Hydrologic-Simulation Bias Correction And Uncertainty Quantification Based On Discrete Principal-Monotonicity Inference

Hydrologic risks such as droughts and floods are increasing as climate change intensifies,threatening the security and development of hydrologic systems.To reduce these risks,hydrological simulation was developed and gradually enriched.Hydrological simulation is an important way to explore and understand hydrological processes,and it can also assist hydrological forecast and analysis.However,the simulation results are always accompanied by system biases;if the simulation results are not corrected,there will be some risks in the water management under its guidance.Meanwhile,bias corrections are challenged by diverse uncertainties(e.g.,hydrologic models,bias-correction schemes,predictor selections,watersheds,streamflow magnitudes,and temporal scales).Therefore,this study attempts to provide comprehensive support for mitigating these risks through correcting biases in hydrologic simulation under these uncertainties.Primarily,the bias correction method of Discrete principal monotone inference(DiPMI)is developed,and the feasibility and performance advantages of the bias correction are explored with the application of Zhongzhou watershed in Guangdong Province and Xiangxi watershed in Hubei Province.The hydrologic models are the hymod model and the Xin'anjiang model,and the correction is compared with the traditional correction method(i.e.,linear regression)(e.g.,such as a variety of correction schemes and combination of predictors).It is found that DiPMI has better performance than LM in all correction conditions.Meanwhile,the original sequence of random sampling(SOR)scheme is better than other schemes in eliminating the bias of monthly scale hydrological simulation.In addition,the hydrological response of the two watersheds has a significant one-day lag.Secondly,to fully understand the bias uncertainties of the whole system of hydrological simulation,an integrated bias-corrected hydro-simulation uncertainty quantification(DiBHUQ)approach is developed based on DiPMI.This approach is based on the previous step,which applies the combination of multiple schemes,multiple prediction factors and multiple indicators for comprehensive analysis.The uncertainties evaluated can be categorized as method uncertainties and system uncertainties.The results show that the impacts of all uncertainties gradually decrease from hydrologic models,bias-correction schemes,watersheds,predictor selections,to indicators.All combinations of quantile intervals(i.e.,runoff magnitudes)and temporal scales follow this rank.The consistency distribution correction scheme(i.e.,SSR)and the consistency daily correction scheme(i.e.,SOR)provide the best bias-correction accuracies.Finally,to further quantify the main effect and interaction effect of uncertainty in the hydrological simulation correction system,Factorial Analysis(i.e.,FA)is used to comprehensively quantify all uncertainties in DiBHUQ.Among them,six uncertainty factors with different levels(i.e.hydrological model,bias correction scheme,predictor selection,watershed,discharge magnitude and time scale)are selected for analysis.It is found that the results of quantitative are basically consistent with the ranking of the uncertainty contribution of DiBHUQ.For example,to achieve the best response to NSE,SSR bias correction method should be used,and different prediction factors can be selected for different conceptualization processes.The non-consistency distribution correction(SSC)scheme can complement the lack of simulation accuracy of SSR scheme at the monthly scale.