MONITORING OF CONTAMINANT TRANSPORT BY USING GEOELECTRICAL RESISTIVITY TOMOGRAPHY P. SEFEROU(1), P. SOUPIOS(1), E. CANDASAYAR(2), N. PAPADOPOULOS(3), A. SARRIS(3) AND H. AKTARAKÇI(4) 1. Department of Natural Resources and Environment, Technological Educational Institute of Crete, Chania, Greece. 2. Department of Geophysical Engineering, Ankara University, Tandogan Kampus, 06100 Ankara, Turkey. 3. Laboratory of Geophysical-Satellite Remote Sensing & Archaeo-environment, Institute for Mediterranean Studies, Foundation for Research and Technology, Hellas (FORTH), Rethymnon, Crete, Greece. 4. Advanced Geosciences Europe (A.G.E.) S.L. – Madrid, Spain.

SUMMARY: Every human activity (industry, agriculture, and others) produces wastes that contaminate and cause continuous devastation of the environment through depletion of resources such as air, water and soil. The recent environmental management trends give emphasis to the control of the wastes' pollution and their deposition. Moreover, the characterization of pollution in public and/or private lands is facilitating the planning of rehabilitation investments in urban and agricultural environments, helping to minimize their consequences in public health. Thus, a robust characterization and monitoring of any released contaminant in the environment is essential for designing effective remediation strategies and for securing reliable water supplies especially in areas where water is scarce. This paper deals with a simulation study of soil pollution using an integrated approach for monitoring a solute transport downward to the subsurface by using modern geophysical methods. Specifically, a medium scale (1m3) experimental tank was constructed and the contaminant flow into porous materials was monitored through cross-hole electrical resistivity tomography.

1. INTRODUCTION Several types of contaminants are used by a range of industrial and/or agricultural activities which contaminate the subsurface in many areas worldwide through spills, leaks, landfills, uncontrolled releases and disposals (Tait et al., 2004). The dissolution of a contaminant into the

water causes serious problems to the groundwater quality. Thus, the uses of real time monitoring methods are necessary to enhance the knowledge about solute transport through the subsurface. Geophysical methods allow the estimation of subsurface physical properties (such as density, resistivity, velocity, magnetic susceptibility, etc) and can prove to be a useful tool for hydrogeological modeling studies since they provide an additional data source for a detailed groundwater investigation (Binley et al., 2002). Among the geophysical methods, the direct current resistivity (DCR) method is proved to be the most frequently used geophysical technique in hydrology as the resistivity of the subsurface strongly depends on the effective porosity, degree of saturation and pore water conductivity. In this method, electrical current is injected by using pair of electrodes (current electrodes) and resulting potential differences is measured between another pair of electrodes (potential electrodes). Apparent resistivity is calculated by using the measured potential differences. The current and potential electrodes can be laid out on surface (surface measurement), in borehole (crosshole measurement) or both (surface to borehole measurement). The observed data set can be used to interpret tracer migration in the subsurface (Osiensky and Donaldson, 1994). 2D or 3D inversion of surface, surface to borehole or crosshole data set is called as Electrical Resistivity Tomography (ERT).ERT is a geophysical imaging method which is used to estimate and present 2D and/or 3D models of the resistivity distribution of the earth interior. Data acquisition, processing and interpretation methodologies are described in the literature (e.g. Slater et al., 2002; Kemna et al., 2002; Cassiani et al., 2006). The collected DCR soundingprofiling data sets are inverted to reconstruct the 2D/3D resistivity distribution of the subsurface by using non-linear inversion methods (Loke and Barker, 1996). Moreover, the regular acquisition of 2D/3D resistivity data in various time slots (time-lapse ERT–3D/4D) can provide a measure of the temporal changes of the subsurface resistivity, assuming a fixed geology. The concept of field or laboratory scale crosshole ERT as a method of investigating transport mechanisms in heterogeneous media is quite simple since the final model is constructed as the difference between an image during tracer injection and an image prior to tracer injection. The main focus of this work is to investigate the transport of olive oil mill wastes (OOMW) in an unsaturated-saturated soil laboratory column using geophysical imaging data as a surrogate for contaminant concentration measurements in order to track the plume’s movement. Timelapse 3D crosshole mode was employed to increase the depth resolution of the final inverted images. This laboratory scale study shows that the use of geophysical techniques can be a useful tool for large scale field applications, where contaminant concentration measurements are scarce and costly. 2. EXPERIMENTAL WORK – METHODS USED 2.1 Preparation of the controlled experiment A tank with dimensions 1mx1mx1m was constructed by using a high performance 1cm thick plexiglass (transparent material) with specific properties such as, high clearance and durability, in order to carry out the medium monitoring controlled laboratory experiment, to monitor the contaminants’ movement and follow any water level changes. The tank was supported by a metallic frame able to support the forces and pressures generated by the tank and containing materials. Nine fluid valves were bolted on the bottom of the tank to be able to discharge it. Four boreholes with twelve electrodes each were constructed and installed into the tank. The boreholes were constructed by using high density PVC tube and having an electrode spacing of 5 cm and they were installed at the center of the tank (Figure 1).

