Application of marginalized particle filter to linear-Gaussian problems with unknown model error covariance structure

0316110 - ÚTIA 2009 CZ eng K - Conference Paper (Czech conference)
Hofman, Radek
Application of marginalized particle filter to linear-Gaussian problems with unknown model error covariance structure.
[Aplikace opomíjeného praktického filtru pro lineárně Gausovské problémy s neznámým modelem chyb kovariantních struktur.]
Sborník workshopu doktorandů FJFI oboru Matematické inženýrství. Praha: ČVUT, 2008, s. 1-12. ISBN 978-80-01-04195-6.
[Doktorandske dny 2008. Praha (CZ), 07.11.2008-21.11.2008]
R&D Projects: GA ČR(CZ) GA102/07/1596
Institutional research plan: CEZ:AV0Z10750506
Keywords : asimilation * linear-Gaussian problems
Subject RIV: DI - Air Pollution ; Quality
http://library.utia.cas.cz/separaty/2008/AS/hofman-application of marginalized particle filter to linear-gaussian problems with unknown model error covariance structure.pdf

The paper presents a scheme for estimation of spatio-temporal evolution of a quantity with unknown model error. Model error is estimated on basis of measured-minus-observed residuals evaluated upon measured and modeled values. Methods of Bayesian filtering are applied to the problem. The main contribution of this paper is application of general marginalized particle filter algorithm to the linear-Gaussian problem with unknown model error covariance structure. Methodology is demonstrated on the problem of modeling of spatio-temporal evolution of groundshine-dose from radionuclides deposited on terrain in long-time horizon.

Článek se zabývá odhadováním spatio-temporální evoluce kvantity s neznámým modelem chyb.
Permanent Link: http://hdl.handle.net/11104/0166132