Journal of Applied Geophysics 99 (2013) 6065

Contents lists available at ScienceDirect

Journal of Applied Geophysics

journal homepage: www.elsevier.com/locate/jappgeo

Using near-surface seismic refraction tomography and multichannel

analysis of surface waves to detect shallow tunnels: A feasibility study
Steven D. Sloan a,, Jeffery J. Nolan b, Seth W. Broadfoot a, Jason R. McKenna c, Owen M. Metheny a
a
XRI Geophysics, LLC, 6207 Highway 80, Vicksburg, MS 39180, USA
b
Kansas Geological Survey, 1930 Constant Avenue, Lawrence, KS 66047, USA
c
U.S. Army Engineer Research & Development Center, 7701 Telegraph Road, Alexandria, VA 22315, USA

a r t i c l e i n f o a b s t r a c t

Article history: Near-surface seismic refraction and surface wave data were collected at a site to determine the feasibility and
Received 23 May 2013 limitations of using these seismic methods to detect and localize a shallow tunnel in unconsolidated sediments.
Accepted 17 October 2013 Data sets were collected both before and after the construction of the tunnel. We were able to detect the air-lled
Available online 26 October 2013
cavity using multichannel analysis of surface waves. The refraction tomography results showed the tunnel
location in the raypath coverage plots, but only small velocity variations were observed. In tandem the two
Keywords:
Near-surface
methods would reduce false positives, but individually the false alarm rate would likely be high due to non-
Seismic uniqueness of the results. In this geologic setting, these methods are not the best choice of geophysical methods
Refraction to detect clandestine tunnels and should be combined with other geophysical techniques to improve and
Surface wave constrain interpretations.
Void 2013 Elsevier B.V. All rights reserved.

1. Introduction Multiple examples of different geophysical techniques have been

Illegal tunnels pose a threat to several nations around the world for
various reasons including narcotic trafcking (USAMexico border),
unregulated trade (EgyptGaza border), and attacks (IsraelGaza
border), to name a few. Research using geophysical methods to detect
these tunnels has been ongoing for multiple decades with much of the
early work focusing on large and deep targets in hard rock environ-
ments on the Korean peninsula (Ballard, 1982; Rechtien et al., 1995)
and more recent work looking at shallower features in unconsolidated
sediments (Llopis et al., 2005; Tucker et al., 2007). Despite the volume
of previous studies, no individual technology or method has been
identied or developed that can detect and localize clandestine tunnels
efciently, consistently, and across a variety of geological settings.
Whereas one method, such as GPR, may work great at a particular site,
it may not work at all at another depending on subsurface properties
such as clay content, and dielectric permittivity, or target parameters
(depth, size) and the same can be said for all geophysical methods,
not just GPR. As with most geophysical studies, appropriate methods
are chosen based on the goal of the study and properties of the site to
be surveyed. The study presented here focuses on the use of seismic
refraction tomography and multichannel analysis of surface waves
(MASW) to determine the potential of the two methods for tunnel
detection at shallow depths.
Corresponding author. Tel.: +1 407 592 2171.
E-mail addresses: Steve.Sloan@xrigeo.com (S.D. Sloan), nolanjj2@gmail.com
(J.J. Nolan), seth.broadfoot@xrigeo.com (S.W. Broadfoot), Jason.McKenna@gmail.com
(J.R. McKenna), owen.metheny@xrigeo.com (O.M. Metheny).
applied to tunnel detection (Cho et al., 2006; Choi and Ra, 1999;
Greeneld et al., 1991; Llopis et al., 2005; Mahrer and List, 1995;
Rechtien et al., 1995; Sloan et al., 2011), including seismic, electromag-
netic, and radar, among others. Near-surface seismic methods in partic-
ular have been used for both general void detection (Branham and
Steeples, 1988; Dobecki, 1988; Inazaki et al., 2005; Peterie et al., 2009)
and tunnel detection (Belfer et al., 1998; Llopis et al., 2005; Rechtien
et al., 1995; Sloan et al., 2011; Tucker et al., 2007; Walters et al., 2007,
2009). From a theoretical standpoint, seismic methods would be a
good choice for void detection due to the drastic change in seismic prop-
erties from a geologic medium to an air-lled void (Sloan et al., 2011).
Sheehan et al. (2005) describe an example of locating an interpreted
water-lled void in a karst environment using seismic refraction tomog-
raphy at a depth of approximately 20 m. The cavity in this case was
much larger than a typical tunnel, but did exhibit noticeable variations
in the P-wave velocity prole. More recent examples have applied
refraction tomography methods to detect voids at depths of 0.6 and
6 m. Hickey et al. (2009) buried a plastic pipe at 0.6 m depth using
cut-and-ll and subsequently completed a seismic refraction survey
orthogonal to the buried pipe. The authors noted reduced P-wave
velocity (VP) around and above the pipe; however, the method of
emplacement also disturbed the overlying material, which would be
expected to produce a similar result. Riddle et al. (2010) used refraction
tomography to detect a concrete tunnel 1m1.6m in size approximately
6 m deep. Their results show subtle changes in raypath coverage and VP
compared to the surrounding material. The main difference between
past studies and the work presented here is that an actual tunnel is
used that was constructed in a similar fashion to illegal cross-border

