitable model settings. In this study, we compare the tsunami inundation in three cities in central Chile (Coquimbo, Viña del Mar, and Valparaíso) using three different models (TsunAWI, Tsunami-HySEA, COMCOT) while varying the parameters such as bottom friction. TsunAWI operates on triangular meshes with variable resolution, whereas the other two codes use nested grids for the coastal area. As initial conditions of the experiments, three seismic sources (2010 Mw 8.8 Maule, 2015 Mw 8.3 Coquimbo, and 1730 Mw 9.1 Valparaíso) are considered for the experiments. Inundation areas are determined with high-resolution topo-bathymetric datasets based on specific wetting and drying implementations of the numerical models. We compare each model’s results and sensitivities with respect to parameters such as bottom friction and bathymetry representation in the varying mesh geometries. The outcomes show consistent estimates for the nearshore wave amplitude of the leading wave crest based on identical seismic source models within the codes. However, with respect to inundation, we show high sensitivity to Manning values where a non-linear behaviour is difficult to predict. Differences between the relative decrease in inundation areas and the Manning n-range (0.015–0.060) are high (11–65%), with a strong dependency on the characterization of the local topo-bathymery in the Coquimbo and Valparaíso areas. Since simulations carried out with such models are used to generate hazard estimates and warning products in an early tsunami warning context, it is crucial to investigate differences that emerge from the chosen algorithms for the tsunami simulations.