Land-Water Boundary Treatment for a Tsunami Model With Dimensional Splitting

摘要

The Method of Splitting Tsunamis (MOST) model adapted by National Oceanic and Atmospheric Administration (NOAA) for tsunami forecasting operations is praised for its computational efficiency, associated with the use of splitting technique. It will be shown, however, that splitting the computations between x and y directions results in specific sensitivity to the treatment of land-water boundary. Slight modification to the reflective boundary condition in MOST caused an appreciable difference in the results. This is demonstrated with simulations of the Tohoku-2011 tsunami from the source earthquake to Monterey Bay, California, and in southeast Alaska, followed by comparison with tide gage records. In the first case, the better representation of later waves (reflected from the coasts) by the modified model in a Pacific-wide simulation resulted in twice as long match between simulated and observed tsunami time histories at Monterey gage. In the second case, the modified model was able to propagate the tsunami wave and approach gage records at locations within narrow channels (Juneau, Ketchikan), to where MOST had difficulty propagating the wave. The modification was extended to include inundation computation. The resulting inundation algorithm (Cliffs) has been tested with the complete set of NOAA-recommended benchmark problems focused on inundation. The solutions are compared to the MOST solutions obtained with the version of the MOST model benchmarked for the National Tsunami Hazard Mitigation Program in 2011. In two tests, Cliffs and MOST results are very close, and in another two tests, the results are somewhat different. Very different regimes of generation/disposal of water by Cliffs and MOST inundation algorithms, which supposedly affected the benchmarking results, have been discussed.