The accurate description of initial conditions in numerical simulation of tropical cyclones usually suffers from a lack of observational data over the tropical oceans which is an important factor limiting the improvement of numerical model simulations and the accuracy of cyclone track predictions. It is therefore, expected that assimilating winds near the sea surface in numerical models will improve the cyclone track and intensity prediction. The microwave scatterometer QuikScat (or QSCAT) was launched in June 1999 with a primary mission of measuring winds near the sea surface. In the present work, we assimilated QSCAT winds in a numerical weather prediction model. Our initial studies shows that a direct assimilation of sea surface winds into an NWP model for a cyclone case does not give a very realistic simulation, since the winds are available at only ocean surface. For a more realistic simulation, a number of cyclonic situations were selected. Three dimensional wind fields generated by the numerical model for these situations were used. A complex EOF analysis of vertical wind profiles was carried out to determine the dominant vertical structure of wind velocity in a cyclonic situation. While assimilating the QSCAT winds for a cyclone, the above structure functions were used to determine an approximate 3-dimensional field of wind. The complete field was assimilated by the method of nudging. It is found that above scheme not only helps in successful evolution of tropical cyclones but also improves cyclone track and intensity prediction.