subsurface tunnel detection using electrical resistivity ...

SUBSURFACE TUNNEL DETECTION USING ELECTRICAL RESISTIVITY TOMOGRAPHY AND SEISMIC REFRACTION TOMOGRAPHY: A CASE STUDY Grey I. Riddle, Department of Physics, University of Alberta, Edmonton, AB, Canada Craig J. Hickey, National Center for Physical Acoustics, University of Mississippi, Oxford, MS, USA Douglas R. Schmitt, Department of Physics, University of Alberta, Edmonton, AB, Canada

Abstract Seismic and electrical methods are two geophysical techniques commonly used to detect rock property changes in the subsurface. Surveys are usually carried out with sources and sensors placed at or near the earth’s surface. Various levels of data processing are applied to construct a map of the subsurface distribution of a physical property. Seismic methods are sensitive to velocity and density changes of the rock, while the electrical response is dependent upon the electrical resistivity of the rock. In this paper, we present an evaluation of using both seismic refraction and electrical imaging to locate a tunnel. Although the contrast in seismic velocity and electrical resistivity between the tunnel and its surrounding material can be large, tunnels remain difficult to detect. The difficulty primarily arises because the spatial resolution of these methods is less than or on the order of the size of the tunnels. Detection is further complicated by heterogeneity of the near surface materials within which many tunnels are constructed. In this paper, we present a case study using ERT dipole-dipole data and seismic refraction data at a tunnel site. Both seismic and electrical surveys were acquired at the same time with a lateral offset of about 3m. The results from both techniques show anomalies at the location of the tunnel. The confidence in predicting the location of the tunnel is increased by using data from both techniques.

Introduction Subsurface voids include any zone in the subsurface having physical properties that are significantly different than the surrounding earth material. Some examples of high-contrast voids are: tunnels, bunkers, cellars, tombs, pipes, tanks, culverts, caverns, and underground mines. In most cases these voids will contain water and/or air and the bulk properties of the void is equal to the properties of the occupying fluid. These voids may have exterior thin shells, as is the case for plastic or metal pipes and tanks, which may produce unique responses of the void. However, despite the large contrasts associated voids, they do not produce easily-detectable geophysical anomalies. The reasons include: (1) the size of the void with respect to the resolution of the geophysical technique, (2) the often irregular shape, and (3) the heterogeneity of the surrounding native material itself, which can produce a significant number of anomalies. The problem of tunnel detection has been an issue for quite some time. Many meetings have been organized over the past 30 years to discuss possible detection methods (USARL, 2009; Daniels and Harmon, 2009; Simmons and Aldridge, 2008; Sabatier and Muir, 2006; Halihan and Nyquist, 2006; McKenna and Ketchum, 2006; Anon., 1988; Anon., 1981). Some of the techniques being considered for the detection and location of tunnels include: acoustic/seismic technologies, electro-magnetic and resistivity technologies, gravity technologies, optical sensing technologies, and radar technologies (Sabatier and Muir, 2006).

