There is no clear separation between the wind-driven circulation and the THC (see Section 7.3.6) because they interact with each other on several time-scales. While there is a great deal of empirical evidence that the ocean and sea surface temperatures co-vary with the atmosphere, this may only indicate that the atmosphere forces the ocean, and it does not necessarily signify a feedback or a truly coupled process that contributes to the variability. Moreover, it is very difficult to establish such coupling from observational studies. This topic was not dealt with thoroughly by the SAR. In the tropics there is clear evidence of the ocean forcing the atmosphere, such as in El Niño (see Section 7.6.5). In the extra-tropics much of what can be seen is accountable through fairly random wind variations; essentially stochastic forcing of the ocean is converted into low frequency ocean variability and gives a red spectrum in oceanic temperatures and currents up to the decadal time-scale (Hasselmann, 1976; Hall and Manabe, 1997). Feedback to the atmosphere is not involved. In the spatial resonance concept (Frankignoul and Reynolds, 1983) there is still no feedback from the ocean to the atmosphere, but oceanic quantities may exhibit a spectral peak through an advective time-scale (Saravanan and McWilliams, 1998) or Rossby wave dynamics time-scale (Weng and Neelin, 1998). In coupled air-sea modes, such as those proposed by Latif and Barnett (1996) for the North Pacific and by Groetzner et al. (1998) for the North Atlantic, there is a feedback from the ocean to the atmosphere. Spectral peaks are found in both the ocean and the atmosphere, and the period of the oscillation is basically determined by the adjustment time of the sub-tropical gyre to changes in the wind stress curl.
Coupled models indicate that, in mid-latitudes, the predominant process is the atmosphere driving the ocean as seen by the surface fluxes and as observed, yet when an atmospheric model is run with specified SSTs, the fluxes are reversed in sign, showing the forcing of the atmosphere from the now infinite heat capacity of the ocean (implied by specified SSTs). Recent ensemble results (Rodwell et al., 1999; Mehta et al., 2000) have been able to reproduce the decadal North Atlantic atmospheric variations from observed SSTs but with much reduced amplitude. Bretherton and Battisti (2000) suggest that this is consistent with a predominant stochastic driving of the ocean by the atmosphere with some modest feedback on the atmosphere, and that the signal only emerges through ensemble averaging.
In the extra-tropics, a key question remains the sensitivity of the mid-latitude atmosphere to surface forcing from sea ice and sea surface temperature anomalies. Different modelling studies with similar surface conditions yield contradictory results (e.g., Robertson et al., 2000a,b). The crude treatment of processes involving sea ice, oceanic convection, internal ocean mixing and eddy-induced transports and the coarse resolution of most coupled climate models, adds considerably to the uncertainty.
Other reports in this collection