Identifier

etd-11172005-162210

Degree

Doctor of Philosophy (PhD)

Department

Civil and Environmental Engineering

Document Type

Dissertation

Abstract

This study developed a coregionalized model to estimate hydraulic conductivity using spatial cross correlation between hydraulic conductivity and borehole geophysical data (a transform of the formation factor). An experimental pseudocross variogram is used instead of a cross variogram because data are not collocated. Experimental variogram uncertainty is investigated using confidence intervals for the experimental variogram calculated assuming variogram sills are lognormally distributed. These intervals are used for sensitivity modeling using kriging, cokriging, simulation and cosimulation. The hydraulic conductivity fields generated by kriging, cokriging, simulation, and cosimulation are then used in a high-resolution groundwater model created using telescopic mesh refinement (TMR) from a regional flow model of the Chicot Aquifer system in southwestern Louisiana. Results are analyzed to assess the significance of adding additional information (i.e., transform of formation factor), the process (i.e., kriging versus simulation and cokriging versus cosimulation) and variogram uncertainty on the groundwater flow model. Spatial images and flow predictions using regionalized models based on sparse conductivity data only are compared with coregionalized models using both conductivity and resistivity data, and the effects on model accuracy and robustness are discussed. Coregionalized model (i.e., cokriging) and simulation process (i.e., cosimulation) significantly affect groundwater flow model prediction. A new approach examines sensitivity of a capture zone groundwater model for the Chicot aquifer parameter uncertainty. Sensitivities to spatial variability of hydraulic conductivity, porosity, and aquifer thickness were investigated. The method calibrated aquifer properties to flow and geophysical data using cosimulation of hydraulic conductivity and formation factor, simulation for porosity, and kriging for aquifer thickness. Geostatistical model uncertainty was analyzed with a Bayesian method. Aquifer property models were scored using integral range to preserve correlation among variogram parameters. Variogram and pseudocrossvariogram models were selected from a lower bound, median, and upper bound of the posterior probability distribution of integral range. A steady-state two-dimensional groundwater flow model of the Chicot aquifer beneath Acadia Parish in Southwestern Louisiana examined capture zone sensitivity to spatial structure of aquifer properties. The capture zone model was insensitive to porosity variability and sensitive to hydraulic conductivity and aquifer thickness. The proposed method demonstrates the importance of model uncertainty compared with fluctuations of a fixed geostatistical model.

Date

2005

Document Availability at the Time of Submission

Release the entire work immediately for access worldwide.

Committee Chair

Clinton S. Willson

Share

COinS