A UPCIYI

auwumwifi V
.3?» mm. 3.. .
bnfi 2+.“ :7 Jun

‘1”; E
fishfmfi ,
42“”. .3 l o
1" fawn

 

. a 3m. .
W...

L an). .3 v
.. I. .s.
. .1...
‘ .‘
.6 n .
.3139...
.2. 2"
7. . .
.3.

.Hfinr
. 7

.3 ‘
rif ..
, 9r

 

 

    

.42 .. r. . L.
mfiukfv Eryn

   

  

 

Tm uflg‘ia... . , ., , . . twig”; u $2 :AJfiPfl! 53‘
‘ i .r ‘ .‘ . Jm‘ . .51.. ....: 1r! 4 1 . 1
. _ ‘ ‘ 1.3.. unitaa.‘ . 2501...: ..

 

mu; .. LIBRARY
W ' Michigan State
University

This is to certify that the
thesis entitled

LABORATORY AND NUMERICAL SIMULATIONS OF
THREE-DIMENSIONAL MICROBIAL TRANSPORT AND
BIODEGRADATION

presented by

Peter A. Lepczyk

has been accepted towards fulfillment
of the requirements for the

MS. degree in Geological Sciences

 

 

fix“)

 

 

MajoKProfessor’s Signature

(/E/ZGO§

Date

MSU is an Affinnative ActiorVEqual Opportunity Institution

 

PLACE IN RETURN Box to remove this checkout from your record.
TO AVOID FINES return on or before date due
MAY BE RECALLED with earlier due date if requested

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 clel Rm.mp.13

LABORATORY AND NUMERICAL SIMULATIONS OF THREE-
DIMENSIONAL MICROBIAL TRANSPORT AND BIODEGRADATION
By

Peter Alexander Lepczyk

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE IN ENVIRONMENTAL GEOSCIENCES
Department of Geology

2005

ABSTRACT

LABORATORY AND NUMERICAL SIMULATIONS OF THREE-DIMENSIONAL
MICROBIAL TRANSPORT AND BIODEGRADATION

By

Peter Alexander Lepczyk

A three-dimensional column experiment was created to provide insight into the
bioaugmentation of carbon tetrachloride by Pseudomonas stutzeri KC. The goals of this
experiment were to increase our understanding of the transport of Pseudomonas stutzeri
KC and to collect high-resolution datasets to conduct a case study using a numerical
model developed by the Departments of Geological Sciences and Civil and
Environmental Engineering at Michigan State University that accounts for the three-
dimensional microbial transport and biodegradation specific to the Schoolcrafi Plume A
project. The laboratory model results exhibit effective transport of the mobile phase
microbe and degradation of CT; however, the results indicate a lag between the transport
of mobile Pseudomonas stutzeri KC and a conservative tracer, which was not produced in
the numerical simulation. The simulated CT degradation exhibited a higher degree of
mass removal than what was observed, though detailed qualitative comparisons could not

be made.

ACKNOWLEDGMENTS

I would first like to thank my wife, Xiomara, whose encouragement, love, and patience
was instrumental in the completion of this thesis. I would also like to acknowledge all of

my family, friends, and co-workers who provided me with additional strength and

support.

Next, I would like to thank my committee members, starting with my advisor, David
Hyndman, without whom this thesis would not have had direction. His comments and
suggestions were always appreciated and more importantly, his belief in me through this
arduous process will never be forgotten. I would also like to thank Phanikumar Mantha,
who was constantly willing to help and provide insight into the numerical modeling
aspects of this work. Lastly, Dave Long’s constructive criticism and flexibility in
meeting my time constraints was greatly appreciated. In addition to my committee
members, I would like to express my gratitude towards the rest of the Department of

Geological Sciences for allowing me this opportunity to learn and grow as a scientist.

Finally, I would like to acknowledge the entire Schoolcrafi laboratory column research
team, including Mike Dybas, Xianda Zhao, Roger Wallace, Georgina Vidal-Gavilan,
Emily King, and Chris Ritchie, all of who played major roles in this project and I am very
appreciative of their hard work and dedication. Specifically, I would like to thank Mike
Dybas, Xianda Zhao, and Roger Wallace for their contributions to the design of this

laboratory experiment, and Xianda Zhao, Georgina Vidal-Gavilan, Emily King, and Chris

iii

Ritchie for the data collection and sample analyses. Working as a member of this
interdisciplinary research group taught me a great deal about teamwork, of which I am

very thankful.

iv

TABLE OF CONTENTS

 

 

 

 

 

 

 

LIST OF TABLES VII
LIST OF FIGURES VIII
1 INTRODUCTION 1
2 SCHOOLCRAFI‘ BACKGROUND 4
2.1 PILOT STUDY ....................................................................................................... 5
2.2 FULL-SCALE BIOCURTAIN ................................................................................... 6
2.3 THREE-DIMENSIONAL MODEL AQUIFER EXPERIMENT ......................................... 7
3 THREE-DIMENSIONAL LABORATORY MODEL 9
3.1 LABORATORY COLUMN DESIGN .......................................................................... 9
3.2 SUPPORT SYSTEMS ............................................................................................. 15
3.3 GROUNDWATER FLow AND CHEMICAL TRANSPORT EXPERIMENTAL DESIGN 17
3.4 HORIZONTAL BROMIDE TRACER TEST ............................................................... 18
3.5 VERTICAL BROMIDE TRACER TEST AND CARBON TETRACHLORIDE SORP’I‘ION
STUDY .......................................................................................................... 19
3.6 SOIL TITRATION ................................................................................................. 21
3.7 INOCULATION AND HORIZONTAL BROMIDE TRACER TEST 2 .............................. 21
3.8 BIOCURTAIN EXPERIMENT ................................................................................. 23
4 3-D LABORATORY MODEL AQUIFER EXPERIMENTAL
RESULTS 25
4.1 HORIZONTAL BROMIDE TRACER TEST — LABORATORY RESULTS ...................... 25
4.1.1 Vertical Bromide Tracer Test and Carbon T etrachloride Sorption Study .. 29
4.1.1.1 Vertical Bromide Tracer Test ............................................................... 30
4.1.1.2 Carbon Tetrachloride Sorption Study ................................................... 35
4.1.2 Soil Titration ................................................................................................ 39
4. 1.3 Inoculation ................................................................................................... 40
4.1.4 Biocurtain Experimental Results ................................................................. 44
5 NUMERICAL MODELING RESULTS 49
5.1 MICROBIAL TRANSPORT AND CARBON TETRACHLORIDE DEGRADATION MODEL
BACKGROUND ............................................................................................... 49
5.1.1 Multi—Component Reactive Transport Model .............................................. 51
5.2 LABORATORY COLUMN STUDY NUMERICAL MODEL CONCEPTUALIZATION ...... 55
5.3 CALIBRATION 0F GROUNDWATER FLOW AND CONSERVATIVE TRACER TRANSPORT
MODEL .......................................................................................................... 56
5.4 SENSITIVITY ANALYSES ..................................................................................... 70
5.5 CONSERVATIVE TRACER TEST DISCUSSION ....................................................... 74
5.6 CARBON TETRACHLORIDE SORPTION SIMULATION ............................................ 75
5.6.1 CT Sensitivity Analysis ................................................................................. 78
5.6.2 CT Sorption Simulation Discussion ............................................................. 82
5.7 INOCULATION SIMULATION ................................................................................ 84

5.8 NUMERICAL BIOCURTAIN SIMULATION ............................................................. 88

 

 

6 DISCUSSION 99
APPENDICES 103
REFERENCES l 17

 

vi

LIST OF TABLES

Table 1. List of numerical model input parameters. ........................................................ 76

vii

LIST OF FIGURES

Figure 1. Map illustrating the location of Schoolcraft, Michigan, the position of plume A,
and hydraulic head contours in feet. Also identified are the locations of the pilot

study and the full-scale biocurtain, reprinted from Hyndman et a1 (2000). ................ 5
Figure 2. Schematic of laboratory model aquifer experiment,modified from Vidal-
Gavilan (2000). ......................................................................................................... 10
Figure 3. Location of sampling ports, horizontal influent and effluent ports, and model
aquifer dimensions, modified from Vidal-Gavilan (2000). ...................................... 12
Figure 4. Location of vertical influent ports, modified from Vidal-Gavilan (2000). ....... 13
Figure 5. Locations of sampling ports, modified fiom Vidal-Gavilan (2000) ................. 15
Figure 6. Distribution of bromide concentrations, in C/Co, for the horizontal bromide
tracer test (90, 120, 150, and 180 minutes). .............................................................. 26
Figure 7. ISO-surfaces generated from kriging bromide concentrations, in C/Co, for the
horizontal conservative tracer test (150 and 180 minutes). ...................................... 27
Figure 8. Effluent bromide concentrations, in C/Co, for the horizontal conservative tracer
test. ............................................................................................................................ 29

Figure 9. Distribution of bromide concentrations in C/Co and corresponding iso-surfaces
generated from interpolation using inverse distance weighting, for the vertical
conservative tracer test (16 and 24 hours). ............................................................... 31

Figure 10. Distribution of bromide concentrations, in C/Co, in row 9 sampling ports and
shaded contours generated through kriging for the vertical conservative tracer test

(30, 31, 32, 33, 34, and 35 hours). ............................................................................ 33
Figure 11. Bromide concentrations, in C/Co, versus time, measured at the combined
vertical effluent ports during the vertical conservative tracer test. ........................... 34
Figure 12. Bromide and carbon tetrachloride breakthrough curves. ..... A .......................... 36
Figure 13. pH versus time measured from the horizontal effluent port during the soil
titration. ..................................................................................................................... 40
Figure 14. Bromide (top) and liquid phase biomass (bottom) breakthrough curves during
the inoculation event (modified fiom Vidal-Gavilan, 2000). ................................... 42

Figure 15. ISO-surfaces generated from kriging bromide (left) and liquid phase biomass
(right) concentrations, in C/Co, at the end of the inoculation event (180 minutes)

(Modified from Vidal-Gavilan, 2000). ..................................................................... 43
Figure 16. Carbon tetrachloride concentrations in row 7 sampling ports during the
biocurtain experiment. .............................................................................................. 45
Figure 17. Carbon tetrachloride concentrations (ppb) 1.5 hours and 12 days into the
biocurtain experiment. .............................................................................................. 47
Figure 18. Simulated versus observed bromide tracer concentrations during the
horizontal tracer test .................................................................................................. 58
Figure 19. Simulated versus observed vertical bromide tracer test results for the
combined vertical effluent port. ................................................................................ 63
Figure 20. Simulated versus observed RT3D simulation of the vertical bromide tracer
test at row 9 sampling ports ...................................................................................... 64

viii

Figure 21. ISO-surfaces generated from the RT3D simulation of the horizontal bromide
tracer test at 150 and 180 minutes, compare to Figure 7 for the observed iso-surfaces
at the same times. ...................................................................................................... 69

Figure 22. ISO-surfaces generated from the RT3D simulation of the vertical bromide
tracer test at 16 and 24 hours. Compare to Figure 9 for the iso-surfaces generated

from the observed data sets. ........................................... _. .......................................... 70
Figure 23. Sensitivity analyses of porosity and dispersivity in the horizontal bromide

tracer study. ............................................................................................................... 71
Figure 24. Sensitivity analyses exhibiting the effects of varying dispersivity on the

horizontal (top) and vertical (bottom) breakthrough curves. .................................... 73

Figure 25. Simulated versus observed concentrations at row 9 using the initial reaction
parameters fiom Phanikumar et al. (2005) and the calibrated parameters from the

tracer simulations. ..................................................................................................... 78
Figure 26. Simulated CT breakthrough curves resulting from perturbations to sorption
parameters. ................................................................................................................ 79
Figure 27. Simulated versus observed carbon tetrachloride concentrations during the CT
sorption study at sampling port 932. ......................................................................... 81
Figure 28. Comparison of simulated versus observed mobile strain KC concentrations
during the inoculation event ...................................................................................... 86

Figure 29. Comparison of bromide and mobile strain KC during the inoculation event. 87
Figure 30. Comparison of simulated concentrations for sampling ports 312 (top) and 722
(bottom) during the biocurtain experiment. .............................................................. 91
Figure 31. Concentrations of CT represented by iso-surfaces at day 12 of the simulation.
................................................................................................................................... 95
Figure 32. Concentrations of mobile strain KC represented by iso-surfaces at day 12 of
the biocurtain simulation ........................................................................................... 96
Figure 33. Concentrations of immobile strain KC represented by iso-surfaces at day 12
of the biocurtain simulation. ..................................................................................... 97

ix

1 INTRODUCTION

Carbon tetrachloride (CT) is a carcinogenic compound that has been widely used as a
degreaser, solvent, fumigant, and fire extinguisher. Unfortunately due to disposal
practices prevalent in the past it has also become a common groundwater contaminant
(Lewis and Crawford, 1999; Hyndman et al., 2000; Dybas et al., 2002). Traditional clean
up methods for removing CT, include pump and treat coupled with activated carbon
sorption or air stripping. However, these techniques are costly and time consuming.
Moreover, these methods only treat aqueous phase constituents and do not address the
contaminants sorbed onto the aquifer media, which may pose a future threat to the site if
the contaminants were released afier the treatment operations have been completed

(Davis, 1995; Fetter, 1999).

Novel strategies, such as bioremediation, are necessary to clean up CT contaminated
groundwater. Three distinct types of bioremediation strategies have been developed:
natural attenuation, biostimulation, and bioaugrnentation. Natural attenuation is the
process by which the native flora are left alone to degrade contaminants and mitigate the
Size and extent of the plume. This may be a reasonable option when the natural rate of
biodegradation is fast enough to prevent significant contaminant migration.
Biostimulation involves stimulating indigenous microbes to transform target
contaminants through the addition of either an electron acceptor, (i.e., oxygen or nitrate),
an electron donor (i.e., acetate or lactate), or both. Bioaugmentation is the process of

adding a non-indigenous microbe into the aquifer to transform target contaminants that

indigenous microbes either could not transform at an acceptable rate or the results of the
transformation are undesirable. Bioaugmentation may also require addition of electron

donors and/or acceptors (Hyndman et al., 2000; Chapelle, 2001; Dybas et al., 2002).

Bioaugmentation offers an advantage over biostimulation and monitored natural
attenuation because it creates a controlled pathway for degradation to occur. This is an
important distinction from the other bioremediation strategies, which may lead to end
products that are more hazardous than the original contaminant. In the case of CT,
Criddle et a1. (1990) demonstrated that chloroform (CF) is a common end product of CT
transformation by many indigenous microbes. Therefore, it was critical to discover a
microbe capable of transforming CT into non-hazardous end products. In 1987, Criddle
et al. (1990) found such a microbe, later named Pseudomonas stutzeri KC (KC), in an
aquifer in Seal Beach, California. This discovery provided the basis for a significant
amount of research on the mechanisms and kinetics of this bioremediation process
(Criddle et al., 1990; Lewis and Crawford, 1993; Tatara et al., 1993 and 1995; Dybas et
al., 1995; Davis, 1995; Mayotte et al., 1996; Witt et al., 1995 and 1998), and led to tests
of strain-KC bioaugmentation of a CT contaminated aquifer in Schoolcraft, Michigan

(Dybas et al., 1998; Hyndman et al., 2000; Dybas et al., 2002; Phanikumar et al., 2005).

This thesis presents a subset of the results of a three-dimensional column experiment
created to provide insight into the full-scale biocurtain used to remediate dissolved phase
carbon tetrachloride at a contaminated aquifer located near Schoolcraft, Michigan

(Hyndman et al., 2000; Dybas et al., 2002). The objectives of this laboratory model

aquifer experiment were to (a) improve our understanding of how well Pseudomonas
stutzeri KC was transported in the subsurface, (b) to visualize the spatial distribution of
biomass after an inoculation event, and (c) to conduct a case study using a numerical
reactive transport model developed by the Departments of Geological Sciences and Civil
and Environmental Engineering at Michigan State University that accounts for the three-
dimensional microbial transport and biodegradation associated with the Schoolcrafi

project.

2 SCHOOLCRAFI‘ BACKGROUND

In 1987, the Michigan Department of Natural Resources (MDNR) began investigation of
an aquifer near Schoolcraft, Michigan (Figure 1), which was impacted with dissolved
phase CT and nitrate. This plume has later been referred to as Schoolcraft Plume A. In
accordance with common technologies at the time, the MDNR proposed a pump and treat
system coupled with air shipping. According to Halliburton NUS Environmental Co.
(1991), the remedial action for this site would take approximately 25 years. As was
previously mentioned, 1987 also marked the discovery of strain KC. Utilizing this
microbe, researchers at Michigan State University conducted batch and one-dimensional
column studies (Mayotte et al., 1996; Witt et al., 1995 and 1999). These experiments
demonstrated the rapid transformation of CT into C02, forrnate and an unidentified non-
volatile product; this transformation required denitrifying conditions, the presence of an
electron donor (like acetate), alkaline pH (7.8 to 8.3), and typical aquifer temperatures (5-
25 °C); (Criddle et al., 1990, Tatara et al., 1993; Lewis and Crawford, 1993; Davis, 1995;
Mayotte et al., 1996; Witt et al., 1995 and 1998; and Dybas et al., 1995 and 1998). It was
later demonstrated that the transformation was mediated by a secreted factor determined
to be pyridine-2,6-bis(thiocarbonxylate) or PDTC, which is produced under the fore
mentioned conditions (Lee et al., 1999). Encouraged by the results, the state of Michigan
funded an interdisciplinary research team to test the field applicability of a
bioaugrnentation system using strain KC at the Schoolcrafi Plume A site. This team
conducted a field pilot test (Dybas et al., 1998), engineered a full-scale bioaugrnentation
system (Hyndman et al., 2000 and Dybas et al., 2002), and later designed this three-

dimensional laboratory model aquifer experiment (V idal-Gavilan, 2000).

    

/ no.0

  
 

Site of Pilot Study

  
 

Site of Full Scale Study
/

 
 
 
 
 

  

/
/
/

   

/

o m .m chsuo

m
V I0” rum ILL
_-—. m!“ CUT”

Figure 1. Map illustrating the location of Schoolcrafi, Michigan, the position of plume A,
and hydraulic head contours in feet. Also identified are the locations of the pilot study
and the full-scale biocurtain, reprinted from Hyndman et a1 (2000).