0926-9851/$ see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jappgeo.2013.10.004
 S.D. Sloan et al. / Journal of Applied Geophysics 99 (2013) 6065 61

tunnels, providing a more representative target and removing extrane-

ous inuences such as overburden that has been disturbed by target
emplacement. This study is the rst that we are aware of to use an
actual tunnel in a controlled environment for research purposes.
The objective of the research presented here was to determine the
feasibility and limitations of using near-surface seismic refraction
traveltime tomography and MASW methods to detect and locate a
small-diameter, shallow tunnel in an environment comprised of dry
unconsolidated materials. Baseline data were collected prior to the con-
struction of the tunnel and are compared with coincident data collected
after the tunnel was completed. Results of this study show that these
methods can be used to detect voids in the right geologic setting; how-
ever, individually the results of one method alone may be inconclusive
and would be best used by combining with other seismic or geophysical
methods to reduce uncertainty and increase condence in the results.
This work was done as part of an undergraduate senior research project.
2. Site description
The tunnel used for this study was dug using a 6 m by 6 m vertical
shaft for entry, exit, and spoil removal. Digging was done using
mechanical hand tools similar to those discovered in recent tunnel
seizures along the southwest US border, such as hammer drills, to
accurately represent the target of interest. Overlying and surface
material was not disturbed during construction. The roof of the tunnel
is located at 3 m depth and the tunnel has a size of 1.25 m width and
1.25 m height (Fig. 1). The tunnel is shored using wooden beams and
the walls and ceiling are also lined with wooden boards.
The site of investigation is located in the northeastern portion of the
Great Basin near the Great Salt Lake, a sub-province of the Basin and
Range province in the United States. The area of investigation is
underlain by a thin layer of Holocene-age eolian sheet-sand deposits
overlying Pleistocene-age lacustrine deposits related to the former
existence of Lake Bonneville. Drilling at the site yielded 0.51.0 m of
eolian sheet sands across the entire site, underlain by ne-grained
lacustrine deposits. The eolian deposits consist of ne-grained, loose
to medium-dense silty sand and sandy silt. The lacustrine deposits are
comprised of alternating layers of silt, sandy silt, and silty sand, overly-
ing gravelly sands and clayey sand toward the bottom of the borings.
3. Methods
Multiple shallow seismic data sets were collected at the site including
refraction tomography (P and S), and MASW over a three day period
in July and again in November of 2010 (Fig. 2). Each survey was
conducted along a coincident line with similar acquisition geometries
Fig. 2. Illustration depicting the layout of the surface wave line (a), S-wave lines (b), and
P-wave lines (c). Note that all data sets were collected along the same coincident line
(a), but have been spread out for illustration purposes as indicated by the dashed arrows.
and parameters (Table 1). Data sets were acquired both before and
after construction of the tunnel, besides the MASW data set which
was acquired only after. The pre-construction P-wave seismic refraction
data were collected using 144 100-Hz vertical-component geophones
with 0.25 m spacing. The source was a 1.36 kg (3 lb) hammer struck
on a steel plate with 0.5 m spacing. The rst shot point was 5 m away
from the rst geophone and the last shot point was 5.25 m beyond the
last geophone. Ninety-three source locations were occupied over a
total distance of 46m. Data were recorded using six 24-channel seismo-
graphs with 24-bit A/D conversion, 0.25-ms sampling interval, and
256-ms trace lengths. Post-construction data were acquired using 144
40-Hz vertical-component geophones with 0.25 m spacing due to
equipment availability constraints, but the other acquisition parameters
remained the same.
Shear-wave refraction data were also acquired using the same
parameters with the exception that 14.5-Hz horizontal-component
geophones were used, with the same hammer impacting a horizontal
shear block for the source. The source and receivers were oriented to
collect horizontally polarized (SH) data. Each data set was collected
during the same visit for both pre- and post-construction surveys. How-
ever, the post-construction S-wave data proved too noisy to reliably
pick rst breaks due to wind noise and required another return visit
one year later in the fall of 2011. Surface wave data were collected
using 96 4.5-Hz vertical-component geophones spaced at 1-m intervals.
An accelerated weight drop provided the input energy every 1 m. The
rst source location was 20 m in front of the rst geophone and ended
8 m past the last geophone. This resulted in a total of 124 shot locations.
Seismic refraction tomography methods typically utilize a grid of
either xed or variable sized cells to represent the subsurface. Forward
modeling methods, such as a nite difference method, are used to pre-
dict ray paths and travel times between source locations and receivers.
Cell velocities are iteratively adjusted until the mist between the
observed and predicted travel times is within some acceptable range.
In this case the wavepath eikonal traveltime (WET) inversion scheme
is used (Schuster and Quintus-Bosz, 1993).

