DEPSCoR: Integrating Hig Perf Comp Tools with FE Hydrodynamic Model for Operational Molding of Tidal Estuaries

The objective of this work is to enhance the predictive capabilities of the Finite Element QUODDY code, by integrating various high performance computing tools to it. The strategies among others include incorporating a rich blend of Fortran90 into QUODDY and in using Domain Decomposition tools. The advanced features available in Fortran90 will bring the code in line with object oriented languages. The parallel strategy will be based on Domain Decomposition principles, and will use Message Passing Interface (MPI) protocols for all interprocessor communication. The physics in the code will be improved by incorporating a nonhydrostatic pressure and spatially varying mixing length modules. Together with a finer grid resolution, this will pave the way for simulating internal hydraulic jumps. The resulting parallel code will be tested on various platforms, both at the University of Kentucky and NOAA, including cluster of workstations. Performance modeling will address to calculating the points of diminishing return and no return. Our goal is to combine powerful algorithms and Krylov Subspace Solvers with current generation high performance computers to enhance the capabilities of QUODDY, both in terms of accelerating the simulation time and as a window for better simulation of the physics of flow. The resulting code will be used for obtaining real time forecasts of tidal waves in Chesapeake and Delaware Bays. This work is conjunction with NOAA officials who oversee the planning and implementation of the real-time semi-operational numerical models. The ideas we discuss are tested" over few existing codes, including the commercial RMA2 code and the results presented. Our idea is to show that a robust parallel code can be developed from any existing sequential code and we hope that this work will stimulate the formulation and application of parallel computing for hydrodynamic flows. This research will be used to train three graduate students in computational environmental fluid mechanics and high performance computing, leading to two PhD degrees and one or more Masters degrees.