Geophysical techniques remain the only ways to remotely and non-destructively sense the earth’s near subsurface and as such have the most promise for rapid and accurate detection of tunnels. Detecting tunnels may be less difficult than studying and characterizing geologically important voids and structures because tunnels are usually somewhat cylindrical in shape and are constructed in otherwise competent media. Furthermore, the tunnel problem is concerned with detection only, whereas geological studies are concerned with characterization. For a successful characterization the geophysical attributes must have a clearly defined correlation to the physical/geological properties. The geophysical attributes used in detection schemes must be scientifically sound but the connection to physical properties does not have to be fully understood for the detection scheme to be successful. Although tunnel detection might be a simpler problem than say, studying karst environments, the tunnel detection problem still has the issues associated with the size of the tunnel with respect to the resolution of the geophysical technique and the heterogeneity of the surrounding native material. Geophysicists performing site characterization for geotechnical applications, mining exploration, and oil exploration employ several techniques in order to alleviate the ambiquity in geophysical interpretation. The construction of a tunnel and the disruption of the material in its vicinity will change a group of physical properties which should manifest in a common set of geophysical signatures. For example, construction of a tunnel will produce a region of lower bulk density, lower seismic speeds, and, in many cases, lower water content. Since electrical resistivity is strongly dependent on water content, clay fraction, and the presence of soluble salts the tunnel should have a high electrical resistivity signature. The elasticity of soil is strongly dependent on the cohesion, degree of cementation, and water content and therefore a tunnel should have a low velocity signature. Crawford et al. (2006) suggested the use of microgravity and electrical resistivity techniques for the detection of caves and tunnels. They state that a low gravity anomaly and a colocated high resistivity anomaly indicates the presence of a cave. In this paper we present some preliminary results on the simultaneous use of active seismic refraction tomography (SRT) and electrical resistivity tomography (ERT) for the purpose of tunnel detection. The SRT and ERT techniques were chosen because: (1) the ERT is sensitive to the electrical resistivity of the ground as opposed to the SRT which is sensitive to the elasticity of the ground; (2) the ability for both SRT and ERT surveys to be conducted simultaneously; and (3) the results are both displayed in terms of tomographic profile images, which facilitates joint interpretation of data. ERT and SRT method In near-surface refraction tomography, the travel times of seismic energy recorded at the surface by multiple source-receiver combinations are used to generate an optimized model of the distribution of seismic velocities in the subsurface (Baker, 2002). The “velocity image map” can be used to infer subsurface units/features having sufficient differences in elastic properties (Baker, 2002). This approach is commonly referred to as travel time tomography and uses the ray approximation to wave propagation. The limits of resolution for travel time tomography are often associated with the break-down of the ray approximation, i.e. when the scale of the perturbations is on the order of a wavelength. However, Williamson (1991) suggests that the smallest feature that can be accurately reconstructed using ray travel time tomography is given by rmin  L

(1)

which is the radius of the first Fresnel zone for propagation over a distance L. Resolution increases as the square root of the frequency and decreases as the square root of the path length. Sheehan et al. (2005) performed a study using three different tomography codes. They showed that although the actual seismic velocity contrast may be abrupt, the resulting image smears the boundary into a gradient. This may be associated with the resolution of the actual seismic wave and/or with the regularization algorithms used for calculating the velocity distributions. The regularization algorithms are based on the assumption that the subsurface properties vary continuously; and, therefore, sharp physical boundaries are smeared into a gradient. Sheehan et al. (2005) also showed that none of the codes were able to “image” the cavity example using the velocity map. All three codes have artefacts located above the cavity, which is most likely associated with low ray coverage. In this study we extend the work of Hickey et al. (2009) who studied sites containing buried pipes. This work showed that although the actual seismic velocity contrast between the pipe and the surrounding earth materials is large, the measured velocity contrast in the seismic refraction tomograms is diffuse. However, these field studies showed that spatial distribution or ray coverage within the subsurface is an attribute that can be used for detecting and locating pipes. The RayfractTM (Intelligent Resources Inc.) seismic tomography code was used to process the seismic data for locating the pipes. This code uses the Wavepath Eikonal Traveltime inversion (Schuster and Quintus-Bosz, 1993), which incorporates a Fresnel volume approach. This approach accounts for the finite wavelength of real waves and the influence of the model in the adjacent vicinity of the ray path. The Fresnel volume approach addresses this issue by taking into account ray paths that arrive within a half period of the fastest waveform. Electrical resistivity methods utilize direct currents or low-frequency alternating currents to investigate the electrical properties of the subsurface. In the resistivity method, the source is artificially-generated electric current introduced into the ground using electrodes. The potential differences are measured at the surface and the pattern of potential differences provides information on the distribution of subsurface electrical resistivity. Factors that affect the resistivity of soil-water mixtures include ionic concentration, porosity, surface conduction, tortuosity, and connectivity of fluid or conductive solid phases. Tunnels constructed in the vadose zone will cause redistribution in soil moisture possibly causing a local reduction in electrical resistivity. Tunnels constructed in saturated soil will also cause a local reduction in soil resistivity associated with the pumping of water in order to keep the tunnel clear. Furthermore, if the tunnel alters the height of the water table, due to water entering the tunnel, this might be an indirect indicator when using the electrical resistivity method. The development of multi-channel resistivity meters and electronic switches enables the use of many electrode configurations which has made data acquisition for tomographic reconstruction feasible (Griffiths and Barker, 1993). In ERT, the forward problem uses the finite-element method to compute the electric potential response of the earth due to a given input electric current. The inverse algorithm iteratively finds the best distribution of subsurface resistivity that best fits the observed data. Daily et al. (2004) presents a concise overview of ERT. ERT has been used for some time to detect caves and tunnels (Spiegel et al., 1980; Van Schoor, 2002; Burger, 1992). The Advance Geosciences, Inc. Website also has several examples of using ERT for finding various types of high contrast voids (http://www.agiusa.com/ 2Dvoids.shtml, accessed January 15, 2010). In this study we use RES2DTM inversion software from Geotomo to calculate the psuedosections for the resistivity data. There are many different types of resistivity arrays with

