2009 Portland GSA Annual Meeting (18-21 October 2009)

Paper No. 12
Presentation Time: 11:10 AM

PARAMETRIC SHAPE-BASED INVERSION IN ELECTRICAL IMPEDANCE TOMOGRAPHY FOR THE CHARACTERIZATION OF SUBSURFACE CONTAMINANT DISTRIBUTIONS


AGHASI, Alireza1, MILLER, Eric Lawrence1, RAMSBURG, Andrew2 and ABRIOLA, Linda3, (1)Electrical Engineering, Tufts University, Department of electrical Engineering, 161 College Ave, Tufts University, Medford, MA 02155, (2)Civil and Environmental Engineering, Tufts University, 200 College Avenue, Anderson Hall, Tufts University, Medford, MA 02155, (3)Department of Civil & Environmental Engineering, Tufts University, 200 College Avenue, Medford, MA 02155, elmiller@ece.tufts.edu

In recent years, there has been considerable interest in the development of geophysics-based methods for characterizing the spatial distribution of subsurface contamination in DNAPL source zones (i.e., source zone architecture) and tracking remediation progress. For such problems, electrical impedance tomography (EIT) is one common sensing modality. Subsurface images formed using EIT however tend to be low resolution reflecting the fact that impedance tomography is a highly ill-posed inverse problem. In particular, EIT images do not typically provide a sharp boundary between the NAPL and the surrounding subsurface.

In this talk, we build on recent advances in geometrically-based inverse methods to improve the information content in EIT images for better characterization of the source zone. Rather than inverting for a dense collection of pixel or voxel values, we use so-called “level-set inverse methods” to directly estimate (a) the boundary separating contaminated regions from the nominal medium as well as (b) a low order representation of the spatial distribution of the contaminant's electrical properties. We introduce parametric representations of the level set function to further reduce the dimensionality of the inverse problem. Using this approach, there is no need for regularization parameters. Applications to numerically generated data sets demonstrate the utility of this modeling and inversion method for pre-remediation DNAPL source zone characterization. Results illustrate how this approach could substantially improve our ability to recover the geometry of the contaminant zone and allows for greater robustness in the face of spatial heterogeneity. Results also demonstrate that the method is computationally efficient, relative to traditional EIT inverse processing. Potential extensions to the tracking of remediation progress tracking are also discussed.