Post-remediation evaluation of a LNAPL site using electrical resistivity imaging

Abstract

Present understanding of the earth’s subsurface is most often derived from samples at discrete points (wells) and interpolations or models that interpret the space between these points. Electrical resistivity imaging techniques have produced an improved capability to map contaminants (especially NAPLs—NonAqueous Phase Liquids) away from traditional wells using actual field data. Electrical resistivity image data, confirmed by drilling, have demonstrated that LNAPL (Light NAPL—less dense than water, such as gasoline) contaminants exist outside of a delineated and remediated area in Golden, Oklahoma. The data also demonstrate that LNAPL exists between monitoring and remediation wells which indicate low contaminant levels when sampled. Additionally, the electrical images provided the drilling location with the highest concentration of hydrocarbon ever found on the site, even after two phases of remediation work had been performed, although the sampling protocols varied. The results indicate that current methods of post-remediation site characterization are inadequate for complete site characterization.

---

Introduction

As of November 2004, over 447 000 petroleum-contaminated sites have been confirmed in the US with the number expected to rise. The movement and degradation of hydrocarbons in groundwater has been studied for more than 20 years, but our capabilities to remediate hydrocarbon-contaminated sites have not increased significantly.¹ Much of our understanding of the natural attenuation of hydrocarbons has been generated from field experiments.² Microbial communities actively metabolize hydrocarbon contaminants in a variety of subsurface environments, but in situ attenuation of hydrocarbons is difficult and expensive to detect and monitor.³ These difficulties most commonly arise because discrete sampling techniques cannot provide the detailed 2- or 3-dimensional data required to monitor heterogeneous formations. To resolve these difficulties, data gathering techniques must be developed so that areas or volumes of the subsurface can be examined. These methods should be economical when compared to alternative techniques.⁴

Existing techniques

Existing methods of site assessment have relied primarily on two detection and monitoring strategies. The first strategy involves discrete point sampling of fluids using wells or multilevel piezometers whose data are interpreted by hydrogeologists, civil engineers, and other scientists. The second strategy uses indirect measurements through surface or borehole geophysical techniques.

The difficulty with point sampling techniques is that sufficient sampling can be expensive because of drilling costs, sampling time, and sample analysis and interpretation time. Point sampling methods can miss contaminants transported on flow paths not sampled by wells, or contaminants stored in clay lenses that are not sampled with a traditional piezometer monitoring grid. This is especially problematic if the contaminants of interest are moving either in a non-uniform manner such as density-driven fingering or in isolated flow paths in heterogeneous media.⁵ In some settings, the very act of probing and monitoring the aquifer can create additional heterogeneity and new preferential flow paths for solutes.

Electrical resistivity imaging

A solution to these contaminant detection and monitoring problems in the vadose and phreatic zones is the utilization of electrical imaging to provide more complete site data coverage. A temporary surface system for site evaluation can be used as an evaluation of a 2-D or 3-D portion of the subsurface, using traditional investigation methods for confirmation. Cables can be permanently installed in shallow trenches or in boreholes for long-term monitoring applications.

Electrical resistivity measurements have been used since the 1830s to interpret the geology of the subsurface.⁶ The technique introduces current into the ground and the potential field is measured. ERT (Electrical Resistance Tomography) is a method of obtaining resistivity measurements that determines the electrical conductivity of the ground using subsurface electrodes.⁷ In contrast, a multielectrode array uses electrodes only on the surface. Electrical Resistivity Imaging (ERI) is a general term used to indicate that an electrical resistivity technique is being used without naming each electrode configuration differently. An ERI image is an inverted model of hundreds to thousands of 4 electrode resistivity measurments. Hundreds of measurements of a site are required to produce a 2-D or 3-D ERI model of the subsurface. This technique is occasionally used for site characterization, but it can be inefficient, expensive, or worse, ambiguous.⁸

