Abstract: The Objective of this study is to carry out a set of four-dimensional variation al (4DVAR) experiments to assess the impact of cloud-drift wind data from the operational Geosynchronous Meteorological Satellite 5 (GMS-5) on improvement of the initial conditionals and numerical track predictions of tropical cyclones (TCs) in the western North Pacific (WNP).The data assimilated were derived from GMS-5 infrared and water vapor imageries and provided by the China National Satellite Meteorological Center. About 70% of the multispectral winds are observed above 400 hPa, and 50% of the data are between 200 and 300 hPa. Experiments were carried out using the Pennsylvania State UniversityNational Center for Atmospheric Research nonhydrostatic Mesoscale Model version 5 (MM5) and its 4DVAR system. A 6-h assimilation window was used to incorporate the clouddrift wind data at the initial and 6 h later. A simple quality control similar to the first-guess check in the ECMWF system was used here.Twentytwo cases were examined for 8 different WNP TCs in 2002. Forecasts up to 48 h were performed with the original and 4DVARassimilated optimal initial conditions. The 4DVAR assimilation of the cloud-drift wind observations led to significant improvements, with the relative reductions in track error by 5% at 12 h, 12% at 24 h, 10% at 36 h and 7% at 48 h on average. But the effectiveness of the assimi lation of the clouddrift wind data apparently varies with TC intensity. If a c entral pressure of 960 hPa is selected as the demarcation between strong and wea k TCs, the mean track error reductions for the 11 strong TCs range from 13% at 12 h to over 20% after 12 h. However, for 11 weak TCs, slight increases in the average track errors were observed. The results suggest that assimilation of the clouddrift wind data for TCs of different initial intensity has different impact on TC track forecast. This is apparently because stronger TCs are affected to a larger extent by the u pper tropospheric circulation, and they usually have a deeper vertical structure . At the same time, the cloud-drift winds associated with weaker TCs are fewer and more disorganized, and they are easier to be influenced by the environmental circulation. Therefore when the model variables are not consistent enough with the model or with each other after assimilating, a negative impact may result.