Overview
A key WOCE goal is to develop ocean models for climate research. These models
should be able to represent the largest-scale overturn and renewal of water-masses in
the global ocean. We examine whether coarse resolution ocean models - such as those
used in climate change studies - can reproduce the ventilation time-scales of ocean
overturn. A number of experiments are examined spanning a range of mixing schemes
(Cartesian, isopycnal, and Gent and McWilliams (1990, GM), as well as combinations
thereof), with a variety of diffusion coefficients. The GM runs include a standard case
without background horizontal diffusion, another in which horizontal diffusion is
maintained, and a third wherein the isopycnal tracer diffusion is set equal to the
isopycnal thickness diffusivity. In the final GM case, the isopycnal thickness diffusivity
is halved. Of particular interest is the role of the tracer mixing schemes in determining
ventilation processes such as mixed layer formation, deep water renewal and abyssal
flow regimes. We employ CFCs, radiocarbon, and idealized tracers to monitor overturning
rates in each model experiment.
Simulated ventilation rates vary greatly between the three mixing schemes. Isopycnal
mixing runs exhibit the most rapid water mass renewal due to strong diffusion effects
and excessive surface convective overturn, particularly in the Southern Ocean. In
contrast, all GM cases show gradual renewal of deep and bottom waters, with limited
vertical convection of surface waters and slower abyssal currents. Background horizontal
diffusion does not significantly alter interior ocean ventilation rates under GM,
whereas it results in spurious diapycnal fluxes in non-GM runs.
Comparison has made between the model simulations of CFCs and radiocarbon with
WOCE measurements in the ocean. All model runs exhibit only gradual and shallow
water-mass renewal in the North Atlantic Ocean, with little NADW outflow below
2000-m. Deepening critical sills in the region does not alleviate this problem. Adjusted
surface thermohaline forcing can enhance outflow rates of NADW, though at the
expense of requiring artificially high surface heat and/or freshwater fluxes. In the
Pacific and Indian Oceans, Cartesian and isopycnal mixing runs are ventilated too
rapidly due to strong convection and water-mass contribution from the Southern Ocean.
In contrast, GM runs simulate spuriously old and radiocarbon-depleted Circumpolar
Deep Water. Overall, none of the model cases reproduce global ocean ventilation rates
over decadal to centennial time-scales. Higher horizontal resolution and a spatially-varying
GM thickness diffusion coefficient may be required before global models
capture ocean ventilation rates with some degree of fidelity.
Experimental design
Experimental design adopted for the present study. The seven cases listed are the Cartesian-mixing
experiments (HOR-Z and HOR=0.75), the isopycnal mixing experiment (ISO) and the Gent and
McWilliams (1990; GM) experiments. AHH refers to the horizontal diffusivity, Ar(z) to a depth-
dependent profile for isopycnal diffusion, and k to the GM isopycnal thickness diffusivity.
| Experiment: | Subgrid-Scale Eddy Parameterisation: |
HOR-Z | Horizontal mixing profile, AHH = 1.0 [surface] to 0.5 [bottom] x 10^3 m2/sec. |
HOR=0.75 | Horizontal mixing (constant AHH = 0.75 x 10^3 m2/sec). |
ISO | Isopycnal mixing profile Ar(z), background AHH = 0.75 x 10^3 m2/sec. |
GM-0 | GM, k = 1 x 10^3 m2/sec, isopycnal mixing profile Ar(z). Zero AHH. |
GM-H | As in GM-0, only AHH is non-zero (0.75 x 10^3 m2/sec). |
GM=ISO | As in GM-0, only Ar = k = 1 x 10^3 m2/sec. |
GM=0.5 | As in GM-0, only k = 0.5 x 10^3 m2/sec. |
Ventilation in the Southern Ocean
The meridional overturn of water masses in the Southern Ocean is limited to
mode and intermediate water formation north of the polar front and bottom water formation over the
Antarctic continental shelf.