Table 1
Summary of the seismic acquisition parameters for the different lines collected.

Seismic acquisition parameters

Seismic line Receiver Source Channels Geophone Source

spacing spacing

Pre-construction
Refraction (P) 0.25 m 0.5 m 144 100-Hz Hammer
Refraction (S) 0.25 m 0.5 m 144 14.5-Hz Hammer

Post-construction
Refraction (P) 0.25 m 0.5 m 144 40-Hz Hammer
Refraction (S) 0.25 m 0.5 m 144 14.5-Hz Hammer
MASW 1m 1m 96 4.5-Hz Weight drop
Fig. 1. Picture of the tunnel used in this study during construction.
62 S.D. Sloan et al. / Journal of Applied Geophysics 99 (2013) 6065

Fig. 3. Representative overtone image and interpreted fundamental mode dispersion curve (indicated by the line with white squares) for the MASW line.

The data were processed using commercial software packages,
including SurfSeis 3 and Rayfract. The refraction data were processed
by inputting raw eld les, dening geometry, and picking the rst
arrivals for each shot. There are a total of 13,392 time-offset pairs from
144 rst arrival picks on each of 93 different shot locations. The cell
size used was 12.5 cm on each side, for a total of 100 cells representing
the tunnel. The surface wave data processing included geometry deni-
tion, overtone analysis, manual picking of fundamental mode dispersion
curves (Fig. 3), and inverting the dispersion curves to calculate shear-
wave velocity. In-depth explanations on the theory and processing for
MASW and refraction tomography can be found in Park et al. (1999),
Xia et al. (1999), and Sheehan et al. (2005). All source and receiver loca-
tions were used for the refraction analysis; however, the surface wave
eld les were cut to mimic an off-end conguration with constant
source-to-receiver offset. We chose a 1-m offset with a 24-channel
spread in this case after testing various parameters.
4. Results and discussion
The pre- and post-construction ray coverage plots for the P- and
S-wave data are displayed in Figs. 4a and 5a, respectively. The pre-
and post-construction VP and VS proles are shown in Figs. 4b and 5b,
top and bottom, respectively. The void is located at 19 m along the
x-axis in Fig. 4 and 17.5 m in Fig. 5, with the top at approximately 3-m
depth (the center of the dashed circle). The lateral discrepancy in the
void location between the P and S data sets is due to the monument
used as a reference point for positioning the lines being accidentally
dislodged between the collections of the pre-construction P- and
S-wave lines. VP remains relatively constant from pre- to post-
construction; however, there is a slight decrease in the post-
construction VP where the void is located, dropping from approximately
550 m/s to 450 m/s. The VP proles exhibit a change in velocity of
approximately 22% when comparing the before and after. VS drops

