See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/251946525
4D GPR tracking of water infiltration in fractured high-porosity limestone
Conference Paper · June 2010
DOI: 10.1109/ICGPR.2010.5550069
CITATIONS
READS
11
276
5 authors, including:
Mark Grasmueck
Pierpaolo Marchesini
University of Miami
Silixa Ltd
66 PUBLICATIONS 1,851 CITATIONS
15 PUBLICATIONS 70 CITATIONS
SEE PROFILE
SEE PROFILE
Gregor P. Eberli
Michael Zeller
University of Miami
Equinor ASA
619 PUBLICATIONS 10,817 CITATIONS
15 PUBLICATIONS 315 CITATIONS
SEE PROFILE
All content following this page was uploaded by Gregor P. Eberli on 27 February 2016.
The user has requested enhancement of the downloaded file.
SEE PROFILE
4D GPR Tracking of Water Infiltration in Fractured
High-Porosity Limestone
Mark Grasmueck, Pierpaolo Marchesini,
Gregor P. Eberli, and Michael Zeller
Division of Marine Geology and Geophysics,
RSMAS – University of Miami
Miami, FL, USA
mgrasmueck@rsmas.miami.edu
Abstract—Three thousand liters of water were infiltrated from a
4 m diameter pond to track flow and transport inside fractured
carbonates with 20-40 % porosity. Sixteen time-lapse 3D Ground
Penetrating Radar (GPR) surveys with repetition intervals
between 2 hrs and 5 days monitored the spreading of the water
bulb in the subsurface. Based on local travel time shifts between
repeated GPR survey pairs, localized changes of volumetric
water content can be related to the processes of wetting,
saturation and drainage. Deformation bands consisting of thin
subvertical sheets of crushed grains reduce the magnitude of
water content changes but enhance flow in sheet parallel
direction. This causes an earlier break through across a
stratigraphic boundary compared to porous limestone without
deformation bands. This experiment shows how time-lapse 3D
GPR or 4D GPR can non-invasively track ongoing flow processes
in rock-volumes of over 100 m3.
4D GPR; Time-Lapse; Hydrology; Fracture; Infiltration; Water
Content Change; Flow; Transport; Vadose Zone
I.
INTRODUCTION
Time-lapse 3D GPR (Ground Penetrating Radar) also
known as 4D GPR is the repeated acquisition of identical
surveys to track dynamic processes in the near surface. As
water content is the main factor controlling the propagation of
electromagnetic waves in geologic materials, 4D GPR can be
used to image and quantify flow and transport in unsaturated
domains. Drilling and insertion of probes in solid rock is
expensive, only renders a limited observation range, and may
disturb natural flow patterns. Non-invasive time-lapse GPR has
been successfully applied to track flow in thin horizontal
fractures [14, 10]. During a pumping test a decrease of GPR
reflection amplitude from a horizontal fracture was caused by
desaturation inside the fracture. This helped define the fracture
drainage pattern [14]. Alternatively saline water was injected
from a borehole into a surface-parallel fracture. Fracture
reflection amplitude changes mapped on a grid of repeated
GPR profiles were attributed to channelization of flow within
the fracture [10]. Both studies were performed in low-porosity
rocks with a single fracture where flow is confined inside the
fracture and no fluid enters the host rock matrix.
Time-lapse GPR studies in porous media have mainly been
reported for soils and sediments e.g. [1, 12, 9, 6, 13]. Time-
Remke L. Van Dam
Dept of Geological Sciences
Michigan State University
East Lansing, MI, USA
lapse GPR data from porous media display amplitude changes
but also significant travel time shifts as water content changes
take place in a large portion of the imaged subsurface volume.
The time shifts are caused by water slowing down the
electromagnetic wave velocity. Comparison of repeated GPR
surveys reveals travel time shifts between corresponding
reflection events complicating the precise localization of zones
affected by water content changes. Subtraction of perfectly
repeated GPR surveys produces fake amplitude anomalies
below hydrologically active zones because also deeper
reflection events are subject to time shifts where no moisture
content changes are happening [1, 12]. While in GPR time
shifts caused by water content changes have been perceived by
many authors as an obstacle to hydrologic characterization,
Time Domain Reflectometry (TDR) uses the delays of
electromagnetic waves to quantify water content in geological
materials by inserting rod shaped probes [8]. The frequency
range of TDR and GPR are very similar. While the use of TDR
in hydrologic studies is widespread and accepted, the use of
GPR time shifts is still rare. Methods have been developed for
mapping the average moisture content in the surface layer
based on the ground wave arrival time [5] or moveout on
multi-offset data [15]. These two methods provide a solution
for 2D mapping of moisture distribution of the top layer (e.g.
topsoil or thaw layer) but do not resolve flow and transport
phenomena at different depths.