An ERI image is taken using an acquisition algorithm that collects data from a series of electrodes placed either on the surface or located in boreholes. Two-dimensional data are collected using a linear array of electrodes or 3-D data can be collected using electrodes placed as a series of 2-D arrays or a 3-D electrode grid. However, 3-D data collection is more expensive and difficult as most contaminated sites do not have sufficient open area available and the acquisition and processing times are longer. Obtaining borehole data is also costly since the majority of boreholes constructed for sampling an aquifer do not consider electrical properties as part of the well design and cannot be used to collect robust ERI data. The size and resolution of the image is defined by the distance between electrodes and their location in space. A rough estimate for a 2-D image is that the resolution is half of the electrode spacing and the image depth is one fifth of the total line length.

The acquisition algorithm defines the set of measurements collected to create an image. The algorithms are generally defined by the spatial geometry of the two current electrodes relative to the two potential electrodes used to collect a single measurement. For example, a dipole–dipole array uses two adjacent current electrodes and two adjacent potential electrodes.⁹ A Wenner arrray uses four equally spaced electrodes with the potential pair inside the current pair. There are many such algorithms available to aquire an ERI dataset in one-, two-, or three-dimensions. The “raw” collected data are referred to as apparent resistivity data.

The image is “developed” using an inversion algorithm. The inversion algorithm uses the collected apparent resistivity data to create a model space of resistivity values that would replicate the collected data. These models suffer from problems of nonuniqueness like many potential methods.⁹

Common geophysical techniques are limited by several factors as outlined by Stollar and Roux.¹⁰ They noted a concurrent loss in signal quality and resolution as the depth to the top of contaminated ground water increased. They also noted that there must be a significant contrast between the contaminated and uncontaminated ground water for earth resistivity surveys to be effective tools. Although the costs of earth resistivity techniques may be lower than point monitoring methods (i.e., wells) for long-term projects, the results are often difficult to correlate with objectives and still require traditional ground water sampling techniques. These problems are exacerbated by a lack of integration among geophysicists, hydrogeologists, and microbiologists. A major problem with the application of electrical techniques to contaminant detection is that many contaminants of interest to site managers, such as NAPLs, are electrical insulators. The ERI method works best for identifying conductors, making it difficult to image these relatively resistive materials in their host soil or rock. A primary objective of this research was to overcome or limit these problems by altering the methodologies used to acquire and process ERI data to create “drillable” images. This paper focuses on the application of the methods.

High resistivity anomalies obtained by surface resistivity methods or ERT can indicate the location of LNAPL. Using direct current electrical resistivity soundings supported by GPR (Ground Penetrating Radar) data, LNAPL was identified from a high resistivity anomaly in Arizona¹¹ as demonstrated. Controlled experiments have also been conducted for evaluating ERT imaging during gasoline spills.¹² In images associated with gasoline release, high resistivity anomalies formed in the upper portion of the saturated zone. Minor resistivity changes were detected in areas below the water table due to the entry of hydrocarbons into the dissolved phase.

In cases where a plume has aged approximately 50 years, an LNAPL can show an elevated conductivity relative to surrounding geologic material.^13,14 The initial free product accumulation may show up as a resistive area only until biodegradation becomes established. After time, the mixed zone and underlying aquifer may show anomalously low resistivity. This suggests that LNAPL sites should be treated as individual cases, with changes dependent on site composition, time, and many other variables specific to the location.¹³

As ERI capabilities progress due to increases in field data acquisition capabilities and computing speed, the question arises as to whether ERI techniques can be employed effectively as a method to detect contaminants and certify a contaminated site as remediated. Is it sufficient to use ERI techniques for rapid investigation of remaining pockets of contaminants after a remediation phase has been completed or must other methods be employed?

This paper addresses the above question and is part of a larger project that evaluated remediation and monitoring techniques at a field site in Golden, OK. This site provided a test case that was difficult to monitor and remediate. After standard pump-and-treat technology failed, surfactant flush techniques were utilized to extract fugitive hydrocarbon from a silty clay soil. Electrical imaging was employed by the Oklahoma Corporation Commission Petroleum Storage Tank Division during March 2003 to evaluate whether the site was fully remediated.

