Effect of Wind Changes During the Last Glacial Maximum on the Circulation in the Southern Ocean

Present-day surface wind stress climatology is manipulated to simulate wind conditions during the last glacial maximum. These estimated wind fields force a onelayer, wind-driven numerical model of the southern ocean to determine if a change in the strength of the surface wind stress can shift the location of the Antarctic Polar Front, which is part of the Antarctic Circumpolar Current. A change in the forcing by a factor of 0.5-2.0 results in a change in the speed of the flow by an identical factor with no change in position. However, if the present-day wind climatology is shifted meridionally, there is a change in both strength of the circulation and spatial pattern. A shift of the wind stress of more than 5 o of latitude is required to produce a shift in the location of the polar front.

shifted toward the equator by at least 4 o to 50 of latitude [CLIMAP, 1976;Dow, 1987].A review of conditions during the LGM is presented by Crowley [1988].
There is a strong temptation to link these facts and attribute the shift in the polar front to the increased wind stress.A possible mechanism for the shift would be the increase in the Ekman transport due to the stronger surface stress which would cause an increased advection toward the equator.However, current understanding of the dynamics of the Antarctic Circumpolar Current (ACC) is that the position of the current is determined by the bottom topography and will tend to follow isopleths of topographic vorticity f/h, which is the ratio of the Coriolis parameter and the local mean depth [Johnson and Hill, 1975].From this point of view, increased surface stress should produce stronger circulation and, perhaps, increased production of mesoscale eddies from instability but not a shift of the location of the current.
The purpose of this study is to consider the question of the effect of changes in the surface wind stress that may have occurred during the LGM on the location and strength of the circulation in the southern ocean.The magnitude of these changes is estimated from a one-layer, wind-driven model for the southern ocean which is driven by several choices of wind stress.For each of these wind stress fields, the model is driven to steady state, and the steady pattern is analyzed fo• changes in the strength and position of the ACC.
The results of a series of simulations are presented.Several cases are considered that are forced with the zonal average of the present-day surface wind stress that is changed by a constant factor, representing an increase (or decrease) of wind stress without a change in the general character of the winds.A second series of simulations considers the effect of meridional shifts in the surface wind stress.
The shifted winds might also represent situations where the wind strength during earlier times changed by different amounts at different latitudes.These simulations are more speculative in that there is no compelling evidence that there was a meridional displacement of the wind field during the LGM.However, this is another mechanism that might cause a lateral shift of the ACC.A final simulation considers a northward shift in the winds together with an increase in strength.

Numerical Model
The wind-driven southern ocean numerical model used in this study is based on the vertically averaged (shallow water) primitive equations on a/• plane with a free surface.
Lateral friction is the only explicitly included mechanism for dissipation of momentum.These equations are solved with an alternating direction implicit scheme [Leendertse, 1967]  The bottom topography, relative to the average depth over the model domain, is reduced to 15% of its true magnitude to account for effects due to stratification.That is, the bottom topography is constructed as the sum of a spatial mean depth H and a departure from the mean r/b; the total depth is h(x, y): H + e 7lb(x, y).The bottom reduction parameter e has a value of 0.15 for these cases.This reduction has the effect of changing the total transport and the general path of the ACC by changing the location of planetary vorticity f/h isopleths.In general, one cannot get a perfect match between transport and path by adjusting the topography, so the reduction parameter is chosen to produce a realistic path for the ACC, which is one that matches the dynamic topography of the surface relative to a deeper pressure level, say 1400 m.Simulations for two choices of the bottom reduction parameter along with the dynamic topography from the Southern Ocean Atlas [Gordon and Baker, 1986], can be seen in the work by Klinck [1992, Figures 5, 6 Previous experience with wind-iYrced circulation in the southern ocean shows that the currents are not strongly influenced by the zonal structure of the winds [Klinck, 1991].In fact, nearly identical circulation patterns are obtained from the one-layer model when forced by the annual average winds or the zonal average of the annual average winds.Because of this insensitivity to the zonal structure, we have chosen to use the zonal average of the annual average wind stress as the basic wind field (Figure 1).This field is shifted and scaled to create the estimated surface wind stress from previous climate regimes.50øS [Peterson and Whitworth, 1989].The rest of the ACC flows generally zonally, with a tendency to slip southward in response to the wind forcing.

