An experiment in which surface wind stress data from the National Center for Atmospheric Research global circulation model (GCM) was used to drive a simple model of the tropical Pacific is described. First, a 15-year integration of the GCM was conducted, forcing the model with observed sea surface temperatures (SSTs) for the period 1958–72 (the FORCED case). A parallel integration was conducted using long-term monthly mean SSTs (the CONTROL case). All other boundary forcing in both GCM simulations was identical. Surface wind stress date from each GCM integration were then used to drive the Florida State University 1½ layer, reduced gravity model of the tropical Pacific. A separate integration of the ocean model was conducted using observed wind stress for the period 1961–72 (the OBSTRESS case).
The goal of this research was to evaluate the response of a sophisticated GCM to tropical Pacific SST variability in terms of the surface wind stress field, and to investigate the sensitivity of a simple wind-driven ocean model to differences between the simulated and observed wind stress data. These are important issues bearing on the potential for accurate modeling of the coupled ocean-atmosphere system over the tropical Pacific. In this paper the results from the ocean model simulations and observations are compared in terms of interannual variability. An earlier paper describes the response of the GCM tropical Pacific surface wind stress field to prescribed SSTs.
The results show that the GCM response to prescribed SSTs produced wind stress anomaly patterns over the tropical Pacific that qualitatively resemble those observed in association with extremes of the El Niño activity, particularly in the central equatorial ocean. These wind stress anomalies produced upper-layer thickness anomalies in the eastern ocean that bore some resembalance to those found in observations and the results of the OBSTRESS integration; i.e., simulated El Niños did occur. In general, however, the El Niño signal in the FORCED case was considerably lower in magnitude and was less organized than in the OBSTRESS simulation. Further, the episode-to-episode changes in magnitude did not agree well with those in the OBSTRESS integration. These results reflect not only important differences between the spatial character of the response of the observed and GCM surface wind stress fields to El Niño SST anomalies, but also the fact that the overall coupling between the GCM atmosphere and the tropical Pacific SST field is not as strong as observed in the real ocean–atmosphere system.
A second interesting result was that quasi-periodic oceanic variability in some ways resembling that associated with El Niñc variability in the OBSTRESS and FORCED experiments was clearly evident in the CONTROL case. Considerations of the response of the model ocean to temporally random atmospheric forcing with large spatial scales shows that such organized low frequency variability may arise from the excitation of preferred resonant frequencies defined by the Rossby wave dispersion relation. This finding may have implications concerning the maintenance and character of the El Niño activity.