2.1 PILOT STUDY

A pilot test was conducted at Schoolcrafi Plume A in 1994 to test the effectiveness of

bioaugrnentation with strain KC under field conditions (Dybas et al., 1998). In general,
the groundwater in Schoolcraft Plume A contains CT, up to roughly 100 ppb, nitrate at
roughly 65 ppm, dissolved oxygen between 2 to 5 ppm, pH between 7.2 and 7.4, and is

phosphorous and carbon limited. The pilot test entailed pulsing base and phosphate

amended groundwater to create conditions more favorable for the grth and survival of
strain KC (Dybas et al., 1998). The aquifer was then inoculated with strain KC, which
was grown in an aboveground reactor, and an additional pulse of acetate and phosphate
was added. In order to provide the required conditions for strain KC, the aquifer was re-
injected with acetate, base, and phosphate. Although favorable, the observed degradation
during the pilot study was less than predicted determined through lab experiments;
dissolved nitrate levels decreased by approximately 85% and dissolved CT
concentrations by 65% and soil borings indicated that sorbed CT concentrations
decreased between 60 to 88%. Dybas et a1 (1998) suggested that aquifer heterogeneities
were responsible for these sub-optimal degradation conditions. These results agreed with
conclusion of Murphy et al. (1997) that aquifer heterogeneity can cause variations in
degradation rates as well as uneven spatial distributions of microorganisms. The
Schoolcrafi Plume A pilot test results confirmed that the success of bioaugmentation
requires an understanding of the aquifer heterogeneities and a remedial design that

accounts for them (Hyndman et al., 2000).

2.2 FULL-SCALE BIOCURTAIN

Following the analyses of the pilot test, a full-scale biocurtain was designed and installed
at the Schoolcrafi site (Hyndman et al., 2000; Dybas et al., 2002). Conceptually, this
entailed creating a biologically active zone perpendicular to natural gradient flow. The
contaminated groundwater flows by natural gradient into the biocurtain and is degraded

when the microbes are stimulated with the appropriate pH, acetate, and phosphorous

concentrations through a series of delivery wells. The objectives of the biocurtain design
were to (a) decrease the concentrations of CT and nitrate to below Michigan Department
of Environmental Quality residential drinking water criteria (5 ppb and 10 ppm,
respectively) from the original concentrations of 100 ppb and 65 ppm, respectively, (b)
develop an effective treatment zone across a vertical transect through the contaminant
plume, and (c) to minimize remediation costs. In order to meet these objectives, the
aquifer at the field site was characterized in detail and numerous delivery grid designs
were examined using numerical models. After conducting a cost optimization algorithm,
the final design entailed 15 closely spaced re-circulation wells (l-m apart) screened
across the entire vertically impacted zone (Hyndman et al., 2000). The wells were used
for initial inoculation and weekly aquifer amendments. Installation of the system began
in spring 1997 and most downgradient monitoring locations indicated between 95 and
100% CT removal after two to three years of operation (Hyndman et al., 2000 and Dybas

et al., 2002).

2.3 THREE-DIMENSIONAL MODEL AQUIFER EXPERIMENT

In order to fiIrther understand the bioaugmentation of CT by strain KC, members of
MSU’s Departments of Civil & Environmental Engineering and Geological Sciences
designed a large 3-dimensional model aquifer. This 3-dimensional model aquifer was
used to create a carefully controlled set of experiments, designed to simulate the
biocurtain at Schoolcrafi Plume A (V idal-Gavilan 2000). This experiment filled a void

between the early one-dimensional column studies by Witt et al (1995 and 1999), and the

full-scale biocurtain presented by Hyndman et a1 (2000) and Dybas et a1 (2002). Through
the use of a three-dimensional aquifer model, many shortcomings of one-dimensional
models like edge effects are reduced. Moreover, complete three-dimensional sampling
and analyses can be accomplished in a model aquifer, where as costs would be

prohibitive to fully characterize a field site in this manner.

3 THREE-DIMENSIONAL LABORATORY MODEL

Column experiments are common practice in hydrogeology and engineering. They allow
investigations to take place under controlled conditions, therefore reducing many
uncertainties found in the field. In turn these experiments are often used as an
intermediate step to evaluate remedial options prior to a field scale application.
Furthermore, the data obtained from these experiments can be used to calibrate, and in
some cases even develop, reactive transport models. Many hydrogeologists and
environmental engineers are discovering that this is a valuable course of action. By

decreasing the uncertainty, one can create a conceptual model with fewer assumptions.

This section discusses the development and design specifications of the 3-dimensional
model aquifer experiment. Furthermore, it describes the series of tests that were

performed to determine specific groundwater flow and aquifer parameters.

3.1 LABORATORY COLUMN DESIGN

The 3-D laboratory model aquifer was composed of a large column representing the
Schoolcrafi aquifer and a series of stages designed to support the laboratory biocurtain.
These stages include pumps, tanks, and a cooling system, to create controls and add
amendments to the system. A detailed schematic of the laboratory model is illustrated in

Figure 2.

Effluent
Tank

 

Overflow

 

 

 

 

4 Efiluent lines

 

 

 

 

 

“L.“ pH pH

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
 

 

 
    

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

[:1
Batch Tank , g be 0 o
(Tank#4) . , . Pm Data 33
" U m Logger oo
Influent
pH Flux Control NW. CT/Br Tank
C: Pu It # 3
3 3 3 3 controller mp (4 channels) (Tan )
Til Tank
3888 CI‘lBr Feed
Pump - Syringe Pump(s)
Flo
Influent . Met:
0 O Refill pH Tank Solenord O 0
Tank (Tank #2) Valve
(Tank #1) (normally closed)
N2 [/1 Conflller ° ° ° 0
Tank 0 D 0 CI C 02 N2
pH Tank Tank
Probe
Recycle Recycle
Pump Pump

Figure 2. Schematic of laboratory model aquifer experiment, modified fi'om Vidal-
Gavilan (2000).

10

The actual model aquifer was a vertical cylinder 3 feet high by 2 feet in diameter and
filled with aquifer soil collected from the site. The sand was mixed to create a
homogeneous, isotropic media. The cell was packed by saturating the sand before filling
the tank with the wet slurry, and ensuring that the height of water in the tank was always
above the surface of the sand, which also aided in the removal of trapped air (V idal-

Gavilan, 2000).

The stainless steel cylinder used for this experiment was both durable and non-reactive.
Inflow and outflow lines were placed at the bottom and top of the tank, as well as at the
Sides. Moreover, 57 sampling ports were installed throughout the cylinder to allow the

collection of 3-dimensional data sets (Figure 3) (V idal-Gavilan, 2000).

The Inflow and outflow lines placed at the bottom and top of the flow cell were used to
create conditions similar to natural gradient flow. The system was designed for water to
be pumped into the tank through the bottom to guarantee full saturation of the aquifer
media. The influent and effluent lines included one center position and six locations that
were evenly spaced around the center, at a radius of 8 inches. However, due to improper
manufacturing, the top lid did not create a watertight seal with the cylinder, thus the top
lid was rotated until a watertight seal was achieved. This resulted in the top lid being
rotated clockwise from the desired position by approximately 30 degrees (Figure 4)

(Vidal-Gavilan, 2000).

11

 

913

933

 

 

 

923

912 00 922 GO 932 00

911 O 921 O 931 O

8120 8220 8320

713 723 733

712 ()0 72200 732 GO

711 O 721 O 731 O

6120 6220 6320

513 523 533

512 GO 522 GO 532 00

511 O 521 O 531 0 91,4 cm
. 412 422 432

Horrzontal O O 0 "mm
Influcnt Effluent
Port 313 323 333 p0,,
——> 31200 322 GO 332 ()0 ———>

311 O 321 O 331 O

2120 2220 2320

113 123 133

112 GO 122 ()0 132 ()0 I 30.48 cm

111 O 121 O 131 O

22.86 cm
15.24 cm 1‘
61 cm

 

 

 

 

Figure 3. Location of sampling ports, horizontal influent and effluent ports, and model
aquifer dimensions, modified from Vidal-Gavilan (2000).

12

 

60.96
48.26
*--- 30.48 —-

4 12.7 V\ Units

60 \ arecm

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

50 O 7
O I +
I 50.80
40 40.64
20.32
30/ l l
10.16 —
(a) Bottom of the 3-D flow cell.
60.96
«— 40.64 ——J
~— 30.48 —.l
+ 20.32
\ are cm
0“ .
5
65 G3—

 

 

 

(b) Top of the 3-D flow cell.

Figure 4. Location of vertical influent ports, modified from Vidal-Gavilan (2000).

13

The horizontal inflow and outflow lines were located 12 inches from the base of the tank
on each side. Their purpose was to Simulate the lateral flow between a set of injection
and extraction wells comprising the Schoolcraft biocurtain. All of the flow lines were
attached to the tank via 1/4 inch, stainless steel fittings. The fittings were housed with
fine mesh in an attempt to prevent the migration of fine grained sand into the influent and

effluent lines.

The 57 sampling ports were positioned throughout the model aquifer and penetrated the
aquifer media at various depths. The sampling ports consisted of stainless steel fittings
with 1/16 inch stainless steel tubing running through the center of each one. The
sampling ports were subdivided into three columns, nine rows and three depths. The
center column was perpendicular to the cross flow influent and effluent lines. The other
two columns were offset 6 inches from the center. Row number 1 was 6 inches from the
bottom of the tank and subsequent rows were 3 inches apart from one another. Odd
numbered rows had sampling ports at three separate depths. These depths are assigned 1,
2 or 3, which were equivalent to 6 inches, 12 inches and 18 inches, respectively. Even
numbered rows had only one sampling depth at 12 inches. Identification of sampling
ports was accomplished through assigning a notation depending on row, column and
depth. Sampling ports will herein be referred to as a 3-digit number, where the first
number is the row, the second number is the column and the third number is the depth.
For example, sampling port 812 is located in row 8, column 1, and depth 2. Refer to

Figure 5 and (V idal-Gavilan, 2000) for details.

14

 

45.72 ——-i

H— 30.48 ——l

—~ 15.24 \ Units

\ are cm
(3

X13 X23 X33

 

 

 

 

 

 

 

Figure 5. Locations of sampling ports, modified from Vidal-Gavilan (2000).

3.2 SUPPORT SYSTEMS

Afler the cylinder was fabricated and wet packed with the soil, the support systems were
connected to the tank. The design and fimction of the support systems is as follows (See

Vidal-Gavilan (2000) for additional details).

Schoolcraft groundwater was introduced into the laboratory model aquifer system via the
refill tank (tank #1). Nitrogen was pumped into the tank at a pressure to just overcome
the hydraulic head in order to strip out any dissolved CT. Water from tank #1 was then
pumped into the pH adjustment tank (tank #2). An overflow line connected tank #2 to
tank #1, thus maintaining a constant hydraulic head in the pH adjustment tank. Nitrogen

was also added to tank #2 to strip out any oxygen that may have been added to the water

15

and create anoxic conditions. Carbon dioxide was added to the water when the pH was
above 7.4 to reduce it to 7.2. This was accomplished by means of a pH probe in tank #2
connected to a solenoid that was normally closed. Both of these tanks were housed in a
cold room, maintained at the ambient groundwater temperature of Schoolcrafi (roughly

12° C).

A gravity feed line was then connected from tank #2 to the amendment tank (tank #3),
located outside of the cold room next to the model aquifer. Tank #3 was where chemical
species were introduced into the water, i.e. bromide or carbon tetrachloride. This tank
had a series of influent ports where syringe pumps could be attached to add a fixed rate of
chemical and thus create a constant influent concentration. This tank rested on a
magnetic stirring plate with a stir bar mixing the water. A batch tank (tank #4) was used
for specific experiments when it was deemed that the engineered support system was
unnecessary. Finally the water was fed into the model aquifer either thrOugh the bottom
of the cell or the side, depending on the nature of the specific experiment. A peristaltic
pump provided the vertical flow via connections to the bottom of the model aquifer,

while a diaphragm control pump provided the horizontal flow component of the tank.

The effluent lines from the top of the tank converged into a single constant head tank
positioned 15 cm above the top of the tank. Prior to the experiments, the vertical influent
and effluent lines were decreased to 4 each (influent ports 1, 3, 5, and 7 and effluent ports
1, 2, 4, and 6) due to plugging problems in a few of the other ports observed during the

leak testing. The horizontal effluent line discharged to the same constant head tank.

16

In order to keep water temperature constant and minimize effects of thermal fluctuations
on the kinetics of the co-metabolic degradation, the tank was insulated and cooled. Metal
tubing was attached to a re-circulating cold water bath kept at 8° C and wrapped around
the outside of the tank. Insulation was then adhered to the outside of the tank. Measured
model aquifer temperatures matched field conditions and were maintained between 10
and 12° C. Water was sampled through syringes screwed onto the sampling ports and
withdrawing 2-5 ml of water for bromide and CT analyses and microbial plate counts,

respectively. See Vidal-Gavilan (2000) for details of the design and system operation.

3.3 GROUNDWATER FLOW AND CHEMICAL TRANSPORT
EXPERIMENTAL DESIGN

Due to scale differences between the field and the model aquifer, laboratory flow rates
varied from those observed on site. Under natural gradient conditions, the average linear
velocity of the Schoolcraft aquifer is approximately 14.8 cm/day (Halliburton NUS
Environmental Co., 1991). We used approximately 4 times this value for the vertical
tracer tests. To achieve this, groundwater was pumped into the 3-D laboratory model
aquifer system at a combined flow rate of approximately 33 ml/min. According to Dybas
et a1 (2002), the flow rate of a single injection well is approximately 19 L/min. In order
to use flow rates more suitable for a laboratory scale model, this horizontal flow rate was
decreased to a value of approximately 280 mein. After establishing the flow rates, a
series of tracer tests were performed to observe the breakthrough curves and use the data

sets to characterize parameters for the numerical modeling, such as porosity and

17

dispersivity. Following these tests, the soil was titrated with NaOH to raise the pH to a
level more suitable for the growth of KC. Lastly, the column was inoculated with KC
and the full-scale biocurtain experiment commenced. The target flow rate for the vertical

natural gradient flow used in the biocurtain experiment was 1.4 ml/min.

3.4 HORIZONTAL BROMIDE TRACER TEST

The first tracer test consisted of a conservative tracer breakthrough study performed
between the single horizontal influent and effluent lines without the presence of a vertical
flow component. The primary objective of this test was to experimentally determine the
amount of time required to achieve breakthrough in the delivery zone and observe if any
preferential flow paths existed. Additional objectives were to collect a data set that could
be used with numerical models in order to estimate the porosity and dispersivity of the

model aquifer.

Bromide was chosen as the tracer for its ease of analyses with a bromide probe. Sodium
bromide was mixed with nitrogen stripped Schoolcraf’t groundwater in tank #4 to create a
bromide solution of 114.5 mg/L. This concentration was assumed constant throughout
the experiment. The bromide solution was pumped into the column at a flow rate of
roughly 280 mllmin, however flow measurements taken from the effluent water stream
indicated that the rate varied between 260 and 300 mein. These variations were
attributed to the flexible tubing inside the peristaltic pump, or partial plugging of flow

lines.

18

Based on preliminary calculations, using an estimated porosity of 37% and flow rate, full
tracer breakthrough was expected after 3 hours. A Cole-Farmer bromide electrode was
connected to a Jenco model 6091 data logger and placed in-line with the effluent stream.
Data was recorded at 5-minute increments for the duration of the test (204 minutes). A
sampling schedule was prepared in accordance with the breakthrough estimate for a
selected number of internal sampling ports. The schedule was created to capture the
tracer progressively as it traveled through the tank. In turn, samples were collected at 90,
120, 150, and 180 minutes after the start of the test. 2-3 ml samples were collected from

these sampling ports and later analyzed with the bromide probe upon test completion.

3.5 VERTICAL BROMIDE TRACER TEST AND CARBON
TETRACHLORIDE SORPTION STUDY

Following the conservative horizontal tracer test, a combined conservative (bromide)
vertical tracer test and Carbon Tetrachloride sorption test was performed. The objectives
of this experiment were to (a) determine the amount of time required to achieve
breakthrough in the vertical dimension, (b) compare the breakthrough curves of Br and
CT and calculate the retardation value of the CT, and (c) utilize the data to determine
physical parameters for the numerical model. This test was performed without the

influence of the horizontal delivery system.

Carbon Tetrachloride and bromide were mixed in tank #3. Two syringe pumps delivered

the chemicals into tank #3 where they were combined with gravity drained groundwater

l9

fi'orn tank #2. The syringe pumps were set to inject a specified concentration of each
chemical constituent at a constant flow rate in order to create constant influent
concentrations. The target concentration of bromide was 80 ppm and the target
concentration of CT was 50 ppb. Influent concentrations were analyzed several times
over the course of the experiment. Results indicated that the influent bromide
concentration varied between 62.0 and 74.6 ppm with an average concentration of 68.9
ppm, and the CT influent concentrations varied between 36.7 and 64.5 ppb with an
average of 49.4 ppb (V idal-Gavilan, 2000). It was later realized that the variations in
concentration were the result of changes in water level in tank #3 from the gravity feed

line.

The target flow rate for this experiment was 33 ml/min. Flow rates were collected
several times over the course of the experiment and measured from the combined
effluent, located 15 cm above the top of the tank. Though the vertical flow system had
been tested prior to this experiment, fluctuations in flow rate were observed between 22
and 36 ml/min. Over the course of the 42-hour bromide tracer test, the flow rate
averaged 34.3 ml/min. During the 10 day CT portion of the experiment the flow rate

averaged 32.7 ml/min.

Water samples were collected fi'om the combined effluent at lS-minute intervals once the
bromide tracer front was observed nearing the effluent. Bromide samples were collected
from selected internal sampling ports at 16, 24, 30, 31, 32, 33, 34, 35, and 36 hours. All

samples were analyzed with the bromide electrode. Water was sampled for CT analyses

20

at 16, 24, 48, 72, 96, 120, 138, 146, 161, 184, and 241 hours. CT was analyzed by gas
chromatography, as described by Vidal-Gavilan (2000). CT addition continued through
the model aquifer for an additional two days to try and homogenize the distribution of

solute in the model aquifer.

3.6 SOIL TITRATION

Upon completion of the vertical bromide tracer and CT sorption tests, the pH of the
model aquifer was adjusted through a soil titration process. The goal of this procedure
was to elevate the pH of the delivery zone from the naturally occurring pH of 7.2 to
approximately 8.2, where Pseudomonas strain KC flourishes. In turn, the adjustment of
the delivery zone was accomplished by injecting CT spiked Schoolcrafi groundwater
with an approximate CT concentration of 45 ppb and an elevated pH of 8.2. Water was
injected into the horizontal influent port at an approximate flow rate of 280 ml/min and
flowed out of the horizontal effluent port into the constant head discharge tank over a 16-
hour period. Data was collected in-line at 5-minute increments from the effluent port and

analyzed with a Jenco model 6091 pH meter and 8 Hanna Instruments pH probe.

3.7 INOCULATION AND HORIZONTAL BROMIDE TRACER TEST 2

Once the pH of the delivery zone was optimized for the growth of strain KC, the model
aquifer was ready to be inoculated. To replicate the field application, the goal was to set-

up a population of strain KC in the delivery zone, thus creating the biocurtain.

