. if.
‘fLAll

as.
«aw... .
aw... kfivr

a .
. . i. x
. . . “fig: a: a
IT: Iii->1 9 . 1.
‘—.v.-»-. c “I... I . ‘
I. :. e. I In... a .
. O. - ‘)A~ Imv’i: .‘U.
2,... .35.... 2.. . .
.1. 34:: 3:. 1.5.1...
3. . fir...H a.a.........xu. ..
I. 43w . . .. a...
15.3“. I 1» “MP

9.

i:

,5 ..

I. 7“.-
.‘.‘~'. . ’-1‘-£JT‘.
“4.1“ in..utut‘l a.“ ¢

éxafiéauv. i .
£15. 1%! ~
z :2 4x:

03......
, z.
sumfififi

!

«wanna...
XII). ’i-Iv-I I.
1:}. 1:36}
XAE’ITUII;

3.3%..-.fi1iki: 3.

$3,. . ..

it}. “E: «i.
dun” 3 It}...

.
O 14.1..”- ‘h a. I.

(t... . it‘s! i 5. . .
lib-l: - i

 

 

 

 

1....

x f. . a? . "I”.I. L.

1.340099. 5%... .... .si. . “a
. ..Lnu.h..§and:n.rr‘lfiu§uir

iLmai...r§,..kiuflnYrrifi§n a... .

:2 g. It 1 u t L

 

£3 £1-
i . l...i?2!3x.)lr
1.13

 

 

‘I5\..i..
3.1.57...‘

 

 

99.3.5. .uuFEh-d ._{ EL... .. . . .32. :

mists

.q/(yo’v

This is to certify that the
thesis entitled
Estimating Natural Attenuation Rates for a
Chlorinated Hydrocarbon Plume in a Glacio-fluviai

Aquifer, Schoolcraft, Michigan

presented by

Brian Andrew Lipinski

has been accepted towards fulfillment
of the requirements for

MaSter'S degree in GCO'OQY

 

 

Major professor

Date January 21, 2002

 

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

 

LIBRARY
Michigan State
University

 

 

 

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRC/DateDuepas-sz

ESTIMATING NATURAL ATTENUATION RATES FOR A CHLORINATED
HYDROCARBON PLUME IN A GLACIO-FLUVIAL AQUIFER,
SCHOOLCRAFT, MICHIGAN
By

Brian Andrew Lipinski

A THESIS

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

MASTER OF SCIENCE
Department of Geological Sciences

2002

ABSTRACT
ESTIMATING NATURAL ATTENUATION RATES FOR A CHLORINATED
HYDROCARBON PLUME IN A GLACIO-FLUVIAL AQUIFER,
SCHOOLCRAFT, MICHIGAN
By

Brian Andrew Lipinski

A plume of chlorinated hydrocarbons in Schoolcrafi, Michigan, that has
undergone abiotic and biotic degradation processes, was examined using numerical
modeling techniques. The flow field was simulated with a three-dimensional steady state
heterogeneous plume scale model that incorporated natural hydrologic boundaries
through a Telescopic Grid Refinement of a three-dimensional steady state regional scale
model that was extended to natural hydrologic boundaries. Contaminant plume evolution
was examined using an RT3D model that includes. aerobic and anaerobic biodegradation
of the source compounds PCE, TCE, TCA, and their daughter byproducts, as well as the
simultaneous abiotic degradation of TCA. The model was used to estimate the
degradation rates of the plume species by comparing the calculated ratio of a species to
its daughter byproduct for observed data and simulated results in a well transect
downgradient of the source. The ratio approach should be applicable to any field site in
which contaminants degrade to form a byproduct, thus circumventing the source term
problem. The plume compounds degraded at a rate between 10'2 day’1 and 10'5 day",
which is similar to natural attenuation rates at the RTDF Dover Site. The rates are too
slow to allow natural attenuation process alone to remediate the plume in a reasonable

time frame.

ACKNOWLEDGEMENTS

Many people have contributed to the completion this thesis, and their help was
greatly appreciated. First, I would like to thank the members of my committee. My
advisor, Dave Hyndman, has provided much support and many insightfiil comments that
led to the completion of the thesis. Committee members Gary Wiessmann and Dave
Wiggert made significant contributions to the manuscript, which made the text flow
much better, while also pointing out fundamental problem areas of the work. Jaime
Graulau-Santiago helped with the assimilation of the plume data; his help was
irreplaceable. The research was funded by the Michigan Department of Environmental
Quality (contract #Y43086).

There were also many people that made my stay here enjoyable and/or were very
helpful. A short list includes Joel, Nick, Linker, Moss, and Chris. Maybe we will actually
catch some steelhead on the Manistee River someday instead of just catching a cold.
Also, we need to practice blackjack before we head back to Reno (note to self, always
double down when you are dealt two aces). I would also like to thank Jackie, Cathy, and
Loretta. Without their guidance, I would still have years left to finish my degree. Lastly, a
good fi'iend once told me that I was not allowed back in Pennsylvania until I had

completed my thesis. Well, it’s done, so I guess I can get back into the Keystone state.

iii

TABLE OF CONTENTS

LIST OF TABLES ................................................................................ vii
LIST OF FIGURES .............................................................................. viii
I. INTRODUCTION AND SCOPE OF WORK .............................................. 1
OVERVIEW ...................................................................................... 1
SCOPE OF WORK ............................................................................... 8

II. THREE-DIMENSIONAL REGIONAL STEADY STATE
GROUNDWATER FLOW MODEL OF SCHOOLCRAFI‘, MICHIGAN ........... 10
INTRODUCTION .............................................................................. 1 O
PREVIOUS WORK ............................................................................ 11
Regional Geology ........................................................................... 12
Regional H ydrogeology ................................................................... 1 9
Regional Hydrology ........................................................................ 21
CONCEPTUAL MODEL ..................................................................... 22
Conceptual Boundaries ................................................................... 22
Conceptual Layers ......................................................................... 24
ARCVIEW ANALYSIS ....................................................................... 28
Boundary Digitization in Arc View ....................................................... 28
Stratigraphic Unit Development ......................................................... 29
IMPLEMENTATION OF THE CONCEPTUAL MODEL .............................. 33
Development of a 3-D Finite Difference Grid .......................................... 33
Layer Elevation Interpolation ............................................................ 34
Model Boundary Assignment ............................................................. 35
Hydraulic Conductivity Assignment ..................................................... 35
MODEL CALIBRATION ..................................................................... 38
Observed Head Data Set .................................................................. 38
Results ....................................................................................... 39
Discussion ................................................................................... 39

III. THREE-DIMENSIONAL PLUME SCALE STEADY STATE
GROUNDWATER FLOW MODEL OF SCHOOLCRAFI‘, MICHIGAN ............45

iv

INTRODUCTION .............................................................................. 45

CONCEPTUAL MODEL ..................................................................... 45
Conceptual Boundaries .................................................................... 45
Conceptual Aquifer Material Types ...................................................... 46

IMPLEMENTATION OF THE CONCEPTUAL MODEL .............................. 50
Development of a 3-D Finite Difference Grid .......................................... 51
Creation of Aquifer Solid Materials in GMS ........................................... 51
Interpolation of the Aquifer Solids to the MODFLOW Grid ......................... 51
Boundary Assignment ..................................................................... 55

MODEL CALIBRATION ..................................................................... 55
Results ....................................................................................... 55
Discussion ................................................................................... 56

IV. THREE DIMENSIONAL REACTIVE TRANSPORT MODEL OF

PLUME G, SCHOOLCRAFT, MICHIGAN ................................................ 60
INTRODUCTION .............................................................................. 6O
ARCO SITE BACKGROUND ............................................................... 60

Source Areas ................................................................................ 61
Previous Remediation Systems ............................................................ 63
DEGRADATION PROCESSES ............................................................. 66
Abiotic Degradation Processes .......................................................... 68
Biotic Degradation Processes ............................................................ 68
Biotic Degradation Process: Occurrence Under Aerobic Conditions .......... 69

Biotic Degradation Processes: Occurrence Under Anaerobic
Conditions .............................................................................. 7O
Biotic Degradation Processes: Rate Sensitivity to Redox Conditions ......... 71

