Simulations of tropical surface sinds : Seasonal cycle and International variability

Abstract

Tropical climate variability is strongly influenced by coupled ocean-atmosphere interactions. The sea surface temperature (SST) influences the atmospheric heating through evaporation and convergence of moisture. The atmospheric heating drives surface winds, which in turn drive upper ocean currents leading to changes in the SST. Consequently, coupled ocean-atmosphere models of varying complexity are being increasingly used to understand and predict climate variability in the tropics on the annual and interannual time scales. The El Niño-Southern Oscillation (ENSO), being one of the most dramatic, energetic, and well-defined patterns of interannual variability in the tropics, has received the largest attention. While coupled general circulation models (CGCMs) have increasingly come in vogue, practical considerations such as climate drifts and their sensitivity to the parameterizations of physical processes in the component systems limit their usefulness. These problems of the CGCMs are related to deficiencies in the atmospheric component (the Atmospheric GCM or AGCM) in simulating the surface winds and in the oceanic component (the Oceanic GCM or OGCM) in simulating the SST. On the other hand, simpler coupled models, such as the Cane-Zebiak model, have proved their usefulness as prediction systems and have been employed to understand various aspects of ENSO. The simple coupled models are, however, not without their faults, and there is scope for improvement of the component systems. As the tropical surface winds are one of the most important components in the coupled ocean-atmosphere interactions of relevance to ENSO, we focus only on the surface stress in this study. The main objective of the thesis is to develop an improved simple atmospheric model, capable of simulating annual and interannual variations of tropical surface stress, for use in simple or hybrid coupled models. Another related problem undertaken involved the intercomparison of the simulation of the annual cycle and interannual variability of the tropical surface winds in a large number of AGCMs. The motivation for undertaking this project arose from the following considerations: (i) such a rigorous intercomparison has not been carried out in the past, (ii) it would provide a perspective to evaluate the simulations of simple models, (iii) it was also realized that such documentation of the strengths and drawbacks of AGCM simulations would provide useful feedback for subsequent improvements of these complex models, (iv) further, it was hoped that the examination of a large number of AGCMs with diverse resolutions and physical parameterizations would provide information about the sources of systematic errors in AGCM simulations. The data for this was adopted from the Atmospheric Model Intercomparison Project (AMIP) and from the Center for Ocean-Land-Atmosphere Interactions (COLA) with results of the thirteen-year integration of their low-resolution (COLA R15 version) AGCM with observed SSTs. This comparison was made in two different ways. The simulation of the annual cycle of tropical surface winds in 17 AMIP AGCMs was documented and intercompared employing suitable statistical indices. These statistical indices were designed to compress and extract only relevant information and were needed because of the volume of information that one has to handle when using such a large number of AGCM simulations. This study brings out some interesting features illustrating different aspects of the performance of current AGCMs in the simulation of the tropical surface stress and its annual cycle (Saji and Goswami, 1997). In contrast to older generation AGCMs that tended to produce weaker than observed annual mean stress, most AGCMs considered here were seen to capture the spatial features of the fields of annual mean and annual variance quite realistically. A common feature of the different model simulations is the weaker than observed surface stress simulated over the Somali and the South East Pacific regions. The underestimated surface stress over the Somali region is linked with the weak precipitation simulated over the land to the east of it, while the problem over the South East Pacific is possibly due to insufficient representation of the boundary layer processes. Another result of concern from this study was the diversity of the different model simulations of surface stress over Somali, Equatorial Indian Ocean, North and South Central Pacific, and the North and South East Pacific. For the interannual variability, we studied the simulation results of tropical Pacific surface winds in the COLA R15 AGCM (Goswami et al., 1995). The COLA AGCM was integrated with observed SSTs for the period Jan 1979 through Mar 1992. This AGCM was noted to be successful in simulating the annual cycle and the interannual variability of tropical Pacific surface winds. The locations of the prominent features such as the trade wind maxima, the Intertropical Convergence Zone (ITCZ), and the South Pacific Convergence Zone (SPCZ) are all well simulated, although the strength of the mean zonal winds in the trade wind maxima were about 20% weaker than observations. The amplitude of the annual cycle of the simulated winds was close to the observations except over the SPCZ and ITCZ regions where it is much weaker. The AGCM was seen to simulate realistically the spatial structure and temporal evolution of the large-scale low-frequency variability. Noteworthy was the simulation of the eastward migration of zonal wind anomalies associated