21

A 57 L fermenter was prepared with filter-sterilized Schoolcraft groundwater and strain
KC cultures, and amended with acetate to a concentration of 800 ppm, 70 ppm nitrate,
roughly 40 ppb CT, phosphate to a concentration of 10 ppm, trace metals, and base to a
pH of 8.2 (refer to Vidal-Gavilan (2000) for a complete description of the strain KC
culture preparation). The final inoculum concentration was approximately 8.8 x 107
cfu/ml. The fermenter was spiked with bromide, to a concentration of roughly 64 mg/L,
and connected to the model aquifer via a diaphragm pump. The inoculation and tracer
test lasted for approximately 3 hours. The flow rate was measured several times and
averaged 272 ml/min. Samples for bromide and mobile phase KC analyses were
collected at 30-minute intervals fi'om the horizontal influent, ports 312, 322, 332, and the
horizontal effluent during the duration of the test. Dissolved CT was measured from the
horizontal influent as well; however, concentrations did not hold steady and fell to below
detectable limits by the first hour. Upon completion of the inoculation, the vertical flow
was initiated and the horizontal flow ceased. Moreover, sampling of rows 1, 3, and 5 for
bromide and mobile phase KC immediately followed the commencement of the
inoculation/tracer injection. All bromide samples were analyzed with the bromide
electrode, as previously described, and biomass concentrations were determined by

Vidal-Gavilan (2000) through plate counting on R2A agar dishes.

22

3.8 BIOCURTAIN EXPERIMENT

After the inoculation event, the bioaugmentation phase of this study began. The goal of
this experiment, and the true impetus of this study, was to simulate the biocurtain at the
Schoolcrafi site. As mentioned above, the horizontal flow component was terminated
once the inoculation event was finished and the vertical flow system was activated,
simulating the natural gradient flow. The experiment lasted two weeks and was broken
up into three stages. The first bioremediation phase lasted 6 days, after which point the
vertical flow was temporary shut off, and a 3 hour feeding event took place between the

horizontal flow system, followed by another 7 days of natural gradient flow.

Schoolcraft groundwater was pumped into the model aquifer at a flow rate of
approximately 14 ml/min. Flow readings were recorded fiom the combined effluent on a
daily basis and indicated that the flow fluctuated between 13 and 14 ml/min. Carbon
tetrachloride was added to the water in tank #3 via a syringe pump. Influent
concentrations of CT were monitored on a daily basis. The average influent
concentration was 37.9 ppb, however fluctuated between 16.8 and 61.7 ppb. Samples
were collected from the combined effluent on a daily basis and analyzed for CT.

Samples were collected fi'om the odd numbered internal sampling ports at three different
times during the experiment, 1.5 hours, 5 days, and 12 days after the start of the
experiment. CT was analyzed by gas chromatography, as described in Vidal-Gavilan

(2000).

23

The feeding event occurred on the 6th day of the biocurtain experiment. The vertical flow
rate was temporarily turned off and the horizontal flow system was activated. The model
aquifer was injected with amended Schoolcrafi groundwater with a pH of 8.2, acetate
(100 mg/L), nitrate (70 mg/L), carbon tetrachloride (53 pg/L), and trace metals. The
feeding event lasted approximately 3 hours. The flow rate was monitored every 30
minutes fiom the horizontal effluent port and averaged 287 mllmin, though fluctuated
between 265 and 301 ml/min. In the field case the natural gradient flow continues during
the feeding event in the field, however this difference from the field situation was
deemed insignificant due to the difference in magnitude of the horizontal and vertical
flow rates. After the feeding event, the horizontal flow system was terminated and the

natural gradient flow resumed.

The experiment continued for another 6 days at which point it was determined that
several problems existed with the experiment and that the resulting data was becoming
less meaningful. This will be discussed in detail in Chapter 7. See Vidal-Gavilan (2000)

for details of the collected data.

24

4 3-D LABORATORY MODEL AQUIFER EXPERIMENTAL RESULTS

The following chapter presents the results and analyses of the 3-D laboratory aquifer
model experiment. Where appropriate, geostatistical techniques were employed to aid in

the visualization of the data. These and other images in this thesis are presented in color.

4.1 HORIZONTAL BROMIDE TRACER TEST — LABORATORY RESULTS

The results of the horizontal bromide tracer test are listed in Appendix A. The data was
normalized by dividing the observed concentration by the influent concentration (C/Co).
Resulting fi'om the Slight fluctuation of influent concentrations and errors associated with
the analyses of the bromide, values of C/Co were occasionally greater than 1.0. When
this occurred they were set to 1.0 and denoted with asterisks in the Appendix. AS was
mentioned in Chapter 2, data was collected from the internal sampling ports at 90, 120,
150, and 180 minutes after the start of the experiment. Figure 6 illustrates the 3-D data
sets in the context of the 3-D model aquifer. The dots indicate the location where the
sample was collected and are shaded in accordance to the concentration gradient listed

next to each figure.

25

as

k I

,, L
1!; gm; .
A 11 I'll

ll"

3.

1111:
1115:,
:‘v1

5
. -=:

":1, i
f l

I

.. u

1‘ I

‘ll ll:
.. .!i
.ll
1“- 1 1

1 I
tfll'l‘lll

 

Figure 6. Distribution of bromide concentrations, in C/Co, for the horizontal bromide
tracer test (90, 120, 150, and 180 minutes).

Adequate samples were collected at 150 and 180 minutes to create 3-D geostatistical
plots. Shaded iso-surfaces were generated from these interpolations to aid in the

visualization of the bromide tracer front (Figure 7). Interpolations were performed inside

26

of Groundwater Modeling System (GMS) 4.0 using kriging. The 3-D finite difference
grid used in the interpolation will be discussed in the numerical modeling section of this
chapter. The interpolations represent statistical models of the data, and in turn should be
used for visualization purposes only. The data indicates that the bromide distribution
radiates fi'om the influent port creating a parabolic shape when viewed in cross-section.
As expected, the fastest breakthrough occurs along the direct flow path between the
horizontal influent and effluent ports. Sampling ports 232, 132, and 131 displayed less
than expected bromide concentrations. This may be due to local heterogeneities in the

vicinity of those ports.

 

ffluent

Figure 7. ISO-surfaces generated from kriging bromide concentrations, in C/Co, for the
horizontal conservative tracer test (150 and 180 minutes).

Data was also collected from the horizontal effluent port at 5-minute intervals. In turn,
this represents the most complete dataset to observe the tracer breakthrough. A graph of

C/Co versus time (Figure 8) suggests some instability in the readings from the bromide

27

electrode. However, since the data was primarily used to look at the slope of the
breakthrough curve and the center of mass arrival time, a little instability in the readings
was acceptable. The maximum normalized concentration reached at the effluent was
0.78, thus indicating that full breakthrough had not been achieved by the end of the
experiment. The center of mass, or 50% breakthrough, arrived at approximately 154
minutes. The data obtained from this experiment was used to calibrate the numerical
groundwater flow model. This will be discussed in detail in the numerical modeling

section of this thesis.

28

Horizontal Effluent Port

C/CO

 

 

 

° 1 l ' I 1 1 1 1 1 1
0 4O 80 1 20 160 200
Time (min) '

Figure 8. Effluent bromide concentrations, in C/Co, for the horizontal conservative tracer
test.

4.1.1 Vertical Bromide Tracer Test and Carbon Tetrachloride Sorption Study

The data from the vertical bromide tracer test and carbon tetrachloride sorption study are
included in Appendix B. Since these two experiments were run simultaneously, the data
needed to be normalized, so that direct comparisons could be made. Once again, if C/Co
was greater than 1.0, it was set to 1.0 and indicated in the appendix. This fluctuation in

influent concentrations resulted in more than half of the sampling ports exhibiting

29

concentrations above 1.0 after full breakthrough occurred. The maximum normalized

concentration observed during this test was 1.16.

4.1.1.1 Vertical Bromide Tracer Test

In order to effectively characterize the bromide breakthrough, the data was broken into
three distinct data sets, illustrating the 3-dimensional movement of the tracer, the 2-
dimensional breakthrough of the tracer as it passed through the plane represented by row
9, and the l-dimensional breakthrough of the combined vertical effluent ports. Sufficient
3-dimensional data was collected at 16 and 24 hours to observe the bromide fiont as it
traveled up through the model aquifer. Once again, geostatistical techniques were
utilized inside GMS 4.0 to aid in the Visualization of these two data sets. Figure 9 shows
the 3-D data sets in the context of the model aquifer and the shaded iso-surfaces that were
generated fi'om interpolating the data using inverse distance weighting With a quadratic
nodal function. As was noted earlier, these interpolations should be used for
visualization purposes only. In general, the iso-surfaces illustrate a roughly radial tracer
front moving up from the influent ports at the bottom of the tank. However, several
locations within the model aquifer indicated less than expected bromide concentrations.

This is further elucidated in the series of 2-D plots of collected data.

30

Vertical Effluent Vertical Effluent

jggu 1 111 II

'Iiiilmi' I

11' 1 lllll
ll 1:;

1 lIliu'
11i|fl11'-

 

Vertical lnfluent Vertical Influent

Figure 9. Distribution of bromide concentrations in C/Co and corresponding iso-surfaces
generated from interpolation using inverse distance weighting, for the vertical
conservative tracer test (16 and 24 hours).

In order to observe the breakthrough as the bromide approached the vertical effluent ports
of the model aquifer, samples were collected and analyzed from all of the row 9 sampling
ports at the following times, 30, 31, 32, 33, 34, 35, and 36 hours. This data was brought
into GMS 4.0 and interpolated in two-dimensional cross sections using kriging (Figure
10). The data and subsequent interpolations, illustrate a non-uniform breakthrough as the
bromide passes row 9. The portion of the model aquifer near sampling port 923 displays

the fastest breakthrough. Sampling port 922, which represents the centerline of the

31

column as it passes through row 9, displays the second fastest breakthrough. The slowest
breakthrough was observed near sampling ports 913 and 921. As can be seen in Figure
10, both of these locations are located near the same X, y coordinates as influent ports,
however are 76.2 cm higher in the z dimension. It was assumed that flow rates were
divided equally amongst the 4 influent and 4 effluent ports and that the model aquifer
was unifonnly packed. However, partial plugging of some of the influent and effluent
lines was observed during this experiment. Direct comparison with the CT breakthrough
data indicated that this plugging effectwas partially responsible for the non-uniform
breakthrough. However, the rotated positions of the influent and effluent ports also
contributed to the non-uniform flow field, as did possible heterogeneities. Regardless of

these less than ideal circumstances, valuable data was obtained.

The time required to reach 50% breakthrough in row 9 averaged 32.3 hours, but varied
between 30.2 and 35.1 hours. These arrival times were calculated throUgh linear
interpolations between time series observations. Using this average for the center of

mass arrival time and the height of row 9 (76.2 cm), an average linear groundwater

velocity of 2.36 cm/hr was determined (V idal-Gavilan, 2000).

32

 

 

 

 

 

 

 

 

 

 

Figure 10. Distribution of bromide concentrations, in C/Co, in row 9 sampling ports and
shaded contours generated through kriging for the vertical conservative tracer test (30,
31, 32, 33, 34, and 35 hours).

33

Bromide samples were also analyzed fiom the combined vertical effluent of the model
aquifer. Figure 11 is a graph of the breakthrough curve. It is apparent that the sample
collection schedule did not firlly characterize the bromide breakthrough. Fortunately,
adequate samples were collected to determine the center of mass arrival time at the top of

the column (approximately 39 hours) and the slope of the breakthrough curve.

 

 

 

 

1 .—
0.8 __
_l
o 0.6 _.
Q
o —4
0.4 — Legend
G—G—O Effluent Concentration
_ E—B—B lnfluent Concentration
0.2 —
_l
o I I I I I I I r
O 10 20 30 40

Time (hrs)

Figure 11. Bromide concentrations, in C/Co, versus time, measured at the combined
vertical effluent ports during the vertical conservative tracer test.

These results were useful for many reasons. First, they provided insight into the non-
uniform nature of the vertical groundwater flow in the model aquifer. Secondly, they

were used to estimate the retardation coefficient of carbon tetrachloride, as will be

34

discussed in the next section. Lastly, the data sets obtained from this experiment were

again used in the calibration of the numerical groundwater flow model.

4.1.1.2 Carbon T etrachloride Sorption Study

The results of the carbon tetrachloride sorption study were useful to visualize the
movement of contaminant in the model aquifer and to determine the retardation factor.
As noted earlier, the CT influent concentrations varied between 36.7 and 64.5 ppb,
averaging 49.4 ppb (refer to Figure 12a for a graph of the normalized influent
concentrations). The 50% breakthrough of CT in row 9 occurred at approximately 79.6
hours, arriving as early as 60.5 hours at port 923 and as late as 120.9 hours at port 921.
These anival times were calculated in the same manner as the vertical bromide arrival
times. Furthermore, utilizing the same technique as was done to determine the average
linear groundwater velocity; one arrives at an average CT velocity of 0.96 cm/hr. Refer

to Appendix C for the data.

Retardation is ofien determined by dividing the velocity of a retarded chemical species by
the average linear velocity (Fetter, 1988). Since both the average linear velocity and the
CT velocity were determined at the same height in the model aquifer, and knowing that
velocity equals distance divided by time, the retardation equation was re-written in terms
of arrival times. Utilizing the Solver feature inside of Microsofl Excel, the optimal
retardation value was calculated. This was accomplished by dividing the 50% CT arrival

time data by the retardation value and minimizing the root mean squared residual

35

between the Br and adjusted CT data sets by solving for the optimal retardation value.
This approach yielded a retardation value of 2.61 . This is consistent with CT isotherm
studies performed on Schoolcraft soil, which yielded a retardation factor of 2.64 (Zhao et
al., 1999). This study also indicated that there was a kinetic component or multi-rate

behavior to the sorption.

Figure 12. Bromide and carbon tetrachloride breakthrough curves.

1.4 m

 

 

0.8 — /
Legend .
[Ge 0.. Br-lnlluerfl

O —Q— Q CT-lnfluent

 

 

 

0.6 I 17 l I l ‘1
0 20 40 60
Normalized Time (hrs)

Figure 123. Influent concentrations versus time for Br and CT.

36

C/CO

 

 

 

0 20 40 60 80
Normalized Time (hrs)

Figure 12b. Br and CT breakthrough curves at row 9, colunm 1, all three depths.

1 _ l
l
0.8 — J
0.6 —
8
o _
0.4 —4
‘ Legend
7 1 . , . Br-921
,‘ O L. Br-922
0.2 ~ f O 0 Br-923
‘ AH cr-921
- , i079 CT-922
e 94 cr-923
0

 

 

 

0 20 40
Normalized Time (hrs)

Figure 12c. Br and CT breakthrough curves at row 9, colunm 2, all three depths.

37

C/CO

 

 

  

8O

40
Normalized Time (hrs)

Figure 12d. Br and CT breakthrough curves at row 9, column 3, all three depths.

 

 

 

1 _
0.8 — E
i
— O
i ,1“ xx,
0.6 e .‘ ,/ t
z /‘
g — ’ I' e 6/
O O/
0.4 — ,4
,0“ 0
— ,3
1‘ Legend
0.2 -— T .’ . r. Br-Effluent
o Q' 0A 0 CT-Effluent
_ g /
/'/
..
° ' l I I ' l ' l . l
0 20 40 60 80 100

Normalized Time (hrs)

Figure 12e. Br and CT breakthrough curves at the combined effluent port.

38

CT concentrations were plotted versus time, along with the corresponding Br
concentrations, to create a series of breakthrough curves at the influent port, all row 9
sampling ports and the combined vertical effluent port (Figure 12). This data was
originally presented in Vidal-Gavilan (2000). In order to visualize the breakthrough of
CT and Br on the same plot, the CT time was divided by the retardation factor. Notice
that the 50% breakthroughs of the row 9 sampling ports and the effluent port are roughly
the same, however the slopes of the curves are different due to the non-equilibrium or
multi-rate nature of the sorption. This experiment was Simulated using a two-site

numerical sorption model as described in section 5.6.

4.1.2 Soil Titration

Data was obtained by X. Zhao at the horizontal effluent port and recorded at 5-minute
intervals. Figure 13 is a plot of pH versus time. The starting pH of the model aquifer
was roughly 7.4. The pH of the influent water was 8.2 and at the end of the 14.5-hour
titration, the effluent pH was 8.0. Thus full titration of the model aquifer was not quite
achieved, illustrating the high buffering capacity of the soils. However, KC flourishes

when the pH is between 7.8 and 8.3.

39

Soil Titration

 

 

8.2—7

a;

378—:
7‘ I l I l I l I l
0 4 8 12 16

Time (hrs)

Figure 13. pH versus time measured from the horizontal effluent port during the soil
titration.

4.1.3 Inoculation

The results of the combined inoculation and bromide tracer test are listed in Appendix D.
Time series data sets of the liquid phase biomass and bromide were collected at sampling
ports 312, 322, 332, and the horizontal effluent. Once again, the data sets were
normalized so that direct comparisons could be made. The Co for bromide was measured
once during this experiment and assumed to be equal to 63.8 ppm throughout the duration
of the test. The influent liquid phase biomass was measured several times over the course

of the experiment and averaged 8.8 x 107 cfu/ml. Once again, values of C/Co were

40

occasionally greater than 1.0. When this occurred they were set to 1.0 and indicated with

an asterisks in the appendix.

Figure 14 shows the breakthrough of the liquid phase biomaSs and tracer. Over the
course of the 3-hour experiment, full breakthrough of the bromide and liquid phase
biomass was achieved at sampling ports 312, 322, and 332. Concentrations decreased
afier full breakthrough occurred, due to fluctuations in the influent concentrations.
Because the flow rate for this test matched the first horizontal tracer test, full
breakthrough was not anticipated at the horizontal effluent for either bromide or liquid
phase biomass. As expected, all of the sampling ports show a lag between the bromide
arrival time and the liquid phase biomass. This could be due to a portion of the liquid
phase biomass attaching to the aquifer media or a retardation of the mobile phase KC.
This differential breakthrough is further illustrated in the data sets that were collected at
the end of the inoculation event. Samples were collected from rows 1 #— 5 and analyzed
for bromide and liquid phase biomass. The data was brought into the 3-D finite
difference grid and interpolated using kriging. Once again, iso-surfaces were generated

for Visualization purposes (Figure 15).

