Professional Documents
Culture Documents
Timothy J. McGraw
A Thesis
MASTER OF SCIENCE
August 2010
Committee:
ABSTRACT
Karst, near surface bedrock cavity, fracture, or cave (solution) features, are present near
the Crystal Rock Cave system in Erie County, north-central Ohio. Previously, a capacitively-
coupled electrical resistivity study of the area was completed to evaluate the methodology for
detecting karst features in the subsurface. Resistivity traverses were completed over known
subsurface features using different transmitter-receiver spacing, varying the penetration depth of
the resistivity survey. The study found that the larger electrode spacing distances and thin, low-
clay soils at the Crystal Rock Cave site provided data that clearly identify known sinkhole and
cavity features. However, the durability and maneuverability of the towed resistivity array posed
To evaluate the findings of the previous resistivity study and assess the ability of ground-
penetrating radar (GPR) to identify the same features, a GPR survey was also completed along
the same resistivity transects over the Crystal Rock Cave system. The use of GPR is a well-
established method of not only detecting subsurface cavities but also mapping the subsurface.
The cart-mounted, 400-MHz antenna, GSSI GPR system identified the karst features by
detecting the adequately contrasting dielectric properties of the soils, bedrock, and associated
air-filled cavities within the shallow subsurface. The shallow (5 m) transects resulted in higher
resolution datasets and profiles that clearly show more detail. Deeper (10 m) transects required
significant processing and manipulation in order to enhance deeper reflections and to retain
solution features were discernable and comparable within the subsurface profiles generated by
the two methods at numerous locations along transects. Comparison of resistivity pseudosections
and GPR profiles during interpretation aided in identifying solution features. The resistivity data
along a traverse are affected by features off line, but the resolution and the three-dimensional
nature of the recorded GPR signal often allows multiple features (i.e., soil-filled cavity
surrounded by bedrock) to be observed within the same space along a transect. In terms of data
collection speed and ease, the GPR proved to be better suited than resistivity. However, with the
valuable subsurface data both methods provide, they are best utilized in conjunction as
companion methodologies.
iv
ACKNOWLEDGMENTS
I would like to acknowledge my thesis Advisor, Dr. Jeffrey Snyder for his enduring
effort, support, and guidance; and my other thesis committee members, Dr. Charles Onasch and
Dr. Sheila Roberts for their guidance and support. Recognition is also due to Bowling Green
State University, the Graduate College, the Geology Department, and associated staff for
providing a venue conducive of academic study and continued support. I would also like to
acknowledge the consulting firm Atwell, LLC and staff for the use of the GPR equipment and
software, for without, this thesis would not have been possible. Finally, I want to acknowledge
my wife and son for their patience and support throughout my years of graduate study.
v
TABLE OF CONTENTS
Page
INTRODUCTION ............................................................................................................ 1
Objectives ............................................................................................................ 3
Background ............................................................................................................ 3
RESULTS ............................................................................................................ 27
Interpretation ................................................................................................. 31
Transect 1a ............................................................................................................ 32
Transect 2a ............................................................................................................ 34
Transect 3a ............................................................................................................ 36
Transect 4a ............................................................................................................ 38
Transect 5a ............................................................................................................ 40
Transect 6 ............................................................................................................ 42
Transect 7 ............................................................................................................ 47
Transect 8 ............................................................................................................ 49
vi
Transect 9 ............................................................................................................ 50
Errors ............................................................................................................ 52
CONCLUSIONS ............................................................................................................ 54
REFERENCES ...................................................................................................................... 56
LIST OF FIGURES
Figure Page
5 Bedrock geology of the site region, modified from the ODNR ................................. 17
LIST OF TABLES
Table Page
1 Relative dielectric constants (εr) and radio wave velocities (V) for a range of geological
4 Soil types present on the site and pertinent characteristics related to GPR suitability 21
1
INTRODUCTION
Ground penetrating radar (GPR) has been a popular tool for identifying and defining
subsurface geological features since the mid-1980s (Reynolds, 1997). Since its initial
development during the mid-1920s, this method of transmitting pulsed electromagnetic (EM)
radiation into the ground and recording the reflected signal has been utilized for a number of
subsurface exploration applications. GPR has also proven to be a useful method in appropriate
the shallow subsurface within engineering and environmental applications (Benson, 1995). These
sorts of applications often directly affect the population that occupies the land and wishes to
understand and utilize shallow subsurface environments. With the rise in population predicted
during the 21st century, a more detailed understanding of the shallow subsurface will be required
if humans desire to manage Earth’s limited resources (Neal, 2004), and utilize this knowledge
Although a versatile and proven method, GPR is only one of several geophysical survey
tools that can be used to identify and define subsurface geological features (Chamberlain et al.,
targets and lists trap structures for oil and gas, mineshafts, pipelines, ore lodes, cavities,
groundwater, and buried rock valleys as specific examples. Subsurface cavities in areas of karst
are often susceptible to ground surface subsidence which can pose a threat to new and existing
development as well as the population that occupies the land (Doolittle and Collins, 1998).
karst will become increasingly important to new and existing development. Other geophysical
techniques (used with varied success) that are suitable for detecting and/or defining karst features
2
al., 2000). However, previously published field trials of GPR suggest it is an effective means of
detecting small (<10 m diameter) caves and fissures in karst terrain (Collins et al., 1994; Benito
et al., 1995; Harris et al., 1995; Chamberlain et al., 2000). With this in mind, and of particular
interest to this study, a suitable application of GPR would appear to include the identification and
characterization of near-surface solution cavities within the karst topography of northern Ohio.
During a previous study, an electrical resistivity survey was performed over a karst
topography known as the Crystal Rock Cave system in Erie County, Ohio (Sabo, 2008). This
geophysical method of analyzing the shallow subsurface. GPR and resistivity are both
considered to be “primary” methods for detecting subsurface cavities that are dependent on the
electrical properties (i.e., permittivity, conductivity, and resistivity) of the substrate (Reynolds
1997). The primary ranking of both methods for detecting subsurface cavities is mainly due their
ability and relative ease of detecting the stark contrast between the air in cavity or voids and the
According to Sabo (2008), while the resistivity method and equipment appeared to work
fairly well at the Crystal Rock Cave site, the durability and maneuverability of the towed
resistivity array contributed to varying data collection speeds. Additionally, the inability to
increase the electrode spacing configuration and thus increase the amount of current induced into
the ground may have limited the signal penetration depth into bedrock. As noted previously,
GPR is known to be well suited in appropriate geological settings for a study of this nature.
Chamberlain et al., (2000) indicates that a GPR with at least a 300 MHz antenna provides
excellent resolution at depths within 5 m of the surface. As configured for this study, the highly
3
maneuverable and durable cart-mounted GPR with a 400 MHz antenna is potentially better
suited to attain consistent data collection speeds as well as sufficient depth of subsurface
penetration. However, the GPR performance is a complicated function of subsurface and target
(i.e., cavity, voids) properties, the system specifications, and application (Plumb et al., 1998).
