You are on page 1of 10

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

net/publication/250377594

A Comparison of Electrical Resistivity, Ground Penetrating Radar and Seismic


Refraction Results at a River Terrace Site

Article  in  Journal of Environmental & Engineering Geophysics · December 2008


DOI: 10.2113/JEEG13.4.325

CITATIONS READS

22 1,296

3 authors, including:

Laurence Bentley Peter Dietrich


The University of Calgary University of Tuebingen
137 PUBLICATIONS   1,368 CITATIONS    364 PUBLICATIONS   3,585 CITATIONS   

SEE PROFILE SEE PROFILE

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

AquiferAnalogue View project

Water and Earth System Science Competence Cluster (WESS) View project

All content following this page was uploaded by Laurence Bentley on 09 June 2015.

The user has requested enhancement of the downloaded file.


325

A Comparison of Electrical Resistivity, Ground Penetrating Radar and Seismic Refraction Results at
a River Terrace Site

Markus Hirsch1,*, Laurence R. Bentley2 and Peter Dietrich3,*


1
UFZ, Helmholtz Centre for Environmental Research – UFZ, Department Groundwater Remediation,
Permoserstrasse 15, 04318 Leipzig, Germany, Tel: ++49 (0) 341- 235-1506, Fax: ++49 (0) 341-235-1837,
Email: markus.hirsch@ufz.de
2
Department of Geoscience, University of Calgary, Calgary, 2500 University Drive, NW, Alberta, T2N 1N4 Canada,
Tel: (403) 220-4512, Fax: (403) 284-0074, Email: lbentley@ucalgary.ca
3
UFZ, Helmholtz Centre for Environmental Research – UFZ, Department Monitoring and Exploration Technologies,
Permoserstrasse 15, 04318 Leipzig, Germany, Tel: ++49 (0) 341- 235-1253, Fax: ++49 (0) 341-235-1837,
Email: peter.dietrich@ufz.de
*
Formerly Center for Applied Geoscience, Applied Geology, University of Tübingen, Sigwartstrasse 10, 72076
Tübingen, Germany

ABSTRACT

Electrical resistivity imaging (ERI), ground penetrating radar (GPR) and seismic
refraction (SRF) profiles were repeated over three lines on a terrace of the Bow River. The
site had a resistive gravel layer overlying mudstone bedrock with horizontal transitions to
lacustrine and overbank deposits. Electrical resistivity results were best for determining changes
in sediment types and detecting boundaries, but the ERI smoothness constraint blurred the
location of the boundaries. The GPR gave the most resolution and showed internal structures
that the other methods did not image. The GPR signal was severely attenuated in several areas
where the surficial sediments became too conductive because of a fine grained component. The
seismic refraction inversion provided good reproduction of the bedrock interface, but it did not
detect changes in the composition of the surficial sediments. It also required the introduction of
a low velocity surficial layer not indicated by the other methods that may be related to the
increase in effective stress with depth. Jointly interpreting the three data sets gives a more
reliable and less ambiguous interpretation than any single method. The data may be useful to
test joint inversion algorithms and are available for download.

