Utilizing monthly average sea surface temperature (SST) data from 1951 to 1994, the correlative periods of SST and severe floods in Hubei Province were identified through lag correlation analysis. The sums of SST at the grid points in key areas were used as forecasting factors intimately associated with the severe floods in Hubei Province. The prediction curves of severe floods for different time periods were built up by using statistic methods. The method worked well in the past eight years in operational application, especially in 1996 and 1998 provided correct trend predictions of floods.