2010 Chile Earthquake Tsunami modeling with adaptive mesh refinement

Abstract

In numerical modelling of tsunami, direct calculation of tsunami propagation and run-up from source regions to the coastal zones with a single model domain results in low accuracy. Therefore, nested calculation with varying grid intervals from open sea to coastal zone are generally applied. High level accuracy for tsunami run-up in coastal zones requires small grid intervals in the order of 10 meters or even less, resulting in the significant increase of computing times. As a result, numerical models with nesting method are difficult to apply in operational practice. In this study, we illustrate the 2010 Chile earthquake tsunami propagation and run-up modeling using non-linear shallow water equation solver with adaptive mesh refinement (AMR) method. In the tsunami modelling, we investigate the efficiency of AMR in terms of computation time and accuracy. GEBCO 30 arc-sec is used for bathymetry and the model results are validated with NOAA DART buoy and coastal sea level measurements. Numerical experiments using a non-linear shallow water equation solver with AMR method have been carried out in three categories, (a) adaptive mesh (66023.4 m ≤ resolution ≤ 2063.2 m), (b) non-adaptive with low resolution (constant grid interval of 16505.9 m) and (c) non-adaptive in high resolution (2063.2 m). In addition, four experiments to investigate the effect of tide on tsunami propagation are tested. The quadratic bottom friction is used with the friction coefficient, Cd, of 0.003. Therefore, in total, 7 numerical experiments are performed. The level of adaptive meshes varies from 7 to 12 which are equivalent to approximately 66023.4 m and 2063.2 m in distance, respectively. The results of numerical experiments illustrate that the experiments with AMR applied show averagely 50 times faster in computational time than those required in non-adaptive highresolution runs, while keeping the comparable accuracy with the non-adaptive high-resolution runs.