Composition du jury

Aaron BOONE, Directeur de recherche, Centre National de Recherches
Météorologiques, Rapporteur
Eleanor BLYTH, Directeur de recherche, Centre for Ecology & Hydrology,
Rapporteur
Laurent Li, Directeur de recherche, Laboratoire de Météorologie Dynamique,
Examinateur
Philippe CIAIS, Directeur de recherche, Laboratoire des Sciences du Climat
et de l’Environnement, Examinateur
Jan POLCHER, Directeur de recherche, Laboratoire de Météorologie
Dynamique, Directeur de thèse
Tao YANG, Perfesseur, Hohai University, Co-directeur de thèse

Résumé

Focusing on different uncertainty sources that can affect the model accuracy of hydrological modeling and impact analysis, this thesis reviews the past studies and provides new approaches for estimating and comparing the uncertainties with their applications concentrated over China. This thesis first proposes a three-dimensional variance partitioning approach that estimates the uncertainty among multiple precipitation products with different types. The new estimation uses full information in temporal and spatial dimensions and thus is a more comprehensive metric for uncertainty assessment especially for multiple datasets. This thesis then proposes a ORCHIDEE-Budyko framework that helps attribute the discharge bias between model simulation (provided by land surface model ORCHIDEE) and observations to uncertainty sources of atmospheric variables and model structure. The framework qualifies the possibility of different uncertainties with physical-based Budyko hypothesis and support of related literatures. This thesis finally reviews the human activities and their impact on river discharge over China regions as well as the related approaches that used for the quantification. The human impact that quantified as the difference between observed river discharge and the naturalized ones is then compared with multi-model simulations driven by different forcing inputs. Results show that the uncertainty in atmospheric variables (e.g., precipitation) is large especially for General Circulation Models (GCMs). Precipitation uncertainty is very likely larger than that of the model uncertainty. The uncertainty in the modeled discharge with different forcing is larger than the magnitude of human impact for most of the regions especially in south China, which impedes the credibility of human impact quantification for those regions. This understanding of uncertainties in the natural water cycle and the management humans impose on it is a prerequisite before attempting to model the anthropogenic pressures.