Introduction Electrical resistivity imaging (ERI), seismic refrac-


tion (SRF) and ground penetrating radar (GPR) profiles
The Pine Creek field site is located on the southern were each collected over three lines (Fig. 1). Several old
limits of the city of Calgary in southwest Alberta, boreholes with geological logs existed along the lines,
Canada (Fig. 1). The site is the location of a future allowing for reasonably good verification of the
waste water treatment facility and potentially a major geophysical interpretations. This control and the repeat-
research facility that requires a large and reliable flow of ed surveys with different methods allowed us to evaluate
water. Several disadvantages to direct withdrawal of the resolution of the methods and the ability to identify
water from the nearby Bow River existed, so a study of changes in lithology. We only present results from Lines
groundwater extraction potential was conducted. A 1 and 2 because Line 3 was only partially covered by
geophysical study was performed in the summer of GPR and these two lines are adequate to demonstrate
2004 in support of the hydrogeologic study of the site. the main points of the paper. All three lines are
The hydrogeologic study showed that only minimal described by Hirsch (2004).
groundwater extraction potential existed because of the In the following, we will present the geologic
elevation of the bedrock and a lack of connectivity of setting of the site, briefly discuss the methods used, and
water bearing gravel to the adjacent Bow River. outline field acquisition parameters and the processing
Nevertheless, results of the geophysical study provide procedures. We will then compare the results of the
a unique data set that illustrates the strengths and different surveys and discuss the strengths and weak-
weaknesses of the geophysical methods and the advan- nesses. Finally, we will show an integrated interpretation
tages of an integrated interpretation. using all three methods.

JEEG, December 2008, Volume 13, Issue 4, pp. 325–333


326
Journal of Environmental and Engineering Geophysics

Figure 1. Location of Pine Creek field site and layout of survey lines.

Field Site deposits. In the west, the terrace is bounded by the


The Pine Creek field site is located on a terrace valley slope. The valley slope is made up of fine-grained
within the Bow River valley. The terrace is mainly lacustrine deposits. In the east, the site is bounded by the
underlain by sand and gravel from channel deposits, present Bow River channel. The terrace ends at a fairly
although some areas contain finer grained overbank steep slope with about five meters drop down to the
327
Hirsch et al.: A Comparison of Geophysical Methods at a River Terrace Site

Figure 2. Conceptual geological model of a west-east profile.

current river channel, which is at or below the bedrock Methods


level underlying the site. The surficial deposits are
underlain by the weathered surface of the Paskapoo Fm. Lines were laid out using a survey chain. Positions
bedrock. The Paskapoo Fm. was formed from fluvial and elevations were surveyed in with a total station and
deposits and contains significant mudstone. The bed- a global positioning system (GPS) receiver.
rock surface has been exposed and eroded and is an
uneven and sometimes weathered surface. The average Electrical Resistivity
depth to bedrock is about 5.5 m, and ranges between The ERI data were collected with a StingH system
2.5 m and 8.5 m. Figure 2 shows the geological concep- and the SwiftH switching box manufactured by Ad-
tual model of the field site that we developed from vanced Geosciences, Inc. The system had 56 electrodes
borehole logs (Thurber Engineering Ltd, 2004) and the in four sets of electrode cables each containing 14
surficial geology setting of Calgary (Moran, 1986). electrodes. Data were collected using a Wenner array
The field site has 39 boreholes and PVC cased (range: min n 5 1, max n 5 17) with multiple roll
piezometers with legacy borehole logs (Thurber Engi- alongs. Roll alongs were conducted by moving one cable
neering Ltd, 2004). The borehole logs provided us with at a time from the rear of the line to the front of the line
information about the elevation of geological boundar- to obtain continuous coverage. In order to obtain both
ies. The profile lines for the geophysical measurements shallow resolution and sufficient depth penetration,
were designed in a way to cross as many borehole lines were run with a 2-m unit electrode spacing and
locations as possible. repeated with a 4-m unit electrode spacing. The 2-m and
The campaign consisted of three lines. Two 4-m data sets were combined and then inverted with the
parallel profile lines (531.5 and 642.0-m length) start processing software RES2DINV (Loke and Barker,
at the western valley slopes and run east towards the 1996). Usually it took about five iterations to obtain an
Bow River. The third profile starts in the southern RMS error below 3%. A smoothness-constraint least-
section of the site and follows the Bow River bank to squares inversion method was applied.
the northern site boundary. This profile intersects with
the profiles 1 and 2 and has a total length of 1,257 m. Ground Penetrating Radar
Only a small portion of the third line was covered by The GPR survey was conducted with a pulse
GPR. The lateral offset from the river bank is about EKKO 100H radar unit manufactured by Sensors &
60 m. The profile lines were collected during the Software Inc. Data were collected with 100-MHz
months of May, June and July 2004. Periodic antennas. Antennas were oriented perpendicular to the
monitoring of groundwater levels showed only minor line and separated by 1 m. Measurement points were
variations throughout this time. The typical ground- separated by 0.5 m. Data processing was done with the
water level was around six meters below ground software REFLEXW (Sandmeier, 2004). An average
surface and site conditions were generally dry and well radar velocity of 0.11 m/ns was estimated from several
drained. common mid point experiments (CMPs) in combination
328
Journal of Environmental and Engineering Geophysics

