Marine Ecosystem Dynamics Modeling Laboratory

Integrated Model

Fig. 1: Schematic of the integrated GoM/GB model system

Led by Dr. Changsheng Chen (Umass-Dartmouth) and Dr. Robert Beardsley (WHOI), the FVCOM development team has developed an integrated model system for the Gulf of Maine (GoM)/Georges Bank (GB)/ and New England shelf (NES) region (Fig. 1). The major components of this system include: 1) the modified fifth-generation community mesoscale atmospheric model (MM5); 2) the unstructured grid Finite-Volume Coastal Ocean circulation Model (FVCOM); 3) a 3-D suspended sediment transport model; 4) a generalized lower trophic level food web model [the Flexible Biological Module (FBM)], and 5) a multi-stage zooplankton model (developed by Cabell Davis at WHOI). We are currently in the process of migrating the meteorological model to the Weather Forecasting and Research Model (WRF) and thus at this time run both models simultaneously in forecast and hindcast modes in order to be able to compmare the results. The meteorological models predict the forcing fields needed to drive the coupled FVCOM and FBM models. FVCOM uses both nudging and advanced Kalman Filters to assimilate sea surface temperature (SST), satellite-derived insolation, remote sensing reflectance (RSR) as well as in-situ oceanographic data (both moored and shipboard hydrographic and current data) in model simulations conducted with realistic forcing and boundary conditions and observed ocean response for specific periods of time. The integrated model system can be run with idealized forcing and boundary conditions to investigate specific processes.  Both FVCOM and FBM have been coded to run efficiently on multi-processor computers, making coupling with GOM MM5 feasible. The coupled ocean model system is quite flexible, allowing increased resolution in different regions depending on the application and direct assimilation of in situ ocean data collected using moorings, ships, coastal radar, and satellites. Inclusion of a sediment transport model based on the Community Sediment Transport Model System is also complete. The GUI software includes model run parameter selection features and visualization tools to examine the model results.

To improve efficiency, a “parallel” version of the FVCOM core code has been developed to run on multiprocessor computers. The SPMD (Single Program Multiple Data) approach uses a message-passing model to perform the necessary inter-processor communication and synchronization. The physical domain is decomposed into sub-domains using the public METIS graph partitioning libraries. Each sub-domain is assigned to a processor for integration of the model equations. The flux at the sub-domain (inter-processor) boundaries is calculated through flow data exchange among processors. The exchange subroutines utilize non-blocking send and receive messages from the MPI (Message Passing Interface) 2.0 library. The efficiency of the parallel code can be measured in terms of its speedup and/or scalability on a multiprocessor computer. These metrics are dependent on a good load balance from the domain decomposition and minimization of the communication overhead. The parallelized code scales well on moder multiprocessor platforms.

Posted on January 17, 2014