Examination of Geophysical Signatures of Ethanol-Water Mixtures in a Homogeneous Sand Column Using Ground Penetrating Radar and Time-Domain Reflectometry

In recent years there has been an increase in the frequency of incidents involving ethanol-blended gasoline affecting the groundwater. Near-surface geophysical methods hold promise for site characterization at ethanol contaminated sites. We attempted to record the broadband dielectric properties of ethanol at various concentrations as it was circulating through a closed tank containing either sand or a sand-clay mixture. Two high frequency GPR antennas were positioned symmetrically on either side of the sample holder to obtain transmission measurements. In the first part of the experiment we used a sample consisting of sand. In the second part of the experiment we used a sample containing sand mixed with 2% clay. Ethanol concentration was increased inside the sample holder by incrementally introducing additional pure ethanol. The sample holder was part of a closed circulation system that allowed the adjustment of ethanol concentration in the 0%-28% range. Prior to starting the experiments we established a dielectric permittivity relationship for liquid ethanol water mixtures in the range of ethanol concentrations between 0% and 100% using TDR. Using this relationship we inferred concentrations of the ethanol-water mixtures circulating through the sample holder by temporarily suspending the flow and collecting TDR measurements of the mixtures entering and exiting the sample. We evaluated the potential for GPR to predict ethanol concentrations based on a CRIM model in sand or sand-clay media. Ethanol concentration in the sand and sand-clay mixtures are well-predicted by the CRIM model when ethanol is added. However, the CRIM model does not succeed in predicting ethanol concentrations in either mixture undergoing ethanol withdrawal. During ethanol withdrawal, dielectric permittivity values exhibit hysteretic effects. Synthetic modeling exercises indicate that this behavior is not the result of a change in the fitting parameter within the CRIM model, and that the hysteresis observed is a physical process warranting further research