U.S. Dept. of Commerce / NOAA / OAR / GFDL *Disclaimer

 

5. OCEANIC CIRCULATION

GOALS


  5.1 WORLD OCEAN STUDIES

     5.1.1 Modeling Eddies in the Southern Ocean

ACTIVITIES FY00

          The Southern Ocean in general, and the Antarctic Circumpolar Current (ACC) in particular, are emerging as centrally important players in both the general circulation of the ocean and the earth's climate system. For this reason, the Ocean Group at GFDL has embarked on an ambitious, multi-year effort to numerically simulate the Southern Ocean, with sufficient horizontal and vertical resolution to resolve the ubiquitous energy containing mesoscale eddies.

          The Southern Ocean is unique in containing latitudes that are not blocked by land. Consequently, the leading-order dynamic balances (e.g., Sverdrup balance) do not obviously dictate the general structure of the ACC, unlike most of the other ocean currents. The northward Ekman transport in the unblocked latitudes must be balanced by the combination of the upwelling of waters from great depth and meridional eddy mass fluxes. The meridional density gradients associated with the ACC are intimately tied to the mean stratification of the entire ocean to the north of the ACC, and it is becoming increasingly evident that the dynamics of the Southern Ocean are crucial in determining the stratification and overturning circulation of the entire ocean (1313, 1595, 1700, 1728).

          Complete understanding of the interplay between the intense mesoscale eddies and the mean ocean circulation, and so of the general oceanic circulation itself, demands the use of numerical models with sufficiently high resolution to explicitly resolve these eddies, typically as small as 10 km in the ACC. These processes are being studied through simulations of the Southern Hemisphere oceans using GFDL's z-coordinate model (MOM) and GFDL's isopycnal-coordinate model (HIM). Fig. 5.1 shows a comparison of instantaneous velocities in the ACC near the tip of Africa. The top two panels show results from HIM and MOM at 1/2° resolution (25 km at 60°S). The bottom panel shows results from HIM at 1/4° resolution (13.9 km resolution at 60°S). The ACC flows toward the east between 40° and 45°S. The strong flow along the coast of Africa is the Agulhas current flowing toward the southwest. The Agulhas current in the real world is known to join the ACC south of Africa by retroflecting back to the east. It is also known to break up into eddies which propagate westward into the Atlantic. Fig. 5.1 shows that both HIM and MOM capture the retroflection at 1/2° resolution, but only the 1/4° model in the bottom panel is able to simulate the production of Agulhas eddies.

          Wind stress perturbations are being applied to both models over a range of resolutions to study the sensitivity of the dynamic balances between the mean circulation, eddy fluxes, and the diabatic circulation in the Southern Ocean. The complete suite of experiments at 1/2° resolution is nearing completion.

          The Southern Ocean provides an ideal test-bed for exploring the effects of eddy-flux parameterization schemes and the still more basic scientific issue of geostrophic turbulence in the ocean. Parameterizing (and therefore understanding) oceanic mesoscale eddies is necessary for properly representing the ocean in long-term coupled ocean-atmosphere climate simulations. The re-entrant channel geometry of the ACC makes it possible to pose the problem in a relatively clean form and, taken together with the inclusion of sub-tropical gyres and intense western boundary currents in the proposed simulations, allows the problem to be studied with unprecedented detail.

PLANS FY01

          Sensitivity studies with both MOM and HIM will be analyzed at 1/2° (isotropic) resolution. Remaining calculations, including wind stress perturbations, at 1/4° resolution will be carried out. A similar series of simulations at a resolution of 1/6° or 1/8° will be initiated. More idealized geometries will also be used to understand the dynamics.

     5.1.2 Southern Ocean Winds and the Circumpolar Current: Theoretical Studies

ACTIVITIES FY00

          A number of theories have been advanced to predict how the Antarctic Circumpolar Current should respond to changes in the wind stress over the Southern Ocean. According to one theory, the strength of the Circumpolar Current should respond to the mean wind stress within the open latitudes of Drake Passage. According to another, the meridional gradient of

the wind stress should be the most important factor. Still another theory predicts that the strength of the ACC is saturated, i.e., additional energy input from the wind simply makes for more energetic eddies. Until recently, none of these theories considered the impact of heating and cooling north and south of the ACC.

          A recent study (1728) suggests that any theory which purports to explain the response of the ACC to the wind must consider the buoyancy forcing north and south of the ACC as it pertains to the north-south pressure gradient across the current. Fig. 5.2 shows schematically

 
that upwelling in the south forces a conversion of dense deep water into lighter thermocline waters. This leads to an increase in the depth of the light-water pool in low latitudes and a larger north-south pressure gradient in the upper part of the water column. This pressure gradient increases the overturning in the Northern Hemisphere leading to warmer and lighter deep waters in the north. Insofar as the density of deep water in the Southern Ocean remains basically unchanged, the result is a larger north-south pressure gradient at all depths across the ACC and an increase in the transport of the Circumpolar Current.

          The picture is complicated by the presence of mesoscale eddies in the ACC. Eddies allow for a southward mass flux of relatively light water which replaces some of the dense water being upwelled from the deep ocean by the winds. Eddies thereby reduce the mass conversion and northern overturning depicted in Fig. 5.2. An idealized theoretical study to examine the effects of eddies has been carried out using a two-layer isopycnal primitive-equation model (nc). The model setup allows for strong topographic effects and the resolution of baroclinic eddies. A key parameter is the ratio between the Ekman flux, which lifts the interface between the two layers, and the diapycnal mass flux, which converts water from one density type to the other. When this ratio is small, i.e., the circumpolar current is relatively weakly forced by the wind, the conversion of dense upwelled water into lighter surface water dominates and the strength of the circumpolar current increases with a stronger wind stress. An increase in the wind stress is partly compensated by increases in the pycnocline depth and overturning at the northern wall, partly compensated by an increase in eddy fluxes. However, when the Ekman flux dominates at the point of instability, i.e., the conversion of dense upwelled water into lighter surface water can't keep up with the Ekman flux, increases in wind stress are compensated almost entirely by increases in the eddy flux.

