Combining Generalized Likelihood Uncertainty Estimation (GLUE) and Bayesian Model Averaging (BMA) to Account for Conceptual Model Uncertainty in Groundwater Modelling

Conceptual model uncertainty is one of the most difficult problems to deal with in the practice of groundwater modelling. In recent years, several methodologies, based on the construction and calibration of alternative models, have been proposed to face this problem. In this article, a more general and flexible approach than those previously developed is described. We achieve this combining the generalized likelihood uncertainty estimation (GLUE) and the Bayesian model averaging (BMA) methodologies. Implementing the GLUE methodology ensures that a large set of acceptable simulators, i.e., conceptual models and parameter sets, are included in the analysis, therefore, avoiding compensation of the conceptual model errors and biased parameter estimates. Implementing the BMA approach allows the inclusion of previous knowledge about the system and the obtaining of consensus multi-model predictions. Preliminary results show that the approach provides a general and flexible framework to account for predictive uncertainty due to the specification of alternative conceptual models.

ROJAS Rodrigo;  FEYEN Luc;  DASSARGUES Alain; 

2008-02-29

Geological Survey of Denmark and Greenland (GEUS)

JRC43542

https://publications.jrc.ec.europa.eu/repository/handle/JRC43542,