Towards simulating landslidegenerated tsunamis using Gerris
Sébastien DelauxNIWA, MetOcean Ltd
New Zealand
Gerris users meeting, Paris, 4-5 July 2011
Outline
“Standard” tsunami modelling using GfsRiver Empirical source module for GfsRiver Simulation of the initial wave with Gerris3D Transport of transport of the initial wave from
Gerris3D to GfsRiver Summary – what next?
“Standard” tsunami modelling
GfsRiver: Nonlinear shallowwater equations Popinet, S. Quadtreeadaptive tsunami modelling,
Ocean Dynamics, 2011. 3 steps:
Initial conditions Propagation Inundation
The case of the September 2009 Samoan tsunami
Initial condition
Analytical rectangular fault model of (Okada, 1984) available in the okada module
GfsInitOkada {} D {
x = 171.8 y = 15.8
strike = 7 dip = 72 rake = 116
depth = 5e3
length = 100e3 width = 50e3 U = 8. } A tsunami source can be modelled as a single or the
sum of several Okada source terms
Propagation - 1
Dynamic and adaptive reconstruction of the bathymetry using the terrain module
Easy to handle kdtree terrain databases built from xyz files
Lon/Lat coordinate system
Wave-height ( -2 m / 2 m)
Mesh refinement (Level 8 / Level 12)
Inundation - 1
Very high resolution needed (1/3 arcsecond ~ 10 m here)
Nested grid + adaptivity Similar domain size
The TOPICS module for landslide -generated tsunamis Implementation in Gerris of the TOPICS module
developed by Dr. P. Watts and included in GEOWAVE
Based on empirical fitting of labscale experiments by Grilli and Watts
Source for submarine landslide, aerial landslide, submarine slump and pyroclastic flows.
Comparison with publishedresults – topics source
(Ioualalan et al., 2006) 1999, Vanuatu tsunami initial wave
Comparison with published results – maximum elevation
(Waythomas et al., 2009), tsunami hazard senario in the Aleutian arc of Alaska
Limitations
Simplistic slide geometry Reasonably large depth needed Froude number > 1 and < 4 A single bathymetry is taken into account
Solid slides
Solid slides are GfsSolidMoving objects Moving solids cannot intersect with other solids but
can intersect with the boundaries of the domain the slope has to be one of the boundary of the →
domain Air and seawater are represented using a Volume of
Fluid approach The motion of the slide is imposed using an
analytical function or a timeseries of position
Solid slides - limitations
The slope has to be one of the boundary of the domain which implies that only simples slope geometry can be considered
Solids have to be fully immersed in on the the phase and cannot intersect the VOF interface yet
The motion of the solid must be known. The behaviour of the ODE module for that type of problem still need to be assessed
Fluid slides
3 phases represented using 2 VOF tracers Bingham fluid
Fully implicit diffusion scheme Water density = 1000 kg/m^3
Slide density = 1500 kg/m^3 Water viscosity = 0.001 Pa.s
Slide viscosity = 75 Pa.s
Abadie et al., 2010 Similar resolution leads to similar results but an
increased resolution does not...
Cremonesi et al., 2011 Tuning of the properties of the fluid would either
give the right amplitude, either the right timing. Increasing the resolution would have little effect.
Liquid slides – success and limitations
Work with more complex slides and bathymetry geometries
Many rheology can be simulated (HerschelBulckley fluids in general...)
High density contrast can lead to convergence problems
Granular rheology would be desirable. Works for avalanches (i.e. not under water)
Transport of the wave from Gerris3D to GfsRiver - 1
Calculate the distance to the VOF interface
VariableDistance {} Dist T { stencil = 0} Store the initial distance to the interface
Init { istart = 0 istep=1 iend=1} {Dist0 = Dist} Output a gridded data file containing the interface
elevation
GfsOutputInterfaceGrid { start = end} interface.cgd T { theta = 26.1 x = 170 y = 40 alpha = 45
sx = 1 sy = 1 sz = 1 } Dist0
Transport of the wave from Gerris3D to GfsRiver - 2
Now 3 cgd file containing water elevation, and the horizontal components of the velocity are available
In GfsRiver, the U, V and the elevation are read
Init {} { U = u.cgd V = v.cgd D = d.cgd }
Summary – what next ? Tsunami propagation and inundation work well
using GfsRiver Available empirical source model for submarine
landslide Simple wavemakers can be done for solid
landslides Liquid slide simulation need to be improved or
more suitable test cases need to be found We rely strongly on the GfsRiver solver, but
ultimately a full NavierStokes simulation is what we want