their respective advantages and disadvantages. For our situation we chose to use the dipoledipole configuration due to its good lateral resolution. The results shown are given as an apparent resistivity and thus are interpreted as local changes within the subsurface and cannot be directly interpreted as true resistivity values. The final inverted model shows the resistivity distribution of the subsurface determined from the apparent resistivity pseudosection.

Field Experiment Geophysical field surveys were performed at a site with a known tunnel and with good surface access for surveying. The tunnel is a ~1 mx1.6 m (3ftx5ft) concrete lined tunnel about 80m (250ft) long. In general the tunnel is approximately 13m below the ground surface. The tunnel has issues of flooding, indicating that the tunnel is below or near the water table at some time in the year. The data presented here was collected in a nearby drainage ditch where the tunnel is at a depth of about 6m. The site layout for both the ERT and the seismic survey is shown in Figure 1. Based on prior knowledge of the tunnel location the surveys are approximately perpendicular to the tunnel and were purposely centered on the approximate location of the tunnel in order maximize the geophysical sampling in the vicinity of the tunnel.

Figure 1: Site layout for the seismic refraction and ERT surveys conducted in a drainage ditch. The seismic line extends between geophone 1 (G1) and geophone 96 (G96). The ERT line is 3m south of seismic line and extends between electrode 1 (E1) and electrode 50 (E50). The location of the tunnel is approximate and intersects the surveys near the midpoint of the lines.