          Figure 5.3 illustrates the importance of this non-dimensional ratio in governing the dynamics of a circumpolar current. It shows two realizations of an idealized ACC, both with the same ratio of Ekman flux to diapycnal flux. The idealized ACC in the top panel is forced with a

strong wind stress and a high rate of thermal damping in the sponge regions to the north and south. The ACC in the bottom panel is forced with a weaker wind stress and less thermal damping. The ACC transport is almost identical in the two runs, but the structure of the ACC and associated eddy fluxes are very different.
 

PLANS FY01

          As illustrated in Fig. 5.3, the eddy mass flux should be more energetic downstream of large topographic features, such as the Scotia Arc or Campbell Plateau. Homogeneous parameterizations of eddy fluxes, like the widely used Gent-McWilliams parameterization, are unlikely to be able to reproduce this asymmetry. It is unclear whether this will have important effects on their ability to predict the large-scale structure and dependence of the Circumpolar Current. Simulations will be carried out in which various parameterizations are included in a coarse-resolution version of the eddy resolving model. Results will be compared with the fully eddy-resolving calculations.

     5.1.3 The ACC and the Ocean's Thermohaline Circulation

ACTIVITIES FY00

           Much of the meridional heat transport into the earth's temperate and sub polar zones is carried by the ocean's thermohaline circulation. It has long been supposed that this circulation is maintained by diabatic heating in the interior brought about by the downward mixing of heat in low and middle latitudes. Given the central role of diabatic heating in the theory, the thermohaline circulation has been thought to carry tropical heat poleward toward both poles. The archetypal thermohaline circulation in today's ocean, seen in the Atlantic Ocean, does not readily conform to expectations. Both the warm upper branch and the cold lower branch of the Atlantic thermohaline circulation extend across the equator into the circumpolar region in the south such that much of the heat carried northward into the North Atlantic comes from outside the Atlantic basin. The cross-equatorial overturning in the Atlantic weakens the poleward ocean heat transport in the Southern Hemisphere, while it strengthens that of the Northern Hemisphere. The old theory requires special factors, e.g., extra salt in the Atlantic, to account for the interhemispheric nature of the Atlantic overturning.

          Earlier studies (1313, 1562) showed that the Atlantic's thermohaline circulation could very well exist without diabatic heating if it operates in conjunction with a wind-driven ACC in the far south. From this perspective, the ocean's dominant thermohaline circulation is interhemispheric by definition, and it owes its existence to Drake Passage, the gap between South America and Antarctica which allows the ACC to exist. An idealized coupled model has been created in which the climatic effects of an open Drake Passage can be explicitly simulated (1714). The model describes a mostly water-covered earth in which a full ocean GCM (MOM) is coupled to an energy balance model of the atmosphere. Land consists of a thin barrier which extends north and south between two Antarctica-like polar islands. The opening of Drake Passage is simulated by removing a section of the barrier near the south polar island. Air temperatures and oceanic SSTs are prognostic variables of the coupled system; no restoring boundary conditions are used.

          Results from the "water planet" model show that all the major features of the Atlantic thermohaline circulation and the ocean's heat transport system suddenly appear with the opening of Drake Passage. The high latitudes of the Southern Hemisphere cool down by some 3°C while the high latitudes of the Northern Hemisphere warm up by about the same amount, in accord with the observed temperature differences between the North and South Atlantic. A relatively warm, salty deep water mass forms next to the north polar island and flows southward across the equator at mid-depths much like North Atlantic Deep Water (NADW) in the real world. The deep water forming next to the north polar island is inherently salty in relation to deep water formed elsewhere.

          Virtually all the extra heat given up to the atmosphere in the high latitudes of the Northern Hemisphere comes from the high latitudes of the Southern Hemisphere, where cold water upwelled by Ekman divergence south of the model's ACC is warmed as it moves northward under the influence of the Southern Hemisphere westerlies. According to the old theory, the thermohaline circulation should cool the tropics as it transports tropical heat poleward. The tropics in the water planet model actually warm slightly and contribute no extra heat to the Northern Hemisphere when Drake Passage is opened. The pattern of warming and cooling generated by an open Drake Passage is proportional to the wind stress applied in the latitude band of the model ACC. Stronger winds thicken the thermocline north of the model's ACC in accord with the simple predictive scheme in (1595). Stronger winds enhance the amount of warm water that comes into contact with the atmosphere near the north polar island and thereby enhance the flux of oceanic heat to the atmosphere.

PLANS FY01

          This approach to the ocean's heat transport suggests that the earth's climate should be particularly sensitive to tectonic changes that open and close critical ocean gateways. It is hypothesized, for example, that the equable climates of the Cretaceous and early Cenozoic were maintained, in part, by the enhanced ocean heat transport caused by a circumglobal current, analogous to the ACC, that circled the earth in the latitude band of the northern tropics. Preliminary work with the water planet model shows that the opening of a Drake Passage-like gap in the northern tropics cools the entire tropical ocean by some 2°C and boosts the northward oceanic heat transport out of the tropics up to 4x1015 W.

