GEOPHYSICS, VOL. 72, NO. 4 共JULY-AUGUST 2007兲; P. F139–F144, 8 FIGS.
10.1190/1.2732994
Comparison of an optimized resistivity array
with dipole-dipole soundings in karst terrain
Jonathan E. Nyquist1, John S. Peake2, and Mary J. S. Roth3
ABSTRACT
We report the results of a field comparison of 2D resistivity
data collected using both the traditional dipole-dipole array
and a new computer-optimized array recently described in
literature. The study was conducted at a karst site in eastern
Pennsylvania. Computer simulations suggested that for a
given line length, the optimal array and the dipole-dipole ar-
ray would be equally effective at imaging shallow targets but
the optimal array would provide better resolution at depth.
Our field test results showed that the two arrays imaged karst
bedrock topography equally well. When the full grid of 2D
lines were combined and analyzed using 3D inversion, how-
ever, the optimal array was able to resolve a crosscutting bed-
rock fracture system that was not visible in the dipole-dipole
data. The existence of the fracture system was confirmed by
drilling. Because the optimal array requires roughly three
times as many measurements per line, we conclude that the
optimal array is preferable to traditional dipole-dipole sound-
ings only when the slight improvement in resolution at depth
is more important than rapid data collection.
INTRODUCTION
Most resistivity arrays used today were developed before micro-
processor-controlled multielectrode resistivity equipment. When
using a conventional four-electrode resistivity system, most of one’s
time is spent moving electrodes. Consequently, in the past it paid to
use arrays that minimized the number of electrodes moved between
measurements. Although many papers were published debating the
relative merits of various arrays 共e.g., Bhattacharya and Sen, 1981;
Bear and Tripp, 1995兲, the electrode measurement sequences com-
monly used today were adopted as much for their efficiency in the
field as for their ability to image the subsurface.
Presented at the Second International Conference on Environmental and Engineering Geophysics, ICEEG. Manuscript received by the Editor August 30, 2006;
revised manuscript received January 18, 2007; published online May 15, 2007.
1
Temple University, Philadelphia. E-mail: nyq@temple.edu.
2
ARM Geophysics, Hershey, Pennsylvania. E-mail: jpeake@armgeophysics.net.
3
Lafayette College, Easton, Pennsylvania. E-mail: rothm@lafayette.edu.
© 2007 Society of Exploration Geophysicists. All rights reserved.
F139
The Wenner array, for example, maintains a constant spacing be-
tween electrodes. It is ideal for profiling because only one electrode
is moved between measurements. Similarly, the Schlumberger array
is efficient for depth sounding because only the two outer current
electrodes are moved between measurements. The spacing between
the central potential electrodes is increased only as required to keep
the signal above noise. An additional advantage to using standard ar-
rays 共before the rise of the personal computer兲 was that interpreta-
tion of these data typically involved matching plots against type
curves, which were published only for the standard arrays.
Practice has demonstrated that conventional resistivity arrays are
effective, but there is no proof that they are optimal. Today, the use of
multielectrode resistivity systems with microprocessor-controlled
data acquisition has become routine. Yet, typically, one of the tradi-
tional arrays is still used, such as Wenner, Schlumberger, or dipole-
dipole. One can speculate whether this practice reflects a cautious
“change-one-thing-at-a-time” attitude on the part of equipment de-
velopers. Or perhaps it is the natural tendency of the practicing geo-
physicist, constrained by budget and schedule, to continue using
standard arrays that have worked reliably in the past. With a modern
resistivity system, however, there is no need to move electrodes be-
tween measurements. With inversion software, type curves are no
longer required for interpretation. So there is no compelling reason
to use arrays that minimize the number of electrodes that must be
moved between measurements.
In recent research, Stummer et al. 共2004兲 use computer simulation
to search for an optimal electrode measurement sequence. They ac-
knowledge that even with modern equipment, it takes too long to ac-
quire data for all possible electrode combinations. So they seek a
trade-off between resolution and measurement time. Figure 1 shows
resolution as a function of the number of measurement combina-
tions. At first, the improvement is dramatic; eventually, however, the
curve begins to level off, a result that signifies a diminishing return
with additional measurements. Note that the word “measurement” in
this context refers to the voltage reading acquired using one pair of
F140
potential electrodes for excitation by one pair of current electrodes.
In practice, with multichannel resistivity systems, it is possible to
make many simultaneous voltage readings for each current injec-
tion, thus decreasing data acquisition time.
In the appendix of their paper, Stummer et al. 共2004兲 show a sam-
ple measurement sequence for a 30-electrode array. They propose
this array sample as a compromise between acquisition time and res-
