oEPA
United States
Environmental Protection
Agency
Monitoring of a Controlled
DNAPL Spill Using a
Prototype Dielectric
Logging Tool
RESEARCH AND DEVELOPMENT
-------
-------
EPA/600/R-06/092
September 2006
www.epa.gov
Monitoring of a Controlled
DNAPL Spill Using a
Prototype Dielectric
Logging Tool
by
Philip J. Brown II1
AldoT. Mazzella2
David L. Wright1
1U.S. Department of the Interior
U.S. Geological Survey
Denver, CO 80225
2U.S. Environmental Protection Agency
Office of Research and Development
National Exposure Research Laboratory
Environmental Sciences Division
Las Vegas, NV89119
Notice: Although this work was reviewed by EPA and approved for publication, it may not necessarily reflect official
Agency policy. Mention of trade names and commercial products does not constitute endorsement or
recommendation for use.
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460 267cmboe
-------
-------
Notice
This research was supported by the United States Geological Survey, Mineral Resources Program, and the
U.S. Environmental Protection Agency under Interagency Agreement DW14937586-01-0. It has been
subjected to the Agency's peer and administrative review and has been approved for publication as an
EPA document. Mention of trade names or commercial products is for identification purposes only and
does not constitute endorsement or recommendation by the U.S. EPA or USGS for use.
in
-------
IV
-------
Acknowledgements
We acknowledge the Lawrence Berkeley National Laboratory and the University of California at
Berkeley for providing the research facility at the Richmond Field Station. A special thanks to Dr. Karl
Ellefsen of the USGS for his assistance with the FDTD modeling, to Dr. Robert Horton of the USGS for
providing the preliminary petrophysical laboratory results, and to John Zimmerman of the U.S. EPA for
the gas chromatograph (GC) tetrachloroethylene (PCE) analysis of the excavated tank samples.
-------
VI
-------
Table of Contents
Notice iii
Acknowledgements v
List of Figures ix
List of Tables xiii
Abstract xv
Introduction 1
The Dielectric Logging Tool 3
The Physical Experiment 7
Data Collection Procedure 9
Data Reduction 10
Numerical Modeling 13
Model Description 13
Numerical Modeling Results 15
Physical Dielectric Logging Results and Interpretation 17
Conclusions 27
References 29
vn
-------
Vlll
-------
List of Figures
Illustration of the USGS prototype slim-hole time-domain dielectric logging system,
not to scale 3
Prototype dielectric logging tool. Panels (b) and (c) show the 7.6-cm spacing between
the Tx and the Rx as it was configured for the RFS experiment. Other tool configurations
replace the metal tubing resulting in 14.6-cm or 29.9-cm Tx and Rx spacings (from
Ellefsenetal.,2004) 4
(a) Photograph of the fiberglass tank cylinder used for the controlled PCE spill
experiment. The tank was located inside two other barriers to prevent any PCE leakage
to the outside environment, (b) Photograph of the completed sand and clay/sand
formations in the tank. North is to the right in the photograph. The west well is at the top
of the photograph, (c) Schematic of the fiberglass tank is not to scale. The dielectric
logging presented in this report was in the west well. For clarity, the location of other
downhole CR probes and seismic wells positioned near the north and south wells are
not shown in the schematic 8
Plot of PCE injection volume verses time for the May 11, 2004 controlled
spill experiment 10
Pre-spill background dielectric logging measurements taken at (a) 104, (b) 57, (c) 5, and
(d) 4 hours before the injection started. Plot (e) is the average of plots (a), (b), (c), and
(d). The color scale represents the recorded voltage, the vertical scale is the logging
depth below the top of the borehole, and the horizontal scale is the recorded trace time.
Note the additional radio noise in plot (a). This is due to these data only having two
stacks per record instead of four stacks 11
Dielectric waveforms at 65-cm depth, at the top of the 3% clay layer, taken at 104, 57, 5,
and 4 hours before the injection started. A maximum variation in the peak amplitude of
3.7% is observed among the four pre-spill measurements. Note the strong radio
interference noise shown in the background waveform taken 104 hours before the spill.
Only two stacks of the data were obtained for this first data set. Four stacks of data were
acquired for subsequent data sets and the noise was considerably reduced 11
(a) Model used for the FDTD simulations in the numerical modeling. This diagram is not
to scale, (b) Close-up of the transmitter part of the model drawn to scale. The model
used for the receiver is identical to the model used for the transmitter but rotated 180
degrees. The electrical properties of the model are listed in Table 2 (after Ellefsen et al.,
2004) 13
Plot of the modeled response for a host material around the borehole having a relative
permittivity of 24, 20, 16, 13, 7, and 4. For comparison purposes, the average
background physical waveform from a depth of 65 cm is plotted as well. Primary
differences in the modeled and received waveform are due to multiple reflected arrivals
not accounted for in the model 15
IX
-------
9 Plot of the four raw pre-spill (background) physical traces from a depth of 65 cm and six
modeled curves. Note how radio noise is more apparent in the 104-hour pre-spill trace.
This is because this trace only stacked individual waveforms two times at each data set
and the others stacked individual waveforms four times. These data suggest that the
dielectric logging tool can detect relative dielectric permittivity changes of about 4 at high
dielectric values of about 20 and changes of about 1.5 at lower dielectric values of about
7. Based upon the BHS model, an overall uncertainty and changes in PCE saturations of
about+/- 5% should be detectable given the repeatability of the system 16
10 (a) Average raw pre-spill data, (b) average raw data four hours into the controlled PCE
spill and (c) the residual of the two images. A color borehole video log taken immediately
after the dielectric log shows approximately a 3-cm thick horizontal layer of DNAPL
against the borehole wall at 62-cm depth (referenced to the top of the borehole). This
location agrees with the bottom of the anomaly shown in the dielectric log and the top of
the 3% clay layer 17
11 Plot of the modeled response for a host material around the borehole having a relative
permittivity of 24, 20, 16, 13, 7, and 4. For comparison purposes, a physical waveform is
plotted from the anomalous zone at 65-cm depth shown in Figure 10b and 10c. The
numerical results suggest a relative dielectric permittivity of 7 for the physical data.
According to the BHS flow chart found in Sneddon et al., 2000, this implies a relative
PCE saturation of approximately 62%. Primary differences in the modeled and received
waveform are due to multiple reflected arrivals not accounted for in the model 18
12 (a) Raw and (b) residual dielectric logging tool data collected 15 hours into the
experiment during injection. The tool response appears to be strongest in the 6% clay
layer 19
13 Raw (a) and residual data (b) collected postinjection approximately 59 hours into the
experiment and approximately 33 hours after injection stopped. Note that the PCE broke
through the 100% clay layer approximately four hours into the experiment. This image
suggests that the PCE has flowed through the 3% and 6% clay layers and that the
dielectric tool has responded to residual PCE left in the pore space. The borehole video
image by this time shows PCE from approximately 60-cm depth to the top of the 100%
clay layer. This correlates with the strongest part of the anomaly shown in the residual
image (b) 20
14 Example of minimum and maximum amplitude traces recorded in the 100% sand zone
33 hours after injection stopped (refer to Figure 13 at approximately 31 and 60 cm).
