NRL and FNMOC are preparing for significant changes to their NWP models and their computer architecture on which to run them in the next 1-5 years. For global modeling, NRL is now engaged in projects that will enable much higher horizontal resolutions and greater vertical extent in NOGAPS. These capabilities can be enabled through the use of semi-Lagrangian advection which will allow for much longer time steps to be used, creating a much more efficient model. Additionally, a new dynamics formulation, using icosahedral grids is also being evaluated, and the stand-alone global aerosol model, NAAPS, will be integrated into NOGAPS. A 4-dimesional variational analysis, referred to as the NAVDAS-Accelerated Representer (NAVDAS-AR), is currently under development, and scheduled for transition to FNMOC in 2008. NAVDAS-AR will enable the Navy to rapidly assimilate the growing number of observations efficiently, while including both observation and model error in the assimilation. A new Ensemble Transform method will be used to construct initial perturbations for the NOGAPS ensembles, and model uncertainty will be included in a larger set of global members through the use of stochastic physics. For the mesoscale, ocean components (circulation and wave) will be added to COAMPS. NRL is now engaged in research to fully couple COAMPS to the NRL Coastal Ocean Model (NCOM) using the Earth System Modeling Framework (ESMF). Follow-on work will enable the WaveWatch III wave model to be coupled with COAMPS, also through ESMF. Both NCOM and WaveWatch III will be initialized using NCODA, which itself will be upgraded to a 3-dimensional variational analysis, similar to the NAVDAS that is currently used for atmospheric analyses. A new capability for mesoscale ensemble prediction will be added using the atmospheric model in COAMPS. This ensemble will include initial, lateral boundary condition, and physics uncertainties. All regional meteorological and oceanographic models will be implemented under the Centralized Atmospheric Analysis and Prediction System (CAAPS), which provides an end-to-end framework for on-demand and rapid configuration and re-location of regional model runs.
To provide the High Performance Computing (HPC) resources necessary to execute the modeling program outlined above, Fleet Numerical will invest in cost-effective Linux-based supercomputing, including the use of specialized compute nodes optimized for this workload. In addition, Fleet Numerical will utilize remote HPC resources provided by the DoD High Performance Computing Modernization Program (HPCMP) for the less time-critical models and applications in its operational suite.