A new method of predicting dryness and wetness based on evolutionary modeling (EM) is presented in this paper. Numerical tests indicate that the basic dynamic characteriscs can be captured by EM and the model obtained by EM is not only able to preferablely simulate historical evolution, but also can exactly predict the future evolutionary trend of a time series. For the model obtained by EM with relatively larger prediction errors, the secondary EM can improve the prediction accuracy obviously.