During ERI work evaluating the entire site, two features of the resulting images generated interest among the regulators and scientists evaluating the site. First, the technique indicated that high resistivity anomalies interpreted as contaminants existed outside the previously delineated plume. Second, the technique indicated that these anomalies existed in isolated pockets between wells instead of in a contiguous plume shape. These features resulted in two experiments being conducted at the site. The location outside of the previously delineated plume area was tested to determine if the electrical resistivity image had detected fugitive hydrocarbons, then a high precision drilling program tested the delineated area where an image indicated hydrocarbons were present between wells. The authors know of no previous case of a remediated site being investigated with this technique using the images on a quantitative basis.

Site description

The study site is located in Golden, OK on an approximately 2 hectare (5 acre) area (Fig. 1). The original gasoline spill was detected in a shallow water well on site in 1993, and a leaking on-site gasoline UST (Underground Storage Tank) was then found. During subsequent investigations, three other UST locations were discovered at the same location between the current gasoline station and the post office (Fig. 1). There is a history of other gasoline stations previously located in this area.

Fig. 1 Map of Golden, OK LNAPL spill site. Irregular outline indicates previously delineated plume area. Dots indicate wells on site. Heavy straight lines indicate location of surface electrode lines used to obtain electrical images. Dark squares indicate locations of USTs removed from site.

The site geology consists of a range of unconsolidated sediments from clay to gravel. The site sediments consist of 0.15 m (0.5 ft) of topsoil. Below this to a depth of 2.7 m (9 ft) a silty clay exists which becomes more porous and sandy from 2.7 to 5.3 m (9–17.5 ft). A clayey gravel mixture then forms a layer below 5.3 m which has been found to have much higher hydraulic conductivity than the layers above, based on results of direct push slug testing. This higher hydraulic conductivity zone has always shown little to no contamination. The hydraulic gradient in the study area is oriented west-northwest.

Initial site investigation and remediation using a pump-and-treat system was attempted, but it was unsuccessful, and surfactant flush techniques were subsequently employed by a second contractor to remediate the site. A total of 92 wells were installed on the site for monitoring and remediation. After remediation was completed, questions remained as to whether the on-site contaminants were reduced to acceptable risk levels.

During site characterization and remediation by the surfactant flushing vendor, the highest total petroleum hydrocarbon (TPH) concentration found on the site was 5984 mg kg⁻¹ in the soils. Results from a well survey in April 2003 found the highest ground water TPH concentration in sampled wells to be 98.0 mg l⁻¹. The majority of soil TPH values were around 100 mg kg⁻¹ after the remediation procedures were completed but prior to the electrical investigations.

In March 2003, ERI surveys were conducted at the site to evaluate the effectiveness of the remediation procedures that were attempted at the site. These 17 surveys were conducted as 2-D surveys over areas of interest in and around the delineated plume. The geometry of the site and project budget limited the ability to acquire a full 3-D dataset. These surveys indicated that areas with resistivities above 150 Ω m were potential locations of hydrocarbons, with some images presenting as completely “clean.” The results led to the set of two experiments described below.

Methods

Two methods were used to monitor and test the Golden site in September 2003 to determine if fugitive hydrocarbons had been removed. First, ERI data were collected and inverted. Second, based on the ERI data, direct push cores were collected and evaluated for TPH. The results of coring and testing were compared with the ERI images.

Electrical images

Two-dimensional data for two ERI images, image 1 and 2 (EI-1-EW and EI-2-NS), were collected using 36 surface electrodes connected to the land surface via stainless steel stakes at a 1.5 m spacing. The east–west line EI-1-EW is oriented approximately parallel to the hydraulic gradient but not in the delineated plume area, and the north–south line EI-2-NS is orthogonal to the hydraulic gradient, but centered in the delineated plume area (Fig. 1).