CURRENT METHODS TO ESTIMATE FIELD SCALE

DEGRADATION RATES ..................................................................... 71
Tracer normalization ...................................................................... 71
Semi-Graphical Method of Buscheck and Alcantar (1995) ........................... 73
Numerical Modeling ...................................................................... 74
Method to Estimate Rates at Plume G ................................................... 75

CONCEPTUAL MODEL ..................................................................... 75
Model Stress Periods ...................................................................... 75
Conceptual Reactions of Interest ......................................................... 79
Conceptual Method of “Calibration ” ................................................... 82

IMPLEMENTATION OF THE CONCEPTUAL MODEL .............................. 88

Creation of an RT 3D Reactions Package ............................................... 88

Assignment of Model Parameters ........................................................ 91
MODEL CALIBRATION ..................................................................... 94
Results .................................................................................... 95

Model Sensitivity to Degradation Rates ................................................ 98
Discussion ................................................................................. 1 08
SIGNIFICANCE AND APPLICATIONS ................................................. 109
Novel Aspects of the Model .............................................................. 109
Applications of the Research at the Site ............................................... 110
RECOMMENDATIONS FOR FUTURE WORK ....................................... 110
APPENDIX A. WELL LOG DESCRIPTIONS FOR MSU WELLS 1-9 ............. 114
APPENDIX B. 3-D SOLID STRATIGRAPHY MODELING IN GMS ............... 117
INTRODUCTION ............................................................................ l 17
GMS BOREHOLE DATA ANALYSIS ................................................... 117
TRIANGULAR INTERPOLATED NETWORKS ...................................... 118
CREATION OF 3-D SOLID OBJECTS ................................................... 118
REFERENCES .................................................................................... 124

vi

TABLE 1

TABLE 2

TABLE 3

TABLE 4

TABLE 5

TABLE 6

TABLE 7

TABLE 8

TABLE 9

TABLE 10

TABLE 11

TABLE 12

TABLE 13

LIST OF TABLES

Chemicals comprising each of the seven currently identified
plumes in the Village of Schoolcrafl, Michigan

(modified from Mayotte, 1991) .............................................

Hydraulic conductivities of the various materials in the regional

model ...........................................................................

Qualitative description of the materials comprising the aquifer

in the plume scale model domain .............................................

Physical characteristics of the plume scale groundwater flow

model (all cells were active) ..................................................

Hydraulic conductivity data (K) for the different aquifer
materials described in Table 3 (observed data from

Mayotte, 1991 ) ..................................................................

Summary of conceptual stress periods for plume G transport

models ...........................................................................

Estimated total aqueous mass of observed compounds at well
transect in shallow (0-13 mbgs, 1.44x105 L), middle (13-19
mbgs, 1.36x105 L), and deep (19-27 mbgs, 1.07x105 L) zones

for June of 1988 .................................................................

Summary of transport model parameters which remained

constant throughout model calibration .......................................

Estimated source concentrations for the model source areas

based on site investigation results assimilated in Everett (1990) ........

Summary of first order decay rates for chlorinated compounds

as compiled by Suarez and Rifai (1999) ....................................

Observed and simulated masses for the three conceptual

contamination zones outlined in the Conceptual Calibration section . .

Observed and simulated contaminant mother/daugher ratios
for the three conceptual contamination zones outlined in the

Conceptual Calibration section .............................................

Summary of calibrated parameters used in the RT3D model

vii

..... 4

...42

...47

...50

...56

...79

...85

...92

...93

....94

96

...97

.97

LIST OF FIGURES

FIGURE 1 Location of the Village of Schoolcraft, Kalamazoo County,
Michigan ............................................................................. 2

FIGURE 2 Location and extent of MDNR plumes A-G in the Village of
Schoolcraft (modified from Mayotte, 1991). Note the location
of the John A. Biewer wood treatment facility and ARCO
Industries Co., which are the source areas for plumes F and
G respectively ....................................................................... 3

FIGURE 3 Bedrock map of the Lower Peninsula of Michigan with
Kalamazoo County enlarged (modified Milstein, 1987).
Note that the Coldwater Shale is Chartreuse, while the
Marshall Formation is gray. Image is presented in color ..................... 13

FIGURE 4 Map of moraines in southwest Michigan and the study area of
Monaghan et. a1. (1986). Note the location of the interlobate
boundary that separates deposits of the Michigan Lobe from
deposits of the Saginaw Lobe (modified from Monaghan
et. al., 1986) ........................................................................ 14

FIGURE 5 Surficial map of Kalamazoo County (modified from
Monaghan and Larson (1982) in Rheaume (1990)). Image
is presented in color ............................................................... 16

FIGURE 6 Michigan Lobe tills correlated to respective moraine(s) in
southwest Michigan (modified from Monaghan et. al., 1986).
Schoolcraft lies between the Outer Kalamazoo Moraine and
the Tekonsha Moraine of the Michigan Lobe ................................. 17

FIGURE 7 Map showing the approximate boundary of the F inkbeiner
(1994) study area (modified from F inkbeiner, 1994). Note the
trace of the Tekonsha Moraine, which is discussed in the text .............. 18

FIGURE 8 Location of the Prairie Ronde Fan presented by Steinmann
(1994) (modified from Kehew et. al., 1996) ................................... 20

FIGURE 9 Location of the significant surface water bodies in the vicinity
of Schoolcraft, Michigan that served as boundaries in the
regional model. The perimeter boundaries that are not
hydrologic boundaries are no flow boundaries interpreted as
topographic divides from the DEM ............................................. 23

viii

FIGURE 10

FIGURE 11

FIGURE 12

FIGURE 13

FIGURE 14

FIGURE 15

FIGURE 16

FIGURE 17

FIGURE 18

FIGURE 19

FIGURE 20

FIGURE 21

Simplified cross-section of the regional geology surrounding

Schoolcrafi used to conceptualize the regional model. Figure

layer thicknesses do not represent actual layer thicknesses.

Image is presented in color ....................................................... 26

Flow chart outlining the steps taken to develop the regional
model in the conceptual model process and summary of the
results ............................................................................... 27

A) Location of the water wells near the regional model.

B) Wells remaining after quality checks were performed,

which are also sorted according to whether the layer 2 till

was identified at the location .................................................... 31

Areal extent of the various conductivity zones used in the
regional groundwater flow model (layer 1 is the top, layer 3

is the bottom) ...................................................................... 32
Variogram for interpolation of layer 1 bottom elevations ................... 36
Variogram for interpolation of layer 2 bottom elevations ................... 36

Interpolated model layer bottom elevations. The location of
the data points from which elevations were interpolated are
shown as triangles ................................................................. 37

Simulated vs. observed heads for the observation wells in
the vicinity of Schoolcraft ........................................................ 40

Head distribution of the calibrated regional model (view is
of the top layer) .................................................................... 41

Head distribution of the calibrated regional model with flow
vectors to show direction of groundwater movement
predicted by the model (view is of the top layer) ............................. 43

Location and orientation of the plume scale model in reference

to the regional model (cells are not shown as they are too

small to be seen at the current scale). Well locations are given

for NUS and MSU wells where lithologic data were available

for aquifer characterization ....................................................... 48

A) Contour map of the bottom of the plume scale flow model,

which is also the bottom of layer 1 in the regional model.