Wind Strength Changes
Several simulations are forced by a zonally averaged wind stress that is created from present-day stress scaled by a factor ranging from 0.5 to 2.0 which represents the change in wind stress from other times relative to presentday conditions.An integrated view of the result of such changes in the wind stress is given by the transport of the ACC for each of these simulations (Figure 3).It is clear from these transports that the response is essentially linear, that is, the transport change is proportional to the wind factor.Furthermore, the resulting surface elevation patterns at steady state have exactly the same shape as Figure 2 but are changed by the same factor as the wind.
In all of these scaling calculations, including the experiments for values well outside the range predicted for conditions during the LGM, the location of the major features of the ACC do not change.The only notable difference is that the total surface elevation across the ACC and the transport are scaled by the same factor as the wind climatology.Even the area east of Drake Passage, which would be a place at which nonlinear effects would appear, responds linearly to the forcing change.a second between 55øS and 60øS.The differences among these cases is not so much that the current has shifted as that there are preferred locations for the current cores and the location of the wind has a strong effect on the strength of a given current core.

Wind Shift and Strength Change
The best estimate of wind structure during the LGM is that the winds both shifted to the north and increased in strength by 70% [Crowley and Parkinson, 1988].Since the increase in wind strength produces an increase in the circulation and the northward shift produces a decrease, there is a question of which effect is dominant if both occur.
A last simulation was run with the winds shifted north by 50 and the winds increased by 70%.The total transport for this case is 280 x l0 s m3s-l(indicated by the solid circle in Figures 3 and 4).This represents an overall decline in the total transport compared to the base case, so the northward shift of 50 has a stronger influence than the increase in wind strength by 70%.Note, however, that the location of the ACC, as represented by the surface elevation at 6.5øW (Figure 5), is north of the present-day ACC by 20 of latitude.

Changes in Wznd Strength
It is generally accepted that the winds were much stronger during the LGM [Crowley, 1988].In the model calculations discussed here, the southern ocean transport increases as the wind stress increases (Figure 3).Fossil records indicate an increased thermal gradient during the LGM [CLIMAP, 1976;Dow, 1987] which is in agreement with the increase in the velocity of the ACC that is found in this study.
The response of the ACC to increased forcing is a proportional increase in the volume transport with no change in ACC location.Thus this study indicates that the northward shift in the ACC during the LGM cannot be explained by simply increasing the surface wind stress to values predicted to exist during the LGM.However, this model neglects part of the dynamics that are active in the southern ocean, specifically, the effect of density variations.In a model that includes the effects of stratification, the stronger wind stress will drive an increase in circulation which will produce stronger density gradients and current shear (both horizontal and vertical).This increase in shear is expected to increase the energy sources for instabilities (barotropic and baroclinic), resulting in an increase in eddy production.The increased eddy activity is not expected to cause the location of the current to shift.However, the presence of stratification will decouple the flow, to some extent, from isobaths, allowing the current to change location.

Changes in Wind Position
The effect of changing the location of the zonally averaged winds is two-fold.The major effect is to move the boundaries of the subtropical and polar gyres by shifting The effect of density stratification is to insulate the flow from topography to some extent and thereby to break the constraint that the flow be along planetary vorticity f/h isopleths.However, the southern ocean is only weakly stratified, so the flow is affected by bottom topography much more than at subtropical latitudes where there is a strong pycnocline.Stratification can allow the circulation to shift location under different wind regimes if the density structure is changed.
The effect of ice cover is more direct in that it interferes with the transfer of momentum between the wind and the ocean.In places where the ice cover is solid, the ocean is shielded from the wind, while in regions with less than complete ice cover, the ice absorbs some of the momentum so that the circulation is not as strongly forced.
These processes are thought to be less important than using a staggered C grid[Mesinger and Arakawa, 1976] which handles the fast external gravity waves implicitly allowing a 1-hour time step.Complete details on the wind-driven southern ocean model, including model geometry and bottom topography, are given byKlinck [1991Klinck [ , 1992]].The model domain extends from 35øS to 75øS and is purposely chosen to extend far enough north so that the northern model boundary, which is free-slip, and dissipation parameters; only the surface wind stress is different.Each simulation is started from rest and integrated to a steady state at which time (about 6000 hours of simulation) the area-integrated potential and kinetic energy reach a steady value (

