In this paper, we report on solutions to a coupled model of intermediate complexity that is confined to the Paciflc basin. For reasonable parameter choices, solutions develop two dynamically distinct oscillations: an ENSO-like interannual signal and a decadal one. The decadal oscillation extends throughout the tropics, and is similar in structure to the observed and modelled variability noted above. Consistent with Nakamura and Yamagata's (1998) second mode it is generated by ocean-atmosphere interactions within the subtropics, not the tropics. It also affects the ocean's northern Subtropical Cell (STC), a shallow, meridional overturning circulation in which water flows out of the tropics in the surface layer, subducts in the North Pacific subtropical gyre, returns to the tropics within the thermocline, and upwells in the eastern equatorial ocean. Variations in the strength of its surface branch efficiently transmit the subtropical decadal signal to the equatorial region, thereby providing an oceanic teleconnection for explaining the linkage noted by Kleeman and Power (1998) and others.
2 MODELE
A useful property of the ocean model is that the process that drives the northern and southern STCs is clearly represented in solutions forced by steady winds (McCreary and Lu, 1994; Lu et al., 1998). Specifically, STC strength is determined by the divergence of meridional transport across particular latitudes equatorward of which subduction is assumed to vanish, inthis case yds = 23｡? and ydn = 23｡?; this divergence drains warm surface (layer-1) water from the tropics, and the resulting mass loss is balanced by upwelling of cool thermocline (layer-2) water along the equator and by convergence of layer-2 water across yds and ydn. In addition, the divergence results almost entirely from the Ekman transports across ydn and yds, so that STC strength is set primarily by the zonal wind stress at these latitudes. Thus, the model contains an explicit process by which subtropical winds affect the amount of thermocline water that upwells along the equator. When the wind oscillates at decadal periods, the balance among these quantities does not hold precisely because of transients associated with slowly propagating Rossby waves. Nevertheless, the relationship between upper-layer divergence and equatorial upwelling is similar to the steady case.
The atmospheric component consists of two linked parts. The first part generates wind-stress anomaies based on SST anomalies from the ocean component. It does this using joint modes of variability between observed SST and wind-stress anomalies, determined from GISST (Parker et al., 1995 and COADS wind-stress (da Silva et al., 1994) data using the statistical technique of singular value decomposition (Kleeman and Power, 1998; Syu et al., 1995). At each time step of the integration, the solution's SST anomaly field is decomposed into a spectrum of these these joint models (called SVDs), and the wind-stress anomaly field associated with the SVDs is then used to drive the ocean component. Only the first 7 SVDs are included in the model atmosphere, the spatial structures of the higher-order SVDs resembling noise. The second part is a simplified model of the atmospheric boundary layer that generates a latent-heat flux anomaly based on SST and wind-speed anomalies, the latter derived from the wind-stress anomalies calculated in the first part (Kleeman and Power, 1995).