Figure 3. Travel time picks for a) profile 1 and b) profile 2 (for reasons of clarity only a subset of the data is shown).
Examples of layer assignments are shown as open squares (layer 1), grey diamonds (layer 2) and black squares (layer 3)
for two shot gathers.

with well log information. This velocity was used for for each 118-m spread. Noise was produced by activity
time to depth conversion. Data processing steps in a gravel pit north of the site as well as from wind. In
included bad trace deletion, alignment of first arrivals order to reduce wind noise, most of the geophones were
to compensate for shifts between sections of lines run at buried in shallow holes. Due to signal attenuation and
different times, dewow, a 20/50 to 120/200 MHz ambient noise, only data for 60 m on either side of the
bandpass filter, manual gain, background removal, shotpoints have been used for the refraction inversion.
average X-Y smoothing filter and topographic correc- A selection of arrival times for profiles 1 and 2 are
tion. For background removal, an average trace is built shown in Figs. 3(a) and 3(b), respectively. For clarity,
from a chosen trace range. It is subtracted from the only a subset of the data is plotted.
section to eliminate consistent noise. X-Y smoothing is Data were processed with the software GLI3D
accomplished by averaging amplitude values within a (Hampson and Russell, 1984). GLI3D requires the
time-space window in order to suppress noise. specification of the number of layers and the velocity of
the first layer. The velocity and the thickness of each
Seismic Refraction layer are adjusted by an inversion algorithm until the
Data were recorded with a 60 channel StrataviewH model travel time is consistent with travel time picks.
seismograph manufactured by Geometrics. Multiple The velocity and thickness in each layer can vary
blows from a 16-pound sledgehammer on a metal plate laterally, but the velocity within a layer cannot vary
provided the energy source. Sixty geophones (Oyo- vertically. The algorithm is best suited to regions where
Geospace GS-20H, 28 Hz) were spaced at two meters. the geology is well modeled by a layered system and is
The geophone spread was moved forward in a varying slowly enough that ray-tracing is appropriate.
‘‘leapfrog’’ manner. Half of the spread was moved Initial velocities were determined from first arrival travel
forward along the profile and the other half that was at times. As will be discussed in further detail, the data
the end of the last spread was kept in place to work as required a three-layer model. Well logs were used to
the beginning of the new geophone spread. Each spread produce an initial estimate of depth of geological units.
(0 to 118 m) used five shots while all geophones were Based on arrival times and the velocity model, GLI3D
recording. Shots were initiated at 21, 29, 59, 89, 119 m then uses a generalized linear inversion algorithm to
329
Hirsch et al.: A Comparison of Geophysical Methods at a River Terrace Site
iteratively update the depth of the layer boundaries, V-m to the underlying lower resistivity of around 50 V-
which are not fixed and free to move. m is most likely not a geologic transition, but due to the
smoothing constraint of the inversion algorithm.
Petrophysics The GPR profile shows a strong reflector at the
gravel-bedrock interface (Fig. 4(b)). The top of bedrock
Before analysing the results, it is useful to review elevation is quite variable, consistent with an erosional
the expected response of the geophysical methods. We surface and the resistivity data. Notably, the depth is
expect that unsaturated gravel will have resistivity in the consistent with the borehole logs. The interpretation of
range of thousands of V-m, whereas silts and clays have the strong reflector between TH 03–05 and TH 03–06A
lower resistivity due to surface conductance and is unclear given that it seems to be coming from above
normally higher moisture content. Past experience the bedrock, as indicated by the borehole logs. Although
indicates that the Paskapoo Fm. has variable resistivity not compelling, there is some indication of a reflection
depending on the shale content. However, our experi- near the water table at piezometers TH 03–10 and TH
ence in the region is that the Paskapoo Fm. tends to be 03–6(A). The signal is attenuated below the bedrock
