Improved ERT imaging with 3-D surface-to-horizontal borehole configurations: relevance to dense non-aqueous phase liquids

SUMMARY

Accurate characterization and monitoring strategies are essential for designing and implementing remedial programs for sites polluted with dense non-aqueous phase liquids (DNAPLs). Electrical resistivity tomography (ERT) is a widely used geophysical technique for mapping subsurface features and processes of interest, and exhibits desirable characteristics for DNAPL sites due to its ability to gather large volumes of continuous subsurface information in a non-invasive, cost-effective and time-efficient manner. However, ERT measured only from the surface suffers from poor imaging quality with depth. Enhanced ERT imaging can be obtained via electrodes deployed on the surface and within horizontal boreholes, but so far it has only been investigated for 2-D imaging. This study evaluates the potential of 3-D surface-to-horizontal borehole (S2HB) ERT configurations for imaging 3-D DNAPL source zones. Laboratory tank experiments were first conducted with a 3-D S2HB ERT configuration, which consisted of a surface grid and a single borehole line of electrodes, being used to monitor DNAPL migration within porous media. Results demonstrate that 3-D S2HB ERT with a single borehole provides improved sensitivity at depth, and therefore enhanced imaging compared to conventional 3-D surface ERT. Further tank experiments were performed to assess the performance of single borehole S2HB ERT when (i) the distance between surface and borehole is increased, and (ii) additional horizontal boreholes are included. The S2HB ERT with a single borehole significantly outperforms surface ERT at larger depths, and performs comparably to S2HB ERT using multiple boreholes. This study suggests that 3-D S2HB ERT with a single borehole can provide the enhanced imaging ability needed to map DNAPLs, while also being relatively practical for implementation at field sites.

1 INTRODUCTION

The rehabilitation of sites contaminated by dense non-aqueous phase liquids (DNAPLs), including chlorinated solvents, creosote and coal tar, remains a challenging environmental task (Pankow & Cherry 1996). DNAPL source zones (SZs) in the subsurface are highly complex and heterogeneous and provide a long-term source for contaminating groundwater and the environment (e.g. Karaoglu et al. 2019; Koohbor et al. 2022). Accurate subsurface characterization is crucial to determine appropriate remediation strategies (Soga et al. 2004), and effective time-lapse monitoring is needed to track the performance of the implemented remedial program (e.g. Chambers et al. 2010).

Conventionally, subsurface characterization and monitoring at contaminated sites rely on invasive methods such as monitoring wells, core sampling and trial pits (e.g. Griffin & Watson 2002; Kueper et al. 2003; Steelman et al. 2020), which can be laborious, costly and provide only sparsely distributed point information. Further, the invasiveness of these methods is particularly problematic at contaminated sites as drilling and excavation can remobilize the contaminants (McMillan et al. 2018). These limitations have motivated the long-standing desire to use geophysical techniques at contaminated field sites (e.g. Brewster et al. 1995; Chambers et al. 2010; Trento et al. 2021). Geophysical techniques are non-invasive, cost-effective and can provide rapid and continuous spatial and temporal information, with the ability to image large volumes of the subsurface data over long time periods (e.g. Binley et al. 2015; DesRoches & Butler 2016; Slater & Binley 2021).

Electrical resistivity tomography (ERT) is a widely applied technique, particularly when it comes to field investigations (e.g. Power & Almpanis 2022; Dimech et al. 2022; Robinson et al. 2022) ERT measures the distribution of electrical resistivity in the subsurface, which corresponds to the variation of materials and processes, including soil type, water saturation and water chemistry (e.g. Loke et al. 2013; Su et al. 2024). Despite the electrical contrast between insulating DNAPL and conductive groundwater, and the long-standing desire to implement it at DNAPL sites, ERT is not commonly used. Due to their highly intricate architectures, DNAPL SZs provide highly challenging targets to accurately resolve (e.g. Kang et al. 2021; Almpanis et al. 2021a), with most success associated with the monitoring of subsurface changes during DNAPL remediation (e.g. Power et al. 2014; Trento et al. 2021).

