New parameterization of atmospheric heating for simple models of relelvance to ENSO

Abstract

The El Niño Southern Oscillation (ENSO) phenomenon, which is one of the strongest signals of interannual variability in the tropics, arises out of strong interactions between the tropical ocean and the atmosphere. Investigations with coupled ocean-atmosphere models provide one of the ways to understand ENSO. However, present coupled models are found to have significant systematic errors (or climate drifts) in simulating this phenomenon. One of the reasons for this could be imperfections in the components of the coupled model, namely the oceanic component and the atmospheric component.

An investigation regarding the performance of the atmospheric components used in some of the present coupled models was undertaken. Both simple atmospheric models as well as complex Atmospheric General Circulation Models (AGCMs) were found to have serious problems in simulating the surface winds and their variability realistically. In the present study, we have undertaken to diagnose the problems with simple atmospheric models and to try to overcome them. One of the prominent problems with simple models has been the simulation of strong spurious easterlies over the eastern Pacific during certain phases of ENSO.

Two of the doubtful aspects of simple models are the assumption of linear dynamics and the way in which atmospheric forcing has been parameterized in terms of sea surface temperature anomalies (SSTAs). Most simple models employ Gill-type linear dynamics, and the atmospheric forcing is assumed to arise from the latent heat release associated with large-scale organized convection. Many studies have suggested that the assumption of linear dynamics is indeed a valid one, and that realistic simulation of surface winds is possible with a linear model if the atmospheric heating is prescribed correctly. Stated otherwise, these studies point out that the parameterization of atmospheric forcing in present simple models is not realistic.

We have critically examined the way in which large-scale organized (or deep) convection has been parameterized in simple models. Many of these models assume that anomalies in deep convection are more or less linearly dependent on SSTAs. As a consequence, the forcing in such models has the structure of the underlying SSTAs. A comparison of deep convection anomalies, parameterized in such a manner, with Outgoing Longwave Radiation (OLR) anomalies (OLR anomalies have proved to be a good proxy for anomalies in deep convection in the tropics) shows that the parameterized heating field differs significantly from the observed field. Another study, in which information about surface winds was fed into a simple linear atmospheric model to infer the forcing field, revealed that the inferred forcing differed significantly from the corresponding SSTA field but resembled the OLR anomaly field rather closely.

We are strongly critical of the assumption of the relationship between SSTAs and deep convection anomalies, based on the increasing amount of observational evidence for the existence of an SST threshold of 28°C for the onset of deep convection. The implication of this threshold is that when the SST is below the threshold, an SSTA would have no effect on deep convection. Convection exists in the first place and can only modulate the deep convection above waters which have SST equal to or greater than the threshold.

We have proposed a new parameterization of deep convection, the novel feature of which is the explicit inclusion of an SST threshold of 28°C for deep convection. We find that the inclusion of the SST criterion has improved the representation of deep convection quite significantly. The parameterized forcing field is now very different from the one in conventional simple models. Another encouraging finding was that this field now closely resembles the OLR anomaly field. We also conducted a comparative study of the variability in the model forcing field with that in the observed OLR anomaly field. The Empirical Orthogonal Function (EOF) analysis was employed for this purpose. This analysis showed that the new parameterization is able to capture well the large-scale features of the variability in the observations. This also made us confident that we would be able to reproduce the large-scale features of the variability of the observed surface wind field with our simple model. This is important because investigations had revealed that it is the large-scale features of the surface wind variability that are crucial for the interannual variability in the oceans.

Our atmospheric model employs simple Gill-type linear dynamics, similar to existing simple models of the tropical atmosphere. The only difference is the new parameterization of deep convection that we have proposed. The model parameterized the heating field for each of the 168 months from January 1970 to December 1987. The model-simulated surface winds were found to bear good resemblance to the observed winds. We also find that the problem of spurious easterlies has been reduced considerably in strength. Further analysis on the observed and simulated surface wind anomalies showed that our model, with the new parameterization, was able to simulate the large-scale part of the observed variability quite well. In fact, it was found to closely resemble, in this respect, results from a similar analysis using a complex AGCM.