Modeling GPR data to interpret porosity and DNAPL saturations for calibration of a 3-D multiphase flow simulation

Abstract

Dense nonaqueous phase liquids (DNAPLs) are a pervasive and persistent category of groundwater contamination. In an effort to better understand their unique subsurface behavior, a controlled and carefully monitored injection of PCE (perchloroethylene), a typical DNAPL, was performed in conjunction with the University of Waterloo at Canadian Forces Base Borden in 1991. Of the various geophysical methods used to monitor the migration of injected PCE, the U.S. Geological Survey collected 500-MHz ground penetrating radar (GPR) data. These data are used in determining calibration parameters for a multiphase flow simulation. GPR data were acquired over time on a fixed two-dimensional surficial grid as the DNAPL was injected into the subsurface. Emphasis is on the method of determining DNAPL saturation values from this time-lapse GPR data set. Interactive full-waveform GPR modeling of regularized field traces resolves relative dielectric permittivity versus depth profiles for pre-injection and later-time data. Modeled values are end members in recursive calculations of the Bruggeman-Hanai-Sen (BHS) mixing formula, yielding interpreted pre-injection porosity and post-injection DNAPL saturation values. The resulting interpreted physical properties of porosity and DNAPL saturation of the Borden test cell, defined on a grid spacing of 50 cm with 1-cm depth resolution, are used as observations for calibration of a 3-D multiphase flow simulation. Calculated values of DNAPL saturation in the subsurface at 14 and 22 hours after the start of injection, from both the GPR and the multiphase flow modeling, are interpolated volumetrically and presented for visual comparison.

A borehole geophysical method for detection and quantification of dense, non-aqueous phase liquids (DNAPL) in saturated soils

“…DNAPL have very low permittivity values compared to water (2-4 as compared to 80) and are electrically resistive, leading to potentially strong GPR reflections off pooled DNAPL, and DNAPL may be detected in both unsaturated and saturated soils. GPR has been used to track DNAPL movement in soils (Brewster et al, 1995) and to interpret DNAPL saturation in soils (Sneddon, 2002). Like all surface geophysical methods, surface GPR surveys exhibit decreasing resolution with depth, and the effective penetration depth of GPR surveys can be fairly shallow, about 1-2 cm/Ωm of soil resistivity (Stewart, 1999).…”

DNAPL mapping by ground penetrating radar examined via numerical simulation

“…However, while the BHS model provides a more thorough treatment of the physics governing dielectric properties, it is only accurate for known geometries and cumbersome to implement for heterogeneous and multiphase materials (Martinez and Byrnes, 2001). This is because it is a two-component model and needs to be applied iteratively to analyse three-material mixtures, such as a DNAPL/ water/sand system (e.g., Hu and Liu, 2000;Sneddon et al, 2002).…”

“…Redman et al (1991) performed a small, controlled release of TCE and successfully monitored the ensuing changes in dielectric stratigraphy. Subsequently, 770 l of TCE was released into an impermeable cell constructed within a saturated sand aquifer with surface dimensions of 9 m × 9 m and with a depth of 3.3 m. The extensive, temporal GPR data set is described in a series of publications (Greenhouse et al, 1993;Kueper et al, 1993;Annan, 1994, Brewster et al, 1995;Birken and Versteeg, 2000;Sneddon et al, 2002). Brewster and Annan (1994) used 200 MHz antennas to collect a time series of GPR images for approximately 2 months after the TCE release.…”

Interpreting DNAPL saturations in a laboratory‐scale injection using one‐ and two‐dimensional modeling of GPR data

“…The initial procedure for 1D modeling of GPR data from the sand tank is similar to the procedures discussed in Sneddon et al (2002). In all the GPR models, the early time data (0 to 3 ns) is not modeled because these data represent near-field effects produced by the antenna layout that does not represent subsurface conditions.…”