Ongoing advancements in ERT instrumentation (e.g. Orlando & Renzi 2015), data acquisition (e.g. von Bülow et al. 2021) and data processing (e.g. Kim et al. 2013) are welcome, with some approaches already used to enhance DNAPL investigations, including 4-D time-lapse inversion (e.g. Power et al. 2014) and deep-learning strategies (e.g. Kang et al. 2021; 2023). ERT imaging can also be enhanced by increasing the proximity of the electrodes to the target. Surface ERT is the most widely used electrode configuration, deploying electrodes only along the ground surface, and while it can reasonably image more simplistic, shallow targets (e.g. Forquet & French 2012; Mohammed Nazifi et al. 2022), it is usually insufficient to adequately resolve the complexity of DNAPL SZs (e.g. Power et al. 2014; Folch et al. 2020). Electrodes have been deployed in vertical boreholes for cross-hole ERT (e.g. Wang et al. 2020; Almpanis et al. 2021b) and borehole-to-surface ERT (e.g. Lévy et al. 2019; Ochs et al. 2022); however, numerous boreholes (and significant effort/cost) are needed for adequate spatial coverage, especially as DNAPL SZs typically have high lateral extent (e.g. Power et al. 2014).

Electrodes can also be deployed within the subsurface along horizontal arrays within tunnels, horizontal boreholes, and via implantation. ERT surveys have been conducted with surface-to-tunnel (e.g. Simyrdanis et al. 2015) and tunnel-to-tunnel (e.g. Danielsen & Dahlin 2010; van Schoor & Binley 2010) configurations, while Kiflu et al. (2016) used direct-push technology to implant electrodes along horizontal lines at depth. For DNAPL studies, Deng et al. (2017) used 13 parallel horizontal survey lines along the full depth of a 2-D sandbox to monitor DNAPL migration; however, this configuration is wholly unrealistic for field application. Power et al. (2015) suggested taking advantage of horizontal remediation wells that are being installed at field sites to maximize the contact between injected remedial fluids and contaminants, with their study demonstrating improved DNAPL mapping by 2-D surface-to-horizontal borehole (S2HB) ERT relative to surface ERT.

Despite demonstrating enhanced resolution of interspatial targets, all previous configurations have only deployed electrodes in two parallel horizontal lines (i.e. 2-D cross-sectional slice). While the assumption of uniformity in the third dimension may suffice for some targets, the complexity and heterogeneity of DNAPL SZs requires 3-D imaging. As 3-D surface ERT typically comprises a surface grid of parallel lines, an ‘ideal’ 3-D S2HB ERT would suggest deploying horizontal borehole lines below every surface line. However, this would be highly laborious and costly, and negate the intended benefit of deploying horizontal boreholes. It is unknown whether adequate 3-D imaging can still be obtained if a 2-D surface grid is combined with a single horizontal borehole, and how this would compare to the ideal but intensive configuration comprising 2-D grids of both surface and horizontal borehole lines.

The objective of this study is to evaluate the performance of novel 3-D S2HB ERT for mapping DNAPLs. The most practical 3-D S2HB configuration, with electrodes along a 2-D surface grid and single horizontal borehole, was first investigated. Laboratory tank experiments were conducted on DNAPL migration through porous media, with the performance of 3-D S2HB ERT evaluated relative to 3-D surface ERT. Further tank experiments were performed for comparative analysis of different configurations of 3-D S2HB ERT, namely a 2-D surface grid combined in turn with one borehole, three boreholes, and a matching grid of boreholes. This study introduces 3-D S2HB ERT with differing configurations and demonstrates its improved imaging of 3-D DNAPL SZs relative to traditional 3-D surface ERT.

2 METHODOLOGY

2.1. 3-D ERT configurations

The most practical 3-D S2HB electrode configuration comprises a 2-D surface grid of parallel lines of electrodes that is underlain at its centre by a single horizontal borehole line of electrodes. Fig. 1 presents an example of this configuration, hereafter referred to as ‘S2HB-1BH’, with 11 parallel surface lines (Lines S1–S11) and one horizontal borehole line (Line HB6) directly below surface line S6. This configuration can be highly versatile as the single horizontal borehole can be combined with any number of easily deployable surface lines.

Other variations of the 3-D S2HB ERT configuration involve additional horizontal boreholes. The most extensive S2HB configuration would consist of a horizontal borehole line directly below each surface line, as illustrated in Fig. 1. This configuration, hereafter referred to as ‘S2HB-FULL’, would be highly complex, laborious, and costly to install many horizontal boreholes in parallel. Nevertheless, despite its impracticality, it is included in this study to showcase the maximum image resolution. A less intensive but more practical configuration for 3-D S2HB consists of three horizontal borehole lines distributed equally below the 2-D surface grid lines, hereafter referred to as ‘S2HB-3BH’ (Fig. 1). Although this study evaluates the comparative performance of S2HB-1BH, S2HB-FULL and S2HB-3BH ERT, their performance relative to 3-D surface ERT (i.e. 2-D surface grid of electrodes only) is also of high significance. It is noted that the 3-D surface ERT surveys were completed with measurements along each individual 2-D line, with no cross-line measurements.