These traces suggest a relative permittivity range in the sand between 24 and 20.
According to the BHS flow chart found in Sneddon et al., 2000, this implies a relative
residual saturation of PCE between 0 and 12%. These have a saturation uncertainty of
about +/- 5%, based upon the pre-spill baseline measurements 21
15 Amplitude traces taken at 33 hours after the injection stopped at depths of 75 cm (middle
of the 3% clay layer) and at 100 cm (middle of the 6% clay layer). These traces suggest
relative permittivity of about 22 and according to the BHS model implies a residual
saturation of about 7%. This has a saturation uncertainty of about +/- 5 %, based upon
the pre-spill baseline measurements 21
x
-------
16 Magnified three-dimensional plot of the anomalous zone displayed in Figure 10b and
10c (four hours after the spill began). Note the large decrease in slope shown at trace 68
at approximately 2.5 ns. In accordance with the numerical simulation, this may be
attributed to a finite vertical layer (-14 cm deep) of PCE against the borehole wall (refer
to Figure 17) 23
17 Plot of the raw waveform average of trace 68 (Figure 15) recorded four hours into the
spill during injection at a depth of 62 cm (Figure 10b and 10c). Also on this plot is a
modeled trace assuming a vertical layer 14 cm wide. For the vertical layer, sr = 4 and for
the water saturated sand beyond the layer, sr = 24. Note the anomalous decrease in
slope on both traces at approximately 2.5 ns for the physical data and 4 ns for the
modeled. Primary differences in the modeled and received waveform are due to multiple
reflected arrivals not accounted for in the model. These results suggest that the tool may
have detected at least 14 cm into the formation around the borehole 23
18 Photographs of PVC wells after excavation of the tank. The clear PVC east and west
wells are shown in Figure 18a. The dielectric tool was logged in the west well at the right
in Figure 18a. The PVC seismic wells are shown in Figure 18b; the south receiver well is
on the left and the north transmitter well is on the right. The variation in PCE distribution
in the layers is evident from the pattern etched and dyed on the surface of the wells from
the direct contact with the PCE 24
19 Plot of PCE concentration as a percentage of pore space verses depth. Samples were
taken at a number of locations at a particular depth in both the 3% and 6% clay layers
and at the bottom of the tank. The range in values at the 55 and 87 cm depths reflect the
inhomogeneous distribution of the PCE in the formations. Note these depths are from
the surface of the sand; 12 cm should be added for comparison to the dielectric logging
measurements (to the top of the west well) 25
XI
-------
Xll
-------
List of Tables
1 Preliminary petrophysical analysis of the materials in the sand tank is shown in Table 1.
The Unimin sand is 100% water saturated with 1.0 mM CaCI2 solution. The 3% clay is
3% SAz-1 clay (dry wt.) mixed with Unimin sand, saturated with 1.0 mM CaCI2 solution.
The high dielectric value (31.6) for the 6% clay is not fully understood. Time-domain
reflectometry measurements during construction of the formation layers indicated
dielectric values of about 24 for the 6% clay. The difference may be due to the different
frequency spectra of the two measurements. The ABC clay is the same 100% block clay
used in the experiment as received. The PCE residual is the percent of the pore volume
retained in the sample after flushing with 1.0 mM CaCI2 with a 5-cm head. A 36%
porosity value was used for the samples. These PCE residual values are comparable to
the dielectric tool BHS analysis and the GC results of the samples from the tank
excavation 9
2 Electromagnetic properties and dimensions of the model used for the FDTD simulation.
Metal is represented as a perfect conductor resulting in the relative dielectric permittivity
with the relative magnetic permeability having no significance 14
Xlll
-------
XIV
-------
Abstract
The U.S. Geological Survey (USGS) utilized its prototype dielectric logging tool to monitor a controlled
Dense Non-Aqueous Phase Liquid (DNAPL) spill into a large tank located at the University of California
Richmond Field Station (RFS) containing multiple sand and clayey sand layers. To assist in the
interpretation of the logging results, finite-difference time-domain (FDTD) numerical simulations were
performed using a model that approximated the physical experiment. Modeling results agree well with the
physical results and demonstrate qualitatively how the tool responded to the DNAPL spilled in the tank.
Logging results show that the tool successfully monitored DNAPL movement throughout the duration of
the experiment and was sensitive to changes in relative DNAPL saturation. Anomalous zones in the data
correspond to areas where DNAPL was observed in images recorded by a color borehole video camera.
Results presented in this discussion suggest that a quantitative interpretation of the dielectric tool data is
possible given necessary system calibration data.
xv
-------
XVI
-------
Introduction
The dielectric permittivity of a material is a measure of a material's ability to store charge when an
electric field is applied (Sheriff, 1991). Often this property is denoted relative to the dielectric permittivity
of free-space, resulting in a dimensionless quantity known as the relative dielectric permittivity, Sr. In the
petroleum industry, dielectric permittivity measurements can be used to distinguish between water and oil
saturated zones because of the large contrast between the relative dielectric permittivity of the two zones.
The relative dielectric permittivity is approximately 80 for water and approximately 2 for oil. When these
fluids are present in the pore space of a typical sand reservoir, a factor of about 8 may be observed
between the relative dielectric permittivity of the water versus oil saturated zones. If assumptions are
made on the dielectric mixing of the components (water, oil, and sand), models can be used to determine
the amount of oil in a formation (Wright and Nelson, 1993; Abraham, 1999a, 1999b).
Non-Aqueous Phase Liquids (NAPLs) comprise one of the largest classes of ground-water contaminants
in the country (Lucius et al., 1992). The irregular and unpredictable migration patterns of the chemicals
make mapping and monitoring the extent of contamination difficult without point measurements
(Sneddon et al., 2000). Previous studies have shown that dielectric measurements can detect and monitor
Light Non-Aqueous Phase Liquids (LNPLs) such as gasoline and Dense Non-Aqueous Phase Liquids
(DNAPLs) such as tetrachloroethylene (PCE) (Wright et al., 1993; Wright et al., 1998; Greenhouse et al.,
1993; Abraham, 1999a, 1999b; Sneddon et al., 2000; and Ellefsen, 2004). Previous controlled spill
experiments have shown that surface and cross-borehole ground penetrating radar (GPR) surveys can also
detect subsurface NAPLs (Greenhouse et al., 1993). The velocity of the radar wave has a strong
dependency upon the dielectric permittivity (Sears et al., 1987). An independent measurement of
dielectric permittivity can provide a better interpretation of GPR data.
Typical oil field dielectric logging tools are much larger than 5 cm (2-in), have a vertical resolution of
about 20 cm (8-in), and are usually operated at a single frequency. This prevents their use for
investigations in the shallow, small diameter wells used for environmental monitoring. For this reason,
the USGS developed a prototype dielectric logging tool for application to environmental problems. The
USGS tool is different from its oil field counterparts in several ways. The USGS prototype dielectric tool
has a 4.45-cm (1.75-in) outer diameter (OD) allowing it to be deployed in the 5.08-cm (2-in) slim-holes
often used for ground-water sampling and monitoring. It has a simple modular design allowing for easy
decontamination. Most importantly, it is a time-domain tool that records data containing broad frequency
content. Theoretically, the broad frequency content allows for the study of frequency dependent dielectric
permittivity. Prior to this study, the USGS tool has been characterized in the laboratory (Wright and
Nelson, 1993; Abraham 1999a, 1999b) and field-tested at the South Carolina Savannah River Site
(Wright et al., 1998). The purpose of the current research is to evaluate the detection limits and
effectiveness of the tool for monitoring DNAPL migration in the subsurface.
