3D-Var data assimilation using a variational autoencoder

Boštjan Melinc (Author), Žiga Zaplotnik (Author)

Abstract

Data assimilation of atmospheric observations traditionally relies on variational and Kalman filter methods. Here, an alternative neural network data assimilation (NNDA) with variational autoencoder (VAE) is proposed. The three-dimensional variational (3D-Var) data assimilation cost function is utilised to determine the analysis that optimally fuses simulated observations and the encoded short-range persistence forecast (background), accounting for their errors. The minimisation is performed in the reduced-order latent space discovered by the VAE. The variational problem is autodifferentiable, simplifying the computation of the cost-function gradient necessary for efficient minimisation. We demonstrate that the background-error covariance (B) matrix measured and represented in the latent space is quasidiagonal. The background-error covariances in the grid-point space are flow-dependent, evolving seasonally and depending on the current state of the atmosphere. Data assimilation experiments with a single temperature observation in the lower troposphere indicate that the B matrix describes both tropical and extratropical background-error covariances simultaneously.

Keywords

meteorologija;asimilacija meritev;strojno učenje;nevronske mreže;variacijski avtokodirnik;3D-Var;meteorology;data assimilation;machine learning;neural networks;variational autoencoder;

Data

Language:	English
Year of publishing:	2024
Typology:	1.01 - Original Scientific Article
Organization:	UL FMF - Faculty of Mathematics and Physics
UDC:	551.5
COBISS:	194564099
ISSN:	0035-9009
Views:	60
Downloads:	35
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	Slovenian
Secondary keywords:	meteorologija;asimilacija meritev;strojno učenje;nevronske mreže;variacijski avtokodirnik;3D-Var;
Type (COBISS):	Article
Pages:	str. 2273-2295
Volume:	ǂVol. ǂ150
Issue:	ǂiss. ǂ761
Chronology:	2024
DOI:	10.1002/qj.4708
ID:	24291058