B) Contour map of the ground elevations in the plume scale

flow model. The contour interval is 0.5 m ..................................... 49

ix

FIGURE 22 Selected cross-sections through the 3-D GMS solids that
represent the aquifer materials in the plume scale model.
The locations of the wells used to create the solids are given
on the inset. Image is presented in color (VE = 20X) ........................ 53

FIGURE 23 Graphical depiction of the results of the three GMS
algorithms to map solid material properties to a hypothetical
MODFLOW model. A) An arbitrary set of solids. B) Cross-
section through A) using Boundary Overlay. C) Cross-
section through A) using grid Overlay. D) Plan view of layer
in A) showing the smoothing effects of Grid Overlay with Keq
(modified from BYU, 2000). Image is presented in color ................... 54

FIGURE 24 Head distribution of the top layer of the calibrated plume scale
model, which shows flow is in a southeasterly direction. The
local heads are shown in grayscale, while the regional heads are
superimposed in black. Inspection between the two distributions
reveals a reasonable match ....................................................... 57

FIGURE 25 Plot of simulated vs. observed heads for the plume scale and
regional scale model ............................................................... 58

FIGURE 26 Location of the five suspected source areas on the ARCO
property (modified from Everett, 1990) ........................................ 62

FIGURE 27 Location of the three purge wells that were part of the
initial pump and treat system to remediate plumes F and G
(modified from Mayotte, 1991) ................................................. 65

FIGURE 28 Biotic and abiotic transformations of PCB, TCE, and TCA
under aerobic and anaerobic conditions (modified from
McCarty, 1997; and Bedient et. al., 1999). All transformations
are biotic unless indicated with a dotted arrow, which is the
abiotic pathway. Heavy arrow indicates preferential reaction .............. 67

FIGURE 29 Nitrate concentrations at monitoring well MSU-5 from
February, 2001. Note that nitrate starts to decrease at about
13 mbgs until about 19 mbgs, where it remains constant.
Results suggest that the aquifer is aerobic from the water table
to 13 mbgs, anaerobic denitrifying from 13 mbgs to 19 mbgs,
and somewhat more anaerobic from 19 mbgs to the clay at
26-27 mbgs ......................................................................... 81

FIGURE 30 Location of downgradient monitoring well transect used to
identify vertical zones of contamination. The numbers correlate
to the NUS number assigned to the wells during the remedial

FIGURE 31

FIGURE 32

FIGURE 33

FIGURE 34

FIGURE 35

FIGURE 36

investigation studies in the late 1980’s (Mayotte, 1991) ..................... 86

Plot of TCE concentrations at a well transect perpendicular to
groundwater flow approximately 650 meters downgradient

of the ARCO facility. The shaded regions are the shallow,

middle, and deep generalized zones of contamination. The

view is looking upgradient towards ARCO. Image is presented

in color .............................................................................. 87

Sensitivity plots of model error (RMSE) of the

mother/daughter ratios (PCB/TCE path) in response to

varying degradation rates (model rate is circled). A) The

anaerobic PCE rate in the middle zone. B) The anaerobic

PCE rate in the deep zone (note, PCE is not currently known

to degrade under aerobic conditions, therefore there is no

aerobic PCE rate in the shallow zone) ........................................ 100

Sensitivity plots of model error (RMSE) of the

mother/daughter ratios (PCB/TCE path) in response to

varying degradation rates (model rate is circled). A) The

aerobic TCE rate in the shallow zone. B) The anaerobic TCE

rate in the middle zone. C) The anaerobic TCE rate in the

deep zone .......................................................................... 101

Sensitivity plots of model error (RMSE) of the

mother/daughter ratios (PCB/TCE path) in response to

varying degradation rates (model rate is circled). A) The

aerobic 1,2-DCE rate in the shallow zone. B) The anaerobic

1,2-DCE rate in the middle zone (dashed line is where 1,2-DCE

rate cannot be less than the VC rate in this zone). C) The

anaerobic 1,2-DCE rate in the deep zone ..................................... 102

Sensitivity plots of model error (RMSE) of the

mother/daughter ratios (PCB/TCE path) in response to

varying degradation rates (model rate is circled). A) The

aerobic VC rate in the shallow zone (dashed line is literature

value boundary). B) The anaerobic VC rate in the middle

zone (dashed line indicates VC cannot be higher than deep

zone. C) The anaerobic VC rate in the deep zone (dashed

line represents boundary where the VC rate in the deep zone

cannot be lower than the middle zone ......................................... 103

Sensitivity plots of model error (RMSE) of the
mother/daughter ratios (TCA path) in response to varying
degradation rates (model rate is circled). A) The aerobic TCA
rate in the Shallow zone. B) The anaerobic TCA rate in the

xi

FIGURE 37

FIGURE 38

FIGURE 39

middle zone. C) The anaerobic TCA rate in the deep zone ............

Sensitivity plots of model error (RMSE) of the

mother/daughter ratios (TCA path) in response to varying
degradation rates (modeled rate is circled). A) The aerobic DCA
rate in the shallow zone. B) The anaerobic DCA rate in the
middle zone (dashed line indicates the rate in the middle zone
cannot be higher than in the deep zone). C) The anaerobic

DCA rate in the deep zone ..................................................

Sensitivity plots of model error (RMSE) of the
mother/daughter ratios (TCA path) in response to varying
degradation rates (modeled rate is circled). A) The aerobic
1,1-DCE rate in the Shallow zone. B) The anaerobic 1,1—DCE
rate in the middle zone (dashed line indicates the 1,1-DCE
rate cannot be lower than the VC rate in this zone). C) The

anaerobic 1,1-DCE rate in the deep zone .................................

Sensitivity plots of model error (RMSE) of the
mother/daughter ratios (TCA path) in response to varying
degradation rates (model rate is circled). A) The aerobic VC
rate in the shallow zone (dashed line is literature value
boundary). B) The anaerobic VC rate in the middle zone
(dashed line indicates VC cannot be higher than deep

zone. C) The anaerobic VC rate in the deep zone (dashed
line represents boundary where the VC rate in the deep zone

cannot be lower than the middle zone .....................................

FIGURE B-l Location of the solid for which an explanation of 3-D solid

FIGURE B-2 Location of the wells used to create the TINS and solids in GMS . . . . .

construction in GMS is presented in Appendix B. The inset
shows the areal extent of the solid with respect to the local
model grid and in reference to the cross-sections. Image is

presented in color (VE = 20X) .............................................

FIGURE B-3 Generalized steps of solid construction from a set of borehole

objects in GMS. A) Creation of a TIN representing the top
of a solid. B) Creation of a TIN representing the bottom
of a solid. C) Top TIN with the interior nodes snapped

to the bottom TIN. D) Resulting solid created by filling

between C) and B). Image is presented in color .........................

xii

....104

....105

....106

....107

....120

121

....122

CHAPTER I

INTRODUCTION AND SCOPE OF WORK

OVERVIEW

The village of Schoolcraft is located in Kalamazoo County Michigan (Figure l).
The town is a rural community with several small industrial sites. The activities at these
sites are the focus of research currently being conducted at Michigan State University.
Groundwater from the glacio-fluvial aquifer in the region is the primary source of
drinking and irrigation water for the farmers (Mayotte, 1991); therefore, contamination of
the aquifer is a severe threat to the livelihood of the community.

Several contaminant plumes were identified in the aquifer beneath Schoolcraft in
the 1970’s and 1980’s. A plume of hexavalent chromium and arsenic was discovered in
the mid 1970’s and the Michigan Department of Natural Resources (MDNR) identified
the source to be the John A. Biewer Company wood treatment facility. In 1985, the
MDNR determined that the aquifer was contaminated with several chlorinated aliphatic
hydrocarbons and ARCO Industries was identified as one of the sources. Subsequently,
Halliburton NUS Environmental Corporation was hired in December of 1986 by the
MDNR to identify and define all sources/plumes of contamination in the aquifer in and
around the village. NUS determined that there were seven plumes in the aquifer beneath
Schoolcraft. The plumes were named A-G (Figure 2) and a brief list of the chemicals
contained within each one is given in Table 1 (Mayotte, 1991).

The groundwater contamination at Schoolcrafi is part of an international problem.

Chlorinated organic compounds, such as tetrachloroethene (PCE), trichloroethene (TCE),

 

42° 25'

 

 

040

I——l

km

 

I
Kalamazoo

 

 

Figure 1. Location of the the Village of Schoolcrafi, Kalamazoo County, Michigan.

 

ELIZA ST WEST AVE

ARCO

l4"I STREET

BIEWER

 

Figure 2. Location and extent of MDNR plumes A-G in the Village of Schoolcraft
(modified from Mayotte, 1991). Note the location of the John A. Biewer wood
treatment facility and ARCO Industries Co., which are the source areas for plumes F
and G respectively.

and trichloroethane (TCA), gained widespread use in industry throughout the 20th century
mainly in the form of solvents to degrease metals and machinery and in the dry cleaning
industry. These chlorinated organic solvents have become ubiquitous groundwater
contaminants throughout the United States and other countries through their subsequent
misuse and mishandling (Jackson and Dwarakanath, 1999). PCE, TCE, and TCA are
currently regulated by the Clean Water Act with Maximum Contaminant Levels (MCL’S)
of 5 ppb, 5 ppb, and 200 ppb respectively because they are suspected human carcinogens.
Another problem is that these compounds undergo a series of dechlorination reactions to
form compounds such as dichloroethene (DCE), dichloroethane (DCA), and vinyl
chloride (VC). All of these compounds are suspected carcinogens and have MCL’s as
follows: 1,1-DCE (7 ppb), cis-1,2-DCE (70 ppb), trans-1,2-DCE (100 ppb), 1,1-DCA (70
ppb), and VC (2 ppb). The transformation of PCB, TCE, and TCA to their daughter by-
products is a topic that has been heavily researched in the literature, a review of which is

given by Vogel (1987) and further by Semprini et. a1. (1995).

