You are on page 1of 13

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/282239001

Surface-to-tunnel electrical resistance tomography measurements

Article  in  Near Surface Geophysics · July 2015


DOI: 10.3997/1873-0604.2015019

CITATIONS READS

5 295

6 authors, including:

Kleanthis Simyrdanis P.I. Tsourlos


Laboratory of Geophysical-Satellite Remote Sensing and Archaeo-environment Aristotle University of Thessaloniki
36 PUBLICATIONS   136 CITATIONS    184 PUBLICATIONS   2,161 CITATIONS   

SEE PROFILE SEE PROFILE

Pantelis Soupios
King Fahd University of Petroleum and Minerals
168 PUBLICATIONS   1,828 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Crowie View project

Ur Project View project

All content following this page was uploaded by Nikos Papadopoulos on 29 September 2015.

The user has requested enhancement of the downloaded file.


Near Surface Geophysics, 2015, 13, 343-354  doi:10.3997/1873-0604.2015019

Surface-to-tunnel electrical resistance tomography


measurements
K. Simyrdanis1,2*, P. Tsourlos1, P. Soupios2, G. Tsokas1, J.-H. Kim3 and
N. Papadopoulos4
1
Aristotle University of Thessaloniki, Thessaloniki 541 24, Greece
2
Technological Educational Institute of Crete, Heraklion 714 10, Greece
3
Korean Institute of Geoscience and Mineral Resources (KIGAM), Daejeon 305-350, Korea
4
Foundation for Research and Technology-Hellas (Forth), Heraklion 711 10, Greece

Received March 2014, revision accepted January 2015

ABSTRACT
In this work we examine the applicability of surface-to-tunnel electrical resistance tomography
measurements for imaging subsurface targets. Various issues of this special arrangement are dis-
cussed and explored by means of numerical model simulations, experimental measurements in a
tank, and a field application. The geometrical factor was considered as a filter during the generation
of the protocols in order to reject electrode combinations and problematic measurements. Different
electrode arrays are examined and compared for various targets, and we concluded that, for the
tested cases, the pole–tripole electrode arrangement produced superior results. Additional tests
showed that the maximum distance between the electrode lines is strongly related to the nature of
the prospected target. Furthermore, the study showed that the tunnel size affects the measured data
and needs to be taken into account when processing surface-to-tunnel data, and the full incorpora-
tion of the tunnel into a 3D model proved to be the optimum among the tested correction schemes.
Finally, we present a real case study involving electrical resistance tomography surface-to-tunnel
measurements, and results are in comparable agreement with the findings of the tests with numeri-
cal and experimental data.

INTRODUCTION examined the application of ERT measurements in horizontal


Electrical resistance tomography (ERT) measurements are rou- boreholes, examining the geological conditions of the rock in
tinely used in order to map the subsurface physical properties front of a tunnel bore machine. They concluded that performing
(Loke et al. 2013). Initial limitations of surface electrode arrays small-scale resistivity tomography between the boreholes yields
regarding the resolution with depth were overcome by introduc- a better image of the geological setting and by which the opera-
ing electrodes in boreholes. Several successful applications of tor could be better prepared. Lile et al. (1994) concluded that
the cross-borehole ERT technique have been reported in the lit- severely fractured zones, which may cause difficulties for tunnel
erature in several fields (Sasaki 1992; Daily and Owen 1991; construction, can be detected with seafloor arrays using elec-
LaBrecque and Ward 1988; Morelli and LaBrecque 1996; Slater trodes inside a tunnel. Another ERT application in case of in-
et al. 2000; Goes and Meekes 2004; Coscia et al. 2011). Bing tunnel mining prospection was made by van Schoor and Binley
and Greenhalgh (2000) proposed different cross-borehole elec- (2010) who studied the applicability of tunnel-to-tunnel ERT for
trode arrays for resistivity tomography. imaging disruptive geological structures ahead of mining in an
Among existing electrode arrangements, one special case is igneous platinum mining environment.
when electrodes are placed inside tunnels. Actually, installing The present study deals with the case of ERT measurements
electrodes in tunnels is an easy way to obtain information of obtained in a combined mode using electrodes on surface and
increased sensitivity by taking advantage of the proximity of the in-tunnel. Such an arrangement is expected to provide an
tunnel to the prospected targets. increased imaging resolution for the area in between the tunnel
Application of electrodes in tunnels has been proposed in and the surface and can be used for geological, geotechnical, and
geotechnical investigations by Danielsen and Dahlin (2010) who mining applications. Only limited application of such a measur-
ing mode has been reported in literature describing surface-to-
*
ksimirda@geo.auth.gr tunnel application for very deep mining (Sasaki and Matsuo