CRETE 2012

2

Figure 1. Left: Simulation of the tank construction. Center: The plexiglass tank on the metallic base and with the four boreholes with electrodes installed is shown. Right: Preparation of the container with the contaminant before the beginning of the experiment (Seferou 2011). 2.2 Contaminant and Soil Material Used The production of olive oil generates a large amount of solid and liquid wastes. The OOMW especially contain high concentrations of phenolic (toxic material) compounds, reaching up to 80 g/L and the electrical conductivity (EC) is 7.6 mS/cm (APHA, 1999; Komnitas 2011 pers. com.). Usually, the OOMW are transferred to specially designed open tanks or directly in the soil posing a high risk to the environmental pollution. Impermeability of these tanks has crucial importance for preventing leaking of the wastes to the subsurface and subsequently possibly to the groundwater. For the preceding reasons, the OOMW that are water soluble was choses as the pollutant material for this experiment. The appropriate quantity (60lt) of the contaminant (OOMW) samples was taken from Keritis river basin, an area near Chania, Crete (Figure 2). The landfill areas should have specific geology such as impermeable rocks, non-tectonic regime, etc, to avoid contamination of the subsurface (unsaturated zone) and the aquifer. Despite that, in most of the cases, the areas are usually composed of grain size materials like sand, gravels, pebbles, etc (Figure 2, Alikianos OOMW site). In order to simulate a real case study, we decided to use grained material for this experiment. Specifically, it was decided to use fine grained materials (k=1.09x10-6m/s) in order to delay the percolation of the OOMW in the subsurface and be able to get continuously time lapse measurements.

Figure 2. The broader study area (Chania Municipality) and the area where the OOMW are collected are presented (Seferou 2011).

CRETE 2012

Figure 3. The acquisition system (IRIS Syscal R1 Plus Switch 48) is shown (Seferou 2011).