Following comparative analysis of various electrode arrays, the pole-tripole array was selected for the S2HB ERT configurations due to its versatile resolving ability on a range of target geometries (e.g. Goes & Meekes 2004; Power et al. 2015; Simyrdanis et al. 2015). The pole-tripole array involves current electrode (A) on one line and current electrode (B) on the opposite line alongside the two potential electrodes (M and N), denoted as A-BMN. The measurement sequence contains A-BMN and then AMN-B. The dipole–dipole array is one of the most effective and widely used in field ERT studies (e.g. Power et al. 2018) and was used for the surface ERT configuration. The sensitivity maps of each S2HB and surface ERT configuration can provide an understanding of their spatial sensitivity and imaging ability (e.g. Dahlin & Zhou 2004).

2.2. S2HB-1BH versus surface ERT: DNAPL migration in porous media

2.2.3. DNAPL target geometries

Two experiments were conducted within a subvolume of a 3-D plexiglass tank (1.2 m × 1.2 m × 1 m), with DNAPL released into porous media and its subsequent migration simultaneously monitored by S2HB-1BH ERT and surface ERT. S2HB-3BH and S2HB-FULL measurements were not collected for these experiments as this would provide excessively long measurement times between DNAPL injection time steps, providing the possibility for unwanted DNAPL migration during measurement acquisition.

The subvolume for both experiments contained a zone of uniform medium sand with a specific geometry, which was fully encased in a lower permeability, uniform fine sand. Upon release into the medium sand zone, the DNAPL would preferentially flow within, and ultimately fill up, this zone, thereby providing control on DNAPL migration and allow its location to be estimated at any time during the experiments. Table 1 presents a summary of these experiments.

The two experiments differed in the geometry of the medium sand (see Fig. 2) to promote different lateral and vertical changes during DNAPL migration and allow a robust assessment of the different electrode configurations used in this study. The first experiment, referred to as Experiment D-1, consisted of a ‘staircase’ with three staggered steps of dimensions 0.22 m × 0.16 m × 0.05 m, while the second experiment (Experimental D-2) consisted of an ‘inverted-T’ with a horizontal bottom bar (0.22 m × 0.28 m × 0.05 m) overlain by the vertical portion (0.22 m × 0.11 m × 0.10 m).

2.2.2. Experiment preparation

The two sands used in these experiments were first sieved to retain only a single mesh size and provide a narrow particle size distribution with high uniformity. As shown in Table 2, the medium sand and fine sand exhibited a mean grain size of 0.58 mm and 0.23 mm, coefficient of uniformity of 1.36 and 1.56, porosity (ϕ) of 0.40 and 0.43, and hydraulic permeability (k) of 2.35 × 10⁻¹¹ m² and 8.07 × 10⁻¹³ m², respectively.

The tank was initially packed with the fine sand in lifts of 0.05 m to the elevation of the base of the experimental subvolume (i.e. 0.24 m depth from the final surface) to permit easier accessibility throughout the experiments. At this depth, the single horizontal borehole for S2HB-1BH ERT was installed, which consisted of a series of small electrodes (0.004 m long stainless-steel nails) attached to the outside of a polyvinyl chloride (PVC) pipe. The electrical cables, which were soldered to each individual electrode, were housed inside the pipe and connected via banana plugs to the resistivity meter strip box. The horizontal borehole was then carefully backfilled with fine sand. At the base of the bottom medium sand step (i.e. 0.20 m depth), two horizontal DNAPL injection wells, constructed from 10 mm diameter PTFE tubes that were only screened within the medium sand, were installed in parallel to each other and the lines of surface and borehole electrodes. Both wells were initially filled with DNAPL to ensure that no air existed in the wells immediately prior to DNAPL injection.

Rectangular stainless-steel dividers were constructed to the size of each segment of the medium sand zones in each experiment, specifically a single step of the ‘staircase’, and the bottom and top portions of the ‘inverted-T’. These dividers allowed for accurate and simultaneous packing of the medium sand and fine sand during the installation of each segment of the staircase and inverted-T DNAPL target zones (see Fig. 3a), after which the dividers were removed. A secondary DNAPL point injection well was placed near the top of each target zone (i.e. 0.05 m depth) to provide an extra source for DNAPL injection if needed. After the medium sand zone was installed, fine sand was packed to the surface, where 2-D grids of surface electrodes were placed (Fig. 3b).