The data were collected using an Advanced Geosciences, Inc. Supersting 8-channel resistivity instrument. Several data acquisition array types were used to collect data in the target areas. Dipole–dipole (DD), Wenner (WN), and Schlumberger Inverse (SI) data arrays⁹ were acquired and inverted. These methods were used as a comparison to a Halihan–Fenstemaker (HF) method¹⁵ (commercially applied as Aestus, Inc. Geotrax Survey™), which is a proprietary data acquisition and processing methodology. The range of array types was used to determine if the targets of interest would differ between ERI images or would alter their location horizontally or vertically. All data were collected using a double measurement protocol providing a measurement error for each data point collected.

The ERI field data were inverted using EarthImager version 1.6.8 (Advanced Geosciences, Inc., 2004). The inversions were performed using a robust inversion method with data values having a repeatability error in excess of 2% trimmed from the datasets prior to inversion. This comprised no more than 6.5% of any dataset. The dipole–dipole arrays were trimmed an additional 1.0–3.0% due to a poor fit in the inversion process. All resistivity values are reported in ohm-metres (Ω m).

Confirmation data

The electrical images were provided to US Environmental Protection Agency (EPA) personnel. The EPA personnel then employed direct push coring and field methanol extraction of core samples to test the electrical images. Most other samples taken by site contractors did not use field methanol extraction techniques. A single core was collected from image transect EI-1-EW, and 6 cores were collected from image transect EI-2-NS at 15 or 30 cm (0.5 or 1 ft) depth increments. The cores were sampled and analyzed at the EPA Ground Water and Ecosystems Restoration Division Laboratory using EPA method 8015M GRO for TPH. The chemical analyses are in mg of TPH per kg of soil sampled.

Results

The EPA personnel used the results of the electrical imaging to dictate the drilling program at Golden. The anomalies present in the two electrical images will be presented followed by the results of drilling.

Electrical images

The results for both ERI images were similar in data statistics. The dipole–dipole array produced the most errors in data collection and the highest inversion root mean square error at 10.3%, with all other arrays below 7% RMS error. The Wenner array produced the lowest inversion error at 2.6% RMS error.

Electrical image EI-1-EW had a single 20 m long high resistivity anomaly that extends from near the surface to a depth of six metres (Fig. 2). The center of the anomaly was between 20 and 25 m from the left-hand side (LHS) of the line at a depth of 3 m. All array types detected the anomaly, but the maximum resistivity values ranged from 411 Ω m (WN) to 686 Ω m (HF).

Fig. 2 Electrical image EI-1-EW. Electrical image collected using HF method contoured on 150 Ω m intervals. Vertical line indicates location of soil boring used to sample high resistivity anomaly.

Electrical image EI-2-NS had multiple high resistivity anomalies (Fig. 3). There was a zone of anomalies from 10–20 m from the LHS of the line that were resolved as either one or two anomalies by the different arrays. A second set of anomalies located at a distance of 25–40 m from the LHS of the image presented the most resistive anomalies detected on the site. The magnitude of the resistive anomalies from the DD array and HF method made the anomalies obvious and easier to interpret as true anomalies. The magnitudes of the anomalies from the SI and WN arrays were sufficiently lower to make the interpretation of the data ambiguous. The anomalies were typically 6 m in lateral extent and 4–5 m in vertical extent. The anomalies between 25 and 40 m indicated peak resistivity values at 4–5 m depth.

Fig. 3 Electrical image EI-2-NS. Electrical image collected using HF method contoured on 150 Ω m intervals. (A) Vertical lines in image indicate the location of monitoring and remediation wells. Dotted line indicates area of inset. (B) Vertical lines indicate the location of soil borings used to sample high resistivity anomalies.

