Quantification of Inter-Tsunami Model Variability for Hazard Assessment Studies

Tuesday, 16 December 2014
Patricio Andres Catalan1,2, Alejandro Alcantar1 and Pablo Ignacio Cortés1, (1)Federico Santa María Technical University, Departamento de Obras Civiles, Valparaiso, Chile, (2)Centro Nacional para la Gestion Integrada de Desastres Naturales, CIGIDEN, Santiago, Chile
There is a wide range of numerical models capable of modeling tsunamis, most of which have been properly validated and verified against standard benchmark cases and particular field or laboratory cases studies. Consequently, these models are regularly used as essential tools in estimating the tsunami hazard on coastal communities by scientists, or consulting companies, and treating model results in a deterministic way. Most of these models are derived from the same set of equations, typically the Non Linear Shallow Water Equations, to which ad-hoc terms are added to include physical effects such as friction, the Coriolis force, and others. However, not very often these models are used in unison to address the variability in the results. Therefore, in this contribution, we perform a high number of simulations using a set of numerical models and quantify the variability in the results. In order to reduce the influence of input data on the results, a single tsunami scenario is used over a common bathymetry. Next, we perform model comparisons as to asses sensitivity to changes in grid resolution and Manning roughness coefficients. Results are presented either as intra-model comparisons (sensitivity to changes using the same model) and inter-model comparisons (sensitivity to changing models). For the case tested, it was observed that most models reproduced fairly consistently the arrival and periodicity of the tsunami waves. However, variations in amplitude, characterized by the standard-deviation between model runs, could be as large as the mean signal. This level of variability is considered too large for deterministic assessment, reinforcing the idea that uncertainty needs to be included in such studies.