Study On The Uncertainty Of Hydrological Model Simulation

There are many uncertain factors affecting hydrological simulation model,using the VIC model as the tool,the typical watershed of Weihe basin in the middle latitude region as the research object,this paper discusses the response of the river basin to various uncertain factors.In this paper,with mathematical support,combined with the hydrology,computer science,number theory,statistics,meteorology,botany,geotechnical engineering and other disciplines.With the combination of theoretical analysis and model simulation,from the analysis of runoff regularity,the uncertainties of the optimal parameters of the model parameters,the uncertainties of the optimal range of the model parameters,the upper boundary conditions of the model(ie,the uncertainty of the climate factors),the lower boundary conditions of the model(ie,the uncertainty of the vegetation change)fully described the causes uncertainty and identification method.The main research results of this paper are as follows:(1)Deterministic research is the basis of uncertainty research.This paper uses several traditional hydrology statistical methods such as correlation analysis,wavelet transform,Markoff technology,Kendall rank test,Mann-Kendall method,R/S analysis method etc for the analysis of Weihe River Basin runoff.This paper argues that any study of uncertainty should proceed from deterministic factors,from known to unknown,as a general step in the study of hydrological uncertainties.(2)Parameter estimation is an important part in hydrological modeling of watershed.In this paper,the problem of the optimal solution of the hydrological model parameters is summed up as the first sort of parametric uncertainties.The substance of the model is finding the optimal solution of the nonlinear model when the objective function is nonnegative.The general method is given in judging the linear / non-linearity of the model,and the principle and method of determining the convexity / nonconvexity of the objective function are given based on the response surface.Based on this conclusion,a method of homogeneity design is introduced into the hydrological model parameter estimation.The method is very suitable for open-sourcehydrological models based on manual and semi-automatic parameter estimation and effectively reducing the computational spatio-temporal complexity.(3)The second kind of parameter uncertainty is based on Bayesian statistics.The essence is to solve the optimal distribution interval of each parameter.In this paper,an information diffusion theory based on small sample is introduced to estimate the parameter interval of the hydrological model,which can effectively solve the bottleneck of the traditional Monte Carlo sampling data and the long operation time.For the VIC model and modeling watershed applied in this paper,the optimal distribution interval of each parameter is characterized by only 27 times in sampling,and it is consistent with the hydrological and physical phenomena.The method is theoretically adapted to the estimation of the parameters of various hydrological models.(4)The hydrological model and climate models are combined to discuss the uncertainty of the upper boundary conditions on the model in this paper.This paper analyze change of the boundary conditions in the basin,namely,the precipitation and temperature in the future,based on the measured data of the CO2 emission from the river basin,and the data of the urbanization rate and the rate of industrialization for many years.The two scenarios(RCP 4.5 and RCP8.5)under the three models(CanESM2?BCC?MIROC-ESM)are used as the basis for the study.The statistical downscaling method,Delta method and Delta-DCSI method are used to reconstruct the future meteorological data,and the result show that the improved Delta-DCSI method can reflect the variability of meteorological factors in the sub-grid,which is closer to the actual situation than the traditional Delta method.The results show that the future runoff of the statistical downscaling method and Delta-DCSI method in three models had a tendency of first decrease then increase in RCP 4.5 scenario,while they had a decrease tendency in the RCP8.5 scenario.(5)The uncertainty of the lower boundary conditions are analyze in this paper,due to the relevance between vegetation coverage with the upper boundary conditions.The RPC4.5 and RCP8.5 scenarios under three climate models(CanESM2,BCC,MIROC-ESM)combined with statistical downscaling and Delta-DCSI method are applied respectively to generate climate factors,and then combined with the Holdridge life zone classification method to predict future watershed vegetation changes,and the prediction results are used to driven VIC model.The increase of greenhouse gas emissions leads to the increase of runoff when the upper boundary condition works.When the upper boundary and the lower boundary condition are coupled,the degradation of vegetation leads to the increase of runoff.The effect of vegetation change is greater than that of greenhouse gas emissions on runoff variation.