     5.1.4 Ocean Eddy Energies, Scales, and Vertical Structure

ACTIVITIES FY00

          Turbulent motions comprise a significant fraction of the oceanic energy budget and transport, yet remain at the edge of what computer models can resolve. Parameterization of such motions is thus a continuing effort, but one for which only intermediate solutions currently exist, all of which have deficiencies. A particular approach pioneered by Held and Larichev (1362) is based on the fundamental aspects of fully developed geostrophic turbulence generated by baroclinic instability, and is thus most likely applicable in regions of highly unstable flow - a fair characterization of much of the world's ocean. A new project extends the Held and Larichev formalism to incorporate the effects of arbitrary stratification and shear. The modified theory specifies a vertical structure for the eddy diffusivity which automatically ensures conservation of the appropriate material properties. In particular, the theory specifies a form for the northward eddy potential vorticity flux, v'q', shown in Fig. 5.4, which always

integrates to 0, thereby ensuring that the parameterization conserves momentum. The modified theory has been tested in a multi-layer homogeneous quasi-geostrophic turbulence model with realistic stratification and mean shear with encouraging results.
 

PLANS FY01

          The proposed parameterization presumes that an eddy producing flow is highly unstable. While this is likely the case, no specific analysis of ocean climatology has addressed this question specifically. Moreover, alternate theories exist which are likely more appropriate for the weakly unstable limit. In order to determine the degree to which regions of the global ocean are stable, weakly unstable or strongly unstable to both baroclinic and barotropic instability, a linearized balanced model appropriate for large scale and frontal regions will be used to assess the regional instability of oceanic currents. By comparing these results to measurements of the four-dimensional spectrum of mesoscale variability throughout the ocean, an assessment will be made regarding the degree to which various regions are controlled by linear or non-linear processes, leading to the determination of appropriate eddy closure schemes for each region. Plans are underway to test the modified theory in an isopycnal primitive equation model (HIM) and to extend the theory to incorporate the effects of non-zonal flow, surface buoyancy flux and topographic roughness.

     5.1.5 Geostrophic Turbulence

ACTIVITIES FY00

          An interesting research project emerged from a graduate class taught in the fall of 1999. In lieu of a final exam, the students worked on various computational/theoretical problems in geostrophic and two-dimensional turbulence. These projects were of sufficient interest and originality that they have been extended and focussed. The common theme of the work is the inverse cascade in forced-dissipative turbulence. The inverse cascade is of particular oceanographic significance, because it is one means whereby mesocale eddies interact with the large-scale circulation, potentially providing natural variability at the gyre scale. Here, we will highlight just a few of the results that have emerged so far in this idealized, homogeneous framework.

          Ocean general circulation models typically use a linear Rayleigh damping on the bottom layer vorticity to simulate Ekman drag. Because baroclinic modes are primarily surface trapped, bottom drag acts primarily to damp the barotropic mode. The degree and mechanism by which drag is involved in setting the observed eddy scales is an open question. Two-dimensional turbulence is an approximation of the ocean's barotropic mode, thus a simplified two-dimensional model can be used to selectively examine the dynamics of linear drag on non-linear transfers. A theory which predicts the steady-state energy spectrum of this system has been devised and tested against a simulation of two-dimensional turbulence in a doubly periodic domain forced at high wave number (Fig. 5.5). Given the highly non-linear character of these flows, it is remarkable that the theory predicts not only the mean scale of

the eddies, but their magnitude and energy distribution as well. Ongoing research will address the combined effects of more realistic quadratic drag, differential rotation and finite radius of deformation on eddy energies and scales, as well as on tracer variance distributions, which approximate baroclinic mode energy at large scales.

  5.2 MODEL DEVELOPMENT

     5.2.1 Modular Ocean Model

ACTIVITIES FY00

          Version 3 of GFDL's Modular Ocean Model (MOM 3) was developed and optimized for use on vector computers such as the CRAY T90. It can also run on distributed memory systems like the CRAY T3E. Model performance suffers in two respects, however, when MOM3 is run on distributed memory systems. First, the amount of memory required per processor is not reduced as more processors are used to attack a problem. This means that large models may not be able to fit in available memory even though the computational domain is divided among many processors. The second deficiency is an inadequate increase in speed when the work load is spread across a large number of processors. Because of the way that arrays are laid out in memory, it is impossible to address these deficiencies within MOM 3.

          A new version of MOM (MOM 4) has been created to address these deficiencies. MOM 4 has been written in Fortran 90 syntax and is structured so that it can use generalized curvilinear coordinates. Early tests indicate that MOM 4 memory requirements and speed scale significantly better than in MOM 3 on the CRAY T3E. Better scaling of memory has been achieved by eliminating MOM 3's memory window and allocating all arrays to the size of the local domain on each processor which is determined at run-time.

          Better scaling of model speed across multiple processors has been realized due to the two-dimensional domain decomposition used in MOM 4. Fig. 5.6 shows a comparison of model

speedup versus number of processors in MOM 3 and MOM 4. The test configuration used 64 rows of latitude, 120 longitudes, and 15 vertical levels. Both versions of the model were run on 2, 4, 8, 16, 32, and 64 processors of the GFDL CRAY T3E and execution times were normalized by the execution time on 1 processor. The circles indicate the speedup over one processor using MOM 4 and the triangles represent the speedup over one processor using MOM 3. The speedup in MOM 3 peaks at 32 processors and then declines while the speedup in MOM 4 on 64 processors is still increasing. In absolute terms, the execution time of MOM 4 on 64 processors is a little less than one half of the execution time of MOM 3 on 64 processors.