 

Table 1. Chemicals comprising each of the seven currently identified plumes in
the Village of Schoolcraft, Michigan (modified from Mayotte, 1991).
Plume Contaminants Present
carbon tetrachloride, chloroform, trichloroethane, and trichloroethene
benzene and 1,2-dichloroethene
trichloroethane, trichloroethene, and methylene chloride
tetrachlorethene, trichloroethene, and trichloroethane
tetrachloroethene, trichloroethene, and trans-1,2-dichloroethene
arsenic and hexavalent chromium
tetrachlorethene, trichloroethene, and trichloroethane

 

 

 

 

 

 

 

 

 

 

Gremcnw>

 

 

Significant attention in government, academia, and private consulting has focused

on developing successful remediation strategies due to widespread contamination of
groundwater with PCE, TCE, and TCA. One of the most common remediation methods is
with “pump and treat”. In a pump and treat system, several wells are typically installed
downgradient of the source area and pumped in a manner as to “capture” the
contaminated groundwater. The captured water is pumped to the surface and directed to
some type of wastewater treatment facility (Fetter, 1993). The treatment facility consists
of either an air stripping tower, a reaction vessel, or a combination of both. The air
stripping tower contains a packing material onto which contaminated water is sprayed as
air is forced through the tower. The organic compounds volatilize as air passes over the
packing material and the vapors are carried out the top of the tower to the atmosphere. If
the contaminant is not very volatile, the captured water is treated with a reaction vessel.
The reaction vessel contains granular activated carbon (GAC) onto which the
contaminants will sorb (Driscoll, 1986). The treated wastewater must then be disposed.
Several drawbacks exist to the pump and treat approach. First, the systems are
typically in operation for a significant time-period and they pump excessive volumes of
water, both of which are very costly endeavors. Second, contaminants that have diffused
into the low conductivity zones are difficult to remove because pumping preferentially
draws water from the high conductivity zones. In addition, the sorbed phase of the
contaminant may not be removed from the aquifer matrix (Fetter, 1993). Also, air
shipping only acts to transfer the contaminant to the atmosphere which potentially creates
an air pollution problem. Lastly, the extracted water must be disposed, which may require
federal and state permits (Fetter, 1993). For these reasons, alternative methods of aquifer

treatment have been sought.

In-situ bioremediation is a method of contaminant remediation that has recently
gained popularity. The general principle is that microbes will directly or indirectly break
down the undesired chemical to harmless by-products, which is a much more desirable
result than the end product of pump and treat systems. There are currently three general
sub-categories of bioremediation: monitored natural attenuation (MNA), biostimulation,
and bioaugrnentation.

MNA is referred to by many other names, such as intrinsic bioremediation,
passive remediation, or natural recovery (Azadpour-Keeley et. al., 2001). Natural
attenuation processes include degradation, dispersion, and sorption. The United States
Environmental Protection Agency (USEPA, 1997) defines MNA as

...the reliance on natural attenuation processes to achieve site-

specific remedial objectives within a time frame that is reasonable

compared to other methods. The “natural attenuation processes ” that are

at work in such a remediation approach include a variety of physical,

chemical, or biological processes that under favorable conditions, act

without human intervention to reduce mass, toxicity, mobility, volume, or

concentration of contaminants in soil or groundwater.

The processes of dispersion and sorption merely dilute contaminant concentrations.
Degradation removes contaminant mass from the system, thus it is the process of primary
interest. It is often not difficult to demonstrate that degradation processes are occurring at
a site, but it is difficult to determine if the rate of the processes will proceed at a pace to
prevent harm to the environment and human health (Azadpour-Keeley et. al., 2001).
MNA is widely used as a treatment solution for hydrocarbon (BTEX) plumes, but
chlorinated organic solvents are often not remediated at an acceptable rate by MNA

(Wiedemeier et. al., 1996).

Biostimulation is a method similar to MNA, except that electron donors, electron

acceptors, and/or nutrients are added to an aquifer to stimulate microbial growth. As a
result, the rate of contaminant degradation is increased significantly and the contaminants
are transformed faster than would be if the microbes were not stimulated. The method
proved successful at site 19, the Edwards Air Force Base in California (McCarty et. al.,
1998). A plume of TCE with concentrations ranging from 500-1200 ppb was treated with
injection of toluene as an electron donor and oxygen as an electron acceptor. TCE
concentrations in groundwater leaving the site were reduced by approximately 97-98%
(McCarty et. al., 1998). The study was conducted in an aquifer that was relatively
homogeneous and bounded above and below by thin aquitards. The remediation system
was engineered for the specific hydrogeologic system, and therefore its application to
other aquifers does not seem likely.

Bioaugrnentation is a process that involves the addition of microbes and the
necessary nutrients to sustain them into an aquifer. The microbes can be indigenous
species that are added to increase the current populations, or non-indigenous organisms if
the species present do not degrade the contaminant at a sufficient rate or produce
undesirable intermediate products. A benchmark study in bioaugmentation was
conducted at plume A in Schoolcraft, Michigan (Table 1 and Figure 2). The plume is
composed of carbon tetrachloride (CT) with concentrations generally less than 100 ppb.
The researchers designed a system to inject a CT degrading bacteria and nutrients into a
zone referred to as a “biocurtain”. The biocurtain extended vertically over the entire
thickness of contamination and groundwater passively moved through the treatment zone
and CT was degraded. CT in plume A has been successfully remediated for over 3 years

(Hyndman et. al., 2000). In addition to the bioaugmentation research currently being

conducted at Schoolcraft plume A, a bioremediation system is being developed at the
Schoolcraft plume G site. The main goal is to design and implement an in situ
bioremediation system that will be applicable to a wide range of hydrogeological
conditions and will be much more cost effective than traditional remediation methods,

such as pump and treat systems.

SCOPE OF WORK
The research contained in this thesis was used to develop a more thorough
understanding of the flow and natural attenuation processes at plume G. The primary
objective was to develop a 3-D reactive transport model at the plume scale and to
estimate the rates of degradation of the contaminants contained within plume G. The
steps used to complete the aforementioned objective were as follows:

0 Chapter II - Construction and calibration of a 3-D, steady state, regional scale,
groundwater flow model for the area surrounding the village of Schoolcraft

0 Chapter III - Construction and calibration of a 3-D, steady state, plume scale,
groundwater flow model using heads derived from the regional model

0 Chapter IV - Construction and calibration of a 3-D reactive transport model of
plume G using the plume scale flow model to estimate the rates of degradation of
the source contaminants and their daughter byproducts

The regional flow model incorporated natural hydrologic boundaries and regional scale
geologic features to more fully understand the hydrogeology of the area and to determine
the spatial pattern of hydraulic heads at the plume site. The plume scale flow model was
constructed with a more refined grid than the regional model, which enabled inclusion of
more heterogeneity to better represent contaminant transport. The hydraulic heads of the

regional model were incorporated in the local model through telescopic grid refinement.

The plume scale flow model served as the basis for the transport modeling activities to
explore the movement and degradation of PCB, TCE, TCA, and their degradation

byproducts.

CHAPTER II

THREE-DIMENSIONAL REGIONAL STEADY STATE GROUNDWATER
FLOW MODEL OF SCHOOLCRAFT, MICHIGAN

INTRODUCTION

Wiedemeier et. al. (1996) strongly recommended the use of detailed groundwater
flow and coupled contaminant transport models as part of the investigation of natural
attenuation processes. A groundwater flow model was created as the first step in the
evaluation of the degradation of chlorinated organics at plume G. While groundwater
flow models can be valuable tools, they are only reliable and thus useful if all aspects of
the modeling process have been critically evaluated and the natural system adequately
represented.

