![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
Uncertainty analysis of a conceptual hydrological model by the Metropolis Hasting algorithm and GLUE method |
VerfasserIn |
Lu Li, Jun Xia, Chongyu Xu |
Konferenz |
EGU General Assembly 2010
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 12 (2010) |
Datensatznummer |
250036738
|
|
|
|
Zusammenfassung |
Uncertainty assessment of hydrological models is one of the most important and difficult
topics in hydrology research now. Many studies using different methods have been carried
out and different results are reported in the literature. Uncertainty assessment methods can be
broadly classified into two groups, i.e. the Bayesian method and the Generalized Likelihood
Uncertainty Estimation methodology (GLUE). Previously published results have shown that
the determination of the likelihood function is the key point in the Bayesian method
which determines the successfulness of the method. While the threshold values and
ranges of parameters are the important factors influencing the results of the GLUE
method.
This study makes a comprehensive evaluation about the parameter and predictive
uncertainty estimated by the GLUE and the Metropolis Hasting (MH) algorithm, which is
based on Bayesian inference, for a well-tested conceptual hydrological model (WASMOD) in
an arid basin of North China. For deriving the likelihood function for the MH algorithm, it is
convenient to have the simulation errors (1) normally distributed with zero mean and constant
variance, and (2) time independent. In this study the Normal Quantile transform (NQT) is
used to transform variable into a Gaussian distribution and the AR (1) Gaussian error model
is used to remove the time dependence of residuals. As for the GLUE method, the
dependence of the results on the threshold values and the ranges of the parameters are
investigated.
The results show that: (1) the parameter posterior distributions estimated by the
Metropolis Hasting algorithm are sharper and narrower than those by the GLUE
method; (2) the best Nash-Sutcliffe efficiency of the estimated discharges derived by
Metropolis Hasting algorithm and GLUE are nearly the same; and (3) however the
posterior distribution of parameters and the 95% confidence intervals of the simulated
discharge by GLUE are significantly impacted by the threshold values and the ranges of
parameters.
Key words: Metropolis Hasting, Bayesian method, GLUE, hydrological model, Normal
Quantile Transform, autoregressive, likelihood function |
|
|
|
|
|