The circulation near 60 S is dominated by wind-driven upwelling of old Circumpolar
Deep Water (CDW). Mean profiles of water-mass age simulated in the Southern Ocean at
the latitude band 55 S-70 S are shown here (0.02 Mb)
. The non-GM runs exhibit spuriously
young deep Southern Ocean water, consistent with previous studies using CFCs. In contrast, the
GM runs have dramatically older CDW. It turns out that the GM runs have a distinct concentration
of young water in only a few model grid points adjacent to Antarctica. This has been cited as one
of the successes of the GM parameterisation; it permits relatively confined downslope flows
without spurious diapycnal dilution. However, the time-scale for downslope flow of AABW is
somewhat too slow under GM. This is demonstrated in panel (b), which compares
the minimum
age of AABW in the Weddell Sea in ISO and two of the GM runs (0.02 Mb)(GM=0.5 and GM=ISO, the
most and least rapidly ventilated GM cases, respectively). Bottom waters in ISO are as young as
23 years, whereas under GM AABW is at least 400 years old, substantially older than estimated
renewal time-scales for this water mass.
Convective mixed layer formation
This diagram shows (a) September mixed layer depth (m) calculated from the
Levitus (1982) climatology. (b)-(d) Maximum depth of wintertime
surface convective overturn (m) in experiments HOR-Z, ISO, and GM-0
(0.03 Mb) [A postscript version (2.7Mb)
is also available]. Convection patterns in the other HOR
and GM experiments are largely similar to those respectively shown in a,c. The HOR and ISO cases both show extensive
deep convection in the Southern Ocean. This is due to unrealistically weak stratification in the upper 1500-m of the
water column at 55 S to 70 S, a consequence of insufficient density of the model-equivalent of CDW. This widespread
deep convection conflicts with observations, which show upwelling of old CDW and shallow surface mixed layers at
these latitudes. Even during winter, surface mixed layers are typically only 50-100m thick at these latitudes. Unlike the
HOR and ISO cases, convection in all GM experiments is largely limited to the Antarctic shelf and north of the polar
frontal zone; both thought to be regions of convective mixed layer formation.
Observed and modelled radiocarbon
Basinwide GEOSECS observations and mean model profiles of radiocarbon (14C) in the
(a) North Atlantic (0 -70 N) (0.01 Mb),
(b) Indian Ocean north of the Equator (0.01 Mb),
(c) North Pacific Ocean (0 -70 N) (0.01 Mb), and
(d) the Southern Ocean at the latitude
band 55 S-70 S (0.01 Mb). Observations of 14C suggest that the North Atlantic Ocean (0.01 Mb)
is relatively well-ventilated to great depth.
In contrast, all model experiments underestimate the depth of NADW penetration; with a clear delineation between upper
well-ventilated NADW and lower 14C-depleted waters (particularly under GM). Additional GM experiments with an
enhanced seasonal cycle of T-S and/or inclusion of the topographic stress parameterisation of Holloway (1992) did not
rectify the problem. In addition, experimental runs with a deepened ridge system in the far North Atlantic showed negligible
change in the depth of NADW overturn.
In the North Pacific Ocean (0.01 Mb),
the simulated ventilation time-scales vary greatly between the GM and non-GM runs; typical
mid-depth North Pacific Deep Water (NPDW) is only 900 years in ISO and 1500-1800 years in the GM experiments.
This is reflected in dramatically different simulations of radiocarbon in NPDW; about -180 ppt in ISO and -260 to
-280 ppt in the GM runs (compared with -230 ppt to -250 ppt in observations). Similar trends can be seen in the
Indian
Ocean (0.01 Mb). Overall, the non-GM runs significantly underestimate the ventilation time-scales for deep water renewal in the
Pacific and Indian Oceans, whereas the GM runs significantly overestimate these time-scales.