© 2015 European Association of Geoscientists & Engineers 343


344 K. Simyrdanis et al.

1993). The objective of their study was not only to demonstrate Experimental tank: setup
the applicability of resistivity tomography to mineral exploration In addition to the simulation results, experimental tests were con-
problems but also to examine its effectiveness in mapping rock ducted in a tank to provide further examination of the theoretical
types and fracture zones. A dipole–dipole configuration was approach (Fig. 1). The capacity of the tank is 1 m3 and can hold
used, and due to practical limitations (large vertical distance weight of more than 2500 kg. During the experiment, the tank was
between surface and tunnel), transmitter and receiver dipoles filled either with drinking or deionized water (depending on the
were separated. They concluded that orebodies lie within the desired contrast between the targets and the medium). By mixing
distinct low-resistivity areas, and this correlation was the most the two liquids, the conductivity of the medium could be altered
important result obtained in their experiment, demonstrating the accordingly. A special apparatus that simulates surface-to-tunnel
applicability of surface-to-tunnel measurement mode to mineral measurements was constructed and placed inside the tank. The
exploration. apparatus consists of three parallel plastic sections with different
Although tunnel-to-surface arrangement shares many similari- colours, each one serving a different purpose. The green pipe con-
ties with the cross-hole ERT mode, the major difference lies in the tains a series of electrodes that are situated on the surface of water
fact that measurements associated with surface electrodes exhibit (“surface” electrodes). The blue coloured part contains electrodes
different sensitivities compared with measurements involving that will be located inside the water (“interior” electrodes), and the
tunnel electrodes. The vertical distance between arrays is crucial, black pipe (with diameter d=0.07 m) contains electrodes (on top of
concerning the resolving ability in the area between the arrays; it) that are in galvanic contact with the surrounding medium (even
hence, this matter is additionally studied in this work for various though the interior of the pipe is sealed off to make it air-filled).
distances. Apart from sensitivity, another issue that is studied is The whole apparatus was built with fully non-conductive (plastic)
the tunnel itself that may be a source of significant effect for the components and has the ability to adjust the vertical distance
measurements. Apart from numerical modelling in two-dimen- between the “surface,” “interior”, and “tunnel” electrodes. The
sional (2D) and three-dimensional (3D) spaces, an experimental probe spacing between electrodes is a=0.02 m with 20 electrodes
water tank was also used to study the research subjects. Finally, a placed in each pipe (60 electrodes in total). The whole apparatus
real study case took place at an existing tunnel to reconstruct the can be attached firmly to the experimental tank to ensure stability
subsurface geoelectrical and geological conditions. throughout the experimental procedure. A single-channel resistiv-
ity meter with the ability to drive up to 48 electrodes was used to
METHODOLOGY collect the experimental data in the tank.
Numerical model simulation
The 2D modelling and inversion of surface-to-tunnel data is
based on an existing proven algorithm that was previously used
to support inversion of geoelectrical data obtained with tradi-
tional arrays with electrodes on the surface or inside boreholes
(Karaoulis 2009; Tsourlos 1995). The software, developed in
MATLAB environment, uses a 2.5D finite-element forward
solver, which is based on an unstructured mesh generator (see
Fig. 2), whereas inversion is based on an iterative smoothness-
constrained scheme. The software has been appropriately modi-
fied in order to accommodate arrays with electrodes on both the
surface and the tunnel, whereas the finite-element mesh is
defined taking into consideration the electrode position.
For the study of the tunnel effect, the 3D modelling and inver-
sion software “DC3D_Pro” (Kim and Yi 2010; Yi et al. 2001)
was used to perform forward calculations where the synthetic
data were produced after constructing an entirely 3D model.
Also, 2 mV, which is the standard deviation of random noise, was
applied to the potential value of each measurement to all simula-
tion data.
Validation of the inversion images was based on the % root- FIGURE 1
mean-square (RMS) error and the correlation factor that was Experimental tank setup: Inside the tank, there is an apparatus with three
used for the synthetic data. Correlation factor compares the parallel arrays (colored green, blue, and black) with 20 equally spaced
known resistivity values of the parameters that comprise the true electrodes at every 0.02 m. Vertical distance between arrays is adjustable.
model and the resistivity values of the parameters after inversion Bottom array (black) is a pipe filled with air, which simulates tunnel
and varies from 0 (worst value) to 1 (best value). conditions.

© 2015 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2015, 13, 343-354
Surface-to-tunnel electrical resistance tomography measurements 345

FIGURE 2
Measurement arrays: (a) bipole–bipole (left), (b) pole–bipole (middle), and (c) pole–tripole (right). Black dots indicate electrode positions.

