The Quantification And Partition Of Source Uncertainties Of Design Flood In Hydrology Risk Inference

Uncertainty widely exists in hydrological frequency analysis,and it is one of the hotspots in flood frequency research.In recent years,the uncertainty findings have mainly focused on model selection and parameter estimation,while less research related to sample uncertainty.It is necessary to quantitatively evaluate the influences of sources in uncertainties of designed flood on data and model factors and propose a reliable design flood.Taking the Thames River as an example,this study considers the uncertainty factors of the sample sizes and the probability distributions to establishes a two-factor multi-level orthogonal experiment with eight probability distributions and five sample size scenarios through random sampling methods.The design floods as response values are computed at a frequency of 0.995.And these values are used in the next whole procedure.This study is dedicated to identifying the influence of model and data factors uncertainties of the design flow through factor analysis and ANOVA,identifying the change characteristics in design values and design interval responses to the different sources of factor uncertainty,and revealing the design main effects and interaction effects of data and model factors in the design values,quantitating the significance of the influence of data and model factors and their interaction in the design values,clarifying the effect of different factor sources in the design value and the length of the design interval.The main results achieved are as follows:(1)New samples generate by resampling,and the parameters of different combinations of sample sizes and distributions were estimated through L-moments.The results of the K-S tests evaluated parameters are all good.The results of MEA and R2 of the goodness of fit evaluation are all small,indicating that the difference between the predicted value and the measured value is less,that also indicating the model fitting effects are all good.The research results show that the probability distributions such as Gev,Glo,Gno,Gpa,Pe3,Ln3,Gam andGum can fit the Thames daily maximum discharge sub-sample..(2)The sample and model factors have an impact on the uncertainty of the design value and design interval.In the same sample size,the design values predicted by each linear model are different.Results show that the Gum performs very well in terms of the length of the non-abnormal extreme value interval,the length of the interquartile range,and the median design flow rate.It is the optimal line type to describe the annual maximum daily flow of the Thames.In the same linear model,the study shows that the uncertainty decreases as the sample size increases.The non-abnormal extreme value and interquartile length of each line type are reduced obviously,and the fluctuation of the design median also decreases.(3)The variance results quantify the influence of different factor sources on the uncertainty of the design value and design interval.The sample factors,model factors and their interaction have significant effects on the uncertainty interval and the design value(P<0.01,except for the design mean sample main effect).The contribution rate of linear factors to the design value dominates,and the contribution rate of the sample factor,that is,of the sample size to the uncertainty interval of the design value dominates.In the division of the contribution rate the division of the contribution rate of the design interval,the contribution rate of the sample sizes to the change of the design interval length is relatively large,and it is close to 80%.