Hybrid Boussinesq-RANS simulation of a solitary wave breaking on coastal topography, where the RANS is used only in the very shallow water (where the water is the solid color). Note that as the wave runs over the land, shown in the lower right, the front face becomes steep and turbulent - the wall of white water - a situation the Boussinesq must parameterize to capture.
Numerical Modeling
The numerical modeling of the tsunami hydrodynamics will utilize a multiple-scale, hybrid modeling methodology. The PI's Lynett and Liu are experienced in this type of modeling, as they are currently leading a research team on a NSF grant to develop a hybrid coastal simulator for wind waves. The simulation backbone of the hybrid tsunami model will borrow from the established OpenSees package, thereby efficiently using already-developed code and working towards eventual tsunami integration into the current geotechnical and structural capabilities of OpenSees.
An Open Source Tsunami Simulator
Previously, we discussed the capabilities and the needs for improvement to existing simulation models for tsunamis. In order to apply these to the prediction of tsunamis, it will also be necessary to couple the models into an integrated framework for community use. The established structural and geotechnical earthquake modeling framework OpenSees (Open System for Earthquake Engineering Simulation http://opensees.berkeley.edu/ ), an NSF-sponsored development maintained by the Pacific Earthquake Engineering Research Center, will be adapted for this purpose. The aggregate tsunami model to be developed here will be called TSUNAMOS (Tsunami Open Source Community Model) and will be made available through the NEES Consortium Inc. simulation repository.
The software framework is conceptually divided into four primary abstractions: model builder components, a model storage component, analysis (solver) components, and output (recording and visualization) components. Model builder components, such as earthquake and bathymetry specifications, are in large part to be user-defined. The graphical user interface will build largely on OpenSees' freely available TCL user interface. The model storage center acts as the backbone of the software framework, storing the setup in-formation, feeding the solver components input data, receiving the solver output, and then sending the output data to visualization and database components. Included in the solver components are major analysis groups to handle tsunami generation, propagation/runup, and interaction with structures. With time, it will be advantageous to the earthquake engineering community to combine the tsunami simulator and the structural/geotechnical simulator, through common fundamental frameworks. In the short term, the visualization components of OpenSees will be directly employed in the tsunami simulator, allowing for combined structural-geotechnical-hydrodynamic visualizations.
Hybrid Hydrodynamic Modeling for Multi-Scale Simulation
In the hydrodynamic solver component of the proposed community modeling system, TSUNAMOS, there are two main classes of governing equations employed, 2D depth-integrated and fully 3D solvers (model classes are discussed in Section 2). The depth-integrated category includes NSW and Boussinesq models. Depth-integrated models have been shown to provide accurate predictions for transoceanic propagation. They are by definition inviscid, however, so special treatment is required to capture physical dissipation processes in the coastal zone, such as wave breaking and bottom friction. Additionally, wave interaction with complex, spatially variable structures is difficult, due to the mild slope assumptions inherent in depth-integrated models. These important aspects of tsunami simulation are best predicted with fully 3D models, even though they cannot simulate, in a computationally practical sense, the large spatial areas traveled by tsunamis. To overcome this dilemma, a major objective here is to develop a parallel, hybrid computational model for tsunami simulation so that we can correctly simulate from the largest physical scale to the smallest. The hybrid model proposed here will employ three different computational modules, consisting of NSW, Boussinesq, and fully 3D models.