In order to take into account the boundary effects of the tank,


we corrected the measured apparent resistivities by using as a
reference the apparent resistivities that were measured for the
tank filled with homogeneous medium of known resistivity (i.e.,
water) but with the absence of any type of target. Subsequently,
correction factors calculated for the homogeneous medium were
applied to all measured apparent resistivities measured under the
presence of targets.

ARRAY CONFIGURATIONS
Among the various array configurations that could be used, we
tested the ones that were proposed by Bing and Greenhalgh (2000)
concerning vertical boreholes. However, due to different array
sensitivities, these arrays were further tested and modified in order
to be adjusted for the case of surface-to-tunnel measurements.
The array configurations that were used are shown in Fig. 2.
These are (i) bipole–bipole (‘bb’), (ii) pole–bipole (‘pb’), and
(iii) pole–tripole (‘pt’). Bipole–bipole and pole–tripole
(Leontarakis and Apostolopoulos 2012) use four electrodes
(A, B current electrodes and M, N potential electrodes), whereas
pole–bipole uses three electrodes (B as “infinite” electrode).

Geometrical factor
The geometrical factor value depends on the relative distance
and position between measurement electrodes. From all the pos-
sible combinations between electrodes A, B, M, and N (Fig. 3),
some exhibit high geometrical factor values.
In particular, the quality of the inversion results when these
measurements were included into our synthetic dataset was very
low, and we attributed this to the increased ill-conditioning that
these measurements were inflicting into the inverse problem. By
removing them using empirically decided geometrical factor
thresholds, inversion results were greatly improved.
Thus, among many possible combinations in each array con-
figuration (bb, pb, pt), we choose only those combinations whose
geometrical factor is lower than a threshold value to complete the
protocol and those that have better signal-to-noise ratios. This
optimum threshold value results after trial-and-error tests, and its FIGURE 3
actual value depends on the distance between the electrodes. Geometrical factor values for protocols bipole–bipole, pole–bipole, and
For bipole–bipole protocol, accepted electrode combinations are pole–tripole using all possible electrode combinations. Red highlighted
AM-BN, AM-NB, MA-BN, and MA-NB (rejected: AB-MN, areas indicate rejected measurements due to large geometrical factor
BA-NM, AB-NM, and BA-MN). For pole–bipole protocol, accept- values (positive or negative).

© 2015 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2015, 13, 343-354
346 K. Simyrdanis et al.

FIGURE 4
Electrical resistivity tomograms using simulation data with bipole–bipole, pole–bipole, and pole–tripole protocols. A total of 40 electrodes, with probe
spacing a=1 m and distance of surface and tunnel arrays D=5 m. Resistivity values are indicated using logarithmic scale.

FIGURE 5
Synthetic models that were tested for comparison between arrays bipole–bipole, pole–bipole, and pole–tripole.

ed electrode combinations are AM-N, MA-N, N-AM, and N-MA The initial model is a rectangular target with high resistivity, simu-
(rejected: A-MN, MN-A, A-NM, and NM-A). For pole–tripole lating a void (‘infinite’, 109 Ωm), which is embedded in a homog-
protocol, accepted electrode combinations are AMN-B and B-AMN enous half-space of 10 Ωm. The apparent resistivity values, using
(rejected: A-BMN and BMN-A). Finally, additional tests using 40 equally spaced electrodes (20 on surface and 20 inside earth)
simulation data with different models in homogenous domains with probe spacing a=1 m and array vertical distance between
showed that the optimum protocols that were used throughout this surface and internal electrodes D=5 m, were calculated using all
study are combination of the following: bb= AM-BN, pb= AM-N+N- tested protocols. From the results, it can be argued that surface-to-
AM, and pt= AMN-B+B-AMN. Note that, when electrodes are tunnel arrays can locate the target with greater success than the
before symbol ‘-’, it means that they are on surface, and after ‘-’, it individual surface or interior arrays for all protocols. Especially
means that they are inside earth (in a tunnel). when the target is closer to tunnel electrodes, the reconstructed
images with surface-to-tunnel measurements are superior with
Model simulation data respect to the surface or interior individual measurements.
Comparison of surface, interior and surface-to-tunnel arrays
The validation of array configurations has been tested using vari- Comparison of array configurations in surface-to-tunnel methods
ous simulation data. A relevant result is presented in Fig.  4. A comparison study was also undertaken between the inversion

