INTERCOMPARISON OF THE SEASONAL CYCLE OF TROPICAL SURFACE STRESS IN 17 AMIP ATMOSPHERIC GENERAL-CIRCULATION MODELS

The mean state of the tropical atmosphere is important as the nature o f the coupling between the ocean and the atmosphere depends nonlinearl y on the basic state of the coupled system. The simulation of the annu al cycle of the tropical surface wind stress by 17 atmospheric general circulation models (AGCMs) is examined and intercompared. The models considered were part of the Atmospheric Model Intercomparison Project (AMIP) and were integrated with observed sea surface temperature (SST) for the decade 1979-1988. Several measures have been devised to inter compare the performance of the 17 models on global tropical as well as regional scales. Within the limits of observational uncertainties, th e models under examination simulate realistic tropical area-averaged z onal and meridional annual mean stresses. This is a noteworthy improve ment over older generation low resolution models which were noted for their simulation of surface stresses considerably weaker than the obse rvations. The models also simulate realistic magnitudes of the spatial distribution of the annual mean surface stress field and are seen to reproduce realistically its observed spatial pattern. Similar features are observed in the simulations of the annual variance field. The mod els perform well over almost all the tropical regions apart from a few . Of these, the simulations over Somali are interesting. Over this reg ion, the models are seen to underestimate the annual mean zonal and me ridional stresses. There is also wide variance between the different m odels in simulating these quantities. Large model-to-model variations were also seen in the simulations of the annual mean meridional stress field over equatorial Indian Ocean, south central Pacific, north east Pacific and equatorial eastern Pacific oceans. It is shown that the s ystematic errors in simulating the surface winds are related to the sy stematic errors in simulating the Inter-Tropical Convergence Zone (ITC Z) in its location and intensity. Weaker than observed annual mean sou thwesterlies simulated by most models over Somali is due to weaker tha n observed southwesterlies during the Northern Hemisphere summer. This is related to the weaker than observed land precipitation simulated b y most models during the Northern Hemisphere summer. The diversity in simulation of the surface wind over Somali and equatorial Indian ocean is related to the diversity of AGCMs in simulating the precipitation zones in these regions.