Open ocean convection occurs every winter at high latitudes when buoyancy loss at the sea surface causes the surface layer to become denser than the water below, and results in highly variable mixing depth as a function of space and time (see recent review by Marshall and Schott, 1999). Convection directly affects the SST locally, and on larger scales indirectly through its effect on water mass properties and circulation. The maximum depth of convection occurs at the end of the cooling season, and depends on the balance between the cumulative air-sea fluxes, including ice-melt and precipitation, and the oceanic advection of buoyancy. During the summer, shallow warm surface mixed layers isolate the newly formed deep water from the atmosphere and mean currents and mesoscale eddies steadily transfer the newly formed deep water into the abyssal ocean.
Deep convective mixing is an essential ingredient of the THC, in particular in the North Atlantic, and is thus important for climate problems. It constitutes a very efficient vertical transfer process, and only a few small regions are needed to offset the slow diffusive buoyancy gain due to vertical (diapycnal) mixing (see also Section 7.3.3; e.g., Winton, 1995). Major sites of known open ocean deep convection are the centre of the Greenland Sea (Schott et al., 1993; Visbeck et al., 1995), Labrador Sea (LabSea Group, 1998) and a small region in the north-western Mediterranean Sea (Schott et al., 1996). However, only the Labrador Sea is in direct contact with the NADW which replenishes the deep waters of the Atlantic, Pacific and Indian Oceans. In the Greenland Sea deep and bottom waters remain local to the Arctic Ocean and deep basins north of Iceland and have no direct influence on the Denmark Strait Overflow (Mauritzen, 1996). Deep water formed in the Mediterranean Sea also never directly outflows through the Strait of Gibraltar. Dense water also forms on the shelf where convection can reach to the sea floor producing a well-mixed layer of dense water. Several mechanisms, such as eddies, flow over canyons and time-varying shelf break fronts, allow the dense water to enter into the deep ocean via descending plumes. This shelf convection' is believed to be fairly widespread around the Antarctic continent and is probably the primary mechanism by which AABW is formed (Orsi et al., 1999). AABW is the densest bottom water and penetrates into all of the three major oceans.
The overall effect of open ocean convection is usually parametrized through simple convective-adjustment schemes (Marotzke, 1991) which have been found to work well (Klinger et al., 1996). More advanced schemes have a somewhat increased performance (Paluszkiewicz and Romea, 1997). Lateral exchange between the deep convective centres and the surrounding boundary currents can significantly alter the convective process (Maxworthy, 1997). Within those small regions, deep mixing is affected by mesoscale eddies in two ways: cyclonic eddies provide an additional preconditioning (Legg et al., 1998), and collectively exchange fluid with the periphery of the deep mixed region. Coarse and medium resolution ocean models have shown significantly improved simulations of the deep convective regions when more sophisticated parametrizations of mesoscale eddies were employed (Danabasoglu and McWilliams, 1995; Visbeck et al., 1997). In particular the unrealistic widespread convective mixing (>500 m) over much of the Southern Ocean was significantly reduced.
Shelf plume convection is more difficult to represent in coarse resolution climate models. The problem is challenging because shelf convection is heavily influenced by the details of the bathymetry, coastal fronts and mesoscale eddies (Gwarkiewicz and Chapman, 1995), as well as entrainment of the ambient water (Baringer and Price, 1997). Several different attempts have been made to parametrize its overall effect (e.g., Beckmann and Döscher, 1997; Killworth and Edwards, 1999). Most ocean models have, however, not yet implemented such schemes and ventilate, e.g., the Southern Ocean by means of extensive open ocean (polynya) convection. The effect on the sensitivity of the current coupled climate models to forcings which involve changes in the convective processes is not known.
The convective activity in the Greenland and Labrador Seas varies inversely on decadal time-scales (see Chapter 2, Section 184.108.40.206). The switch of the convective activity from the Greenland Sea to the Labrador Sea has been attributed to changes in the index of the North Atlantic Oscillation (Dickson et al., 1996). The magnitude of the corresponding change in THC intensity is, however, controversial. While model results suggest moderate fluctuations (10 to 15%), it has been claimed from analysis of hydrographic observations that these changes might be much larger, with heat transport changes of more than 0.3 PW at 48°N, and in excess of 0.5 PW at 36°N (Koltermann et al., 1999).
In summary, the representation of oceanic convection in current climate models is satisfactory to simulate the convection changes observed over the last decades. It is, however, not certain that current schemes will work equally well in situations that involve substantial changes in the THC.
Other reports in this collection