Page Personnelle de Christophe Baehr

Christophe Baehr's Homepage

Météo-France/CNRS Researcher member of the CNRM/GAME URA1357, French National Centre for Meteorological Research, Toulouse

Associated member of the Laboratory of Statistics and Probability University of Toulouse III - Paul Sabatier

Office: CNRM-GMEI, B035

Postal address:
Météo-France
CNRM/GMEI
42 Avenue de Coriolis
31057 Toulouse cedex
France

Phone: +33 561 079 641

E-mail: christophe.baehr_{at}_meteo_{dot}_fr, christophe.baehr_{at}_math.univ-toulouse.fr .




Recherche / Research

Stochastic Filtering of Observations on Random Media. Application to Measurements of Turbulent Fluids

Stochastic Filters for Data Assimilation. Application to Atmospheric Data Assimilation

En quelques mots / In some words:
The filtering of experimental measurements carried out on a turbulent fluid was done until now by linear digital techniques. In order to turn to a non-linear filtering, we have developed stochastic modelings of measurement on a fluid. By studying the measurements made with a mobile sensor in a random medium, we have defined and give the properties of the acquisition process of a vector field along a random path. Then, we have deeply modified the Lagrangian models of fluids proposed by the physicists to make them compatible with the problem of filtering. These models initiated by S.B. Pope belong to the class of McKean-Vlasov equations with mean field. The closure of these equations was obtained by conditioning the Markovian dynamics to the observation and to the acquisition process along the sensor path.

For the stochastic filtering of the conditioned acquisition process, we have proposed new algorithms of non-linear filtering for mean-field processes and for some various types of laws. We prove the convergence of the particle approximation for each new algorithms of estimates we gave.

The resulted filtering algorithms are based on the dynamics of genealogical trees where the processes interact by genetic selections and by their mean-field law.

Finally, this innovative work allowed us to filter velocity measurements of a turbulent fluid. We present several applications of our methods by using some 1D, 2D or 3D measurements, simulated or real. Our techniques make it possible to obtain high frequency estimates of the fluid velocities as well as quantities characterizing turbulence, and we proceed to a systematic study on numerical errors produced by our methods of calculation.

The second part of our research is linked to data assimilation. We study different non-linear techniques used in stochastic engineering to suggest new algorithms appropriate to high dimensional problems occuring in atmospheric data assimilation. In the end, we study the mathematical properties of this new estimators.


ANR Research Projects 2009-2011 :

ANR PREVASSEMBLE; Forecasting and Data assimilation. Subject : Particle Filter in Geosciences Applications with Pierre Del Moral


SESAR Joint Undertaking 2009-2016 :

SESAR J.U., Meteo-France affiliate of the DGAC/DSNA/DTI ; Work Package 4.7.1 and 4.7.2. Subject : New algorithms and stochastic estimators for the weather environment in the optimization of the commercial aircraft trajectory prediction problem


Publications

Workshop

Talks

Lectures


Background

Researcher employed by Meteo-France at the National Research Center, Toulouse since 2001.

Ph.D. degree, Paul Sabatier University, Toulouse, Speciality : Applied Mathematics, option Probability, 2008.

M.Sc in Applied Mathematics, Probability Option, Université Paul Sabatier, Toulouse, 2004.
Maitrise in Mathematical Engineering, Université Paul Sabatier, Toulouse, 2003.
Maitrise in Mathematics and Fundamental Applications, Université Pierre et Marie Curie, Paris 6, 1998.
B.Sc in Pure Mathematics, Université Pierre et Marie Curie, Paris 6, 1996.