Objectives
The objective of this study is to utilize a GPR survey system to analyze the solution
cavities and associated near-surface features previously analyzed with electrical resistivity within
the karst topography of the Crystal Rock Cave site in Erie County, Ohio (Site). These data will
be compared to the previous towed resistivity array survey for their abilities to identify and
Background
Karst are mainly caused by groundwater interacting and reshaping bedrock and surface
topography. Some of the first documented studies of karst, occurring between the late 1770s and
early 1790s involved the recognition of collapsed surface depressions which were later
determined to have been caused by the profound effect of groundwater dissolution and transport
of soluble rocks (Easterbrook, 1999). Groundwater is typically slightly acidic and reacts
chemically by dissolving highly soluble rocks such as limestone, dolomite and some evaporates
(i.e., halite) to form closed depressions, caves, collapse features, and underground diversion of
As noted by White (1926) and Verber and Stansbery (1956), the solution cavities and
cave features at the Crystal Rock Cave Site were formed by at least two karst or cavity-forming
processes. At the site, dolomite, halite, and anhydrite/gypsum predominate as rock types (White,
1926; and ODGS, 2004). Karst or cavities in bedrock around the site can be attributed to jointed
or fissured carbonate and sulfate rocks that were prone to dissolution, and the diagenic process of
hydration. One karst or cavity forming process that occurred at the site was fissure or joint-plane
dissolution of the carbonate bedrock that occurred as groundwater and/or rain water (H2O,
potentially enriched with humic acid from decaying vegetation), moved through the fissures or
joints and combined with the carbon dioxide (CO2) in the atmosphere to form carbonic acid
(H2CO3), which dissolved the carbonate bedrock (White, 1926). Subsequent circulation of water
The second karst or cavity-forming process produces the floor and roof of Crystal Rock
Cave to be exact casts of one another that seem to “fit” together (White, 1926). This process
began as the jointed/fissured dolomite or limestone, with inter-bedded anhydrite, gypsum, and
halite comes in contact with the circulating rain or groundwater as previously described. This
primary formation mechanism of the Put-In-Bay style cave is then suggested by Kraus (1905)
and Cottingham (1919) to be caused by the hydration of anhydrite (CaSO4) to gypsum (Ca(SO4)
2H2O). As of result of this hydration, a significant expansive force is generated within the
gypsum layer, increasing the volume by 30 to 60-percent as it replaces anhydrite (White, 1926).
The growth or expansion of the Ca++ to Ca2H2O++ destroys the orthorombic symmetry of
anhydrite (Easterbrook, 1999). According to the ODGS (1999), the swelling lifts the overlying
carbonate strata, opens fractures, and creates massive passageways for groundwater flow through
the bedrock. The circulation of water through the now existing cavities and voids accelerates the
5
solution process. When the gypsum, now expanded to larger volume within the stratified bedrock
layers, is later removed by solution, the displaced rock layers over the resulting void space forms
the typical Put-In-Bay style cave feature. With caves of this type, the overlying carbonate
electromagnetic (EM) energy or wave fields are transmitted into the ground at high frequency
(typically 10-1000 MHz) from a transmitting antenna, and received by an antenna that records
the electromagnetic energy or wave field reflections from the subsurface (Benson, 1995; and van
der Kruk, et al., 1999). When the transmitted electromagnetic energy propagates through the
subsurface and encounters a surface or interface where the electric and/or magnetic properties
change, part of the transmitted energy is reflected back to the receiver while some of the energy
is able to pass through the interface (van der Kruk, et al., 1999). The depth to the object or
interface is then determined by the amount of time it takes for the electromagnetic wave to travel
from the transmitting antenna, through the subsurface to an object or interface, and back to the
receiving antenna. The amount of time recorded is referred to the two-way travel time.
materials (composition and water content), which affect the capacity and speed of the
energy after it is transmitted (Reynolds, 1997). The depth of penetration of the electromagnetic
6
wave and the GPR system is dependent on the frequency of the system’s transmitting antenna.
Higher radar antenna frequencies cannot penetrate the subsurface to the extent lower frequency
antennas are capable. However, with the greater penetration depths achievable with lower
Figure 1. Simplified configuration of a GPR unit and operation. Modified and redrawn from
Reynolds (1997).
According to Reynolds (1997), the velocity of a radio wave in a material (Vm) is reliant
on speed of light in free space (c = 0.3 m/ns), the relative dielectric constant (εr), permittivity of
free space (εo = 8.854 x 10-12 farads (F)/m) and the relative magnetic permeability (µr = 1 for
non-magnetic materials). In common materials, εr and V can range from 1-30 and 33-300,
V
Material εr (mm/ns)
Air 1 300
Water (fresh) 81 33
Sand (dry) 3-6 120-170
Sand (wet) 25-30 55-60
Silt (wet) 10 95
Clay (wet) 8-15 86-110
Clay soil (dry) 3 173
Average 'soil' 16 75
Granite 5-8 106-120
Limestone 7-9 100-113
Dolomite 6.8-8 106-115
Shale (wet) 7 113
Sandstone (wet) 6 112
Quartz 4.3 145
Concrete 6-30 55-112
Asphalt 3-5 134-173
Table 1. Relative dielectric constants (εr) and radio wave velocities (V) for a range of geological
and man-made materials. Some materials omitted, modified from Reynolds, 1997.
Factors that can be attributed to signal loss are reflection/transmission losses in the
substrate, signal scattering losses caused by objects equidimensional with the radar signal
wavelength, absorption (EM energy converted to heat), and geometrical signal spreading during
propagation (Reynolds, 1997). These factors are accounted for by the loss factor (P), where P =
σ/ωε, when σ is the conductivity, ω = 2πf where f = wavelength frequency, ε is the permittivity =
εr εo) and can be expressed as: Vm = c /{( εr µr /2)[(1 + P2) + 1]} 1/2 , or more simply, for the
The attenuation factor (α), which is primary cause of reduced signal energy, is the
cumulative loss of energy due to electric conductivity (σ), magnetic permittivity (µ), the
dielectric properties (ε) of the material through which the signal is introduced. Neal (2004)
indicates that for low-loss materials, α can be expressed as: α = σ/2 μ⁄ . The expression of α
8
indicates that the conductivity of materials exerts the greatest control over α (Theimer et al.,
For EM waves, the depth by which the signal amplitude (A) has decreased to 37% (or
1/e) of the initial value is known as the skin depth (δ) (Reynolds, 1997). As an electromagnetic
wave travels through a material, the A declines exponentially from its initial value (Ao) as it
travels a distance (z) (Neal, 2004). This is expressed by Neal (2004) as: A = Ao e- αz, if α is
constant (such as in low-loss materials where P ≈ 0). Figure 2 illustrates factors that contribute to
Figure 2. Factors that lead to a reduction in GPR signal strength. (redrawn from Reynolds, 1997).
With the assumptions made and expressed in Vm (or V) and α, several statements
regarding earth materials can be made. According to Olhoeft (1981) and Neal (2004), freshwater
has an elevated ε in comparison to air and common rock-forming minerals (Table 2).
materials (Topp et al., 1980; Davis and Annan, 1989; and Neal, 2004). As ε is a primary
influence on the velocity of EM waves through the subsurface, it is clear that water content and
the presence of other high-conductivity substances (i.e., clay-rich soils, magnetic minerals) in the
subsurface profoundly affect the assumptions made earlier to describe V and α. Without such
interferences, as a general rule, lower ε results in higher V, and lower σ results in lower α
(Reynolds, 1997; and Neal, 2004). When determining the two-way travel time (TWT) and the
Relative
dielectric EM wave
permittivity velocity Conductivity Attenuation
Medium (εr) (V, m/ns) (σ, mS/m) (α, dB/m)
Air 1 0.3 0 0
Fresh water 80 0.03 0.5 0.1
Unsaturated sand 2.55-7.5 0.1-0.2 0.01 0.01-0.14
Saturated sand 20-31.6 0.1-0.2 0.01 0.01-0.14
Unsaturated sand and gravel 3.5-6.5 0.09-0.13 0.007-0.06 0.01-0.1
Saturated sand and gravel 15.5-17.5 0.06 0.7-9 0.03-0.5
Unsaturated silt 2.5-5 0.09-0.12 1-100 1-300
Saturated silt 22-30 0.05-0.07 100 1-300
Unsaturated clay 2.5-5 0.09-0.12 2-20 0.28-300
Saturated clay 15-40 0.05-0.07 20-1000 0.28-300
Unsaturated till 7.4-21.1 0.1-0.12 2.5-10 not available
Saturated till 24-34 0.1-0.12 2.5-10 not available
Bedrock 4-6 0.12-0.13 0.00001-40 0.000007-24
Table 2. Table with values for common materials at 80-120 MHz. Note the significant change in
conductivity with respect to several saturated versus unsaturated materials. Modified from Neal
(2004).