          Coding complexity has also been reduced by removal of the memory window and by the use of averaging and derivative operators. Further improvements in speed are expected to follow once operators have been optimized for better use of cache and registers. MOM 4 is compliant with the Flexible Modeling System (FMS) and will be added to the FMS repository once the driver for interfacing MOM 4 to the atmosphere and ice models has been written. After MOM 4 has been added to the FMS repository, MOM 4 will be available for addition of physics modules.

PLANS FY01

          Future plans include further optimization of speed once the new GFDL computer system is in place. Parameterizations such as Redi-diffusion and Gent McWilliams skew flux will have to be rewritten. An adjoint of MOM 4 will be built using the Giering TAF compiler. MOM 4 will be coupled to ice and atmosphere models using the FMS exchange grid.

     5.2.2 Isopycnal Coordinate Model Development

ACTIVITIES FY00

          Isopycnal coordinate ocean models offer several potentially important advantages over traditional level models. But disadvantages (such as the difficulty of representing the nonlinear equation of state, the need to use a nonlinear and nonlocal treatment of vertical mixing especially in dense flows over sills, and the difficulty of representing mixed layer dynamics in the detraining phase) have in the past been limitations for this class of model. Many of these difficulties have now been addressed.

          GFDL's isopycnal coordinate ocean model (HIM1.0) was officially released to the community this past year. New technical features include run-time specification of many parameters, flexible registration style temporal averaging and determination of the output fields, full support of NetCDF-based input and output, and run-time determination of parallel processor count. HIM may be run in parallel using a one- or two-dimensional domain decomposition, and this past year HIM was ported to 8 different parallel computers without having to change a single line of code. Also, the calling interface of HIM1.0 is now essentially compatible with that of MOM 4. HIM development is now managed with modern version control software.

          The effects of the nonlinear equation of state of seawater are particularly tricky to incorporate into a model which uses a density-like vertical coordinate. In the past year, all of the effects of the nonlinear equation of state, including thermobaricity and cabbeling, have been incorporated into HIM. Potential density referenced to an interior pressure (typically 2000 dbar) is used as the vertical coordinate. But various other terms involving density gradients use different, locally appropriate measures. Gradients of in situ density (adjusted by a function of pressure only) are used to calculate pressure gradient accelerations. Locally referenced potential density is used to estimate the shear-Richardson-number-dependent diapycnal mixing (1707). Cabbeling is included by advecting potential temperature and salinity, and adjusting the vertical mixing to restore the coordinate density of each layer towards its target value. Mixed layer dynamics use potential densities referenced to the surface to evaluate the mixed layer turbulent kinetic energy balances (although convective adjustment is also done if the coordinate variable is unstably stratified). In short, by separating the role of density as the coordinate variable from its other dynamic roles, all of the effects of the nonlinear equation of state of seawater can be described within an isopycnal coordinate ocean model.

PLANS FY01

          Improved treatments of the mixed layer dynamics will be explored. These may include depictions of Ekman-driven mixed layer stratification, baroclinic instability within the mixed layer, penetrative short-wave radiation, and rotational constraints on convective efficiency. The use of HIM as the ocean component of a coupled model will be tested. Hybrid pressure-density vertical components and alternative horizontal grids will also be examined.

  5.3 COASTAL OCEAN MODELING AND PREDICTION

     5.3.1 East Coast and North Atlantic Modeling and Forecasting

ACTIVITIES FY00

          A 6-year simulation of the Atlantic Ocean, west of 55°W, has been carried out using the sigma coordinate Princeton Ocean Model with a curvilinear grid. A comparison between experiments with and without surface fluxes shows that the effect of the surface wind stress and heat fluxes on the Gulf Stream path and separation is closely related to the intensification of deep circulations in the northern region. The separation of the Gulf Stream and the downslope movement of the Deep Western Boundary Current (DWBC) are reproduced in the model results. The model DWBC crosses under the Gulf Stream southeast of Cape Hatteras and then feeds the deep cyclonic recirculation east of the Bahamas. The model successfully reproduces the cross-sectional vertical structures of the Gulf Stream, such as the asymmetry of the velocity profile and the eddy activity of the Gulf Stream. Entrainment of the upper layer slope current into the Gulf Stream occurs near the cross-over point; the converging cross-stream flow is nearly barotropic.

          Using the same model, the Loop Current and deep circulation in the Gulf of Mexico has been described. A deep cyclonic circulation, bounded by the deep basin in the eastern Gulf, is shown to be spun up by the Loop Current. The Loop Current also induces deep anticyclonic columnar eddies in the eastern Gulf which decouple from the upper layer Loop Current. The westward translation speed of a Loop Current Ring is about 2.5-6 cm/s. Lower layer eddies have a higher speed and lead the rings into the central Gulf. The time-average surface circulation of the Gulf of Mexico basin is anticyclonic, mainly due to the transport of anticyclonic vorticity by Loop Current Rings in the surface layer. An average lower layer cyclonic circulation occurs along the continental slope of the basin.

           The Data Assimilation and Model Evaluation Experiments in the North Atlantic Basin (DAMEE-NAB), supported by the Office of Naval Research (ONR), has been completed with the publication of a special issue that includes an intercomparison between six different ocean models (1716). Sensitivity studies evaluate the effect of open boundary conditions, horizontal diffusivity, and model resolution on model variability and Gulf Stream dynamics (1715).

PLANS FY01

          Future research will focus on the dynamics of the Loop Current and the deep circulation in the Gulf of Mexico. A coastal forecasting system based on a higher resolution version of the extended east coast model will be tested.

