Research On Hydrological Process Variability And Reservoir Operations Under Uncertain Environment

Reservoir operation,as one extremely complex system engineering,is the primary way to reallocate the time and space of streamflow and solve the imbalance between water supply and demand.As a whole,research on reservoir operations be launched from four levels,i.e.determining the variabiliity of hydrological cycle,simulating water r-esources,computing water requirments for different water sectors,and identifying the appropriate operations for water projects.Additionally,uncertainty accompanies the above-mentioned research,which induces significant impact on related results.The conductive and cumulative effect inside the system enhances the unascertainty of reservoir operation system.To improve the arrnti-jamming capability of reservoir operation system,reduce the risk of reservoir operation and enhance the reliability and practicability of research results,this study focuses on the above mentioned four levels and builds suitable model for quantifying the effects of different uncertainties on the basis of uncertainty sources recognition.Present study selects multiple catchments located in the Yellow River Basin as study regions,such as Jinghe River Basin,Weihe River Basin,the upper catchments of Yellow River Basin and so on.Primary contents of this study contain:(1)analyzing evolution characteristics of the relationship between hydro-meterological element:s;(2)studying bivariate return periods for extreme precipitation and flood peak;(3)revealing the impact of parameter w uncertainty of Budyko-based model on water availability assessment;(4)quantifying the impacts of model uncertainty and parameters uncertainty on streamflow simulation;(5)uncovering impacts of precipitation amount,times of precipitation events and randomness of precipitation process on crop irrigation water requirements;(6)building reservoir operations model and analyzing the impacts of parameters uncertainty of hydrologic model and randomness of precipitation process on reservoir operations and related benefits.Main conclusions drawn from this study are shown as follows.(1)Adopting Copula function-based logarithmic likelihood ratio method for detecting variability of precipitation-runoff relationship,and developing integrated ARMA-GARCH and Partial Copula model to identify the relationship variability between hydro-meterological factors.Results indicate that the Copula function is appropriate to model the nonlinear correlation between hydro-meterological factors;change point of precipitation-runoff relationship in the Jinghe River Basin occurs in 1996;ARMA-GARCH model is suitable to eliminate the effects of autocorrelation,conditional heteroscedasticity,and covariates.In addition,results show that detection results of evapotranspiration-runoff relationship change significantly after eliminating the effect of precipitation on evapotranspiration and runoff.(2)Applying the cross-wavelet method to revealing the changes of extreme precipitation-flood relationship,and presenting a univariate and copula-based bivarnate hydrological rnsk framework under uncertain environment which incorporates both flood control and sediment transport.Results indicate that the 1994-2014 and 1981-2014 are identified as periods with transformed flood-generating mechanism in the Weihe River Basin and Jinghe River Basin.Moreover,there is considerable sampling uncertainty affecting the univariate and bivariate hydrologic risk evaluation,which greatly challenges measures of future flood mitigation.The developed flood risk assessment framework can potentially help decision-makers in flood control and mitigation by providingvaluable insights toward assessing flood risk.(3)Developing non-parametric Bootstrap based method for disclosing parameter w uncertainty of Budyko model,and deriving analytical expression for describing impact of parameter w uncertainty on water avalibility assessment as a function of climatic conditions.The results indicate that there exists remarkable uncertainty in parameter w,which are 2.52(95%CI:[2.44,2.62])and 3.14(95%CI:[3.01,3.28])in the upper Yellow River Basin and Weihe River Basin,respectively.Impact of parameter w uncertainty on water avalibility assessment initially increases quickly with the increasing PE/P,and later decreases slowly as PE/P decreases.The impact reaches maximum when the ratio of potential evapotranspiration to precipitation(denoted as PEIP)are 1.37 and 1.2 in the UYRB and WRB,respectively.(4)Applying the MCMC algorithm and DREAM algorithm under Bayesian framework to investigate the parameters uncertainty of hydrological models,introducing conditional rando forest method for parameters sensitivity analysis,and adopting the one way repeated-measures anova to disclose impatcs of model uncertainty and parameters uncertainty on runoff simulation.Results show that parameters sensitivity of HyMod,GR4J and SIMHYD model varies with simulation period and the magnitude of runoff.Moreover,compared to the parameter uncertainty,impact of model uncertainty on runoff simulation is more pronounced.(5)Employing the compound Possion process to stochastically simulate daily precipitation,and carrying out the factorial design analysis of variance method to quantify the effects of precipitation amount,times of precipitation events and interaction effects between them on crop irrigation water demands compared to the randomness of daily precipitation.Results indicate that precipitation amount has the greatest effect on crop irrigation water demands,times of precipitation events following.The interaction effect of precipitation amount and times is little,compared to the effect of randomness of daily precipitation.Impact of randomness of daily precipitation on crop irrigation water demand is noteworthy,which causes large deviatation(197 mm)between the maximum and minimum crop irrigation water demand.(6)Building the integrated water resources management for food and power generation,and employing the probability box to deal the mixed uncertainty in runoff simulation,food and power generation,and reservoir operations.The Dongzhuagn reservoir located in the Jinghe River Basin is selected as the srtudy region.Results indicate that probability of runoff unexceeding 100 m3/s in 1998/07/10 is[0.55,0.65]at 95%confidence level;the randomness of daily precipitation process has greater effect on corn yield compared to the winter wheat,while the power generation varies 2.17?2.83 ×108 kW·h in 1998/07?1999/06.Probability of water level under different uncertainty sources are determined for the Dongzhuang reservoir.