olution 共which we subsequently call the optimal array兲. Their paper
includes a simple field example, but their conclusions are based
largely on computer simulations.
The principal difference between the Stummer et al. 共2004兲 opti-
mal array and conventional resistivity arrays is the addition of mea-
surements that use widely spaced current electrodes or widely
spaced potential electrodes. This technique is not just a matter of in-
creasing the maximum dipole spacing. The optimal array includes
unusual combinations, including measurements in which the current
electrode pair overlaps the potential electrode pair 共both with separa-
tions more than half the total array length兲. Figure 2 compares the
pseudodepths for a standard dipole-dipole array and the optimal ar-
ray for a 28-electrode sounding. There are several different ways to
plot pseudodepths; however, because the optimal array is nonstand-
100
80
Relative model resolution (%)
60
40
20
0 defined as
0 2000 4000 6000
Number of measurements
Figure 1. Computer simulation results show that the resolution
decreases only slightly at first 共as the number of measurements is re-
duced兲, but then resolution falls off rapidly. The arrow shows the se-
lected compromise between resolution and data collection time 共af-
ter Stummer et al., 2004兲.
a) 0
Depth (m)
10
0 10 20 30 40
Distance (m)
b) 0 model inversion algorithm, which is based on the work of Constable
Depth (m)
10
0 10 20 30 40
Distance (m)
Figure 2. 共a兲 Pseudodepths for the traditional dipole-dipole array. 共b兲
Pseudodepths for the optimal array. The two arrays are identical near
the surface, but the optimal array includes many more widely spaced
electrode pairs, a technique resulting in increased coverage at depth.
Nyquist et al.
ard, we elected to place the pseudodepth at one-fifth the maximum
electrode separation. Although the dipole-dipole array and optimal
array are identical at shallow depths, the optimal array includes
many more measurements that correspond to deep pseudodepths.
Based on the pseudodepths 共Figure 2兲, we hypothesized that the
optimal array would offer significantly better resolution at depth
than conventional arrays. Our tests of this hypothesis consisted of
comparing resistivity results for dipole-dipole and optimal array
soundings with drilling logs. We also used information accumulated
from previous geophysical investigations as well as depth of investi-
gation 共DOI兲 simulations.
SIMULATIONS — DEPTH OF INVESTIGATION
Many methods have been proposed to determine the DOI of DC
resistivity arrays. Oldenburg and Li 共1999兲 propose a technique for
estimating DOI that takes advantage of modern, commercially avail-
able resistivity inversion software. The idea is simple: Resistivity in-
version typically starts with an initial guess for the subsurface resis-
tivity structure 共typically, a homogeneous earth兲. Resistivity values
of the model blocks are then adjusted iteratively to improve the fit to
the measured data. For model blocks that are well constrained by the
measurements, the solution will converge to the same final resistivi-
ty value 共regardless of the starting value兲. For regions that are poorly
constrained by the data, the resistivity inversion results will remain
virtually unchanged from the starting value. The depth of investiga-
tion can be estimated by inverting the data using different starting
models and then comparing the results.
Marescot and Loke 共2004兲 describe a simple implementation of
this algorithm. First, an inversion is performed using a starting mod-
el of a homogeneous earth, using a reference resistivity qA, typically
the average of the logarithm of the measured apparent resistivities. A
second inversion is then performed using a different reference resis-
tivity qB, typically 10 times larger than qA. The inversion results
共from the two different starting models兲 are used to calculate the nor-
malized difference between the resistivities RAB for each model cell,
8000 10,000
qA共x,z兲 − qB共x,z兲
RAB共x,z兲 = 共1兲
qA − qB
Where the model is well constrained by the data, RAB will be close to
zero. But where there is poor data coverage, RAB will approach one.
We conducted our DOI analysis using the survey planner option
of AGI Geoscience’s EarthImager® 2D. This option allows the user
to simulate field surveys over hypothetical earth models. EarthIm-
ager 2D discretizes the subsurface model into either a finite-differ-
ence or a finite-element grid. Then it solves the partial differential
equation for a 3D point current source over a 2D earth, the so-called
50 60 70 80 2.5D problem 共Yang and Lagmanson, 2003兲. For the nonlinear in-
version of the simulated data, we used EarthImager 2D’s smooth
et al. 共1987兲.
We investigated a variety of geologic scenarios and found that, in
general, the simulations predicted a greater DOI for the optimal ar-
50 60 70 80
ray. Figure 3 shows sample DOI results for dipole-dipole and opti-
mal arrays. The first row in Figure 3 shows the EarthImager 2D fi-
nite-element model, which was identical for both array types. The
simulations are for an array of 28 electrodes spaced 3 m apart for a
total line length of 81 m. We used a 0.5-m uniform mesh that corre-
sponds to six elements between electrodes at the surface. The geo-
 Optimized versus dipole-dipole array F141