For both Experiment D-1 and Experiment D-2, 11 parallel surface lines of electrodes were deployed with an inline electrode spacing of 0.04 m, and an interline spacing of 0.04 m (i.e. Lines S1–S11 in Fig. 1). The experiments only differed in terms of the number of electrodes per line. Experiment D-1 contained 16 electrodes (0.004 m long stainless-steel nails) on all surface lines and the single horizontal borehole line, providing line lengths of 0.6 m, while all lines in Experiment D-2 contained 21 electrodes for line lengths of 0.8 m. In each experiment, the single horizontal borehole line was centred below the surface grid (i.e. underlies Line S6). The 0.24 m distance between the surface and horizontal borehole lines is six times the electrode spacing. This lies within the rule of thumb proposed by Simyrdanis et al. (2015), who investigated the optimal depth for buried electrodes within tunnels, which is analogous to horizontal boreholes. It is noted that the minimum distance between the electrodes and tank walls/base (i.e. five times the electrode spacing) generated negligible boundary effects on the measurements.

The packed dry sands were then saturated via the slow injection of tap water into the base of the tank over 24 hr. A small concentration of salt was added to provide water conductivity resembling groundwater (12.8 Ωm; e.g. Power et al. 2015). Canola oil (Unico, Canada) was used in this experiment as a non-toxic surrogate for toxic industrial DNAPLs. The density of the canola oil was 920 kg m⁻³, meaning it was actually lighter than water. Therefore, while it is not a conventional DNAPL, it is referred to as a DNAPL throughout this study as it was used exclusively within the saturated zone (i.e. below the water table). The key property of canola oil in this study is that, like most DNAPLs, it is an insulating fluid and was used to displace the more conductive water in the pore spaces of the saturated zone. The oil was essentially neutrally buoyant, and migration was expected to be negligible when the injection pump was off during the short experiment duration (24 hrs total). The canola oil was dyed with Oil Blue N powder (Sigma-Aldrich, Canada) to record its final distribution more easily during post-experiment excavation (Fig. 3c). The properties of the water and canola oil used are summarized in Table 2.

2.2.3. Data acquisition and processing

In both experiments, DNAPL migration was separated into approximately equal volume intervals to progressively displace the porewater out of the staircase. The dyed DNAPL was placed in large syringes and injected at a flow rate of 10 ml min⁻¹ using a syringe pump (Fisherbrand, USA). Experiment D-1 consisted of six time steps: the first time step (T1) of the clean sand (i.e. 0 ml) and the second to sixth time steps (T2–T6) mapping the five successive injections of 300 ml until high DNAPL saturations resided within the pore volume of the staircase. Experiment D-2 consisted of five time steps: T1 at 0 ml, and T2–T5 for DNAPL injections of 420, 420, 360, and 300 ml, respectively. The dyed DNAPL was placed in large syringes and injected at a flow rate of 10 ml min⁻¹ using a syringe pump (Fisherbrand, USA).

The multichannel Syscal Pro Switch 48 resistivity meter (IRIS Instruments, France) was used to record the measurements of apparent resistivity. This instrument allows for high productivity measurements with a precision of 0.2 per cent and threshold voltage of 1 μV. The system features an internal switching board for 48 electrodes, and an internal 250 W power source. The acquisition time for resistivity measurements was 0.5 s, while strong ground coupling (ground resistance: <1 kΩ) was attained at all electrodes. The dipole-dipole (surface) sequence consisted of 157 and 266 measurements along each line in Experiments D-1 and D-2, respectively, while the pole-tripole (S2HB) sequence consisted of 496 and 583 measurements, respectively.

ResIPy is a user-friendly, open-source package that comprises a range of forward modelling and inversion capabilities (Blanchy et al. 2020) and can accommodate a range of surface electrode and borehole electrode based configurations (e.g. Moser et al. 2023; Yan et al. 2023). The recorded data in this study were inverted with the 3-D inversion code within ResIPy, which is finite element based. The domain was discretized into an unstructured tetrahedral mesh, with fine elements constructed around the electrodes and the mesh becoming coarser with distance from the electrodes. The resolution of the mesh was defined by a characteristic length of 0.02 m (i.e. 1/2 electrode spacing) and a growth factor of 4. The overall mesh was extended beyond the extent of the survey area, which is necessary for the half space problem.

