A 21/2-layer ocean model is developed to investigate the role of the first two baroclinic modes in determining the interannual variations of the sea surface temperature (SST) associated with the El Niño–Southern Oscillation (ENSO) phenomenon. Rather than simply adding an additional mode to the ocean component of the Zebiak–Cane coupled atmosphere–ocean model, it proved necessary to completely rethink all parts of the model. This allowed the external parameters to be specified more realistically. For example, the drag coefficient used in calculating the surface wind stress in the model is now consistent with that empirically derived, and the temperature of the water entrained in the surface layer that affects SST is now more carefully parameterized.
When forced by observed wind stress anomalies for 1961–93, the ocean model reproduces the interannual variations of SST satisfactorily. The quantitative discrepancies between the model hindcast and observed SST anomalies are limited to an excessive cooling of 0.5–1°C in the eastern/central Pacific during the period of 1989 to early 1991, and weaker warm phases in the central/western Pacific than observed. Both of the two gravest baroclinic modes are shown to be important in affecting the interannual variability in SST. A critique of the ocean model is presented at the end of this work.
When the ocean model is coupled with a simple atmosphere model, the resulting model exhibits quasi-periodic ENSO cycles with a period of ∼5 years. The variability in the coupled model is sensitive to the strength of the coupling and to the model parameterization of subsurface temperature. This model provides an opportunity to gain a better insight into the instability and variability of large-scale, low-frequency phenomena in the coupled atmosphere–ocean climate system and to bridge the gap between the simple Zebiak–Cane model and the more complex and computationally intensive coupled general circulation models in which more vertical modes are present.