Some energy from an EM wave will be reflected after propagating through the subsurface
and encountering a significant discontinuity with respect to ε, µ, or σ (Neal, 2004). The strength
The reflection, or amount of energy reflected is referred to as the reflection coefficient (R),
the assumption that µ, or σ are insignificant (Neal, 2004). Similarly, the ε components of the
above expression can be replaced with V values of adjacent layers to obtain the same R
coefficient which will be between +1 and -1 (Neal, 2004). The values for ε and V in Table 2
would suggest that R is heavily dependent on the physical and chemical make-up the materials of
adjacent/overlying layers in the subsurface. Therefore, relatively small changes in the physical
The radar reflections are recorded on a fixed, invariant time base so the depth to a
reflector surface can only be estimated if the velocities of the radar waves are known (Neal,
2004). As well, it would be imprecise to interpret a 2-D radar reflection profile by assuming that
each recorded radar reflection was caused by a reflector surface or point directly beneath the
survey point. Radar antennae radiate and receive EM energy in a complex 3-D cone which
indicates a reflection on a radar survey trace could have originated from anywhere on the radar
wave front (Neal, 2004). This effect is complicated by the fact that GPR antennae transmit EM
energy across a range of frequencies (Conyers and Goodman, 1997), and higher frequencies
wavelengths (Jol, 1995 and Bano, 1996). So, the notional center radar frequency (e.g., 400 MHz)
for a GPR unit will be different from the most common or return center frequency detected by
the receiving antenna (Neal, 2004). Therefore, enhancing the high frequency (preferentially
attenuated) data during post-processing will increase the vertical resolution of the survey trace
(Neal, 2004).
Vertical and horizontal resolutions determine the ability of GPR to identify and define a
dependent on the wavelength (λ), wavelet sharpness, and pulse width (Knapp, 1990). Therefore,
11
vertical resolution increases as frequency increases (λ = V/f). This relationship becomes relevant
during post-processing attempts to enhance the preferentially attenuated high frequency data.
When the return center frequency (or most common frequency detected) is higher, λ decreases,
and vertical resolution increases (Neal, 2004). Sheriff (1977) indicates that the best vertical
resolution that can be attained is one-quarter the dominant wavelength. Maximum attainable
vertical resolution for low-loss materials (e.g., sand, gravel) with a high frequency antenna is
between 0.02 and 0.08 m (Neal, 2004). Reynolds (2007) reports a range from 2-23 cm depending
Bedrock
wavelength (cm) 92 22 12
resolution (cm) 23 5.5 3
the ground takes the form of a cone with a finite-sized footprint (Reynolds, 1997). That said,
horizontal resolution is dependent on the center wavelength frequency (λ), the width of the first
Fresnel zone, and the average εr (Neal, 2004). The first Fresnel zone is a function of frequency
dependent wavelength and depth (D) to a reflector surface and describes the minimum horizontal
area in which features with smaller dimensions will not be imaged (Sheriff, 1977; and Reynolds,
1997). The depth (D) is important as radiated EM energy expands laterally as it propagates
through the subsurface indicating that horizontal resolution decreases with depth (Neal, 2004).
However, during a GPR survey the EM energy propagates downward in a cone elongated in the
12
direction of the survey. With the depth to a reflector surface in mind, the larger the first Fresnel
zone is (indicating longer λ), the lower the horizontal resolution (Reynolds, 1997). This
dimension radius footprint, λ = center frequency wavelength, D = depth to reflection surface, and
attenuation factor) (Daniels et al., 1988); therefore, horizontal resolution decreases over low-loss
Figure 3. Determining radar footprint size, modified from Conyers and Goodman (1997) and
Neal (2004)
When a GPR antenna is placed directly on the ground, the transmitted EM wave form is
not exactly reproduced as it propagates into the ground. This EM wave distortion is known as the
13
ground coupling effect or ‘ground coupled’ (Reynolds, 1997) signal. The ground coupling effect
acts as a filter that changes the form of and decreases the energy of the EM wave transmitted into
the subsurface. With ground coupling and as previously indicated, higher frequencies become
Ground coupling effects become important when evaluating vertical and horizontal resolution
with closely-spaced reflector surfaces. For the same reasons vertical and horizontal resolution are
important considerations, the ground coupled signal effect can mask subsurface reflectors when
higher frequencies are preferentially attenuated, providing additional justification for enhancing
The EM energy reflections are what the GPR receives, records, and displays as an
individual trace as a function of time. The character of these reflections as the energy propagates
through the subsurface to a particular target and then back to the surface are what make the GPR
survey method work. However, the character of these reflections, as well as reflection
interferences, and interpretational complications related to the structure and type of subsurface
materials present and are also why numerous problems arise during data interpretation, and why
limitations related to the GPR system configuration and capabilities in a given subsurface
environment exist. With any application of GPR, the user must have a thorough understanding of
the principles underlying the GPR technique, the effects of data collection and system
configuration, effects of survey area variations, vertical and horizontal resolution, depth of
penetration, causes of reflections, and the values of different forms of data post-processing (Neal,
2004).
14
As noted by the previous Sabo resistivity study, the sandy clay soils that overly the
solution cavity features at the Crystal Rock Cave site appeared to have had only a minor impact
on the ability of the method to detect the subsurface solution cavity features. Incidentally, a
second site surveyed by Sabo using the resistivity method was found to have dense clay-rich till
that appeared to adversely impact the ability of the method to identify the solution cavity
features. Thick overlying clay-rich soils with a high electrical conductivity (Doolittle, et al.
2007), could potentially impact the ability of the GPR to detect the solution cavity features. The
performance of GPR in soils of high electrical conductivity can be diminished as radar signal
conductivities such as clay-rich soils and/or soils containing dissolved salts (Doolittle et al.,
2007). For these reasons, the sandy clay soils at the Crystal Rock Cave site may be less of an
obstacle for the GPR. GPR and resistivity methods are both considered suitable geophysical
characteristics such as permittivity and conductivity, and resistivity is dependent on the resistive
properties of subsurface materials (Reynolds 1997). With both methods deemed potentially
suitable, it appeared as though this would be a well-suited field site to test the GPR against the
resistivity method.
Accurately interpreting GPR reflection profiles requires a detailed knowledge of the local
environment (Bristow and Jol, 2003). It was beyond the scope of this study to conduct ground-
15
truth excavations or subsurface coring; however, the following information is provided in order
The Crystal Rock Cave site is located in the far northwest corner of Erie County, Ohio to
the northwest of the intersection of Wahl Road and Crystal Rock Road, approximately 0.8-
kilometers to the south of the town of Crystal Rock. The field area is approximately 0.13-square
kilometers in total size, with the north and central portions of the site occupied by a wooded area,
and the south portion of the site occupied by structures and other improvements (Figure 4). The
site is generally flat, with relatively low relief, and exhibiting slightly rolling topography typical
Physiographically, the field site is located just west of the Columbus Escarpment within
the Huron-Erie Lake Plains Section of the Central Lowland Province of the Interior Plains
Division of Ohio (ODGS, 1998). The bedrock underlying the site that bears numerous solution
cavity features is the Silurian Salina Group undifferentiated (ODGS, 1990). According to White
(1926), ODGS (1990, revised 2000, 2004), and Sabo (2008), the bedrock within the Silurian
Salina Group is further differentiated as the Put-In-Bay Dolomite within the Bass Island
Formation and is composed of massive, fine-grained dolomite interbedded with halite, anhydrite,
or shale and predominately overlies, but in some locations is intertonguing with, the Tymochtee
and Greenfield Dolomites. Sparling (1970) documents the Bass Island unit to be about 750-feet
of strata deposited under hypersaline conditions. The ODNR describes the bedrock at the site as
the Salina Group undifferentiated (Upper and Lower Silurian) and composed of dolomite,
anhydrite, gypsum, salt, and shale; with the dolomite further described as gray to brown, very
finely crystalline, mostly in thin to medium beds and laminae (with shale, anhydrite, and/or
gypsum laminae), with brecciation in the upper one-sixth of the unit (Figure 5). The USGS
describes the bedrock at the site as the Salina Group which is laminated to thinly bedded; gray,
yellow-gray to olive-gray dolomite with occasional thin bed and laminae of dark gray shale and
1958). Easterbrook (1999) indicates that the amount of substitution of magnesium for calcium
can be highly variable within a given carbonate species; with purer limestone (60-90% CaCo3)
more prone to form karst. Dolomite is less prone to form karst due to a typical lower
Although several classifications exist, the first described Put-In-Bay Dolomite within the
Bass Island Formation may be more appropriate, as subdividing (differentiating) the Upper
Silurian Salina and Bass Islands Groups (as in Figure 5) is only routinely feasible in the
subsurface with surface outcrops of the units often impossible to differentiate (Sparling, 1970).