In May 2004, a controlled spill of the DNAPL PCE was conducted in a water-saturated tank containing
multiple sand and clayey sand layers at the University of California Richmond Field Station (RFS). The
experimental goal was to evaluate the detection limits of various surface and downhole geophysical
techniques for monitoring the movement and location of the PCE during the controlled spill. This would
allow a direct comparison of the sensitivity of the different methods and subsequent possible joint
inversion (data fusion) of multiple methods for better interpretations. Techniques evaluated included
surface ground penetrating radar, cross-borehole radar, directional-borehole radar, cross-borehole high
frequency seismic measurements, borehole and cross-borehole complex resistivity (CR), borehole self-
potential (SP), surface high-frequency frequency-domain electromagnetic soundings, surface very-early-
-------
time time-domain electromagnetic measurements, borehole video, and dielectric logging. This discussion
focuses on the dielectric logging results.
-------
The Dielectric Logging Tool
The dielectric logging tool consists of eight main components (Figure 1). These are a fiberglass, brass-
capped sonde containing the transmitter (Tx) and receiver (Rx), a sheathed 36.58 m (120 ft) long bundle
of three low-loss coaxial cables, a draw-works containing the transmitter pulser, a draw-works motor
controller, an optical encoder mounted to a sheave wheel, a Tektronix TDS 820 two channel digitizing
oscilloscope, and a personal computer (PC).
Sheave wheel and optical depth encoder
PC used for data storage
and system control
Logging
Cable
Draw-works
Motor Controller
Tektronix Digitizing
Oscilloscope
Draw-works and
Transmitter Pulser
Figure 1. Illustration of the USGS prototype slim-hole time-domain dielectric logging system, not to scale.
-------
Figure 2 shows images and a cross section of the dielectric tool. The dielectric tool's transmitter and
receiver are circular plate voltage gaps that act as radial waveguides. The Tx is designed to transmit a
short duration pulse, about 6 nanoseconds (ns) wide, with a very fast nanosecond rise-time into the
medium surrounding the tool. A pulsed time-domain design was chosen rather than a frequency-domain
design in order to rapidly obtain broad-spectrum data. A voltage gap Tx was chosen because it is difficult
to drive a current loop at high frequencies since the impedance of a current loop increases with increasing
frequency (Kraus, 1950).
(a) Exterior View
Logging
cable
Cablehead
Top
cap
Fiberglass
sheath
Bottom
cap
(b) View with Fiberglass
Sheath, Bottom Cap,
and Top Cap Removed
cq
LT>
co
Coaxial cables
from transmitter
and receiver
Metal tubing
(39.4 crn long)
Transmitter
Metal tubing
(7.6cm long)
Receiver
Metal tubing
(39.4 cm long)
(c) Cross Section of Transmitter
and Receiver
Coaxial Cables
Rx
-— Tx monitor
— Metal tubing
— Metal disk
— Plastic Ring
—Metal disk
Air
L:JUU
— Metal tubing
Air
Metal disk
Plastic Ring
Metal disk
— Metal tubing
Figure 2. Prototype dielectric logging tool. Panels (b) and (c) show the 7.6-cm spacing between the Tx and the Rx as it was
configured for the RFS experiment. Other tool configurations replace the metal tubing resulting in 14.6-cm or 29.9-cm Tx
and Rx spacings (from Ellefsen et al., 2004).
Both the Tx and Rx consist of two metal disks and a plastic ring used to keep the disks apart and parallel
to each other. Attached to each of these disks is a coaxial cable, one used to drive the Tx and another to
receive the signal from the Rx. The Tx also has an additional attached coaxial cable that is used to
monitor the transmitter driving voltage. All three coaxial cables have a characteristic impedance of 50
ohms. Between the Tx and the Rx is a metal tube that can be one of three different lengths. This allows
-------
for a transmitter and receiver spacing of 7.6, 14.6, or 29.9 cm (measured from the midpoints of the Tx and
Rx). The RFS spill experiment was logged with the tool using the 7.6-cm Tx to Rx spacing. For a more
comprehensive discussion of the dielectric tool design, refer to Abraham, 1999a.
During data collection, a voltage pulse is generated by the electronics contained within the system's draw-
works at the ground surface. This pulse is transmitted via a coaxial cable contained within the sheathed
logging cable and then along the coaxial cable contained within the sonde of the tool. When the pulse
reaches the transmitter, it is radiated from the voltage gap. This pulse then propagates along the tool as
well as into the borehole and strata radially out from the borehole. The direct radiated field and reflections
are picked up at the Rx producing a voltage pulse that is transmitted to the surface via the Rx coaxial
cable in the tool and the logging cable. At the surface, the analog signal is digitized by the 16 bit
Tektronix TDS 820 oscilloscope and recorded by a PC. Recording of the received signal starts
approximately 1.5-ns before Tx turn-on and lasts until approximately 21.5 ns after Tx turn-off. Each trace
consists of 2,500 samples with a sample interval of 0.01 ns, giving atrace duration of 25 ns. For amore
complete discussion of the wave propagation modes set up by the dielectric logging tool, refer to Ellefsen
et al., 2004.
-------
-------
The Physical Experiment
The current experiment was conducted in a nonmetallic, fiberglass tank that was housed in a special
building at the University of California Richmond Field Station that had been designed to minimize
electromagnetic coupling and electrical noise interference for geophysical research. The tank is about 2.4
meters in diameter and 1.8 meters in depth and was constructed with PCE-resistant resin (Figure 3). Two
other barriers were in place outside the tank to prevent any PCE leakage to the outside environment. A
multilayer formation consisting of sand and clayey-sand layers was engineered and constructed in the
tank. It was necessary to consider various factors in the construction of the formation. These included a
balance between the presence of clay needed for the induced polarization response for the CR method and
high formation resistivity needed for the GPR method. In addition, the layers had to have good hydraulic
conductivity to allow the PCE to penetrate into the formation within a reasonable time period of a few
days and the ability to effectively allow the removal of air in the formation pore space for good seismic
signals. The final configuration was developed after a number of experimental laboratory studies were
conducted.
The layers in the tank were constructed with a washed, well-sorted medium grain sand, Unimin
20/30. This is composed of over 99% SO2. Laboratory measurements showed that the packed sand had a
porosity of about 36%. A number of time-domain reflectometer (TDR) measurements indicated a range in
porosity from 35 to 39% at different locations about 7 cm apart in a water saturated packed sand cylinder.
