Air–Sea Feedback in the North Atlantic and Surface Boundary Conditions for Ocean Models

Extratropical sea surface temperature (SST) and surface turbulent heat flux monthly anomalies in the central and eastern part of the North Atlantic are considered for the period 1952–92 on a 5° × 5° grid. In this region where the mean surface current is small, the SST anomalies are well simulated by a simple one-dimensional mixed layer model that is stochastically forced by the day-to-day changes in the local air–sea fluxes. A statistical signature of the stochastic model is that the cross correlation between surface heat flux and SST anomalies changes sign between negative and positive lags when the heat flux feedback is negative. This is observed at each grid point of the domain for the turbulent heat flux, which thus contributes both to generating the midlatitude SST anomalies and to damping them, once they are generated. Using properties of the lag covariance between SST and heat flux anomalies, the turbulent heat flux feedback is estimated from the observations. It averages to about 20 W m−2K−1in the investigated domain, increasing toward the northwest and the northeast and decreasing southward. It also varies seasonally, being generally largest in the fall and smaller and more uniform in summer. There is no indication that it can become significantly positive. A negative turbulent heat flux feedback is also suggested by the lag relation between the dominant modes of SST and turbulent heat flux variability over the whole North Atlantic, and it is found that the spatial patterns of the associated SST and turbulent heat flux anomalies are remarkably similar whether the atmosphere leads or lags, with only a change of heat flux sign between lead and lag situations. This analysis provides some observational support for the use on short timescales of a restoring condition for SST in ocean-only simulations, but the coupling coefficient should be weaker than usually assumed and a function of latitude and season. The associated SST–evaporation feedback has little effect on interannual surface salinity changes. It should be significant on longer timescales, but then the restoring temperature should be allowed to vary and nonlocal influences should be considered.