Sparling (1970) also documents numerous investigations over the past 100 years of the site area
regarding the often disputed depositional sequence and dolomitization of the bedrock units of the
site area.
dolomitization, and when and how it occurred. Although the nature of limestone/dolomite, and
its proclivity for karst is of prime importance when understanding the nature of subsurface voids
and cavities, a more thorough evaluation of the depositional environment of the bedrock in the
Figure 5. Bedrock geology of the site region, modified from the ODNR. ODNR bedrock geology
of 7.5 Minute Castalia, Ohio Quadrangle (1969) and overlaid on the 1969 USGS Castalia, Ohio
Quadrangle.
18
According to USDA (1992), the soils at or near the surface of the Crystal Rock Cave site
consist of the Castalia very channery (i.e., coarse bedrock fragments) loam, 12 to 18-percent
slopes (CcD); Marblehead loam, 0 to 6-percent slopes (MbB); Ritchey loam, 0 to 2-percent
slopes (RhB); Sandusky loam, 0 to 1-percent slopes (SaA), and Milton silt loam, 2 to 6-percent
slopes (MnB). The USDA Soil Survey (1992) indicates the SaA and MnB occupy small,
northern slivers of the site (nearly off-site); the CcD occupies a northeast to southwest trending
tongue in the north portion of the site; the RhB occupies the central portion of the site; and the
MbB occupies the south of the site and the majority of the north areas of the site with exception
to the areas occupied by the other soils already identified (Figure 6).
USDA (1992) documents the CcD to have a clay content ranging from 12-20%, low
organic content (0-6%), increasing CaCO3 with depth (5-70%), and a decreasing cation exchange
capacity (CEC, expressed as milliequivalents per 100 grams of clay, or meq) at shallow depths
ranging from 0 to 58 cm below the ground surface (bgs). The USDA Soil Survey documents the
MbB to have a clay content ranging from 10-20%, an organic content ranging between 3 and 8%,
increasing CaCO3 with depth, and an increasing CEC at shallow depths ranging from 0 to 25 cm
bgs.
USDA (1992) documents the RhB to have a clay content ranging from 18-35%, an
organic content ranging between1 and 3%, increasing CaCO3 with depth, and a variable CEC at
depths ranging between 0 and 25 cm bgs. The USDA Soil Survey documents the SaA to have a
clay content ranging between 10-50%, an organic content ranging between .1 and 8%, decreasing
CaCO3 with depth, and a variable CEC at depths ranging from 0 to 2 m bgs. The USDA Soil
Survey documents the MnB to have a clay content ranging 14 to 50%, organic content ranging 1
to 3%, increasing CaCO3 with depth, and increasing CEC at depths ranging from 0 to 68 cm bgs.
19
Figure 6. Soils present on the site. Modified from USDA (1992). Soils classified as ice-
deposited, clayey tills of Late Wisconsinan – Late Woodfordian age (18-14 kA) and described as
poorly-drained, very-flat, lake-planed moraines, planed by waves of glacial lakes, present with
small patches of sand, silt, and clay at the surface (Pavey, et al., 1999).
20
As noted by Doolittle et al. (2007) and Daniels (2004), the penetration depth of GPR is
determined by antenna frequency and the electrical conductivity of the earth materials being
profiled. As described earlier, subsurface materials with high electrical conductivity rapidly
attenuate GPR signal energy and limit the penetration depth of GPR. For the purposes of this
study, the variable clay content, CEC, and CaCO3 component (USDA, 1992) will be considered
when evaluating the GPR suitability over a given area of the site that is occupied by a specific
soil type.
Clays are more adsorptive, have more surface area, and therefore retain more water than
silt and sand fractions of soils (Doolittle, et al., 2007). As previously indicated, saturated soils (or
soils with higher water content) are more conductive and therefore more attenuating to the GPR
signal. The conductivity of a soil is also dependent on concentration of dissolved salts within the
soil (Doolittle, et al., 2007). Each clay mineral within the soil has a different CEC, or ability to
adsorb cations (Easterbrook, 1999). A soil with a higher CEC would be expected to have a
higher concentration of dissolved salts (i.e., calcium, magnesium, sodium) which would increase
the soil conductivity and attenuate the GPR signal. Soils with a higher CaCO3 component can
also increase the electrical conductivity of the soil and attenuate the GPR signal (Grant and
Shultz, 1994). Therefore, the lower the clay content, CEC, and CaCO3 component of a given soil
type, the more suited the soil type is for deeper GPR penetration depth. The different soil types at
the site will have an impact on the ability of the GPR to achieve high penetration depths (Table
2, Table 4). Soils with high, saturated clay components were earlier determined to dramatically
Soil Type * Clay (%) CEC (meq/100 g) CaCO3 (%) Depth (inches bgs)
CcD 12-20 5-24 5-70 0-23
MbB 10-20 0-21 0-55 0-8
MnB 14-50 10-30 0-5 0-27
RhB 18-35 13-23 0-20 0-14
SaA 10-50 4-45 10-110 0-80
Table 4. Soil types present on the site and pertinent characteristics related to GPR suitability.
Modified from the Erie County Soil Survey
According to Doolittle et al. (2007), soils with a low clay component (< 15%) are
favorable to deep GPR penetration, while soils with a high clay content (≥ 35%) are restrictive to
deep penetration with GPR. Sandy clay soils are reported to exist at the site. With sufficient
fractions of adequately dry or unsaturated sand present within the soils at the site, GPR signal
attenuation should be kept to a minimum as unsaturated sand is reported to have a low σ (Table
2). However, with the variable nature of the site’s soil characteristics with depth and spatial
variability (as well as water content), the depth of penetration (through the surface soils) across
the site will also be variable. A more detailed analysis (i.e., probing, sampling, testing) of the
soils at the site was beyond the scope of this study; however, a preliminary assessment of the
data in Table 4 indicates the MNB and SaA soils, where present and with upper range clay
contents (50%, also with high CEC and CaCO3) would be the poorest suited for GPR. These soils
are also the deepest soils at the site and present in areas with known cave and other solution
cavity features identified by Sabo (2008) (Figure 6). However, in an ideally unsaturated scenario,
the electrical properties of the soils at the site appear to possess sufficiently low σ with which to
Dolomite is shown to possess relatively low electrical conductivity (σ), which has been
shown to have the greatest influence over GPR attenuation (Figure 2). The interbedded halite,
anhydrite, and shale bedrock reported at the site is also shown (if unsaturated) to potentially have
a low σ (Table 2). Data from Olhoeft (1981) and Schön (2004) indicate that halite, anhydrite, and
shale have low εr values which, if unsaturated, would indicate lower σ in comparison to
overlying soils within ideal conditions. Furthermore, any air within cavities or voids in the
bedrock would notionally have a σ of zero (also indicated in Table 2) and not contribute to GPR
signal attenuation. A more detailed analysis (i.e., coring, sampling, analysis) of the bedrock at
the site was beyond the scope of this study. However, in an ideally unsaturated scenario, the
electrical properties bedrock and associated air-filled cavities at the site appear to possess
sufficiently low and/or contrasting σ with which to allow ample GPR penetration depth and
target identification.