A calcium montmorillonite clay, SAz-1, obtained from the National Clay Repository at Purdue University
was used for the development of the clay/sand layers. The central portion of the tank, with a diameter of
about 1 meter, was the primary area for the experiment. The formation in this area consisted of three
layers, a 53-cm thick upper sand layer, a 20-cm thick 3% clay/sand layer overlying a 30-cm thick 6%
clay/sand layer (dry weights). Beneath this, a 3-cm thick solid clay barrier was placed to prevent the PCE
from migrating directly to the bottom of the tank. Below the solid clay barrier, a sand layer about 47 cm
thick extended to the bottom of the tank. The horizontal extent of these layers was confined with a thin
vertical plexiglass sheet that formed a cylinder around the volume that was about 70 cm high and 100 cm
in diameter. The sheet was embedded about 1 cm into the solid clay barrier. The formations were fully
saturated during construction with a 0.001 M calcium chloride distilled water solution. The conductivity
of the water was about 230 microseimens/cm. Preliminary results of a petrophysical analysis of the
materials in the tank can be found in Table 1. A number of clear polyvinyl chloride (PVC) and clear
acrylic closed wells and probes were placed in the formation during construction to accommodate the
different geophysical methods (Figure 3). The east and west cased wells ran from approximately 12 cm
above the sand to the bottom of the tank and had an inner diameter (ID) of 7.62 cm (3 in). The north and
south observation wells ran from about 27 cm above the sand and had an ID of 5.08 cm (2 in). The two
larger ID east and west wells were continuously logged with geophysical instruments and all of the wells
were logged using borehole video equipment. The dielectric logging presented in this paper was
conducted in the west well.
After construction, the formation was monitored and allowed to stabilize for about two months. About 10
days before the injection experiment began, monitoring with the complex resistivity and seismic
measurements established the formation stability. About five days before the injection began, monitoring
with all the geophysical methods established pre-spill background baselines and noise levels.
In May 2004, the injection experiment began. Eighty-five (85) liters of PCE was injected over a period of
26 hours into the subsurface at a fairly constant rate of about 3.7 liters/hour. In order to track the
migration of the PCE with downhole video and subsequent excavation, a dye, Oil Red O, was added to
-------
(a)
(b)
jjj
n
>
it 'i
0)
T3
c
O
(0
en
w/
ro
0)
'x
o
0.
j
'!
'
' '
( i
* t
S i!
M
1 J
i
i
1 100% Sand
i
"^
.
n it l> ii
f 3%
g 6%
I
Clay
i
i
Clay
i
!.
1.
.
i
TJ
5
x
^Q
^^
^^ 1
CO
O
5'
Q.
Q
M
^
i
i 100% Sand
j
!|
Ii
'l
aaei* '
Side View
Plan View
(c)
Figure 3. (a) Photograph of the fiberglass tank cylinder used for the controlled PCE spill experiment. The tank was
located inside two other barriers to prevent any PCE leakage to the outside environment, (b) Photograph of the
completed sand and clay/sand formations in the tank. North is to the right in the photograph. The west well is at the top
of the photograph, (c) Schematic of the fiberglass tank is not to scale. The dielectric logging presented in this report was
in the west well. For clarity, the location of other downhole CR probes and seismic wells positioned near the north and
south wells are not shown in the schematic.
the PCE at a concentration of 3.0 g/L. The PCE was injected through a tube in the center of the tank at a
depth of about 6 cm below the surface. The injection tube extended 24 cm above the surface. This
allowed a sufficient head to develop to overcome surface tension effects and the PCE flowed into
formation. A plot of the PCE injected volume as a function of time is shown in Figure 4. Before the
injection began, the water table had been established at about 1 cm above the surface of the sand. In order
to maintain a constant fluid level in the tank, the water displaced by the PCE during injection was
8
-------
SAMPLE
Unimin sand
3% clay
6% clay
ABC
Two-
electrode
resistivity
(ohm-m)
100Hz
178
75.8
61.1
-
Two-electrode
resistivity (ohm-m)
10MHz
178
46.7
34.2
-
8r
10 MHz
23.9
23.4
31.4
7.56
PCE residual
percentage of
pore volume
%
16.4
13.3
10.6
-
Table 1. Preliminary petrophysical analysis of the materials in the sand tank is shown in Table 1. The Unimin sand is
100% water saturated with 1.0 mM CaCl2 solution. The 3% clay is 3% SAz-1 clay (dry wt.) mixed with Unimin sand,
saturated with 1.0 mM CaCl2 solution. The high dielectric value (31.6) for the 6% clay is not fully understood. Time-
domain reflectometry measurements during construction of the formation layers indicated dielectric values of about 24
for the 6% clay. The difference may be due to the different frequency spectra of the two measurements. The ABC clay is
the same 100% block clay used in the experiment as received. The PCE residual is the percent of the pore volume
retained in the sample after flushing with 1.0 mM CaCl2 with a 5-cm head. A 36% porosity value was used for the
samples. These PCE residual values are comparable to the dielectric tool BHS analysis and the GC results of the samples
from the tank excavation.
siphoned off to an external reservoir. Because of the limited space and interference effects between
different systems, a schedule was established when data with the different geophysical methods would be
obtained.
Pre-spill data were obtained with the borehole dielectric tool at 104, 57, 5, and 4 hours before the
injection began. Post spill data were collected about 4, 15, 59, 79, 90, 112, 137, and 158 hours after the
injection started. The first two measurements (at 4 and 15 hours) were obtained during the injection
period and the last six were after the injection stopped. These were the start times for the logging; it took
about 30 minutes to acquire a data set.
Data Collection Procedure
All dielectric logging was performed in the west well of the tank (Figure 3c). For all tests, the
experimental setup remained the same with all electronics and required support equipment in the same
position relative to the tool. Logging speed was maintained by the computer-interfaced electronic draw-
works and held constant at approximately 1 mm per second. Zero depth was set for the optical encoder
when the Rx was at the same level as the top of casing. The tool was run to the bottom of the hole and
stopped. The depth level was then checked to ensure it was consistent with other logging runs. For any
one logging run, the tool was never more than 5 mm off in depth from any other logging run. Data were
recorded while the tool was being pulled out of the hole and the computer automatically stopped the
system when the tool reached zero depth. The logging sonde was kept in the middle of the hole by
polystyrene centralizers affixed to the head, tail, and midpoint of the tool. All data were collected with the
tool running in continuous data acquisition mode as opposed to a start and stop mode. To reduce radio
interference in the received signal from various sources at the site, each final record is an average of
two repeat logging runs of four-stack data.
