A simple method of calculating the optimal step size in 4DVAR technique

In four-dimensional variational data assimilation (4DVAR) technology, how to calculate the optimal step size is always a very important and indeed difficult task. It is directly related to the computational efficiency. In this research, a new method is proposed to calculate the optimal step size mor...

Full description

Saved in:
Bibliographic Details
Published inAdvances in atmospheric sciences Vol. 17; no. 3; pp. 433 - 444
Main Authors Yunfeng, Wang, Rongsheng, Wu, Yuan, Wang, Yinong, Pan
Format Journal Article
LanguageEnglish
Published Dordrecht Springer Nature B.V 01.09.2000
Subjects
Online AccessGet full text

Cover

Loading…
More Information
Summary:In four-dimensional variational data assimilation (4DVAR) technology, how to calculate the optimal step size is always a very important and indeed difficult task. It is directly related to the computational efficiency. In this research, a new method is proposed to calculate the optimal step size more effectively. Both nonlinear one-dimensional advection equation and two-dimensional inertial wave equation are used to test and compare the influence of different methods of the optimal step size calculations on the iteration steps, as well as the simulation results of 4DVAR processes. It is in evidence that the different methods have different influences. The calculating method is very important to determining whether the iteration is convergent or not and whether the convergence rate is large or small. If the calculating method of optimal step size is properly determined as demonstrated in this paper, then it can greatly enlarge the convergence rate and further greatly decrease the iteration steps. This research is meaningful since it not only makes an important improvement on 4DVAR theory, but also has useful practical application in improving the computational efficiency and saving the computational time.
Bibliography:ObjectType-Article-2
SourceType-Scholarly Journals-1
ObjectType-Feature-1
content type line 23
ISSN:0256-1530
1861-9533
DOI:10.1007/s00376-000-0034-5