In summary, the soils, bedrock, and associated air-filled cavities at the site appear to have
differentiation. To accomplish a main objective of this study, these factors become critical for
The GPR unit utilized during field work and data collection portions of this study
consisted of a Geophysical Survey Systems, Inc. (GSSI), TerraSIRch SIR System-3000 (SIR
3000) ground penetrating radar (GPR) with a 400 MHz transducer/receiver antenna mounted on
a three wheel cart configured to be operated with the GPR antenna in the place of the front cart
wheel (this configuration, often utilized for applications over uneven ground, replaces the front
23
wheel of the cart with the GPR antenna case). The data collection was performed via single-fold,
fixed-offset reflection (bistatic) profiling (Bristow and Jol, 2003). The GSSI GPR (Model 5103)
unit consists of a 400 MHz ground coupled antenna, with an advertised depth of viewing window
of approximately 4 m (assuming a dielectric constant of 5), range: 50 ns, samples per Scan: 512,
resolution: 16 bits, number of gain points: 5, vertical High Pass Filter: 100 MHz, vertical Low
Pass Filter: 800 MHz, scans per second: 64, and transmit Rate: 100 KHz (GSSI, 2003). The GPR
unit consisted of a digital control unit with operating system, GPR antenna assembly (in plastic
transit case), 10.8 V lithium-ion battery, wheeled cart, and operator manual. The control unit
consisted of a keypad, display screen, connector panel, battery slot, and indicator lights. The
connector panel had six slots/connectors for such things as a compact flash memory card, AC
power, serial port, Ethernet, and/or USB connections. The GPR control unit has an internal
memory of 512 MB (if no other external memory devices were utilized) (GSSI, 2003).
The physical set-up of the GPR unit (discussed in the following sections, modified from
GSSI, 2003) began by assembling/connecting the GPR cart frame components and securing the
components with metal pins per manufacture guidelines. The cart wheels were then mounted on
the cart by inserting the wheel axles into the cart axle. The GPR survey wheel apparatus, located
against the left rear wheel, was positioned so that it made contact with the GPR cart wheel. The
GPR antenna case was then mounted in place of the front cart wheel by assembling the front
mount kit and attaching the antenna unit with fasteners and instructions provided by the
manufacturer. The front mount assembly and antenna case were then attached to the front fork of
the assembled two-wheel cart. Mounting the antenna assembly on the cart was completed once
confirmed that the bottom of the antenna case was be able to contact the ground surface. The
female end of the antenna control cable was then fastened to the control port on the antenna case
24
and the lead wire from the survey wheel apparatus was connected to the survey port and both
hand-tightened.
The GPR control unit is attached to the upper mounting brackets of the assembled cart
frame (near the cart handle) and secured with manufacturer provided thumb screws (hand-
tightened). To connect the GPR antenna assembly to the control unit, the male end of the antenna
control cable was then attached to the control unit by inserting the antenna control cable
connector into the large, 19-pin protruding connector at the back of the GPR control unit and
hand-tightening. The lithium-ion battery is inserted into the GPR control unit, which causes the
GPR control unit to turn on automatically. The GPR unit mode set-up is then configured for use
on the site. Prior to collecting data at the Crystal Rock Cave site, the GPR equipment was
assembled, configured, and tested at a remote location to ensure proper operation (Figure 7).
Data collection, survey design, and target selection are important components of a
successful GPR survey (Bristow and Jol, 2003). It is important to consider the goal of a GPR
survey in order to determine what data are required to accomplish the objectives. In the case of
this study, the contrasting EM energy amplitude reflections from air-filled cavities or sediment-
filled collapse structures (former solution cavities) and the surrounding native soils and bedrock
were primary targets. These same objectives also dictated the size of the surveyed area, depth of
sampling intervals, and locations of survey lines (Bristow and Jol, 2003).
Ground-penetrating radar quantifies entire waveforms per unit volume, giving it its three-
dimensional character; and increasing electromagnetic interference from cell phones and other
sources complicate this character and cause unpredictable fluctuations in recorded signal
amplitude (Ernenwein and Kvamme, 2008). Reflected signal amplitudes and ‘breaks’ in the EM
25
Figure 7. GSSI 400 MHz cart-mounted antenna GPR, as configured for Crystal Rock Survey
wave trace are what are interpreted in the field and post-processed to create subsurface profiles
of the target features. These recorded signals are of primary importance to this study.
Environmental operating errors such as uneven ground and surface vegetation can also jar the
antenna causing unpredictable fluctuations in recorded signal amplitude. Therefore, during the
survey (and when feasible), appropriate steps were taken to minimize surface reflector influence
(i.e., trees, fences), environmental operating errors, and outside electromagnetic interference
A thorough evaluation of the site area was conducted during the field work to identify
any of the above sources of interference and error. The locations of several of the transects
26
precaution, no cell phones were used and the locations of surface obstacles were annotated for
RESULTS
The GPR survey data were collected on April 1, 2010. The survey data consisted of nine
transects within the study area. Transects 1 through 5 were completed in the south portion of the
site (near Winkel’s Cave) and transects 6 through 9 were completed in the north, wooded portion
of the site in the vicinity of Crystal Rock Cave (Figure 8). All transects were first collected at the
5m depth setting. Transects 1 through 5 were surveyed again using a depth setting of 10m. To
differentiate between the two during discussion, they are referred to as Transect 1a and 1b (5m
and 10m, respectively). An additional, 10m depth transect was also collected in the vicinity of
Crystal Rock cave and referenced accordingly as a ‘b’ transect when referenced.
Topography along transects 1a through 5b, in the south portion of the site was relatively
flat; therefore, no topographic data were necessarily applied during processing and interpretation.
Other transects in the north portion of the site had significant topographic variation; however,
A main objective of this study was to evaluate the GPR data over karst topography and
identify solution cavities and associated features. Initially, the radar reflections and diffraction
hyperbolae were evaluated in the field prior to data processing in order to qualitatively evaluate
the data and identify areas of interest. Some initial data processing during data acquisition was
accomplished using manufacturer- configured default high and low-pass filters and gain settings.
28
Figure 8. GPR transect locations. The red “x” marks the approximate location of Crystal Rock
Cave and the yellow “x” marks the location of Winkel’s Cave (Sabo, 2008)
29
Post-processing the data after collection was also used sharpen the detail and correct the
profile. In order to avoid losing or not collecting valuable data during the acquisition phase, the
default settings on the GPR unit were not manipulated and left as broadband as possible
(Reynolds, 1997). The data collection mode and settings used were provided in the “Geology
Following the data collection phase, the data were downloaded and transferred to a PC
for initial review prior to processing (Figure 9). The data were processed using GPRSoft
software available from Geoscanners AB. This software provides the capabilities of viewing the
Figure 9. GPR data processing and analysis. Modified from Bristow and Jol, 2003
Data Processing
The data were first time-zeroed to account for air and ground wave reflections and time
windows. This essentially selects the surface within the profile and reallocates the two way travel
time information through the profile. The data were then ‘dewowed’ or filtered using default
30
software parameters. This step will correct the data for false reflections caused by the GPR
equipment or other nearby surface object interference. Gain adjustments were then implemented
across the profiles. This step involves an attempt to equalize the reflection amplitudes all the way
down each trace. This step varied by profile; however, gain was typically applied linearly or
logarithmically across the time window of the profile to enhance deeper reflections and/or
attenuated signal energy. This is known as spherical and exponential gain compensation (SEC)
which applies a linearly and exponential increasing time gain (Bristow and Jol, 2003). An
attempt to sharpen the image by filtering the data is often a primary step of data processing and
often sufficient to locate subsurface features (Reynolds, 1997). Filtering is the most basic and
most powerful post-processing step and can include forms of signal averaging (Bristow and Jol,
2003) and high and low pass (frequency) filters. This processing step estimates the velocity and
dissipation of energy (Sensors and Software, 1999). However, the processing completed by this
point had resulted in discernable parabolic reflections and velocity ‘pull-up’ anomalies indicative
of the target features. Consequently, no additional data filtering was necessary. The next step is
typically to restore correct subsurface geometrical relationships through the processing technique
subsurface features are in the exact locations as viewed on the unprocessed GPR scan display.
However, migration was not implemented due to the relatively shallow nature and discernable
scale of the target features. It is important to remember that both unprocessed data and highly
processed data can lead to incorrect interpretations. In most cases, only basic processing
techniques are needed to reveal valuable information within the GPR profile.
