Modeling the 26 December 2004 Indian Ocean tsunami: Case study of impact in Thailand

Abstract

[1] The devastating 26 December 2004 Indian Ocean tsunami stressed the need for assessing tsunami hazard in vulnerable coastal areas. Numerical modeling is but one important tool for understanding past tsunami events and simulating future ones. Here we present a robust simulation of the event, which explains the large runups and destruction observed in coastal Thailand and identifies areas vulnerable to future tsunamis, or safer for reconstruction. To do so, we use an accurate tsunami source, which was iteratively calibrated in earlier work to explain the large-scale tsunami features, and apply it over a computational domain with a finer grid and more accurate coastal bathymetry in Thailand.Computations are performed with a well-validated numerical model based on fully nonlinear and dispersive Boussinesq equations (FUNWAVE) that adequately models the physics of tsunami propagation and runup, including dissipation caused by bottom friction and wave breaking. Simulated runups in Thailand reproduce field observations with a surprising degree of accuracy, as well as their high degree of along-coast variation: a 92% correlation is found between (58) runup observations and computations, while the model explains 85% of the observed variance; overall, the RMS error is approximately 1 m or 17% of the mean observed runup value (skill of the simulation). Because we did not use runup observations to calibrate our coseismic tsunami source, these results are robust, and thus provide a uniquely accurate synoptic prediction of tsunami impact along the Andaman coast of Thailand, including those areas where no observations were made.

The Papua New Guinea tsunami of 17 July 1998: anatomy of a catastrophic event

“…Day et al, 2005;Greene et al, 2005;Ioualalen et al, 2007) suggest that maximum inundation and maximum runup may be caused at some locations by collisions (or interactions) of edge waves travelling in opposing directions along the shoreline Waythomas and Watts, 2003). Despite being present in our previous simulation results, edge wave interactions have received little attention in the context of tsunami hazards, specifically with regard to maximum runup.…”

“…10d). A similar sensitivity analysis was carried out by Ioualalen et al (2007) With regard to the beach slope (in a 2-D vertical plane), NSW wave theory predicts that this is an important control on tsunami wave runup (e.g. Kanoglu and Synolakis, 1998).…”

“…Notwithstanding, they made the surprising claim that a NSW model was necessary to reproduce the overland inundation correctly, a claim that contradicts the results of Tappin et al (2001), Synolakis et al (2002), as well as the results presented here. With its accurate inundation algorithm on dry land, using a slot method, FUNWAVE has successfully reproduced coastal inundation for all tsunami events studied to date (see Watts et al, 2003Watts et al, , 2005bWaythomas and Watts, 2003;Fryer et al, 2004;Day et al, 2005;Ioualalen et al, 2007). The choice of numerical model is thus critical in reproducing data acquired by onshore field surveys and from survivor observations.…”

“…The use of FUNWAVE for tsunami simulations has been well validated by case studies, based on a pyroclastic flow (Waythomas and Watts, 2003), underwater landslides Day et al, 2005;Greene et al, 2005), earthquake generated tsunamis (Day et al, 2005;Grilli et al, 2007;Ioualalen et al, 2006Ioualalen et al, , 2007, and a debris flow (Walder et al, 2005). The inclusion of both nonlinear and dispersive terms in Boussinesq models eliminates the excessive shallow water steepening, and corresponding early offshore wave breaking and dissipation, that take place in NSW wave models (such as TSUNAMI-N2 or MOST) and hence allows for tsunami runup to occur onshore.…”

“…Of course, this result is not generally true for all tsunamis. For instance, recent simulations by Ioualalen et al (2007) of tsunami impact in Thailand during the 26 December 2004 event showed that dispersion affected runup values very little, which was expected given the longer wavelength of earthquake tsunami waves. Indeed, we found nearly identical NSW and Boussinesq simulation results for the earthquake tsunami generated by the main shock during the PNG event.…”