102
ANNE-MARIE TREGUIER
finds a dramatic improvement in the representation of restratification
after convection in the Labrador Sea between a 18 km grid and a 3-4 km
grid (1/3 ◦ to 1/15 ◦ ). Certainly, a full GM parameterization would be
justified in the 1/3 ◦ model, in the Labrador Sea but not elsewhere. The
eddy fluxes in Chanut’s high resolution case correspond to GM coefficient of up to 800 m 2 .s −1 . If such a high value is used in the subtropical
gyre it destroys the eddy activity in the Gulf Stream. The spatially variable form proposed by Visbeck et al. (1997) does not help in that case,
because baroclinic instability growth rate is higher in the Gulf Stream
than in the Labrador Sea based on the resolved flow field. Studies are
under way to propose variants of the GM parameterization that would
”switch on” when needed.
Considering a case where the first Rossby radius is well resolved (say,
a 5 km grid where R 1 = 40 km), how should the dynamical effect of
sub-mesoscale eddies be parameterized? Is the unresolved part of the
spectrum mainly controlled by baroclinic instability? Roberts and Marshall (1998) advocate the use of a biharmonic GM parameterization,
based on their wish to eliminate diapycnal mixing in the surface layers.
However, this requirement may not be physically defensible, considering that eddies do perform diapycnal (horizontal) mixing across surface
fronts (Treguier et al., 1997). As was the case for in the two previous parameterizations we considered (isopycnal diffusivity and lateral
viscosity) we still lack observational evidence and theory to justify parameterizations of the submesoscale effects.
5.2
Other mesoscale eddy effects
Another dynamical effect of mesoscale eddies is the so-called ”Neptune” effect (see for a review Alvarez and Tintor´ e, 1998). In the presence
of bathymetry and β-effect, quasigeostrophic eddies have the tendency
to generate mean flows along f /H contours. This additional mean flow
must be forced by a parameterization when eddies are not represented
in a model. The problem is that we do not have enough knowledge of
the strength of this effect in realistic ocean circulations and neither do
we know the vertical structure of the generated mean flows. Certainly,
adding a parameterization forcing barotropic currents along f /H contours would help Atlantic models to improve the strength of their deep
western boundary currents. However, if the models do not represent the
overflow correctly, the Neptune parameterization could have the effect
of generating spurious transport of water with the wrong properties.
Finally, mesoscale eddies tend to exist as coherent structures that
carry water far from their generation region (well-known examples are
ANNE-MARIE TREGUIER
finds a dramatic improvement in the representation of restratification
after convection in the Labrador Sea between a 18 km grid and a 3-4 km
grid (1/3 ◦ to 1/15 ◦ ). Certainly, a full GM parameterization would be
justified in the 1/3 ◦ model, in the Labrador Sea but not elsewhere. The
eddy fluxes in Chanut’s high resolution case correspond to GM coefficient of up to 800 m 2 .s −1 . If such a high value is used in the subtropical
gyre it destroys the eddy activity in the Gulf Stream. The spatially variable form proposed by Visbeck et al. (1997) does not help in that case,
because baroclinic instability growth rate is higher in the Gulf Stream
than in the Labrador Sea based on the resolved flow field. Studies are
under way to propose variants of the GM parameterization that would
”switch on” when needed.
Considering a case where the first Rossby radius is well resolved (say,
a 5 km grid where R 1 = 40 km), how should the dynamical effect of
sub-mesoscale eddies be parameterized? Is the unresolved part of the
spectrum mainly controlled by baroclinic instability? Roberts and Marshall (1998) advocate the use of a biharmonic GM parameterization,
based on their wish to eliminate diapycnal mixing in the surface layers.
However, this requirement may not be physically defensible, considering that eddies do perform diapycnal (horizontal) mixing across surface
fronts (Treguier et al., 1997). As was the case for in the two previous parameterizations we considered (isopycnal diffusivity and lateral
viscosity) we still lack observational evidence and theory to justify parameterizations of the submesoscale effects.
5.2
Other mesoscale eddy effects
Another dynamical effect of mesoscale eddies is the so-called ”Neptune” effect (see for a review Alvarez and Tintor´ e, 1998). In the presence
of bathymetry and β-effect, quasigeostrophic eddies have the tendency
to generate mean flows along f /H contours. This additional mean flow
must be forced by a parameterization when eddies are not represented
in a model. The problem is that we do not have enough knowledge of
the strength of this effect in realistic ocean circulations and neither do
we know the vertical structure of the generated mean flows. Certainly,
adding a parameterization forcing barotropic currents along f /H contours would help Atlantic models to improve the strength of their deep
western boundary currents. However, if the models do not represent the
overflow correctly, the Neptune parameterization could have the effect
of generating spurious transport of water with the wrong properties.
Finally, mesoscale eddies tend to exist as coherent structures that
carry water far from their generation region (well-known examples are