Before the spill began, four background measurements were logged in the tank over a 2-day period. Each
final background measurement was the average of two logging runs of four-stack data with the exception
of the first background measurement that was only two-stack data. This was done to get a base level to
compare to the monitoring measurements made during and after the spill as well as to demonstrate the
-------
PCE injection
Injection starts May 1 1 , 2004 at 21 :35
no
san
7O
JR «n
J{ DU
*"' 5O
u ou
3 40
1
111 ^o
m JU
* 20
m
n <
X
S
y
/
f
/
/
r
^^
/^
A
^~-*^
/>
0 5 10 15 20 25 30
Time (hours)
Figure 4. Plot of PCE injection volume verses time for the May 11, 2004 controlled spill experiment.
reproducibility of the dielectric tool data. These data are plotted in Figures 5 and 6. The depths indicated
are from the top of the well casing which was about 12 cm above the surface of the sand. The data in
Figure 5 are shown as color amplitude plots for the recorded trace time at each depth for each of the four
background measurements. An average of these four data sets was used as the pre-spill background
baseline and is shown on the right of Figure 5. The data plotted in Figure 6 are the voltage amplitude
verses recorded trace time for each of the four background measurements at a depth of 65 cm (53 cm
below the surface); this is at the top of the 3% clay/sand layer. Note the noise shown in the oldest
background measurement (Figure 5 and 6; 105 hours pre-spill); this is believed to come from a local AM
radio transmission tower located about 1 mile from the site. This noise is larger on this first data set
because only two stacks were acquired. All subsequent data represent four stacks resulting in the noise
being considerably reduced. A maximum variation of 3.7% is observed in the peak amplitude among the
four different background measurements acquired over the 2-day pre-spill period.
Data Reduction
Once data are recorded, very little data processing is required. The time base for the system is stable so
that no traces need to be adjusted temporally to account for late or early triggering of the Tx or Rx. From
trace to trace, there is a small static offset in the record before turn-on. To correct for this offset, the
average voltage for the first 70 samples of the record are subtracted from the entire record normalizing the
pre turn-on part of the trace to approximately zero volts. After the traces have been static shifted,
correction factors are applied to account for attenuators that exist at the draw-works. No correction is
made for the attenuation of the received signal through the 36.58 m of logging cable. Data from four
10
-------
repeat logging runs were then averaged to further reduce the radio noise that was seen in the data. After
the data were averaged, the data were gridded using the minimum curvature method and a grid cell size of
2 mm. Once the data were gridded, the average background data were subtracted from the data recorded
after the PCE injection began. This produces residual data where changes occurring due to the PCE are
more apparent.
Time (ns)
Time (ns)
Time (ns)
Time (ns)
Time (ns)
-0.25 -0.2G -0.17 -0.14
Amplitude (Volts)
Figure 5. Pre-spill background dielectric logging measurements taken at (a) 104, (b) 57, (c) 5, and (d) 4 hours before the
injection started. Plot (e) is the average of plots (a), (b), (c), and (d). The color scale represents the recorded voltage, the
vertical scale is the logging depth below the top of the borehole, and the horizontal scale is the recorded trace time. Note
the additional radio noise in plot (a). This is due to these data only having two stacks per record instead of four stacks.
1.50
1.00
O 0-50
0.00 *
•0.50 -
104 Hours Pne-splll
57 Hours Pre-spill
• 5 Hours Pre-spill
• 4 Hours Pre-spill
10
15
20
Time (ns)
Figure 6. Dielectric waveforms at 65-cm depth, at the top of the 3% clay layer, taken at 104, 57, 5, and 4 hours before the
injection started. A maximum variation in the peak amplitude of 3.7% is observed among the four pre-spill
measurements. Note the strong radio interference noise shown in the background waveform taken 104 hours before the
spill. Only two stacks of the data were obtained for this first data set. Four stacks of data were acquired for subsequent
data sets and the noise was considerably reduced.
11
-------
12
-------
Numerical Modeling
Model Description
To understand the influence of changes in DNAPL saturation on tool response, numerical simulations
were performed. This modeling was accomplished using a finite-difference time-domain (FDTD) method
that simulates the propagation of electromagnetic waves in time. To optimize the modeling speed, the
numerical simulation takes advantage of tool and borehole symmetry by utilizing a cylindrical coordinate
system where the z-axis is aligned with the borehole axis. Given this symmetry, computations only have
to be performed in the r and z-directions, greatly reducing the computation time. Each model took
approximately 5 hours to run on a standard PC using a 1.3-GHz AMD processor. A detailed description
of the FDTD numerical simulator can be found in Ellefsen et al., 2004.
Figure 7 is a diagram of the numerical model. The physical properties used in the modeling can be found
in Table 2. The upper and lower bounds used for the relative dielectric permittivity of the sand, water, and
PCE mixtures are based upon research done by Sneddon et al., 2000. From Table 2 and Figure 7, it is
apparent that the model is a simplification of the physical dielectric logging tool. It omits the monitoring
coaxial cable found at the transmitter, the top and bottom brass caps on either end of the sonde, the cable
head, and the logging cable as well as the polystyrene centralizers. Additionally, the tool is assumed to be
as long as the entire borehole. Nonetheless, the FDTD model still represents important characteristics of
the dielectric logging tool. It can be used to qualitatively understand the tool's early-time physical
response to different borehole environments without the complications of reflections from the ends of the
tool.
Annular :
Space .
•Transmitter
' Receiver—
Tx Coaxial Cable
— Inner Conductor
I— Dielectric Sheath
i— Outer Conductor
(a)
(b)
1 cm
Figure 7. (a) Model used for the FDTD simulations in the numerical modeling. This diagram is not to scale, (b) Close-up
of the transmitter part of the model drawn to scale. The model used for the receiver is identical to the model used for the
transmitter but rotated 180 degrees. The electrical properties of the model are listed in Table 2 (after Ellefsen et al., 2004).
13
-------
Table 2. Electromagnetic properties and dimensions of the model used for the FDTD simulation. Metal is represented as i
perfect conductor resulting in the relative dielectric permittivity with the relative magnetic permeability having no
significance.
Item
Inner conductor, Tx and Rx
coaxial cables
Plastic Rings in Tx and Rx
gap
Outer conductor, Tx and Rx
coaxial cable
Metal cylinder above Tx and
below Rx
Metal disk above Tx and
below Rx
Air in Tx and Rx gap
Metal disk below Tx and
above Rx
Metal cylinder below Tx and
above Rx
Metal cylinder
Annular Space
Fiberglass Tool Casing
Borehole Casing
Formation
Material
Metal
Plastic
Metal
Metal
Metal
Air
Metal
Metal
Metal
Air
Plastic
Plastic
100%
water
saturated
sand or
PCE water
saturated
sand
Relative
Dielectric
Permittivity
3.6
1
...
1
4.5
4
24 for wet
sand; 20,16,
13,10,8,6
and 4 for
PCE and
water
saturated
sand
Relative
Magnetic
Permeability
1
1
...
1
1
1
1
Electrical
Conductivity
(Sim)
00
0
OO
OO
OO
0
OO
OO
OO
0
0
0
0
Dimension (mm)
Outer radius: 0.5
Inner radius: 12.6
Outer radius: 20.6
Inner radius: 1.5
Outer radius: 2.0
Inner radius: 2.0
Outer radius: 20.6
Inner radius: 2.0
Outer radius: 20.6
Axial length: 5.0
Inner radius: 0.5
Outer radius: 20.6
Axial length: 6.4
Outer radius: 20.6
Axial length: 5.0
Outer radius: 20.6
Outer radius: 20.6
Inner radius: 21.9
Outer radius: 38.1
Inner radius: 20.6
Outer radius: 21.9
Inner radius: 38.1
Outer radius: 40.4
Inner radius: 25.4
Outer radius: 1 ,500
14
-------
Numerical Modeling Results
Figure 8 shows the calculated voltage response at the end of the coaxial cable connected to the receiver.