Confirmation data

A portion of the transect imaged by EI-1-EW was drilled to confirm the presence of hydrocarbon co-located with a high resistivity (>450 Ω m in HF array) anomaly (Fig. 2). The location of the anomaly was not in the previously delineated plume region, defined by the original contractor, and thus was investigated to determine the presence of detectable hydrocarbon. The line was sampled by coring 20.0 m from the LHS. Subsequent chemical analyses of the samples indicate hydrocarbon is present at a maximum concentration of 60.8 mg kg⁻¹. Higher levels would be expected further east in this transect according to the ERI values.

The transect imaged by EI-2-NS was drilled to confirm the presence of hydrocarbon in a known area of contamination (Figs. 3 and 4). The drilling indicated that (a) the highest resistivity anomaly provided the highest TPH concentration on the site before or after remediation, although the sample was collected with a field methanol extraction technique, (b) the hydrocarbon concentrations were not in a contiguous plume shape, but in discrete zones or “blobs” located both inside and outside the previously delineated plume area, and (c) the high resistivity anomalies correspond in a semiquantitative manner with the TPH values found in the cores. The relationship is not linear, but has a general resistivity increase with TPH concentration.

The cores were divided into three categories, high TPH (>10 000 mg kg⁻¹ in 1 core—EPA 3.10), medium TPH (peak concentration >10 mg kg⁻¹ in 3 cores—EPA 3.8, EPA 3.11, and EPA 3.15) and low TPH (peak concentration <10 mg kg⁻¹ in 3 cores—EPA 3.13, EPA 3.14, and EPA 3.16). The high TPH concentrations were detected by all electrical image types with peak resistivities above 200 Ω m (Fig. 4A). The DD and HF images peak resistivity values corresponded to the sampled peak TPH depth. Both the SI and WN images predicted that the contaminants should be approximately 1 m deeper.

Fig. 4 Core TPH concentrations obtained from direct push samples compared to resistivity profiles obtained using 4 different array methods. (A) High concentration TPH core EPA 3.10. (B) Medium concentration TPH core EPA 3.15. (C) Low concentration TPH core EPA 3.16. Core locations shown in Fig. 3B.

The medium concentration anomalies were poorly or not observed by the WN image (Fig. 4B). The peak value for the core area was approximately 1 m above the correct location and there was not a significant anomaly in the profile. The SI and DD images detected anomalies above 200 Ω m, but the SI predicted a shallower location similar to the WN image. The DD and HF images predicted the correct depth, but the DD image peak value was still very high, indicating higher levels of contaminant than were observed. This result is likely due to the low image sensitivity of the DD image at depth and not the result of inadequate soil sampling.

At low concentrations, only the HF image provided a detectable anomaly above 150 Ω m in these cores (Fig. 4C). The other arrays did not detect low concentrations of hydrocarbon.

Discussion

The results of comparing the core analyses and electrical resistivity images indicates that ERI is a good technique for detecting hydrocarbons in the shallow subsurface. It should be noted that the ERI images do not provide a signal such that the presence of hydrocarbon is the only explanation for any given anomaly. Some confirmatory drilling is almost always required. Locating anomalies and thus potential hydrocarbon contamination more precisely using ERI techniques allows the subsurface to be drilled and sampled much more effectively and comprehensively than any other currently available technique. This site was reasonably simple in that there was only one contaminant of interest and the resistivity signature for hydrocarbon contrasted clearly with any geologic signatures. If the hydrocarbon was further aged resulting in conductive features along with resistive features, or if the geology had strongly resistive features, the work may not have been as successful.

The experiment also indicates that several assumptions regarding hydrocarbon contaminated sites may not be correct. The first is that a single plume is generated on a site. The images combined with data collected during initial site characterization indicate that some hydrocarbon is independent from a primary contaminated area and was not found during the initial investigations. After characterization of the site using electrical imaging techniques, a number of non-contiguous contaminated areas were detected and sampled. These areas were small (<10 m long and <5 m wide) and had not previously been discovered by an intensive boring/well installation program (92 wells over 2 hectares). This may indicate that the capillary trapping mechanism divides NAPL into separate “blobs” on the site scale as well as the pore scale.¹⁶

Wells that have been used as part of a remediation scheme are unlikely to present a true picture of subsurface contamination. The ERI images indicated that the majority of the contaminants that were found by the images and subsequent coring were not located near the wells. It is possible that the act of remediation caused the remnant hydrocarbons to be positioned approximately midway between wells. This may be the reason that sites that are declared ‘clean’ from well sampling data are found to be contaminated at a later date when hydrocarbon re-enters a well used for site monitoring following remediation efforts.