© 2015 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2015, 13, 343-354
Surface-to-tunnel electrical resistance tomography measurements 347

results of the simulation data produced by the three protocols this also depends on the relative vertical distance of the two elec-
(bb, pb, and pt) for four models, which are shown in Fig.  5. trode lines. As the vertical distance increases, the resolving ability
Correlation factor values between the original and inversion decreases, and apparently, there is a threshold distance above which
models are shown in the bar chart in Fig. 6. surface-to-tunnel imaging becomes ineffective. The effect of the
Further, inversion results of model ‘a’ (Fig. 5) are shown in distance (D) between the surface and tunnel electrodes was studied,
Fig.  7, where a resistive (10000 Ωm) and a conductive (1 Ωm) which is expressed as a ratio related to the unit electrode spacing
target are embedded in a half-space of 100 Ωm. As a conclusion (a). Different ratio values between the arrays (D) and probe spacing
from the comparison, it can be said that systematically pole–tri-
pole protocol has better results than the other two protocols and
bipole–bipole is the second best protocol to use. In particular, for
target ‘a’, as shown in Fig. 7, the pole–tripole protocol is able to
reconstruct equally good both targets, although when using the
other protocols, the conductive target is more difficult to depict.

Experimental tank
A number of different targets were tested with the experimental
tank. One such target was a metal plate placed vertically in
between surface (green pipe) and interior (blue pipe) electrodes
embedded in the tank, which was filled with water having
100‑Ωm resistivity, as shown in Fig. 8.
The vertical distance between surface and interior electrodes
was adjusted to D=15 cm apart. In Fig. 9, we can see the inverted
results from the raw data. Comparing surface, interior, and surface-
to-tunnel arrays, it can be argued that surface measurements locate
the upper part of the target, and, respectively, interior measure-
ments locate the lower part of the target. Using surface-to-tunnel
measurements, bipole–bipole reconstructs the middle part of the
target. Pole–bipole and pole–tripole protocols reconstruct the target
in its full extent and length. In all protocols, the superiority of the
surface-to-tunnel array, compared with that of surface or interior
measurements, is evident and pole–tripole seems to exhibit the best
performance regarding the reconstruction of the metal plate.
FIGURE 7
OPTIMUM VERTICAL DISTANCE Resistivity inversion models for the simulation data, using bipole–bipole,
Surface-to-tunnel measurements can have an increased resolving pole–bipole, and pole–tripole protocols. A total of 40 electrodes, equally
capability between the surface and interior electrodes. However spaced (a=1 m) and distance of arrays D=5m. Resistivity values are
indicated using logarithmic scale.

FIGURE 8
FIGURE 6 A metal plate was used as one of the targets at the experimental tank
Correlation factor values for different synthetic models (shown in Fig. 5) tests. The tank was filled with water whose resistivity value is measured
comparing three different protocols (bb, pb, pt). a priori (100 Ωm).

© 2015 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2015, 13, 343-354
348 K. Simyrdanis et al.

FIGURE 9
Resistivity inversion results of a metal plate in water in the experimental tank using protocols bb, pb, pt (columns), testing surface (first row), interior
(second row), and surface-to-tunnel (third row) arrays. Resistivity values are indicated using logarithmic scale.

FIGURE 10
Inversion simulation results using a resistive (5000 Ωm) and a conductive (1 Ωm) target in homogeneous medium (100 Ωm) for different distance ratio
D/a (‘D’ vertical distance between horizontal arrays, ‘a’ probe spacing). Resistivity values are indicated using logarithmic scale.

(a) are studied for various targets (Fig. 10). In this figure, inversion It is shown that, as the ratio increases, the inversion results are
results are shown for a resistive and a conductive target using pro- deteriorating, but this is not happening in a uniform way since
tocols ‘bb’, ‘pb’, and ‘pt’ for distance ratios: D/a= 3, 5, and 7.5. inversion results depend also on the electrode array used and on

© 2015 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2015, 13, 343-354
Surface-to-tunnel electrical resistance tomography measurements 349

