Distributed hydrological modelling in California semi-arid shrublands: MIKE SHE model calibration and uncertainty estimation

Uncertainty assessment of integrated distributed hydrological models using GLUE with Markov chain Monte Carlo sampling

“…The general procedure in GLUE is to retain as behavioural solutions a given percentage of the sampled parameter sets (Lamb et al, 1998) or the solutions with a likelihood above a certain threshold (Freer et al, 1996;Mertens et al, 2004;McMichael et al, 2006). The latter method does not always provide uncertainty bounds capturing a satisfactory percentage of the observations.…”

“…Data requirements can be an obstacle for a proper validation, since, even if distributed measurements are usually available to physically characterize the catchment, spatially-distributed time series of observations of all the variables of interest are rare. This is one of the main reasons why, also in the cases where space-validation was conducted, distributed models have usually been calibrated and validated against discharge data only (Andersen et al, 2001;McMichael et al, 2006;Engeland et al, 2006). This method is inconsistent with spatially distributed modelling and it constitutes a limit to the performance of such models (Rosso, 1994).…”

“…The Generalized Likelihood Uncertainty Estimation (GLUE) technique by Beven and Binley (1992) is an alternative procedure, which has been extensively used for simultaneous calibration and uncertainty assessment of different distributed models (Lamb et al, 1998;Freer et al, 2004;Mertens et al, 2004;Muleta and Nicklow 2005;Cho and Beven, 2006;McMichael et al, 2006). Compared to other methods, GLUE has the advantages of being easy to implement and, at the same time, of allowing the simultaneous assessment of the total uncertainty present in all the components of the modelling process.…”

“…The latter feature is only of concern when running computationally expensive models, such as integrated, distributed hydrological models, since it limits the maximum number of simulations to be run (McMichael et al, 2006). The stochastic nature of the sampling scheme employed in GLUE (usually the Latin Hypercube Sampling, LHS, approach) can deplete considerably the goodness of the statistics inferred from the retained solutions, if the initial sampling of the parameter space is not dense enough to include many good solutions.…”

Nitrogen and phosphorus transformations and balance in a pond-ditch circulation system for rural polluted water treatment

“…GLUE has been applied in much of the practical simulation research of MIKE SHE (Christiaens and Feyen, 2002;McMichael and Hope, 2007;McMichael et al, 2006;Blasone et al, 2008;Vázquez et al, 2008Vázquez et al, , 2009. Currently, research on uncertainty analysis has significantly increased (Refsgaard and Henriksen, 2004;Rosbjerg and Madsen, 2005).…”