基于小波和ARIMA的黄山市年降水量分析及预测

为研究区域降水时间特征，提高年降水量预测精度，采用Morlet小波对黄山市1957－2016年的年降水量周期进行分析，并基于1957－2011年的年降水量构建了ARIMA模型和小波与ARIMA组合模型，分别对该市2012－2016年的年降水量进行了预测及对比分析。结果表明：黄山市近60 a年降水量主要受28 a、13 a、5 a的周期波动影响；采用ARIMA及组合模型预测2012－2016年年降水量的平均相对误差绝对值分别为19.8%和12.3%，组合模型的拟合和预测效果更优；两种方法对2012年、2015年和2016年的年降水量预测误差均较大，可能是这几年降水受ENSO事件影响，降水机制异于常年，致模型预测误差较大。结果可为区域中长期水文预报提供科学依据，对区域旱涝灾害预警管理具有一定应用价值。 In order to identify the temporal characteristics of regional rainfall and improve the prediction accuracy of annual precipitation,the periodic variation of annual precipitation in Huangshan City from 1957 to 2016 was analyzed based on Morlet wavelet.A series of annual precipitations of Huangshan City from 1957 to 2011 was adopted to establish the ARIMA model and the combined model of ARIMA with wavelet,and both models were applied to predict the annual precipitation of Huangshan City from 2012 to 2016.The results show that the annual precipitation in Huangshan City is mainly affected by the periodic fluctuations of 28 a,13 a and 5 a.The mean absolute percentage error of the predicted annual precipitations for the period of 2012-2016 is 19.8% by applying the ARIMA model, while that for the same period is 12.3% by applying the combined model,indicating better accuracy in simulating and predicting the annual precipitation in Huangshan City by applying the combined model.Significant errors can be found in predicting the annual precipitations of Huangshan City in 2012,2015 and 2016 as different mechanism of precipitation in these years with normal years, probably resulting from an impact of ENSO events on the precipitations.As providing the scientific foundation for the regional long-and mid-term hydrological forecasting,the achievement is of certain value in management of early warning of the regional flood/drought disasters. 国家重点研发计划(2017YFC1502405);国家自然科学基金(51579060;51509065;51779067)