On the Orthogonalization of Bred Vectors

On the Orthogonalization of Bred Vectors
复制标题

关于育种向量的正交化

DOI:
10.1175/2010waf2222334.1
复制
发表时间:
2010
影响因子:
2.9
通讯作者:
Rhodin A.
Rhodin A.
中科院分区:
地球科学3区
文献类型:
--
作者:
Keller;Kornblueh L;Rhodin A.

文献摘要

参考文献

被引文献

相似文献

提高集合预报质量评估固有流动不确定性的关键是初始集合扰动的选择。为了产生这样的扰动,在过去的二十年里,人们一直使用生长模式培育方法。这里,估计初始模型状态的最快增长误差模式。然而,由此产生的繁殖向量(BV)主要指向领先Lyapunov向量的相空间方向,因此倾向于误差增长的一个方向。为了克服这一特点,并获得指向Lyapunov向量的增长模式,基于边界值之间的相似性,采用了一种正交化方法,实现为奇异值分解。这种转换类似于NCEP目前在操作中使用的集合转换技术,但在使用的度量和实现方面有一定的差异。在这项研究中,给出了使用基于德国气象局全球数值天气预报模式GME的集合预报系统(EFS)产生的BVS的结果。通过使用评估集合可靠性、方差和分辨率的不同概率预报分数,显示了正交化BV初始化所获得的预报性能的增益。对于2007年夏季的3个月期间,将结果与相同集合预报系统的简单BV初始化产生的预报以及来自ECMWF和NCEP的业务集合预报进行比较。正交化极大地提高了GME-EFS的得分,使他们与其他两个中锋竞争。
The key to the improvement of the quality of ensemble forecasts assessing the inherent flow uncertainties is the choice of the initial ensemble perturbations. To generate such perturbations, the breeding of growing modes approach has been used for the past two decades. Here, the fastest-growing error modes of the initial model state are estimated. However, the resulting bred vectors (BVs) mainly point in the phase space direction of the leading Lyapunov vector and therefore favor one direction of growing errors. To overcome this characteristic and obtain growing modes pointing to Lyapunov vectors different from the leading one, an orthogonalization implemented as a singular value decomposition based on the similarity between the BVs is applied. This transformation is similar to that used in the ensemble transform technique currently in operational use at NCEP but with certain differences in the metric used and in the implementation. In this study, results of this approach using BVs generated in the Ensemble Forecasting System (EFS) based on the global numerical weather prediction model GME of the German Meteorological Service are presented. The gain in forecast performance achieved with the orthogonalized BV initialization is shown by using different probabilistic forecast scores evaluating ensemble reliability, variance, and resolution. For a 3-month period in summer 2007, the results are compared to forecasts generated with simple BV initializations of the same ensemble prediction system as well as operational ensemble forecasts from ECMWF and NCEP. The orthogonalization vastly improves the GME–EFS scores and makes them competitive with the two other centers.
迈向 GME 集合预报系统:使用育种技术进行集合初始化
DOI: --
发表时间: 2008
期刊:
影响因子: --
作者:
J. Keller;L. Kornblueh;A. Hense;A. Rhodin
通讯作者: A. Rhodin
混沌遍历理论
DOI: --
发表时间: 1985
期刊: Topical Meeting on Optical Bistability (OB3)
影响因子: --
作者:
D. Ruelle
通讯作者: D. Ruelle
DOI: --
发表时间: 2008
期刊:
影响因子: --
作者:
T. Palmer;R. Buizza;R. Hagedorn;A. Lawrence;M. Leutbecher;Lenny Smith
通讯作者: Lenny Smith
混沌遍历理论
DOI: --
发表时间: 2005
期刊:
影响因子: --
作者:
W. Perruquetti;J. Barbot
通讯作者: J. Barbot
繁殖向量与李亚普诺夫向量相同吗
DOI: --
发表时间: 2002
期刊:
影响因子: --
作者:
E. Kalnay;M. Corazza;M. Cai
通讯作者: M. Cai