Driving the Ocean's Overturning: Analysis with Adjoint Models
The focus of this thesis is the sensitivity of the strength of the meridional overturning circulation to surface forcing and mixing on climatological time scales. An adjoint model is used to gain new insights into the spatial characteristics of the sensitivity patterns.
Adjoint models provide the sensitivity of a diagnostic, often called cost function, to all model parameters in a single integration. In contrast, traditional sensitivity analyses are performed by repeated integrations of the so-called "forward" model, perturbing slightly the value of a single parameter at each integration. The results of the adjoint model allows us to calculate global maps of sensitivity. These maps provide a geographic picture of where on the ocean heat and freshwater flux, wind stress and diapycnal mixing perturbations have the greatest impact on the meridional overturning and its heat transport.
The adjoint model provides clear identification of the physical mechanisms which can influence the meridional overturning on times scales of years to decades. Boundary and equatorial Kelvin waves and equatorially trapped Rossby waves carry information around the boundaries of the basin and across the equator in less than a decade for a basin of the size of the Atlantic. Advection of buoyancy perturbations has an important influence on the meridional overturning on the decadal time scale. Diffusion is important in determining the final equilibrated state of the meridional overturning on the centennial scale.
The role of diapycnal mixing in determining the overturning’s strength is confined to the regions near the lateral boundaries in the Northern hemisphere and to the tropical region in both hemispheres. The important role played by the tropics in setting the overturning’s strength seems to confirm the thermodynamic principles outlined by Sandström (1908), Jeffreys (1925) and Munk and Wunsch (1998): upward advection of heat is balanced by downward diffusion. The strength of the meridional is then determined by the power available to return the fluid to the surface across the ocean’s stratification. Because the ocean is most strongly stratified in the tropics, the mixing process is most efficient in that region. Along the eastern boundary in the extratropics, the importance of diapycnal mixing is confined to a shallow layer at the base of the thermocline. The large vertical temperature contrast between the western and deep western boundary currents induces efficient mixing in that region. Surface wind stress has two effects on the ocean’s stratification which concentrate the sensitivity in the eastern equatorial region. Ekman suction increases the stratification along the equator while Ekman pumping decreases it in the rest of the tropics. The equatorial easterlies lift the thermocline on the eastern side of the basin, further increasing the stratification and the efficiency of the vertical mixing process in that region. These processes are similar in the results from a coupled model. Atmospheric feedbacks do, however, allow vertical mixing in the Pacific to play a role as important as mixing in the Atlantic in determining the overturning’s strength. The large uncertainties in the global value of the diapycnal mixing in the ocean, estimated here at κv = 3·10-5 ± 2 ·10-5m2s-1, translate into an uncertainty of approximately 6 Sv in the maximum value of the meridional overturning streamfunction.
The role of surface buoyancy forcing on the overturning’s strength depends on the formulation of the surface boundary conditions. The sensitivities are confined to high latitudes and the vicinity of convection sites when the surface forcing is prescribed as restoring the sea surface salinity or temperature towards observations. When the forcing is prescribed as a flux of heat or freshwater, advection allows buoyancy perturbations in the Atlantic basin to play an important role in determining the evolution of the meridional overturning. For annual and decadal time scales, heat flux perturbations in the North Atlantic are likely to have the greatest impact on the meridional overturning. On climatological time scales, it is the uncertainty in the precipitation and evaporation fields in the tropics which have the greatest impact on the uncertainty in the streamfunction, the latter can be estimated at: ψMAX = 29 ± 4 Sv. Over the intermediate time scale of climate change, the overturning is likely to weaken at first because of warming and freshening in high latitudes. It will, however, eventually recover as positive salinity anomalies are advected northwards from the tropics.
The sensitivity of the overturning to the wind stress forcing is also dependent on the surface boundary conditions. Under restoring boundary conditions, large positive sensitivities are observed in the Antarctic Circumpolar Channel in a pattern reminiscent of the so-called Drake Passage effect. According to that hypothesis, upwelling of North Atlantic Deep Water takes place predominantly in a branch of the Deacon cell in the Drake Passage region. The importance of wind in the Drake Passage vanishes when the surface buoyancy fields are less tightly constrained, for example in the model forced by mixed boundary conditions or in the coupled model. The Agulhas Plateau, the Chilean coastline and the Indonesian throughflow play an important role in setting the overturning’s strength in the ocean model forced by mixed boundary conditions. These "gateways" act as a regulator of the salinity of the Atlantic basin. The wind stress determines the balance between the inflow of relatively salty Indian Ocean water through the Agulhas current, the inflow of fresher Benguela current water southwest of Africa and the flow of very cold and fresh water through the Drake Passage. A wind stress of perturbation of ±0.03 N m-2 over the Agulhas Plateau would have a significant impact on the meridional streamfunction’s maximum, estimated at ψMAX = 29 ± 0.5 Sv. Both Drake Passage and gateway effects disappear almost completely in the coupled version of the model, which shows the strongest positive sensitivities to wind stress in the region of equatorial Ekman upwelling.
Our study shows that, in a climatological ocean model, the choice of air-sea boundary conditions is crucial in determining the sensitivity of the meridional overturning circulation. The climatology of the forward ocean model is credible and quite similar in all scenarios. However, including interactive atmospheric transports of heat and moisture changes the manner in which the ocean model state adjusts in wind stress, heat flux and diapycnal mixing. Considering the role of both the atmosphere and the ocean when studying the climatological behavior of the MOC is, therefore, clearly important. Models which keep one of the components fixed can lead to a very different conclusions from models in which both components are represented.