An idealized, three-dimensional model, of size comparable to the Atlantic, is used to study interdecadal variability of the thermohaline circulation (THC). In most of the model experiments, salinity is kept uniform and constant, the model being driven by surface heat flux only. When the model is driven by the surface heat flux diagnosed from a restoring spinup experiment, no oscillations occur. Driving the model by a time-independent, surface heat flux, obtained by applying a ''small'' zonal redistribution to the diagnosed flux, leads to strong interdecadal oscillations; ''small'' means that the modification to the diagnosed flux is within the error bars on estimates of surface heat flux based on observations. The model sea surface temperature (SST) anomalies are similar to the observed pattern of SST anomalies in the North Atlantic and to the SST anomalies associated with the interdecadal oscillation in the GFDL fully coupled ocean-atmosphere model. For redistributions that weaken the east-west variation of the flux, the mean THC, and the amplitude/period of the oscillation do not depend strongly on the amount of redistribution, once the threshold, beyond which oscillations occur, has been reached. If the east-west variation is enhanced, then the mean THC and the oscillation amplitude/period are very sensitive to the amount of redistribution. Coupling to a simple model of the atmosphere, it is found that redistributing the divergence of the atmospheric heat transport diagnosed from a spinup can lead to interdecadal oscillations. An experiment is included that incorporates freshwater flux, wind forcing, and idealized, non-flat bottom topography to show the robustness of our results. This case exhibits interdecadal variability in the vertically integrated transport of the model Gulf Stream. The transport variability is apparently driven by bottom pressure torque variations induced by the variable thermohaline circulation. Greatbatch et al, have suggested that the transport of the Gulf Stream was reduced in 1970-74 compared to 1955-59 and that the reduction in transport was driven by bottom pressure torque.
