Adaptive Mesh Refinement Applied to Tsunami Modeling: TsunaFLASH

Abstract

The devastating Sumatra-Andaman tsunami of December 2004, is a milestone for the internationalcommunity striving to introduce measures to prevent hazards from (future) tsunamis. One of themeasures is numerical modeling which plays a key role for predictions as well as inundation mappingdevelopments.Numerical modeling has been used as a tool for analyzing and reconstructing tsunamis foralmost 40 years. Nowadays, many tsunami codes are available as open-source or free-ware andwidely used in the tsunami modeling community. Many numerical methods have been applied suchas finite difference, finite element and finite volume. The same applies to gridding methods, such asstructured and unstructured non-adaptive types.This thesis introduces a new triangle-based adaptive mesh finite element model for tsunamipropagation (and inundation) simulations. TsunaFLASH combines a numerical method developed inthe framework of the unstructured triangular element, yet non-adaptive, tsunami model TsunAWIwith adaptive mesh refinement capabilities provided by the library amatos. The methods are wellsuited for accurate resolution of localized features, maintaining computational efficiency in terms ofthe number of computations and the required memory.In this first developments of TsunaFLASH, a number of experiments have been performed fortesting. There are: An experiments on various initial conditions, from an analytical source up to acoupling to the sophisticated rupture generator RuptGen; experiments with diverse error estimatorsfor testing refinement criteria and adaptation algorithms; benchmark experiments using analyticalsolutions and field observations where the analytical solution is derived from the first benchmarkproblem from The International LongWaves reference; and the sea surface elevation in the field dataexperiment is used from the satellite tracks of Jason-1 and Topex for verification of the Sumatra-Andaman mega-tsunami 2004 event, while the water level reading of DART 23401 is used for the verification of the Andaman minor tsunami 2009.Some additional studies have been conducted to assess the physical background of tsunamissimulation and test proper supporting tools for TsunaFLASH. There experiments comprise sourcemodel reconstructions and simulations based upon these in the context of a probable worst casetsunami simulation for Padang; experiments to test the influence of different types of topographieson the inundation behavior; and an investigation of the most representative source model for theAndaman-Sumatra mega-tsunami of 2004. Finally, the ambiguity of the arrival time of the Javaminor tsunami 2009 is experimentally investigated.