the target characteristics (resistive or conductive). For ratio 3/1 results are overlapped in Fig. 11 for a comparison with the ana-
arrays, ‘bb’ and ‘pt’ have comparative inversion results with ‘pb’ lytic solutions. The resistivity of the background medium was set
appearing slightly worse. For greater ratios, pole–tripole proto- to 102 Ωm, and the tunnel was modelled by a high resistivity,
col has the best results, and generally, it can be said that the 104 Ωm.
resistive target is less sensitive to the vertical distance of the The highest numerical error, about 10%, is registered when
arrays. On the basis of the correlation factor values, it can be said the distance between the current and potential electrodes is the
that the ratio 5/1 seems to be the limit beyond which reconstruc- shortest in the dipole–dipole array. The remaining parts are
tion of targets is problematic. Apparently, the decreasing resolu- almost under 5%. We interpreted that these discrepancies are
tion resulting from an increasing distance between the surface mainly due to simulating the tunnel, an insulating object, with a
and tunnel electrodes could be countered to some extent by finite high resistivity. Although the numerical error increases as
increasing the inter-electrode spacing a. the spacing between current and potential electrodes is shorter,
the 3D FEM modelling scheme would have enough accuracy to
TUNNEL EFFECT study the effects of a tunnel in an arbitrarily heterogeneous
In the surface-to-tunnel arrangement, the tunnel itself is a body medium.
of infinite resistivity, which can seriously affect the measure-
ments. In practice, the tunnel is a highly resistive target that is In a heterogeneous medium
located next to the interior measuring electrodes. The tunnel In an attempt to further examine the tunnel effect, a 3D simula-
response can contaminate the geoelectrical signal coming from tion model was created with a tunnel having diameter d=6 m (see
the subsurface targets. Depending on the tunnel’s size and the configuration details in Fig.  13). Using 40 electrodes placed
electrode arrangement that is used, the extent of tunnel effect can every 1 m (20 on surface and 20 in-tunnel), the tunnel effect is
vary from negligible to extremely important. calculated for homogeneous ground for the three protocols (bb,
Consequently, the tunnel effect needs to be studied, and its pt, and pb) using the program ‘DC3D_Pro’. The % estimated
effect needs to be taken into account when processing surface-to- error (the divergence from homogeneity of the apparent resistiv-
tunnel ERT data. This is not always an easy task since the tunnel ity values due to the presence of a tunnel) for all protocols using
is a fully 3D body, often with irregular dimensions, while the all measurements is shown in Fig. 12.
geometry used in the described tunnel-to-surface arrangement is As shown in Fig.  12a, the bipole–bipole dataset exhibits a
practically 2D. Here, the tunnel effect is studied for the case of a significant but less variable amount of error compared with the
homogeneous (without target) and a heterogeneous (using a tar- other protocols. This is due to the fact that the array has two
get) medium. electrodes (AM) on surface, which are less sensitive to the tun-
nel (at least the measurements with small separation), whereas
In a homogenous medium the “tunnel” electrodes BN are highly sensitive to the existence
The study of the tunnel effect for homogeneous ground is based
on an analytical solution proposed by Wait (1982) for a single
current electrode placed at the wall of a cylindrical tunnel. We
calculated the apparent resistivity of several in-tunnel arrays for
different inter-electrode spacings (a) in relation to a tunnel of a
fixed diameter (d), which is expressed as a ratio a/d: i.e., a ratio
of 1 suggests that the array inter-electrode spacing is equal to the
diameter of the tunnel. The results are depicted in Fig. 11 where
the tunnel diameter is expressed as a ratio of the inter-electrode
spacing for five arrays: pole–pole, pole–dipole, dipole–dipole,
Schlumberger, and Wenner arrays. The figure clearly shows that
the tunnel effect cannot be ignored particularly for smaller ratios
(i.e., a/d <5). Among the five different kinds of the tested con-
ventional arrays, the tunnel effects are smaller for the pole–pole
array and more significant for the dipole–dipole array. The
effects in the Schlumberger array data are exactly same as those
in the pole–dipole array. FIGURE 11
For studying the responses of arbitrarily heterogeneous media Changes of apparent resistivity with respect to the ratio between the
involving tunnels, we utilized a 3D finite-element method (FEM) distance of the positive current and potential electrodes to tunnel diam-
modelling routine of the software ‘DC3D_Pro’. To study the eter. Simulation results by using an analytic solution are shown as solid
feasibility of using the numerical modelling scheme, the same and dashed lines and those by the 3D modeling routine of ‘DC3D_Pro’
model was generated using the software and the simulation as symbols.

© 2015 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2015, 13, 343-354
350 K. Simyrdanis et al.