3

2.3 Description of the experiment At the beginning of the experiment, an effort was made to get a reference 3D resistivity dataset under dry (unsaturated) conditions without any released contaminant. During the rising of the water level through the central valve, local soil subsidence among the boreholes was observed and some compacted cores (lenses) were created (known as Brazil nut effect) and imaged by the 3D resistivity of measured DCR data. When the tank was fully saturated, all the valves were opened to create a semi saturated and fully saturated conditions in the tank. The contaminant container (60 lt) was also installed on the top of the experimental tank as it is shown in Figures (1, 3). The purpose was to release the contaminant at the center of the tank and between the monitoring boreholes (Figure 3). During the first data set, two instruments were used for the acquisition to ensure the accuracy and repeatability of the measurements. All the data files were collected by using the IRIS Syscal R1 Plus Switch 48 and the crosschecking was done by using the IRIS Syscal Pro (Figure 3). After 15 min (the beginning of the 2nd data file), the OOMW was released with a controlled flow rate. The measurements were continuously taken every 15 min. In total 77 2D cross-hole resistivity data were acquired. After more than a day (25 hours and 13 min), 8 supplementary 3D resistivity cross-hole data files were collected. In order to keep the water level stable (water level changed during the contaminant release), four piezometers were installed to discharge the water (when needed) from the valves (Figure 3). 130 hours after the beginning of the experiment, the last DCR data set was collected and the acquisition procedure was terminated. The data quality control was checked by applying noise reduction and filtering, bad datum removal before the 2D and 3D inversion. 2.4 2D inversion and 2D Time lapse crosshole geophysical modeling Before the beginning of the experiment and the release of the contaminant onto the soil column, crosshole resistivity data were collected between each pair of the borehole (302 measurements) and these data sets combined (1812 measurements) to use in 3D inversion for reconstructing an accurate reference resistivity model which was later used as a constrain model for time-lapse inversion. The crosshole reference resistivity dataset was collected in two different directions in order to define any inhomogeneity of the experiment and to determine the optimum direction for continuing the measurements based on RMS error and sensitivity analysis. The direction that was finally selected for acquiring the rest of the measurements was the B-B direction (Figure 4). After the release of the contaminant, 75 crosshole geoelectrical data sets were acquired for about 25 hours and every 15 min on average. In order to assure the reliability of the resulted crosshole tomographic images, two different resistivity inversions software packages were used: EarthImager 2D/3D from Advanced Geosciences Inc (AGE) and DC_2DPro created by Prof. Jung-Ho Kim (Korea Institute of Geoscience and Mineral Resources-KIGAM). The optional of 2D Time Lapse Inversion module of AGI was used for this experiment. The main idea is that the acquisition geometry is installed at fixed (permanent) locations (into boreholes in our case study) during the period of monitoring to facilitate image comparison and high sensitivity of small changes in the ground (EI2D, 2009). The Time Lapse Inversion algorithm takes advantage of the reference model and instead of inverting the monitor data set alone, the algorithm inverts the difference (Difference Inversion) between the monitor and reference data set. Supplementary, the reference resistivity model is used as the a-priori model in the time lapse inversion. Therefore, the time lapse inversion converges faster than standard inversion and is more sensitive to small subsurface changes.

CRETE 2012

4

Figure 4. A plan view of the 3D experimental geometry is presented. The 2D resistivity tomographies as part of the 3D resistivity acquisition experiment are depicted by using the red thick lines (Seferou 2011).

Figure 5. The cross section of the experimental geometry is presented. The electrodes were installed at 3.2 cm from the soil surface and till the depth of 0.582 cm and had a spacing of 5 cm (12 electrodes per borehole) (Seferou 2011).

2.5 3D crosshole resistivity tomography 48 electrodes were used for the 3D/4D resistivity modeling (Figure 4). Specifically, 4 boreholes separated 30 cm each in X and Y directions were installed. Each borehole comprised of 12 steel electrodes separated every 5 cm. Distances from the sides/bottom of the experimental tank were fixed in such a way to avoid boundary effects. The boreholes were installed at 35 cm away from the sides of the plastic tank and the last (the deeper) electrodes were installed at the depth of 0.582 cm (0.418 m from the bottom) (Figure 5). The numbering of the 48 electrodes followed the counter-clockwise direction (Figure 4) starting the counting from borehole B1. The acquired resistivity data were processed by using the EarthImager 3D (EI3D) commercial package by AGI (EarthImager 3D, 2008). The cross-hole bipole-bipole acquisition configuration was used as suggested by Bing and Greenhalgh (2000). All the possible combinations between the boreholes (B1-A1, B1-B2, B1-A2, A1-A2, A1-B2, A2-B2) and the electrodes (1-48) were used for collecting resistivity data by using the bipole-bipole configuration (Figure 4). Each 3D data set consisted of 1812 apparent resistivity readings which were taken in 90 min in total. The average RMS of all the data set was less than 3% after 6-10 iterations. The Sand Box boundary condition (EI3D module) was used to achieve the optimum forward solution and jacobian matrix calculations.