shaly and the resistivity is often less than 50 V-m. surface, consistent with the low resistivity of the fine
GPR reflections are caused by contrasts in the grained Paskapoo bedrock. The gravel layer appears to
dielectric permittivity. Reflections are often associated pinch out towards the upland starting at around 150 m.
with changes in soil texture partially due to the variations Although the general trend is consistent with the ERI
in moisture content with texture. We also expect a major data, the details appear different. The GPR signal is
change in dielectric permittivity at the sediment-bedrock attenuated by the lower resistivity silts and there is no
boundary. In addition, GPR signals are attenuated more useful signal between 0 and 80 m. Within the gravel, the
rapidly as the electrical resistivity decreases. GPR profile contains many internal reflectors.
Near-surface unconsolidated sediments are expect- The three-layer refraction inversion model is
ed to have low seismic velocity, often less than the speed shown in Fig. 4(c) and the inversion velocity model is
of sound in air. Velocity will generally increase with shown in Fig. 4(d). The top layer, layer 1, has a velocity
depth as increasing effective stress causes an increase in that varies between 260 and 330 m/s. Layer 2 is within
the bulk and shear moduli. Previous experience indicates the gravel layer and its velocity is approximately 600 m/s.
that the Paskapoo Fm. velocity is generally greater than Layer 3 is the bedrock and its velocity varies between
2,000 m/s. 3,800 and 4,100 m/s. The gravel-bedrock interface is
consistent with the piezometer data. However, the surface
Results is smoother than either the ERI or GPR images. The thin,
low velocity layer 1 is required to get reasonable
Line 1 convergence of the inversion. The ERI image does not
Figure 4 shows the results of Line 1. Piezometer show any surficial layer and the GPR has only
locations are shown on the profiles. The bottom of the intermittent shallow reflections. Our current hypothesis
piezometers indicates the location of the bedrock is that this layer is due to increasing effective stress with
interface and the level of the water table is indicated depth; as the overburden stress increases with depth, the
by blue filling. Lack of blue filling means that the velocity will increase. The increase is most rapid in the
piezometer is dry. ‘‘near surface’’ where the effective stress is low. We
The ERI inversions clearly show the high resistiv- interpret the layer 1-layer 2 interface to be the location
ity gravels overlying the lower resistivity bedrock where velocity increase with depth is no longer rapid.
(Fig. 4(a)). The resistivity indicates that the soil cover The combined image is shown in Fig. 4(e). For
is extremely thin and, during electrode placement, it was most of the line, the three methods and the borehole
seen that the soil layer was often only a few centimeters data are in reasonable agreement. Between 30 and
thick. The thin layer led to some difficulties with 130 m the refraction image shows inconsistencies with
electrode contact resistance. To the southwest, the the GPR and ERI images, especially in the transition
gravel thins and a surface layer of more conductive soil zone between the thick gravel package and the silty
is observed between 0 and 100 m. We interpret this less upland region from 32 to 50 m. It may be that the
resistive sediment as silt that has been deposited from velocity difference between the gravel and the uncon-
the adjacent upland. The resistivity contour of approx- solidated silt is not large, making SRF insensitive to the
imately 2,000 V-m is correlated with the gravel-bedrock sediment change. It also may be that the geology is
interface. As the line starts up the slope, there are changing too rapidly for the GLI3D algorithm, which is
indications of an increase in the elevation of the bedrock best suited for slowly changing subsurface conditions.
beneath the valley wall. The transition zone from 2,000 At less than 32 m, in the upland area, the refraction and
330
Journal of Environmental and Engineering Geophysics

Figure 4. Results of geophysical measurements on profile 1 : a) ERI, b) GPR, c) seismic refraction layer boundaries, d)
seismic refraction velocity as a function of down line distance for each layer, e) combined image (X = location where
correlation is vague).

