To accomplish this, we have designed an ocean modeling and model/observation intercomparison program, including data assimilation experiments. We have chosen a mature ocean general circulation model, the Princeton Ocean Model or POM that highlights the proper representation of regional processes that, when taken together, define the basin scale pathways and fluxes of the MOC through the subtropical/subpolar/Labrador and Nordic Seas domain. More specifically, our approach is designed to provide a quantitative dynamical description of the upper (warm water) circulation limb of the MOC through the subpolar region as well as improve understanding of the relationships between this upper limb, its directly forced vs. natural (internal) seasonal-to-decadal variability, and sea surface temperature (SST) and salinity (SSS) anomalies. The experiment hierarchy is designed as a coordinated program of model intercomparisons and process, simulation and data assimilation experiments - all towards achieving our fundamental objective. It has also been designed in collaboration with that of Dr. Mike Spall's (WHOI) ACCE program so that together we can establish a consistent modeling framework for placing the two year ACCE field program within the context of a multi-decadal North Atlantic (and full Atlantic basin) climatology.
There are many processes that require accurate representation for determining the warm water pathways of heat and salt into and out of the subpolar gyre, the division between transport by the mean flow and that by eddies, and the water mass transformation processes that support this circulation. Only after a detailed program of study can one determine if a certain model is properly representing all of those processes and so, together with Drs. Spall and Chassignet, we initially set out to compare three mature ocean general circulation models against each other and available observations, with diagnoses directed towards the ACCE objectives. We have selected the POM (Blumberg and Mellor 1987), which has been recently designed to properly represent all of the important subpolar physical processes. We further augment the POM with an atmospheric boundary layer model (to more accurately represent surface heat fluxes) and an ice model (to properly simulate regional fresh water fluxes). Dr. Spall will work with the NCAR version of the GFDL model; Dr Chassignet wil work with the MICOM. Together, an important phase for all three programs will be a model intercomparison study that will define necessary improvements to the models. Once the basin scale simulations and model intercomparisons are complete, our analyses will target processes in the upper and intermediate-depth subpolar regimes.
Furthermore, our group will begin to obtain a better appreciation for the impact of assimilating various components of the ACCE (and other historical) observations for improving model climatology. A series of experiments will be performed that assumes uncertainties in surface fluxes of momentum and heat and in initial and open boundary conditions and sets out to determine which set of model assimilated ACCE observations will be most important for improving model simulations. These experiments will additionally help us understand the sensitivity of targeted processes to controlled changes in surface fluxes and initial conditions and, together with our process-oriented model diagnoses, will contribute in placing the ACCE field program within the context of longer term, basin scale variability.
Our initial set of general working hypotheses are:
The POM is a primitive equation ocean general circulation model with an embedded turbulence closure scheme which has been proven in numerous studies to be a mature process, and ocean general circulation, model. It already includes several essential components that will enable us to model the subpolar gyre while accounting for the complicated topography/geometry and physics of the region: an orthogonal, curvilinear grid for flexibly increasing horizontal resolution near both geographic (topography and sills) and geophysical (western boundary currents and frontal) features; a bottom-following, vertical sigma-coordinate which places resolution exactly where it is needed: over sills and topography; a turbulent plume-entrainment model located near sill overflow regions; a well-tested turbulence closure model that determines vertical mixing coefficients for momentum and scalar variables (important for realistic SST and SSS fields; Mellor and Yamada 1982); the ability to incorporate a coupled ice model (Kantha and Mellor 1989); and a free-surface allowing for assimilation of sea surface height data.