The inversion in ResIPy uses an ‘Occam's’ type solution, which is based on a regularized objective function combined with weighted least-squares. It aims to yield a smooth distribution of electrical resistivities that fits the data, with the data misfit expressed as a root mean square. The code can also provide the 3-D cumulative sensitivity, which is the diagonal component of [J^TW^TWJ], where W is the data weight matrix and J is the Jacobian matrix (Binley & Kemna 2005; Binley & Salter 2020).

2.3. Variations to S2HB-1BH ERT configuration

To allow a comprehensive and flexible assessment of S2HB ERT, a second suite of experiments was performed with plastic boxes suspended within water. This approach still provided a changing resistive target (analogous to DNAPL) within a more conductive surrounding environment, but now the ERT experimental setup and/or targets could be quickly altered to more easily explore all 3-D S2HB configurations (i.e. S2HB-1BH, S2HB-3BH and S2HB-FULL) for a range of conditions. For these experiments, a plastic mesh platform was used to suspend the various plastic targets within the water (see Fig. 3d). A PVC frame was constructed to hold all electrode lines in place within the water, while incorporating flexibility for vertical and lateral movements. A summary of these experiments is shown in Table 1.

2.3.1. Deeper horizontal borehole

The inverted-T DNAPL target (Fig. 2b) was reconstructed with plastic containers suspended within water, which were gradually altered to provide three time steps: T1 consisted of only water (including the PVC frame and mesh platform), T2 added a plastic container to represent the bottom horizontal portion of the inverted-T, while T3 added the vertical portion to complete the inverted-T (see Fig. 3d).

The evolving target was first re-imaged with the same 3-D S2HB-1BH and surface ERT configurations used in Experiment D-2 with a borehole depth of 0.24 m. This was then repeated with the borehole depth now increased to 0.32 m (note: the target also increased in depth to maintain 5 cm distance from the base of the target to the borehole). This represents the scenario at field sites where the horizontal borehole would be installed as close as possible to the bottom of the DNAPL SZ. Deeper DNAPL SZs would normally rule out traditional surface ERT, so this experiment helps determine whether S2HB ERT would still be beneficial.

2.3.2. Additional horizontal boreholes

These experiments evaluated the performance of S2HB ERT when additional horizontal borehole lines were deployed. For the three-borehole configuration (S2HB-3BH), a borehole was added to either side of the central borehole used in S2HB-1BH. The horizontal boreholes were equally spread beneath the grid of surface lines, being placed directly underneath surface lines S3, S6 and S9 (see Fig. 1). For S2HB-FULL, a horizontal borehole was placed directly underneath every surface line.

The evolving inverted-T target was imaged by S2HB-1BH, S2HB-3BH and S2HB-FULL, with a borehole depth of 0.24 m. Although S2HB-1BH measurements combined each surface line (S1–S11) successively with the single borehole (i.e. S1–S11 with HB6), S2HB-3BH combined four surface lines with each successive borehole, ensuring overlap between these segments (i.e. S1–S4 with HB3, S4–S7 with HB6, S7–S11 with HB9). S2HB-FULL measurements combined each surface line with its corresponding borehole line (i.e. S1 with HB1, S2 with HB2).

2.3.3. Alternative target geometry

The inverted-T target was rotated 180 degrees to become a ‘standard-T’ shaped target. As the wider horizontal portion of the target is now shallower, this provides an ideal target to assess whether 3-D S2HB ERT still provides superior imaging to 3-D surface ERT on shallow targets. This target was gradually altered to provide three time steps: T1 consisted of only water, T2 added a plastic container to represent the bottom vertical portion of the standard-T, while T3 then added the top horizontal portion to complete the inverted-T.

3 RESULTS AND DISCUSSION

3.1. Sensitivity patterns of ERT configurations

Fig. 4 presents the 3-D cumulative sensitivity maps for each S2HB and surface ERT configuration within a homogeneous half space with the same dimensions of the experimental domain (i.e. 0.8 m × 0.4 m × 0.25 m). Each sensitivity map is normalized to the maximum sensitivity value across all datasets (log sensitivity = 3.75).