Anderson and Woessner (1992) indicated that one of the most important aspects
of modeling is the appropriate choice of boundary conditions. Reilley (2001) indicated
that the key in boundary selection is to choose model boundaries that correspond to
features in the model system that minimize artificial approximations. Anderson and
Woessner (1992) stated that the best boundaries to use are physical boundaries, such as
impermeable rock units (or units that have a hydraulic conductivity that is two orders of
magnitude less than the modeled unit). If physical boundaries are not readily present in
the system, then hydraulic boundaries, such as regional groundwater divides, are the next
best type to use (Anderson and Woessner, 1992).

Physical and/or hydraulic boundaries were not immediately found in Schoolcraft
on the small scale that was needed to complete a detailed analysis of natural attenuation

at plume G, but past studies by Buxton and Reilley (1986), Miller and Voss (1987), and

10

Ward et. a1. (1987) outlined a method to overcome these difficulties. Simply stated, a
large “regional scale” flow model with a coarse grid can be created that encompasses
physical and hydraulic boundaries that are far removed from the local study site, yet
control its groundwater flow. Specified head or specified flow boundaries can be
interpolated from the regional model for an embedded smaller scale refined local flow
model. The process is commonly referred to as “telescopic mesh refinement” (Anderson
and Woessner, 1992) or “telescopic grid refinement” (BYU, 2000). Therefore, a regional
scale model was built to estimate the spatial distribution of flow for a plume scale local

flow model.

PREVIOUS WORK

The site geology, hydrogeology, and hydrology of the Schoolcraft area were
assessed to develop the regional model. The bedrock is the Mississippian Coldwater
Shale, which is subsequently overlain by Pleistocene glacial deposits (Martin, 1957).
Investigations by Martin (1957), Dorr and Eschman (1970), Ibrahim (1970), and
Lillienthal (1978) provided descriptions and maps of the Coldwater Shale. The glacial
deposits were initially described and interpreted by Leverett and Taylor (1915).
Subsequent work by Martin (1957), Monaghan et. al. (1986), Monaghan and Larson
(1986), and Finkbeiner (1994) provided further insight into the complex glacial history of
the region surrounding Schoolcraft. The hydrogeology and geochemistry of the aquifers
in the region were evaluated by several authors due to agricultural and industrial
contamination (Barrese, 1991; Steinmann, 1994; Kehew et. al. 1996; Hyndman et. al.,

2000). The conceptual model was based on the results of the literature review.

11

Regional Geology

The bedrock underlying Kalamazoo County is mostly composed of the Coldwater
Shale except for a small portion of the northeast comer of the county that is composed of
the Marshall Formation (Figure 3). The Coldwater Shale was deposited as a sequence of
fine-grained muds in the Early Mississippian Period of the Paleozoic Era when the region
was an offshore marine environment (Dorr and Eschman, 1970). The formation is about
150 meters thick near Kalamazoo County and generally dips to the northeast (Martin,

195 7). The shale is gray to blue-gray in color, contains small amounts of limestone,
dolomite, siltstone, and sandstone, and is separated from the underlying Ellsworth Shale
by a 3-6 meter thick layer of red argillaceous limestone/dolomite commonly referred to
as the Coldwater Redrock (Martin, 1957). Ibrahim (1970) conducted extensive
geophysical mapping of the surface of the Coldwater Shale underlying Kalamazoo
County, as did Forstat and Sorenson (1982). Both showed that the bedrock surface in the
area of interest generally slopes in a WNW direction. The Coldwater Shale is considered
to be a low conductivity hydrostratigraphic unit and aquifers are subsequently only found
in the overlying glacial sediments (Kehew et. al., 1996). The Marshall Formation was not
examined since it was approximately 40 km from Schoolcraft and was not in the model
area.

Kalamazoo County’s bedrock is overlain by thick deposits of glacial sediments
(Leverett and Taylor, 1915). The western three-quarters of the county are sediments
deposited by the Lake Michigan Lobe of the Laurentide Ice Sheet, while the eastern
quarter of the county is underlain by deposits of the Saginaw Lobe of the Laurentide Ice

Sheet (Figure 4) (Leverett and Taylor, 1915). A surficial geologic map modified by

12

 

 

.IDCK GEOUXIY OF
[OWE]! PHINSUIA

m Bans
mum ERMA‘I'DN
snormw rem-mu

- MYPOIII' UMBS'IDNE

- mute»: FORMATION

- mm romxmn

- comma: SHAH!
smaunv sum:
mum s; a man at

- IDPORD sum
uswormr SHALE
m SHALE
mvm snow
mu. smut
mm mansion:

- canon mm mow

- mm mrommn
- uss numcnouv
- sum mom

 

KALAMAZOO COUNTY

 

 

 

km

 

 

>2

   
  

 

Figure 3. Bedrock map of the Lower Peninsula of Michigan with Kalamazoo County
enlarged (modified from Milstein, 1987). Note that the Coldwater Shale is Chartreuse,

while the Marshall Formation is gray. Image is presented in color.

 

 

Z
V:
E9
E.)
5'
Lu
M
<1:
..I

 

é ’ / .
$05“ / l‘ .’ . 2'” dz “Prams

 

 

 

 

 

 

 

// [DA
2% \AM
/!5 . . \C%@
I'/ .

 

 

\
/ \ \
.\ + drumlin
I W A, 4“ {27 county lines
’ I / 6,, ‘V a _._.interlobate
e°~ , boundary

 

 

 

 

r

/ / l’

G“ {3“ \ // 0 15

\ ._____.
0 as} 43 km

0 Schoolcraft

 

 

 

 

MICHIGAN LOBE MORAINES SAGINAW LOBE MORAINES

LBM - Lake Border Moraine TkM - Tekonsha Moraine
IVM - Inner Valpraiso Moraine KM - Kalamazoo Moraine
OVM - Outer Valpraiso Moraine ChM - Charlotte Moraine
IKM - Inner Kalamazoo Moraine LaM - Lansing Moraine

IKM - Outer Kalamazoo Moraine

StM - Sturgis Moraine

 

TkM - Tekonsha Moraine

 

 

Figure 4. Map of moraines in southwest Michigan and the study area of
Monaghan et. al. (1986). Note the location of the interlobate boundary that
separates deposits of the Michigan Lobe from deposits of the Saginaw Lobe

(modified from Monaghan et. al., 1986).

14

Rheaume (1990) from Monaghan and Larson (1982) for the county is provided in Figure
5. The surficial geology is generally characterized by outwash plains of sands and gravels
lying between northeast-southwest trending moraines of undifferentiated till (Monaghan
and Larson, 1982). Schoolcraft lies within the Galesburg-Vicksburg Outwash Plain.

The complex glacial stratigraphy in southwestern Michigan was studied by
Monaghan et. al. (1986). The authors correlated three till sheets to moraines from the
Michigan Lobe of the Laurentide Ice Sheet. Figure 4 outlines the study area of Monaghan
et. al. (1986) depicting the moraines of interest, the interlobate boundary that separates
deposits of the Michigan and Saginaw Lobe, and the location of Schoolcraft. The till
sheets were correlated based on clay mineral analyses of samples from moraines, till
plains, and multi-till exposures. The authors determined that the Outer Kalamazoo
Moraine of the Michigan Lobe was composed of the Saugutuk Till, while the Tekonsha
Moraine of the Michigan Lobe was composed of the Ganges Till (Figure 6) (Monaghan
et. al., 1986). No attempt was made to correlate the Glen Shores till inland or speculate
on its stratigraphic significance since it was not seen at any other site in the area. The till
stratigraphy was important since Schoolcraft is located between the Outer Kalamazoo
Moraine and the Tekonsha Moraine, thus the Ganges till was expected to lie beneath the
village.

Finkbeiner (1994) conducted an extensive investigation of the subsurface glacial
geology between the Tekonsha and Kalamazoo Moraines in Kalamazoo County (Figure
7). F inkbeiner (1994) concluded that three till units existed in the area, informally named
till 1, till II, and till III, based on cross-sections constructed from water well logs and

several wells drilled and logged by individuals at Western Michigan University. Till I

15

 

I Schoolcraft

 

A

 

 

Galesburg-Vicksburg outwash plain-Mostly medium to very coarse
sand and gravel. Pebble to cobble gravel common.

Climax-Scotts outwash plain-Mostly medium to very coarse
sand and gravel. Pebble to cobble gravel common.

Downcut glacial drainage channel-Mostly medium to very coarse
sand and gravel. Deposits from a deeply incised valley train system.

