Hydrological Uncertainty Quantification and Propagation in Multimodel Approaches
Author: Edom Melesse Moges
Publisher:
Published: 2018
Total Pages: 174
ISBN-13:
DOWNLOAD EBOOKModel uncertainties and inaccuracies can limit the application of hydrological modeling as decision making tool. Analysis and insight derived from uncertain models can significantly undermine the implication of their results, recommendations and conclusions. This thesis intends to deal with the various sources of uncertainty in hydrological modeling, particularly multi-modeling approaches, by using different statistical, computational and physically-based diagnostic measures. The uncertainty and the proposed approaches are evaluated using various hydrologic problems including -- extreme event frequency analysis, rainfall-runoff modeling, and coupled surface and subsurface models. First, the significance of model averaging, particularly Bayesian Model Averaging (BMA), is demonstrated by exploring extensive data, fundamental theory, and systematic diagnostic measures. Second, the study integrated hydrological signature measures and a multi model integration approach - Hierarchical Mixture of Experts (HME), in order to reduce structural uncertainty. Third, the study developed uncertainty quantification and propagation framework for coupled hydrological models that can readily be transferred to other coupled models. Using the framework, the study explored uncertainty propagation and their interplay in coupled hydrological models. The findings from this study -- in terms of developing a systematic uncertainty quantification framework and model diagnostic approaches -- are expected to improve the applications of hydrological and environmental models in understanding the underlying physical processes and making improved predictions.