Modeling Hurricane Waves and Storm Surge using Integrated Tightly Coupled High Performance Computations

　Joannes J. Westerink
　Department of Civil Engineering and Geological Sciences, University of Notre Dame
　303A Cushing Hall, Notre Dame, Indiana 46556, USA

　Abstract

Coastal Louisiana and Mississippi are characterized by the tremendous complexity in their geography, topography, bathymetry and surface roughness. In addition, this system is characterized by the wide range of geometric scales and the resulting range of highly energetic flow scales. The geography of the region is defined by low lying topography and a network of sounds, estuaries, bays, marshes, lakes, rivers and inlets that permit widespread inundation during hurricanes. The rapid evolution of data collection systems (e.g. very high resolution Lidar and Satellite based photos and land use maps) allow the physical system to be accurately defined, and the rapid evolution of unstructured grid computational models allows the physical characteristics of the system to be numerically resolved. These enhanced capabilities have led to vastly improved models that compute the generation and propagation of wind waves and the movement and levels of waters in the coastal ocean and adjacent floodplain.

In the wake of recent Gulf of Mexico hurricanes (Katrina, Rita, Gustav and Ike), efforts to accurately model storm surge in the Gulf of Mexico have intensified. A modeling system has been developed that simulates hurricane winds, wind-waves, storm surge, tides and river flow in this complex region. This is accomplished by defining a domain and computational resolution appropriate for the relevant processes, specifying realistic boundary conditions, and implementing accurate, robust, and highly parallel unstructured grid algorithms for both the wind waves and the long wave current/storm surge/tide model. A basin to channel scale implementation of the UnSWAN wave model and of the ADCIRC continuous Galerkin hydrodynamic model has been developed to compute the long wave circulation in the region. The associated grid resolves features down to 30 meters and contains 4.2 million nodes and has a solution computed every half second. This code is run on up to 16,384 processors and requires as little as 25 minutes of wall clock time per day of simulation time.

The present modeling system indicates that localized resolution is key and that, in fact, select areas require even higher resolution, on the order of a meter. Furthermore, the models need to be able to automatically adapt to resolve not only their own scales but the energetic scales of the other models. In order to accomplish this we need to evolve the algorithms that drive the models as well. We are currently developing a class of Discontinuous Galerkin (DG) solutions to these problems that will allow robust and accurate solutions to these highly energetic flow problems. DG solutions are ideal for propagation and advection dominated problems, allow for h-p adaptivity, are elementally mass conservative and allow for very high parallel efficiency ideally suited for the new Petaflop computers that are currently in the planning stages.