广西夏季异常降雨量的前馈网络预报模型

以1957～2001年广西37个基准站的夏季(6～8月)平均降水量为基础,将夏季降水量距平百分率大于等于20%、小于等于-23%作为异常多雨和少雨年,建立广西夏季降水量"0、1"化的异常序列,利用前馈网络的非线性映射技术,构造广西夏季降水异常预报模型.通过对该模型的预报检验分析发现,该预报模型不仅能准确地报出历史样本的异常多雨和异常少雨年,并且对2002～2004年的独立样本预报也全部正确.这为异常降水的短期气候预测业务工作提供了新的思路和方法.

A Feedforward Network Prediction Model of the Summer Precipitation Anomaly in Guangxi

Based on the mean precipitation in summer(June-August) on 37 benchmarking sites between 1957 and 2001,a prediction model of the abnormal precipitation is established by use of the nonlinear mapping technology of the artificial neural network.Years in which precipitation was 20% greater than the mean or above that and 23% smaller than the mean or below that are classified as years with abnormally high precipitation and years with abnormally low precipitation respectively.When it is used to analyze data,this model could not only accurately identify years with abnormally high precipitation as well as low precipitation,but also can make accurate independent prediction for the 2002-2004 samples.This study provides new perspective and method to short-term forecast of abnormal precipitation.