ERI images both show two layers and shallow bedrock where it is overlain by lower resistivity silt from the
beneath the slope. We interpret this as the gravel having upland. In this case, there are two distinct high
pinched out and a relatively thin section of silt overlying resistivity zones interpreted as gravel channels (paleo-
the bedrock. channels) that are separated by a lower resistivity area
centered at approximately 250 m. In the northeast, the
Line 2 gravel is abruptly truncated at around 500 m. From
Line 2 was run parallel to Line 1, 450 m to the 500 m on, the near-surface resistivity is lower and we
south (Fig. 1). The ERI profile shows the high resistivity interpret these sediments as overbank deposits with
gravels overlying the low resistivity bedrock (Fig. 5(a)). perhaps a higher resistivity sand or sandstone lens at
Similar to Line 1, gravel pinches out to the southwest approximately 5 m depth.
331
Hirsch et al.: A Comparison of Geophysical Methods at a River Terrace Site

Figure 5. Results of geophysical measurements on profile 2: a) ERI, b) GPR, c) seismic refraction layer boundaries, d)
seismic refraction velocity as a function of down line distance for each layer, e) combined image (Y = location where
correlation is vague).

The GPR profile does not have the strong bedrock little to no penetration from 500 m to the end of the line
interface reflector seen in Line 1 (Fig. 5(b)). The due to the change from gravel to overbank deposits.
bedrock location is correlated with a loss of energy. Careful inspection of the GPR profile in the vicinity of
From 0 to 120 m the GPR has no signal, presumably 480 m shows a shallow diffraction pattern originating at
due to low resistivity silt deposits originating from the the gravel-overbank deposit boundary, indicating that
upland region attenuating the signal. Similarly, there is the boundary is quite distinct.
332
Journal of Environmental and Engineering Geophysics

Figure 6. Geological interpretation for profile 2.

The refraction model shows good agreement with GPR gives extremely good location of the
the depth to bedrock of the piezometer data except at the boundary between the gravel and the bedrock when
location of TH 04–14 (Fig. 5(c)). The velocity of layer 1 is there is sufficient penetration. In addition, the GPR
approximately 200 m/s, layer 2 is 540 to 770 m/s and provides some indication of the internal structure of the
layer 3 is 2,500 to almost 4,000 m/s (Fig. 5(d)). Although gravel and has the best resolution of the three methods.
the depth to bedrock is well represented, there is no However, the signal is lost when the surface resistivity
indication of the transition from gravel to overbank at decreases. The method cannot resolve the transition
480 m. There may be a small velocity contrast between from gravel to lacustrine sediments, nor the transition to
the overbank and the gravel deposits, or conditions are overbank deposits on the northeast end of Line 2.
changing too rapidly for the GLI3D algorithm, which is The refraction method provides excellent resolution
best suited for slightly varying geology. of the bedrock boundary in most locations. However, it
Up until the transition at 500 m, all the methods does not resolve the lithologic transition at 500 m on Line
show good agreement with the piezometer data 2. Furthermore, the SRF method cannot provide a
(Fig. 5(e)), although GPR reflections are not always satisfactory resolution of the transition from gravel to
present throughout the line. There are some variations upland on Line 1. Most likely this is due to similarity of
in the results of the individual geophysical approaches in velocity between silt and gravel and the transition from a
the transition zone around 500 m, but both the ERI and three-layer velocity structure to a two-layer velocity
refraction results show the bedrock interface to be structure. In addition, some question remains as to the
deeper than indicated at TH 04–14. TH 04–14 is offset interpretation of Layer 1 that is required for stable
from the line by 10 m and the mismatch may be due to inversions. It may be that a refraction tomography
the offset. Another possibility is that the velocity is algorithm would better resolve the lateral changes in
incorrectly modeled due to the presence of a low velocity velocity than the layered model used in this study.
horizon or too much lateral smoothing of layer 2. Putting all the methods together, along with the
Certainly, there is a large variation in the inverted model borehole information, gives a high resolution interpre-
velocity of layer 3. Finally, we must always be concerned tation of the subsurface (e.g., Fig. 6). Having results
about the possibility of out of plane features distorting from more than one survey reduces the non-uniqueness
the ERI results (Bentley and Gharibi, 2004). of the interpretation. Additionally, in the future multiple
data sets may be simultaneously inverted to improve the
Discussion final images and interpretations.