Surface ERT exhibits highest sensitivity near the surface electrodes before diminishing with increasing depth (i.e. red to dark blue). S2HB-1BH provides increased sensitivity at depth. The highest sensitivity is concentrated around the horizontal borehole, with sensitivity diminishing with increasing lateral distance from the borehole (i.e. lowest sensitivity in the corners). Additional boreholes provide increased lateral sensitivity, with S2HB-FULL exhibiting the highest cumulative sensitivity, as expected. As expected, all configurations exhibit reduced sensitivity at each end of the survey lines (x-distance = 0.0 and 0.8 m). The percentage of the subsurface area enclosed with sensitivity values greater than one (i.e. positive log sensitivity in Fig. 4) for surface, S2HB-1BH, S2HB-3BH and S2HB-FULL is 17, 23, 26 and 32 per cent, respectively.

3.2. DNAPL monitoring: 3-D S2HB-1BH versus 3-D surface

3.2.1. DNAPL staircase

Fig. 5 presents the inverted resistivity images from surface and S2HB-1BH ERT surveys at each time step of DNAPL migration within the staircase target zone. At T1 (i.e. 0 ml DNAPL), the small variability imaged by surface ERT and S2HB ERT is associated with the difference between the medium sand target zone and the surrounding fine sand. The fine sand has a greater porosity (0.43) than the medium sand (0.40), thereby having slightly lower resistivity.

Fig. 5a shows that surface ERT shows little change in resistivity until T4 (i.e. 900 ml of DNAPL injected), before the change increases in magnitude and extent at T5 and T6 as the DNAPL continues to get shallower and reside within the depth of investigation of surface ERT (i.e. ∼1/6 of line length = 0.1 m). S2HB-1BH ERT indicates resistivity changes at every time step, with the changing body coinciding with DNAPL gradually saturating the staircase from bottom to top (Fig. 5b).

Fig. 6 presents the time-lapse resistivity difference (in percentage) images between T1 and each of the five subsequent time steps (i.e. T2-T1, T3-T1, T4-T1, T5-T1 and T6-T1) to more clearly highlight the changes produced by DNAPL migration. Cross-sectional slices are presented, along with 3-D isovolumes representing the most significant changes (i.e. 100 per cent and above).

As shown in Fig. 6(a), surface ERT is unable to resolve any resistivity changes at T2 and T3 (i.e. 300 and 600 ml DNAPL, respectively) since the DNAPL first fills up the deeper, bottom step of the staircase. At T4 and T5, the DNAPL starts to migrate into the middle portions of the staircase and become more amenable to detection by surface ERT; however, it significantly underestimates the magnitude of the change, as shown by the small change in the corresponding isovolumes. It is only at T6, where DNAPL now resides in the shallow top step, that surface ERT accurately captures the resistivity change. In contrast, S2HB-1BH ERT shows enhanced resolving ability, capturing the extent and magnitude of the DNAPL changes at all time steps in both the 2-D cross-sectional images and 3-D isovolumes.

The post-experiment excavation confirms that the DNAPL invaded the entire volume of the staircase, except for the left-hand side of the top step. This corresponds to the lower resistivity change in this area (x-distance = 0.15 m) in Fig. 6 (T6-T1). The saturation of DNAPL ranges from 70 to 75 per cent throughout the staircase.

The T6-T1 images show the total resistivity change after all DNAPL was injected. Fig. 7 presents a bar chart that indicates the percentage of the total target volume (i.e. staircase) that is sufficiently captured by surface ERT and S2HB-1BH ERT. It is based on a resistivity change of at least 100 per cent (i.e. resistivity has doubled) to ensure only strong magnitude changes are considered, and matches the 3-D isovolumes for T6-T1 shown in Fig. 6. As shown, S2HB-1BH ERT captures a strong change in resistivity over a significantly greater portion of the target zone than surface ERT (82 per cent versus 56 per cent, respectively).

3.2.2. DNAPL inverted-T

Fig. 8 presents the surface ERT and S2HB-1BH ERT per cent difference images between the baseline (T1) and each subsequent time step (T2–T5) of DNAPL migration within the inverted-T target zone. Fig. 8a indicates that small resistivity changes are first detected by surface ERT after 840 ml DNAPL was injected (T3-T1), which coincides with the upward invasion of DNAPL into the vertical portion of the target zone. Subsequent injections of DNAPL and migration to shallower portions of the target zone are captured by surface ERT, although the magnitude of changes is severely underestimated. For instance, the 3-D isovolumes of 100 per cent resistivity difference indicates that only a small isovolume of significant change exists in the final time step (T5-T1).