Fig. 1 .Fig. 2 .
Fig. 1.Zonal average of the eastward wind stress (newtons per square meter) from the climatology created from the European Centre for Medium-Range Weather Forecasting twice daily analyses [Trenberth et al., 1989].
Fig. 3.The transport (in units of 1 x 106 m 3 s -1) at steady state as a function of increase in wind strength.The average wind is multiplied by the factors 0.5, 0.75, 1.0, 1.25, 1.5, and 2.0 in constructing this figure (indicated by open circles).The solid circle is a simulation with wind increased by 70% and shifted north by 5 ø .
the lines at which the curl of the wind stress vanishes.The strength of the circulation in the southern ocean is determined by the relative location of the gyre boundaries and the continents which block the flow.As the wind pattern is shifted to the south, less of the circulation is blocked by the continent of South America and the transport of the ACC is larger.The minor effect of the wind stress shift is due to the fact that the strength of the flow is determined by the component of the surface stress that is tangential to isopleths of planetary vorticity f/h [Johnson and Hill, 1975].As the wind pattern is shifted, the strength of the circulation along each of these lines changes.These effects are the same, since the southward shift moves the forcing to planetary vorticity lines that are not blocked by continental walls, allowing much stronger circulation.The first effect of the wind shift is the change in transport in proportion to the amount of shift (Figure 4); the second effect is the shift in the location of the ACC (Figure 5).If the ocean had a fiat bottom and no continental boundaries, then the circulation in the channel would shift meridionally exactly in response to a change in the location of the wind forcing.If the continental boundaries are added but the bathymetry remains fiat, then some of the latitudes (all of which are parallel to planetary vorticity isopleths) would be blocked by land and others would not; the flow along the unblocked latitudes would be nmch stronger than the flow in the blocked latitudes.When the wind stress pattern is shifted, the transport in the ACC would increase or decrease depending on whether the strongest winds occurred over the unblocked latitudes or not.Finally, when the bottom topography is included, the planetary vorticity lines follow some path around the southern hemisphere.The circulation at any place depends on the projection of the wind stress onto the path and on whether the path is blocked by a land boundary.It is interesting to note that a shift in the winds of as little as a 20 can produce a 50 x 10 e m a s-• change in the total volume transport.The surface elevation in the South Atlantic (Figure 5) illustrates this last point in that the path of the circulation in the southern ocean is dictated by the topography, and this feature of the southern ocean does not change with different climatological regimes.However, if the location of the wind stress changes, then the forcing along each of the planetary vorticity lines changes, easiest to include as a mask over the onelayer model to represent the percentage of ice cover and thereby the reduction in the wind stress.The problem is then the estimation of the extent of ice in different climatic regimes and the parameterization of the effect of the ice on momentum transfer.Both of these problems with the ice effects are soluble but increase the number of processes being estimated in this model and are, therefore, not considered at this time.The inclusion of stratification is also possible, but it adds considerably to the cost of the model calculation and is not pursued at this time.SUMMARY The simple wind-driven model used in these calculations represents present-day southern ocean circulation with reasonable accuracy but fails to show that a change in the strength of the surface wind forcing, such as occured during the Last Glacial Maximum, produces a latitudinal displacement of the ACC.The model results support the conclusion that increased wind stress creates an increased transport of the ACC and hence an increase in the oceanic temperature gradients, as has been estimated from sediment cores.Meridional shifts of the wind stress offer a plausible mechanism to create the recorded shifts in the ACC positions.The basic mechanism is that the speed of flow along planetary vorticity isopleths is dependent on the wind strength tangential to the path.In order to create a shift in the location of the ACC, it is necessary to change the wind strength by different amounts in different locations.One way to do this is to shift the present-day structure of the wind stress.A second way is to have the surface wind stress change by different amounts at different latitudes (essentially the same result).Either of these mechanisms seems to be able to produce a shift, in the location of the ACC in response to changes in the surface wind stress.