The seismic refraction survey was performed using 96 GS-20DM 14Hz OYO geophones with a Geometrics GeodeTM configuration and multichannel takeouts. The geophone layout was centered in the drainage ditch with the first channel at the east end of the line. A geophone spacing of 0.5m was used for a total spread length of 47.5m. The general rule for refraction tomography is that the depth of investigating is approximately 1/4 th of the spread length, so a 47.5m spread length should give a depth of penetration of 10-12m. The tunnel is estimated to be about 5-6m deep. The source consisted of a 3.7 kg (8lb) sledgehammer impacting a 10cmx10cm aluminum plate. The shot locations were centered between the geophones at 1m intervals. This shot–receiver configuration results in 9312 seismic rays. The ERT survey was acquired before the seismic acquisition; this was done by running the resistivity survey while setting up for seismic acquisition. The ERT system consisted of a Scintrex SARISTM automated imaging system, with two 25 takeout smart cables. The SARIS can produce up to 1A of current into the ground and 100W of power. The electrode layout consisted of 50 electrodes in a 1m dipole-dipole configuration for a total 49m spread length. The electrodes were one foot stainless steel rods and were planted near the edge of the ditch where the sand had slightly higher moisture promoting better coupling. The depth of investigation and the sensitivity of an ERT survey depend on the electrode spacing. The depth of investigation increases as the spacing between the potential and current dipoles increases. However, as the electrode dipole offset increases a higher current is needed to obtain a measureable voltage. ERT Data The data was edited to eliminate the improper lineations and large RMS data errors at the edge of the tomograms due to lack of current at larger offsets. The ERT results after three iterations are shown in Figure 2,. Figure 2a is the measured apparent resistivity at the site location; the results are then displayed in block format and seen in the pseudosection. Figure 2a is the final apparent resistivity pseudosection corresponding to the resistivity model shown in Figure 2c. The accuracy of the final model is evaluated by comparing the m Figure 2b to the measured apparent resistivity pseudosection shown in Figure 2a. For this case the forward modeled apparent resistivity was iteratively calculated until it the RMS error between it and the observed apparent resistivity differed by less than 2%. Figure 2c is the inverse model resistivity section representing the final distribution subsurface resistivity for this survey. Due to the small size of our data set a finite element forward model with trapezoidal model blocks was acceptable. Smooth features near the edges of the measured apparent resistivity are created during the inversion and results near these sides must be ignored. In the final resistivity model (Figure 2c) , two high resistive anomalies with values reaching 60 Ωm are centered near at 20 m and 25.5 m. The two anomalies could represent a void buried within a fairly conductive weathered sands, in general, the resistivity of concrete ranges from 30-100Ωm and depends on a number of factors including porosity and grain composition. The removed data points seen at 30m-40m was taken out due to being ten times larger than the standard deviation of surrounding material. Not all data was taken out since the anomaly of interest is at 25.5m, seen as E1 in (Figure 2C). This would skew the results and force the anomaly to fit if all bad data was removed. Another anomaly is seen at 3m depth and is shown as E2 in (Figure 2C). For this region we have an approximate water table depth of 5m, this is known due to prior information of the void filling full of water. Low resistivity region directly below and above the two anomalies are evidence for this water table. The resistivity deceases at

a depth of about 5-6m and is associated with the groundwater table. The resistive anomaly at E2 is unknown seen, while the anomaly at E1 corresponds with the approximate known location of the tunnel.

Figure 2: Resistivity tomography for the survey at a true tunnel site. A.): raw measured pseudosection of the apparent resistivity. B): calculated apparent resistivity from model resistivity, 3 iterations were used for the inversion. C): final model resistivity, this is the inverted model for the measured apparent resistivity section. Anomaly of interest is at E1, while E2 is not known. Seismic Refraction The sesimic field data was imported into RayfractTM and 9312 first break arrival times were picked from the data. Starting with a simple 1-D gradient velocity model the program then determined the distribution of subsurface velocities that resulted in a best fit to the arrival times. The results were then displayed in a color scale cross-section referred to as a velocity tomogram and its corresponding ray coverage plot. The velocity tomogram and its associated ray coverage map for the complete 96 geophone spread after 20 iterations are shown in Figure 3a and figure 3b, respectively. The majority of the rays (i.e. densest seismic sampling in red color) are concentrated above a depth of 12m. This is consistent with the usual assumption that the depth of investigation of a seismic refraction survey is about ¼ the spread length. The velocity tomogram exhibits a general