of the tunnel. As a result, the overall sensitivity to the tunnel is the tunnel since electrodes are mostly on surface while the oppo-
a weighted average of the sensitivities of the surface (AM) and site holds for the B-AMN configuration. Further, for the AMN-B
tunnel (BN) electrode pairs. As the AM and BN spacing part of the data, the increase in the error due to the tunnel effect
increases, the surface electrode pair (AM) will become more is associated with configurations of increased spacing: as surface
sensitive to the tunnel, but at the same time, the tunnel elec- electrode spacing increases, measurements can “sense” more of
trode pair (BN) will become less sensitive to the tunnel, as the underground tunnel. Conversely, for the B-AMN configura-
shown in the example in Fig.  12. It follows that the overall tions, the higher errors are associated with measurements having
sensitivity is not highly changing with the increase in electrode small electrode separations since these configurations are more
spacing. sensitive to the tunnel, as it was already shown in the tunnel
In Fig. 12b, the pole–tripole dataset shows two different lev- effect study in Fig. 12b. The same observations hold also for the
els of errors. Data collected with the AMN-B configuration (left errors that the “pb” array exhibits, which are shown in Fig. 12c.
part of Fig.  12b) are, on average, significantly less sensitive to The ability to model the tunnel effect by simulating the tunnel
in a fully 3D mode can be used in three ways.
(i) The simplest possible way is to assume that the tunnel effect
is linearly added to the total effect and try to remove it. This
can take place by calculating the % error that the tunnel
effect is introducing into the apparent resistivity measure-
ments for homogeneous ground and then correct the original
dataset using the obtained error factors. This type of
approach has been used in geoelectrical prospecting in order
to correct for the topography effect (Tsourlos, Szymanski,
and Tsokas 1999). However the linearity assumption does
not hold when the tunnel effect is significant. This was veri-
fied by several tests that were performed (not shown here),
and we concluded that this type of correction can easily
result into invalid results; hence, it cannot be recommended.
(ii) Rather than using the calculated % error introduced by the tun-
nel to correct the data, we could use it as an indicator to reject
data from the entire dataset. After calculating the % error that
the tunnel effect is introducing into the apparent resistivity
measurements, threshold error values can be defined: measure-
ments exhibiting errors above the threshold values are excluded
from the dataset. As it can be easily concluded by the analysis
presented in the previous section (see Fig.  12), the measure-
ments that are rejected are either the ones with large spacing
when it comes to configurations with surface electrodes or the
ones with small spacing when the configurations having tunnel
electrodes are involved. The technique can be easily applied
although it has two main disadvantages: first, there is no objec-
tive way to define the threshold value, which therefore needs to
be defined empirically or based on the parameters of the used
configuration (electrode spacing and number of electrodes);
second, since data are excluded, useful information regarding
the near-tunnel region is also rejected.
(iii) The third and most optimum way to reduce the tunnel effect
is to approach the data using a 3D modelling and an inver-
sion program, where the tunnel is simulated using its actual
dimensions in the 3D space. In this way the tunnel is fully
FIGURE 12 incorporated into the inversion procedure so its response is
Estimated percentage error (absolute values) of the apparent resistivity taken directly into account. The particular correction
values calculated for a homogeneous medium with and without the pres- procedure requires the application of a fully 3D scheme (see
ence of a tunnel for protocols (a) bb (AM-BN), (b) pt (AMN-B+B- Fig. 14). On the other hand, given the measurement arrange-
AMN), and (c) pb (AM-N+N-AM). ment, the inverse problem could be approached in the 2D

© 2015 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2015, 13, 343-354
Surface-to-tunnel electrical resistance tomography measurements 351

FIGURE 13
Simulation (created at ‘DC3D_Pro’ Kim and Yi 2010; Yi et al. 2001) seen as y-z, x-z, and x-y cross views respectively, with high resistive body
(109 Ωm) above a tunnel embedded in homogenous half-space (100 Ωm). Black dots indicate electrode positions.

space: the 3D parameters need to be defined with a large


(i.e., infinite) length across the Y-axis in order to cope with
the inherent 2D nature of the borehole-to-surface inverse
problem (see Fig. 14).
The techniques described above were tested with a synthetic
model showing a rectangular high resistive target (109 Ωm) that
is placed above a tunnel of a diameter d=6 m (see configuration
details in Fig. 13) using 40 electrodes placed every 1 m (20 on
surface and 20 in-tunnel).
Figure 15 depicts the inverted images when inverting raw data
(first row), after correction of data (second row), and inverting
data using a 3D modelling and inversion program. Corrected data
(second row) result after rejecting noisy measurements accord-
ing to Fig.  12 (rejected measurements with small spacing for FIGURE 14
bipole–bipole protocol, rejected measurements with large spac- 3D model using 2.5D parameters with the presence of a tunnel. White
ing for combinations AM-N and AMN-B, and small spacing for dots indicate electrode positions.
combinations N-AM and B-AMN for protocols pole–bipole and
pole–tripole, respectively). of the Roman period that was constructed for transferring water
The inversion results confirm that the less sensitive protocol is was chosen as a test site to apply surface-to-tunnel measure-
bipole–bipole (but it is also less sensitive to variations of resistiv- ments developed in this work (Fig. 16).
ity due to target appearance). On the contrary, pole–bipole seems The tunnel’s diameter varies between 2-m and 3-m heights
to be the most sensitive protocol in the presence of the tunnel. and 2-m width with approximately 40-m length (Fig. 17). Above
Pole–tripole protocol is also sensitive to tunnel effect. After cor- the tunnel, there is a layer of recent deposits with a thickness of
rection, pole–tripole presents better results compared with bipole– D=8 m. In total, 60 electrodes were simultaneously used (30 on
bipole or pole–bipole although there are some artefacts close to surface and 30 inside the tunnel), where the first and last elec-
the target and the tunnel electrodes. When inverting data using the trodes (inside the tunnel) were placed 5 m away from the tunnel
3D modelling program, despite the fact that % RMS error has edges to avoid errors related to side effects. Probe spacing was
higher values than the inversion results of the corrected data, the chosen to a=1 m, and a ten-channel resistivity meter was used to
target and the background area are reconstructed in a better way acquire data.
(less artefacts close to the target and tunnel electrodes). Figure 18 shows inverted results comparing surface and sur-
face-to-tunnel measurements. When using both (surface and
FIELD APPLICATION in-tunnel) arrays, the stratigraphy of the area is clearer than the
An ERT field survey was conducted at the archaeological site of results obtained using only surface electrodes. Additional infor-
the ancient city of Eleftherna (Crete, Greece). An ancient tunnel mation close to the interior electrodes can be provided in the

