When a tsunami strikes the results can be catastrophic.  Tsunami or 'harbour waves' are the result of earthquake-type excitation of the sea bed generating a fast moving disturbance of the surrounding fluid.  After travelling across the ocean, tsunami waves increase dramatically in height in progressively shallower waters generating a massive breaking wave that brings destruction to all in its path at coastlines.  While the mechanisms that generate tsunami flows are understood generally, the ability to predict the flows they produce at coastlines still represents a formidible challenge.  Theoretical apporaches are extremely difficult to apply, hence fully 3-D numerical models are essential.

The animations on this page were presented at the 3^rd International Workshop on Longwave Propagation (2004).  The aim of the workshop was to compare different numerical schemes for the same benchmark problems.  The animations presented here are for benchmarks problems 3 & 4. The 2-D cases were simulated with the solver described on my SPH-SPS page, while the 3-D case was simulated with an old particle-type boundary condition.  While the results look promising (all simulations were run on a single processor machine), there is clearly still much work to be done!

Presentation:  JHU_SPH_updated.ppt

#3 - Tsunami generation and runup due to a 2-dimensional landslide
This test involved comparison with a linear analytical solution (Analytical solutions for forced long waves on a sloping beach by Liu, Lynett and Synolakis, Journal of Fluid Mechanics, 478, 101-109, 2003).  The comparison with other nonlinear wave propagation models is very good except  at the very tip of the moving shoreline - avi file (3.0MB):


#4 - Tsunami generation and runup due to a 3-dimensional landslide
This aim of this test is to compare numerical results to the experimental data of  "Waves and run-up generated by a three-dimensional sliding mass", by Synolakis and Raichlen, in Submarine Mass Movements and Their Consequences, J. Loquat and J. Mienert Editors, Kluwer, 2003.  The block's motion was prescibed by the time series given.  Below are both 2-D and 3-D simulations. 

The massive drawdown in the coarse 2-D cases really demonstrate how 2-D simulations are indequate for such three-dimensional phenomena.  A convergence study (not shown here) does not alleviate the lack of three dimensionality.

2-D Run 30 - Initially partially submerged block (avi: 7.8MB)

2-D Run 32 - Initially fully submerged block (avi: 8.4MB)

3-D Run 30 - Initially partially submerged block (avi: 2.7MB)
This is clearly suffering from lack of resolution and problems with the particle-type boundary.