Upland moraine, kame field, or highly collapsed outwash-Mostly
sandy till. Boulders and cobbles common on the surface.

Till plain-Mostly coarse sandy and cobbly till. Clay-rich in areas.
Boulders common on the surface

 

 

Figure 5. Surficial map of Kalamazoo County (modified from Monaghan and Larson
(1982) in Rheaume (1990)). Image is presented in color.

16

 

 

 

 

w . . , E
b" 8" é o 0°
s s o o O
‘bo Q70 ‘5" A?” ‘9’ 5‘9,
:9 e s s s’ e e «i
.y 2’ f: A .5” 2’ e” e?
0 £4 0 £2 0 <9 0
<5“ 0° ¢° 0° x5“ 0° 03$ «£3:
a SAUGATUCK TILL
E
.2
2
435 GANGES “LL
15
A S O STILL (I) l
GLENN km

 

Figure 6. Michigan Lobe tills correllated to respective moraine(s) in southwest
Michigan (modified from Monaghan et. al., 1986). Schoolcraft lies between the
Outer Kalamazoo Moraine and the Tekonsha Moraine of the Michigan Lobe.

l7

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

e: *E‘ ". -'. N
0 ° ° ‘°
0 o .' °
C) o . '.
t: o o o . .
23 :1 .° ' °
:3 . o 0 o.
_ m g 1. ‘ 0 o o
5 '26 .'
>4 M -'
r ‘r- --- - x-
0.. \ o: ‘ \ \ \
J :° \ \ 1
l 5 I
) 3 l
/ 3 ‘
I : \
/ . _ . ‘
/ ,2." / \
I, ’3' Schoolcraft )
o I
.- J
f 53' l”
K\ ~ Kalamazoo County
' ‘~ ~. “ ‘ St. Joseph Countyl
\ \ _J
~ I_ _ _—
0 8 --- Boundary of F inkbeiner’s Study Area
. km ' — Trace of the Tekonsha Moraine
- ° - 0 Outer Kalamazoo Moraine

Figure 7. Map showing the approximate boundary of the Finkbeiner (1994) study area
(modified from Finkeiner, 1994). Note the trace of the Tekonsha Moraine, which is

discussed in the text.

18

was areally extensive and found to lie directly above bedrock. F inkbeiner (1994)
interpreted that till I may be pre-Wisconsinan in age based on research conducted on an
apparent paleosol (possible Sangamon Paleosol) above till I by Passero (1981). Till H
was located above a layer of outwash that was deposited on till I and is areally extensive;
however till H is missing west of Schoolcraft. Till III was a discontinuous diamicton
layer that was deposited above another layer of outwash that covers till 11 and was only
seen in the northern half of the study area, about 6 kilometers north of Schoolcraft
(Finkbeiner, 1994). Leverett and Taylor (1915) originally hypothesized that the Tekonsha
moraine may have been buried in southeastern part of Finkeiner’s study area and the
author also delineated and interpreted an unnamed till as being a buried arm of the
Tekonsha moraine (Figure 7). Kehew et. al. (1996) reiterated that this buried till sheet
may be either a buried extension of the Tekonsha moraine, a buried margin of a
drumlinized till sheet, or a combination of both. The tills were never interpreted in the
context of the tills identified by Monaghan et. al. (1986). Review of both studies suggests

that F inkbeiner’s (1994) till II corresponds to the Ganges Till of Monaghan eta]. (1986).

Regional Hydrogeology

The hydrogeology and geochemistry of the aquifer underlying Schoolcraft were
investigated by Barrese (1991), Steinmann (1994), and the subsequent results were
presented by Kehew et. al. (1996). A discrete glaciofluvial deposit named the Prairie
Ronde Fan that encompasses the Schoolcraft area was identified (Figure 8). The fan was
delineated based on topographic and geomorphic criteria. It is composed of sands and

gravels deposited by meltwaters of the stagnant ice margin that formed the Kalamazoo

19

 

 

 

 

 

 

   

 

 

 

 

 

 

 

 

2.- ? '. '. 5 N
a :3 .0 o. . .
8 . - - '~- A
U U : .0
c: 8 ' °
2 N °' ° ' 3
cc . .
__ 5 § ,. -."n ~-
51. E .-'
3.... /I
.~.' I". Kalamazoo County
‘ St. Joseph County
0 g I: Prairie Ronde Fan
' I _ Trace of the Tekonsha Moraine

 

km
' ' ° ‘ Outer Kalamazoo Moraine

Figure 8. Location of the Prairie Ronde Fan presented by Steinmann (1994) (modified

from Kehew et. al., 1996).

20

Moraine of the Michigan Lobe. Kehew et. al. (1996) notes that there are scarps in the
Kalamazoo Moraine that mark the partially collapsed channel that served as the outlet for
meltwaters that created the fan. The sediments are thickest near the Kalamazoo Moraine
and thin as one moves east-southeastward until it terminates against the margin of the
buried till sheet described by Finkbeiner (1994) and Kehew et. al. (1996).

The aquifer in the Prairie Ronde Fan was studied by construction of water table
contour maps, geochemical analyses, and through slug tests. Steinmann (1994) showed
that the recharge area for the aquifer was the topographically high area of the Kalamazoo
Moraine and that the discharge area for the aquifer was the Howard-Barton chain of lakes
to the southeast of Schoolcraft (Figure 9). Barrese (1991) determined from tritium studies
that flow within the aquifer was stratified. The author identified an upper flow system
that interacted with the ponds and wetlands in the area, and a deeper flow system that did
not interact with the surface waters until it discharged at the Howard-Barton Lake system.
Even though both systems were somewhat separated, both displayed properties of
unconfined aquifers indicating that there was no laterally extensive confining unit present
in the system. Steinmann (1994) also stated that in general, the outwash in the lower
aquifer was composed mostly of medium grained sands, while the upper aquifer was
composed of coarser grained sands. Slug tests in the aquifer indicated that conductivities

in the outwash were on the order of 10'2 cm/s.

Regional Hydrology

Rheaume (1990) investigated the regional hydrology of Kalamazoo County. The

author outlined three major drainage basins, which were the St. Joseph River basin, the

21

Paw Paw River basin, and the Kalamazoo River basin. Schoolcrafi lies in the St. Joseph
River basin. Several surface water bodies occur near the village: Sugarloaf and
Gourdneck Lakes to the north, Sunset Lake to the east, the Howard-Barton lake system to
the southeast, Spring Creek to the southeast, and F lowerfield Creek to the southwest
(Figure 9). Several wetlands also exist in the area, most notably are Kramer Marsh,
Harrison Lake, and Schug pond, all to the northwest of the village (Figure 9). The
Kalamazoo Moraine to the northwest of Schoolcraft provides a significant topographic
divide, and the groundwater divide has been shown to closely mimic this surface water
divide (Figure 9) (Rheaume, 1990). Rheaume calculated groundwater recharge from
precipitation for the county to be 23.7 cm/yr based on the method of baseflow separation

described in Freeze and Cherry (1979).

CONCEPTUAL MODEL
The regional model was developed using the conceptual model approach (Bear et.
al, 1992; Anderson and Woessner 1992). The boundaries and model layers were
conceptualized based on the results of the intensive literature review of previous
investigations of the geology, hydrogeology, and hydrology of Schoolcraft. The

numerical model was based on the conceptual model described below.

Conceptual Boundaries
Choosing the model boundaries was a critical step in the flow model design since
the flow pattern is heavily determined by the boundary conditions (Anderson and

Woessner, 1992). Starting to the north of Schoolcraft and moving in a clockwise

22

 

 

 

 

 

 

 

 

 

 

 

 

\ / - 4/ Sunset Lake

 

 

 

 

 

 

 

 

 

   
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

L l / K K Portage Creek
Mud Lake ~—-‘ A \
Howard Lake
Harrison Lake 0 All?" '
__ r Barton Lake
Schug Pond " l
Schoolcrafi ‘ '\
Spring Creek
F lowerfield Creek

 

 

/\/ Roads N Rivers/Stream W/A Moraine

 

I "x ’ Drain a Wetland Lake
0 4
2
km

 

Figure 9. Location of the significant surface water bodies in the vicinity of Schoolcraft,
Michigan that served as boundaries in the regional model. The perimeter boundaries
that are not hydrologic boundaries are no flow boundaries interpreted as topographic
divides from the DEM.