increase in velocity with depth that is typical for poorly consolidated materials. The saturated materials beneath the water table usually have seismic velocities of about 1500-1800m/s; and from Figure 3a this would infer that the water table is at a depth of about 10m from the bottom of the drainage ditch. The tunnel location is estimated to be near the center of the spread and at a depth of about 6m. The anomaly corresponding to the location of the tunnel is labeled S1. As expected, there velocity tomogram shows only a slight downward bend in the velocity contour; however, ray coverage plot shows a zone of low ray coverage at this location. In order to focus in on shallower depths, a subset of the travel times using only geophones 24 through 72 was processed (Figure 4). The velocity tomogram shown in Figure 4a shows a slight lowering in the velocity contour in the general location of the tunnel (labelled S1), it does not have a strong anomaly associated with the tunnel. Figure 4b shows the ray coverage over the complete range; a region of low coverage encircled by high coverage is evident near the location of the tunnel. This type of signature was observed by Hickey et al. (2009) over various pipes. In the enhanced ray coverage mapping (Figure 4c), the local minimum in ray coverage is at a depth of 5.5m and at station location of 13.5. In terms of the complete survey this corresponds to the location of geophone 49. The center of this region of lower ray coverage is in excellent agreement with the known location of the tunnel. With the correlation between low ray density and tunnel position in mind, and returning to the complete seismic spread of Figure 3, another region of low coverage and a slight dip in the velocity contour is seen at 44m (near geophone 88). This anomaly is labelled S2. This type of anomaly could possibly be associated with another tunnel. The anomaly is close to the end of the seismic line and additional data to west of the current spread is need to further investigate this anomaly. ERT and Seismic Refraction The tunnel produces visible anomalies in both the seismic refraction and the ERT data. In the ERT, the tunnel coincides with one of the high resistivity anomalies, but a second, shallower, resistive anomaly of unknown provenance appears just to the east. Based on the results of the seismic survey it would be difficult or impossible to interpret the existence of the tunnel on the basis of the velocity tomogram even with the slight dip in the velocity pattern near the known tunnel location. However, the location agrees well with a zone of low seismic ray density. There is no seismic anomaly corresponding to the shallower anomaly in the ERT data at location E2. In this instance the use of both methods would suggest that this second ERT anomaly is not a tunnel and illustrates how the use of both seismic refraction and ERT can be used to increase the reliability of detecting tunnels. A second anomaly is also present near the end of the seismic line seen at location S2. However, in this case neither the ERT nor the seismic surveys reached sufficiently far enough west to eliminate this anomaly as a possible tunnel.

a.

4

0

8

12 16 20 24 28 32 36 40 44 48 52 56 60 64 68 72 76 80 84 88 92 96

m/s

6000 5500 5000

Depth(m)

-5

S1

4500

S2

4000

-10

3500 3000

-15

2500 2000 1500

-20

1000 500

-25 0

4

8

12 16 20 24 28 32 36 40 44 48 Location (m)

b.

# of rays

0

1600 1500 1400

Depth (m)

-5

1300

S1

S2

-10

1200 1100 1000 900 800 700

-15

600 500 400

-20

300 200 100

-25

0

0

4

8

12 16 20 24 28 32 36 40 44 48 Location (m)

Figure 3: The measured a.): velocity tomogram and the corresponding,b.): ray coverage map* from the 96 geophone seismic line. The tunnel is located near the center of the spread and at a depth of about 6m. The anomaly S1 is associated with the tunnel. The anomaly S2 requires further investigation.

a.

m/s

25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71

0

2400 2200

Depth(m)

-2

2000 1800

-4

1600 1400

-6

1200

S1

1000 800

-8

600 400

-10

200

0

2

4

6

8

10 12 14 16 18 20 22 24

Location (m) Re 11.5m b.

# of rays

0

260 240

Depth(m)

-2

220 200 180

-4

160 140 120

-6

S1

100 80

-8

60 40 20

-10

0

0

2

4

6

8

10 12 14 16 18 20 22 24

Location (m) Re 11.5m c.

# of rays

0

90 88

Depth(m)

-2

86 84

-4

82 80

-6

78 76

-8

74 72

-10

70

0

2

4

6

8

10 12 14 16 18 20 22 24

Location (m) Re 11.5m Figure 4: The data from the central part of the survey, from geophone 24 to geophone 72, processed using Rayfract. A.):the measured velocity tomogram, b.):the corresponding ray coverage map, and c.): the ray coverage map with a threshold.

