Tsunami generation by submarine landslides: comparison of physical and numerical models

Abstract

An idealised two-dimensional laboratory model of tsunamis generated by submarine landslides is described. The experimental configuration corresponds to the benchmark configuration suggested by other researchers in the international tsunami community. It comprises a semi-elliptical rigid landslide with a height to length ratio of 0.052 sliding down a 15° slope. The initial landslide submergence and specific gravity are varied, the second of which primarily determines the initial landslide acceleration. In these experiments the landslide motion is generally well approximated as consisting of two periods of constant acceleration. The first phase of positive acceleration finishes as the landslide reaches the base of the slope, while the second period of a slower deceleration continues until the landslide comes to rest along the horizontal base of the tank. A novel experimental technique, which utilises laser-induced fluorescence (LIF), is employed to measure the free surface displacement over the entire space and time domains. This enables the wave potential energy field to be computed directly and provides a vivid picture of the wave generation and development process. Particle tracking velocimetry provides detailed information on the landslide motion and also some data on the sub-surface velocity field. Experimental runs require multiple repeats (typically 35–50) of the same setup in order to capture the entire wave field with the desired resolution. Thus high level experimental repeatability is required, and this is demonstrated. A range of parameters relevant to hazard management are presented and discussed. Maximum crest and trough amplitudes of the offshore propagating waves are shown to be approximately proportional to the initial landslide acceleration and somewhat less strongly dependent on the initial landslide submergence. The maximum wave run-up experienced at the shoreline is shown to depend almost linearly on the magnitude of a high deceleration that occurs for a short period when the landslide nears the toe of the slope. The initial submergence and initial acceleration do not directly determine the maximum wave run-up, although for these experiments they impact indirectly on the magnitude of the deceleration. The efficiency of the energy transfer from the landslide potential energy to the wave field potential energy reaches values of up to 6% and is found to be strongly dependent on the initial submergence. However because of the link between the landslide mass and its acceleration, this efficiency is almost completely independent of the initial acceleration. The results from a numerical model based on linear, inviscid and irrotational wave theory, and solved with the boundary element method, are compared with the data from the experimental program. The numerical model accurately produces the generated sequence of wave crests and troughs, but slightly overpredicts their phase speed by between 2 and 4%. For all other parameters the numerical model predictions are within 25% of the experimental values, although this includes both under- and overprediction for the range of independent parameters covered.