基于增长模繁殖法的初始分析误差计算及集合预报试验  

Computation of initial analysis error and ensemble forecast based on the method of "BGM"

作　　者：关吉平[1] 张立凤[1] 马环宇 

机构地区：[1]解放军理工大学气象学院,南京211101 [2]北京2433信箱6分箱,北京100081

出　　处：《气象科学》2009年第3期317-322,共6页Journal of the Meteorological Sciences

基　　金：国家重点基础研究项目(2004CB418304);解放军理工大学气象学院博士启动基金

摘　　要：首先用成对繁殖的方法对NCEP再分析资料与T106分析场中的快速增长分析误差进行了计算和比较,然后把它们叠加到各自的初始场中进行了多初值集合预报试验。结果表明,用不同分析资料计算的达到饱和的快速增长误差,两者在全球分布上既有许多相似的地方,也存在一些不同,多初值集合是综合考虑它们的有效方法;考虑了两种分析误差的集合预报比只考虑一种分析误差的集合预报优越;在集合总成员数不变的前提下,发现引入另一种分析资料的集合成员数并不是越多越好,就预报技巧的最大改善而言,其存在一个阀值。The fast-growing analysis error in NCEP reanalysis data and T106 analysis field is computed and compared using the method of ＂twin-breeding＂, then multi-initial-value ensemble forecasts are done by adding them to their initial field. Results show that for the saturated fast-growing error computed using different analysis data, their distribution has much similarity and some difference and the multi-initial-value ensemble forecast is an effective method to consider them synthetically; the ensemble forecasts considering two kinds of analysis error are superior to those only considering one kind of analysis error; under the condition of fixed number of members, this is not true that the more the number of members from another analysis data, the better the result, for the maximum improvement of forecast skill, there must be a threshold.

关 键 词：增长模繁殖法 初始分析误差 集合预报 

分 类 号：P435.1[天文地球—大气科学及气象学]