     5.3.2 Turbulent Boundary Layer Modeling

ACTIVITIES FY00

          An improved surface boundary layer formulation has been developed to improve the Mellor-Yamada boundary layer model (ky). Unlike the situation in three-dimensional simulations or in the real ocean, the kinetic energy in one-dimensional surface layer models can build up and artificially enhance local mixing. Adding a sink term to the momentum equations counteracts this behavior. The sink term is a surrogate for energy divergence available to three-dimensional models, but not to one-dimensional models. The new sink term tends to exacerbate problems with overly warm summertime surface temperatures. A Richardson number dependent dissipation term yields a favorable improvement in the comparison between model calculations and observations.

          Further testing of the modified Mellor-Yamada turbulence scheme with a three-dimensional North Atlantic Ocean model (1717) shows improvement in the simulations of the seasonal mixed-layer depth. The mixed layer in the three-dimensional model is shown to be strongly influenced by errors in the surface heat flux, the frequency of the forcing wind stress and most importantly, by penetration of short wave radiation, making the evaluation of the new turbulence scheme in three-dimensional models much more difficult than it is in one-dimensional models.

PLANS FY01

          Future work will incorporate the effects of surface waves in both the surface boundary layer and bottom boundary layers in shallow water.

     5.3.3 Princeton Ocean Model Development and Testing

ACTIVITIES FY00

          The Princeton Ocean Model (POM) users group has continued to grow. It now includes more than 700 users from 54 countries. User support and code development continues (for more detail consult the POM web page - www.aos.princeton.edu/WWWPUBLIC/htdocs.pom). Some of the new features that have been tested with POM during the last year include: a) a modified Mellor-Yamada turbulence scheme (1717, ky) that improves the simulation of vertical mixing and surface layers; b) a vertical generalized coordinate system that can accommodate z-level, sigma-level or other vertical discretizations (ip); c) a multidimensional positive definite advection transport algorithm; and d) a sixth-order accuracy combined compact difference scheme that significantly reduces pressure gradient errors. A parallel version of POM is currently being tested.

PLANS FY01

          A new ONR-supported initiative will foster a collaboration between the POM developers and other ocean model developers from Rutgers University in order to develop and improve terrain-following ocean models and community support.

     5.3.4 Climate Variability Studies with POM

ACTIVITIES FY00

          Sigma coordinate ocean models, originally developed for regional coastal studies and prediction, are now being used for long-term climate simulations. Attention is given to surface and bottom boundary layers which may play an important role in surface heat exchange and in deep water formation processes. Observed decadal variations in the North Atlantic Ocean during the period 1950-1989 have been successfully simulated with an Atlantic Ocean version of POM (1663) and compared favorably with a simple wind-driven Rossby wave model (ny). The simulations demonstrate the important role played by westward propagating planetary waves in affecting the subtropical gyre and Gulf Stream variations. An interesting finding was that, on decadal time scales, the ocean model responds in a linear fashion to the combined effect of surface temperature and wind stress anomalies. Simulations of the response of the ocean to heat and fresh water flux anomalies in high latitudes, a result of possible future climate change, show an adjustment process that takes a few decades (nz). Spatial climatic changes in circulation patterns and in coastal sea level will be studied in detail.

PLANS FY01

          Research will focus on the interaction between climatic changes in the open ocean and variability in the coastal ocean. The effect of bottom boundary layers and the way they are formulated in ocean models on deep water formation will be further studied.

     5.3.5 Coastal Models of the West Coast and the Gulf of Mexico

ACTIVITIES FY00

          Research on the coastal observing systems off the west coast of the U.S. have started up recently as part of the National Ocean Partnership Program (NOPP). This work shows that details of the wind field, particularly the local wind curl, can be an important forcing to coastal circulation (1611). Studies based on coastal observations and models that continue to deploy sparse arrays of wind stations and coarse-resolution wind products (e.g., NCEP or ECMWF) may therefore be incomplete. Hindcasts of the coastal variability in the Santa Barbara Channel, carried out in collaboration with the Scripps Institute of Oceanography, show that the high-resolution wind fields give the best results.

          Observations have suggested that episodic subsurface current events in the Gulf of Mexico are caused by topographic Rossby waves (TRW) forced by Loop Current pulsation (north/south extrusion and retraction) and eddy shedding in the eastern Gulf. However, the existence of TRWs in coastal forecast models has never been rigorously established. A ten-year simulation has been analyzed to isolate the TRWs. Over 70% of the simulated subsurface energy was observed to reside in the 20 to 100 day periods in narrow bands over the continental slope and rise. Bottom intensification has been shown to exist in these high-energy bands. While the modeled TRWs are unambiguously forced by variability induced by the Loop Current and Loop Current Eddies, precise mechanism(s) through which energy is transmitted to lower layers is not yet understood.

PLANS FY01

          High-resolution simulations of coastal variability along the west coast and Gulf of Mexico will continue.

  5.4 GLOBAL BIOGEOCHEMISTRY AND THE CARBON CYCLE

     5.4.1 Terrestrial Carbon Cycling

ACTIVITIES FY00

          Through a detailed analysis of forest inventory data, the relative contribution of land use, carbon dioxide fertilization, and nitrogen fertilization to carbon sequestration in U.S. forests has been quantified. Results show that land-use is the primary factor governing the rate of carbon sequestration in U.S. forests. This conclusion is further supported by a second analysis that compares U.S. carbon budgets obtained through inventories of forests and other components of the terrestrial carbon cycle with U.S. carbon budgets obtained through inverse modeling of CO2 flask sample measurements. The findings of this study support the conclusion that changes in land-use account for most of the carbon being sequestered within the U.S., and reconciles a longstanding perceived inconsistency in the estimates obtained by these different methods.