The different curves represent the modeled response to formations having different relative dielectric
permittivity. Figure 8 illustrates that with decreasing relative dielectric permittivity of the formation, the
received voltage decreases in amplitude.
For comparison purposes, Figure 8 also displays an average of the four traces recorded in the sand portion
of the RFS tank at a depth of 65 cm before the spill began. In order to match amplitudes with the physical
data, the modeled data had to be scaled by a factor of 0.54. This is because the model simulates voltage at
the receiver and the tool records voltage at the end of a 36.58-m cable. Such an amplitude loss in voltage
over the length of the cable agrees with tests performed by Abraham, 1999a. To be consistent with the
petrophysical analysis of the wet sand (Table 1), a scaling factor of 0.54 was chosen because it
normalized the amplitude of the data to the modeled curve where Sr = 24. A comparison of the variation
in the pre-spill data (3.7% maximum variation in the peak amplitude) with the modeling results is shown
in Figure 9. These results suggest that relative dielectric permittivity changes of about 4 can be
determined at high dielectric values of about 20 and changes of about 1.5 can be determined at lower
dielectric values of about 7.
Some characteristics of the physical data are not seen in the numerical data. The large differences are due
to multiple reflections constructively and destructively interfering with each other. These reflections
arrive at the receiver from the ends of the tool and the ends of the logging cable. These arrivals are not
observed in the model for two reasons. The modeled tool is significantly longer than the physical tool so
that these arrivals would theoretically arrive at a different time in the record and the edges of the model
are absorbing boundaries that have practically no reflection, meaning that the model emulates a grid
extending to infinity.
1.50
1.00
0.50
0.00
-0.50
10
15
20
Time (ns)
Figure 8. Plot of the modeled response for a host material around the borehole having a relative permittivity of 24, 20, 16,
13, 7, and 4. For comparison purposes, the average background physical waveform from a depth of 65 cm is plotted as
well. Primary differences in the modeled and received waveform are due to multiple reflected arrivals not accounted for
in the model.
15
-------
1.50 •
CO
20
Time (ns)
Figure 9. Plot of the four raw pre-spill (background) physical traces from a depth of 65 cm and six modeled curves. Note
how radio noise is more apparent in the 104-hour pre-spill trace. This is because this trace only stacked individual
waveforms two times at each data set and the others stacked individual waveforms four times. These data suggest that the
dielectric logging tool can detect relative dielectric permittivity changes of about 4 at high dielectric values of about 20
and changes of about 1.5 at lower dielectric values of about 7. Based upon the BHS model, an overall uncertainty and
changes in PCE saturations of about +/- 5% should be detectable given the repeatability of the system.
16
-------
Physical Dielectric Logging Results and Interpretation
The average of all four background measurements can be found in Figure lOa. The horizontal white lines
on this plot indicate the boundaries between various formation zones in the sand tank. Approximately four
hours into the spill, two logging runs were made back to back with the dielectric tool and one run was
made using a color borehole video logger. The dielectric data in Figure lOb show an anomaly that
encompasses the 63-cm depth where the red-dyed PCE is first seen breaking through to the borehole wall
on the video log. The plot in Figure lOc, which shows the difference between the data of Figures lOb and
lOa, emphasizes this anomaly further. The thickness of the anomaly in the dielectric log is about 17 cm
starting at roughly 48-cm depth and ending approximately at the contact between the 100% sand and 3%
clay layers at 65-cm depth. The apparent thickness of the anomaly is probably greater than the actual
thickness because of the finite Tx-Rx spacing. As predicted by the computer models, most of the traces in
the anomalous zone decrease in amplitude. This is illustrated in Figure 11, which shows the modeling
results compared to a trace recorded in the anomalous zone of Figures lOb and lOc. Comparing the
physical and modeled data in Figures 8 and 11 suggest an average Sr = 7 for the formation around the
borehole. Using the Bruggerman-Hanai-Sen (BHS) mixing formula flow chart presented in Sneddon et
al., 2000, this permittivity corresponds to a PCE concentration of 62% of the pore space. Based upon the
uncertainty in the four pre-spill dielectric measurements (3.7%) and the BHS mixing formula, the PCE
saturation can be determined with an overall uncertainty of about +/- 5% from the dielectric logging
measurements.
(c) Residual Rx Data
. : !?-!/
-0.4993
-0.5438
Amplitude
volts
Time (ns)
Amplitude
volts
Time (ns)
Amplitude
volts
Time (ns)
Figure 10. (a) Average raw pre-spill data, (b) average raw data four hours into the controlled PCE spill and (c) the
residual of the two images. A color borehole video log taken immediately after the dielectric log shows approximately a 3-
cm thick horizontal layer of DNAPL against the borehole wall at 62-cm depth (referenced to the top of the borehole). This
location agrees with the bottom of the anomaly shown in the dielectric log and the top of the 3% clay layer.
17
-------
1.50
1.00
0.00
-0.50
-1 0
10
15
Time (ns)
Figure 11. Plot of the modeled response for a host material around the borehole having a relative permittivity of 24, 20,
16, 13, 7, and 4. For comparison purposes, a physical waveform is plotted from the anomalous zone at 65-cm depth shown
in Figure lOb and lOc. The numerical results suggest a relative dielectric permittivity of 7 for the physical data.
According to the BHS flow chart found in Sneddon et al., 2000, this implies a relative PCE saturation of approximately
62%. Primary differences in the modeled and received waveform are due to multiple reflected arrivals not accounted for
in the model.
Additional dielectric logs were recorded approximately 15, 59, 84, 90, 112, 137, and 158 hours after PCE
injection started. The video logs confirm that the tool did track the progress of the PCE as it passed
through all the layers of the tank and eventually reached the tank bottom. It can be seen that during
injection, approximately four hours into the spill, the PCE is pooling on top of the 3% clay layer (Figure
lOc). Approximately 15 hours into the experiment, PCE has made its way down to the 100% clay layer
and the tool response to the PCE is strongest in the 6% clay layer (Figure 12a and 12b). Approximately
10 hours into the experiment, the video log showed PCE from approximately 60-cm depth to the top of
the 100% clay layer (115-cm depth from the top of the well casing). This corresponds to the higher
amplitude parts of the residual signal displayed in Figure 12a and Figure 12b. Figure 13 shows the data
taken at 59 hours into the experiment, 33 hours after the spill stopped. A decrease is observed in the
residual amplitudes between Figure 12b and Figure 13b; it appears that the PCE concentration close to the
borehole in Figure 13 (at 59 hours) has decreased from the PCE concentration shown in Figure 12 taken
approximately two days before.
