Estimating biofuel contaminant concentration from 4D ERT with mixing models

Highlights

• Evaluate popular mixing models to estimate biofuel (ethanol) concentration from 4D electrical resistivity tomography.

• Observe the nature of EtOH as injected into a closed hydraulic system.

• Estimated EtOH concentration within 5% from temporal ERT data with a formation factor-based Lichtenecker-Rother model.

Abstract

We present the results of a lab-scaled feasibility study to assess the performance of electrical resistivity tomography for detection, characterization, and monitoring of fuel grade ethanol releases to the subsurface. Further, we attempt to determine the concentration distribution of the ethanol from the electrical resistivity tomography data using mixing-models. Ethanol is a renewable fuel source as well as an oxygenate fuel additive currently used to replace the known carcinogen methyl tert-butyl ether; however, ethanol is preferentially biodegraded and a cosolvent. When introduced to areas previously impacted by nonethanol-based fuels, it will facilitate the persistence of carcinogenic fuel compounds like benzene and ethylbenzene, as well as remobilize them to the ground water. These compounds would otherwise be retained in the soil column undergoing active or passive remediation processes such as soil vapor extraction or natural attenuation. Here, we introduce ethanol to a saturated Ottawa sand in a tank instrumented for four-dimensional geoelectrical measurements. Forward model results suggest pure phase ethanol released into a water saturated silica sand should present a detectable target for electrical resistivity tomography relative to a saturated silica sand only. We observe the introduction of ethanol to the closed hydraulic system and subsequent migration over the duration of the experiment. One-dimensional and three–dimensional temporal data are assessed for the detection, characterization, and monitoring of the ethanol release. Results suggest one-dimensional geoelectrical measurements may be useful for monitoring a release, while three-dimensional geoelectrical field imaging would be useful to characterize, monitor, and design effective remediation approaches for an ethanol release, assuming field conditions do not preclude the application of geoelectrical methods. We then attempt to use predictive mixing models to calculate the distribution of ethanol concentration within the measurement domain. For this study we examine four different models: a nested parallel mixing model, a nested cubic mixing model, the complex refractive index model (CRIM), and the Lichtenecker-Rother (L-R) model. The L-R model, modified to include an electrical formation factor geometry term, provided the best agreement with expected EtOH concentrations.

Introduction

Ethanol is a preferentially biodegraded compound and therefore delays the natural attenuation of other contaminants including the harmful BTEX compounds: benzene, ethyl-benzene, toluene, and xylenes (Caprio et al., 2007; Corseuil et al., 1998; Da Silva and Alvarez, 2002; Firth et al., 2014; Ma et al., 2013; MacKay et al., 2006; Powers et al., 2001; Ruiz-Aguilar et al., 2007). Ethanol also has cosolvency effects on existing non-aqueous phase liquids (NAPL) allowing transport and partitioning of harmful and otherwise immobile chemicals in the subsurface (Da Silva and Alvarez, 2002; Gomez and Alvarez, 2009; McDowell et al., 2003). Additionally, the degradation of ethanol results in methane production at potentially hazardous levels (Frietas et al., 2010; MacKay et al., 2006). Given the potential negative impacts of an ethanol release in conjunction with the increased consumption of ethanol worldwide, an effective means of detection, characterization, and monitoring potential releases is necessary. Soil and groundwater sampling through borehole investigations are costly and result in discrete single point data that can be difficult to interpret spatially. Geophysical measurements provide the opportunity to connect the direct but sparsely acquired spatial information, with indirectly measured high-resolution information resulting in a more complete picture for optimal site characterization and monitoring at minimal additional cost (Binley and Slater, 2020; Comenga et al., 2013; Glaser et al., 2021; Rucker et al., 2009a, Rucker et al., 2009b; Slater and Glaser, 2003; Ustra and Elis, 2019). Over the last couple of decades, it has become common for geoelectrical methods to be used for monitoring of NAPL natural attenuation which is reflected in the literature (Ajo-Franklin et al., 2004; Boleve et al., 2011; Halihan et al., 2017; Wang et al., 2020; Werkema et al., 2003). With sufficient contrasting electrical properties between background soil conditions and contaminant properties, NAPL extents both vertically and laterally can be mapped. Ideally, the indirect and direct methods are deployed concurrently and iteratively such that the geophysical surveys are informed by the borehole investigations, and vice versa, to constrain and maximize the subsurface information.

Relative to the resistivity of a saturated sand, ethanol, gasoline, and their mixtures are initially electrically resistive prior to biogeochemical alteration(Kirk, 1983; Personna et al., 2013a). Thus, in the presence of a conductive background, like saturated Ottawa sand, a measurable contrast for geophysical detection should be realized. Glaser et al., 2012, demonstrate results from a sister experiment designed to examine vadose zone migration of ethanol (EtOH) using ground penetrating radar (GPR). Here we inject ethanol (EtOH) mixed with a brilliant blue dye (BB), into a well characterized, saturated Ottawa sand in an instrumented, laboratory-scaled tank. While this study does not use gasoline, the ethanol-dye mixture is a close electrical analogue for E85, a commonly available alternative fuel comprised of 85% Ethanol and 15% gasoline (Kirk, 1983). Therefore, the physical property differences between ethanol mixtures and saturated Ottawa sand should allow geoelectrical methods to provide a means of differentiating between water-saturated pore spaces and ethanol-saturated pore spaces in the subsurface (Lucius et al., 1992). Further we attempt to estimate the concentration and distribution of ethanol in the sand tank through various mixing models using the geophysical data (Glaser, 2021).