Water Resources Research Act Program

Details for Project ID 2010LA76G

Hierarchical Multimodel Saltwater Intrusion Remediation and Sampling Designs: A BMA Tree Approach

Institute: Louisiana
USGS Grant Number:
Year Established: 2010 Start Date: 2010-09-01 End Date: 2013-08-31
Total Federal Funds: $217,645 Total Non-Federal Funds: $228,868

Principal Investigators: Frank Tsai, Jeff Hanor

Abstract: The goal of the project is to develop concurrent remediation designs and sampling designs to mitigate saltwater intrusion problems. In spite of being surrounded by abundant surface water, Louisiana vitally relies on high-quality, low-cost groundwater to sustain its economic development. Groundwater serves more than a half million Louisiana residents via privately owned individual wells and almost three million residents via groundwater-supplied community water systems. Excessive withdrawal for various economic activities have caused significant saltwater intrusion in Louisianas major aquifer systems, including the Chicot aquifer system extending to east Texas, the Sparta aquifer system extending up into southern Arkansas, and the Southern Hills aquifer system that covers the Baton Rouge Capital area and surrounding parishes. The project focuses on saltwater intrusion remediation and sampling designs in the Southern Hills aquifer system in the Baton Rouge area. Due to limited amounts of data, robust remediation designs need to account for model uncertainty at high cost in order to reduce risk and failure of recommendations. The project introduces Bayesian model averaging (BMA) statistics to combine multiple saltwater intrusion models in probabilistic optimal remediation designs, which include a saltwater intrusion barrier (SIB) system (injection wells) and scavenger well operation (SWO) system (extraction wells). Specific studies for remediation designs include: (i) multimodel saltwater encroachment prediction in the designs, (ii) BMA trees of prediction and uncertainty analysis for hierarchical BMA models, (iii) multimodel uncertainty propagation to model prediction, and (iv) BMA chance-constrained formulation for risk analysis. To reduce model uncertainty and remediation cost, concurrent sampling designs are recommended to collect new hydrological and geophysical data to improve model structure accuracy. To reduce BMA within-model uncertainty as well as between-model uncertainty, the project proposes (1) optimality designs and (2) model discrimination designs. The optimality designs aim to reduce within-model uncertainty. Specially, the project will consider D-optimality and A-optimality for information maximization to reduce model parameter uncertainty, which accounts for fitting errors, parameterization schemes, and prior parameter uncertainty. Through the uncertainty propagation, within-model uncertainty is reduced. The model discrimination designs aim to reduce between-model uncertainty. We will focus on solving the model identifiability problem by collecting new data to distinguish models from each other. Specifically, we will consider interval identifiability and Occams window to discard poor models. The two types of sampling designs involve location searching and sampling schedule. Research, management practices, and education will benefit from this project. The project outcomes will provide information to visualize and predict performance of clean-up scenarios before implementing the actual work. Moreover, the management model will be distributed to user groups who are coping with saltwater intruded areas. The project will directly benefit the university researchers and the USGS through collaborative activity and has a potential positive impact on education, public awareness, and government administrations. The project will assist the USGS in developing cost-effective strategies for water availability information to Congress and the public. Also, the management model will serve as an educational and research tool to promote the understanding of the critical importance of groundwater management to undergraduate and graduate students.