Home > News > Degree Defenses > Sylvain Kuppel's PhD Defense (LSCE)

PhD Defense

Sylvain Kuppel (LSCE)

Title : Assimilation de mesures de flux turbulents d'eau et de carbone dans un modèle de la biosphère continentale

Date and time : The 14-12-2012 at 14h00

Type : thèse

Université qui délivre le diplôme : UVSQ

Location : LSCE - Orme des Merisiers, Bat. 701, Pièce 17C - Gif sur Yvette
Members of jury :

Mme. Sylvie Thiria (UVSQ, Versailles) : Examinatrice
M. Jean-Christophe Calvet (CNRM-Meteofrance, Toulouse) : Rapporteur
M. Sönke Zaehle (Institut Max Planck, Jena) : Rapporteur
M. Olivier Talagrand (LMD, Paris) : Examinateur
M. Christophe François (Université Paris Sud, Orsay) : Examinateur
M. Frédéric Chevallier (LSCE, Gif-sur-Yvette) : Directeur de thèse
M. Philippe Peylin (LSCE, Gif-sur-Yvette) : Invité

Summary :

Assimilating terrestrial ecosystem data allows to fill the gap between experimental observations and the ecophysiological theory as embodied in process-based land surface models. In this thesis, we have sought to use the information content of daily-averaged eddy covariance measurements of net carbon dioxide (NEE) and latent heat (LE) fluxes made at the sites of the FLUXNET global network, aiming at optimizing key parameters of the ORCHIDEE vegetation model and at refining the uncertainty budget of the modeled fluxes. A Bayesian statistical approach has been chosen, so as to account for knowledge of the uncertainties related to the prior model parameters, the model equations, and the measurements. The major line of work has been to develop a data assimilation method where the observations from several flux sites are simultaneously used as a constraint, in order to optimize a common set of parameters considered as generic within a given broad class of ecosystem. We found that this "multi-site" approach significantly improves the model-data agreement at seasonal and annual time scales, with performances fairly similar to those brought by site-scaled, independently optimized parameters ("single-site" approach). We notably discuss the fact that the multi-site parameters' values are not always a mere average of their single-site counterparts. Making use complementary data, we have also evaluated the ability of the multi-site optimizations to improve the simulation of the gross carbon fluxes (photosynthesis and respiration) at the site level. At the global scale, the phenology of the leaf cover has been compared to that brought by satellite measurements from the MODIS instrument. The seasonality of the global net biospheric carbon fluxes has been evaluated after optimization using ground-based measurements of the atmospheric CO2 concentration, after applying a transport model. Lastly, we have developed a method to rigorously infer the statistical structure of the error stemming from inadequate and/or missing process representations in a global vegetation model such as ORCHIDEE. This last work indicates that this "model error" must be explicitly accounted for in the uncertainty budget related to carbon cycle modeling.

Contact :
Return to list of degree defenses