Uncertainty analysis of a hydrological model using a multi-criteria likelihood measure within the GLUE method

摘要

A single-objective criterion was often used to assess the adaptability of a model in the past two decades, but this methodology can not appropriately show all the characters of hydrology via observed data at the same time. The GLUE (Generalized Likelihood Uncertainty Estimation) methodology for uncertainty analysis of hydrological models is a good example. Usually the single-objective criterion is the Nash-Sutcliffe coefficient. In this paper, a multi-criteria likelihood measure is presented within GLUE methodology, which consists of peak forecast error, runoff error, peak-time error and Nash-Sutcliffe coefficient. Based on the Xinanjiang Model, it was applied to LiangHui Reservoir. Results show that this multi-criteria likelihood measure has improved the real uncertainty of the hydrological model, which is very valuable for model calibration and uncertainty study. Comparison confirms that this multi-criteria GLUE methodology is superior to the single-objective criterion GLUE methodology.