The objective of this paper is to extract local time shifts
between precisely repeated high-resolution 3D GPR surveys
for detection of water content changes in three dimensions. The
goal is to relate the GPR time shifts to zones of wetting,
saturation and drainage in fractured rock. The field site for the
experiment is a fractured high-porosity limestone exposed in an
abandoned quarry in central Italy. We recorded a total of
sixteen high-resolution 3D GPR repeat surveys before and after
a 3000 liter ponded water infiltration. The following
paragraphs give more detailed descriptions of the quarry
location and geology, design of the infiltration experiment, and
the GPR system used for rapid and precise data acquisition.
The results show to our knowledge for the first time a 3D
snapshot of ongoing water bulb propagation during a 3hr time
window and how flow is affected by fractures and stratigraphy.
II.
FIELD SITE DESCRIPTION
The Madonna della Mazza quarry located in central Italy
(Figure 1a) is cut into a succession of rudist-derived grain
stones of upper Cretaceous age. The porosity ranges from 20 –
40%. The quarry is 64 m long (east–west), and 50 m wide
(north–south) with walls rising to a maximum height of 12 m.
The box shaped quarry was created by chain- and disk saws
cutting out regular grids of brick sized blocks used for
construction of buildings.
The stratigraphy in the quarry is rather uniform and
bedding dips gently to the NE. The bedding surfaces are often
represented by thin fine-grained layers. The outcropping wall
and floor of the quarry have been studied in detail by structural
geologists [11]. The fracture network consists of faults and
deformation bands. Faults are open fractures with
displacement. Deformation bands are defined by typically 0.2 –
0.5 cm thick subvertical sheets of crushed grains and
compaction without an open fracture. Deformation bands are
typical for porous rocks. The hydrological character of
deformation bands and faults is opposite: Deformation bands
often produce a reduction in permeability and porosity,
whereas open fractures cause a permeability increase. The
ground water table in the quarry is well below the GPR
investigation depth as the nearest creek draining the area is
over hundred meters below the quarry floor.
III.
PRELIMINARY INFILTRATION TESTS AND 4D GPR SITE
SELECTION
To estimate duration of the infiltration, the size of the
temporary pond, and amount of water we performed small
scale infiltration experiments on rock samples and directly on
the quarry floor. These experiments revealed that a natural
(microbial?) coating on all weathered rock surfaces reduced
infiltration to 3 mm/hr water column when compared to fresh
cuts with 25 mm/hr. The conclusion of these tests was that the
infiltration through the untreated quarry floor would take too
long. The whole experiment including acquisition of all repeat
3D GPR surveys had to be completed in 10 days. The partial
removal of the coating with steel brushes increased the
infiltration rate to an acceptable 9 mm/hr. At such an
infiltration rate the temporary pond had to have a diameter of at
least 4 m to infiltrate 3000 liters of water in less than 48 hrs.
The small scale infiltration tests also revealed large (214
mm/hr) sustained infiltration rates into open fractures and
faults that are superficially sealed with a dark about 1 cm thick
organic crust and filled with light brown sand. Grass bushes
and small plants grow along these cracks, with roots extending
into the fractures. The plants serve as a good indicator were
preferential flow paths exist. The entire quarry floor must be
drained with an efficient system of preferential flow paths. If
the infiltration pond is set up on such an open fracture there is a
risk of pond water rapidly moving below the 12 m maximum
GPR imaging depth. With the aid of a 3D GPR survey acquired
in the previous field season the circular pond area was
positioned to include a portion of intact porous limestone, a
zone with a cluster of deformation bands, and part of a fault
(Figure 1b). The open fracture of the fault was sealed at the
surface with cement to prevent direct entry of water from the
pond
IV.