3. RESULTS 3.1 2D crosshole resistivity tomography and Time Lapsed Inversion The 3D crosshole resistivity tomography of the reference model shows that the average resistivity values were about 40 ohm.m (Figure 6, left). Additionally, two high resistance areas with resistivity values ranging around 50-70 ohm.m at the depths of 23 and 40 cm were depicted. The very low resistivity (3-8 ohm.m) areas can be assumed as inversion artifacts since they were found at the edges of the model. These results are in agreement with the 2D reference crosshole tomographic images (Figure 6, center, right) which were collected on 28th of June. At the same depths, two resistive lenses were also found. The resulted crosshole resistivity images after the beginning of the controlled experiment by using the KIGAM and AGI software are presented in Figure (6). Generally, the KIGAM and CRETE 2012

5

AGI software provided similar results. The contaminant in all models reached the depth of 12-13 cm and a high resistive anomaly was found at the depth of 17-23 cm and a conductive zone was depicted at the bottom of the model (below 45 cm) which can be safely assumed as edges effects of non linear inversion. In order to overcome the problem of uncertainty in 2D tomographic interpretation, and to get robust estimates of the final crosshole tomographic images, time lapse 2D inversion was applied to the collected data and more data (1812 reading instead of 302 measurements during the 2D modeling) were collected and interpreted using 3D modeling.

Figure 6. Left: 3D crosshole resistivity tomography before the beginning of the experiment is presented. Only low resistivity isosurfaces are visible (8-45 ohm.m), indicating the possible presence of two high resistive (impermeable ???) bodies at the average depths of 23 and 40 cm. Center: 2D resistivity crosshole tomography in B-B’ direction (rms=1.5%) by using the KIGAM algorithm. Right: Crosshole tomographic model as created by the application of EI2D, AGI software by applying different regularization parameters (Seferou 2011). Figure (7) presents the time lapse inversion result in specific time windows after the beginning of the experiment. The average RMS of the models is less than 3% and the final models were reconstructed after 4 iterations. The background change is 0%, but at the top a low resistivity area is presented. Figure 7b shows that, 164 min after the contaminant release, the solute’ concentration is gradually increasing and moves downwards reaching the depth of 9 cm. In agreement with the aforementioned, Figure (7c, d) shows that more than 8 hours and 22 hours after the beginning of the experiment, the contaminant moves downward reaching at the end the depth of 14 cm. The two resistive zones as depicted in Figure (6, left) are interpreted and presented as a widening resistive zone at the center of the model ranging from 20-33cm in depth.

a. 1 hour and 20 min

b. 2 hours and 44 min

c. 8 hours and 12 min

d. 22 hours and 58 min

Figure 7. Crosshole time lapsed tomographic images as resulted from difference inversion (Seferou 2011).

CRETE 2012

6

3.3 3D crosshole resistivity tomography The final 3D resistivity tomographic models are shown in Figure (8a-f) by using the isosurface presentation. Six different time steps are presented to show the contaminant movement over time, The first 3D image (Figure 8a) shows that 52 hours after the beginning of the experiment (and the Release of the Contaminant onto the Surface-RCS) the contaminants' front reached the depth of 19 cm. The resistivity of the isosurface was about 20 Ohm.m since the upper part of the material has higher resistivities. Figure (8b) shows that the contaminant reached the depth of 22 cm (63 hours after RCS), and a high (around 70 Ohm.m) resistivity was defined at the depth of 23 cm. The anomalies at the sides/bottom of the model can be discarded assuming that they were artifacts (numerical instabilities) from the application of the inversion algorithm. Figure (8c) indicates that after 82 hours the contaminant moved down to the depth of 27 cm moving around the high resistive (impermeable??) lens. Figures (8 d,e,f) show the contaminants' movement till the depth of 38 cm. It is worth to mention that the contaminant passing around the impermeable (high resistive) lens is unified again (Figure 8f) and moves as a high viscosity liquid mass.

a. 52 hours

b. 63 hours

d. 94 hours

e. 106 hours

c. 82 hours

f. 118 hours

Figure 8. Final 3D resistivity models presented as isosurfaces. Only low (5-20 Ohm.m) resistivity surfaces (the contaminant) are visible (Seferou 2011).