          The first implementation of a new terrestrial biosphere model, the Ecosystem Demography model (ED model), has been documented. A second implementation of the off-line model that examines the impact of human land-use on the terrestrial carbon cycle of North America is currently being completed. The results of this study show how historical changes in patterns of land-use across the continent during the past three centuries account for the spatio-temporal distribution of carbon stocks and fluxes during this period and how these effects are likely to continue into the next century.

          The ED model has also been successfully coupled to a mesoscale model of the atmosphere (MM5 V3.3), and significant progress has been made towards integrating ED into GFDL's new FMS. Energy and moisture exchange parameterizations in MM5 have been modified to account for the sub-grid heterogeneity in vegetation distribution predicted by ED. The sub-grid scale heterogeneity includes ecosystem age and species type, and vertical profiles of plant canopy characteristics, in particular, the Leaf Area Index (LAI) and stomatal closure. The coupled ED-MM5 model is being used to explore the role of vegetation structure on diurnal patterns of fluxes between land surface and the lower atmosphere. Preliminary results obtained using the original OSU-ETA soil hydrology model show that the diurnal mode is able to capture the diurnal pattern of moisture and heat fluxes between atmosphere and biosphere in the tropical region.

          A canopy interface that couples the atmosphere with vegetation and soil processes modeled by ED has also been developed. In this parameterization, sensible heat and moisture fluxes from leaves and soil surface are calculated using a multiple resistance parameterization and turbulent mixing inside the canopy is assumed to be dominated by gusts with a size characteristic of the height of canopy. Mass and energy are transported directly from leaf surfaces to the top of the canopy, while diffusion between leaf layers is assumed to be small and thus neglected. Effective canopy air temperature and moisture content near the top of the canopy are calculated as linear combinations of temperature at the bottom layer of the atmosphere, ground temperature and leaf temperature through the canopy.

PLANS FY01

          The ED model will be integrated with forest inventory and land-use history data to complete a global, off-line implementation. The MM5-ED model will be used to: 1) analyze the impact of anthropogenic land use practices on the regional distribution and dynamics of CO2 sources and sinks; and 2) understand the role of plant rooting depth on the biophysical feedback to local climate. This research will explore a range of perturbations to the land atmosphere system that represent scenarios of land-use and vegetation succession in different regions. Once incorporated into FMS, ED will form the terrestrial component of a global, land-atmosphere-ocean model that we will use to study the importance of biospheric feedbacks for global climate. A particular focus will be the identification of regions where ecosystem dynamics have a significant impact on the regional climate.

     5.4.2 Inverse Modeling of Carbon Isotopic Ratios of CO2 in the Atmosphere

ACTIVITIES FY00

          Over the past two years, the steady-state distribution of carbon isotopic ratios in atmospheric CO2 has been modeled using the GFDL Global Chemical Tracer Model (GCTM), which uses a year-long record of wind fields generated previously by a general circulation model. Prescribed air-sea isotopic fluxes without seasonal variability are used to force the model over the oceans. The annual isotopic fluxes due to terrestrial net primary production and respiration were estimated for three land regions (Eurasia, North America, and the tropics and Southern Hemisphere) by inversions of atmospheric observations of the isotopic ratios. The inversion results suggest for the 1993-1995 period the presence of a terrestrial carbon sink in the midlatitude Northern Hemisphere and a terrestrial carbon source in the tropics. The terrestrial carbon source and sink were estimated previously for the 1980s and 1990s from inversions of atmospheric CO2 mixing ratios.
 

PLANS FY01

          The goal of this work is to estimate, using inverse models, the isotopic carbon fluxes between air and ocean, as well as between air and land. The isotopic fluxes are caused by net oceanic CO2 uptake and by the existence of an air-sea disequilibrium due to a lagged isotopic Suess Effect in the ocean. On-going global ocean general circulation and biogeochemistry model simulations by the Carbon Modelling Consortium (CMC) will be used to predict the net CO2 uptake and the disequilibrium isotopic carbon fluxes from the pre-industrial times to the present. The inversion results will be combined with ocean models to improve the estimate of the oceanic uptake of CO2 and the estimate of the terrestrial carbon fluxes for different geographical regions. A time-dependent inverse model of atmospheric carbon dioxide and its isotopic ratios may be developed towards this goal.

     5.4.3 Mixing Parameterizations, Large-Scale Ocean Circulation, and Global
              Biogeochemical Cycles

ACTIVITIES FY00

          It has been shown that the large-scale pycnocline depth can be explained by multiple combinations of wind stress, Southern Ocean eddy fluxes, and low-latitude diffusion (1595). However, different combinations of vertical and lateral mixing result in very different pathways for vertical exchange. A suite of model runs was performed in which the vertical and lateral diffusion were changed in a way that produced a similar pycnocline depth. In one case (with low vertical and lateral mixing) relatively little upwelling of deep water occurred through the low-latitude pycnocline. In a second case (with high vertical and lateral mixing) about 18 Sv of upwelling occurred through the low-latitude pycnocline. The high and low mixing cases are quite similar in their thermal structure, but the high mixing case has slightly higher uptake of CFCs and anthropogenic CO2, and a new production which is twice that of the low mixing case (pc). This is because the high mixing model has more upwelling at low latitudes and more convection at high latitudes. Additional model runs were made with high vertical mixing only in the Southern Ocean. In these runs, the amount of convection in high latitudes increased, but the low-latitude upwelling was essentially unchanged.
 