logic scenario modeled here represents a narrow 共10 m兲 limestone
ridge with a thin mantle 共0.5 m兲 of conductive clay soil, flanked by
10-m-deep conductive clay valleys. This scenario is appropriate to
our field site, as we discuss in the next section. The next two rows of
Figure 3 show the inversion results for the two arrays, starting with a
uniform resistivity of 50 ohm-m 共Figure 3, row 2兲 and a uniform re-
sistivity of 500 ohm-m 共Figure 3, row 3兲. The bottom row of Figure
3 shows the DOI results calculated using equation 1.
There are a number of interesting features in the DOI results. First,
note that the resistivity results for model blocks right at the surface
are poorly resolved. In practice, the apparent-resistivity pseudosec-
tion is commonly used as the starting model. This practice over-
comes the resolution problem because apparent resistivities for the
smallest electrode spacings are a good starting guess for the true re-
sistivities of the top layer. Second, note the large difference in the in-
version results for the elements at the boundary between the conduc-
tive soil layer and the top of the resistive limestone ridge. This zone
is seen for both array types, and it persisted when we modeled a
thicker soil layer over the ridge. But it was not seen in a simple lay-
ered-earth model. Marescot and Loke 共2004兲 point out that the DOI
analysis can also be thought of as a region of investigation analysis,
not just a depth of investigation. Apparently, the transition zone be-
tween the top of the limestone ridge and the soil layer is poorly re-
solved, with the optimal array actually performing slightly worse.
As expected, the benefit of the optimal array appears in the deeper
portions of the survey. For a starting guess of 50 ohm-m, the inver-
sion of the dipole-dipole data does not image the deeper bedrock.
However, if the starting guess is 500 ohm-m, bedrock is detected.
The DOI analysis shows that this array has little coverage at the edg-
es of the array. Analysis shows that coverage falls off starting at a
depth of about 7 or 8 m. In contrast, the DOI result for an optimal ar-
ray shows good data coverage right to the base of the section at 15 m,
and the bedrock contact at 10 m is detected independent of the start-
ing model.
We also modeled the case of a homogeneous, 1000-ohm-m half-
space and a 1-m-thick, 10-ohm-m layer over a 100-ohm-m half-
space. DOI simulation for the dipole-dipole array showed reliable
results up to a depth of 11 m 共14% of the line length兲 for the homoge-
neous half-space. However, a conductive soil overburden just 1 m
thick rendered suspect the inverted resistivity values below 7 m
共only 9% of the line length兲. Although simulations for the optimal ar-
ray showed a similar pattern, the optimal array had reliable results
extending to a depth of 17 m 共21% of the line length兲 for a homoge-
neous half-space and extending to 13 m 共16% of the line length兲 for
the case of a 1-m-thick soil overburden.
These DOI estimates are likely to be overly optimistic because
they do not account for noise in the data. However, in all cases, the
simulations predicted that an optimal array would provide signifi-
cantly better resolution at depth. The optimal array has a practical
drawback, however, because it requires nearly three times as many
measurements. We conducted field tests to determine whether, in
practice, the improvement in resolution is sufficient to justify the ad-
ditional data acquisition time.

