A case study evaluating the predictive capability of an upper layer circulation model of the northwest Indian Ocean is presented. The model is a nonlinear, reduced gravity model incorporating realistic boundary geometry and is forced by observed winds. Model results for the fall of 1985 are compared with and evaluated against U.S. Navy bathythermograph and NOAA satellite data collected during August–November 1985. An assessment is made of the model's ability to simulate correctly the circulation structure. Ship wind observations are converted to wind stress for model forcing by a procedure developed by Legler and Navon (1988). The model is only moderately successful in reproducing the structure of the large, rather homogeneous pool of water located off the Arabian Peninsula in September. However, the model behaves remarkably well in the dynamically active region around Socotra. Major fronts and eddies frequently observed in the region during the transition period between the southwest and the northeast monsoon appear in the 1985 model results and compare well, both temporally and spatially, with the observational data. Thus given accurate wind information, the model appears highly effective in dynamically active regions and demonstrates potential as a useful prognostic tool for evaluation of the Arabian Sea when real time winds become available.