41

 

 

 

 

 

 

 

 

 

/'//// \x‘0
L nd
0.8 H —O (79% Br-312
' ' ~2 Br-322
I ‘ -- V -—\7 Br-332
X“ X Br-Hor_ett
0.6 _.
E -i /
° /
0.4 — //)<
0.2 ~ ’1 "Pi/
l / .
\ /
-1 / \\)/
' /
0 ‘4 I I l I T I l I 1
0 4° 30 120 160
Time (min)
1 _1 /D O o O 0______ Q1 V
0.8 — \/
‘ Legend
0 " O W 0 8104312
/ BNZZ
0'6 T V V V Bio-332
X X "X Bio-Hor_efl
c
Q _
0
0.4 ~
0.2 -
,.- ,,,,, x
0 —T i] 1* I X , It .._- ixT—x/"i x T‘ _. L Vii-xi’jl j
0 4° 30 120 160
Time (min)

Figure 14. Bromide (top) and liquid phase biomass (bottom) breakthrough curves during the

inoculation event (modified from Vidal-Gavilan, 2000).

42

 

Figure 15. ISO-surfaces generated from kriging bromide (left) and liquid phase biomass
(right) concentrations, in C/Co, at the end of the inoculation event (180 minutes)
(Modified fi‘om Vidal-Gavilan, 2000).

As can be seen, both the tracer and biomass show parabolic breakthroughs emanating
from the horizontal effluent port. These figures also clearly display the lag time between
the tracer and liquid phase biomass. More importantly, the spatial distribution of the
liquid phase biomass is visualized. A concentration gradient was established with higher
concentrations of liquid phase biomass near the influent port and very low concentrations

at the horizontal effluent port.

The immobile phase biomass was not measured during this experiment. However, when
using this dataset, Vidal-Gavilan (2000) applied the clean-bed filtration theory to estimate
the amount of solid phase biomass remaining in the model aquifer. The parameters used

in the filtration theory were previously determined from a ID column study investigating

43

the soil loading capacity of KC in Schoolcrafi soil (Radabaugh, 1998). In general, Vidal-
Gavilan’s findings stated that once KC reached a concentration of 3x107 cfu/g in the soil,
full liquid phase biomass breakthrough was observed. Since 100% breakthrough was
observed at ports 312, 322, and 332, it can be assumed that the soils have reached their

loading capacity.

As is discussed in the next section, the distribution of biomass is key to understanding
how the CT was degraded during the biocurtain experiment. Furthermore, the mobile
phase biomass data was useful to test the predictive capability of a user-defined reactive

transport model, as presented in Chapter 6.

4.1.4 Biocurtain Experimental Results

Although the Simulated biocurtain was monitored through the analyses of dissolved
carbontetrachloride, acetate and nitrate, liquid phase biomass, and pH, only limited
dissolved phase CT data was available for use in this thesis. This sparse dataset made it
difficult to analyze the bioremediation experiment. Dissolved phase CT data was
obtained fi'orn the vertical combined effluent ports at 1.5 hours, 2.2 days, 3.2 days, 5
days, and 12 days, the odd numbered internal sampling ports at 1.5 hours, 5 days, and 12

days, and the horizontal effluent port during the feeding event (Appendix E).

Resulting fi'om the limited data and experimental oversights, the results are difficult to
interpret. The results suggest that the simulated biocurtain was actively degrading carbon

tetrachloride. Based entirely on the concentrations of CT in rows 7 and 9 and the

combined vertical effluent, it appears as if the laboratory biocurtain was actively

degrading CT, for those sampling ports all show a decrease in concentration over time. A

graph illustrating this decrease has been prepared for row 7 (Figure 16). Recalling the

distribution of liquid phase biomass at the end of the inoculation, it was expected that

maximum CT removal would occur closer to the horizontal influent and decrease near the

horizontal effluent. This may be evident in the data obtained fi‘om the row 7 ports.

 

 

 

 

 

 

 

45 —IB\
b- ‘x
\ »\\
A \
40 —7 ‘\ \\
\ \\\g\
‘13“ —-~.\.
35 *4 X T‘ \ _,,,_ -,_ _,
\ ‘WK‘m—n D
1:1“ \ o
3 3O —\
a. . \ L
3 \\ 1 \ \ \
O \
'0 : \
'C 25 —
.9
5 \
«I
a \
I- 20 —" \ \
c \ \
\ f \\
g Legend
3 \ c—e—c or 711 \
15 —“ +—+—+ cr-712 \
\
\
0
A
\_L.__ 4 L“ +—_.__. —V— O
I
300

13an of Inoculation (0 hrs)

 

Time (hrs)
Feeding Event (144 hrs)

. Figure 16. Carbon tetrachloride concentrations in row 7 sampling ports during the
biocurtain experiment.

45

Rows l — 5 exhibit mixed results, with many of the sampling ports indicating increases in
CT concentrations over time. This could be a result of the initial distribution of CT in the
model. Since the CT included in the inoculum was degraded within an hour of the onset
of the inoculation event, initial concentrations of CT between the horizontal influent and
effluent ports were low. Once again, 3-D interpolations were generated using kriging
inside of GMS 4.0 to illustrate the CT distributions afier 1.5 hours and 12 days (Figure
17). Figure 17a, illustrates the CT distribution 1.5 hours afier the end of the inoculation
event and the beginning of the CT spiked natural gradient flow. As can be seen in Figure
17a, the CT distribution looks like the inverse of the biomass distribution at the end of the
inoculation (Figure 15b) with the lowest CT concentration near the horizontal influent
port. Figure 17b illustrates the distribution of the CT after the biocurtain has been
operating for 12 days. As can be seen in this figure, concentrations are elevated near the
vertical influent ports and towards the horizontal effluent port. Concentrations are less in
row 7, primarily on the side proximal to the horizontal influent port. This suggests that
either the biocurtain was most active near row 5 and that the bioactive zone migrated
fiom the delivery zone due to the natural gradient flow vertically transporting the
microbes or that the low CT concentrations in that portion of the column were a result of
the low concentration CT zone that developed during the inoculation event migrating
vertically. Based entirely on advective flow, the low concentration CT zone emanating
fiom the horizontal influent port should have migrated to the top of the tank within 3
days. Therefore, the data suggests that the majority of the CT decrease over time was the

result of biodegradation.

46

Vertical Effluent Vertical Effluent

     
   

   

 

Horizontal

1 N Effluent

Horizontal : if Effluent Horizontal Il'
Influent Influent "

~11 ':

Vertical Influent Vertical lnfluent

Figure 17. Carbon tetrachloride concentrations (ppb) 1.5 hours and 12 days into the
biocurtain experiment.

Based on analytical results from row 7, the average decrease in concentration as
contaminant moved through the biocurtain was 31% at 1.5 hours, 68% at 5 days, and
77% at 12 days. However, using the median of the analytical results fiom row 7, the
decrease in CT concentrations is 19% at 1.5 hours, 95% at 5 days, and 98% at 12 days.
The median values are more representative when determining the remediation
percentages, for they are not influenced by outliers or in the case the ports proximal to the
horizontal effluent where it was already shown that the biomass concentrations were
much lower. These results compare favorably to what was observed in the field (Dybas

et al., 2002).

47

Moreover, the results indicate that in order to uniformly treat a contaminant as it moves
through a biocurtain, one needs to make certain that the biomass is evenly distributed
between the injection and extraction wells comprising the biocurtain. It is therefore
necessary to ensure that firll breakthrough of the liquid phase biomass has occurred
between the injection and extraction wells and that heterogeneities in the aquifer material

have been accounted for.

In addition to the sparse dataset, many unforeseen problems occurred during the
experiment that compromised portions of the data. In particular, the highly variable
influent concentrations of CT added a complexity to the experiment, as did the lack of pH
control. pH dropped significantly over the course of the experiment; according to
Gavillan (2000), one week afler the feeding event, the pH decreased to roughly 7 in the
biocurtain, at which point a second titration of the model aquifer occurred. This titration
lasted 72 hours and required 1,214 liters of re-circulated pH adjusted groundwater to
bring the pH up to 8.2 at the horizontal effluent. Based on these problems, the research
team decided to end the experiment early. Fortunately sufficient data was collected to
utilize the results in a series of numerical reactive transport models aimed at simulating

portions of the laboratory biocurtain experiment.

48

5 NUMERICAL MODELING RESULTS

Groundwater Modeling System (GMS) 4.0 and 5.0, the numerical modeling software
used in this project, includes modules for geostatistics, site characterization, groundwater
flow modeling, reactive transport modeling, calibration, and visualization. GMS
possesses common numerical modeling programs, of which MODFLOW 2000
(Harbaugh et al., 2000) was selected for the finite-difference flow model and RT3DV2.5
(Clement, 1997) was chosen for the finite-difference reactive transport modeling. Inside
MODFLOW 2000, layer property flow (LPF) was selected as the flow package with the
preconditioned conjugant gradient (PCGZ) solver. The following packages were utilized
in RT3D: the advection package with hybrid method of characteristics (HMOC),
dispersion, source Sink mixing, the generalized conjugate gradient solver (GCG, to
implicitly solve the dispersion and source/sink portions of the reactive transport
equation), and the user-defined reaction module. The user-defined reaction module
allows users a great deal of flexibility to describe site specific reactions, including in-situ
chemical oxidation and bioremediation. Given the broad application of this model,
researchers at Michigan State University utilized RT3D to create a numerical model that

accounts for the CT bioaugmentation associated with the Schoolcraft project.

5.1 MICROBIAL TRANSPORT AND CARBON TETRACHLORIDE
DEGRADATION MODEL BACKGROUND

The extensive research associated with the Schoolcraft Plume A resulted in increased

understanding on biodegradation processes and rates. This understanding of the processes

49

allowed for the creation of a three dimensional microbial transport and biodegradation
model. This model is unique to the co-metabolic degradation of CT by strain KC that
occurs under denitrifying conditions with nitrate as the electron acceptor and acetate as
the electron donor. The model is comprised of seven, partial differential, coupled mass
balance equations that account for the primary reactions involved in the transport,
sorption, and biotransformation of CT. Specifically, the modeled processes include the
transport of CT, nitrate, acetate, bromide, and mobile phase strain KC with considerations
to advection, dispersion, sorption, and chemical reaction, as well as, microbial growth,
decay, attachment and detachment (Phanikumar et al., 2002a, Phanikumar et al., 2002b,

Phanikumar and Hyndman, 2003, Phanikumar et al., 2005).

This model was first applied to one-dimensional column studies conducted at Michigan
State University, which investigated microbial transport and CT degradation in two
intermittently fed aquifer columns (Phanikmnar et al., 2002a). Utilizing starting
parameters gleaned from literature and independent experiments, the authors conducted a
series of parameter optimizations whose resulting solution provided reasonable fits to CT,
nitrate, acetate, and early time mobile KC data. The modeling results indicated that the
CT degradation rate for a small lab column was an order of magnitude less than what was
estimated from a batch study. Furthermore, the modeling suggested that another source
of nitrate consumption was present through the existence of indigenous flora in the 1-D

columns.

50

The model was then applied to data obtained fi'orn the one-dimensional column study to
explore the interaction of sorption and bioavailability with pulsed nutrient injections
(Phanikumar and Hyndman 2003). The results indicated the strong influence of
bioavailability on degradation and the necessity to include degradation terms for sorbed
phase mass, without which retardation is an insensitive parameter. Moreover, the authors
demonstrate that using a pulsed nutrient injection strategy is much more efficient than
continual pumping. These observations have wide application to the field of reactive

transport modeling and remedial system design.

Most recently, Phanikumar et al. (2005) utilized the reactive transport model to predict
solute concentrations across the Schoolcrafi site prior to and during the biocurtain
operation using the optimized parameters obtained from the one-dimensional column
study (Phanikumar et al., 2002a). The results fit the observed acetate and nitrate
concentrations well; however they over predict the amount of CT degradation using the
lab optimized parameters. In order to better predict the CT concentrations, the authors
decreased the CT degradation rate by half and suggested that the difference in laboratory
rates and field rates may be due to limitations in electron acceptor availability. The study
also provided insight into the dynamics of the field biocurtain and the influence of

heterogeneities and pumping on microbial transport and solute concentrations.

5.1.1 Maui-Component Reactive Transport Model

As described above, the multi-component reactive transport model is comprised of seven

mass balance equations accounting for the primary reactions involved in the transport,

51

sorption, and biotransformation of CT. Equations (1) through (6) are coupled to account
for the dependency of the substrate utilization, growth and decay of strain KC, and
bioavailability. Equation (1) relates the transport and degradation of aqueous phase CT
with sorption. Equation (2) accounts for the growth, decay, and attachment of mobile
strain KC cells and the detachment of immobile strain KC cells, while Equation (3)
describes the coupled processes for the immobile phase strain KC cells. Equation (4)
describes the utilization of the electron donor (acetate) and equation (5) describes the
utilization of the electron acceptor (nitrate). Equation (6) accounts for the concentration
of sorbed CT, based on a two-Site sorption model. Equation (7) describes the transport of

the non-reactive tracer (bromide) (Phanikumar et al., 2005).

(1) Carbon tetrachloride concentration

6C 6 6C 6
RECI— = 5;;[D4 axT]—E(VICCT)_I(XMCCT -%k'[(l‘f)KdCCT _SCT]+QSCCTS

 

(2) Mobile-phase strain KC concentration

6X" :3— D17 6X" _i(V1Xm)+ I‘m Ca C" “bro 1_ Cd _Ka' X"
at ax; ax] &i Ca + Ksa Cu + Kan C + K
C.
C, + K

30

 

 

 

 

 

+Kde[l- 1X1.“ +Q3Xms

(3) Immobile-phase strain KC concentration

91%: pm Ca CI 4;,“ 1— Ca -K,,,1— C“ Xim+KaiXm+QsXms
a: Ca+Km Cn+Km C +K C +K

a 30 a M

 

 

 

52

(4) Acetate concentration

8C 6 6C a y C S
a:_ D. a _— m a X” X Cas
6t 811,1 38le ax.(‘C ..-) Y, C,+K,, C, K ( + "0+9

 

 

 

(5) Nitrate concentration

 

 

 

33~=,3[D 01—1. c.)-—— Ca (x. + x...)

“6): YC,+KM:‘R,,,CK

 

6,, C C, s
— —1————“—— + X, +X + C,8
[mi C,+K,,) 7C,+,K,1( ”") Q

(6) Sorbed phase CT concentration

6S
3C1- : k[(1—f)KdCCT ‘SCTI—kX S

(7) Bromide tracer concentration

 

a__C,, =_ a D. 6C,
a: ax, I'ax,

)— 5376.9.» Q‘CB.S

Refer to the Table 1 for a complete description of the above symbols. The variables on
the left hand of the equations are the concentrations of all of the species being modeled.
CCT, Xm, Xim, C", C., SCT, and C3, are the dissolved concentrations of carbon
tetrachloride, mobile phase strain KC, immobile phase strain KC, acetate, nitrate, sorbed
CT, and dissolved bromide, respectively. The term R in equation (1) is the retardation
factor which is determined through R = 1 + (prD/B), where KD is equal to the partition
coefficient. This equation has been modified from the original retardation equation, as
presented in Freeze and Cherry (1979) to include the fraction of equilibrium sites (f), as
described in Zhao et a1 (1999), which limits the amount of equilibrium sorption that

occurs (Phanikumar et al., 2005).

53

The first group of terms in the equations for CT, mobile phase KC, acetate, nitrate and
bromide account for the transport of solute based on dispersion, where the dispersion
coefficient (D) is equal to ot'U and U is equal to the average linear velocity and a is the
longitudinal and transverse dispersivities. The second group of terms in the CT, mobile
phase KC, acetate, nitrate and bromide equations account for the transport of solute based
on advection. When used alone, these two groups of terms account for the basic
advection-dispersion equation as described in Freeze and Cherry (1979). The third group
of terms in equation (1) accounts for the degradation of dissolved CT by means of the
mobile phase KC and a second order reaction rate. The fourth group of terms in equation
(1) accounts for the degradation of dissolved CT on the equilibrium sorptive sites. The
concentration of sorbed phase CT is controlled by the kinetic sorption, as described in
equation (6). The fifth term in equation (1) is the CT source term resulting from
recirculation. This term was necessary for the field application of the‘model, since the
extracted groundwater contained solutes, which were re-inj ected into the subsurface
(Phanikumar et a1, 2005). This process did not occur in the laboratory column study;

therefore, all of the source terms denoted with QS are zero.

The third group of terms in equation (2) accounts for the growth, decay and attachment of
the mobile phase biomass. These terms are related through the acetate and nitrate
(substrate) concentrations in the form of Monod terms. The fourth set of terms in

equation (2) accounts for the detachment of immobile phase biomass. Equation (3)

54

functions similarly to equation (2), except that it describes the linked processes of the

immobile phase strain KC.

The third group of terms in equations (4) accounts for the depletion of the electron donor
(acetate) through the grth of mobile and immobile phase biomass. Again growth is
dictated by Monod terms. A similar process is described in the nitrate utilization
equation, equation (5), which describes the nitrate sink that results from the transfer of
electrons from donor to acceptor and prompts microbial growth. The fourth group of
terms in equation (5) accounts for nitrate that is consumed during processes such as
microbial decay and utilization by indigenous flora or endogenous respiration. Readers
are encouraged to refer to Phanikumar et al. (2002a, 2002b, 2003, and 2005) for

additional information pertaining to the reactive transport model.

5.2 LABORATORY COLUMN STUDY NUMERICAL MODEL
CONCEPT UALIZATION

The first step in the model development was to create the finite difference grid
representing the laboratory model aquifer. The 3-D grid consisted of 25 layers, and 41
cells in the x and y dimensions. Cells were de-activated outside of the 61 cm diameter
tank to create the cylindrical shape of the model aquifer. Once the 3-D grid was created,

the fore mentioned interpolations were completed inside the geostatistics module.

The laboratory aquifer was modeled as a confined aquifer since the potentiometric

surface was always greater than the top of the tank. The influent ports were modeled as

55

wells. Furthermore, since all of the effluent ports were connected to tubing that
discharged to the same reservoir, they were modeled as constant head cells. Even though
the laboratory column Showed signs of heterogeneities, sufficient hydraulic head and
permeability data was not available to account for this. Therefore, for the purposes of the
numerical modeling, the laboratory column was assumed to be homogenous. The
hydraulic conductivity was held constant at 0.027 cm/sec. The hydraulic conductivity of
the Schoolcraft aquifer was characterized using constant head permeameter tests on
hundreds of aquifer samples obtained from cores collected across the delivery zone (Zhao

et al., 2005; Dybas et al., 2002; Hyndman et al., 2000).