PLANS FY01

          Analysis of the suite of model runs will continue. Particular attention will be paid to the effect of nutrient fertilization on anthropogenic CO2 uptake and low-latitude production. Development of the ocean general circulation model will continue with a focus on improving the deep circulation.

     5.4.4 Air-Sea Fluxes of O2 and CO2 Determined by Inverse Modeling
              of Ocean Bulk Measurements

ACTIVITIES FY00

          A method has been developed to estimate the air-sea fluxes of gases from their observed distributions in the interior of the ocean. An ocean circulation model is used to establish a relationship between a unit air-sea flux in a given region of the surface ocean and the concentration at any point in the interior of the ocean. The surface ocean is divided into 15 regions, each of which contributes to the interior concentration at any location in the ocean. The tracers are linear (i.e., additive), so the model predicted concentration in any given region is equal to the sum of the 15 components. In a final step, the Singular Value Decomposition is used to estimate the actual magnitude of the air-sea fluxes in each of the 15 regions by requiring that the predicted interior concentrations fit the observations.

          The method has been applied to estimating the air-sea flux of heat, water, O2, and CO2. Before the method could be applied to O2 and CO2, the observed concentrations had to be corrected for the changes in water column chemistry that are due to biological cycling. This was done using the observed phosphate, nitrate, and alkalinity distributions, and assuming that the stoichiometric oxygen and carbon to phosphate ratios in organic matter are constant. The observed CO2 distribution must also be separated into a pre-industrial component, which is assumed to be at steady state, and an anthropogenic component. This separation enables estimation of both the pre-industrial air-sea flux distribution, and anthropogenic carbon uptake.

PLANS FY01

          Estimates of CO2 uptake will be made as more estimates of anthropogenic CO2 become available from the World Ocean Circulation Experiment.

     5.4.5 An Ecosystem Model for Biogeochemical Studies

ACTIVITIES FY00

          A simplified ecosystem model has been developed for use in biogeochemical studies. The ecosystem model is simple in the sense that it has few state variables, yet advanced in the sense that it captures fundamental features concisely. The production model consists of two state variables representing small and large producer organisms. These are parameterized in a way that captures the difference between small and large phytoplankton, as well as the production and temperature dependent behavior of the f-ratio. The model is simple enough that four of the parameters needed for its specification can be estimated from the f-ratio data, and the other two can be estimated from fits to Levitus nutrient data.

PLANS FY01

          The ecosystem model simulations will be compared with local patterns of seasonal variation in nutrients and chlorophyll at long term study sites (the Bermuda Atlantic Time Series and Ocean Weather stations), as well as satellite chlorophyll data and primary production estimated from these data. These comparisons give a measure of the ability of the model to reproduce seasonal and interannual patterns of production, as well as identifying regions where additional biogeochemical processes such as iron limitation will need to be added. Future development of the model will include splitting of production among taxonomic categories (diatoms, calcium depositors, and other algae) to match observed latitudinal patterns in silicate, phosphate, and alkalinity.

     5.4.6 Oceanic Nitrogen Cycle

ACTIVITIES FY00

          An analysis of nitrogen cycling in the Pacific has recently been completed based on nutrient data from the World Ocean Circulation Experiment (WOCE). The analysis uses the tracer N* (nitrate - 16phosphate) to identify regions where nitrogen fixation and denitrification occur (1497). The analysis shows the basin to contain sources and sinks of fixed nitrogen with magnitudes that are significant on the global scale (about 50 TgN/yr of water column denitrification - a sink, and about 60 TgN/yr of newly fixed nitrogen).

PLANS FY01

          Work is currently under way to prepare gridded fields from the high quality nutrient dataset provided by the WOCE program, with the aim of diagnosing the spatial patterns and rates of nitrogen fixation and denitrification by damping an OGCM toward those observed nutrient fields. This will allow exploration of a number of hypotheses regarding what controls the rates and locations of nitrogen inputs and outputs to the ocean. It will also enable diagnostic models of ocean carbon uptake to account for carbon transports that are associated with source and sink pathways in the marine nitrogen cycle.

     5.4.7 Global Patterns of Marine Silicate, Nitrate, and Alkalinity Cycling

ACTIVITIES FY00

          The large-scale field of nutrients has been used to identify regions where particular functional groups of plankton (in particular, silicifying organisms such as diatoms) are important. Typical diatoms have a silicate to nitrate molar uptake ratio of approximately 1:1. This observation is used to define a tracer Si* = (Si) - (N) of excess silicate concentration relative to the needs of such diatoms (Fig. 5.7). The dominant feature of deep ocean Si* is an increase along the pathway of the deep thermohaline circulation from a low of 0 to 10 mol kg-1 in the North Atlantic to highs in excess of 100 mol kg-1 in the deep North Indian and Pacific Oceans. This increase results from the well-known preferential dissolution of silicate in the deep ocean.

The high Si* waters of the abyss reach up to the surface in the Southern Ocean south of the Antarctic Polar Front. There the silicate is preferentially stripped out by diatoms with a Si:N ratio of approximately 5:1, consistent with this being an iron limited region. This generates a broad 20°-wide band of negative surface Si* around the entire Southern Ocean that subsequently penetrates to the north at the depth of the Subantarctic Mode Water (SAMW). The negative Si* supplied by the SAMW is the dominant feature of the upper kilometer of all ocean basins except the North Pacific, which is the only other place in the world where high Si* waters of the deep ocean are able to reach the surface. Consequently, most areas of the ocean outside the North Pacific and Southern Ocean have a deficit of silicate relative to nitrate. Within the North Atlantic, the low Si* water is swept downward with North Atlantic Deep Water. A more detailed examination of Si* shows how other processes such as shallow remineralization of nitrate relative to silicate modify the above overview. Of particular interest is the fact that there is no clear indication of a high Si:N drawdown in the iron limited North Pacific.
 

