2013
DOI: 10.1002/2013jd019838
|View full text |Cite
|
Sign up to set email alerts
|

Conditions for successful data assimilation

Abstract: [1] We show, using idealized models, that numerical data assimilation can be successful only if an effective dimension of the problem is not excessive. This effective dimension depends on the noise in the model and the data, and in physically reasonable problems, it can be moderate even when the number of variables is huge. We then analyze several data assimilation algorithms, including particle filters and variational methods. We show that well-designed particle filters can solve most of those data assimilati… Show more

Help me understand this report
View preprint versions

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1

Citation Types

4
58
0

Year Published

2013
2013
2023
2023

Publication Types

Select...
7
2
1

Relationship

2
8

Authors

Journals

citations
Cited by 37 publications
(62 citation statements)
references
References 50 publications
4
58
0
Order By: Relevance
“…The present paper complements the work in Chorin and Morzfeld [23], where it was shown that, in physically reasonable vector-valued data assimilation problems, the various one-time components of the noise must be correlated; in particular, they must be spatially correlated if the several components describe physical variables estimated at different spatial points. In the present paper we show that the noise components must be correlated in time as well.…”
Section: Discussionsupporting
confidence: 59%
“…The present paper complements the work in Chorin and Morzfeld [23], where it was shown that, in physically reasonable vector-valued data assimilation problems, the various one-time components of the noise must be correlated; in particular, they must be spatially correlated if the several components describe physical variables estimated at different spatial points. In the present paper we show that the noise components must be correlated in time as well.…”
Section: Discussionsupporting
confidence: 59%
“…On the other hand, gradient-15 based methods are more prone to getting stuck in local minima and, furthermore, they require the complex calculation of the gradient of the cost function with respect to all parameters. With idealised models, Chorin and Morzfeld (2013) have shown that the assimilation can be optimal with particle filters or variational methods, under the condition that the dimension of the problem is not excessive. The choice of the minimisation method was shown to have little impact on the overall optimisation efficiency for relatively simple ecosystem models (Trudinger et al, 2007;Fox et al, 2009).…”
Section: Introductionmentioning
confidence: 99%
“…We note here that Chorin and Morzfeld (2013) have investigated a different, but related, effective dimension of a Gaussian data assimilation problem. In particular, they define a ''feasibility criterion'' to be the Frobenius norm of the steady-state posterior covariance matrix (which can be exactly calculated in the linear, Gaussian regime.)…”
Section: Introductionmentioning
confidence: 99%