This makes sense given that PCE was observed in the video logs below the 100% clay barrier
approximately 4.5 hours into the experiment, meaning the PCE had a pathway to migrate below the 100%
clay barrier during the injection period and then drain after the injection of the PCE stopped. It also agrees
with the preliminary petrophysical analysis that shows that a PCE residual stays in the sand even after
flushing with water (Table 1). Comparison of maximum and minimum amplitude traces recorded in the
sand at depths of 31 and 60 cm in Figure 13a to the computer models suggests a relative dielectric
permittivity that ranges between 24 and 20, respectively (Figure 14). According to the BHS flow chart in
Sneddon et al., 2000, such a permittivity range implies a relative residual PCE saturation of between 0
and 12% of the pore space of the 100% sand. A similar analysis was conducted on the dielectric data at
the 75-cm depth (middle of the 3% clay layer) and at the 100-cm depth (middle of the 6% clay layer).
These results are shown in Figure 15. Based upon the peak amplitude, dielectric values of about 22 are
derived. Ignoring the presence of clay for the modeling and BHS mixing-law, bulk PCE residual
saturations of about 7% were derived in these layers. Based upon the four pre-spill dielectric
measurements, these PCE saturation values have an uncertainty of about +/- 5%. These data confirm that
18
-------
1.6359
1.5914
1.5470
1.5025
1.4580
1.4135
1.3690
1.3245
1.2801
1.2356
1.1911
1.1466
1.1021
1.0576
1.0132
0.9687
0.9242
0.8797
0.8352
0.7907
0.7463
0.7018
0.6573
0.6128
0.5683
0.5238
0.4793
0.4349
0.3904
0.3459
0.3014
0.2569
0.2124
0.1680
0.1235
0.0790
0.0345
-0.0100
-0.0545
-0.0989
-0.1434
-0.1879
-0.2324
-0.2769
-0.3214
-0.3658
-0.4103
-0.4548
-0.4993
-0.5438
Amplitude
volts
(a) Raw Rx Data
0.1113
0.0495
0.0395
0.0331
0.0278
0.0237
0.0210
0.0184
0.0160
0.0140
0.0125
0.0112
0.0100
0.0089
0.0080
0.0071
0.0063
0.0057
0.0049
0.0042
0.0035
0.0029
0.0024
0.0018
0.0012
0.0006
-0.0001
-0.0007
-0.0014
-0.0019
-0.0026
-0.0032
-0.0039
-0.0046
-0.0054
-0.0063
-0.0071
-0.0080
-0.0088
-0.0099
-0.0113
-0.0128
-0.0147
-0.0168
-0.0200
-0.0262
-0.0377
-0.0505
-0.0720
-0.0999
(b) Residual Rx Data
4 9 14
Time {ns}
Amplitude
volts
Figure 12. (a) Raw and (b) residual dielectric logging tool data collected 15 hours into the experiment during injection.
The tool response appears to be strongest in the 6% clay layer.
19
-------
1.6359
1.5914
1.5470
1.5025
1.4580
1.4135
1.3690
1.3245
1.2801
1.2356
1.1911
1.1466
1.1021
1.0576
1.0132
0.9687
0.9242
0.8797
0.8352
0.7907
0.7463
0.7018
0.6573
0.6128
0.5683
0.5238
0.4793
0.4349
0.3904
0.3459
0.3014
0.2569
0.2124
0.1680
0.1235
0.0790
0.0345
-0.0100
-0.0545
-0.0989
-0.1434
-0.1879
-0.2324
-0.2769
-0.3214
-0.3658
-0.4103
-0.4548
-0.4993
-0.5438
(3) Raw Rx Data
I
Amplitude
volts
1 _•
-14 B M '9 24
0.1113
0.0495
0.0395
0.0331
0.0278
0.0237
0.0210
0.0184
0.0160
0.0140
0.0125
0.0112
0.0100
0.0089
0.0080
0.0071
0.0063
0.0057
0.0049
0.0042
0.0035
0.0018
0.0012
0.0006
-0.0001
-0.0007
-0.0014
-0.0019
-0.0026
-0.0032
-0.0039
-0.0046
-0.0054
-0.0063
-0.0071
-0.0080
-0.0088
-0.0099
-0.0113
-0.0128
-0.0147
-0.0168
-0.0200
-0.0262
-0.0377
-0.0505
-0.0720
-0.0999
(b) Residual Rx Data
Amplitude
volts
14 19 24
Time (ns)
Figure 13. Raw (a) and residual data (b) collected postinjection approximately 59 hours into the experiment and
approximately 33 hours after injection stopped. Note that the PCE broke through the 100% clay layer approximately
four hours into the experiment. This image suggests that the PCE has flowed through the 3% and 6% clay layers and that
the dielectric tool has responded to residual PCE left in the pore space. The borehole video image by this time shows PCE
from approximately 60-cm depth to the top of the 100% clay layer. This correlates with the strongest part of the anomaly
shown in the residual image (b).
20
-------
1.50 -
1.00 -
O 0.50 -
0.00
-0.50 -
-1 0
10
15
20
Time (ns)
Figure 14. Example of minimum and maximum amplitude traces recorded in the 100% sand zone 33 hours after injection
stopped (refer to Figure 13 at approximately 31 and 60 cm). These traces suggest a relative permittivity range in the sand
between 24 and 20. According to the BHS flow chart found in Sneddon et al., 2000, this implies a relative residual
saturation of PCE between 0 and 12%. These have a saturation uncertainty of about +/- 5%, based upon the pre-spill
baseline measurements.
1.53
O 3.53
3.33
-3.53 -
. Mrxlelen ;
• 1 L-J :_:m
• 7^ r.m
13
15
23
Time (ns)
Figure 15. Amplitude traces taken at 33 hours after the injection stopped at depths of 75 cm (middle of the 3% clay layer)
and at 100 cm (middle of the 6% clay layer). These traces suggest relative permittivity of about 22 and according to the
BHS model implies a residual saturation of about 7% This has a saturation uncertainty of about +/- 5%, based upon the
pre-spill baseline measurements.
21
-------
the dielectric logging tool can be used to monitor movement of PCE around a borehole and that the tool is
sensitive to changes in the relative saturation of PCE.
Closer analysis of the data show that some traces in the anomalous PCE saturated zones have a dramatic
decrease in slope approximately 1.5 ns after the peak amplitude. This is illustrated by trace 68 (62 cm)
labeled in Figure 16. At first this was troublesome to the authors because such a change in slope was not
seen in the modeled data. In an attempt to reproduce this response in the numerical simulation, a two-
dimensional model was run that had a 100% PCE saturation at the borehole wall to a radius of 14 cm
away from the borehole wall. The result of this modeling along with trace 68 for comparison can be
observed in Figure 17. The modeled data agree qualitatively with the physical data. The tool may have
detected at least 14 cm or more into the formation surrounding the borehole. Quantitative interpretation of
the dielectric tool data is complicated by possible asymmetry and inhomogeneity of the PCE saturation
surrounding the borehole.