The three geophysical methods respond to differ- Conclusions


ent petrophysical properties, with each method contrib-
uting different information to the interpretation. The Electrical resistivity, ground penetrating radar and
ERI provides good stratigraphic information and is seismic refraction profiles were repeated over three lines
particularly useful in discriminating between sand and on a Bow River terrace. Electrical resistivity results were
gravel and overbank deposits. It gives the least best for determining changes in sediment types and
ambiguous picture of the transition between the gravel detecting boundaries. However, the ERI smoothness
and the silt of the upland area. However, the bottom constraint blurred the location of the boundaries. The
boundary of the gravel deposits is not precisely located GPR gave the most resolution and showed internal
because of the blurring of the transition caused by the structures that the other methods could not image. On
smoothing constraint. Although this could be improved the other hand, the GPR signal was severely attenuated
by adjusting inversion parameters, it is typical of the in several areas where the surficial sediments became too
method. conductive due to a fine grained component. The seismic
333
Hirsch et al.: A Comparison of Geophysical Methods at a River Terrace Site
refraction inversion provided good reproduction of the References
bedrock interface, but it did not detect changes in the
composition of the surficial sediments. It also required Bentley, L.R., and Gharibi, M., 2004, Two and three-
the introduction of a low velocity surficial layer that dimensional electrical resistivity imaging at a heteroge-
does not appear to represent a change in sediment type. neous remediation site: Geophysics, 69, 674–680.
Jointly interpreting the three methods gives a more Hampson, D., and Russell, B., 1984, First-break interpretation
reliable and less ambiguous interpretation than any using generalized linear inversion: Journal of the
single method. Canadian Society of Exploration Geophysicists, 20,
40–54.
Presently, methods are being developed for the
Hirsch, M., 2004, Geophysical survey on Pine Creek field site,
joint inversion of different types of geophysical data.
Calgary, AB, Canada: M.Sc. thesis, University of
The data used in this study were high quality and Tübingen, Tübingen, Germany.
demonstrated the complimentary nature of the different Loke, M.H., and Barker, R.D., 1996, Rapid least-squares
surveys. Consequently, these profiles are potentially inversion of apparent resistivity pseudo-sections using
good data sets for testing joint inversion algorithms. To quasi-Newton method: Geophysical Prospecting, 44,
that end, the data are available for download at: https:// 131–152.
dspace.ucalgary.ca/handle/1880/44835 Moran, S.R., 1986, Surficial geology of the Calgary urban
area: Bulletin No. 53, Alberta Research Council, Ed-
Acknowledgments monton, Canada.
Sandmeier, K.J., 2004. REFLEXW Vers. 3.5, manual,
The authors wish to thank the City of Calgary for Karlsruhe, Germany.
funding support and access to the site. The work was also Thurber Engineering Ltd., 2004, Geotechnical Investigation
supported by the Natural Sciences and Engineering Research Pine Creek wastewater treatment plant – stage 1,
Council (NSERC) of Canada. Report, Calgary, Canada.

You might also like