A Bayesian perspective on input uncertainty in model calibration: Application to hydrological model ''abc''

摘要: [1] The impact of input errors in the calibration watershed models is a recurrent theme water science literature. It now acknowledged that hydrological are sensitive to measures precipitation and those bias model parameters estimated via standard least squares (SLS) approach. This paper presents Bayesian uncertainty framework allowing one account for input, output, structural (model) uncertainties model. Using this framework, we study on ‘‘abc.’’ Mostly academic interest, ‘‘abc’’ has response linear its closed form integration nuisance variables under proper assumptions. analytical solutions compute posterior density parameters, some interesting observations can be made about their sensitivity errors. We provide an explanation identified SLS approach show error context prior ‘‘true’’ value significant influence parameters’ density. Overall, obtained from method more accurate, over them realistic than with SLS. method, however, specific models, while most display strong nonlinearities. Further research thus needed demonstrate applicability commonly used models.