31
Interpretation
Interpreting GPR data is a complex operation that can be subjective at times and
relatively unambiguous at others. Identifying karst features, such as joints, fractures, cavities,
voids, and collapse structures in the subsurface was primary objective of this study. Following
processing, the 2-D profiles were evaluated for any of the phenomena associated with karst. The
fact that radio waves travel about three times faster in air than in solid material is one critical
component of the interpretation. This contrast results in the pronounced velocity ‘pull-up’ effect
in association with a significant void in the subsurface (Reynolds, 1997). Reynolds (1997) also
indicates that resonance within the void (also discernable in a 2-D profile) occurs when the
incident EM energy wavelength is the same as or shorter than the dimensions of the void. The
velocity ‘pull-up’ effect can result in high (bright) amplitude reflections at the interface between
rock/soil and the air-filled cavity. A macro-scale hyperbolic reflection pattern (i.e., stacked
hyperbolas or stacked reflections) taking an inverted chevron shape can indicate the presence of
soil or air-filled void, while a reflection-free pattern may indicate homogeneous materials such as
clay-rich soils, bedrock, or groundwater containing highly conductive dissolved minerals that
attenuates the most of the EM energy (Bristow and Jol, 2003). Discontinuities along banded
reflections may indicate the presence of joints or fractures . A mottled reflection pattern may
indicate fractured bedrock or soil or that gravel fill materials are present that scatter EM energy.
In instances where cavities/voids/caves were interpreted as present, due to the resonance of the
EM energy within the void and likely irregular and unknown shape of the bottom of these
features, the horizontal extent and depth to the upper interface of bedrock and void were most
discernable.
32
Transect 1a
Transect 1a is 170 m in length and was surface-corrected (13.7 ns time window) and
filtered or dewowed (Figure 10). A 10 dB gain compensation was first logarithmically applied
through the profile between 0 and 81 ns, then linearly applied through the profile beginning with
5 dB at 0 ns through 5 dB at 80 ns.
The GPR profile depicts karsted bedrock between 0 and 80 m along the transect at depths
between 1 and 3 m. Soil and air-filled cavities are present throughout this same section of the
transect between depths of 1.5 and 2.5 m. Several discontinuities are visible along internal
bedrock reflections that may be indicative of fractures. Relatively competent bedrock (free of
karst) is depicted between 80 and 115 m and 135 m and the end of the transect. Between 115 and
135 m, a large, soil-filled collapse structure is present from the top of the profile to a depth of at
least 4.5 m. An undulating (and karsted) bedrock surface is present around the area that appears
In comparison to the Sabo (2008) resistivity pseudosection, the GPR Transect 1a and
particular, the soil-filled karst near the end of the transect. Additional detail is present and
interpretable within the GPR profile between 0 and 80 m, that is present but was not interpretable
in the resistivity pseudosection. Overall there is generally more detailed information within the
Figure 10. Comparison of GPR and resistivity profiles along Transect 1a. (a) GPR profile and interpretation. (b) Resistivity
pseudosection from Sabo (2008). (c) Sabo’s interpretation of pseudosection.
34
Transect 2a
Transect 2a is 140 m in length and was surface-corrected (12.8 ns time window) and
filtered or dewowed (Figure 11). A 10 dB gain compensation was first logarithmically applied
through the profile between 0 and 81 ns, with 5 dB then linearly applied through the profile.
indicate the presence of air-filled cavities between 0.75 to 3.75 m depth. Within the same regions
of the transect, lower amplitude reflections within the bedrock indicate the presence of soil-filled
karst features, with discontinuities along reflection horizons potentially indicating fractures.
Between 65 and 92 m along the GPR transect, a soil-filled collapsed karst feature is present to a
depth of 4.5 m.
In comparison with the resistivity pseudosection (Figure 11), the GPR Transect 2a and
along the GPR transect may correlate with the same feature in the resistivity pseudosection, but it
is not immediately obvious due to scale. The interpreted bedrock surface from 60 to 180 m along
the resistivity pseudosection is similar to the bedrock surface interpreted in the GPR profile.
Overall, there is generally more detailed information within the GPR profile, which facilitates
structural interpretation. Differences in the profiles are likely due to the three-dimensional,
highly irregular nature of karst features and the fact that the start and end points and exact
locations of the transects were not exactly reproducible during the GPR survey.
35
Figure 11. Comparison of GPR and resistivity profiles along Transect 2a. (a) GPR profile and interpretation. (b) Resistivity
pseudosection from Sabo (2008). (c) Sabo’s interpretation of pseudosection.
36
Transect 3a
Transect 3a is 147 m in length and was surface-corrected (13.2 ns time window and
filtered or dewowed (Figure 12). A 10 dB gain compensation was first logarithmically applied
through the profile between 0 and 80 ns, with 5 dB then linearly applied through the profile.
The GPR profile shows karsted bedrock with several air-filled cavities between 0 and 70
m, 75 and 90 m, and 100 to 147 m along the transect and vertically from 1 to 4.5 m depth. These
features are indicated by high-amplitude reflections. Within these same areas, competent bedrock
is present beneath the karst features, and indicated by a relatively reflection-free profile. Two
soil-filled collapsed karst are present at 70 and 95 m along the GPR transect. These features
extend from the surface to 4.5 m depth. Between those two features, karsted bedrock with air-
filled cavities is present. The bedrock surface is relatively continuous along the first 65 m of the
profile; however, the bedrock is heavily karsted and present with several collapsed karst features
The GPR and corresponding resistivity pseudosection (Figure 12) exhibit several
similarities. The bedrock surface is similarly interpreted. Several of the soil-filled karst collapse
structures are also similarly interpreted between profiles, however not in terms of location along
the transect. The large soil-filled collapse structure is also identified in both profiles but is shown
to be different structurally. Overall there is generally more detailed information within the GPR
Figure 12. Comparison of GPR and resistivity profiles along Transect 3a. (a) GPR profile and interpretation. (b) Resistivity
pseudosection from Sabo (2008). (c) Sabo’s interpretation of pseudosection.
38
Transect 4a
Transect 4a is 127 m in length and was surface corrected (12.9 ns time window and
filtered or dewowed (Figure 13). A 10 dB gain compensation was first logarithmically applied
through the profile (0-81 ns), with 5 dB then linearly applied through the profile.
This transect shows relatively competent bedrock from 0 to about 20 m along the
transect. A soil-filled collapsed karst feature is present at about 30 m that extends from 1.5 to 4.5
m depth. An air-filled cavity is present at the bottom of this feature at 3 m depth. From about 35
m to the end of the transect, relatively competent bedrock is depicted, although an air-filled
cavity is present at 45 m, 1.5 m depth. The bedrock from 35 m to the end of the transect appears
relatively competent but may contain some soil-filled cavities (with several small, possibly soil-
filled karst features). Several soil-filled collapsed karst features are present at 90 and 110 m
along the transect to a depth of 2.5 m. These features give the bedrock surface an undulating
appearance.
similarities (Figure 13). The bedrock surface is again similarly interpreted as undulating,
especially between 30 and 150 m. The soil pipe at 40 m in the resistivity pseudosection
corresponds with the soil-filled collapse karst in the GPR profile at 30 m along the transect. The
majority of the air-filled cavities identified in the resistivity pseudosection were not identified in
the GPR transect. As previously described, differences between the two profiles are likely due to
the three-dimensional, highly irregular nature of karst features and the fact that the start and end
points and exact locations of the transects may have been different. Several of the soil-filled karst
39
Figure 13. Comparison of GPR and resistivity profiles along Transect 4a. (a) GPR profile and interpretation. (b) Resistivity
pseudosection from Sabo (2008). (c) Sabo’s interpretation of pseudosection.
40
collapse structures are also similarly interpreted between profiles. Overall there is generally more
detailed information within the GPR profile in terms of the structural extent of karst features.
Transect 5a
Transect 5a is 142 m in length and was surface corrected (13.3 ns time window) and
filtered or dewowed (Figure 14). A 10 dB gain compensation was first logarithmically applied
through the profile between 0 and 81 ns, with 5 dB then linearly applied through the profile.
Custom gain adjustments (between -3 and 8 dB) were then applied throughout the profile to
enhance features.