Fig. 4. P-wave velocity proles (right) and ray coverage proles (left) produced by the seismic refraction tomography surveys. The top proles are pre-construction velocity and the
bottom proles are post-construction. There is a slight decrease in velocity and change in raypath coverage after construction of the void. The middle of the circle indicates the position
of the tunnel.
 S.D. Sloan et al. / Journal of Applied Geophysics 99 (2013) 6065 63

Fig. 5. S-wave velocity proles (right) and ray coverage proles (left) produced by the seismic refraction tomography surveys. The top proles are pre-construction and the bottom proles
are post-construction. There is a slight decrease in velocity and change in raypath coverage after construction of the void. The middle of the circle indicates the position of the tunnel.

from approximately 400 to 325 m/s, which is a 23% change. Intuitively
we expected a greater change in the S-wave data compared to the
P-wave data, but the percent change is very similar. Although both the
P-and S-wave results show changes greater than 20%, seismic velocity
variations of that scale are not uncommon in the shallow subsurface.
Even within a short ~35-m long line there are changes equal to or
greater than those observed at the tunnel location and do not appear
anomalous.
The left-hand side of Figs. 4 and 5(a) shows raypath coverage plots
for the pre- and post-construction cases (top and bottom, respectively)
for the P- and S-wave data. Seismic waves propagate in accordance with
Fermat's principle of least time, so it is expected that fewer rays pass
through the void area due to the decreased velocity. This response is
expected since the excavation of geologic material has left behind an
air-lled void with a velocity of approximately 335 m/s, which is less
than that of the surrounding medium. Comparing the before and after
plots in Fig. 4 shows a decrease in the number of rays from ~1400 to
600, representing a decrease of 57% in the P-wave data. The raypath
coverage of the S-wave data drops from approximately 1780 to 730,
marking a 41% change. In this example, the raypath coverage plot
does the best job of highlighting the inuence of the void and presents
a noticeable contrast in both the P- and S-wave data. Both pre- and
post-construction plots are comparable, containing similar features in
both. The contrast between the surrounding medium and the void
would become even more pronounced in an environment with higher
seismic velocities.
Fig. 6 shows difference plots calculated by subtracting the post-
construction grid les from those of the pre-construction. The difference
in ray coverage is on the top and velocity is on the bottom. P-wave data
are on the left in Fig. 6a and S-wave data are on the right in Fig. 6b. As
with the pre- and post-comparison plots, the change in ray coverage is
the most evident, indicated by the areas within the dashed circles. The
high events to the left of the void location and the low events beneath
it in the P-wave data are not representative of the void, but are a
byproduct of the differencing where the rays have traveled around the
void location. The velocity difference plot shows minor changes in the
velocity where the tunnel is located, but nothing that sticks out anoma-
lously compared to the surroundings.

Fig. 6. Difference plots for ray coverage (top) and velocity (bottom), calculated by subtracting the post-construction data from the pre-construction data for the P- (left) and S-wave (right)
data. The middle of the circle indicates the position of the tunnel.
64 S.D. Sloan et al. / Journal of Applied Geophysics 99 (2013) 6065

