On the value of conditioning data to reduce conceptual model uncertainty in groundwater modeling

in Water Resources Research (2010), 46(8), 08520

Recent applications of multi-model methods have demonstrated their potential in quantifying conceptual model uncertainty in groundwater modeling applications. To date, however, little is known about the ... [more ▼]

Recent applications of multi-model methods have demonstrated their potential in quantifying conceptual model uncertainty in groundwater modeling applications. To date, however, little is known about the value of conditioning to constrain the ensemble of conceptualizations, to differentiate among retained alternative conceptualizations, and to reduce conceptual model uncertainty. We address these questions by conditioning multi-model simulations on measurements of hydraulic conductivity and observations of system-state variables and evaluating the e ffects on (i) the posterior multi-model statistics and (ii) the contribution of conceptual model uncertainty to the predictive uncertainty. Multi-model aggregation and conditioning is performed by combining the generalized likelihood uncertainty estimation (GLUE) method and Bayesian model averaging (BMA). As an illustrative example we employ a 3-dimensional hypothetical system under steady-state conditions, for which uncertainty about the conceptualization is expressed by an ensemble (M) of 7 models with varying complexity. Results show that conditioning on heads allowed for the exclusion of the two simplest models, but that their information content is limited to further diff erentiate among the retained conceptualizations. Conditioning on increasing numbers of conductivity measurements allowed for a further reffinement of the ensemble M and resulted in an increased precision and accuracy of the multi-model predictions. For some groundwater flow components not included as conditioning data, however, the gain in accuracy and precision was partially o ffset by strongly deviating predictions of a single conceptualization. Identifying the conceptualization producing the most deviating predictions may guide data collection campaigns aimed at acquiring data to further eliminate such conceptualizations. Including groundwater flow and river discharge observations further allowed for a better diff erentiation among alternative conceptualizations and drastic reductions of the predictive variances. Results strongly advocate the use of observations less commonly available than groundwater heads to reduce conceptual model uncertainty in groundwater modeling. [less ▲]

Conceptual model uncertainty in groundwater modeling: Combining generalized likelihood uncertainty estimation and Bayesian model averaging

in Water Resources Research (2008), 44

Uncertainty assessments in groundwater modeling applications typically attribute all sources of uncertainty to errors in parameters and inputs, neglecting what may be the primary source of uncertainty ... [more ▼]

Uncertainty assessments in groundwater modeling applications typically attribute all sources of uncertainty to errors in parameters and inputs, neglecting what may be the primary source of uncertainty, namely, errors in the conceptualization of the system.Confining the set of plausible system representations to a single model leads to underdispersive and prone-to-bias predictions. In this work, we present a general and flexible approach that combines generalized likelihood uncertainty estimation (GLUE) and Bayesian model averaging (BMA) to assess uncertainty in model predictions that arise from errors in model structure, inputs, and parameters. In a prior analysis, a set of plausible models is selected, and the joint prior input and parameter space is sampled to form potential simulators of the system. For each model, the likelihood measures of acceptable simulators, assigned to thembased on their ability to reproduce observed systembehavior, are integrated over the joint input and parameter space to obtain the integrated model likelihood. The latter is used to weight the predictions of the respective model in the BMA ensemble predictions. For illustrative purposes, we applied the methodology to a three-dimensional hypothetical setup. Results showed that predictions of groundwater budget terms varied considerably among competing models; despite this, a set of 16 head observations used for conditioning did not allow differentiating between the models. BMA provided average predictions that were more conservative than individual predictions obtained for individual models. Conceptual model uncertainty contributed up to 30% of the total uncertainty. The results clearly indicate the need to consider alternative conceptualizations to account for model uncertainty. [less ▲]

Stochastic modelling of the hydrogeological environment in low permeability sediment

in Proceedings of IAMG 2003 (2003)

In Belgium, the Boom Clay is being considered as a potential host formation for the disposal of nuclear waste. Part of the safety assessment and feasibility studies of a potential nuclear waste disposal ... [more ▼]

In Belgium, the Boom Clay is being considered as a potential host formation for the disposal of nuclear waste. Part of the safety assessment and feasibility studies of a potential nuclear waste disposal consists of hydrogeological modeling. In order to model the groundwater flow and possible radionuclide transport in the clay, the spatial distribution of the hydraulic conductivity of the clay has to be assessed. In this study, geostatistical methods are used to characterize the hydraulic conductivity field. More specific, direct sequential simulation of the hydraulic conductivity is carried out, using measurements of hydraulic conductivity and 4 types of soft data or secondary variables: resistivity logs, gamma ray logs, grain size measurements and descriptions of the lithology. The primary and secondary information is analyzed with geostatistical tools and combined to generate 100 fields of the hydraulic conductivity of the Boom Clay. Next, each field is input to a groundwater flow model to predict the advective travel time of constituents released from the disposed waste in the Boom Clay to the aquifers surrounding the Boom Clay. Statistical analysis of the ensemble of model predictions results in a predictive distribution for the advective travel time. This distribution reflects the uncertainty of the advective travel time that results from the uncertainty of the spatial distribution of the hydraulic conductivity of the Boom Clay [less ▲]