PLANS FY01

          The global analysis will be extended to look at the effects of calcifying organisms. The goal is to identify regions where coccolithophorids are dominant producers using the water mass distribution. Results will be compared with satellite estimates for calcification under way at other institutions. A key product will be an estimate of the fraction of new production associated with these functional groups. Simulations which evaluate the effects of changes in the relative fraction of carbonate and silicate producing organisms on the carbon cycle will then be carried out.

     5.4.8 Analysis of the Ocean's Carbon Pumps

ACTIVITIES FY00

          It is almost axiomatic that temperature anomalies imposed on the ocean near the poles will have a greater effect on the partitioning of CO2 between the ocean and atmosphere than temperature anomalies imposed in low latitudes. This is true because most of the ocean's volume comes into contact with the atmosphere through cold polar outcrops. Two influential papers published during the last two years1,2 have shown that this polar temperature sensitivity is not nearly as strong in GCMs as it is in box models. Broecker and Archer suggest that this is true because vertical mixing in GCMs allows the solubility effect of warm surface temperatures to be felt well down in the interior of the ocean. Broecker and Archer imply that the mixing effect in GCMs carries over to the other "carbon pumps" in the ocean, namely those involving the production and remineralization of organic particles and CaCO3.

          A new method has been devised for disaggregating the carbon pumps in ocean models to explore these issues. The results show that the critical distinction between box models and GCMs has less to do with mixing than with the relative areas over which atmosphere-deep ocean communication is allowed to occur. GCMs have less polar sensitivity than box models because the high-latitude gas exchange that allows CO2 to move between the deep ocean and atmosphere is restricted by limited areas of convection and deep water formation. The polar boxes in box models tend to be fairly large, typically occupying 5-10% of the overall ocean area. With such a large area, the effect of gas exchange in box models is less limiting. It is found, furthermore, that the effect of limited gas exchange is different for the thermal solubility effect than it is for the CO2 pumped down into the deep ocean by organic particles. This is because the critical region where limited gas exchange influences CO2 solubility is in the North Atlantic while the critical region for the organic carbon pump is in the Southern Ocean.
 

PLANS FY01

          The new method for disaggregating carbon pumps described above has been applied so far only to one GCM-based biogeochemistry simulation (1617). In the coming year, the new method will be used to examine other GCM simulations and data from the real ocean.

     5.4.9 Response of Ocean Biology to Future Climate Change

ACTIVITIES FY00

          In collaboration with modeling groups at CSIRO (Australia), the Hadley Center (United Kingdom), Max Planck Institute (Germany), the Institute Pierre et Simon Laplace (France), and NCAR, six different coupled climate model simulations of future climate change are being examined to determine the range of behavior of those aspects of global warming simulations that are relevant to the ocean biological response. The overall response inferred from examining the physical response of the ocean to global warming is decreased biological production in low latitude upwelling regions and the poleward half of the subtropical gyres, and increased production in the polar regions.

          Wind-driven upwelling is the dominant mechanism of nutrient supply along the highly productive western margins of the continents and in the equatorial regions. Models predict widely varying results, but the general tendency is towards a reduction of upwelling in these regions, from which it is inferred that biological production would decrease. The dominant mechanism for nutrient supply in the subtropical gyres poleward of the subtropical convergence zone is wintertime convection. These regions tend to become more stratified with future climate change, which reduces the depth of wintertime mixing. The expectation, supported by model predictions, is that this would result in reduced biological production. The polar regions generally have a high supply of nutrients due to upwelling and convection, but can suffer from low productivity due to low light supply in deep mixed layers. Increased stratification, which occurs in most models, though with a complex pattern, would thus tend to increase biological production. Exceptions to this would be where low levels of micronutrient supply by dust limit the production, such as is thought to be the case in the Southern Ocean and North Pacific, or where the decreased mixing reduced the nutrient supply to less than the potential biological uptake.

          The mechanism of nutrient supply to regions between the equatorial upwelling bands and subtropical convergence is poorly understood and poorly simulated in most models. It is difficult to determine how these regions will respond to future climate change. The changes described will also very likely lead to changes in ocean ecology, as the major phytoplankton groups such as diatoms, coccolithophorids, flagellates, Phaeocystis, and nitrogen fixers are sensitive to water column stratification, as well as nutrient content.

PLANS FY01

          The analyses made to date are local analyses, and do not consider impacts of changes in one part of the ocean on other parts of the ocean. In the coming year, special attention will be paid to this question. A particular area of interest is the effect of increased high latitude production on nutrient supply to the low latitudes.


1. Broecker, W. S., J. Lynch-Stieglitz, D. Archer, M. Hofman, E. Maier-Reimer, O. Marchal, T. Stocker, and N. Gruber, How strong is the Harvardton-Bear constraint?, Global Biogeochem. Cycles, 13, 817-820, 1999.

2. Archer, D., G. Eshel, A. Winguth, W. Broecker, R. Pierrehumbert, M. Tobis, and R. Jacob, Atmospheric CO2 sensitivity to the biological pump in the ocean, Global Biogeochem. Cycles, in press.

*Portions of this document contain material that has not yet been formally published and may not be quoted or referenced without explicit permission of the author(s).