23

 

direction, the following were chosen as natural hydrologic boundaries: Sugarloaf Lake,
Gourdneck Creek, Gourdneck Lake, Sunset Lake, Barton Lake, Howard Lake, Spring
Creek, Flowerfield Creek, and the Kalamazoo Moraine (Figure 9). Natural hydrologic
boundaries were defined at surface water bodies that were shown to be locations of
groundwater discharge (Steinmann, 1994; Rheaume, 1990) and along the Kalamazoo
moraine, which was shown to be a groundwater divide (Rheaume, 1990; Steinmann,
1994). Gaps that existed between these boundaries were filled by interpreting topographic
highs that were likely no flow boundaries. Several wetlands occur within the model area
that were shown to play a significant role in the movement of groundwater in the
unconfined aquifer by Steinmann (1994) through examination of water table maps of the

aquifer. These were also represented as hydraulic boundaries in the model.

Conceptual Layers

The previous studies that were completed in the region provided sufficient
information to determine the number of layers needed to describe the model area. Based
on the investigations conducted by Barrese (1991), Steinmann (1994), F inkbeiner (1994),
and Kehew et. al. (1996), the region can be separated into three main layers in the
vertical plane (Figure 10). The base of the model was the top of the till sheet that lies on
the Coldwater Shale that was described and informally named as till I by Finkbeiner
(1994) and presented in Kehew et. al. (1996). The till was described as being
predominantly clayey and thus provided a good barrier to downward flow. Kehew et. al.
(1996) indicated that the till layer may not cover the bedrock over the entire region, but in

any case, the Coldwater Shale lies directly below the till and would also serve as a barrier

24

to water flow in the absence of the till unit. The entire bottom layer was modeled as a
layer of outwash. It was unreasonable to think that these deposits were homogeneous, but
no other data were available to further delineate the zone. The middle layer coincided
with the layer of till that occurs approximately 27 meters below the Village of
Schoolcraft and acts to retard the downward migration of plumes A-G, which
corresponds to Finkbeiner’s (1994) till H. The till was not found in the entire region,
especially in the middle of the model area where outwash was present. The deposits of
the middle layer were delineated later during an investigation of the water well logs in the
region. The top layer extended to the ground surface and was predominantly outwash
deposits except for the sediments of the Kalamazoo Moraine to the northwest of
Schoolcraft. The outwash deposits of the top and middle layer were modeled as being
slightly coarser and thus more conductive than the outwash found in the bottom layer
based on well log descriptions and slug tests presented in Steinmann (1994).

The conceptual model was the most important step taken when the regional model
was developed. The boundaries were carefully chosen based on the results of the
literature review. The model layers were determined from the literature review as well.
The data was assimilated and summarized to help ensure accuracy of the model
construction. A flow chart that presents the relevant results of the conceptualization

process is depicted in Figure 11.

25

3.8 E voEomoa mm owns: 3805—25 cobw— 338 38038 8: cc 830535 Hob: onE .388 3:032
06 0533388 9 wow: fine—Snow wage—am .328» Ecomwoc 05 mo 8308-390 Banana .3 0.5mm."

 

Beau

22m Leeaaeo I see any: .38“ 2 3:85 r eséeo

3608 05 .
E 32 2.. E 8E5 I E5 _E .033 I segue as. ass 3.82 I said I

Eamoa on .38 man—o 05.332 33:32 I

.5. was» as... 382 r _E 22:2 25 Beam i gees: 85.8.3 I

 

 

 

 

%

MM<A DM<BOE

egoaoomom \1
maéoz cogs

26

 

Purpose and Scope: Develop a regional scale model in the area surrounding the

Village of Schoolcraft, Michigan, which incorporates hydrologic boundaries that

can then be modified through Telescopic Grid Refinement to model transport and
decay of chlorinated solvents at the plume scale.

 

 

 

 

Surficial Geology: Located
in an outwash plain between
the Kalamazoo Moraine to
the NW and the Tekonsha
Moraine to the East

 

 

 

 

 

 

 

 

 

 

 

 

    

Bedrock Geology: Subsurface Glacial
Composed of the low " Geology: At least 2 till
conductivity Coldwater sheets are present, one on
Shale; aquifers limited to the top of bedrock, and one
overlying glacial sediments. between outwash layers
KAkAMAZOO MORAINE

\

l
2

  
   

OUTWASH
OUTWASH

 
   

  
   

  

La

   
   
 

TILL

 

 

 

Lower Flow System: Deep Upper Flow System: Local
regional system is recharged flow system interacts with
at Kalamazoo Moraine and with wetlands in the area,
discharges at Howard- 7 not much interaction with
Barton Lake system. deep flow regime

 

 

 

 

 

 

 

Boundaries: groundwater
divides closely mimic
topographical divides; lakes
and streams serve as true
hydrologic boundaries

 

 

 

Figure 11. Flow chart outlining the steps taken to develop the regional model in the
conceptual model process and summary of the results.

27

ARCVIEW ANALYSIS
Several preparatory steps were taken before the conceptual model was
implemented as a flow model. First, the model boundaries identified in the Conceptual
Boundaries section were digitized with the GIS software ArcView (ESRI, 1998). Second,
water well logs for the region were assimilated and examined. The well logs were used to
identify and delineate the various conductivity zones defined in the Conceptual Layers
section, and to determine layer elevation points to interpolate to the model. Both

processes enabled large amounts of spatial data to be easily applied to the model.

Boundary Digitization in Are View

The streams that served as boundaries in the model were delineated and digitized
in ArcView. As a starting point, the 30 In United States Geological Survey (USGS)
Digital Elevation Models (DEM’s) were imported into ArcView and contoured to
visualize the topography. The streams in the region were available in digital format as
ArcView shapefiles from the Michigan Resource Inventory System (MiRIS) database
(MiRIS, 2000). Examination of the streams surrounding Schoolcrafi revealed that some
of the MiRIS digitized streams appeared to flow uphill. Boutt et. al. (2001) found that
MiRIS streams/rivers appeared to not consistently flow downhill, which was clearly
unrealistic. This is the result of spatial averaging of DEM’s and slight error during the
digitization process of the MiRIS coverages. Consequently, the Hydrologic Modeling
System of ArcView was used to delineate possible drainage networks based on the DEM.
First, the FLOW DIRECTION command determined the direction that water flowed on

the DEM land surface. Next, the FLOW ACCUMULATION command determined the

28

number of cells that drained into each cell of the DEM. Lastly, the stream network was
identified using the DELINEATE STREAMS command. A threshold value of 4000 cells
accumulating into a DEM cell was found to provide a reasonable match of ARCVIEW
delineated streams to MiRIS streams; therefore, the ArcView streams were used in the
model.

The lakes, wetlands, and topographic divides that were the model boundaries
were delineated and digitized in ArcView. The lakes and wetlands were digitized from
the MiRIS files as separate polygon shapefiles so that each could be manipulated
separately during flow model construction. Significant gaps existed between the surface
water boundaries and the topographic divide of the Kalamazoo Moraine. Where gaps
existed, no flow boundaries were created by tracing topographic divides interpreted from
the DEM. After all surface water boundaries were identified and digitized, a polygon that
represented the physical boundaries of the model was digitized by incorporating the
streams, lakes, and the interpreted topographic divides that existed where significant

surface water features were absent.

Stratigraphic Unit Development

Development of the stratigraphic units was aided by a large database of well logs
for the domestic water wells in the area. Over 13,000 well logs were available from the
MiRIS database for the southern half of Kalamazoo County and the northern quarter of
St. Josephs County. Approximately 2100 wells were located in the immediate vicinity of
the model area (Figure 12A). A series of quality checks were performed to increase the

reliability of the well log information. First, the ground elevations listed in the well logs

29

were checked against the elevation of the DEM. Wells with ground elevations that
differed from the DEM elevations by more than +/- 1 m were removed from the database
because the data are suspect due to inaccurate recorded location or elevation. Second,
since the modeled area was over 90 m thick in some areas and the depth to the base of the
top layer in the vicinity of Schoolcraft was between 27—30 m below ground surface (bgs),
wells that were under 27 m in depth were removed from the database. The two quality
check procedures reduced the number of wells from approximately 2100 to only 72
(Figure 128).

