Quantifying hydrological modeling errors through a mixture of normal distributions

Details

Serval ID
serval:BIB_14007D89F4E9
Type
Article: article from journal or magazin.
Collection
Publications
Title
Quantifying hydrological modeling errors through a mixture of normal distributions
Journal
Journal of Hydrology
Author(s)
Schaefli B., Balin Talamba D., Musy A.
Publication state
Published
Issued date
01/2007
Peer-reviewed
Oui
Volume
332
Number
3-4
Pages
303-315
Language
english
Abstract
Bayesian inference of posterior parameter distributions has become widely used in hydrological modeling to estimate the associated modeling uncertainty. The classical underlying statistical model assumes a Gaussian modeling error with zero mean and a given variance. For hydrological modeling residuals, this assumption however rarely holds; the present paper proposes the use of a mixture of normal distributions as a simple solution to overcome this problem in parameter inference studies. The hydrological and the statistical model parameters are inferred using a Markov chain Monte Carlo method known as the Metropolis-Hastings algorithm. The proposed methodology is illustrated for a rainfall-runoff model applied to a highly glacierized alpine catchment. The associated total modeling error is modeled using a mixture of two normal distributions, the mixture components referring respectively to the low and the high flow discharge regime. The obtained results show that the use of a finite mixture model constitutes a promising solution to model hydrological modeling errors in parameter inference studies and could give additional insights into the model behavior.
Keywords
Gaussian mixtures, Modeling error, Parameter uncertainty, Bayesian inference, Rainfall-runoff models, Metropolis algorithm
Create date
24/02/2009 10:27
Last modification date
20/08/2019 12:42
Usage data