Séminaire: Simulation of the 2007 Chehalis Lake landslide-generated tsunami with SPH

The 2007 Chehalis Lake landslide and subsequent tsunami are simulated in the frame of a multi-phase SPH formulation. The geometry is built from the British Colombia Hydro (BCH) shores and sliding area topography, and lake bathymetry surveys. Numerical results are validated comparing run-up height all around the lake to BCH’s wave marks surveys.

 

The three million cubic meters volume of rocks, that rushed into the lake, is treated as a continuum assumed to behave as a shear-thinning fluid. To that end, a rheological law similar to the \uD835\uDF07(I) rheology is used. The slide effective viscosity is thus a function of the mean scalar rate of strain, and of the yield stress. The latter is calculated according to the pressure-dependent Drucker-Prager criterion. Therefore, it depends on the internal friction angle and on the effective stress within the material. Under the lithostatic assumption, effective stress is computed solving a Laplace problem within the granular material domain. The water and the granular phases are treated as immiscible continua, discretized in two distinct sets of SPH particles with different mass and behaviour laws. In all other respects, both phases are treated similarly.

 

The simulation domain is restricted to the North part of the lake where the landslide occurred. In this area, the lake is 2.8 km long, 1 km wide and 108 m deep. The overall simulation includes around 8 million SPH particles, corresponding to spatial discretization of 3 meters.

 

The model is implemented within the open-source code GPUSPH that runs entirely on GPUs with CUDA. Thanks to a high level of optimization and of the multi-GPU capability, GPUSPH can be applied to real-world 3D cases involving several million of particles with a reasonable computational time.