5.3 CALIBRATION OF GROUNDWATER FLOW AND CONSERVATIVE
TRACER TRANSPORT MODEL

Traditionally, the calibration process involves varying input parameters, such as
hydraulic conductivity, and repeatedly running the model until the reSidual error between
observed and computed hydraulic heads and fluxes are within an acceptable level of
accuracy (Anderson and Woessner, 1992). In addition to minimizing residual error it is
often necessary to examine the hydraulic gradient, groundwater flow direction, and water
mass balance (Mandle, 2002). The calibrated parameters for the tracer test model were
vertical anisotropy, porosity, and dispersivity. In this case, the calibration process
entailed minimizing the residual error between Simulated and observed conservative
tracer test data collected at the effluent port. Hydraulic conductivity did not need to be
calibrated. In accordance with Darcy’s Law, since the fluxes were held constant during

numerical simulations, changing the hydraulic conductivity would result in an equal

56

change in the gradient, thus creating no net change in the flux. In order to alter the rate of
transport for the simulation, one needs to adjust the effective porosity. This is seen in the
relationship describing the average linear velocity, where:

v =o (th/dl)/0
Therefore, decreasing the porosity would effectively increase the simulated average

linear velocity, which is what dictates the advective portion of solute transport.

Prior to calibrating the model, a working MODFLOW simulation was generated.
Simulations were created that modeled the vertical and horizontal conservative tracer
tests. The first calibration was performed on a model that simulated the horizontal
conservative bromide tracer test. The average measured flow rate was used as a positive
flux into the well representing the horizontal influent port. Starting heads were uniformly
assigned as 115 cm across the grid. A steady state MODFLOW model was then run.
Once the flow field had been solved, the RT3D simulation was set-up. A single stress
period was created that matched the duration of the horizontal tracer test. All of the input
parameters were assigned unifome across the model domain. This included values for
porosity, dispersivity, the ratio of transverse dispersivity to longitudinal dispersivity, the
ratio of vertical dispersivity to longitudinal, and the effective molecular diffusion
coefficient. Since temporal three-dimensional data was sparse for this particular
experiment, residuals were determined at the horizontal effluent port. Parameters were
varied over a range of values and successive model iterations were simulated and their
corresponding residuals were recorded. The resulting calibration can be visualized in the
graph of simulated versus observed concentrations at the horizontal effluent port (Figure

18).

57

 

 

 

 

 

0.8 # Horizontal Effluent Port
‘i— ‘+—i' Observed I
[:lw—Fl—F] Simulated / +
q
I figifi
0.6 —T / 1&th _
air /
o 1 111‘ iflI‘l'J
0 _. l, 111 i '
\ “If:
U
0.4 “m [A] Xz IR":
11.; I ,
l/‘
_ ,1_ . L4
71/113311
. . 1
0.2 —4 F111 1
++
1‘1
_. .1 _/
1 1 .1-
first 1* ++ li/i/
O 50 100 150 200 250
Time (min)

Figure 18. Simulated versus observed bromide tracer concentrations during the
horizontal tracer test.

AS can be seen in Figure 18, the slope of the simulated cross flow tracer test is similar to

the slope of the

observed data. The slope of the breakthrough curve is primarily

controlled by dispersivity. A steeper slope would demand a lower dispersivity value or a

reduction in the

amount of numerical dispersion. Dispersivity was tested over a wide

range of values (0.001 -10 cm) during the calibration process. Decreasing the dispersivity

below 0.1 cm did not improve the fit between simulated and observed much, as this value

is likely in a range to ntunerical dispersion. This less than perfect fit could have been

58

attributable to one of a number of reasons, including heterogeneities in the model aquifer,
varying influent concentrations and flow rates, and numerical dispersion. During the
calibration process attempts were made to decrease the numerical dispersion by
decreasing the Peclet number from 15 to 10, an acceptable Value according to Huyakom
et al (1983). This was accomplished by increasing the number of grid cells, however
little change was observed in the goodness of fit between simulated and observed.
Moreover, increasing the grid cells drastically slowed down simulation times, which

extends the times necessary to calibrate the model.

The vertical anisotropy was also varied during the calibration process and the results
were visually noted. As mentioned earlier there was inadequate spatiotemporal data to
compare the simulated and observed at most locations; however, data was available that
allowed for the comparison of the bromide distribution at the end of the tracer test. It was
noted that if vertical anisotropy was ignored (Kh/Kv=1), a slightly better visual fit existed
between simulated and observed data in rows 8 and 9; however, the goodness of fit
between the simulated and observed horizontal effluent data decreased. Since the data
obtained at the horizontal effluent represents the average concentration of solute leaving
the model aquifer it was determined that maintaining the goodness of fit between the
simulated and observed values at this location was more important. This was achieved

when the vertical anisotropy was equal to (Kh/Kv=2).

Other factors that may have contributed to the less than ideal fit between the simulated

and observed breakthrough curve include, the number of particles in the numerical

59

advection package, the type of numerical advection package used, the fluctuations in the
influent concentrations and flow rates, and heterogeneities in the model aquifer. The
number of particles in the numerical advection package was increased to try and decrease
the amount of numerical dispersion; however, this did not improve the goodness of fit
between simulated and observed, nor did using a different numerical advection package,
such as method of characteristics (MOC). Attempts were made to simulate fluctuations
in the influent concentrations and flow rates, however insufficient data was available to
adequately describe these variations and the resulting simulations were not an
improvement. Once a set of values had been estimated that created the minimum residual
error, the parameters were used in a model mimicking the vertical conservative tracer

test.

The vertical tracer test was set-up with 4 wells, representing the influent ports, which
were assigned a quarter of the average observed flow rate. 4 constant head cells (115 cm)
were established at the top of the model aquifer, representing the effluent ports. Once
again, vertical anisotropy, porosity, and dispersivity were tested over a range of values
and the residuals were recorded for each simulation. In this case, the residuals and
breakthrough curves were calculated and compared to the average row 9 concentrations.
The simulated breakthrough curves were also compared to the measurements for each of
the row 9 sampling ports. The starting parameters that were solved for the horizontal
tracer simulation resulted in reasonable fits between for the vertical tracer simulation;
however, minor changes improved the match between simulated and observed values.

Specifically, a slightly lower porosity was required.

60

Calibration was achieved once a set of parameters had been identified that created low
residual error for both the horizontal and vertical simulations. In order to match the slope
of the observed vertical breakthrough curves, both the dispersivity and the vertical
anisotropy (Kb/Ky) were tested. It was discovered that increasing the vertical anisotropy
resulted in a better goodness of fit between the simulated and observed values and that
the slope of the observed values could not be achieved by altering the dispersivity alone,
supporting what was determined during the horizontal bromide tracer calibration. While
packing the physical model with sand, efforts were taken to homogenize the aquifer
media, however, it is possible that settling occurred, creating layers of lower conductivity
and heterogeneities in the model aquifer. The calibrated values for vertical anisotropy

(Kn/Ky), porosity, and dispersivity are 2.0, 0.31, and 0.1 cm, respectively.

Graphs of simulated versus observed concentrations have been generated for the vertical
tracer models (Figures 19-20). Figure 19 illustrates the simulated versus observed
breakthrough curves for the average row 9 concentrations. The fit between these two
data sets is very good. Figure 20 shows the plots of simulated versus observed values for
all of the row 9 sampling ports. These fits are also reasonable, however the breakthrough
of the simulated bromide appears to be consistently a little ahead of the observed
breakthrough for ports 913, 921, and 933. These ports are proximal to the x, y
coordinates of the influent ports, however 76 cm higher. In turn, the numerical
Simulation appears to be over predicting the hydraulic head in the vicinity of the influent
ports. Attempts were made to develop a better fit in ports 913, 921, and 933, by

increasing the amount of transverse dispersion and adjusting the relative flow through the

61

influent ports (i.e. weighting certain influent ports while maintaining a constant influent
flow rate), however the model appeared to be fairly insensitive to adjustments in the
influent flow rates and changing the transverse dispersivity resulted in worse fits for the

other sampling ports.

The observed breakthrough curves for ports 911, 923, and 931 are all ahead of the
Simulated breakthrough curves. The observed data suggests that more flow went through
this region of the model aquifer. This is visible in the series of 2-dimensional
interpolations that were presented in Figure 10. Again, attempts were made to try and
weigh the amount of flow entering through certain influent ports, however this did not

provide a solution.

62

 

 

 

 

 

 

 

0.8 —r
11'
0.6 —
Combined Effluent Port Iii
o i— +—+ Observed I
g 7 W EJ~EJ Simulated n
l
0.4 —~ 1
ti]
1/
0.2 — L"
Li]
.1
"‘ +/ L]
/ II
.. . fiLfi‘lT
0 1 illii‘fiimrrfipi well i 1 1 1 i
0 10 20 30 4O 50
Time (hrs)

Figure 19. Simulated versus observed vertical bromide tracer test results for the
combined vertical effluent port.

63

Figure 20. Simulated versus observed RT3D simulation of the vertical bromide tracer
test at row 9 sampling ports

 

 

 

 

 

 

 

 

1 — ++
/ f“
. 1 ;,
0.8 _. 1 7;
I 1 111
0.6 ’T , I D
9 Sampling Port 911 l 1
Q - ~1— +—I Observed D
0 in [Ii—El Simulated] l 1
0.4 - 1 U
,1
I ii
0.2 a l ,
- Li
/ K ”1;:
O ————,—Enn;rmmnmnwmfimnrmlmmsmml$’ I i
0 1O 20 30 40
Time (hrs)
1 — I}; m
- Iii"
+’
0.8 — ;
_ l
l
0.6 —+ 1,
0 Sampling Port 912
Q - 4— +—+ Observed [1
0 El -— Ll—a Simulated "’
0.4 e l
‘ .1
0.2 e 16'
4‘3
_ / ,/ g;
0 “ ' I I l
0 10 20 30 40
Time (hrs)

64

 

Figure 20 (continued)

 

 

 

 

 

 

_ a?
0.8 ~ ' d“
d [1]
0.6 — , l
o Sampllng Port 913
Q _ +— +—+ Observed I
o D D—U Simulated 31 1
OAi— ’ I
_ u f
0.2 _ [tr /+
_ / *\ [EH/4'
0 r uwmmmmfimimmwwfi l l l
0 1O 20 30 40
Time (hrs)
1 __1 EWMMM
0.8 m I
~ E
0.6 4 . I t
0 Sampling Port 921 U
Q _ -l——- +—+ Obselved 1
0 Emma Simulated] D 1
0.4 — :1 7
q t
.1
0.2 4 t1 /
t A 2’ i
/ a +1
0 _—r—mm;rmmrpmfinw1iwnflmwg l l l

 

O

 

1O 20 30 40
Time (hrs)

65

C/CO

CICO

 

Figure 20 (continued)

 

 

 

 

 

 

 

1 _ fWfllfllflm
‘ it
0.8 ~ .5,
_ 1 1
[1‘] 1
0.6 T [1
Sampling Port 922 1 1
_ +— +—t Observed
[II-F Eli—U Simulated] L; I
0.4 m
- .1 1
Tl
0.2 -—~ :1 +
J 1*
/ a
/ + of"
0 _———y—‘HR1T1TRFH¥HHTTTWHMFBMW l 1 l
O 1 0 20 30 40
Time (hrs)
1 _ T+++ Wm
0.8 — l a
g1
_ 1 T
0.6 — . 1 n
Sampllng Port 923 l
a 1— +——+ Observed 1 U
[a D—Ll Simulated] + 1.1
0.4 —~ / l
a / 55‘
0.2 ‘T / 1;]
_ / / 1 ;J
0 “—1—“1?”“lW“*“*“"‘*‘”’"‘"W“'"“*““’*““m’+¥r ‘ l
0 1 O 20 30 40
Time (hrs)

66

C/CO

C/CO

Figure 20 (continued)

 

 

 

 

 

 

 

 

1 — r+ .
1 try
u / .115“
0.8 -— 1 1?
a l f
0.6 ~ _ I 1
Sampling Port 931 ;1
u +— +—+ Observed I
Le D—EJ Simulated] ;
0.4 ~ I
a l 1
0.2 — l J,
s » ifE
, / +/ ti
0 _—fi—“""l'm‘""'"“*"‘"""'l‘”“"""”i“““m1 ' l
O 10 20 3O 40
Time (hrs)
1 — #Wm
1 f“ '
0.8 —— l1
[5.]
a 1; 1
0.6 a , Tl
Sampllng Port 932
a +— +—+ Observed 7)
l3— D» D Simulated U
0.4 a ;I
u 11
0.2 _ [£1
_ +\ 1; 1
0 —»————.——-mm1rmnmemfinmfnmmm@@ T I l
O 10 20 30 40
Time (hrs)

67

Figure 20 (continued)

 

 

 

 

1 — WWW”
/
‘ 5’1
o 8 a ‘ .1 1
- l‘ 1
0.6 a T 1
0 Sampling Port 933 1
Q - +— +—t Observed l
0 [LJ [3 E] Simulatej [I] .1
0.4 - [1
u l 1
0.2 — 1 I
g It 1
/ KE4
0 —————I—q1m1rmmmmmmnnrmnlumqfl1gr I I I
0 10 20 30 40
Time (hrs)

Iso surfaces were also generated for the horizontal and vertical simulated tracer tests
(Figures 21 and 22) to compare them to the interpolated iso-surfaces created from the
observed data (Figures 7, 9). Figure 21 illustrates the iso-surfaces of the simulated cross-
flow tracer test at 150 and 180 minutes. Figure 22 illustrates iso-surfaces of the
simulated vertical tracer test at 16 and 24 hours. As can be seen from both of these
figures, the simulated iso-surfaces exhibit an idealized distribution of chemical mass with
greater transport occurring proximal to the influent port(s) then what was observed in the
laboratory model. The observed iso-surfaces suggest more spreading of the tracer fronts
as they move towards the effluent port(s). This could be in part attributable to the

interpolations used to create the iso-surfaces or heterogeneities in the model aquifer.

68

However, as was previously mentioned, there was not enough data to account for

possible heterogeneities in the column.

 

Figure 21. Isa-surfaces generated from the RT3D simulation of the horizontal bromide
tracer test at 150 and 180 minutes, compare to Figure 7 for the observed iso-surfaces at
the same times.

69

Vertical Effluent Vertical Effluent

 

IIIIIllImII'
ll III .. “I

I IIHIIIII. ' :1: 'l' I.” III IIIII""11M
'111IIIII II! I.!!.. lll||lll
:II IIII‘!lIIlll ‘

' Lilli! ll.
lIIlllllIlIl’

 

I" ll 1
'1 IIiiIIiIl '1’

Vertical lnfluent Vertical lnfluent

Figure 22. Iso-surfaces generated from the RT3D simulation of the vertical bromide
tracer test at 16 and 24 hours. Compare to Figure 9 for the iso-surfaces generated from
the observed data sets.

5.4 SENSITIVITY ANALYSES

The objective of a sensitivity analyses is to demonstrate the uncertainty in a calibrated
model by varying the input parameters over a reasonable range of values (Anderson and
Woessner 1992). The evaluation often entails noting the change in average hydraulic head
from the calibrated model. In this case, parameter sensitivity was evaluated by looking at

the root mean squared error (RMS) between simulated and observed bromide

70

concentrations at the horizontal effluent port of the cross-flow tracer test and the resulting
simulated versus observed breakthrough curves for the horizontal and vertical bromide
tracer studies. Porosity and dispersivity were tested over a range of values (50% above
and below the calibrated values). The absolute difference between the RMS calculated
from the calibrated model and the RMS determined for each sensitivity model was

plotted against the change from the calibrated parameter (Figure 23).

    

 

 

 

 

 

 

0.25 '—1
T
E. 0.2 _
E
3
0’ ‘1
(I)
c Legend
g 0 15 a I—I——I— Porosity
§ ' G— 0—0 Dispersivity
a:
To z
3
.12
2
g 0.. _
"6
O
_=.
g
9
2
8 0.05 _.
2
/
° Q/ 7 1
~60 -4o -20 o 20 4o 60

Change From Calibrated Value, In Percent

Figure 23. Sensitivity analyses of porosity and dispersivity in the horizontal bromide
tracer study.

71

As can be seen, porosity is the more sensitive parameter in the horizontal bromide tracer
study. This observation prompted an additional sensitivity analyses to determine if
dispersivity was as insensitive in the vertical tracer study as it was in the horizontal tracer
study. Dispersivity was tested over a wide range of values (0.1 — 3.0 cm); however
instead of calculating the RMS error between the two simulations, comparisons were
made between the resulting breakthrough curves (Figure 24a and b). The resulting curves
indicate that dispersivity is fairly insensitive in the horizontal tracer study, yet very
sensitive in the vertical tracer study. This is discussed further in the next section. The
sensitivity analyses supports the estimated value of dispersivity of 0.1 cm, arrived at

during the calibration.

72

0.8 —

 

Legend
+ + + Observed

—— “e Simulated.D=0.1cm
— ‘ —- Simulated,D=1.0 cm ”A
0.6 — —- — Simulated, D=3.0 cm +1] /

z a?
o , / +
Q 0.4 — 1
0 \

 

 

 

 

    

 

 

 

 

 

”\J‘I'
02 ,/++
I ++
_ // .+
//+
/ 7+
I‘ItIC—If tJéJrif III
0 —- 1 1 1 1 l 1 l I l 1
O 50 100 150 200 250
Time (min)
1 _
d f
Legend
0.8 — + + + Observed
———~*r --— Simulated,D=0.1cm
a — . ~ Simulated. D=1.0 cm
K— -— Simulated,D=3.0 cm
0.6 —
0
Q ..
0
0.4 —
0.2 fl
0 _—l 1 1 fl
0 1O 20 30 40 50

Time (hrs)

Figure 24. Sensitivity analyses exhibiting the effects of varying dispersivity on the
horizontal (top) and vertical (bottom) breakthrough curves.

73

5.5 CONSERVATIVE TRACER TEST DISCUSSION

The modeling of the conservative tracer tests provided insight into the nature of solute
transport through the model aquifer. Values of porosity, dispersivity and vertical
anisotropy were estimated based on the groundwater transport model. These parameters
are critical for development of a reactive transport model. Often times environmental
professionals use generic parameters as input into numerical models, which in turn create
solutions that have no bearing on reality. Since the basic flow and transport parameters
have been estimated through the bromide tracer tests, these parameters can be used with
more confidence in the biocurtain numerical model simulations. An additional benefit of
using a numerical model is that it is very easy to adjust parameters and observe the effect,
helping the user to understand the connectivity of different processes. For example,

adjusting the porosity and observing how it affects the breakthrough curve.

The modeling of the conservative tracer indicated that the physical aquifer model was
more complicated than expected. It is clear that treating the aquifer media as
homogenous and pumping rates equal in each of the injection/extraction ports is an
oversimplification; however, as was previously mentioned accounting for heterogeneity
was outside of the scope of this work. The numerical modeling results also indicated that
the numerical model was more complicated than expected. The results exhibited more
chemical mass proximal to the influent port(s) than what was observed in the laboratory
model. More importantly though were the observations drawn fiom the sensitivity
analyses. The results indicated that dispersivity is relatively insensitive in the horizontal

flow regime; however, very sensitive in the vertical flow. This insensitivity may be a

74

result of numerical dispersion. The cause of the enhanced numerical dispersion in the
horizontal test could be due to the divergent, convergent flow that occurs between the
single horizontal injection and extraction ports. The resulting velocity field produces
hydraulic gradients that are diagonal to the finite difference grid cells, along with higher
local velocities than in the vertical test. Having identified these shortcomings, one
expects the data generated from the carbon tetrachloride sorption simulation, inoculation,

and biocurtain simulations to also exhibit these characteristics.