Synthetic resistivity model Synthetic resistivity model

0 18 27 36 45 54 63 72 81 Ohm-m 0 18 27 36 45 54 63 72 81 Ohm-m
0 0
Depth (m)
Depth (m)

5000 5000
5 5

10 100 10 100
15 15

Log 10 resistivity Log 10 resistivity

Starting model: Homogeneous 50 ohm-m (ohm-m) Starting model: Homogeneous 50 ohm-m (ohm-m)
0 4 0 4
Depth (m)

Depth (m)

–5 –5
3 3
–10 –10
2 2
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Position (m) Position (m)

Log 10 resistivity Log 10 resistivity

(ohm-m) Starting model: Homogeneous 500 ohm-m (ohm-m)
Starting model: Homogeneous 500 ohm-m
0 4 0 4
Depth (m)
Depth (m)

–5 –5
3 3
–10 –10
2 2
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Position (m) Position (m)

Percent Percent
difference Optimal array depth of investigation difference
Dipole-dipole array depth of investigation
0 100 0 100
Depth (m)

Depth (m)

–5 50 –5 50
–10 –10
0 0
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Position (m) Position (m)

Figure 3. Example DOI simulation results for a dipole-dipole array 共left column兲 and an optimal array 共right column兲. The top row shows the fi-
nite-element model 共0.5-m elements兲 for the subsurface — identical for both arrays — simulating a resistive limestone ridge with a thin mantle
of clay soil flanked by 10-m-deep clay valleys. The second and third rows show the inversion results for the different starting models. The bottom
row compares the DOI results for the two arrays.
F142 Nyquist et al.

FIELD INVESTIGATION
Metzgar field data collection
Lafayette College and Temple University have an ongoing coop-
erative research program to evaluate the reliability of electrical resis-
tivity surveys as a geotechnical site characterization method in karst.
To date, we have used 2D and 3D multielectrode arrays to investi-
gate bedrock depths and to locate karst solution features 共Jenkins
and Nyquist, 1999; Mackey et al., 1999; Maule et al., 2000; Roth et
al., 2000; Roth et al., 2002; Nyquist and Roth, 2003; Roth and Ny-
quist, 2003; Roth et al., 2004; Nyquist and Roth, 2005; Nyquist et al.,
2005; Peake, 2005兲.
The research area is Metzgar Field, an athletic complex owned by
Lafayette College. The geology is characterized by a thin mantle of
clay soils underlain by limestone 共Epler Formation, Lower Ordovi-
cian兲. On average, 15 new sinkholes open each year in the complex
Soil
Bedrock
Figure 4. Conceptual model of the bedrock topography beneath
Metzgar Field for a cross section perpendicular to strike. The bed-
rock ridges and soil extend parallel to strike.
共Thomas and Roth, 1999兲. Although the sinkholes that develop in
this part of Pennsylvania are typically small 共only 1–2 m in diame-
ter兲, they are a nuisance for groundskeepers and a serious problem
for local developers.
Our study area is an approximately 1-ha portion in the northwest
corner of the site. Geophysical testing and drilling show that the bed-
rock in the test area has multiple high-resistivity limestone ridges
共roughly 3–5 m wide兲 that run parallel to geologic strike 共N 71° E兲,
the result of karst weathering of the dipping bedrock 共45° southeast兲
共Figure 4兲. Bedrock depths range from less than 1 m over the ridges
to more than 10 m between the ridges. The bedrock is also known to
have soil-filled and open fractures as well as larger voids. Because of
the large contrast in resistivity between the limestone bedrock and
the overlying soil, DC resistivity is a reasonable choice for charac-
terizing bedrock topography. Thus, one point of comparison be-
tween the dipole-dipole and optimal arrays is a method’s relative
ability to determine depth to bedrock.
A second, more difficult task, is locating cavities in the limestone
bedrock. The cavities of interest are in the unsaturated zone. Locat-
ing an air-filled cavity in limestone involves the difficult task of
sounding through conductive clay overburden to locate an ul-
traresistive target within a resistive host rock. This difficulty ex-
plains our interest in finding a new array that promises improved res-
olution at depth.
We collected resistivity data using both the dipole-dipole and op-
timal arrays along 56 lines: 28 lines perpendicular to strike and 28
lines parallel to strike. We used an Advanced Geosciences Super-
Sting eight-channel resistivity system with a 28-electrode cable
共Figure 5兲, with 3 m spacing between the electrodes along the lines
as well as between the lines. We then inverted the data using Earth-
Imager 2D.
Metzgar field results and discussion
At our field site, the dipole-dipole and optimal arrays were equally
effective in mapping the karst bedrock topography. Figure 6 com-
pares the two resistivity array results for NS-24, a line perpendicular
to strike. Superimposed on the figure are the bedrock depths ob-
tained by augering holes down to auger refusal along the line. There
is reasonable agreement between the resistivity data and the auger
results but little to distinguish between the two resistivity surveys. In

