The study of the relationship between weather conditions and the unit yield of late rice shows that weather factors, especially air temperature and sunshine, are essential to the unit yield change of late rice for two years in succession. According to the needs of operational service, a method of forecasting dynamically the trends of the late rice yield for the whole country is presented, based on the theory of synthesize cluster analysis and such data as the late rice unit yields of the main planting provinces in China, average air temperature and ten-day sunshine duration. The method can be used to dynamically predict the unit yield of late rice after having sowed for some time, being simple and practical and having high accuracy. It is difficult to screen out the predicting factors within a short time by using the common regression method; therefore, this method is of better value in operational application.