Seasonal and Interannual Variability in a Hybrid Coupled GCM

Abstract

A hybrid coupled model for the tropical Pacific ocean-atmosphere system is used to simulate El Niño Southern Oscillation (ENSO) interannual variability and to investigate the role of coupling in the seasonal cycle. An ocean GCM (OGCM) is coupled to an empirical atmospheric model that specifies a wind stress field from a given sea surface temperature (SST) field. The stress is estimated by singular value decomposition of the covariance between observed surface wind stress and SST fluctuations. Two versions of the atmospheric model are employed: one includes only spatial patterns of the atmospheric feedbacks associated with interannual variability, whereas the other also includes spatial patterns associated with the annual cycle. In the latter version, wind stress coupling in the seasonal cycle is modeled on the same basis as in the interannual variability. The seasonal cycle enters through prescribed beat flux and is modified by momentum-flux feedbacks. In the OGCM, two vertical mixing schemes-Philander-Pacanowski (PP) and a modified scheme-are used. Simulated ENSO anomalies have a reasonable spatial structure compared to observations, and the form is not strongly sensitive to the atmospheric model or mixing scheme. SST anomalies evolve largely as a standing oscillation, though with some westward propagation; heat content evolution is characteristic of subsurface memory, consistent with a mixed SST-ocean dynamics mode regime. In the absence of the seasonal cycle, the ENSO period is affected by vertical mixing: about 2.3 years for the modified scheme and slightly less than 2 years for the PP scheme. Indications of irregular or multifrequency behavior are also found. Interaction with the seasonal cycle frequency locks the interannual signal to a quasi-biennial period. The seasonal cycle in the eastern Pacific is well simulated by the coupled model. Wind stress feedbacks are an important part of the cycle near the equator but are not the sole factor in producing westward propagation along the equator. The seasonal cycle in the western Pacific shows great sensitivity to the mixing scheme. With the PP scheme, small errors in the uncoupled simulation are amplified by coupling; with the modified scheme, great improvements are obtained. These differences also provide an example of nonlinear interaction between ENSO and the coupled seasonal cycle. With the PP scheme, the amplitude of the ENSO signal increases with coupling, but at strong coupling competition with ENSO can decrease the amplitude of the seasonal cycle in the cold tongue region. However, with the modified scheme, although the irregularity of interannual variability is increased, stronger coupling does not affect the amplitude of the coupled seasonal cycle in equatorial SST. Simulating the seasonal cycle on the same basis as interannual variability thus provides much stronger constraints on subgrid-scale parameterizations than simulating ENSO alone.