THE 4D GPR MONITORED INFILTRATION EXPERIMENT
Once the location of the 4 m diameter pond was determined
we first removed the superficial coating with steel brushes. A
first pair of dry 3D GPR surveys was acquired to image the
pre-infiltration condition. The data were recorded with a dual
channel DAD K2 control unit (Ingegneria dei Sistemi, Italy)
with two 200 MHz GPR antennae using a 3D rotary laser
positioning system (Figure 2a-c) to achieve centimeter precise
position repeatability between surveys [4]. The antenna
operator pushing and pulling the antenna cart like a
lawnmower at 1m/s was guided by a LED guidance system
mounted on the antenna cart to follow the parallel survey lines.
The line spacing was 0.05 m. The maximum two-way time
recorded was 300 ns. Each 20x20m GPR survey consisting of
401 profiles took between 105-180 minutes to acquire
depending on the walking speed of the operator. The 8 m wide
rim between pond wall and 3D GPR survey edge allows lateral
movement of the infiltrated water and is also necessary for the
3D migration aperture, especially at larger depths. The 0.05 m
line spacing was chosen in order to properly sample the higher
than 200MHz frequency content of the GPR antennae [2, 7].
After completion of the pre-infiltration 3D GPR survey
pair, the 4 m diameter plastic pond wall was sealed to the
quarry floor (Figure 2d). As there was no running water in the
quarry, 1000 liter industrial liquid containers were used to
transport and temporarily store the water 12 m above the
quarry floor. The pressure head was necessary to drive a
precision turbine flow meter, installed at the end of a 50 m
hose, to measure the exact amount of water supplied to the
pond. The pond was filled with a 9 cm water head at the lowest
point. Once the water level dropped to 8 cm, water was added
to reach the 9 cm head again. Infiltration of 1 cm water head
took about 1.5 hours. This procedure was followed to
approximate a constant head in the pond. After 30 hrs the
infiltration of 3000 liters of water was completed. As soon as
there was no more standing water, the pond walls were
removed and the first post-infiltration GPR survey was
acquired. We continued to record a total of seven 3D GPR
survey pairs over the next 5 days. We gradually increased the
interval between the GPR surveys as water movement slows
down due to the expansion of the water bulb
V.
DATA PROCESSING AND WARP TIMESHIFT EXTRACTION
The sixteen 3D GPR surveys are processed with identical
steps and parameters: Data fusion, gridding, dewow (7 ns
window), time zero correction, gain application, background
removal, and 3D migration. Data fusion assigns xyz
coordinates to every trace. The whole survey area is then
regularized onto a 0.05 x 0.025 m grid and populated with the
GPR trace acquired closest to the center of each bin grid cell.
The dewow filter is applied to remove the low frequency noise
and DC offset for each trace. Time zero correction included
first break alignment of all 16 surveys to the same level and
NMO correction to account for the 0.19 m antenna offset
between transmitter and receiver. This last step is in
preparation for the 3D migration which assumes zero-offset
Figure 1. Overview of the Madonna della Mazza quarry located near the village of Pretoro in central Italy. a) A temporary infiltration pond was installed on the
quarry floor at the center of the 20x20m 3D GPR time-lapse survey area. b) Map view of pond location with structural and geological interpretation based on
quarry floor observations and shallow horizontal slices extracted from 3D GPR data.
Figure 2. a) Acquisition of the 20x20m 3D GPR survey with dual 200 MHz antenna cart and 0.05 m line spacing. b) Centimeter precise positioning was
achieved with a RLPS system consisting of 4 spinning laser beacons transmitting infrared and laser pulses to a small detector mounted in the center of the GPR
cart. c) With 20 position updates per second the operator of the cart is guided in real-time by two linear LED arrays along the profile lines. Average walking speed
is 1 m/s. d) The 4 m diameter infiltration pond was filled with a maximum of 9 cm of water.
data. Gain correction is based on the average energy decay
curve computed with raw data from wet and dry surveys. The
same gain is applied to all data volumes to preserve relative
amplitude information. The gained data are 3D phaseshift
migrated in Promax3D (Landmark Graphics, USA) with a
constant velocity of 0.09 m/ns. The average velocity is
determined with interactive hyperbola fitting in ReflexW
(Sandmeier Software, Germany) applied to the best imaged
diffractions on 2D Profiles extracted from the unmigrated 3D
GPR volume. The same velocity is also used for the NMO
correction. Experience has shown that even for the 3D surveys
acquired during the infiltration experiment this dry velocity
still adequately reduces the diffractions so the subsequent
warping mostly correlates reflection events. The warping step
extracts the 3D volume of vertical time shifts necessary to
match two repeat 3D GPR surveys to each other. The 4D Warp
(Landmark Graphics, USA) routines originally developed for
4D seismic processing, correlate small 3D subvolumes and
compute the optimum vertical time shift necessary to match up
corresponding GPR events.