the tunnel location, marked by the high-velocity halo. These high-

Fig. 7. A shear-wave velocity prole produced by the MASW method. The high-velocity
halo is noticeable in the locality of the void, indicated by the red-dashed circle. The middle
of the circle indicates the position of the tunnel.
The surface wave data in Fig. 7 also show an anomaly in the shear-
wave velocity prole that is coincident with the known tunnel location.
The location of the void is at station number 1023 in this prole
(marked by the red-dashed circle). The MASW plot shows an increase
in velocity above the void, overlying a zone of decreased velocity at
velocity haloes have been observed above voids in other studies and
are interpreted to be caused by zones of increased stress due to the
removal of material and subsequent increased load borne by the side-
walls and roof (Sloan et al., 2011), since VS is directly related to stress.
Refraction tomography data presented here were acquired using
0.25-m receiver spacing and 0.5-m source spacing; however, this
high-density data acquisition is unrealistic outside of a research
environment, especially when covering large areas. To determine the
feasibility of using this type of method in a production mode using
larger spacings, the P-wave data set was decimated to simulate 0.5,
1.0, and 2.0-m spacings. Fig. 8ae shows velocity plots for 0.5-m receiver
and 0.5-m source spacing, 1.0-m and 1.0-m, 2.0-m and 1.0-m, and 2.0-m
and 2.0-m receiver and source spacing, respectively. Fig. 8fj shows the
corresponding ray coverage plots. The velocity plots do not noticeably
change as the receiver and source spacings are altered; however, the
ray coverage plots exhibit a noticeable degradation as the number of
data points is reduced. Based on a qualitative comparison of the plots
displayed, the optimum receiver/source spacings with respect to both
subsurface coverage and labor required to acquire the data would be
1.0-m receiver and 1.0-m source spacing, which is also the same as