© 2015 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2015, 13, 343-354
352 K. Simyrdanis et al.

FIGURE 15
Inversion for protocols bb, pb, and pt with raw data (first row), with corrected data (second row) and with data inverted with the 3D program
‘DC_3DPro’ (third row). Resistivity values are indicated using logarithmic scale on first two rows and normal scale at last row.

Figure 19 shows inverted results where we can see a superfi-


cial high-resistivity layer (200 Ωm–300 Ωm) of about 2 m–3 m
corresponding to the recent deposits. Below that, a layer of
Miocene marl and marly limestone is registered with relatively
lower resistivity signature (100 Ωm–160 Ωm). The tunnel itself
has been constructed in this geological formation.

CONCLUSIONS
The effectiveness of the approach using surface-to-tunnel elec-
trode array is demonstrated using several modelling scenarios
(with numerical simulation and experimental data). A criterion
based on the values of the geometrical factor for generating
protocols is suggested. The tests showed that pole–tripole
seems to be most appropriate protocol to use. The vertical dis-
tance between surface and tunnel arrays is crucial for the
resolving capabilities of the array, and the extensive test
showed that the optimum ratio is D/a=5/1 for conductive and
7/1 for resistive targets, respectively, where ‘D’ is the vertical
distance between surface and tunnel electrodes and ‘a’ is the
probe spacing.
It is clear that the above conclusions are highly dependent on
FIGURE 16 the selection of the electrode array and overall measurement
(a) Location of field site (Eleftherna, Crete). (b) Surface electrodes. geometry. Given the recent significant progress in the field of
(c)  Tunnel electrodes. geoelectrical array optimization (Stummer, Maurer, and Green
2004; Wilkinson et al. 2006; Wilkinson et al. 2012), it follows
proximity to the tunnel. On the other hand, the surface electrodes that optimized arrays are certainly worth testing in the frame-
alone do not provide sufficient information in greater depth. work of the surface-to-tunnel configurations. Given the existing

© 2015 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2015, 13, 343-354
Surface-to-tunnel electrical resistance tomography measurements 353

experience of the application of optimized arrays in other meas- dure are proposed. It is shown that the optimum solution to
uring modes (surface, cross-hole, etc.), we expect them to pro- compensate for the tunnel effect is to invert the data using a 3D
duce improved results for the case of surface-to-tunnel measure- inversion modelling program where the tunnel is simulated in a
ments; therefore, we are targeting our future research towards 3D space with its actual dimensions. Overall, the above approach
this direction. forms a proposed procedure for selecting situation-based meas-
It is demonstrated that the tunnel itself can affect the measure- urement protocols for performing surface-to-tunnel electrode
ments as it is an ‘object’ with infinite resistivity, so it should be measurements. The presented real case study verified the
incorporated during the post-processing of the data. Furthermore, findings of the numerical simulations and demonstrated the
ways for including the tunnel effect into the processing proce- applicability of the proposed approach.

FIGURE 17
A schematic that shows electrode
position and tunnel dimensions of
survey area.

FIGURE 18
Inversion results comparing surface (left) and surface-to-tunnel measurements using protocol bipole–bipole. Resistivity values are indicated using
logarithmic scale.

FIGURE 19
Inversion results using bb, pb, and pt (columns) protocols indicating raw (top row) and corrected (bottom row) data. Resistivity values are indicated
using logarithmic scale.

© 2015 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2015, 13, 343-354
354 K. Simyrdanis et al.

ACKNOWLEDGEMENTS Morelli G. and LaBrecque D. 1996. Advances in ERT inverse modeling.