Based upon video logging, it is estimated that by May 18, 120 hours after the spill stopped, about 72 liters
of PCE had migrated to the bottom of the tank (85% of the injected volume). The remainder was retained
in the layers as residual saturation. The tank was monitored by some of the other geophysical methods
until March 2005. Little geophysical or video logging changes were observed during the period from the
end of May 2004 until March 2005. In March 2005, the tank was drained and the formations were
excavated. Based upon the Oil red O dye, significant fingering and stratification in PCE concentration
were observed throughout both the 3% and 6% clay layers. Photographs of the excavated PVC wells are
shown in Figure 18. Variation in the PCE distribution in the layers is evident from the distribution pattern
left from the PCE etched and the Oil Red O dyed surface of the excavated PVC wells. During excavation,
a number of samples of the formations were collected at about every 10-20 cm in depth. The samples
were analyzed on a gas chromatograph and the results are shown in Figure 19. In the sand above 62-cm
depth, the residual PCE is less than 5% of the pore space, in the 3% clay layer in ranges from 2 to 26% of
the pore space, and in the 6% clay layer in ranges from 2 to 20%. Below the solid clay barrier, little
residual PCE was observed, except at the very bottom of the tank (from 4 to 10%) where it had drained.
Given the significant inhomogeneity of PCE concentration over short distances, these results are in
reasonable agreement with the preliminary BHS analysis of the dielectric tool.
22
-------
Trace 68
Figure 16. Magnified three-dimensional plot of the anomalous zone displayed in Figure lOb and lOc (four hours after the
spill began). Note the large decrease in slope shown at trace 68 at approximately 2.5 ns. In accordance with the numerical
simulation, this may be attributed to a finite vertical layer (~14 cm deep) of PCE against the borehole wall (refer to Figure
17).
1.00
0.50 -
0.00
-0.50
Modeled
Physical
Time (ns)
Figure 17. Plot of the raw waveform average of trace 68 (Figure 15) recorded four hours into the spill during injection at
a depth of 62 cm (Figure lOb and lOc). Also on this plot is a modeled trace assuming a vertical layer 14 cm wide. For the
vertical layer, Sr = 4 and for the water saturated sand beyond the layer, Sr = 24. Note the anomalous decrease in slope on
both traces at approximately 2.5 ns for the physical data and 4 ns for the modeled. Primary differences in the modeled
and received waveform are due to multiple reflected arrivals not accounted for in the model. These results suggest that
the tool may have detected at least 14 cm into the formation around the borehole.
23
-------
sand
3%
clay
6%
clay
sand
(a)
(b)
Figure 18a, b. Photographs of PVC wells after excavation of the tank. The clear PVC east and west wells are shown in
Figure 18a. The dielectric tool was logged in the west well at the right in Figure 18a. The PVC seismic wells are shown in
Figure 18b; the south receiver well is on the left and the north transmitter well is on the right. The variation in PCE
distribution in the layers is evident from the pattern etched and dyed on the surface of the wells from the direct contact
with the PCE.
24
-------
35 n
30 -
8 25
n
o 20 -
o
Q.
•5 15
u
I 10
o
Q.
5 -
PCE Residual Concentration
A
A
20 40 60 80 100
Depth (cm)
120
140
160
Figure 19. Plot of PCE concentration as a percentage of pore space verses depth. Samples were taken at a number of
locations at a particular depth in both the 3% and 6% clay layers and at the bottom of the tank. The range in values at
the 55 and 87 cm depths reflect the inhomogeneous distribution of the PCE in the formations. Note these depths are from
the surface of the sand; 12 cm should be added for comparison to the dielectric logging measurements (to the top of the
west well).
25
-------
26
-------
Conclusions
The dielectric logging tool indeed responded to varying degrees of PCE saturations in the tank and
successfully monitored PCE movement throughout the experiment. Little attempt was made to interpret
the dielectric tool data quantitatively. This is because the tool response in different borehole environments
is not completely known at the present time. Interpretation is further complicated by possible asymmetry
and inhomogeneity of the PCE saturation around the borehole. Despite these unknowns, numerical
simulations that were performed using a simplified model and the FDTD method qualitatively were in
agreement with the modeled data. This suggests that relative interpretation of the dielectric tool data is
possible. The tool is an excellent anomaly detector and it can determine whether there is a relative
increase or decrease in dielectric permittivity in thin zones around the borehole. The modeling also
suggests that the tool can detect changes as deep as 14 cm away from the borehole. This is consistent with
tests performed by Abraham, 1999a that show the tool having a depth of investigation as far as 25 cm. As
a result, the dielectric tool may be helpful in detecting changes of PCE saturation around a borehole
during remediation of a contaminated site.
27
-------
28
-------
References
Abraham, J.D. 1999a. Physical modeling of a prototype slim-hole time-domain dielectric logging tool.
M.Sc. thesis. Colorado School of Mines.
Abraham, J. D., 1999b. "Physical modeling of a prototype slim-hole time-domain dielectric logging tool."
In Proceedings of the Symposium on Application of Geophysics to Engineering and
Environmental Problems. Environmental and Engineering Geophysical Society, 503-512.
Ellefsen, K.J., Abraham, J.D., Wright, D.L. and Mazzella, A.T. 2004. "Numerical study of
electromagnetic waves generated by a prototype dielectric logging tool." Geophysics (69): 64-77.
Greenhouse, J., Brewster, M., Schneider, G., Redman, D., Annan, P., Olhoeft, G., Lucius, J., Sander, K.,
and Mazzella, A. 1993. "Geophysics and solvents: The Borden experiment." The Leading Edge
(12): 261-267.
Kraus, J.D. 1950. Antennas. McGraw-Hill Book Company, Inc., New York.
Lucius, J.E., Olhoeft, G.R., Hill P.L. and Duke, S.K., 1992. Properties and Hazards of 108 Selected
Substances (1992 ed.). U.S. Geological Survey Open File Report 90-527.
Sears, F.W., Zemansky, M.W., and Young, H.D. 1987. University Physics. Addison-Welsley Publishing
Company, Reading.
Sheriff, R.E. 1991. Encyclopedic Dictionary of Exploration Geophysics. Society of Exploration
Geophysics, Tulsa.
Sneddon, K.W., Olhoeft, G.R., Powers, M.H. 2000. "Determining and mapping DNAPL saturation values
from noninvasive GPR measurements." In Proceedings of the Symposium on Application of
Geophysics to Engineering and Environmental Problems. Environmental and Engineering
Geophysical Society, 293-302.
Wright, D.L. and Nelson, P.H. 1993. "Borehole dielectric logging: a review and laboratory experiment."
In Proceedings of the 5th Int. Symp. on Geophysics for Minerals, Geotechnical and
Environmental Applications. Minerals and Geotechnical Logging Society, paper T.
Wright, D.L., Abraham, J.D., Ellefsen, K.J., and Rossabi, J. 1998. "Borehole radar tomography and
dielectric logging at the Savannah River Site." In Proceedings of the 7th Inter. Conf. on GPR,
539-544.
29
-------
30
-------
-------
&EPA
United States
Environmental Protection
Agency
Office of Research
And Development (8101R)
Washington, DC 20460
Official Business
Penalty for Private Use
$300
EPA/600/R-06/092
September 2006
www.epa.gov
Please make all necessary changes on the below label,
detach or copy, and return to the address in the upper
left-hand corner.
If you do not wish to receive these reports CHECK HERE D;
detach, or copy this cover, and return to the address in the
upper left-hand corner.
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT No. G-35
Recycled/Recyclable
Printed with vegetable-based ink on
paper that contains a minimum of
50% post-consumer fiber content
processed chlorine free
------- |