In the Southern Ocean (0.01 Mb),
the model equivalent of CDW is too young in the HOR and ISO cases. The GEOSECS measurements
of radiocarbon indicate a near-uniform Southern Ocean value of -160 ppt, whereas the HOR and ISO cases
simulate radiocarbon to be only depleted to about -120 ppt. This is due to rapid overturn of 14C-rich surface waters at this
latitude band. In contrast, all GM cases overestimate the radiocarbon depletion of CDW, indicating that this water mass
is erroneously old under GM. This is particularly the case in the runs with k = 1.0 x 10^3 m2sec-1, where deep 14C values
approach -220 ppt. Slow downslope flows and weak interior currents under GM explain these spuriously depleted levels
of 14C.
Observed/modelled overturn and density
Oxygen depletion in the deep ocean indicates older water masses that have not been exposed to the
sea surface for some time. This can be compared qualitatively with simulated age in the model
experiments. Here we show the
observed latitude-depth section of dissolved oxygen (mL/L) and
potential density (kg/m3) at 180 W (top panel). For comparison, we include (b) ISO age (years)
and potential density at 180 W, and (c) as in (b), but for experiment GM-0
(0.75 Mb). The 180 W section
exemplifies typical density, oxygen, and water-mass age in the Southern Ocean in both observations
and the model runs.
Surface water overturn (0.03 Mb) is clearly stronger
in ISO than in GM in the Southern Ocean. This is due to
excessive convective mixed layer formation and increased meridional overturn under ISO. In
addition, spuriously steep isopycnal surfaces under ISO provide an efficient pathway for downward
mixing of surface waters near 60 S. In contrast, GM has quite realistically sloped density
surfaces in the region. This is due in part to the GM isopycnal thickness diffusion term which acts
to flatten isopycnal surfaces. The flatter isopycnal surfaces under GM result in a reduction in
interior geostrophic currents, which partly explains the reduced overturning rates in these runs.
Conclusions
A schematic overview is shown of the major water mass
renewal processes and age distributions in
the HOR, ISO and GM runs (0.05 Mb). Only the Southern, Pacific and Atlantic Oceans are shown. The
Indian Ocean behaves in a similar way to the Pacific, only with a reduced northern extent.
Very rapid ventilation of the interior takes place in the Southern Ocean in both HOR and ISO
owing to widespread convective overturn, strong AABW production, and in the case of ISO,
along-isopycnal mixing in a region of steeply sloping density surfaces. In contrast, under GM,
AABW production is weaker and confined to only a few model grid points, isopycnal surfaces
become flatter, and convection is all but shut down over the subpolar waters, yielding much older,
denser and 14C-depleted CDW.
In the deep Atlantic too much water of Southern Ocean origin ventilates the north in all model
experiments, yielding erroneously old and 14C-depleted Lower NADW (0.01 Mb).
NADW outflow is confined
to the upper 2500-m in ISO and HOR, and only the upper 1600-m under GM. In addition,
flattened isopycnal surfaces in the GM runs reduce the strength of interior geostrophic flows as
compared to ISO and HOR. This acts to reduce the speed of flow in the deep western boundary
current in an already viscous model with sluggish deep currents. This may not be a problem with
the GM scheme per se, but with our choice of the thickness diffusion parameter k, which may be
more appropriate for the upper Southern Ocean than for deep ocean flows. Nevertheless, none of
the model cases considered reproduces global ocean ventilation rates over centennial time-scales.
List of Figures
| Observed and modelled convection |
Observed oxygen vs. modelled Age |
Schematic of model ocean ventilation |
|
|
|
Model vs. observed radiocarbon distributions:
Modelled water-mass age distributions:
A version of this paper is in press with The Journal of Physical
Oceanography.
For a copy of the manuscript, please
e-mail M.England@unsw.EDU.AU
Return to
Matthew England's home page
WOCE Conference, Halifax, Canada, May 1998