Analysis Of Parameter Uncertainty And Its Transfer Applicability In Hydrological Modeling Using Multi-criteria GLUE Method

Hydrological modeling aims to represent the water,energy,and vegetation interactions at a required scale through governing mathematical equations,which is widely used in many fields such as the development and utilization of water resources and the study of ecohydrology.The parameters in these equations are one of the main sources of model simulation uncertainty.On the other hand,the lack of watershed data in the Predictions in Ungauged Basins(PUB)Global Program is also an important source of uncertainty in the simulation of hydrological models.As the main method of model uncertainty analysis,the Generalized likelihood uncertainty estimation(GLUE)method is widely used in the research of uncertainty.GLUE method based on multi-objective function is a research hotspot in recent years.In this paper,taking Tangmazhai Station as an example,the Liao River Basin is selected as the research area with the Hydrological,Ecological and water Quality Model(HEQM)model,and ambiguity function strategy.The Pareto strategy are used to analyze the multi-criteria GLUE which contains four objective functions(Nash-Sutcliffe Coefficient:NSCE,water balance:WB,root mean square error:RMSE and certainty coefficient:R~2).Combining these objective functions to form different likelihood functions,considering the influence of likelihood functions and likelihood function thresholds on the parameter uncertainty of multi-criteria GLUE.On the basis,the hydrological parameters of the ungauged regional in the Liao River Basin were filled by the method of parameter transplantation,and then the multi-criteria GLUE method was used to analyze the applicability and accuracy of the parameter transplant in the Liao River Basin.The results shows:two likelihood functions is better than the three results.The best combination is|WB|and R~2 with thresholds of 0.35 and 0.75;Likelihood function has a great impact on multi-criteria GLUE.The most sensitive one is WB,and the least sensitive is R~2.The best threshold of|WB|is equal to 0.35 when likelihood functions are the same and the thresholds are different,the lower the threshold requirements of NSCE and R~2,the better the uncertainty of the multi-criteria GLUE evaluation parameters,but the result of WB is opposite;In the ambiguity function theory,under the same conditions of the threshold and the likelihood functions,the unequal weight under each scenario is better than the equal-weight result,indicating that the weight distribution in the ambiguity theory is very important.The Pareto strategy GLUE method of the four objective functions that are not affected by the threshold value is superior to the GLUE method of the ambiguity function theory of the two objective functions,indicating that the threshold has a great influence on the uncertainty of the GLUE analysis parameter,and RMSE is a sensitive objective function.Attribute similarity method selects seven attributes based on watershed attributes and hydrological characteristics,performs runoff simulation in Liao River basin,and applies the multi-criteria GLUE method to verify the applicability of attribute similarity method.Uncertain interval evaluation results show that except for Tongliao station,the results of the other five stations such as Tangmazhai and the Xingjiawopeng were better,which proved that the attribute similarity method was applicable to the Liao River Basin.At the same time,the spatial distribution map of stable infiltration rate and runoff coefficient in the whole basin verified once again that the attribute similarity method is applicable to the Liao River Basin.