31 6 Model concept Tsunami forecast models for the Pacific Ocean and the Mediterranean Sea have existed for a long time. Models have also been developed for the North Atlantic Ocean and the Norwegian Sea, and have been used in relevant studies. Of course, work on this subject has intensified after the tsunami event of December 2004 (Geist et al. 2006). Even with respect to the German coast, the question has been raised: what are the capabilities of existing models, especially those used for water level predictions at the BSH, and what would be the best concept for modelling water levels on the German North Sea coast, including hypothetical tsunami? The most commonly used method of tsunami prediction is MOST (Method of Splitting Tsunami, Titov et al. 1997). It uses different models for the three tsunami phases: generation, propagation, and modification close to the coastline. Also hindcasts of tsunami in the Atlantic Ocean (Gjevik et al. 1997, Fine et al. 2005) and Norwegian Sea (Harbitz 1992, Bondevik et al. 2005) are based on this principle. The modelling requirements for a simulation of the tsunami impact on the German North Sea coast are slightly more differentiated: Generation of an initial distribution of the surface elevation (6.1.1) Propagation and modification in the near field (6.1.2) Propagation in the deep ocean (6.1.3) Modification on the continental slope (6.1.4) Propagation and modification on the shelf (6.1.5) Propagation and modification near the coast (water depth smaller than 20-10 m) and in estuaries (6.1.7) Run-up and inundation (6.1.8) 6.1 Concepts of existing models and their relevance to BSH models 6.1.1 Generation of an initial distribution of the surface elevation From the seismic data of an earthquake-generated tsunami, fault parameters are derived and used to construct surface elevations (Okada 1985, Titov 1997, Smallman 2006). The computations have been satisfactory in general, but secondary faulting and other causes may lead to major errors (Shuto 2003, Geist et al. 2006). Basic studies were also carried out with a temporally variable bottom topography (Androsov et al. 2005). Other mechanisms of tsunami generation are understood less clearly. Models of submarine slides have been described by Rubino (1998), Harbitz (1992), Fine et al. (2005), and Ward (2001). More complicated simulations are those involving landslides, during which large volumes of material slip into the water from great heights, and tsunami triggered by meteoritic impacts (Gisler et al. 2003, Glimsdal 2007, Weiß 2006). The models used are physically and numerically highly demanding. However, after a relatively short time, a status is reached which is also accessible to simpler, continuous model equations. It does not appear necessary to have models of the generation phase available at the BSH. An initial distribution of water levels and boundary conditions for more precise, exemplary studies should be computed in collaboration with other institutions using their models. For a realistic warning situation, boundary values for the North-East Atlantic model and/or the North Sea/Baltic Sea model would have to be constructed from observations (e.g. real-time gauge station data). 6.1.2 Propagation and modification in the near field To simulate the modification and propagation of a tsunami in the near field, the non hydrostatic equations for viscous media have to be solved in order to reproduce the