Validation and inter-comparison of models for landslide tsunami generation

Abstract

The Mapping and Modeling Subcommittee of the US National Tsunami Hazard Mitigation Program convened a workshop in January 2017 to evaluate the present state of numerical models for the simulation of tsunamis generated by submarine or subaerial landslides. A range of benchmark tests were provided to participants, with three tests emphasized: (i) a laboratory submarine solid slide in a 2D horizontal tank, (ii) a laboratory submarine granular slide in a 1D flume, and (iii) a field case based on submarine slides which occurred in Port Valdez, AK during the 1964 Alaska earthquake. Nine landslide tsunami models configured with 21 different combinations of physical options were benchmarked, including: (1) hydrostatic models with no frequency dispersion, which include the nonlinear shallow equation models traditionally used for modeling coseismic tsunamis; (2) Boussinesq or one-layer weakly dispersive models; (3) Multi-layer or non-hydrostatic (i.e., dispersive) models; and (4) Full Navier–Stokes models. Model/data comparison indicates that the inclusion of frequency dispersion in model formulations is critical to obtaining physically reasonable results for the test cases considered. Because the importance of dispersive effects is unknown a priori for any given simulated event, the central recommendation from this work is that a model with at minimum a leading-order representation of frequency dispersion effects be used whenever possible for landslide tsunami simulations.