The selection of likelihood function has great influence on the uncertainty of model parameters. In this study, taking the control basin of Dapoling Hydrological Station upstream of Huaihe River as an example, the GLUE method was combined with the integrated water system model (HEQM) and the optimization algorithm SCE-UA to analyze the influence of different likelihood functions (i.e., Nash-Sutcliffe coefficient and water balance coefficient) on the model parameter sensitivity, value distribution and uncertainty. The results showed that: From the sensitivity analysis, when the Nash-Sutcliffe coefficient is taken as the likelihood function, all the nine parameters selected are sensitive parameters, and when the water balance coefficient is a likelihood function, there is only one sensitivity parameter. The Nash-Sutcliffe coefficient is more suitable as a likelihood function；Seeing from the evaluation results of uncertainty interval, water balance coefficient was more suitable as a likelihood function, because its coverage and symmetry were better than those of Nash-Sutcliffe coefficient. This study was expected to not only provide some references for selecting likelihood function in GLUE method, but also to give the theoretical basis for weight assignment of the multi-objective GLUE method.