Log 10 resistivity
a) 0 SE NW
(ohm-m)
4.5
–2
Depth (m)

4
–4
–6 3.5
–8 3
–10 2.5
–12 2
0 10 20 30 40 50 60 70 80
Position (m)
Log 10 resistivity
b) 0 SE NW
(ohm-m)
4.5
Depth (m)

4
–5 3.5
3
–10 2.5
0 10 20 30 40 50 60 70 80 2
Position (m)

Figure 6. A comparison, with augering results, of the resistivity data

Figure 5. A resistivity sounding being performed at Metzgar Field.
The flat, groomed surface belies the rugged underlying bedrock to-
pography.
for line NS-24 for the 共a兲 dipole-dipole survey and 共b兲 the optimal ar-
ray. The dashed horizontal line represents the approximate DOI for
the dipole-dipole array 共refer to Figure 3兲. The DOI for the optimal
array extends to the bottom of Figure 6. The vertical black lines show
the depths to auger refusal, which in most cases in our research cor-
respond to a resistivity of roughly 1000 ohm-m on the inverted resis-
tivity section.
 Optimized versus dipole-dipole array F143

both cases, the soil-bedrock interface roughly corresponds to the in the dipole-dipole resistivity results. The optimal resistivity data
1000 ohm-m contour. Karst bedrock, however, can be highly pin- show a possible right-lateral strike-slip fault, based on the offset be-
nacled and pitted. The few points where the augering and resistivity tween the resistivities on either side. There are also many conductive
data disagree dramatically 共for example, 45 and 58 m兲 are probably anomalies within the ridges seen in the resistivity data collected
narrow, soil-filled dissolution features — bedrock weathering on a along the fault line, a result consistent with weathering along a zone
scale too fine to be resolved. It would be difficult to say that the of weakness.
match between resistivity and bedrock topography is superior for the Nine borings in this zone encountered segments of void space,
optimal array. eight on the ridges. Only one other boring encountered a void 共out of
Figure 7 compares the results for another line 共NS-14兲 collected
perpendicular to strike. This line is of interest because it passes di- a) NS-14 NS-24
80
rectly over a cavity intercepted by drilling at a depth of 7 m. The
cavity extends roughly 4 m along strike and 2 m perpendicular to R1
70
strike; it is 2 m high at its highest point 共Roth et al., 2004兲. Surpris-
ingly, although the cavity is air filled, it appears in the resistivity data V1

Perpendicular to strike (m)

60
as a conductive break in the resistive limestone ridge. This conduc-
tive break is even more apparent in the resistivity line parallel to R2
50
strike 共not shown兲, which is the direction of maximum cavity length.
We speculate that the cavity appears as a conductor because of
40
perched water held by capillary pressure in fractures above the roof
of the cavity 共Manney et al., 2005兲. Again, the dipole-dipole array R3
30
and the optimal array produced similar results, although the cavity is
more clearly within the conductive zone in the results from the opti-
20
mal array. V2
Although we did not make direct measurements to confirm the
10
presence of perched water, there is indirect evidence of fractures in
the bedrock above the cavity — fractures that could act as a capillary
0
barrier to unsaturated flow. We drilled three holes into the cavity to 0 20 40 60 80
allow us to use laser triangulation to determine its geometry 共Ny- Parallel to strike (m)
quist and Roth, 2003; Roth et al., 2004兲. After drilling the first bore-
hole, we used a downhole camera to watch the drilling of the two
subsequent boreholes from inside the cavity. In both cases, the water b) Log 10 resistivity
(ohm-m)
that circulated during drilling rained down from the roof of the cavi- NS-14 NS-24
80 4
ty when the drill bit was still several meters away from breeching the
cavity. Clearly, the rock above the cavity is fractured and capable of
70
storing water in unsaturated conditions.
The optimal array did produce results superior to the dipole-di-
Perpendicular to strike (m)

