基于西安市1951年－2013年降水实测资料，利用距平、累计距平和Mann-Kendall检验的方法对降水特征进行分析，用BP神经网络模型对年降水量进行模拟预测。研究结果表明：西安市年降水量呈下降趋势但趋势不显著（置信度95 %），四季中除夏季表现出微弱的上升趋势外，其他季节均呈现下降趋势，且春季降水下降趋势显著（置信度95 %）；年降水量在1965年发生突变；BP神经网络预测结果为2014年与2016年降水偏丰，而2015年降水偏枯。研究成果对于合理科学利用降水资源具有一定指导意义。

Based on the observed data of precipitation from 1951 to 2013 in Xi’an city, its precipitation characteristics were analyzed using anomaly, accumulated anomaly and Mann-Kendall test method, and then annual precipitation was simulated and predicted by BP neural network. The results showed that annual precipitation of Xi’an city was descending, but the trend was insignificant (confidence level was 95%), all seasons presented descending except summer which showed a slight increasing trend, and precipitation in spring had a significant declining trend (confidence level was 95%). The abrupt change of annual precipitation occurred in 1965. The forecasting results by BP neural network showed that annual precipitations of 2014 and 2016 were abundant and 2015 was low. The research has certain significance for guiding scientific and rational use of precipitation resource.

水利部公益性行业科研专项（201301084）