State Water Resources Research Institute Program (WRRI)

Details for Project ID 2009LA61B, 2010

Bayesian Model Averaging for Saltwater Intrusion Management under Model Uncertainty

Institute: Louisiana
Start Date: 2009-03-01 End Date: 2010-07-31
Total Federal Funds: $16,909 Total Non-Federal Funds: $34,230

Principal Investigators: Frank Tsai

Abstract: The project goal is to study the saltwater intrusion problem in the “1,500-foot” sand aquifer, Louisiana under the consideration of model uncertainty. Model uncertainty contains model structure uncertainty and model parameter uncertainty. Due to lack of hydrogeological data, model structure uncertainty leads to non-unique conceptual ground water models. We often find a number of possible conceptual ground water models that explain the same hydrogeological processes. Conceptual models can be different in terms of boundary conditions, initial conditions, sink and source terms, and model geometry. Model parameter uncertainty arises from insufficient measurement data for estimating distributed aquifer parameters in ground water models. Parameterization methods are often used to estimate parameter values at unsampled locations. However, parameterization methods are not unique and traditional methods are lack of flexibility. Failing to recognize model uncertainty in the saltwater intrusion management problem could have a detrimental impact and may mislead remedial actions. To cope with model uncertainty in the management problem, the objectives of the proposal are (1) to develop a generalized parameterization method to increase parameterization flexibility and (2) to develop a multimodel approach to address the non-uniqueness problem. The project proposes to develop a generalized parameterization (GP) method specially to improve the hydraulic conductivity estimation. The GP method will increase the flexibility of traditional parameterization methods, incorporates other data types, and characterize spatial correlation of the estimated hydraulic conductivity field. Moreover, the project will adopt a Bayesian model averaging (BMA) method to address the non-uniqueness problem in saltwater intrusion models and parameterization methods. The BMA will integrate multiple saltwater intrusion models and multiple GP methods for prediction and management of saltwater intrusion. The project will implement the proposed methodology to the on-going saltwater intrusion problem in the “1,500-foot” sand aquifer in the Baton Rouge area.