In a forecast of tropical cyclone, the most important aspect is to be able to determine its movement (track) as accurate as possible. However, lack of observed wind data around the storm made it difficult to achieve this goal. In addition, uncertainties in numerical weather prediction models and in observed data make it even more difficult to have an accurate track forecast. In this paper, an asymmetric wind model of tropical cyclone is developed and used to generate bogus wind of tropical cyclone XANGSANE (0615). Track forecasts of the storm by a single level primitive equation model show that the appropriate pressure level for the application of the bogus wind is 850hPa. To reduce uncertainties in the forecast tracks, ensemble forecasts based on the singular vector method are performed to generate additional 50 forecasts from slightly different perturbed initial conditions. With a suitable size of perturbation that is used to create the perturbed initial conditions, the ensemble forecast tracks are well spread and cover the observed track.