Prediction accuracy of precipitation has an important influence on evaluation of groundwater resources that is recharged directly or indirectly by the precipitation.In former literature, many authors usually adopted the frequency of precipitation in the past to predict the precipitation in the future, and brought it into the models which evaluated the quantity and/or quality of groundwater.But this method is rather conservative.In this paper, we combine time series method with the dispersed Markov Chains theory of stochastic process, and present a Markov Model based the time series analysis for predicting the precipitation.In the course of modeling, first we select one-dimension nonstationary time series model to predict the precipitation in light of characteristics of the precipitation series.The result shows that the time series prediction is feasible as a whole, but there exists bigger errors when predicting the variables on the tops of the curves.In order to improve the prediction accuracy of the model, especially to the data with stronger fluctuation, we use the method of Markov’s state change probability matrix to fit them again.Then we attain the fitting values.To test the time series-Markov model, we apply it to predict the precipitation in the section of Xuzhou in Jiangsu Province as an example.Results show that the time series-Markov Model is efficient, and it has higher accuracy than that of the single model of time series.It enlarges applied scope of the time series model, and it is of important practical values and theoretical magnificence to the evaluation of groundwater resources that is finally recharged by the precipitation.