This work is supported by the EU project: “Operational Program European Journal of Environmental and Engineering Geophysics 1,
Educational and Lifelong Learning, investing in knowledge soci- 171–186.
Sasaki Y. 1992. Resolution of resistivity tomography inferred from
ety, Ministry of Education & Religious Affairs, Lifelong Learning
numerical simulation. Geophysical Prospecting 40, 453–463.
and Religious Affairs, NSFR, 2007-2013”. Sasaki Y. and Matsuo K. 1993. Surface-to-Tunnel Resistivity Tomography
at the Kamaishi Mine. Butsuri-Tansa 46, 128–133.
REFERENCES Slater L., Binley A.M., Daily W. and Johnson R. 2000. Cross-hole elec-
Bing Z. and Greenhalgh S.A. 2000. Cross-hole resistivity tomography trical imaging of a controlled saline tracer injection. Journal of
using different electrode configurations. Geophysical Prospecting 48, Applied Geophysics 44, 85–102.
887–912. Stummer P.B., Maurer H. and Green A. 2004. Experimental design:
Coscia I., Greenhalgh S.A., Linde N., Doetsch J., Marescot L., Günther electrical resistivity data sets that provide optimum subsurface infor-
T. et al. 2011. 3D crosshole ERT for aquifer characterization and mation. Geophysics 69(1), 120–139.
monitoring of infiltrating river water. Geophysics 76, G49–G59. Tsourlos P.I. 1995. Modeling, interpretation and inversion of multielec-
Daily W. and Owen E. 1991. Cross-borehole resistivity tomography. trode resistivity survey data. PhD thesis, Department of Electronics,
Geophysics 56, 1228–1235. University of York, U.K.
Danielsen B.E. and Dahlin T. 2010. Numerical modeling of resolution Tsourlos P.I., Szymanski J.E. and Tsokas G.N. 1999. The effect of terrain
and sensitivity of ERT in horizontal boreholes. Journal of Applied topography on commonly used resistivity arrays. Geophysics 64(5),
Geophysics 70, 245–254. 1357–1363.
Goes B.J.M. and Meekes J.A.C. 2004. An effective electrode configuration van Schoor M. and Binley A. 2010. In-mine (tunnel-to-tunnel) electrical
for the detection of DNAPLs with electrical resistivity tomography. resistance tomography in South African platinum mines. Near Surface
Journal of Environmental and Engineering Geophysics 9, 127–141. Geophysics 8, 563–574.
Karaoulis M. 2009. Inversion algorithm development for time lapse geoe- Wait J.R. 1982. Geo-electromagnetism. Academic Press Inc.
lectrical data. PhD thesis, Aristotle University of Thessaloniki, Greece. Wilkinson P.B., Chambers J.E., Meldrum P.I., Oglivy R.D. and Count S.
Kim J.-H. and Yi M.-J. 2010. DC3DPro: 3D Inversion of ERT data. User’s 2006. Optimization of array configuration and geometries for the
Manual. Korea Institute of Geoscience and Mineral Resources, Korea. detection of abandoned mineshafts by 3D cross-hole electrical resis-
LaBrecque D. and Ward S. 1988. Two-dimensional inversion of cross- tivity tomography. Geophysical Journal International 167, 1119–
borehole resistivity data using multiple boundaries. 58th SEG annual 1126.
international meeting, Expanded Abstracts, 194–197. Wilkinson P.B., Loke M.H., Meldrum P.I., Chambers J.E., Kuras O.,
Leontarakis K. and Apostolopoulos G.V. 2012. Laboratory study of the Gunn D.A. et al. 2012. Practical aspects of applied optimized survey
cross-hole resistivity tomography: the model stacking (MOST) tech- design for electrical resistivity tomography. Geophysical Journal
nique. Journal of Applied Geophysics 80, 67–82. International 189, 428–440.
Lile O.B., Backe K.R., Elvebakk H. and Buan J.E. 1994. Resistivity Yi M.-J., Kim J.-H., Song Y., Cho S.-J., Chung S.-H. and Suh J.-H. 2001.
measurements on the sea bottom to map fracture zones in the bedrock Three-dimensional imaging of subsurface structures using resistivity
underneath sediments. Geophysical Prospecting 42, 813–824. data. Geophysical Prospecting 49, 483–497.
Loke M.H., Chambers J.E., Rucker D.F., Kuras O. and Wilkinson P.B.
2013. Recent developments in the direct-current geoelectrical imaging
method. Journal of Applied Geophysics 95, 135–156.

© 2015 European Association of Geoscientists & Engineers, Near Surface Geophysics, 2015, 13, 343-354

View publication stats

You might also like