Skip to Main content Skip to Navigation

Assimilation multi-échelle dans un modèle météorologique régional.

Abstract : Nowadays, most limited area meteorological models have their own data assimilation system. Theses analyses generaly mix a first-guess, which is taken from the forecast model, with observations in order to freshen up the patterns described by the limited area model. Nevertheless, the coupling model can also be of some interest. Indeed, the coupling model is generaly a global model, that benefits from top-of-the-range data assimilation techniques, and thus has a good description of the larger scales. The goal of this PhD thesis is to bring information from the coupling model directly into the 3-dimensional variational assimilation (3D-VAR) of the limited area model (LAM), as a new source of information. In concrete terms, the input information vector in the LAM assimilation is the concatenation of various sources of information : a first-guess from the model, observations and larger scales from the coupling model analysis. This formalism uses a measure on the uncertainty of the sources of information, which is described by the covariances between the errors of the sources of information. Some simplifications on the cross-covariances between the sources are proposed, so that these developments are suitable for an easy implementation in the current analysis software. A first evaluation of this new formulation is performed within the framework of a unidimensional “shallow water” model, using both coupling and coupled models. These experiments show a neutral to a positive impact, depending on the setting of the experiments, which is limited by the simplified framework of this academic model. In the framework of the application of thismethod in the models used atMétéo-France (global model ARPÈGE and LAM ALADIN), the statistics on the errors of the sources of information have been evaluated. Firstly, the scales to be taken from the global model analysis have been selected, in order to keep only the larger patterns (roughly 240 km). Then, using previous studies based on ensemble methods to sample the errors, statistics have been computed, which enabled a description of the error covariances (standard deviations, isotropy, etc.) and to quantify the error induced by the proposed simplifications. The implementation of the method in ALADIN has been evaluated on 15-day long assimilation cycles, which resulted in a slightly positive impact of the introduction a the larger scales from the global model analysis on objective scores. Nevertheless, in spite of visible and systematic differences due to the new source of information, no specific case study on diagnostic fields, such as precipitation, illustrates the benefit on present weather or on particular meteorological phenomenon. This PhD thesis introduces an innovative assimilation technique in a LAM which takes into account information from the coupling model, in addition to the observations, to correct the model first-guess. It is an open way for further researches, for instance, adding the time dimension or through a modification of the scale selection of the patterns taken from the coupling model.
Document type :
Complete list of metadatas

Cited literature [36 references]  Display  Hide  Download
Contributor : Vincent Guidard <>
Submitted on : Friday, February 25, 2011 - 11:16:30 AM
Last modification on : Tuesday, February 18, 2020 - 10:48:04 AM
Long-term archiving on: : Thursday, May 26, 2011 - 2:34:35 AM


  • HAL Id : tel-00569483, version 1



Vincent Guidard. Assimilation multi-échelle dans un modèle météorologique régional.. Océan, Atmosphère. Université Paul Sabatier - Toulouse III, 2007. Français. ⟨tel-00569483⟩



Record views


Files downloads