Theoretical, observational and numerical studies have shown that a tropical storm's motion depends on the steering current, and its structure and size. The initial structure of a storm usually cannot be resolved due toa lack of observations. Furthermore, these storms often form over data sparse oceanic regions so that even the large scales of motion are not well analyzed. A model initial state based on such an analysis may have storm structure and steering current that are quite different than those in the real atmosphere. Based on available observations and climatological data procedures have been developed to improve the initial structure in the operational NCEP's Medium Range Forecast (MRF) model. MRF forecasts using two storm specification procedures will be compared. Evaluation of both predicted intensity and track will be presented