A great deal of attention has beed paid to improving the POM for both North Atlantic regional process studies and full basin-scale Atlantic simulations. A subtropical gyre configuration (Ezer and Mellor 1994a) finds meridional heat fluxes which closely resemble the observations. Ezer and Mellor (1992) and Ezer (1994) demonstrate the importance of resolving small-scale topography and using realistic forcing in order to simulate realistic Gulf Stream separation, important for properly setting inlet conditions for the NAC. A fully coupled ice-ocean version of the POM has been used to simulate deep convection in the Greenland Sea (Hakkinen et al. 1992) and the seasonal variability in the Arctic and Nordic Seas (Hakkinen and Mellor 1992). The outflow of the Mediterranean over the sills of Gibraltar and into the Gulf of Cadiz has been reproduced (Ezer et al. 1995) as well as this saline plume's spreading into the intermediate layers of the North Altantic. Finally, ongoing simulations with the POM include a basin-scale, 80S to 80N model of the Atlantic that examines seasonal and interannual variability (Ezer, et al. 1995). All of these process and simulation studies using the POM enable us to begin with a very solid modeling foundation from which to meet our objective.
An important model component that will be added to the POM is a new formulation for surface heat fluxes. As argued by Seager et al. (1988), it is not logical to specify air temperature and humidity in the heat flux parameterization if SST is to be predicted since these specifications, to a large degree, pre-determine the SST. For global applications, the Seager et al. (1995) model now includes wind advection for calculating surface latent and sensible heat fluxes in terms of the model SST and winds. This model does an excellent job of capturing, for example, the enhancement of both latent and sensible heat fluxes observed over the Gulf Stream. This new flux formulation appears important for a more faithful, unbiased specification of surface fluxes for studying the evolution of SST and oceanic heat budgets and is being included in our numerical formulation.
The role of fresh water fluxes is critical in the creation of deep and intermediate waters in the subpolar gyre, Labrador Sea, and the Nordic Seas. High latitude evaporation rates are small; salt expulsion during freezing is the only effective manner in which surface salinity can be increased. To properly account for surface fluxes of fresh water due to formation and melting of sea ice, we couple a well-tested ice model (Mellor and Kantha 1989) with the POM.
Although our area of immediate interest is the subpolar region, it is important to model regional inflows and outflows correctly. Since the subpolar basin is not easily isolated, we first evaluate a computational domain configured to extend southward to 10S to include the subtropical gyre, and north to include the Labrador Sea and the Nordic Seas (but see the next paragraph). Northern boundaries will likely be the Davis Strait in the west and a line from Greenland-Spitsbergen-Norway in the east. The resolution in the subpolar region will be eddy-resolving to better capture observed levels of eddy kinetic energy already found by surface drifters and subsurface floats in the region (as part of Rossby/Rothstein's North Atlantic Current Experiment) and in resolving sites of potentially important transformations (fronts, interleaving regions and strong currents associated with the NAC).
The full-Atlantic basin (80S--80N) experiments that have recently been completed by Mellor and Ezer using the POM are important to our program and we will analyze our model Atlantic domain against the backdrop of these results, especially in verifying open boundary conditions. Important differences and priorities between the two groups include generally better resolution (and dynamical diagnostics) that targets subpolar processes, our use of an atmospheric boundary layer model for simultaneously obtaining realistic SST and ocean heat budgets, and our assimilation program. The collaborations with Mellor/Ezer on issues of common interest will help place the dynamical diagnoses of the regional processes in our experiments within the larger scale context of the full Atlantic MOC variability.
Our final intent is to evolve a numerical representation of the subpolar domain that is consistent with the observed historical database, the new ACCE observations and the model physics within the known errors of each of these. To begin, we ask the following: Which of the available ACCE observations contain most information (and which are redundant) in counterbalancing acknowledged deficiencies in the surface fluxes and initial and open boundary conditions? To accomplish this we first employ the POM in a program of synthetic data assimilation in preparation for assimilation of real ACCE, satellite and historical observations. We will evaluate a number of methods, including spatial objective analysis techniques (based on a statistical interpolation scheme solved as an equivalent variational problem; Derber and Rosati 1989) for assimilating direct ocean measurements (e.g. temperature and salinity) and optimal interpolation approaches (for assimilating sea surface height from synthetic or real satellite tracks; Ezer and Mellor, 1994b). Please see our data assimilation program for TOGA-COARE to get an idea of the technques that we will employ for our ACCE program.