5.6 CARBON TETRACHLORIDE SORPTION SIMULATION

The carbon tetrachloride sorption simulation was set-up in the same manner as the
vertical bromide tracer simulation, with 4 wells representing the influent ports, which
were assigned a quarter of the average observed flow rate, and 4 constant head cells (115
cm) assigned to the effluent ports at the top of the tank. The simulation utilized the
calibrated parameters fiom the bromide tracer tests. Concentrations of CT were assigned
at each influent port and set at the average observed influent concentration entering the
laboratory model. Concentrations of the other reactive species were left at zero since
they were not included in the laboratory experiment, with the exception of bromide,

which was modeled separately and presented earlier.

75

 

 

Parameter Description Value(s) Source
calibration of bromide tracer tests,
or longitudinal dispersivity, cm 0.1, 3.0 CT sorption study

arm, ratio of transverse to longitudinal dispersivity 0.3

amok.

fifygxfieffggff<\ffi§

an
Table 1.

ratio of vertical to longitudinal dispersivity 0.1

sediment porosity 0.31
hydraulic Conductivity, cm sec'1 0.027
microbial decay rate, day'l 0.13
fiaction of exchange sites 0.437
nitrate utilization coefficient, day-1 18.89
attachment rate, day-l 0.9
distribution coefficient, L/mg 3.9 x 10'7
detachment rate, day -1 0.04

half-saturation coefficient for acetate, mg/L 1.0

half-saturation coefiicient for nitrate, mg/L 12.0
second-order CT reaction rate, Umg-day 0.12
first order kinetic (de) sorption rate, day -l 0.36

maxinnun specific growth rate, day -1 3.11
soil bulk density, mg/L 1.63 x 106
yield for acetate, mg cells/mg substrate 0.4

yield for nitrate, mg cells/mg substrate 0.25
yield for biomass, mg cells/mg substrate 0.46

List of numerical model input parameters.

calibration of bromide tracer tests

. calibration of bromide tracer tests

calibration of bromide tracer tests
Dybas et al.(2002) and Hyndman et al.
(2000)

Phanikumar et al. (2005)
Phanikumar et al. (2005)
Phanikumar et al. (2005)

Phanikumar et al. (2005)
Zhao et al. (1999)
Phanikumar et al. (2005)
Phanikumar et al. (2005)

Phanikumar et al. (2005)
Phanikurmr et al. (2005)
Phanikumar et al. (2005)

Phanikumar et al. (2005)
Phanikumar et al. (2005)

Phanikumr et al. (2005)
Phanikumar et al. (2005)
Phanikurmr et al. (2005)

Recalling equations (1) and (6) it was necessary to define a set of reaction parameters for

this simulation, specifically values for bulk density (p), first order kinetic (de) sorption

rate (1c), fraction of exchange sites at equilibrium (f), and the distribution coefficient

(1(4). The initial parameters were gleaned from the recent work conducted by the

Departments of Geological Science and Civil and Environmental Engineering

(Phanikumar et al., 2005), in which they applied their biodegradation model to the

Schoolcraft site (refer to Table 1 for parameters). A simulation was executed and the

76

results were compared to the observed data. Figure 25 illustrates the average row 9
simulated versus observed data utilizing the initial parameters. As can be seen, the 50%
breakthrough of the simulated CT matches fairly well with the 50% breakthrough of the
observed data. This was also the case with the vertical effluent port. Moreover the
simulated CT breakthrough curves reach asymptotic values of C/Co, which were fairly
consistent with the observed data. However, the slopes of the simulated and observed
breakthrough curves do not agree. Therefore, a sensitivity analysis was performed on
parameters thought to have an effect on the slope of the breakthrough curves. Values of

K, f, and K, were perturbed from their initial conditions and the effects were noted.

77

Average Row 9 Concentrations

 

 

 

 

 

08 —~ \ ,
/ K .
0 6 — J
a .. f /
o
0.4 —
/
1
0.2 —~ Legend
/ [i— +——i‘ Observed
- Simulated
° r r" l I l T l ' l
0 50 100 150 200
Time (hrs) .

Figure 25. Simulated versus observed concentrations at row 9 using the initial reaction
parameters from Phanikumar et al. (2005) and the calibrated parameters from the tracer

simulations.

5. 6. 1 CT Sensitivity Analysis

The first parameter that was adjusted in an attempt to better fit the observed the

breakthrough curve was the first order kinetic (de) sorption rate. This parameter was

reduced to a fifth of its initial value and the simulation was run. The effect of decreasing

K appears to increase the asymptotic value of the dissolved phase CT; however, does not

have an effect on the slope of the breakthrough curve (Figure 26). Conversely increasing

this parameter decreased the asymptotic value of C/Co. All of the following simulations

78

were terminated prematurely for the sake of time, when it was evident that the parameter

perturbation did not produce the desired effect.

‘7

—<

0.8 ——*

C/CO
l

 

 

~_.- q... - - ,

_~ - — ~—

Average Row 9 Coneentratio ns

“ n 5 cv— h

 

 

 

 

  

Legend

Simulated (A), iritiel parameters

Simulated (B), decreased k by 5x
Simulated (C), decreased f by 2x
Simulated (D), decreased retardation
Simulated (E), increased Kd by 5x
Simulated (F), increased Kd by 2x

 

J

 

 

150

Time (hrs)

Figure 26. Simulated CT breakthrough curves resulting from perturbations to sorption

parameters.

The second parameter that was perturbed was the fiaction of exchange sites at

equilibrium. This parameter was reduced by half, thus reducing the retardation value

fiom 1.89 to 1.45 as well, since f is also a term in the determination of R. The effect of

this perturbation was that it reduced the time required to achieve breakthrough and

decreased the asymptotic value of C/Co; however, no change to the slope was observed

79

(Figure 27). A simulation was also run that utilized the initial parameters, but reduced
the retardation value to 1.45. The result indicated that a reduction in the retardation
translates the breakthrough curve to arrive earlier without affecting the slope or the

asymptotic value of C/Co (Figure 26).

The third parameter that was investigated was the distribution coefficient. According to
Dybas et al. (2002) values of K1 vary over three orders of magnitude in the field,
therefore there is a great deal of uncertainty in this parameter and thus perturbations to
this parameter are justified by field data. The first simulation increased Kd by a factor of
five and thus the retardation value was increased to 5.48. The effect of this change was
that it drastically decreased the slopes of the breakthrough curves and never achieved
asymptotic values (Figure 26). The results looked favorable though, since this was the
first parameter perturbation that resulted in a change of slope. Therefore, additional
simulations were run using increased values of K1 (2x and 1.4x the initial value);
however, neither simulation produced the desired effect. In both cases the slope of the
breakthrough curves increased dramatically and did not match the observed data, the
center of mass arrivals were late, and the asymptotic values of C/Co were too low.
Simulations were also explored that varied several parameters at the same time and set-up
multiple stress periods corresponding to the observed fluctuations in influent
concentrations and flow rates; however, none of these simulations resulted in better fits
between the simulated and observed. Therefore, one last parameter was investigated that

is known to affect the shape of a solute breakthrough, that being dispersivity.

80

 

 

 

 

I 7
. 71
a f \11
“ML-WW- - -_- --
__ J’_::"W-———+
0.6 — //
/
_ /
0.4 _ 1/
/ Sampling Port 932
A %— +-—+ Observed
———— Simulated
0.2 -— /
//
_ //
o _.._..-_-. -_.-. ,1 z/ 1 I 1 I I I I I 1
0 4o 80 120 160 200

Figure 27. Simulated versus observed carbon tetrachloride concentrations during the CT
sorption study at sampling port 932.

Recall that dispersivity was calibrated during the horizontal and vertical bromide tracer
simulations; however, this parameter was re-visited to determine if increasing its value
would create the desired effect on the slope of the CT breakthrough curves. Utilizing the
initial reactive transport parameters that resulted in acceptable fits between the simulated
and observed center of mass arrival times and the asymptotic values reached by the CT
breakthroughs, several simulations were run with increased values of dispersivity. The
results indicated that a dispersivity of 3 cm creates the desired affect on the slope of the

CT breakthrough curves and that a decrease in the retardation value to 1.58 was required

81

to match up the arrival time of the breakthroughs. Moreover, the initial reaction
parameters appear to adequately describe the CT transport and sorption. Refer to Figure
27 for an example of simulated versus observed at sampling port 932 and Table l for the

reaction parameters used in this simulation.

5.6.2 CT Smption Simulation Discussion

The results of the CT sorption simulation sensitivity analyses were very informative in
regard to the nature of the two-site sorption model developed by Zhao et a1 (1999) and
incorporated into the RT3D model developed by Phanikumar et al (2002a, 2002b, 2003,
and 2005). The expectation was that by adjusting the first order kinetic (de) sorption rate,
fraction of exchange sites at equilibrium, or the distribution coefficient, one would
achieve a better fit between the simulated and observed CT breakthroughs. This was not
the case when the parameters were independently adjusted; however, if a multi-parameter
optimization was performed one may arrive at a better fit. In order to produce a better fit
between the simulated and observed an adjustment of the dispersivity and retardation was
required. Conceptually this approach was flawed, for though adjusting the dispersivity
created the desired effect; in general the amount of dispersion is ofien considered the
same for a non-reactive solute as it is for a reactive solute. Unless diffusion or size
exclusion plays a significant role in the amount of dispersion, then differential tracer
breakthroughs can be expected, as indicated during microcosm studies conducted at
MSU. (Ewer, 1997; Bremen, 2004). However, neither of these factors should have

influenced the CT transport.

82

Another possibility is that the current model does not adequately account for the kinetics
of the CT sorption. Observing the laboratory CT breakthrough curves it is apparent that
changes in the slope of the curves occurred that are inconsistent with the bromide
breakthrough curves. This could be an effect of the uncertainty in the dataset (fluctuating
influent concentrations and flow rates) or a factor that was unaccounted for in the
sorption model, like spatial variability in the CT sorption rate. Exploring the idea of a
multi-rate sorption model has been explored by Haggerty and Gorelick (1995) and may

require additional work through the use of 3-dimensional sorption studies.

An additional, and perhaps more plausible explanation for the observed CT breakthrough
curves, is that there is a spatial distribution to the distribution coefficient. The packing of
the column likely caused an uneven distribution of fraction organic carbon, for as
sediment was wet packed into the model and fell through the column of water, the coal
particles may have settled out more slowly, based on their grain size and density.
Evidence of this can be seen in the initial arrival times of bromide and CT. If there were
an equal amount of retardation throughout the column, then the initial arrival of bromide
and CT would not match. The fact that it does, suggest that fastest flow path within the
column may not have had any fraction organic carbon to allow sorption to occur. In turn,
by varying the distribution coefficient throughout the model and utilizing inverse
modeling techniques, one may arrive at a solution that more closely agrees with the

observed data.

83

The numerical modeling exercise moved forward utilizing the initial CT sorption reaction
parameters with the adjusted retardation value and increased dispersivity. The next

simulation conducted was the inoculation of the laboratory model.

5.7 INOCULATION SIMULATION

The inoculation simulation was set-up in the same manner as the horizontal bromide
tracer study. A horizontal flow rate was established at the horizontal influent port that
matched the average flow rate observed during the inoculation. Concentrations of
bromide, acetate, nitrate, and liquid phase strain KC matched the measured influent
concentrations used in the laboratory study, refer to section 4.7. CT was not included as
an injected species since concentrations were negligible due to CT degradation in the
inoculum prior to inoculation. Concentrations were simulated in the units of mg/L.
According to Phanikumar et al. (2002), one cfu/mL of strain KC is roughly equal to 1.67
x 10'7 mg/L; therefore, the measured biomass concentrations were converted to ppm.
The simulation lasted approximately 3 hours and utilized all of the initial reaction
parameters as presented in Phanikumar et a1 (2005), with the exception of the retardation
value, which was altered based on the CT sorption simulation results. The model was
also run with the increased dispersivity value, as determined through the CT sorption
study and the porosity, as determined through the bromide tracer tests. Refer to Table 1

for a complete list of all of the reactive transport parameters.

84

The results indicate a poor agreement between simulated and observed mobile strain KC
concentrations (Figures 28). Both the simulated mobile strain KC and bromide exhibit
early breakthroughs along the centerline sampling ports (312, 322, and 332). This was
expected based on the results of the horizontal bromide tracer simulation, in which the
flow and non-reactive transport model was calibrated to the horizontal effluent
concentrations. Recall that this approach was chosen to try and match the average solute
concentrations leaving the laboratory model aquifer in lieu of attempting a calibration to
the sparse data collected from the internal sampling ports. Though the simulated
concentration histories appear far earlier than the observed, the slopes of the
breakthrough curves at 312 and 322 are reasonable; however, the slope of the simulated
breakthrough curve at 332 is too gradual. Comparisons were not made at the horizontal
effluent, since mobile strain KC was only just starting to arrive at the end of the

inoculation.

85

1.2 ‘7

     

 

0.8 — /

Leg end _
O“- O 690 one 312 O
": /> 038 322 /
V—-V V 088- 332

 

 

 

o
8 A —-~— Sim-312
Sim-322
J k————— Sim-332J XI
0.4 —

 

. _.I_-- . .-

 

Time (min)

Figure 28. Comparison of simulated versus observed mobile strain KC concentrations
during the inoculation event.

The most important observation to come from this simulation involved the relative timing
of breakthroughs. As was discussed in Section 5.1.3 there was a significant lag time
between the mobile strain KC and the bromide breakthrough curves; however, this
differential breakthrough did not occur in the inoculation simulation. Instead, the
simulated breakthrough curves exhibit almost simultaneous arrivals of Br and mobile
phase KC (Figure 29). This simultaneous arrival had previously been observed during
past model applications (Phanikumar et al., 2002a, Phanikumar et al., 2005). The only

difference was that instead of seeing mobile phase KC and bromide breakthroughs

86

occurring together, the authors displayed graphs of mobile phase KC and acetate
concentrations versus time. In fact in the application of the model to the one-dimensional
column studies (Phanikumar et al., 2002a), the simulated mobile phase KC is transported
faster than acetate. The comparison between mobile strain KC and acetate and mobile
strain KC and bromide can be made, since in the above referenced articles, the authors

noted that acetate and bromide exhibit similar transport patterns.

 

 

   

 

 

 

 

 

1.2 ——
_ ~v __ '/—,w - — ——
0.8 —
Port-332
O
Q a
0
0.4 —~ Legend
— : Mobile KC
—— _fl'—-‘ Bromide
0 “i 1 l j l 1 1
O 80 120 160

Time (min)

Figure 29. Comparison of bromide and mobile strain KC during the inoculation event.

Previous studies (Murphy et al., 1997; Johnson et al., 2001; Phanikumar et al., 2005)

have noted the effect of aquifer heterogeneities on the timing of microbial transport.

87

Though the laboratory model may have contained small-scale heterogeneities, due to
settling of sediments during wet packing, they would have existed as layers parallel to the
centerline sampling ports between the horizontal influent and effluent. Therefore, the
velocity field established directly between the horizontal influent and effluent should not
have been affected by the occurrence of possible heterogeneities. Another factor must be
responsible. A simulation was run with an increased attachment rate (K...) to see if
increasing that parameter would delay the transport of strain KC. This idea had been
used previously by Phanikumar et al. (2005) to account for the increased attachment rate
due to the inoculation of a flocculated culture; however, the perturbation only increased
the asymptotic value reached during the breakthrough. Another simulation was run
utilizing the optimized reaction parameters solved for in the one-dimensional column
study (Phanikumar et al., 2002a). This simulation utilized a higher value for the
microbial decay rate; however, no change was observed in the mobile strain KC
breakthrough curves or the relative breakthroughs of the mobile strain KC and bromide.
Therefore, the data suggests that there is an additional factor controlling the lag time
between the mobile KC and a conservative tracer, perhaps something similar to a
retardation factor. Evaluating an additional reaction parameter is beyond the scope of

this thesis; though should be considered as microbial transport models continue to evolve.

5.8 NUMERICAL BIOCURTAIN SIMULATION

The biocurtain simulation was designed to mimic the 3 stages of the laboratory model

biodegradation study following the inoculation event, as described in Section 4.8. This

88

included the first biodegradation period, the feeding event, and the second biodegradation
event. The reactive parameters used in this simulation were the same as were utilized in
the recent model application to the Schoolcraft site (Phanikumar et al., 2005), with the
exception of porosity and dispersivity. Refer to Table 1 for a complete list. Starting
concentrations of acetate, nitrate, dissolved CT, sorbed CT, mobile phase biomass, and
immobile phase biomass were established that corresponded with the concentrations at
the end of the inoculation event. It was the hope to utilize the results of the inoculation
simulation as the starting concentrations for the biocurtain simulation; however, ensuing
fi'om the poor fit between simulated and observed the author decided to use the available
laboratory data instead. The only exception was for the initial concentrations of the
immobile phase biomass, which did utilize the concentrations from the end of the

inoculation simulation, for a good analog did not exist.

The observed bromide data, which had previously been kriged to the finite difference grid
(Figure 15), was used as a proxy for the distribution of the starting acetate and nitrate
concentrations by multiplying the C/Co Br values in each cell by the injected
concentrations of nitrate and acetate. The kriged mobile strain KC at the end of the
inoculation was input as the starting concentrations for this simulation. The kriged CT
concentrations at 1.5 hours after the start of the biocurtain study (Figure 17) were utilized
as the starting concentration of dissolved CT in each grid cell. The starting sorbed phase
CT concentration in each grid cell used the above distribution of the dissolved phase CT
concentrations and assumed equilibrium sorption. Therefore, the value in each grid cell

was multiplied by the distribution coefficient and the fraction of exchange sites.

89

Results of the biocurtain simulation indicate a high degree of CT degradation throughout
the column. The simulated CT concentrations exhibit a much higher degree of mass
removal than what was observed; however, given the numerous difficulties in both the
laboratory model (fluctuating CT concentrations over time, decrease in pH versus time,
etc.) and the numerical model, a quantitative analyses of these results would be entirely
speculative. Instead, it is more appropriate to observe general distributions of mass and

relative timing of species transport in qualitative terms.