As shown in Fig. 8b, S2HB-1BH ERT captures the entire invasion of the DNAPL through the target zone. The difference image after the first injection of 420 ml DNAPL (T2-T1) shows a distinct resistivity change at the depth of the bottom portion of the target zone. This change mainly occurs to the left of this bottom portion, suggesting that the DNAPL only flowed out of the initial perforations in the horizontal injection wells. Subsequent injections of DNAPL are all well-resolved by S2HB-1BH ERT, tracking DNAPL as it gradually fills up the inverted-T shape from the bottom left to the uppermost parts of the target zone. S2HB-1BH ERT provides more accurate imaging on the extent and magnitude of the DNAPL at all time steps, with the 3-D isovolumes confirming resistivity changes of at least 100 per cent at all time steps. S2HB-1BH ERT captures 78 per cent of the total DNAPL change (i.e. outline of inverted-T volume), while surface ERT only captures 25 per cent (see Fig. 7).

The S2HB-1BH ERT difference images indicate DNAPL changes outside the target zone at the bottom left of the target. These images also indicate that DNAPL does not fully invade the right-hand side of the bottom portion and the uppermost part of the top portion. The post-experiment excavation indicates that the final part along the length of the injection wells was unintentionally blocked by sand, and DNAPL only entered the target zone from the left-most part of the wells. As a result, the DNAPL was able to migrate upwards through the vertical portion of the target without needing to first saturate the entire bottom portion. Furthermore, due to the now localized injection in the bottom left, the injection pressure forced some of the DNAPL into the underlying fine sand, which is also confirmed by the excavation. This unintended loss of DNAPL volume from the target area also meant that the bottom right and top of the target did not have sufficient DNAPL volume to become fully saturated.

3.3. Variations to S2HB-1BH ERT configuration

3.3.1. Deeper horizontal borehole

Fig. 9 presents the results of the plastic and water experiments for the inverted-T target with a horizontal borehole depth of 0.32 m. Surface ERT is now completely unable to resolve either the addition of the bottom portion (T2-T1) or the subsequent addition of the top portion (T3-T1) of the inverted-T target zone. In contrast, S2HB-1BH ERT provides reasonable imaging of the resistivity differences for T2-T1 and T3-T1.

Fig. 7 indicates that the volume enclosed by at least 100 per cent resistivity change imaged by S2HB-ERT is 50 per cent of the volume of the target region (i.e. outline of inverted-T volume), while surface ERT only captures 1 per cent. It is acknowledged that even for S2HB-1BH ERT, the targets are not as well-resolved and pronounced as the 0.24 m borehole depth. Therefore, for even deeper DNAPL targets that would further increase the distance between surface and horizontal boreholes lines, another horizontal borehole line could be added above the target in a borehole-to-borehole configuration (e.g. Danielsen & Dahlin 2010).

3.3.2. Additional horizontal boreholes

Fig. 10 presents the S2HB-1BH, S2HB-3BH and S2HB-FULL ERT imaging of the evolving inverted-T target with plastic boxes and water. S2HB ERT imaging improves as the number of borehole lines increases from one borehole to three boreholes, and then to ‘full’ boreholes with a horizontal borehole underlying each surface line. As shown by the changes at T2, S2HB-1BH can identify the centre of the bottom portion of the target zone but underestimates its full magnitude and extent compared to S2HB-3BH and S2HB-FULL. For the full inverted-T target (T3), S2HB-1BH can again detect its general location and shape but underestimates its overall lateral extent, which is most evident in the 3-D isovolumes. In contrast, S2HB-3BH and S2HB-FULL provides improved characterization of the extent and magnitude of the target.

Fig. 7 indicates that S2HB-1BH captures 71 per cent of the total DNAPL change (i.e. outline of the inverted-T volume), while S2HB-3BH and S2HB-FULL captures 76 per cent. While the S2HB-1BH does not perform as highly as S2HB-3BH and S2HB-FULL, it still provides adequate characterization of the target. It is a trade-off between the more realistic and practical deployment of a single horizontal borehole with adequate resolving ability, or obtaining improved resolving ability but with impractical numbers of boreholes needed, specifically in the case of S2HB-FULL. It is evident that S2HB-3BH provides very similar results to S2HB-FULL, suggesting that a full array of boreholes may be unnecessary and that three boreholes may provide the additional image performance required for some sites.

