Author(s): Xiankui Zeng; Dong Wang; Jichun Wu; Xiaobin Zhu; Lachun Wang
Linked Author(s):
Keywords: Groundwater model; Conceptual model; Uncertainty; Model averaging; Markov Chain Monte Carlo
Abstract: Parameter and conceptual model are the main sources of groundwater modelling uncertainty. Based on Markov Chain Monte Carlo (MCMC) and Bayesian model averaging (BMA), conceptual model uncertainty is assessed in this study. A Mean Euclidean distance based BMA method (MCMC-MED) is proposed, and is compared with the traditional method: summation based BMA method (MCMC-Sum), based on a synthetical model. Results demonstrate that the MCMC-Sum is more effective to identify the true model structure than MCMC-MED. However, MCMC-MED method obtains more consistent posterior probabilities of conceptual models than MCMC-Sum.
Year: 2013