In order to observe the relative timing of species transport at two locations in the column,
graphs of the concentrations of each of the main reactive components have been prepared
for sampling ports 312 and 722 (Figure 30). Sampling Port 312 was selected to illustrate
concentrations proximal to the horizontal influent port along the centerline of the
horizontal velocity field. Sampling port 722 was chosen to exhibit concentrations

towards the top of the column along the centerline of the vertical velocity field.

90

Figure 30. Comparison of simulated concentrations for sampling ports 312 (top) and 722
(bottom) during the biocurtain experiment.

Sampling Port 312
55 —

50 *‘ .
Legend
45 — <> 0 0 CT-Observed

—---— CT-Simulated
a

 

 

335E
g30— 0

£25—

8 2o——-

1..

0
15—
10—

5_.

 

 

 

0 I} 1 1 r I

100 200 300
1 Time (hrs)
End of Inoculation (0 hrs) Feeding Event (144 hrs)

—>o

Sampling Port 722

 

25 _9 Legend
<> <> 0 CT-Observed
-———— CT-Simulated

 

CT Concentration (ppb)
5: B
1 .L

8
l

- .-...M..----Q. --_-
o T r
o 100 200

1‘ Time (hrs)
End of Inoculation (0 hrs) Feeding Event (144 hrs)

 

’I/Ir
5 a / \
_ \\"ev\._- ,E
' 1

 

91

CICO

C/CO

Figure 30 (continued)

 

 

 

 

 

 

 

Sampling Port 312
1
I I
I Legend
. Acetate
0.8 ‘d - --—-— Nitrate
0.6 —
0.4 —<
0.2 _
\.-_ I \--
0 i I i T i
100 200 300

1

Time (hrs)

End of Inoculation (0 hrs) Feeding Event (144 hrs)

1..

 

 

 

 

Sampling Port 722

 

 

 

 

 

Legend
~ *- Acetate

0.8 4”” ~ Nitrate
0.6 ~ \

0.4 ~

0.2 fl

0 I i 1:\IV 1? 1 \\i\ 1
100 200

i

1

Time (hrs)

End of Inoculation (0 hrs) Feeding Event (144 hrs)

92

Figure 30 (continued)

 

 

 

 

 

 

 

 

 

 

 

 

Sampling Port 312
35 _
w H \
25 —
e \
gm“ ‘
g
E 15 -
8 I Legend
KC-MOBILE
10 A KC-IMMOBILE
5 __.
0 r‘f'flf l 1 —i— 1
0 100 200 300
1‘ Time (hrs)
End of Inoculation (0 hrs) Feeding Ever! (144 hrs)
Sampling Port 722
35 —
30 .._.
25 — Legend
E "- KC-MOBILE
a v.-. —-——~—— KC-IMMOBILE
5 2° “
g
8 \‘x
l T “i“ If 7 ' “l
100 200 300
I Time (hrs)

 

End of Inoculation (0 hrs) Feeding Event (144 hrs)

93

Nitrate and acetate concentrations were normalized by the initial concentrations for each
phase, so that they could be viewed together. CT concentrations were expressed in ppb
and strain KC was expressed in ppm. Based on the CT concentrations at port 312 and
722, the model predicts a high degree of CT degradation, which occurred very quickly, as
indicated by the mass removal of CT after the inoculation event. Concentrations of
acetate exhibit a decline that corresponds with the growth of immobile strain KC. The
electron donor was completely consumed after 100 hours of the experiment; at which
point the immobile strain KC concentrations began to decrease, indicating microbial
decay. The nitrate concentrations exhibit a sharp decline soon after the inoculation and
feeding event. In the case of sampling port 722, the simulation exhibits almost no nitrate
reaching this location after the feeding event. This almost instantaneous decline in nitrate
has also been noted in the field (Dybas et al., 2002, Phanikumar et al., 2005). Based on
these model results, it appears as if this electron acceptor is a rate-limiting factor. And
though the model still demonstrated CT mass removal without nitrate in the system, the
microbial population was declining, which would eventually halt the ability to transform
CT except near the upgradient source from the bottom of the column. Ifnitrate data was
available for use during this thesis, one may find that a decrease in the nitrate utilization
coefficient, that accounts for indigenous microbial activity or a decrease in the growth
rate, is appropriate. Lastly, the simulated microbial concentrations support what has
previously been noted by Dybas et al. (2002) and Phanikumar et al. (2005), that the
immobile microbes exhibit much higher concentrations than the mobile microbes.
Furthermore, the sharp decline in the concentration of mobile microbes is coincident with

the sharp decline in the nitrate concentrations.

94

43.8 Vertical Effluent

12:2 lI'IIIII :lllIlllfll

. 111
6.3 1111 111 IIIIIIlIiIIIIIII Illlilllilll 1i
3-1 '11-1llll'" 1 1111 lllllllIIIlII In

I "1 11111111111111

II -- ‘ II
III-.1111 IlllI
1Illll'

IIII ! Horizontal
'III .
IIIIIIII .. .lI Effluent

 

Horizontal
lnfluent

 

i Vertical lnfluent
x

Figure 31. Concentrations of CT represented by iso-surfaces at day 12 of the simulation.

A plot of iso-surface concentrations was prepared illustrating the CT concentrations at the
end of the 12 days (Figure 31). The distribution of CT remaining in the column agrees

with the general observed distribution of CT, as indicated in Figure 17. Recall that when

95

evaluating the observed data it was speculated that the greater degree of degradation near
the horizontal influent port resulted from a more biologically active zone. This agrees
with the simulation, as illustrated by iso-surfaee plots of the simulated mobile and
immobile strain KC after 12 days (Figure 32 and 33). Therefore, as stated earlier, in
order to create an effective biocurtain, it is critical to ensure that there is a sufficient
enough distribution of microbes that the bioactive zone completely covers the region of
the aquifer between injection and extraction wells.

01% div! (PM)
0.7 Vertical Effluent

Horizontal
Effluent

Horizontal
lnfluent

 

1 Vertical lnfluent
x

Figure 32. Concentrations of mobile strain KC represented by iso-surfaces at day 12 of
the biocurtain simulation.

96

M295; 2 div! (PM)
2518 Vertical Effluent

IIII!“"'1IIII:
74 III” 11' "'i:IIIlI"::I

Horizontal
Effluent

Horizontal
Influent

 

i Vertical Influent

Figure 33. Concentrations of immobile strain KC represented by iso-surfaces at day 12
of the biocurtain simulation.

Though there was inadequate data for a more complete comparison between simulated
and observed reactive species, the application of the 3-D flow and transport model to the
laboratory biocurtain experiment provided insight into the nature of the remediation of

Schoolcrafi Plume A. The interdependence of nitrate, acetate, mobile and immobile

97

biomass, on the degradation of dissolved and sorbed carbon tetrachloride was made clear
through the numerical modeling. Unfortunately, due to issues with the laboratory
experiment and the difficulties in calibrating the numerical flow and transport model, a
detailed analysis of the reaction parameters was not possible. Therefore, one cannot state
if the parameters formerly optimized for the one-dimensional column studies by
Phanikumar et al. (20023 and 2003) are applicable for this three-dimensional laboratory
model. The results do indicate that the numerical simulation over predicted the amount
of CT degradation, which suggest that a decrease in the second order CT reaction rate
may be appropriate. When this model was applied to the field-scale biocurtain,
Phanikumar et al. (2005) also noted an over prediction of CT degradation, though an
acceptable prediction of acetate and nitrate. They found that reducing the CT degradation
term by half produced a much better fit between simulated and observed. Again, this was

not investigated here, given the inadequacies in the datasets.

98

6 DISCUSSION

The three-dimensional model aquifer experiment was designed to mimic the Schoolcraft
biocurtain and serve as a link between spatially limited one-dimensional column studies
and the field scale application. The specific goals of this thesis were to (a) improve our
understanding of how well Pseudomonas stutzeri KC was transported in the subsurface,
(b) to visualize the spatial distribution of biomass after an inoculation event, and (c) to
conduct a case study utilizing the user-defined reactive transport model developed by the
Departments of Geological Sciences and Civil and Environmental Engineering at
Michigan State University that accounts for the three-dimensional microbial transport and
biodegradation associated with the Schoolcraft project. Although there were many
problems encountered during the study, the three-dimensional model aquifer experiment
did produce several sets of usefirl data to evaluate the effectiveness of microbial transport

and to examine the user-defined reactive transport model.

The laboratory model aquifer study began with a series of tests designed to characterize
the aquifer media and gain insight into parameters effecting the flow and transport of
dissolved species. Horizontal and vertical conservative tracer tests were conducted, as
was a CT sorption study. The results of these tests were evaluated and later modeled
using MODFLOW and a user defined RT3D model accounting for the three-dimensional
reactive transport. The numerical model provided estimates for the porosity and
dispersivity, which were later used in the inoculation and biocurtain simulations. Of key
importance was the discovery of sensitivity of dispersivity in the vertical velocity field,

but not in horizontal flow direction. This may be the result of the unique velocity field

99

generated by the divergent/convergent flow between the horizontal influent and effluent
ports. This flow regime exhibited signs of numerical dispersion, resulting from the
primarily non-orthogonal flow across the finite difference grid cells. Discovering this
problem through the conservative tracer test simulations was important, for this
phenomenon would be present in the inoculation and biocurtain simulations as well;
however, would have been more difficult to identify since there are many more
parameters to consider when modeling these scenarios. The indication of a directional
insensitivity to dispersivity should be fimher investigated if multi-directional flow and

transport models continue to be utilized.

Another interesting aspect of this project was the CT sorption study. Based on the
comparison between the bromide and CT breakthrough curves there appears to be a
process that is currently not accounted for in the reactive transport model. As noted
earlier, there may be a need to investigate a multi-rate sorption model or a model that

accounts for a spatial variability of the distribution coefficient.

The inoculation event was perhaps the most important aspect of this model aquifer
experiment, for according to the literature there is a lack of spatiotemporal microbial
transport data Our data clearly demonstrates that strain KC can be effectively distributed
through saturated porous media; though, the results indicate that transport is slower than a
conservative tracer. This is an important distinction to note, since the effectiveness of
bioaugmentation depends largely on the distribution of the microbes, therefore when

designing a biocurtain, engineers and scientists should be mindful of the biomass

100

concentrations and not assume complete coverage of a bio-active zone through the

detection of a conservative tracer.

The inoculation data also provided the opportunity to teSt how well the user—defined
microbial transport and biodegradation model predicted the transport of strain KC.
Simulations predicted an almost simultaneous arrival of the conservative tracer and
mobile phase biomass using past reactive transport parameters gleaned from the
literature, which was clearly not the case for the observed data. This suggests that there
is an additional factor controlling the observed lag time between the mobile strain KC
and a conservative tracer, such as a retardation factor. This observation should be

considered as microbial transport models continue to progress.

The laboratory biocurtain study demonstrated successful mass removal of carbon
tetrachloride and provided an interesting case study for application of the microbial and
biodegradation model. Resulting from the scarcity of available datasets for comparative
purposes and the problems encountered during the laboratory study, in-depth qualitative
analysis of the biocurtain simulation was not possible. Instead the simulated biocurtain
was used to provide insight into the rates and processes involved with bioaugmentation.
The interdependence of the reactive species, as described by the coupled partial
differential equations, is clearly visualized in the results of the simulated biocurtain.
Furthermore, the simulation demonstrates the need for effective distributions of electron
acceptors and donors, as well as microbial concentrations to maintain successful

bioaugmentation. Though there is little data to compare the simulation to, the results do

101

suggest an over-prediction, which would suggest the need to decrease the CT degradation

rate.

Many problems arose during the laboratory experiment and subsequent numerical
modeling. The first problem encountered involved the ill-fitting top of the flow cell,
which required a 30 degrees rotation to create a watertight seal, thus contributing to the
non-uniform velocity field in the vertical flow direction. A second problem encountered
was the inability to maintain constant influent flow rates. This was due to occasional
partial plugging of influent lines and tubing wear during peristaltic pump use. More
problematic though, were the fluctuating influent concentrations caused by changes in the
volume of water within the solute amendment tank. Lastly, the lack of pH control during
the biocurtain experiment proved to be the problem that truly ended the laboratory study.
Many of these problems could be prevented in future three-dimensional laboratory
aquifer studies, as described in Vidal-Gavilan (2000). Specifically additional attention
should be placed on measuring and maintaining flow rates, measuring hydraulic heads
across the velocity field to generate groundwater flow gradients, utilization of pumps
instead of reliance on gravity, placement of finer screens in influent and effluent ports,

closed-loop recirculation, etc.

102

APPENDICES

103

APPENDIX A

Horizontal Tracer Test

Bromide Concentrations at Internal Sampling Ports (C/Cn)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Location X Y 2 30min 60min 90min 120 min 150 min 180 min
111 15.24 15.24 15.24 0.77 0.80 0.97
112 15.24 30.48 15.24 0.79 0.80 1.00'
113 15.24 45.72 15.24 0.79 0.87 1.00“
121 30.48 15.24 15.24 0.77 0.84 0.86
122 30.48 30.48 15.24
123 30.48 45.72 15.24 0.74 0.96
131 45.72 15.24 15.24 0.10 0.45
132 45.72 30.48 15.24 0.09 0.66
133 45.72 45.72 15.24
212 15.24 30.48 22.86 0.80 0.67 0.66 0.77 0.87 1.00*
222 30.48 30.48 22.86 0.57 0.59 0.83 0.80 1.00*
232 45.72 30.48 22.86 0.23 0.77 0.93 0.96
311 15.24 15.24 30.48 0.90 0.80 0.66 0.77 0.91 1.00
312 15.24 30.48 30.48 0.96 0.70 0.75 0.83 0.91 0.97
313 15.24 45.72 30.48 0.78 0.78 0.72 0.77 0.84 1.00'
321 30.48 15.24 30.48 0.20 0.61 0.79 0.84 1.00*
322 30.48 30.48 30.48 0.75 0.69 0.77 1.00'
323 30.48 45.72 30.48
331 45.72 15.24 30.48 0.07 0.19 0.93 0.96
332 45.72 30.48 30.48 0.72 0.86 1.00 0.96
333 45.72 45.72 30.48
412 15.24 30.48 38.1 0.78 0.78 0.66 0.83 0.91 0.93
422 30.48 30.48 38.1 0.78 0.72 0.83 0.80 1.00'
432 45.72 30.48 38.1 1.00“
511 15.24 15.24 45.72 0.77 0.87 0.86
512 15.24 30.48 45.72 0.79 0.84 0.93
513 15.24 45.72 45.72 0.86 0.77 1.00*
521 30.48 15.24 45.72 0.83
522 30.48 30.48 45.72 0.86 0.93 1.00"
523 30.48 45.72 45.72 0.83 0.97 0.89
531 45.72 15.24 45.72
532 45.72 30.48 45.72
533 45.72 45.72 45.72
612 15.24 30.48 53.34 0.84 1.00‘
622 30.48 30.48 53.34 0.90
632 45.72 30.48 53.34

 

 

 

 

 

 

 

 

 

 

104

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Location X Y Z 30 min 60min 90min 120 min 150 min 180 min
711 15.24 15.24 60.96 0.37 1.00"
712 15.24 30.48 60.96 0.73 0.93
713 15.24 45.72 60.96 0.80 1.00‘
721 30.48 15.24 60.96 0.05 0.09
722 30.48 30.48 60.96 0.10 0.34
723 30.48 45.72 60.96 0.09 0.09
731 45.72 15.24 60.96 0.06 0.05
732 45.72 30.48 60.96 0.05 0.04
733 45.72 45.72 60.96
812 15.24 30.48 68.58 0.96
822 30.48 30.48 68.58 0.04
832 45.72 30.48 68.58 0.03
911 15.24 15.24 76.2 0.10
912 15.24 30.48 76.2
913 15.24 45.72 76.2 0.03
921 30.48 15.24 76.2 0.03
922 30.48 30.48 76.2 0.02
923 30.48 45.72 76.2
931 45.72 15.24 76.2
932 45.72 30.48 76.2
933 45.72 45.72 76.2

Hor_lnfluent 0 30.48 30.48 1.00' 1.00* 1.00" 1.00* 1.00* 1.00’

Hor_Effluent 60.96 30.48 30.48 0.02 0.02 0.09 0.28 0.41 0.67

 

 

 

 

 

 

 

 

 

 

105

 

Horizontal Tracer Test
Bromide Concentrations at Effluent Sampling Port (C/Co)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Tlme Quin) Horizontal Effluent Port
0 0.014
4 0.017
9 0.016 '
14 0.012
19 0.015

24 0.017
29 0.018
34 0.015
39 0.023
44 0.023
49 0.021
54 0.021
59 0.022
64 0.022
69 0.024
74 0.033
79 0.064
84 0.089
89 0.094
94 0.162
99 0.162
104 0.224
109 0.237
1 14 0.279
1 19 0.294
124 0.347
129 0.347
134 0.408
139 0.347
144 0.408
149 0.480
154 0.507
159 0.565
164 0.565
169 0.431
174 0.665
179 0.597
184 0.630
189 0.665
194 0.665
199 0.783
204 0.702

 

106

APPENDIX B

Vertical Tracer Test
Bromide Concentrations at Internal Sampling Ports (C/Co)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0hr 16hr 24hr 30hr 31 hr 32hr 33hr 34hr 35hr 36hr

111 1.00* 1.00*
1 12 1.00* 0.90
1 13 1 .00* 0.97
121 1.00* 0.94
122 1.00“ 0.90
123 1.00* 0.94
131 1.00" 0.99
132 1 .00‘ 0.99
133 0.98 0.94
212

222

232

311 1.00* 0.97
312 1.00* 1.00*
313 0.52 0.94
321 0.88 0.94
322 1.00* 0.90
323 1.00* 0.97
331 0.98 0.94
332 0.92 0.94
333 0.95 0.94
412

422

432

51 1 0.22 1.00*
512 0.14 0.97
513 0.09 0.74
521 0.21 0.80
522 0.73 0.97
523 0.82 0.94
531 0.32 0.99
532 0.14 0.94
533 0.12 0.94
612

622

632

 

 

 

 

 

 

 

 

 

 

 

107

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Location 0hr 16hr 24 hr 30hr 31 hr 32 hr 33 hr 34hr 35 hr 36hr
711 0.14
712 0.17
713 0.12
721 0.16 0.13
722 0.09 0.57
723 0.07 0.83
731 0.51
732 0.22
733 0.10
812 0.70 0.96 1.00* 1.00" 1.00*
822 0.80 0.95 1.00* 1.00* 1.00’ 1.00‘
832 0.51 0.73 1.00 1.00" 1.00* 1.00*
911 0.09 0.20 0.60 0.88 1.00* 1.00“
912 0.09 0.35 0.58 0.85 1.00' 1.00'
913 0.22 0.31 0.48 0.70
921 0.13 0.05 0.05 0.14 0.26 0.39 0.58 0.82
922 0.07 0.19 0.54 0.86 1.00* 1.00” 1.00‘
923 0.08 0.43 0.76 1.00* 1.00* 1.00”” 1.00*
931 0.06 0.12 0.35 0.76 0.95 1.00' 1.00’
932 0.07 0.03 0.03 0.20 0.52 0.59 0.83 1.00*
933 0.07 0.04 0.18 0.44 0.71 0.83 0.96 1.00'
Vert_lnfluent1.00* 0.98 0.90 1.00*
Vert_Effluent 0.02