with the 1982-83 El Niño. Successful performance of this AGCM, as compared to some of the older generation AGCMs, was seen to be linked to its ability to realistically simulate the tropical Pacific precipitation fields. This was substantiated using a simple linear model. This model was forced with the GCM simulated convective heating anomaly field and was successful in simulating the large-scale aspects of the GCM simulated surface wind anomaly fields. Thus, precipitation was demonstrated to be important for forcing surface wind variability on the interannual time scales in this model. The precipitation field was seen to be an important forcing function for the tropical Pacific surface wind interannual variability in the AMIP simulations too. The studies described above helped understand the problems associated with the simulation of tropical surface winds in current AGCMs. It also revealed some grey areas that needed further attention. Some of these 'problem regions' such as the Central Pacific are sensitive regions where the ocean and the atmosphere are strongly coupled. Hence, rectification of these discrepancies in AGCM surface winds is essential for correct CGCM simulations. Like the Central Pacific Ocean, another region where AGCMs performed poorly was the Indian Ocean. The poor simulations were associated with the unrealistic simulation of the precipitation over the monsoon region in the AGCMs. Considering the interactions between the monsoon and the ENSO, this is another crucial area where AGCM simulations need to be improved. The South East Pacific, where most AGCMs underestimated the surface wind stress, is another domain of concern. We hypothesize that the problem in the South East Pacific arises due to the insufficient representation of the boundary layer processes. This hypothesis is tested and its relevancy substantiated using a simple model for the tropical Pacific surface winds. Besides their potential as diagnostic tools for addressing such and related problems in AGCMs, such simple models also form efficient tools for sensitivity and predictability studies. Another concern associated with the use of AGCMs in studies of ENSO variability is the poor simulation of the high-frequency part of the interannual variability. Although the low-frequency part of the interannual variability is well simulated by the AGCMs, the unrealistic high-frequency part is capable of degrading the simulation of SSTs in ocean models. However, two aspects of simple models have been questioned in the past. These pertain to (i) the assumption of linear dynamics and (ii) parameterization of atmospheric forcing as a function of sea surface temperature (SST). Most simple models employ Gill-type linear dynamics and the atmospheric forcing arises solely from the latent heat release associated with large-scale organized convection. Some recent studies have shown that the linear dynamics assumption is reasonable and realistic simulation of surface winds is possible through a linear system, provided correct atmospheric heating is prescribed. Hence, a major thrust of the research and the starting point of model development was to develop a realistic parameterization of atmospheric heating within the linear framework. Another hypothesis that was tested is regarding the importance of the direct effects of SST gradients in forcing the tropical surface winds. As tropical convection is governed by nonlinear dynamical and thermodynamical processes, it is a daunting task to parameterize tropical convection accurately within the framework of a linear model. Hence, the problem was simplified by employing the observed relationship between deep convection and SSTs in the tropical Pacific. An empirical parameterization of organized tropical deep convection was then developed. This new parameterization was tested and was found to be successful in capturing the large-scale temporal and spatial features associated with tropical convective variability. Another important feature of the model developed was the inclusion of the direct effects of SST gradients on the tropical surface winds. The refined simple model was then integrated with observed SSTs for a 17-year period (1974–1991), and the improvement in the simulation of the interannual variability evaluated using observed surface winds from the Comprehensive Ocean-Atmosphere Data Set (COADS) and the European Centre for Medium-Range Forecasting (ECMWF) analyses (Saji and Goswami, 1996a). The model also realistically simulates the annual cycle of the tropical Pacific surface winds. The model was then applied to study the relative significance of the effects of deep convection and SST gradients in forcing the tropical surface wind variability and help understand the questions posed for improving AGCM performance. The improved simulation of the interannual variability and the annual cycle, besides the computational efficiency of this simple model, makes it an ideal component for use in simple or hybrid coupled models for the study of tropical ocean-atmosphere interactions. PUBLICATIONS: 1. Goswami, B. N., Krishnamurthy, V., & Saji, N. H. (1995). Simulation of ENSO-related surface winds in the tropical Pacific by an AGCM forced by observed SST. Monthly Weather Review, 123, 1677-1694. 2. Saji, N. H., & Goswami, B. N. (1996a). An improved linear model of tropical surface wind variability. Quarterly Journal of the Royal Meteorological Society, 122, 23-54. 3. Saji, N. H., & Goswami, B. N. (1997). Intercomparison of the seasonal cycle of tropical surface stress in 17 AMIP Atmospheric General Circulation Models. Climate Dynamics (To appear).