3.3.3. Alternative target geometry

Fig. 11 presents the resistivity difference and 3-D isovolume images for the standard-T target zones. S2HB-1BH provides similar images to S2HB-FULL of this target, with both providing superior results to surface ERT. For example, S2HB-1BH and S2HB-FULL are able to resolve the bottom portion of the target at T2 with similar accuracy, whereas surface ERT is unable to resolve it. At T3, surface ERT is now able to resolve the top horizontal portion of the target with similar ability to S2HB-1BH and S2HB-FULL; however, it remains inadequate for resolving the complete standard-T target. Fig. 7 indicates that the volume representing at least 100 per cent resistivity change imaged by S2HB-1BH ERT and S2HB-FULL is 79 per cent of the volume of the target region (i.e. outline of standard-T), while the volume imaged by surface ERT is 70 per cent of the target.

Due to the geometry of this target and the respective sensitivity of each configuration, the results match expectations. While S2HB-FULL exhibits the most sensitivity over the entire experimental domain, S2HB-1BH has its most sensitive areas where the standard-T target is located (i.e. wide at the top, narrow at the bottom) (see Fig. 2). Therefore, S2HB-1BH is able to provide similar resolving ability. The shallow high sensitivity areas of surface ERT also coincide with the now shallow horizontal portion of the target, meaning it is better able to resolve it. However, the limited depth resolution is still evident as it completely underestimates, or even misses, the lower portion of the target.

4 CONCLUSIONS

This study evaluates the potential of 3-D S2HB ERT for imaging sites contaminated by DNAPLs, which provide a highly complex, heterogeneous 3-D target that can exist at large depths. Traditional ERT deploys electrodes along the surface, and while it is more convenient and requires less effort, it can be limited by declining sensitivity and resolution with depth. The use of electrodes within vertical boreholes for cross-hole ERT is the most common alternative; however, since the depth of vertical boreholes, and consequently the electrode line length, is relatively small, the inter-borehole distance is limited, so even for 2-D surveys, numerous vertical boreholes are usually needed. The use of horizontal boreholes is interesting as they can be drilled at much longer lengths to extend image resolution from the borehole, with a single horizontal borehole replacing the need for multiple vertical boreholes. Previous studies have shown the potential of S2HB ERT for improved imaging ability, but until now, it was only assessed in two dimensions.

Laboratory tank experiments were conducted to investigate 3-D S2HB ERT performance relative to 3-D surface ERT on different subsurface targets. Different electrode configurations for 3-D S2HB ERT were proposed with a 2-D surface grid of electrodes combined in turn with one borehole, three boreholes, or a matching grid of horizontal boreholes. The most practical 3-D S2HB ERT configuration contains one borehole, and results suggest it can provide superior imaging of all subsurface targets and time-lapse changes compared to 3-D surface ERT. Increasing the depth of the horizontal borehole from six times the electrode spacing (0.24 m) to eight times (0.32 m) did diminish the performance of single borehole 3-D S2HB; however, it was still able to moderately resolve the target while surface ERT was unable to resolve any portion of the target. The inclusion of more horizontal boreholes improved imaging but only slightly, with the single borehole S2HB ERT still providing highly comparable results, at least for the cases considered in this study. It should be noted that while a single borehole is the more realistic and practical option for field site application, it will be challenging to ensure the borehole is optimally located below the target zone, and multiple boreholes may be necessary.

Due to both its performance and more practical implementation, this study suggests that 3-D S2HB-ERT with a single borehole can bring superior characterization and monitoring capabilities to DNAPL field sites. While the focus of this study was on DNAPL contaminant zones, the findings are applicable for improved ERT imaging of a range of contaminants, features, and processes within the subsurface.

Acknowledgement

The authors would like to thank the editor and anonymous reviewers for helping to improve this manuscript. We also want to thank Lauren Ing (Western University) for helping with the construction of the experimental apparatus, and to Dr Jason Gerhard (Western University) and Dr Panagiotis Tsourlos (Aristotle University of Thessaloniki, Greece) for brainstorming the early ideas on DNAPL mapping by surface-to-horizontal borehole ERT. Our dear friend Jason - this manuscript is dedicated to you. Special thanks to Western Research and the Faculty of Engineering at Western University for providing funding through the Western Strategic Support (WSS) for NSERC Success Seed Research Grant.

DATA AVAILABILITY

The data that support the findings of this study are available from the corresponding author, Christopher Power ([email protected]), upon request.

References