 

 

 

 

 

 

 

 

 

 

 

108

 

Vertical Tracer Test
Bromide Concentrations at Effluent Sampling Port (00;)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Time (hrs) Vertical Effluent Port
24.00 0.02
36.58 0.14
36.83 0.21
37.08 0.26
37.33 0.30
37.58 0.32
37.83 0.36
38.08 0.39
38.33 0.40
38.58 0.47
38.83 0.47
39.08 0.51
39.33 0.57
39.58 0.55
39.83 0.60
40.08 0.67
40.33 0.70
40.58 0.70
40.83 0.72
41.08 0.78
41.33 0.85
41.58 0.82
41.83 0.92

 

 

 

109

APPENDIX C

Carbon Tetrachloride Sorption Study
CT Concentrations at Internal Sampling Ports (C/Co)

 

Location 24 hr 48hr 72 hr 120 hr 146.5 hr 160.75 hr 184.25 hr 241 hr

 

 

 

 

 

 

 

 

 

 

 

 

111 0.65 0.77 0.70 0.80 0.68 0.82
112 0.92 0.76 0.85 0.85 0.79 0.78
113 0.84 0.75 0.90 0.77 1.00* 0.75
121 0.69 0.77 0.81 0.79 1.00* 0.78
122 0.79 0.87 0.83 0.93 0.72 0.88
123 0.89 0.91 0.81 0.79 0.69 0.77
131 0.71 0.69 0.75 0.67 0.56 0.82
132 0.69 0.74 0.61 0.66 0.85 0.84
133 0.72 0.72 0.87 0.90 0.85 0.85
212

222

232

 

311 0.54 0.61 0.91 0.82 1.00'

 

312 0.76 0.75 0.88 0.88 1.00‘

 

313 0.37 0.72 0.76 0.97 0.97

 

321 0.62 0.86 0.73 1.00' 0.89

 

322 0.76 0.88 0.63 0.91 1.00'

 

323 0.77 0.73 0.73 0.90 1.00‘

 

331 0.67 0.73 0.85 0.92 1.00'

 

332 0.60 0.65 0.78 0.90 1.00“

 

333 0.66 0.73 0.66 0.85 1.00'

 

412

 

422

 

432

 

511 0.25 0.75 0.74 0.98 0.95

 

512 0.20 0.76 0.73 0.95 1.00'

 

513 0.07 0.81 0.81 0.91 0.93

 

521 0.12 0.70 0.64 0.87 0.89

 

522 0.54 0.67 0.73 0.90 1.00'

 

523 0.55 0.74 0.56 0.93 0.83

 

531 0.37 0.76 0.68 0.92 1.00‘

 

532 0.29 0.74 0.81 1.00" 0.98

 

533 0.35 0.66 0.82 0.91 1.00'

 

612

 

622

 

 

 

 

 

 

 

 

 

 

632

 

110

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Location 24 hr 48 hr 72 hr 120 hr 146.5 hr 160.75 hr 184.25 hr 241 hr

711 0.04 0.60 0.74 0.79 0.91

712 0.03 0.47 0.59 0.80 0.88

713 0.01 0.31 0.46 0.63 0.81

721 0.09 0.65 0.63 0.86 0.88

722 0.05 0.43 0.54 0.79 0.86

723 0.04 0.65 0.38 0.95 0.89

731 0.12 0.60 0.73 0.90 0.76

732 0.05 0.52 0.68 0.73 0.72

733 0.04 0.56 0.66 0.79 0.77

812

822

832

911 0.04 0.26 0.47 0.66 0.70 0.78 0.80

912 0.04 0.24 0.49 0.62 0.88 0.79 0.80

913 0.02 0.09 0.46 0.50 0.80 0.76 0.67

921 0.04 0.24 0.44 0.49 0.87 0.58 0.58

922 0.04 0.13 0.62 0.68 0.89 0.81 0.75

923 0.05 0.35 0.64 0.62 0.93 0.86 0.78

931 0.06 0.35 0.58 0.52 0.85 0.83 0.75

932 0.05 0.23 0.62 0.62 0.75 0.71 0.85

933 0.05 0.25 0.54 0.54 0.83 0.75 0.68

VerLlnfluent 0.88 1.00’ 1.00* 0.97 0.74 0.85 0.97 1.00'
Vert_Effluent 0.10 0.36 0.50 0.50 0.57 0.66 0.63

 

 

 

 

 

 

 

 

 

111

 

APPENDIX D

Inoculation
Bromide Concentrations at Internal Sampling Ports (C/Co)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Location 0 min 15 min 30 min 45 min 60 min 75 min 90 min
312 0.94 1.00* 1.00* 1.00‘
322 0.15 0.18 0.73 1.00* 1.00* 0.10
332 0.16 0.19 0.20 0.22 0.67

Hor_Effluent 0.19 0.19 0.06
Location 105 min 120 min 135 min 150 min 165 min 180 min
312 1 .00* 0.98 1 .00* 1.00* 0.98 0.86
322 1 .00* 1 .00" 1 .00" 1 .00* 0.98 0.86
332 1 .00‘ 1 .00* 0.98 0.98 0.94 0.83
Hor_Effluent 0.27 0.43 0.70
Inoculation

Mobile KC Concentrations at Internal Sampling Ports (C/Co)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Location 0 min 15 min 30 min 45 min 60 min 75 min 90 min
Hor_lnfluent 0.93 1 .00* 0.41 0.77
312 0.46 1.00“ 0.74 0.89 1 .00’ 1 .00"
322 0.00 0.00 0.01 0.08 1 .00* 1 .00"
332 0.00 0.00 0.00 0.00
Hor_Effluent
Location 105 min 120 min 135 min 150 min 165 min 180 mln
Hor_lnfluent 0.92 1 .00* 1 .00*
312 1.00* 1.00" 1 .00* 1.00' 0.76
322 1 .00* 1 .00* 1 .00* 1 .00* 0.95 0.63
332 0.04 0.12 0.23 0.49 1 .00* 1 .00‘
Hor_Effluent 0.00 0.01 0.04

 

 

 

 

 

 

 

 

 

112

 

 

APPENDIX E

Biocurtain
Carbon Tetrachloride Concentrations at Internal Sampling Ports (ppb)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Location 0 hr 1 .5 hr 53.75 hr 77.75 hr 1 20.25 hr 144.25 hr 144.75 hr
1 1 1 0.00 23.02
1 12 0.00 30.00
1 13 0.16 8.84
121 5.56 30.44
122 1 .49 18.86
123 1 .67 21 .17
131 42.91 18.85
132 31 .67 26.26
133 42.56 26.25
31 1 0.00 35.24
312 0.00 29.89
313 0.00 30.46
321 0.17 28.00
322 0.00 29.39
323 2.00 13.78
331 21 .29 40.60
332 4.86 40.86
333 10.97 39.20
51 1 0.00 7.64
512 0.00 4.27
513 0.00 1 .27
521 1 .75 15.26
522 0.38 8.96
523 3.10 1 .09
531 33.16 39.68
532 1 1.09 41.57
533 40.63 43.97

 

113

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Location 0 111' 1.5 hr 53.75 hr 77.75 hr 120.25 hr 144.25 hr 144.75 hr

71 1 12.38 1 .06
712 2.97 1 .47
713 2.50 0.81
721 40.65 2.27
722 25.57 1.28
723 31.49 1.80
731 44.88 36.19
732 43.43 37.59
733 30.55 26.80
91 1 37.84 5.23
912 44.44 4.04
913 39.47 13.70
921 30.55 4.13
922 32.46 14.1 1
923 40.55 1 1.84
931 44.83 35.76
932 49.98 31.34
933 38.56 34.76

Vert_lnfluent 39.02 61 .72 43.19 16.75

Verl_Eff|uent 25.68 27.52 22.20 10.82

Hor_Effiuent 28.35 30.10

 

114

 

Biocurtain
Carbon Tetrachloride Concentrations at Internal Sampling Ports (ppb)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Location 145.25 hr 145.75 hr 146.25 hr 146.85 hr 288.25 hr
1 1 1 25.34
1 12 32.22
1 13 ' 16.26
121 30.79
122 30.40
123 35.28
131 34.55
132 33.80
133 30.93
31 1 24.77
312 31 .95
313 0.00
321 25.25
322 29.88
323 26.92
331 35.59
332 31 .70
333 32.49
51 1 23.77
512 17.26
513 0.00
521 7.36
522 28.93
523 ‘ 17.15
531 30.85
532 31 .85
533 32.97

 

 

 

 

 

 

 

115

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Location 145.25 hr 145.75 hr 146.25 hr 146.85 hr 288.25 hr

71 1 0.84
712 0.00
713 0.00
721 ' 3.08
722 0.00
723 0.00
731 33.48
732 31.35
733 9.13
911 0.00
912 0.00
913 0.00
921 0.00
922 0.00
923 0.00
931 23.18
932 25.98
933 25.98

Vert_lnfluent 28.75

Vert_Effluent 3.23

Hor_Effluent 29.47 30.37 34.62 37.1 1

 

 

 

 

 

 

 

116

REFERENCES

Anderson, MP. and W.W. Woessner. 1992. Applied Groundwater Modeling: Simulation
of Flow and Advective Transport. Academic Press, San Diego, 381 p.

Brennan, MA. 2004. Effects of small-scale heterogeneity on tracer transport. MS.
Thesis, Department of Geological Sciences, Michigan State University.

Chapelle, RH. 2001. Ground-Water Microbiology and Geochemistry. John Wiley &
Sons Inc., New York, 477 p.

Clement, T.P. 1997. RT3D-A modular computer code for simulating reactive multi-
species transport in 3-dimensional groundwater aquifers, Battelle Pacific Northwest
National Laboratory Research Report, PNNL-SA-28967, September, 1997.

Criddle, C.S., J.T. Dewitt, D. Grbic-Garlic, and PL. McCarty. 1990. Transformation of
Carbon Tetrachloride by Pseudomonas sp strain KC under Denitri fication Conditions.
Appl. Environ. Microbiol. 56 (11): 3240 - 3246.

Davis, N. 1995. Added Microbe Converts Suspected Carcinogen. Centerpoint, 2 (4 —— 5).

Dybas, M.J., G.M. Tatara, and CS. Criddle. 1995. Localization and characterization of
the carbon tetrachloride transformation activity of Pseudomonas sp strain KC. Appl.
Environ. Microbiol. 61: 758 -— 762.

Dybas M.J, M. Barcelona, 8. Bezborodnikov, S. Davies, L. Fomey, O. Kawka, T.
Mayotte, L. Sepulveda—Torres, K. Smalla, M. Sneathen, J. Tiedje, T. Voice, D.
Wiggert, M.E.Witt, C.S. Criddle. 1998. Pilot Scale Evaluation of Bioaugmentation
for in-situ Remediation of a Carbon Tetrachloride Contaminated Aquifer. Environ.
Sci. Technol., 32 (22): 3598 — 3611.

Dybas, M.J., D.W. Hyndman, R. Heine, J. Tiedje, K. Linning, D. Wiggert, T. Voice, X.
Zhao, L. Dybas, and CS. Criddle. 2002. Development, Operation, and Long-Term
Performance of a Full-Scale Biocurtain Utilizing Bioaugrnentation. Environ. Sci.
Technol. 36 (16): 3635 -— 3644.

Ewer, K.A. 1997. Three-dimensional groundwater flow and contaminant transport in
medium scale highly heterogeneous environments. MS. Thesis, Department of
Geological Sciences, Michigan State University.

Fetter, CW. 1988. Applied Hydrogeology. Second Edition. Macmillan Publishing
Company, New York, 592 p.

Fetter, CW. 1999. Contaminant Hydrogeology. Second Edition. Prentice-Hall, Inc., New
Jersey, 500 p.

117

Freeze, RA. and J .A. Cherry. 1979. Groundwater. Prentice-Hall, Inc., Englewood Cliffs,
New Jersey, 604 p.

Haggerty, R. and S.M. Gorelick. 1995. Multiple-rate mass transfer for modeling diffusion
and surface reactions in heterogeneous media, Water Resources Research. 31(10):
2383-2400.

Halliburton NUS Environmental Co. 1991. Remedial Investigation and Feasibility Study
of Remedial Alternatives, Plume A and Plume F, Village of Schoolerafl, Kalamazoo
County, Michigan.

Harbaugh, A.W., E.R. Banta, M.C. Hill, and MG. McDonald. 2000. MODFLOW-2000
The US. Geological Survey modular ground-water model — user guide to
modularization concepts and the Ground-Water Flow Process: US. Geological
Survey Open-File Report, 00-92, 121 p.

Huyakom RS. and GP. Pinder. 1983. Computational Methods in Subsurface Flow.
Academic Press, 473 p.

Hyndman, D.W., M.J. Dybas, L. Fomey, R. Heine, T. Mayotte, M.S. Phanikumar, G.
Tatara, J. Tiedje, T. Voice, R. Wallace, D. Wiggert, X. Zhao, and CS. Criddle. 2000.
Hydraulic Characterization and Design of a Full Scale Biocurtain. Ground Water, 38
(3): 462 — 474.

Johnson, W.P., P. Zhang, M.E. Fuller, T.D. Scheibe, B.J. Mailloux, T.C. Onstotte, M.F.
Hubbard, S.S. Radtke, J. Kovacik, W.P. Holben. 2001. Ferrographic tracking of
bacterial transport in the field at the Narrow Channel focus area, Oyster, VA.
Environ. Sci. T echnol. 35(1): 182-191. '

Lee, C.H., T.A. Lewis, A. Paszczynski, R.L. Crawford. 1999. Identification of an
extracellular catalyst of carbon tetrachloride dehalogenation from Pseudomonas
Stutzeri strain KC as pyridine-2,6-bis(thiocarboxylate). Biochem. Biophys. Res.
Commun, 261(3): 562-566.

Lewis, T.A. and R.L. Crawford. 1993. Physiological factors affecting carbon
tetrachloride dehalogenation by the denitrifying bacterium Pseudomonas sp strain
KC. Appl. Environ. Microbial. 59: 1635-1641.

Lewis, T.A. and R.L. Crawford. 1999. Chemical Studies of Carbon Tetrachloride
Transformation by Pseudomonas sp strain KC. Navel Approachesfar Bioremediatian
of Organic Pollutants. R. F ass, Y. Flashner, and S. Reuveny Editors. Kluwer
Academic, New York, pp. 1-10.

Mandle, RJ. 2002. Groundwater Modeling Guidance. Groundwater Modeling Program,
Michigan Department of Environmental Quality.

118

Mayotte, T., M.J. Dybas, C.S. Criddle. 1996. Bench-scale evaluation of bioaugmentation
to remediate carbon tetrachloride-contarninated aquifer materials. Ground Water,
34(2): 358-367.

Murphy, E.M., T.R. Ginn, A. Chilakapati, C.T. Resch, J .L. Phillips, T.W. Wietsma, and
CM. Spadoni. 1997. The influence of physical heterogeneity on microbial
degrdataion and distribution in porous media. Water Resources Research, 33(5):
1087-1 103.

Phanikumar, M.S., D.W. Hyndman, D.C. Wiggert, M.J. Dybas, M.E. Witt, and CS.
Criddle. 2002. Simulation of microbial transport and carbon tetrachloride

biodegradation in intermittently-fed aquifer columns. Water Resources Research, 38
(4), 10.1029/2001WR000289.

Phanikumar, MS. and D.W. Hyndman. 2003. Interactions between sorption and
biodegradation: Exploring bioavailability and pulsed nutrient injection efficiency.
Water Resources Research, 39 (5), 1122,10.1029/2002WR001761.

Phanikumar, MS. and D.W. Hyndman. 2003. Correction to “Interactions between
sorption and biodegradation: Exploring bioavailability and pulsed nutrient injection
efficiency.” Water Resources Research, 39 (10), 1281,10.1029/2003WR002327.

Phanikumar, M.S., D.W. Hyndman, X. Zhao, and M.J. Dybas. 2005. A Three-
Dimensional Model of Microbial Transport and Biodegradation at the Schoolcrafl,
Michigan Site. Water Resources Research, 41(5)10.1029/2004WR003376.

Radabaugh, PD. 1998. Factors Affecting Transport of Pseudomonas Stutzeri KC, MS.
Thesis, Department of Civil and Enviromnental Engineering, Michigan State
University.

Tatara, G.M., M.J. Dybas, and CS. Criddle. 1993. Effects of Medium and Trace Metals
on Kinetics of Carbon Tetrachliride Transformation by Pseudomonas sp Strain KC.
App. Env. Microbiology, 59 (7): 2126 — 2131.

Tatara, G.M., M.J. Dybas, and CS. Criddle. 1995. Biofactor-mediated transformation of
carbon tetrachloride by diverse cell types. In Bioremediation of Chlorinated Solvents;
Hinchee, R.E., Leeson, A, Semprinit, L., Eds; Battelle Press: Colombus, OH, pp. 69-
76.

Vidal-Gavilan, G. 2000. Multidimensional Transport of Pseudomonas Stutzeri Strain KC
in a Model Aquifer System: Predicting CT Remediation. MS. Thesis, Department of
Civil and Environmental Engineering, Michigan State University.

Witt, M.E., M.J. Dybas, R.L. Heine, S. Nair, and CS. Criddle, and DC. Wiggert. 1995.
Bioaugrnentation and Transformation of Carbon Tetrachloride in a Model Aquifer.

Bioremediation 3: Bioaugmentation for Site Remediation, Battelle Press, Columbus,
221 — 227.

119

Witt, M.E., M.J. Dybas, D.C. Wiggert, and CS. Criddle. 1999. Use of Bioaugrnentation
for Continuous Removal of Carbon Tetrachloride in Model Aquifer Columns. Env.
Eng. Sci, 16(6): 475 — 485.

Zhao, X., M.J. Szafranski, M.A. Maraqa, and T.C. Voice. 1999. Desorption and
Bioavailability of Carbon Tetrachloride in a Low Organic Content Sandy Soil.
Environmental Chemistry and T axicalagy, 18 (8): 1755 — 1762.

Zhao, X., R.B. Wallace, D.W. Hyndman, M.J. Dybas, and T.C. Voice. submitted.
Heterogeneity of Chlorinated Hydrocarbon Sorption Properties in a Sandy Aquifer.
Journal of Contaminant Hydrology.

120

«I,1,..I'.

III

1293 O 736

3