VI.
RESULTS
A. Survey Nomenclature
The pre-infiltration surveys are numbered DRY1 and
DRY2, according to the sequence of acquisition. The fourteen
post-infiltration surveys are labeled with the time after the start
of the infiltration. As no surveys could be recorded during the
30 hr infiltration due to the pond wall and standing water, the
first post-infiltration survey is labeled with WET32hr meaning
that the acquisition of the survey started 32 hrs after the start of
the infiltration. It took two hours to remove the pond wall and
setup the 3D GPR system before the first data trace was
acquired.
B. Repeatabilty
For the ideal case of perfectly repeatable 3D GPR surveys,
subtraction of data volumes should highlight only the
incremental changes due to dynamic processes and suppress
the stationary geological structures. To benchmark repeatability
of our surveying and processing technique we acquired two
“identical” surveys just before the injection experiment began.
As seen in Figure 3 for sample Inline 186 extracted from the
center of the 3D survey, the subtraction of the two migrated
pre-injection surveys contains random noise and remains of
very low amplitude geological reflections. Exceptions are a
dipping shallow event and parts of the first break at the very
top of the profile. These signals do not cancel out during
subtraction. The cause can be either a difference in the survey
track or the signal amplitudes are clipped in one of the repeat
surveys by exceeding the 16 bit dynamic range of the GPR
analog to digital converter. The gray color scale used for
plotting the data is identical for all 3 panels allowing a direct
visual comparison of relative amplitudes. In the good data
quality range (15-100 ns two-way time) the root-mean-square
(RMS) amplitude level of the difference cube is 5–8 times
lower than in the original cubes. This indicates low
repeatability noise and high sensitivity to changes in subsurface
water content [3].
C. Comparison of Post-infiltration 3D GPR Surveys and
Timeshift Visualization
The comparison of pre-infiltration survey (Figure 3 a) with
first post-infiltration survey (Figure 4 a) shows pronounced
amplitude changes and time shifts. The visual differences
between the WET32hr and WET35hr post-infiltration survey
pair are subtle. However the warp time shifts between these
two surveys show a coherent negative anomaly below the
infiltration pond (Label A in Figure 4 c). The core of the
anomaly has a -1.2 ns maximum shift, meaning that the
reflection events of the 35 hr survey are pulled-up by 1.2 ns
relative to the corresponding events on the 32 hr survey. The
pull-up is caused by the saturated water bulb sinking deeper
and draining the rock volume above the bulb. The time shifts
outside the influence of the pond infiltration can be used to
determine the noise level of the time shift data volume. Here,
time shifts are positive and negative and absolute values are
below 0.2 ns. These random time shifts can be caused by slight
deviations in data acquisition from the ideal survey grid,
migration processing noise and warp uncertainty. One small
isolated positive anomaly (Labeled M in Figure 4b and c) is
due to a curved migration artifact. In order to better visualize
the time shift anomaly caused by water content changes, time
shift values ≥(-0.2) ns are set completely transparent in Figure
5. The stepped colorbar creates contours helpful in visualizing
gradients within the time shift data. The reason for this is that
time shift gradients are a direct indicator of local water content
changes [13].
D. Interpretation of the Timeshift Data
The WET32hr∆WET35hr time shift volume provides a
snapshot of the changes in the water bulb which occurred in the
3 hr interval between the two time-lapse surveys. Over this
short period the total amount of water within the survey
volume did not change assuming evaporation is negligible.
Therefore the positive (water content increases) and negative
gradients (water content decreases) are balanced producing
closed shape anomalies. If water enters or exits the 3D GPR
volume the anomalies would extend across the boundaries of
the 3D GPR volume. The time shift anomaly in Figure 5 is
closed but strongly asymmetric both in terms of extent and
gradients. Comparing the opposite sides (see also labels R, S,
P, and Q on Figure 5) of the anomaly yields some interesting
insights about the ongoing water bulb propagation processes:

The vertical extension of the infiltration time shift
anomaly defined by the -0.6 ns contour is more than
double at label R compared to S. The core of the
anomaly with time shifts larger than -0.6 ns is
relatively flat indicating no change in water content.
