滑动平均-马尔可夫模型在托什干河流域降水预测中的应用

Application of Moving Average-Markov Chain in the Prediction of Precipitation in Tuoshigan River Basin

中长期降水量的预测是气象科学的一个难点问题，也是水文学中的一个重要问题。根据降水过程存在大量不确定性的特点，通过聚类分析建立降水序列的分级标准，采用规范化的各阶自相关系数为权重，用滑动平均的马尔可夫链模型，通过状态转移概率矩阵预测未来时段的降水状态，并根据模糊集理论中的级别特征值计算具体的降水量，最后以新疆托什干河沙里桂兰克水文站44年的降水资料为实例，对该方法进行了具体的应用，预测精度较高，为提高中长期降水量预报的精度提供了一条值得探索的途径。

The prediction of the medium and long term precipitation is a difficult point in meteorology. And it is also an important one in hydrology. According to the uncertain characteristics in precipitation procession, the graduation standard of the precipitation serial was set through cluster analysis. The normalized different autocorrelation coefficient was used as weight, and the moving average-Markov chain was used to predict precipitation state in a future period with state transition probability matrix, then the particular precipitation was calculated according to the class designated value in unresolved set theory. At last,the precipitation material in 44 years of the Shaliguilanke hydrological station in Tuoshigan river was used as a model and found out that the precipitation accuracy was satisfied. So the moving average-Markov chain model can be used in the prediction of the medium and long range precipitation, which provides a channel to be searched.