60 3.5
pole array when we combined all of the 2D lines and inverted the
data using EarthImager 3D software, which generalizes the smooth
50
model inversion approach 共used in EarthImager 2D兲 to three dimen-
sions. In both cases, a horizontal slice taken through the 3D results at
40 3
a depth of 6 m 共Figure 8兲 clearly shows the karst ridge and valley
bedrock topography. The results of inverting the optimal array data,
30
however, also show what appears to be a crosscutting fracture or dis-
rupted zone perpendicular to strike. This area is not readily apparent
20 2.5

Log10 resistivity
a) SE NW
(ohm-m) 10
0 4.5
Depth (m)

4
–5 3.5 0
3 2
–10 Cavity 2.5 0 20 40 60 80
2 Parallel to strike (m)
0 10 20 30 40 50 60 70 80
Position (m)
Log10 resistivity
b) (ohm-m) Figure 8. 共a兲 Horizontal slice 共plan view兲 at 6 m below the ground
SE NW
0 4.5 surface through the 3D inversion results for the combined 56 2D di-
Depth (m)

4 pole-dipole resistivity lines. V1 and V2 indicate soil valleys; R1, R2,

–5 3.5
3 and R3 indicate bedrock ridges. Note the locations of lines NS-24
–10 Cavity 2.5
2
共refer to Figure 6兲 and NS-14 共refer to Figure 7兲. The small black el-
0 10 20 30 40 50 60 70 80 lipse points out the location and extent of the cavity. The x-axis is in
Position (m) the direction of geologic strike. The resistivity color scale is the same
in the top image as in the lower image. 共b兲 The same depth slice
Figure 7. Comparison of the 共a兲 dipole-dipole resistivity results with through the 3D inversion results for the combined 56 2D optimal ar-
共b兲 the optimal array results for a line perpendicular to strike. Shown ray resistivity lines. Drilling later confirmed a fault running perpen-
in black are the location and dimensions of an air-filled cavity. dicular to strike 共marked with the large black ellipse兲.
F144 Nyquist et al.
the more than 30 borings drilled into bedrock within the survey
area兲. From the DOI analysis, we know that a 6-m depth slice is near
the penetration limit for the dipole-dipole with the geometry used. It
appears that the greater penetration of the optimal array was benefi-
cial in this case.
CONCLUSIONS
Although data collected using the optimal array did help us to lo-
cate a fault zone at depth that was not apparent in the dipole-dipole
data, the overall results for the two arrays were quite similar. We con-
clude that despite the predictions of the DOI simulations, for most
applications, the traditional dipole-dipole array provides resolution
comparable to the optimal array of Stummer et al. 共2004兲. The di-
pole-dipole array is preferred because the data acquisition is nearly
three times faster. We speculate that the reason the optimal array is
only slightly better is that the resolution of the resistivity method is
inherently poor for large electrode separations, a limitation that can-
not be overcome by additional electrode pair combinations 共the mea-
surements tend to blur together at depth兲. The additional time and ex-
pense associated with optimal array might be justified under excep-
tional circumstances where the target of interest is at the limit of the
depth of investigation or where limited access precludes using a
longer array.
ACKNOWLEDGMENTS
This work was part of ongoing research sponsored by the National
Science Foundation under collaborative grants CMS-0125601 and
CMS-0201015. This support is gratefully acknowledged.
REFERENCES
Beard, L. P., and A. C. Tripp, 1995, Investigating the resolution of IP arrays
using inverse theory: Geophysics, 60, 1326–1341.
Bhattacharya, B. B., and M. K. Sen, 1981, Depth of investigation of collinear
electrode arrays over homogeneous anisotropic half-space in direct cur-
rent methods: Geophysics, 46, 768–780.
Constable, S., R. L. Parker, and C. G. Constable, 1987, Occam’s inversion: A
practical algorithm for generating smooth models from electromagnetic
sounding data: Geophysics, 52, 289–300.
Jenkins, S. A., and J. E. Nyquist, 1999, An investigation into the factors caus-