This transect shows a soil-filled karst feature at 10 m along the transect, extending to a
depth of about 1.5 m. Between 17 and 75 m, several high-amplitude reflections are present
between 0.75 and 1 m depth. These features overlie a significant cluster of continuous linear
high-amplitude reflections that may indicate the presence of a cave extending from 2.3 m to the
bottom of the profile. A soil-filled collapsed karst feature is present at 75 m, overlying karsted
bedrock below. Between 80 and 130 m, competent bedrock is present. From 130 m to the end of
the transect, a large soil-filled collapse structure is present and extends from the surface soil
down to 4.5 m.
similarities (Figure 14). The bedrock surface is similarly interpreted as undulating, often
indicating the presence of soil-filled collapse structures. The clay fill at 25 m and 160 m along
the resistivity transect correspond with similar features in the GPR profile. Several of the areas in
the resistivity pseudosection that were interpreted as bedrock with air-filled cavities also
41
Figure 14. Comparison of GPR and resistivity profiles along Transect 5a. (a) GPR profile and interpretation. (b) Resistivity
pseudosection from Sabo (2008). (c) Sabo’s interpretation of pseudosection.
42
correspond with significant high-amplitude reflections in the GPR profile that were also
interpreted as air-filled cavities. Overall, there is more detailed information within the GPR
Transect 6
Transect 6 is 274 m in length and was the longest transect completed at the Site (Figures
15, 16, and 17). The profile was surface corrected (11.5 ns time window) and filtered or
dewowed. A 10 dB gain compensation was first logarithmically applied through the profile
between 0 and 81 ns, with 5 dB then linearly applied through the profile. An additional 5 dB of
gain was again applied between 10 and 25 ns to enhance the profile. Due to the length of this
This transect shows a soil-filled collapse feature between 0 and 20 m along the transect to
a depth of 2.25 m (Figure 15). Relatively competent bedrock (with only a few air-filled cavities)
is present between 20 and 90 m. A large air-filled cavity is present at 60 m along the transect.
Beginning around 90 m and stretching to about 182 m down the transect, significant high-
amplitude reflections are present at depths ranging from 0.75 m to 5 m that coincide with the
known location of Crystal Rock Cave (Figure 16). Several of these reflections were detected in
areas away from the known cave location and may be indicative of additional karst features
associated with the cave. Another soil-filled collapse feature that overlies an air-filled cavity is
present just after the known cave location (Figure 16). Between 178 m and the end of the
transect, the profile depicts relatively competent bedrock, with the exception of a significant,
high-amplitude reflection from 210 to 230 m along the transect line at a depth of about 3.75,
similarities (Figures 15, 16 and 17). The clay-fill and air-filled cavities from 20 to 120 m along
the resistivity pseudosection are similarly depicted in the GPR profile. As well, the location of
Crystal Rock Cave is highly visible within both the resistivity and GPR profiles. The large air-
filled cavity at 310 m along the resistivity pseudosection correlates well with an air-filled cavity
signature in the GPR profile. Both data profiles are in agreement in most cases. However,
overall there is generally more detailed information within the GPR profile regarding smaller
Figure 15. Comparison of GPR and resistivity profiles along Transect 6. Split into three sections
due to length. (a) GPR profile and interpretation. (b) Resistivity pseudosection from Sabo
(2008). (c) Sabo’s interpretation of pseudosection.
45
Figure 16. Comparison of GPR and resistivity profiles along Transect 6. Split into three sections
due to length. (a) GPR profile and interpretation. (b) Resistivity pseudosection from Sabo
(2008). (c) Sabo’s interpretation of pseudosection.
46
Figure 17. Comparison of GPR and resistivity profiles along Transect 6 (split into three sections
due to length). (a) GPR profile and interpretation. (b) Resistivity pseudosection from Sabo
(2008). (c) Sabo’s interpretation of pseudosection.
47
Transect 7
Transect 7 is 47 m in length and was surface corrected (11.1 ns time window) and filtered
or dewowed (Figure 18). A 10 dB gain compensation was first logarithmically applied through
the profile between 0 and 83 ns, with an additional 5 dB applied between 11 and 24 ns to
This transect shows several air-filled cavities from the beginning of the transect to about
15 m (Figure 18). Relatively competent bedrock (with several reflections indicative of small air
or soil-filled cavities) is present from 14 m to the end of the transect (Figure 18). The large air-
filled cavities shown in the first 14 m of the profile appear over the northern portion of Crystal
The GPR Transect 7 and corresponding resistivity pseudosection and interpretation are
similar (Figure 18). Most of the air-filled cavities indentified the resistivity pseudosection were
also identified within the GPR profile. Relatively competent bedrock was also identified (within
both profiles) between 50 and 100 m along the resistivity transect. While both profiles show
general agreement, there is generally more detailed information within the GPR profile related to
Figure 18. Comparison of GPR and resistivity profiles along Transect 7. (a) GPR profile and interpretation. (b) Resistivity
pseudosection from Sabo (2008). (c) Sabo’s interpretation of pseudosection.
49
Transect 8
Transect 8 is 25 m in length and was surface corrected (11.3 ns time window) and filtered
or dewowed (Figure 19). A 5 dB gain compensation was logarithmically applied through the
This transect shows several large, air-filled cavities along the profile at depths of 1 to 3
m. These features are likely associated with Crystal Rock Cave, as they are present to the west of
the cave entrance. No resistivity pseudosection is available for comparison with this transect.
Transect 9
Transect 9 is 214 m in length and was surface corrected (20.2 ns time window) and
filtered or dewowed (Figure 20). This transect was collected at the 10 m depth setting. The
logarithmically applied through the profile (0 to 166 ns); (2) inverse logarithmic gain application
of 15 dB applied through the profile to enhance detail in the upper 4 m; (3) a custom (manual)
gain adjustment of 4 dB was then applied at 22, 43, 70, and 90 ns to equalize amplitude
reflections throughout the profile; and (4) 5 dB of gain linearly applied throughout the profile for
additional enhancement.
As a consequence of the profile length and low reflection resolution, only the profile
section from 22 to 107 m (near Crystal Rock Cave) was selected for discussion. From about 22
to 46 m, several air-filled cavities are present at depths between 0 and 4 m (Figure 20). These
features are possibly karst features associated with Crystal Rock Cave. From about 46 to 125 m,
a long continuous band (similar to Transect 6) of high amplitude reflections is present at depths
between 2 and 4 m. These reflections again coincide with the known location of Crystal Rock
Cave, and may be indicative of karst features associated with the cave. The resolution of the
profile is diminished, and the shallow karst features identified in the 5 m depth transect are not as
easily discernable. No resistivity pseudosection was created for comparison with this transect.
51
Transects 1b, 2b, 3b, 4b, and 5b were additional transects collected with 10 m depth
settings (Appendix A). The profiles were first processed the same as the initial 5 m profiles.
Following those steps, numerous custom gain corrections within selected time windows were
applied to enhance deeper features while retaining resolution at shallow depths within the profile.
Transects 1b, 2b, and 4b do not show any additional information or deeper features than
described during the transect 1a, 2a, and 4a interpretations. Transects 3b and 5b appear to
provide some additional information regarding several karst features present at depths greater
than 5m.
In addition to the features interpreted from transect 3a, transect 3b better shows the
apparent vertical extent (6 m depth) of the air-filled cavities identified at depth within the
transect. In addition to the features interpreted from transect 5a, and similar to the interpretation
of transect 3a, transect 5b shows the apparent vertical extent of several karst features. Transect
5b shows the possible cave feature identified in the central portion of the transect to terminate at
5 m depth.
Errors
The transects over the southern portion of the site were conducted over open, even
ground whereas transects in the northern portions were conducted over rolling, wooded terrain
with surface obstacles such as rocks and tree roots. The unavoidable surface obstacles along the
transects in the north jar the GPR antenna often causing the depth profile to appear mottled.
53
Surface topography in the north portion of the site was somewhat variable. Elevation data
for transects in the north portion of the site were not collected or obtained; therefore, only depths
relative to the surface can be determined, not absolute depths. Interpreting the transect profiles in
the north portion of the site was also hindered by to the inability to directly tie significant
The transects collected with the default 5 m depth setting required minimal post-
processing to obtain interpretable profiles. The deeper, 10 m scans required significant post-
process, especially multiple linear, logarithmic, and custom (manual) gain adjustments, in order
to enhance deeper reflections (Appendix A). By doing this, the shallow features (readily
observed in the shallow 5 m profiles) were often lost and unidentifiable in the deeper 10 m
profiles.