In the top ten metres of most sites, electrical images collected from the surface can provide a good template for directing subsurface investigations. Subsequent drilling can be well constrained and small “blobs” of contaminant can be detected. If used in a transient mode, the images can not only detect a contaminant, but can be used to illustrate visually when the contaminant has been removed by collecting data during and after remediation. Although electrical imaging would cost in the order of $5000 US dollars per day at present, it could substantially reduce both the number of monitoring wells or borings on a site and the time required to determine a site is remediated.

Conclusions

A dense network of wells did not provide adequate monitoring of concentrations of LNAPL at a contaminated site which has been remediated in the past. Electrical imaging and subsequent core samples were successfully used to look between existing wells and provided an improved evaluation of the site. These techniques are mobile and can easily be applied to other sites. Long term monitoring of wells, which can be expensive to install and sample, may not be appropriate due to the physics of NAPL movement. Electrical imaging may provide a cheaper and better long term solution for monitoring LNAPL sites.

Acknowledgements

The authors would like to acknowledge the staff of the Petroleum Storage Tank Division of the Oklahoma Corporation Commission, the staff of the EPA GWERD laboratory in Ada, OK, and the residents of Golden, OK.

References

1. B. Bekins, B. E. Rittmann and J. A. MacDonald, Eos Trans. Am. Geophys. Union, January 30, 2001, 82(5), 53, 57–58.
2. I. M. Cozzarelli, B. Bekins, M. J. Baedecker, G. R. Aiken, R. P. Eganhouse and M. E. Tuccillo, J. Contam. Hydrol., 2001, 53(3–4), 387–406.
3. E. K. Nyer and G. Boettcher, Ground Water Monit. Rem., 21, 42.
4. National Research Council (US) Committee on Innovative Remediation Technologies, Innovations In Ground Water And Soil Cleanup: From Concept To Commercialization, National Academy Press, Washington, DC, 1997, p. 292.
5. C. T. Simmons and J. M. Sharp, Jr., J. Contam. Hydrol., 2001, 52(1–4), 245–275.
6. R. G. Van Nostrand and K. L. Cook, US Geol. Survey Prof. Paper 499, US Geological Survey, Reston, VA, 1966.
7. W. Daily, A. Ramirez, A. Binley and D. LaBrecque, Leading Edge, 2004, 23(5), 438–442.
8. A. Ramirez, W. Daily, D. LaBrecque, E. Owen and D. Chesnut, Water Resour. Res., 1993, 29, 73.
9. J. M. Reynolds, An Introduction To Applied And Environmental Geophysics, John Wiley, Chichester; New York, 1997, p. 796.
10. R. L. Stollar and P. Roux, Ground Water, 1975, 13, 145.
11. A. K. Benson and N. B. Mustoe, Environ. Geosci., 1998, 5(2), 61–67.
12. W. Daily, A. Ramirez, D. LaBrecque and W. Barber, J. Appl. Geophys., 1995, 33, 227.
13. W. A. Sauck, J. Appl. Geophys., 2000, 44, 151–165.
14. E. A. Atekwana, W. A. Sauck and D. D. Werkema, Jr., J. Appl. Geophys., 2000, 44, 167.
15. T. Halihan and T. Fenstemaker, Proprietary Electrical Resistivity Imaging Method, Oklahoma State University Office of Intellectual Property, Stillwater, OK, 2.0 edn., 2004.
16. I. Chatzis, N. R. Morrow and H. T. Lim, Soc. Petrol. Eng. AIME, 1983, 23(2), 311–326.

---