COI Funded Project: Implementation of an Atmospheric Mesoscale Model in the Gulf of Maine


Project Duration: 6/1/00-12/31/01
Key Words: Gulf of Maine, Computer Simulation & Modelling, 3-D Numerical Modeling, Coastal Circulation, Waves & Currents

Proposed Research

Modeling or forecasting coastal ocean dynamics requires accurate synoptic estimates of atmospheric forcing fields. We will implement and evaluate a high resolution regional atmospheric model focused in the Gulf of Maine. The model uses moderate resolution model data as initial conditions and is capable of assimilating near real-time buoy and ship measurements. The resulting implementation will provide a framework for improving coastal ocean models and subsequently enhance research in coastal physical and biological dynamics.

Final Report

Oceanic variability on the continental shelf is closely tied to temporal and spatial variability in atmospheric forcing. To accurately model or forecast the coastal ocean response to this forcing, synoptic estimates of air-sea heat, freshwater and momentum fluxes are required as surface boundary conditions. No observing system yet exists to provide all of these flux measurements, so ocean modelers often rely on the surface fields from Numerical Weather Prediction (NWP) models as realistic surface forcing input. The U.S. National Centers for Environmental Prediction (NCEP) run both global and regional NWP models from which surface fields are available for use in ocean research. The NCEP NWP regional model fields are attractive for use in coastal ocean modeling because they can provide near real-time gridded estimates of the air-sea fluxes at relatively high spatial and temporal resolutions over much of the U.S. continental shelf.

The use of regional NWP products for synoptic meteorological or air-sea flux fields is sure to increase as interest in the coastal zone continues to grow. While the regional NWP model surface fields have great potential for ocean research, use of these fields must be accompanied by an understanding of their errors.

We used the fifth Fifth-Generation NCAR / Penn State Mesoscale Model (MM5) was originally developed at Penn State nearly 30 years ago. MM5 is a community model designed for analysis of limited areas requiring higher resolution. The current version is a state of the art model that includes nonhydrostatic dynamics, data assimilation, nested grids and parallel optimizations. The MM5 model requires the output from a large-scale atmospheric model for initialization and lateral boundary conditions. The model uses continuous four-dimensional data assimilation, which will enable the use of in situ measurements to improve the meteorological forecast. The high-resolution model analysis combined with in situ measurements will provide a mechanism to study weakly forced or rapid events that are typically not captured in operational forecasts.

Our goal was to acquire and implement a research quality numerical weather forecast system to produce better and higher resolution forcing fields in the Mid-Atlantic Bight south of Cape Cod . This allowed us to ascertain the strengths and weaknesses of numerical weather prediction data, particularly over the ocean where in situ boundary conditions are sparse. The resulting implementation provides a framework for improving coastal ocean models and subsequently enhanced research in coastal physical and biological dynamics. The model is currently being run in a 24 hour forecast mode using the two nested domains described below.

Model Implementation
We acquired the MM5 model from the National Center for Atmospheric Research (NCAR), built and ran the Storm of the Century test case on an SGI workstation. This test case was used to exercise the various components of the model and verify the results against simulations run at NCAR.

The model requires a number of pre-processing routines to be executed prior to the model forecast. The first program is the terrain program and is only run once. Terrain sets up the elevation and land use characteristics for the model grids. We used the US Geological Survey’s 30-second (0.925 km) data for our analysis. The second program is the regrid program and is comprised of a suite of programs called pregrid and an interpolation program called regridder . The pregrid routines ingest data from a variety of large-scale models and produce an intermediate format suitable for the regridder program. The regridder program reads this intermediate file and produces gridded data suitable for the MM5 modeling system. The third program is rawins and takes the output of regridder and improves the first guess using observations. The final pre-processing routine, interpf , takes this data and interpolates it in a consistent manner onto the model grid. When these procedures are complete, the model is run.

The entire process was automated on a two processor Linux workstation at WHOI. The latest National Centers for Environmental Prediction (NCEP) 32 km ETA forecast is downloaded every night. This forecast is used as initial and boundary conditions for the nested model we have implemented here. A 24-hour model run requires about 90 minutes of CPU time to complete. The model results and the NCEP input data are then archived.

Several software packages were available for plotting the results of the MM5 model runs. However, since these were oriented toward atmospheric analysis, a program was written to extract surface variables from the model output. We are able to use this output with existing analysis and plotting programs to study the results.

Model Domains
The model was tested on the east coast with two nested grids focusing on the Mid-Atlantic Bight southwest of Martha's Vineyard . The initial nested grid is shown in Figure 1. The red dots show the regional domain at 27 km resolution extending from Chesapeake Bay through the Gulf of Maine . Figure 2 shows a closer view the first nested grid at 9 km covering the Mid-Atlantic Bight. The 3 km fine resolution grid centered near Martha’s Vineyard is shown in Figure 3.

The Mid-Atlantic Bight MM5 model is currently run in a semi-operational mode and produces a 24-hour forecast with output every 3 hours at each of the 3 grid resolutions. This is primarily dependent upon the availability of the workstation resources combined with the network accessibility of the NCEP ETA input data. We are currently capable of supplementing research programs needing synoptic views of small-scale atmospheric features anywhere around the globe. Our work has facilitated analysis of an MM5 model run at NCAR for NSF GLOBEC sponsored research in the Southern Ocean.

Typical results from the model run on August 18, 2003 are shown in Figures 4-6. Figure 4 shows the large-scale grid surface air temperature overlaid with surface wind vectors. The forecast was initiated on August 18, 2003 at 00:00 GMT and the forecast time was August 18, 2003 at 09:00 GMT.

The 9 km grid is shown in Figure 5. The increased resolution can be seen in the delineation of the surface temperature at the coast.

The 3 km fine resolution grid is shown in Figure 6. At this resolution, the land sea temperature differences are apparent. At this resolution, surface fluxes can be compared with near shore in situ observations at the WHOI Martha’s Vineyard Coastal Observatory to improve model physics at the surface boundary layer and provide adequate input for coastal ocean models.

The mid-Atlantic bight region MM5 has been run in support of CBLAST and other ongoing operations in the region.  The model was run nearly continuously from June 2001 through June 2002 due to hardware failures.  Three-hours surface flux fields have been archived and can be made available for retrospective studies.  Acquisition of new hardware enabled the continuation of the MM5 system in April 2003 and it is still operational today.  There are no immediate plans to stop the system which will continue to operate largely unattended as long as hardware and software configurations remain stable.


We have acquired and implemented a semi-operational mesoscale atmospheric model in the Mid-Atlantic Bight. We have advanced our knowledge of the strengths and limitations inherent in mesoscale models and handling the large data sets associated with those models. This knowledge has facilitated ongoing research based in other geographic regions such as the Southern Ocean.  In addition, it has enabled us to develop competitive proposals to use MM5 and other model derived surface fluxes in support of air-sea interaction research within WHOI.