ing sinkhole development at a site in Northampton County, Pennsylvania:
Proceedings of the Seventh Multidisciplinary Conference on Sinkholes
and the Engineering and Environmental Impacts of Karst, 45–49.
Mackey, J. R., M. J. S. Roth, and J. E. Nyquist, 1999, Case study: Site charac-
terization methods in karst, in G. Fernandez and R. Bauer, eds., Geo-engi-
neering for underground facilities: ASCE Geotechnical Special Publica-
tion, 90, 695–705.
Manney, R., M. J. S. Roth, and J. E. Nyquist, 2005, Exploring directional dif-
ferences in resistivity results in karst: Proceedings of the Symposium for
the Application of Geophysics to Environmental and Engineering Prob-
lems, Environmental and Engineering Geophysical Society, 1117–1124.
Marescot, L., and M. H. Loke, 2004, Using the depth of investigation index
method in 2D resistivity imaging for civil engineering surveys: Proceed-
ings of the Symposium for the Application of Geophysics to Environmen-
tal and Engineering Problems, Environmental and Engineering Geophysi-
cal Society, 589–595.
Maule, J., J. E. Nyquist, and M. J. S. Roth, 2000, A comparison of 2D and 3D
resistivity soundings in shallow karst terrain, Easton, PA: Proceedings of
the Symposium for the Application of Geophysics to Environmental and
Engineering Problems, Environmental and Engineering Geophysical So-
ciety, 969–977.
Nyquist, J. E., and M. J. S. Roth, 2003, Application of a downhole search and
rescue camera to karst cavity exploration: Proceedings of the Symposium
for the Application of Geophysics to Environmental and Engineering
Problems, Environmental and Engineering Geophysical Society, 841–
848.
——–, 2005, Improved 3D pole-dipole resistivity surveys using radial mea-
surement pairs: Geophysical Research Letters, 32, L21504.
Nyquist, J. E., M. J. S. Roth, S. Henning, R. Manney, and J. Peake, 2005,
Smoke without mirrors: A new tool for the geophysical characterization of
shallow karst cavities: Proceedings of the Symposium for the Application
of Geophysics to Environmental and Engineering Problems, 337–343.
Oldenburg, D. W., and Y. Li, 1999, Estimating depth of investigation in dc re-
sistivity and IP surveys: Geophysics, 64, 403–416.
Peake, J., 2005, A comparative analysis of the detection of subsurface cavi-
ties in karst systems using two dimensional and three dimensional electri-
cal resistivity: M.S. thesis, Temple University.
Roth, M. J. S., J. R. Mackey, C. Mackey, and J. E. Nyquist, 2002, Acase study
of the reliability of multielectrode earth resistivity testing for geotechnical
investigations in karst terrains: Engineering Geology, 65, 225–232.
Roth, M. J. S., and J. E. Nyquist, 2003, Evaluation of multi-electrode earth re-
sistivity testing in karst: ASTM Geotechnical Testing Journal, 26,
167–178.
Roth, M. J. S., J. E. Nyquist, A. Faroni, S. Henning, R. Manny, and J. Peake,
2004, Measuring cave dimensions remotely using laser pointers and a
downhole camera: Proceedings of the Symposium for the Application of
Geophysics to Environmental and Engineering Problems, Environmental
and Engineering Geophysical Society, 1307–1314.
Roth, M. J. S., J. E. Nyquist, and B. Guzas, 2000, Locating subsurface voids
in karst: A comparison of multi-electrode earth resistivity testing and grav-
ity testing; Proceedings of the Symposium for theApplication of Geophys-
ics to Environmental and Engineering Problems, Environmental and Engi-
neering Geophysical Society, 359–365.
Stummer, P., H. Maurer, and A. G. Green, 2004, Experimental design: Elec-
trical resistivity data sets that provide optimum subsurface information:
Geophysics, 69, 120–139.
Thomas, B., and M. J. S. Roth, 1999, Evaluations of site characterization
methods for sinkholes in Pennsylvania and New Jersey: Engineering Ge-
ology, 52, 147–152.
Yang, X., and M. B. Lagmanson, 2003, Planning resistivity surveys using nu-
merical simulations: Proceedings of the Symposium for the Application of
Geophysics to Environmental and Engineering Problems, Environmental
and Engineering Geophysical Society, 488–501.