摘要
用MJ0指数RMM1、RMM2、振幅1—25日平均代替月平均,用上月月平均RMM1和RMM2、振幅构造为右场,下月长江流域降水场为左场,SVD分析两场的关联,借助最优化技术,在降水场预测距平与实况距平同号总站数最大意义下确定系数,建立估计公式,由右场时间系数估计左场时间系数,最后反演降水场。尽管多数的月第一模态相关并不显著,但实际预测效果较好。
RMM1, RMM2 and amplitude averaged from 1 to 25th are used to take the place of the average monthly. Right field consists of RMM1, RMM2 and amplitude of the preceding month, and left field con- sists of monthly rainfall over the Yangtze River Valley. The relationships between the two fields are stud- ied by SVD. With optimization technique the time coefficients of the left field are estimated on the basis of the time coefficient of the right field according to the total number of stations which have the same sign be- tween estimated anomaly and observed anomaly. Rainfall fields are retrieved by linear combination of the time coefficients and the vectors. Even though few of first models of SVD are above the significant level, the actual predictions achieve good results.
出处
《气象》
CSCD
北大核心
2013年第9期1217-1220,共4页
Meteorological Monthly
基金
中国长江电力股份有限公司"长江流域中长期气象预报方法研究"
"关键期天气气候特征分析和预报方法研究"
中国气象局2013年小型基建项目"极端天气气候事件监测预测业务平台建设(二期)"
武汉区域气象中心科技发展基金项目
关键词
气候学
MJO
奇异值分解
长江流域降水
climatology, Madden-Julian oscillation, singular value decomposition, rainfall in the Yangtze River
