Abstract: A one-way nested barotropic model used for predicting tropical cyclone track tendency is developed. The so-called “relay forecasting” time integral method is suggested. That is the grid data supplied by ECMWF through GTS are used not only for determining the boundary value of fine-mesh forecasting domain, but also for updating large scale circulation during integral time. Five different “relay forecasting” schemes are designed and are all tested on tropical cyclones No. 8507 and No. 8509. After comparing and analyzing the results obtained from various forecasting schemes, the main factors which caused the forecast errors are identified. The experimental results show that “relay forecasting” method has great capability for predicting tropical cyclone track tendency. It may forecast not only the normal tracks more accurately than the operational models, but also the anomalous tracks and erratic paths of tropical cyclones. The longer the integral time is, the more obvious the superiority of this method is. This shows that the “ relay forecasting” method has not only suggested a new possible way for predicting tropical cyclone track tendency, but also has made a first encouraging step to develop a mediumrange forecasting method of tropical cyclone tracks. Furthermore, the results from various forecast schemes indicate that they can be used to facilitate the operational forecast.