门限自回归模型在梅雨降水量预测中的应用  

Application of Threshold Auto-Regressive Model to Meiyu Rainfall Forecasting

作　　者：卞正奎 刘文军[1] 刘方[1] 

机构地区：[1]江苏省泰州市气象局,泰州225300

出　　处：《气象科技》2006年第1期18-21,共4页Meteorological Science and Technology

基　　金：江苏省气象局科研开发经费(No.200105)资助

摘　　要：针对梅雨量的分布特性,提出应用门限自回归模型建立一套简便实用的梅雨量丰枯预测方案.利用相关分析取得恰当的延迟特性,根据梅雨量在X轴上的分布特点选取门限值,并通过改进遗传算法和高斯-牛顿法优化回归系数项和回归系数,最后采用分级预报方法对梅雨量做5级预报.该方案在1967～2002年泰州地区梅雨量丰枯预测中取得了很好的效果,其历史拟合率和预报准确率分别达到87%和80%.The forcasting of Meiyu rainfall is very important to industry and agriculture in Taizhou. According to the distributing features of Meiyu in the area, a convenient and practical method is put forward. A Meiyu predication experiment using the threshold auto-regressive model, advanced genetic algorithm, and related techniques was carried out in Thizhou. The graded forecasting method was adopted in predicting Meiyu rainfall and the statistical results are provided. The result shows that the fitting-rate and the accuracy are 78 % and 80 %, respectively.

关 键 词：门限自回归模型 改进遗传算法 分级预报 梅雨 

分 类 号：P426.62[天文地球—大气科学及气象学] O212.1[理学—概率论与数理统计]