Fig. 8. Plots showing the change in velocity (left) and ray coverage (right) for decimated source/receiver spacing combinations of 0.5-m/0.25-m (a, f), 0.5-m/0.5-m (b, g), 1.0-m/1.0-m
(c, h), 1.0-m/2.0-m (d, i), and 2.0-m/2.0-m (e, j). The middle of the circle indicates the position of the tunnel.
 S.D. Sloan et al. / Journal of Applied Geophysics 99 (2013) 6065
the parameters used to acquire the MASW data. In that regard, a single
data set collected with these parameters could be used for both refrac-
tion tomography and MASW.
In this study the MASW and raypath coverage plots were the most
indicative of the tunnel location. The refraction velocity plots show
subtle changes in velocity that can be attributed to the void; however,
these small variations would not serve as a suitable indicator of a
subsurface anomaly independently without previous knowledge of
the location and depth of the tunnel and are not anomalous with respect
to geology related velocity variations. From a practical application
standpoint, neither of these methods alone would be the optimal choice
to locate a small-diameter tunnel due to the high number of anomalies
that would likely be observed because of the small change in velocity
from unconsolidated sediments to an air-lled cavity. Ideally multiple
seismic and/or other geophysical methods would be applied and as
more methods are combined to rene the interpretation, more positive
detections would accumulate at a common anomaly to further dene its
location. In this case, the true-positive percentage would likely increase,
along with the reduction of false positives, by using multiple techniques
for redundancy and cross-validation, which also increases the con-
dence in results.
5. Conclusions
In conclusion, voids in the shallow subsurface, including tunnels, are
challenging to successfully image due to the non-uniqueness inherent
to all geophysical methods and limited resolution. In both of the seismic
refraction tomography surveys described here, we attempted to detect a
man-made tunnel measuring 1.25-m wide, 1.25-m tall, and 3-m deep.
The refraction tomography data show a drop in a velocity of 2223%
and a noticeable decrease in ray coverage at the tunnel location,
which is expected since the void is air lled and has a lower seismic
velocity compared to the surrounding medium. MASW results from
the same site also show a low-velocity zone at the tunnel location. The
results from these data are in good agreement with what we would
expect based on theory.
In most instances it is a given that the applications described here,
including shallow void or tunnel detection, would not afford the luxury
of having pre-void data sets to compare to. However, for the purposes of
this studydetermining the feasibility of using refraction traveltime
tomography and MASW methods to detect a shallow tunnel in uncon-
solidated sedimentsusing pre- and post-construction data sets further
demonstrates the non-uniqueness of the results and the limitations of
the methods when the number of potential anomalies is expected to
be small. In this case, only subtle variations are observed and are no
more anomalous than other changes in velocity related to subsurface
heterogeneity. Future work may include time-lapse surveys to observe
property variations with time, cross-hole tomography, and applying
65
the same techniques to deeper targets to determine depth limitations
in similar environments.
Acknowledgments
The authors wish to thank Ron Elliston, Kenny Mclaughlin, and Ryan
Strange for assistance in the eld and the reviewers for their construc-
tive edits and suggestions.
References
Ballard, R.F., 1982. Tunnel detection. Technical Report GL-82-9. U.S. Army Engineer
Waterways Experiment Station, pp. 194.
Belfer, I., Bruner, I., Keydar, S., Kravtsov, A., Landa, E., 1998. Detection of shallow objects
using refracted and diffracted seismic waves. J. Appl. Geophys. 38, 155168.
Branham, K.L., Steeples, D.W., 1988. Cavity detection using high-resolution seismic reec-
tion methods. Min. Eng. 40, 115119.
Cho, S., Kim, J., Kim, C., Sung, N., 2006. Tunnel detection using borehole radar survey. SEGJ
Annu. Meeting (Abstract).
Choi, H.K., Ra, J.W., 1999. Detection and identication of a tunnel by iterative inversion
from cross-borehole CW measurements. Microw. Opt. Technol. Lett. 21, 458465.
Dobecki, T.L., 1988. A rapid seismic technique for detecting subsurface voids and
unmapped mine workings. SAGEEP Proceedings, pp. 666690.
Greeneld, R.J., Cameron, C.P., MacLean, H.D., Moran, M.L., 1991. Interpretation of cross-
borehole radar signals propagating along a tunnel containing a wire. SEG International
Meeting, pp. 411413 (Abstract).
Hickey, C.J., Schmitt, D.R., Sabatier, J.M., Riddle, G., 2009. Seismic measurements for detect-
ing underground high-contrast voids. SAGEEP Conference Proceedings, pp. 929936.
Inazaki, T., Kawamura, S., Tazawa, O., Yamanaka, Y., Kano, N., 2005. Near-surface cavity
detection by high-resolution seismic reection methods using short-spacing type
land streamer. SAGEEP Proceedings, pp. 959970.
Llopis, J.L., Dunbar, J.B., Wakeley, L.D., Corcoran, M.K., 2005. Tunnel detection along the
southwest U.S. border. SAGEEP Proceedings, pp. 430443.
Mahrer, K.D., List, D.F., 1995. Radio frequency electromagnetic tunnel detection and delin-
eation at the Otey Mesa site. Geophysics 60, 413422.
Park, C.B., Miller, R.D., Xia, J., 1999. Multichannel analysis of surface waves. Geophysics 64,
800808.
Peterie, S.L., Miller, R.D., Steeples, D.W., 2009. Diffraction imaging versus reection pro-
cessing for shallow void detection. SEG International Annu. Meeting, pp. 14211424
(Abstract).
Rechtien, R.D., Greeneld, R.J., Ballard, R.F., 1995. Tunnel signature prediction for a cross-
borehole seismic survey. Geophysics 60, 7686.
Riddle, G.I., Hickey, C.J., Schmitt, D.R., 2010. Subsurface tunnel detection using electrical
resistivity tomography and seismic refraction tomography: a case study. SAGEEP
Proceedings, pp. 552562.
Schuster, G.T., Quintus-Bosz, A., 1993. Wavepath eikonal traveltime inversion: theory.
Geophysics 58, 13141323.
Sheehan, J.R., Doll, W.E., Mandell, W.A., 2005. An evaluation of methods and available
software for seismic refraction tomography analysis. J. Environ. Eng. Geophys. 10,
2134.
Sloan, S.D., Peterie, S.L., Ivanov, J., Miller, R.D., McKenna, J.R., 2011. Void detection using
near-surface seismic methods. Advances in Near-surface Seismology and Ground-
penetrating Radar. SEG Geophysical Developments Series, 15, pp. 201218.
Tucker, R.E., McKenna, J.R., McKenna, M.H., Mattice, M.S., 2007. Detecting underground
penetration attempts at secure facilities. Engineer 37, 3134.
Walters, S.L., Miller, R.D., Xia, J., 2007. Near surface tunnel detection using diffracted
p-waves: a feasibility study. SEG International Meeting, pp. 11281132 (Abstract).
Walters, S.L., Miller, R.D., Steeples, D.W., Xia, J., Zeng, C., 2009. Detecting tunnels and
underground facilities using diffracted P-waves. SAGEEP Proceedings, pp. 937942.
Xia, J., Miller, R.D., Park, C.B., 1999. Estimation of near-surface shear-wave velocity by
inversion of Rayleigh waves. Geophysics 64, 691700.