This is the center of the water bulb which is always
saturated during the 3 hr observation period. The
saturated zone is better developed in the porous host
rock (R) than in (S) where deformation bands are
present. The water content in the saturated core does
not change and the time shifts hover around -0.9 ns
(Figure 4c). The upper slope of the anomaly is the
drain zone as the water bulb sinks deeper. The slope on
underside of the anomaly represents the wetting front,
the transition from initially dry condition to fully
saturated.

Time shifts along vertical column P have steeper
gradients and reach a higher maximum than in column
Q. The higher the maximum time shift the higher the
total amount of water content change. More water is
therefore moving in the porous limestone without
deformation bands. However the vertical extension of
the entire time shift anomaly measured inside the -0.2
ns contour is larger for Q than for P. Apparently the
deformation bands facilitate wider spreading of water
but the local water content changes are smaller. The
deformation bands also help transport water across the
stratigraphic boundary. In P the water bulb has not yet
Figure 3. Inline 186 extracted from the center of the 3D migrated DRY cubes. a,b) For quality control of survey repeatability two pre-infilration 3D GPR
surveys where acquired. c) Subtraction of the two DRY surveys produces low random noise and cancels most geological reflections. Exceptions are a shallow
dipping strong amplitude event and parts of the first break arrivals exceeding the 16 bit dynamic range of the A/D converter. Display gain is identical for all 3
panels.
Figure 4. Inline 186 extracted from the center of the 3D migrated WET cubes. a,b) Comparison of the first 2 repeat surveys acquired just after completion of the
infiltration. c) The local time shifts WET32hr∆WET35hr extracted with warp processing show a strong negative anomaly (labeled A) below the infiltration area.
Label M denotes a positive timeshift anomaly caused by a migration artifact.
entered the deeper layer. In the porous limestone
without deformation bands the thin fine grained
bedding surface presents a flow barrier at this stage of
the infiltration experiment. Here the wetting bulb
anomaly is confined by this stratigraphic boundary.
VII. DISCUSSION AND CONCLUSION
Direct verification of time-lapse GPR results in rocks is
difficult due to lack of accessibility to the subsurface for direct
observation of flow processes. However the excellent exposure
of strata and fractures in the Madonna della Mazza quarry
allows indirect verification and reasoning. After the completion
of the infiltration, the porous limestone between the
deformation bands stayed damp for several hours similar to
wicks transporting moisture by capillary force to the exposed
rock surface. On the other hand, the deformation bands dried
up in the sun within minutes, confirming their low hydraulic
conductivity. Based on these outcrop observations and sample
analyses, deformation bands are thin subvertical sheets with
reduced hydraulic conductivity due to grain crushing and
compaction. It is therefore no surprise that the extent and
gradients of the time shift anomaly caused by the propagation
of the wetting bulb are strongly influenced by the presence of
deformation bands. The GPR time-lapse data show how they
reduce the time shift gradients and therefore the local water
content changes. At the same time the deformation bands cause
a faster and wider spreading of the wetting bulb and facilitate
transport across a stratigraphic boundary. To our knowledge
this is the first time such an ongoing flow process can be
measured and observed with a non-invasive method within a
rock volume of over 100 m3 at an observation interval of 2-3
hrs.
Miami and the National Science Foundation (Grant No.
0323213 and No. 0440322). The University of Miami
acknowledges the support of this research by Landmark
Graphics Corporation via the Landmark University Software
Grant Program.
REFERENCES
[1]
Figure 5. Semi-transparent rendering of contoured local timeshift anomaly
over conventional 3D GPR display. The superimposed interpretation of
fractures and stratigraphy shows how the shape of the waterbulb anomaly is
influenced by the steep deformation bands. The impermeable deformation
bands diffuse watercontent changes and facilitate crossing of the stratigraphic
boundary (marked green). Front face of cube displayed is part of Inline 186.
Please refer to main text for explanations of labels R,S, P and Q.
These preliminary results are encouraging but the 16 timelapse 3D GPR volumes recorded during the experiment still
contain a wealth of information that has yet to be exploited. For
example the role of the fault intersecting the infiltration pond at
the southern boundary is not clearly defined in the WET32hr
and WET35hr survey pair. However a total of 105
combinations of pairs of repeated surveys can be used to
calculate local time shifts and track flow with time increments
between 2 hours and several days. The practical problem is that
time-lapse 3D GPR processing is computationally very
demanding and the average time for a full-density warp
calculation of one survey pair is 8 days on a single core 3GHz
CPU. In order to reduce the compute times by a factor of 10 or
more we are in the process of implementing the warp code on a
parallelized GPU processing unit. Another issue to be resolved
are migration artifacts like the example pointed out in Figure 4.
