The water mass structure in the deep ocean is largely dominated by the flows across a few shallow sills and through straits. The role of such flows for climate change arises from their influence on the THC which in turn can affect surface conditions. Once water has crossed a sill, it descends the continental slope as a dense gravity current and provides the water source for the downstream basin. The water mass properties are determined by the entrainment of and mixing with ambient water (Baringer and Price, 1997). For the NADW which is at the heart of the sinking branch of the global conveyor, the overflow of cold water across the Greenland-Iceland-Scotland ridge is the principal source. Observations show intense flow with speeds up to 1 m/s, in a layer of less than 100 m thickness above the bottom and within 20 km of the continental slope. Model calculations suggest that an interruption of this overflow would lead to a breakdown of the Atlantic THC and the associated heat transport within less than a decade (Döscher et al., 1994). In the past decades the overflow appears to have been fairly steady. Another prominent feature is the Indonesian Throughflow which has substantial contributions to the variability of the THC and heat transport in the Indian Ocean (Godfrey, 1996).
In coarse and medium resolution climate models, the overflow across the Greenland-Iceland-Scotland ridge has been found to be highly sensitive to the precise geometry used. For example, changes in topography by as little as one grid cell resulted in gross changes not only to the amount of cross-ridge flux but also to its location, and to the composition of the water mass actually crossing the sill. Thus, a 50% change in heat flux at the Greenland-Iceland-Scotland ridge latitude could be achieved in a model by the addition or subtraction of a single grid box (Roberts and Wood, 1997). Even eddy-permitting simulations give flows through sills which are very sensitive to model details (Willebrand et al., 2001). The physical processes involve hydraulic control and are not properly represented in climate models, and it is unclear whether a situation with substantial overflow changes can be modelled correctly. Parametrizations for the flux across a sill are only available for the simplest process models (Killworth, 1994; Pratt and Chechelnitsky, 1997) that do not include mixing or unsteadiness, both of which are known to be important at sills (Spall and Price, 1998).
Climate models which are based on depth co-ordinates obtain far too much mixing near a sill. This is caused by both poor mixing parametrizations in such models and by excessive diapycnal mixing resulting from the staircase'-like representation of bottom topography. As a result, water mass structure downstream of a sill is poorly represented in climate models. Isopycnic models are free of this erroneous mixing, and addition of Richardson number dependent mixing to such models results in realistic tongues of dense water moving downslope, mixing at a rate consistent with observations (Hallberg, 2000). A promising new development is the explicit description of the turbulent bottom boundary layer in climate models which yields more realistic flow of dense water down slopes (Beckmann and Döscher, 1997; Killworth and Edwards, 1999).
In summary, the uncertainties in the representation of the flows across the Greenland-Iceland-Scotland ridge limit the ability of models to simulate situations that involve a rapid change in the THC.
Other reports in this collection