摘要
经SVD分析,截取足够多的预报场和因子场时间系数,使其相互关系代表两场的大尺度联系,预报场时间系数与其奇异向量线性组合估计场能反映原场主要特征。利用最小二乘法得到数值上最接近原场的初值,借助最优化技术,确定合理的系数,建立预测公式,由因子场时间系数预测预报场时间系数,同时订正预报场时间系数a1,a2,……,aN本身的误差和反演过程中分析误差造成的场格点趋势预测的误差。最后将预测的预报场时间系数和对应奇异向量反演为整个场的预报。预报过程重点考虑可预报的大尺度变化,滤去不可预报的小扰动,依据两场主要耦合关系,预测预报场未来的主要变化。
With the SVD (singular value decomposition) sufficient amount coefflcients of predicted field and factor field are truncated so that their relationships represent the two fields and estimate of linear combination of factor coefflcients and the vector can indicate the original. With the optimization technique the whole field is predicted with estimating coefflcients of the predicted field according to the coefflcients of the factor field and linear combination of the coefflcients and the vectors of the predicted field. Mainly predictable large-scale variations are analyzed and unpredictable disturbance are canceled. Large-scale variations of predicted fields are predicted with main coupled relationship of the two fields.
出处
《热带气象学报》
CSCD
北大核心
2004年第4期383-390,共8页
Journal of Tropical Meteorology
基金
湖北省气象局<华中地区月
季尺度降水场预报技术研究>课题资助
关键词
最优化技术
奇异值分解
场估计
Optimization technique
singular value decomposition
estimate of field