In this case the associated warp anomaly has the opposite sign
of the wetting bulb response and therefore can be isolated.
Once the local time shifts within high-resolution 3D GPR
time-lapse surveys can be efficiently and reliably extracted, the
next step will be to apply petrophysical transfer functions such
as the Topp equation or a Mixing Model [8] to compute fields
of in situ water content change and water bulb mass balances.
ACKNOWLEDGMENT
This research is supported by the Sponsors of the
Comparative Sedimentology Laboratory at the University of
View publication stats
Birken, R., and R. Versteeg, Use of four-dimensional ground penetrating
radar and advanced visualization methods to determine subsurface fluid
migration, J. Appl. Geophys., 43, 215– 226, 2000.
[2] Grasmueck, M., R. Weger, and H. Horstmeyer, Full-resolution 3-D GPR
imaging, Geophysics, 70, K12-K19, 2005.
[3] Grasmueck, M., and D. A. Viggiano, First look at 4D GPR imaging of
permeable zones around the borehole, in Expanded Abstracts of the 75th
Annual International Meeting: Society of Exploration Geophysics, pp.
2440– 2443, Soc. of Explor. Geophys., Houston, Tex, 2005.
[4] Grasmueck, M. and D.A. Viggiano, Integration of Ground-Penetrating
Radar and Laser Position Sensors for Real-Time 3D Data Fusion, IEEE
Transactions on Geoscience and Remote Sensing, Vol.45, no.1., 130137, 2007
[5] Grote, K., S. Hubbard, and Y. Rubin, Field-scale estimation of
volumetric water content using ground-penetrating radar ground wave
techniques, Water Resour. Res., 39(11), 1321, doi:10.1029/
2003WR002045, 2003.
[6] Loeffler, O., and M. Bano, Ground penetrating radar measurements in a
controlled vadose zone: Influence of the water content. Vadose Zone J.
3:1082–1092, 2004.
[7] Marchesini, P., Grasmueck, M., Impact of Spatial Sampling and
Antenna Polarization on 3D GPR Fracture Detection, Proceedings of
GPR2010.
[8] Robinson, D.A., S.B. Jones, J.M. Wraith, D. Or, and S.P. Friedman, A
review of advances in dielectric and electrical conductivity
measurements in soils using time domain reflectometry. Vadose Zone J.
2:444–475, 2003.
[9] Schmalz B., B. Lennartz, and D. Wachsmuth, Analyses of soil water
content variations and GPR attribute distributions. Journal of Hydrology,
267:217–226, 2002.
[10] Talley, J., G. S. Baker, M. W. Becker, and N. Beyrle, Four dimensional
mapping of tracer channelization in subhorizontal bedrock fractures
using surface ground penetrating radar, Geophys. Res. Lett., 32, L04401,
doi:10.1029/2004GL021974., 2005.
[11] Tondi, E., M. Antonellini, A. Aydin, L. Marchegiani, and G. Cello, The
role of deformation bands, stylolites and sheared stylolites in fault
development in carbonate grainstones of Majella Mountain, Italy.
Journal of Structural Geology, v. 28, p. 376-391. 2006.
[12] Trinks I., D. Wachsmuth, and H. Stümpel. Monitoring water flow in the
unsaturated zone using georadar. First Break 19:679–684, 2001.
[13] Truss S., M. Grasmueck, S. Vega, and D.A. Viggiano. 2007. Imaging
rainfall drainage within the Miami oolithic limestone using highresolution time-lapse ground-penetrating-radar. Water Resources
Research, 43:W03405, doi:10.1029/2005WR004395
[14] Tsoflias, P. G., T. Halihan, and J. M. Sharp, Jr., Monitoring pumping
test response in a fractured aquifer using ground-penetrating radar,
Water Resour. Res., 37, 1221– 1229, 2001.
[15] Wollschläger, U., H. Gerhards, Q. Yu, and K. Roth, Multi-channel
ground-penetrating radar to explore spatial variations in thaw depth and
moisture content in the active layer of a permafrost site, The Cryosphere
Discuss., 3, 919-946, 2009.