domain and resolution of three CGCMS are summarized in Table 1. All of the three modelswere integrated for thirty or longer years and their model climatology and interannual variability were investigated (Nagai et al., 1992; Nagai et al., 1995; Mechoso et al., 1995).
Fig. 1 shows annual mean SSTs of the three models and the observation. All of the three models could produce a basic feature of the equatorial Pacific SST field, i.e., warm water in the western part and the cold water in the eastern part. The details are, however, different from each other. The SSTs in the warm pool in the PACIND and the GLOB are much higher than the observed, whereas those in the PAC are modest as the observed. Cold SSTs on the equator tends to penetrate in the western Pacific so that SSTs in the western Pacific are not so uniform as the observed. In the PAC and PACIND, the southeastern tip of the warm SST extends too far to the east, so that the cold tongue pattern is not simulated, whereas this is improved in the GLOB. The models can simulate some aspects of the seasonal variations over the tropical Pacific, however, several disagreements with the observation are found in common. For example, biannual variations are dominant in the SSTs in the eastern equatorial Pacific in the PAC and PACIND. This variability is weakened in the GLOB but is still recognized.
Temporal and spatial evolution of the model ENSOs are investigated with the EOF analysis of the SST, VAT (vertically averaged temperature for the upper 300m), and zonal wind stress. The leading modes of VAT for three models are shown in Fig. 2 with that for the OGCM stand-alone run. All of Fig. 2 indicate the pile-up pattern of the oceanic heat storage prior to El Nino. The spatial patterns of the SST and zonal wind stress in the models are modified slightly due to the difference of the model climatology with the observed one.
Figure 1.Annual mean SSTs for (a) the model PAC, (b) the model PACIND, (c) the model GLOB, and (d) the observation (1960-1990).