The traverse locations from the previous resistivity study were not accurately known for
the present study. During GPR data collection, the property owner, who was also present during
the resistivity survey, identified where the transects were collected to the best of his knowledge.
This uncertainty resulted in a potential lack of coincidence between the resistivity and GPR
transects. In some cases, significant offset (start and stop points, and overall transect length) was
present which precluded a direct comparison with significant amounts of resistivity data. In
addition, due to the three-dimensional nature of target features and the inconsistent transect
locations, it was often difficult to directly compare interpretations between the two methods.
Ground truthing was beyond the scope of this study. The exact depth of soils and depths
to bedrock were not obtained and not reliably available from other publicly available sources.
Without these data, interpreting the GPR profiles was more challenging and less precise.
54
CONCLUSIONS
The objective of this study was to utilize GPR to analyze karst and associated near-
surface features and compare findings with the towed electrical resistivity survey. Upon review
and analysis of the GPR transect profiles, the karst features were readily identifiable. As
previously postulated, there was sufficient contrast in the dielectric properties of the soils,
bedrock, and associated air-filled cavities at the site to allow identification and differentiation of
target features. The shallow (5 m) transects resulted in higher resolution datasets that clearly
showed more detail that aided interpretation. The deeper (10 m) transects required significant
processing and manipulation in order to enhance deeper reflections while still showing the detail
in shallow reflections (as achieved in the 5 m depth transects). In addition, the GPR and
associated antenna and operating system used were manufacturer-configured for shallow 5 m
transects, and increasing the depth to 10 m appears to have produced lower quality data when
In terms of the speed and ease of data collection within the physical environment at the
field site, the GPR proved to be well suited. The previous resistivity was also determined to be a
suitable method of identifying karst features. However, as described by Sabo (2008), the
durability and maneuverability of towed resistivity array led to operational difficulties and
inconsistencies during data collection. Upon comparison of resistivity and GPR profiles, there
were several cases along multiple transects where the profiles were in agreement in terms of
subsurface features present. Overall, the GPR survey provided more detailed structural
information, which often allowed for significant differences in interpretation when compared to
resistivity data. Due the three-dimensional nature of the GPR signal and high resolution of the
reflection pattern, it often appeared as though multiple features (soil-filled collapse feature
55
surrounded by bedrock) could be viewed simultaneously. However, due to the high contrast
between resistivity values of subsurface materials, cavities, and voids, the processed resistivity
pseudosections were more useful in determining the exact nature of the materials present, while
GPR provided a high level of structural insight. With these findings in mind, both methods
appear to be suitable for identifying karst features. The speed and ease of GPR data collection is
likely more desirable than a towed resistivity array, which has maneuverability and durability
issues. However, given the valuable data both methods can provide, it is likely they are best
REFERENCES
Bano, M., 1996. Consistent dielectric losses of ground-penetrating radar waves. Geophysical
Benito, G., Perez del Campo, P., Gutierrez-Elorza, M., Sancho, C, 1995. Natural and human-
Benson, A.K., 1995. Applications of ground penetrating radar in assessing some geological
Bristow, C.S., Jol, H.M., 2003. Ground penetrating radar in sediments. Geological Society,
Chamberlain, A.T., Sellers, W., Proctor, C., Coard, R. 2000. Cave detection in limestone using
Collins, M.E., Cum, M., Hanninen, P, 1994. Using ground-penetrating radar to investigate a
Conyers, L.B., Goodman, D., 1997. Ground penetrating radar. An introduction for
Cottingham, K., 1919. The origin of caves of Put-In-Bay, Ohio. The Ohio Journal of Science 20,
38-42.
57
Daniels, D.J., Gunton, D.J., Scott, H.F., 1988. Introduction to subsurface radar. IEE Proceedings
Daniels, D.J., 2004. Ground penetrating radar, 2nd edition. The institute of electrical engineers,
Davis, J.L., Annan, A.P., 1989. Ground-penetrating radar for high-resolution mapping of soil and
Doolittle, J.A., Minzenmayer, F.E., Waltman, S.W., Benham, E.C., Tuttle, J.W., Peaslee, S.D.,
2007. Ground-penetrating radar soil suitability map of the conterminous United States.
Doolittle, J.A., Collins, M.E., 1998. A comparison of EM induction and GPR in areas of karst.
Easterbrook, D.J., 1999. Surface processes and landforms, 2nd edition, Prentice-Hall, Inc., 352
pp.
Ernenwein, E.G., Kvamme, K.L. 2008. Data processing issues in large-area GPR surveys:
Grant, J.A., Shultz, P.H., 1994. Erosion of ejecta at Meteor Crater: constraints from ground
Radar, Waterloo Centre for Groundwater Research and the Canadian Geotechnical
Society, 789-803.
58
Geophysical Survey Systems, Inc., (GSSI), 2003. TerraSIRch SIR System-3000 User’s Manual,
Harris, J.G., Mylorie, J.E., Carew, J.L., 1995. Banana holes-unique karst features in the
Herman, J.S., White, W.B., 1985. Dissolution kinetics of dolomite: effects of lithology and fluid
Jol, H.M., 1995. Ground penetrating radar frequencies and transmitting powers compared to
penetration depth, resolution and reflection continuity. Geophysical Prospecting 43, 693-
709.
Knapp, R.W., 1990. Vertical resolution of thick beds, thin beds and thin-bed cyclothems.
Kraus, E.H., 1905. On the origin of Put-In-Bay caves. American Geologist 35, 167-171.
Neal, A., 2004. Ground-penetrating radar and its use in sedimentology: principles, problems and
Ohio Division of Geological Survey, 1990 (revised 2000, 2004). Generalized column of bedrock
units in Ohio: Ohio Department of Natural Resources, Division of Geological Survey, 1p.
Natural Resources, Division of Geological Survey, page-size map with text, 2 p., scale
1:2,100,00.
Ohio Division of Geological Survey, 1999 (rev. 2002, 2006). Known and probable karst in Ohio:
Olhoeft, G.R., 1981. Electrical properties of rocks, In: Touloukian, Y.S., Judd, W.R., Roy, R.F.,
Eds., Physical properties of rocks and minerals, McGraw-Hill, New York, 257-330.
Plumb, R.G., Noon, D.A., Longstaff, I.D., Stickley, G.F., 1998. A wave-form performance
Reynolds, J.M., 1997. An introduction to applied and environmental geophysics, John Wiley &
Sabo, S.H., 2008. Evaluation of capacitively-coupled electrical resistivity for locating solution
cavities overlain by clay-rich soils. MS Thesis, Bowling Green State University, 43 pp.
Schön, J.H., 2004. Physical properties of rocks - Fundamentals and principles of petrophysics.
Elsevier, Amsterdam.
Sensors and Software, 1999. Technical manual 29: PulseEKKO Tools, User’s Guide v2.0.
Shrake, D.L., Swinford, M., Schumaker, G.A., Larson, G.A., Slucher, E.R., 2007 (revised).
Sheriff, R.E., 1977. Limitations of resolution of seismic reflections and geologic detail derivable
Sparling, D.R., 1970. The bass islands formation in its type region. The Ohio Journal of Science
70, 1-33.
Theimer, B.D., Nobes, D.C., Warner, B.G., 1994. A study of the geo-electrical properties of
Topp, G.C., Davis, J.L., Annan, A.P., 1980. Electromagnetic determination of soil water content:
United States Department of Agriculture, Natural Resource Conservation Service, Erie County
United States Geological Survey, 1969. Castalia quadrangle, Ohio, 7.5-minute series, scale
1:24,000.
61
United States Geological Survey, 1999. Ground Water Atlas of the United States, internet web
Verber, J.L., Stansbery, D.H., 1956. Caves in the Lake Erie islands. The Ohio Journal of Science
53(6), 358-362.
White, G.W., 1926. The limestone caves and caverns of Ohio: The Ohio Journal of Science, 26,
73-116.
Zumberge, J.H., 1958. Elements of Geology, 2nd edition. John Wiley & Sons, Inc.
62