4. CONCLUSIONS This research work focused on the use of geophysical imaging techniques (time lapse geoelectrical measurements) to monitor the transport of olive oil mill wastes in a soil laboratory column under unsaturated-saturated conditions. This controlled experiment signifies that geophysical methods can provide valuable information for large scale environmental problems. Moreover, such laboratory experiments can be repeated applying different soil types, contaminants, configurations and other parameters controlling the effect of each of them into the collected resistivity data.

CRETE 2012

7

ACKNOWLEDGEMENTS The work is co-funded by the European Social Fund and National Resources in the framework of the project THALIS (32-4-4):‘‘GEODIAMETRIS–Integrated Geoinformatics Technologies for Time-Lapse Monitoring of Land Pollution from the Disposal of Olive-Oil Mills Waste’’.

REFERENCES APHA, 1999, American Public Health Association, Standard Methods for the Examination of Water and Wastewater, method 5530, 20th edition, Washington D.C., 1999. Binley A., Cassiani G., Middleton R., Winship P., 2002. Vadose zone flow model parameterisation using cross-borehole radar and resistivity imaging, Journal of Hydrology 267, 147–159 Bing, Z. and Greenhalgh, S. 2000, Cross-hole resistivity tomography using different electrode configurations . Geophysical Prospecting, 48: 887–912. Cassiani G., Bruno V., Villa A., Fusi N., Binley A. M., 2006, A saline trace test monitored via time-lapse surface electrical resistivity tomography, Journal of Applied Geophysics, Volume 59, Issue 3, July 2006, Pages 244-259 EarthImager 2D, 2009, Resistivity and IP Inversion Software Manual, Version 2.4.0, Advanced Geosciences, Inc. 2121 Geoscience Drive, Austin, Texas 78726 EarthImager 3D, 2008, Resistivity Inversion Software Manual, Version 1.5.3 Advanced Geosciences, Inc. 2121 Geoscience Drive, Austin, Texas 78726 Kemna A., J. Vanderborght, B. Kulessa and H. Vereecken, 2002. Imaging and characterisation of subsurface solute transport using electrical resistivity tomography (ERT) and equivalent transport models. Journal of Hydrology, 267, pp. 125–146. Loke M.H. and R.D. Barker, 1996. Rapid least squares inversion of apparent resistivity pseudosections by a quasi-Newton method. Geophysical Prospecting, 44, pp. 131–152. Osiensky J.L. and P.R. Donaldson, 1995, Electrical flow through an aquifer for contaminant source leak detection and delineation of plume evolution. Journal of Hydrology, 169 , pp. 243–263. Seferou Paraskeyi, 2011, Characterization of unsaturated flow and transport in porous deposits with hydrogeophysical laboratory methods, MSc Thesis, Georg-August Universitat – Gottingen, Hydrogeology and Environment Geosciences (HEG), Germany, S2011. Slater L., R. Versteeg, A. Binley, G. Cassiani, R. Birken and S. Sandberg, 2002. A 3D ERT study of solute transport in a large experimental tank. Journal of Applied Geophysics, 49, pp. 211–229. Tait N.G., R.M. Davison, J.J. Whittaker, S.A. Leharne and D.N. Lerner, 2004. Borehole optimisation system (BOS): a GIS-based risk analysis tool for optimising the use of urban ground water. Environmental Modelling and Software,19 , pp. 1111–1124.

CRETE 2012

8