基于加权马尔科夫链的新疆近十年年降水量预测效果分析
Prediction of Annual Precipitation Based on the Weighted Markov Chain in Xinjiang in Recent 10 Years

采用加权马尔可夫链模型原理，以乌鲁木齐市为例，依据1961~2008年年降水量资料建立加权马尔可夫链模型，计算了2009~2018年的降水量状态和年降水量，对预测降水量和实况降水量进行了对比分析；在此基础上，计算了新疆88个气象站2009~2018年年降水量，通过对预测降水量距平百分率的分析，可知：各站预测效果差异很大，但整体上北疆和天山山区预测偏差小的年份明显多于南疆大部；各站实际降水量的多少明显影响偏差在1成以内的预测效果，常年降水量在300毫米以上的站点最为适用；6个代表站的预测结果也验证了加权马尔科夫链模型在降水量相对偏多的北疆和天山山区站点预测效果好于南疆站点且效果更加稳定。
The weighted Markov Chain was established by using annual precipitation data from 1961 to 2008 in Urumqi and the states and the quantities of the annual precipitation from 2009 to 2018 were calculated. The prediction and the observation were compared. Based on this, annual precipitation of 88 meterological stations in Xinjiang from 1961 to 2008 was calculated. The results show that the relative errors of prediction in different stations are quite different from each other, but as a whole, the years with lower relative errors are more in Northern Xinjiang and Tianshan Mountain than that in Southern Xinjiang. The prediction within 10% relative errors is obviously affected by the annual precipitation, which means it is the most fitful in stations with more than 300mm annual precipitation. The result of prediction in 6 representative stations has also testified that the effect of the weighted Markov Chain is much better in Northern Xinjiang and Tianshan Mountain than that in Southern Xinjiang and is steadier.