Conclusion The study shows how both seismic and electrical techniques can be used to detect tunnels and agrees with previous studies over buried pipes. Seismic and electrical properties are sensitive to different types of rock properties and anomalies are present in the data at corresponding to the tunnel location. Seismic tomography requires dense spatial sampling in order to obtain adequate ray coverage. The velocity tomographic image is inadequate for tunnel detection as the smoothing inherent to the tomographic calculations results in only slight changes in velocity near the tunnel location. Instead, the ray coverage density mapping associated with ray tracing displays small regions of low coverage associated with the tunnel. The tunnel should show up in the electrical imaging as regions of high resistivity since both the concrete and air of the tunnel are higher resistivity than the conductive weathered rock. In practice, the resistive anomaly of the tunnel gets smoothed out and is larger than the actual tunnel.. From this field study, the seismic data and the electrical both show an anomaly at the center of each spread that correlate with the location of the tunnel. The electrical data also shows another anomaly near the middle of the spread which is not present in the seismic data. Similarly near the edge of the seismic spread a drop in velocity is seen and a possible void site is suggested. Using results from both surveys together allows for a more confident detection of this tunnel.

References Advanced Geosciences, Inc, 2007., Case and void detection, accessed January 15, 2010 from http://www.agiusa.com/2Dvoids.shtml. Anon., Proceedings of the second technical symposium on tunnel detection, Colorado School of Mines, Golden, Colorado, July, 1981. Anon., Proceedings of the third technical symposium on tunnel detection, Golden, Colorado, January, 1988. Burger, H. R., 1992. Exploration Geophysics of Shallow Subsurface Prentice-Hall, Eaglewood Cliffs, NJ, 489pp Daily, W., Ramirez, A., Binley, A., LaBrecque, D., 2005, Electrical Resistance TomographyTheory and Practice. Investigations in Geophysics; no. 13, 525-550. Daniels, J. and Harmon, R. "Special Session on Cavities and Tunnels", EEGS Annual Meeting, Forth Worth, TX, March 9-April2, 2009. Griffiths, D.H., Barker, R. D., 1993. Two-dimensional resistivity imaging and modeling in areas of complex geology. J. Appl. Geophys. 29, 211-226 Halihan, T. and Nyquist, J.E. “ Detection of voids, tunnels and collapse features,” Session No. 218-T65, 2006 Philadelphia Annual Meeting of the Geological Society of America, Philadelphia, PA, October, 2006. Hickey, C.J., Schmitt, D.J., Sabatier, J.M, and Riddle, G. “ Seismic Measurements for Detecting Underground High-contrast Voids”, Proceedings of Symposium on Applications of Geophysics to Environmental and Engineering Problems, Fort Worth, TX, May 30-April 3, p. 929-936, 2009b. McKenna, J.R. and Ketchum, S.A. “Tunnel detection, monitoring, and modeling,” Session NS21A, 2006 AGU, GS, MAS, MSA, SEG, UGM Joint Assembly, Baltimore, MD, May 2006.

Sabatier, J.M. and Muir, T.G. “ Workshop on real-time detection of clandestine shallow tunnels,” National Center for Physical Acoustics, Univ. Of Mississippi, NCPA report HB0306-01 for US Army Research Office, Grant No. W911NF-06-1-001, April, 2006. Sheehan, J.R., Doll, W.E., Mandell, W., 2005, An evaluation of methods and available software for seismic refraction tomography analysis, JEEG, 10(1), 21-34, 2005. Simmons, J. and Aldridge,D. "Special session on methods for tunnel detection," Sesimo. Soc. Am Annual Meeting, Santa Fe, Nm, April 16-18, 2008.Van Schoor, M., 2002, Detection of sinkholes using 2D electrical resistivity imaging., J. Appl. Geophys. 50,393-399. US Army Research Lab, "Human, Light Vehicle, and Tunnel Detection Workshop" Beltsville, MD, June 16-17, 2009. Williamson, P.R., 1991, A guide to the limits of resolution imposed by scattering in ray tomography, Geophysics, 56(2), 202-207.

Acknowledgements This material is based upon work supported by the U.S. Department of Homeland Security under Grant Award Number 2007-ST-108-000003 and pending there approval. Field data acquisition was assisted greatly by C.R. Schmitt, S. Taylor, J.D. Heffington, and G. Brasnett.