Low-frequency atmospheric variability and seasonality of the mean circulation

Some studies have argued that basic attributes of low-frequency variability (LFV), like the wintertime centers of variance found over the northern oceanic basins and the existence of preferred low-frequency patterns, rely on organized forcing of the atmosphere by phenomena like El Nino. Other investigations have indicated that these attributes are largely a dynamical consequence of the structure of the mean circulation. As a means of sorting out these two possibilities, we consider seasonal changes in the attributes of LFV and determine whether they result from seasonality of forcing or mean state.

To accomplish our goal we first document various aspects of the seasonal cycle of LFV in nature, including the geographical distribution of its variance, scale and structure. One pertinent finding is that divergence of momentum fluxes associated with low-frequency anomalies tend to weaken the climatological waves, behavior often found in perturbations that are reacting to ambient gradients.

Next we examine the structure of LFV and its seasonality in ensemble integrations of a GCM forced with observed SSTs. We find that, even though they are forced by very different processes, the LFV of internally and externally generated anomalies has similar structure and seasonality, a further indication that forcing structure is not an essential contributor to low-frequency attributes.

Finally, to unambiguously test whether LFV would be organized and have seasonality even if its forcing had no preferred structures and were independent of season, we consider the low-frequency behavior in two classes of linear models. In both classes the models are driven by seasonally independent noise, so that any organization in the resulting perturbations cannot be attributed to the forcing. The first class consists of linear inverse models, one for each month of the year. Every aspect of LFV found in the GCM and observations is reproduced in this class, including its seasonality. The second class consists of the barotropic vorticity equation linearized about climatological, monthly mean states. Though not as accurate as the linear inverse model, these too produce LFV much like that observed. Most importantly, as the background states is changed to correspond to various seasons, the stochastically driven LFV changes in a manner like the observed seasonality. Analysis indicates that this occurs because the perturbations adjust to the mean state of each season in such a way that they a) are nearly stationary, b) can barotropically extract energy from the mean state, and c) have momentum fluxes that damp background gradients.