The well log database and the surficial map for the area were used to differentiate
conductivity zones for the model layers and to determine layer elevations. Layer 1 (the
top layer) was composed of predominantly outwash deposits, but the end moraine
materials were delineated from the current surficial geologic map of Kalamazoo County
(Monaghan and Larson, 1982). The lithologic data of the 72 wells gleaned from the
MiRIS database were imported into the borehole module of GMS for visualization
purposes. Layer 2 was composed of till and outwash and the extent of each was
delineated based on Figure 128, which depicts wells where the till was present and wells
where outwash was present. Layer 3 was modeled as an entire layer of outwash;
therefore, no zones were delineated. The various conductivity zones were digitized as
polygons in ArcView so that the coverages could be used to adjust conductivity values
during model calibration (Figure 13). Coordinates representing the bottom elevation of
the three model layers were then chosen from the well logs and saved in files for

geostatistical interpolation to the model grid.

30

 

.5982 05 “a coma—BE 83 z: N LP?— 2: 350:3 8 $68on 3:8 8? 2m :05? 6088.289
So? 3023 3:28 Saw @8888 £63 8 .388 _mzommou 05 So: £33 683 05 .«o 3:804 2 .N~ onE

 

Eo—
H
v o

wows—82 83 N ~98— Eob £3328 20:3 :85

Basso“: 83 N 29?— an :3 20:3 :03

m

E
H
v o

283304 :0? 83¢an maaz -

 

<

 

 

 

 

 

 

31

.293— dou 2: mm fl his: 38.: Son 83:65on _mcomme 06 5 com: moaou 33:26:00 got? 05 .«o 325 :82 .3 853m

€38 05 a m

 

Eo—
H

v o

exam \/\
€5,330 8:602 E

Eo—
H
v. o

exam >\

EH a
£3380 3:30 I

Sa—
H
v c

exam >\
0:502 E .\\\\\\x
530 338 I

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

N MES

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

32

IMPLEMENTATION OF THE CONCEPTUAL MODEL
The conceptual model was implemented as a steady state finite difference

groundwater flow model. The USGS code MODFLOW (McDonald and Harbaugh, 1988)
was used to solve the 3-D steady state saturated groundwater flow equation. The
Groundwater Modeling System (GMS) (BYU, 2000) preprocessor program was used to
create the necessary input files for MODFLOW and to visualize the results.
Development of the model in GMS consisted of several steps. In general, the process was
to build a 3-D grid based on GIS data, interpolate the layer elevations from well log data
using geostatistical methods, assign model boundaries using previously delineated GIS
data, and calibrate the model to field data. Calibration was achieved through a simple trial
and error process of manipulating conductivity values until an acceptable match was

obtained between simulated and observed head values.

Development of a 3-D Finite Difference Grid

A 3-D finite difference grid was developed in GMS for the regional scale model.
The DEM was imported as a 2-D grid into GMS and a 3-D grid was then constructed so
that the model boundaries previously defined in ArcView fit within. The grid was 611
cells by 645 cells by 3 layers in the X (East) and Y (North) and Z (vertical) directions
respectively. The cells were 30 m by 30 m in the horizontal plane but variable in
thickness in the vertical plane. The DEM elevations were then interpolated to the 3-D
grid as the top elevation of the model. The grid was oriented in a manner such that the
MODF LOW cells directly coincided with the cells of the DEM to insure direct

interpolation of the DEM values. The ArcView polygon that represented the model

33

boundaries was then imported into GMS and the cells within the polygon were activated.
The model was then split into three layers in preparation for geostatistical interpolation of

the model layer bottom elevations.

Layer Elevation Interpolation

The layer elevations were interpolated using 2-D geostatistical methods in GMS.
The layer elevations were more difficult to interpolate for the lower layers than the top
layer because there were fewer wells penetrating the deeper sediments; therefore, less
data were available to use in the interpolation process. The number of sample points
representing layer elevations decreased from layer 1 to layer 2 and layer 2 to layer 3. For
layer 1, 47 wells were identified that showed the contact between the bottom of layer 1
and the top of layer 2. Many of these points, approximately 25%, were within the limits
of the plume scale model, which is described in Chapter H1. The elevations were
interpolated using ordinary kriging in GMS with a spherical model having an isotropic
range of 6700 m and a variance of 35 (Figure 14). For layer 2, only 37 wells were
identified that showed the contact between the bottom of layer 2 and the top of layer 3.
The elevations were also interpolated using ordinary kriging in GMS with an exponential
model having an isotropic range of 6700 m and a variance of 136 (Figure 15). Lastly, the
top of the till representing the bottom of the model was only found in 11 wells since very
few wells penetrated to these depths (60-90 m). Therefore, a different algorithm was
chosen as the interpolation scheme because a reasonable variogram could not be
developed for kriging. The bottom elevations of layer 3 were interpolated

using the Inverse Distance Weighted algorithm in GMS. The final model interpolated

34

layer elevations are depicted in Figure 16.

Model Boundary Assignment

The boundaries that were identified in the conceptual model were applied to the
numerical model. The surface of the top layer was a recharge boundary upon which a
value of 23.7 cm/year of recharge was assigned (Rheaume, 1990). The base of the model
was represented as a no flow boundary since it was the top of a till layer. The streams and
lakes were assigned as constant head boundaries in all three layers. The value of head in
each lake cell was equal to the value of the DEM, which was checked against the most
recent topographic map to ensure accuracy. The value of head that was assigned to each
stream constant head cell was a value equal to the DEM minus 0.6 m (2 ft). The stream
constant head values were lowered because the streams were smaller than the 30 m DEM
cells, and therefore a stream that is downcut through an area would have an elevation
lower than the averaged elevation that was assigned to the DEM cell. . The wetlands were
designated as general head boundaries in the top layer only. The justification was based
on research presented in Kehew et. a1. (1996) that showed that the wetlands in the area
typically only affected the local flow system and not the deeper regional systems. The

F ollmer drain is located to the east of Schoolcraft and it was assigned as a drain boundary
in the top layer of the model. All of the aforementioned boundaries were applied to the

model through the map module of GMS with coverages that were created in ArcView.

Hydraulic Conductivity Assignment

Several conductivity zones were identified in the model area (Figure 13) and

35

 

 

 

6O {
O.)
.2
a:
2% 40--
Eb
-:: 20*-
a:
>
0 2000 4000 6000

Lag Distance (m)

Figure 14. Variogram for interpolation of layer 1 bottom elevations.

150" ..

100“

50‘"

Variogram Value

 

 

 

0 ' 2000 ' 4000 6000

Lag Distance (m)

Figure 15. Variogram for interpolation of layer 2 bottom elevations.

36

 

Elevation (m)

   

Layer 1 ‘ Data Point

 

Elevation (111)
-

      

 

 

Layer 3 ‘ Data Point

 

 

 

Figure 16. Interpolated model layer bottom elevations. The location of the
data points from which elevations were interpolated are shown as triangles.

37

these parameters were applied to the model. Initially, representative conductivity values
for the different aquifers materials were assimilated from a variety of sources, most
notable were Mayotte (1991), Kehew et. al. (1996), and Hyndman et. al (2000). The
conductivities were assigned using coverages created in ArcView (Figure 13) that were
mapped to MODF LOW using the MAP to MODFLOW command of the map module of
GMS. Assigning conductivities in this manner allowed for easy manipulation of values

during the calibration process.

MODEL CALIBRATION
The regional model was calibrated to ensure that the model reasonably
represented the natural flow system. A data set that was specific to the plume site was
used to compare model results. The model conductivities were varied and the simulated
head data were analyzed until a reasonable match was obtained with the observed data.
The objective function was to minimize the root mean squared error (RMSE) of

simulated vs. observed heads.

Observed Head Data Set

The observed head data that were available for this project consisted of a set of
approximately 50 observations made in monitoring wells in and adjacent to the village of
Schoolcrafi. The data were collected June 23—27, 1989 by the NUS corporation during
their RI/FS study of plumes A and F. Additional data were available in the form of static
water levels from the water wells in the area that were recorded when they were installed.

The water well data may not be very reliable since they were recorded over a broad time

38

span and consequently a wide range of climatic conditions. In addition, the head readings
were recorded immediately after the wells were drilled and installed, therefore, the
aquifer was disturbed and head readings may be inaccurate. The calibration process
consisted of matching the simulated heads to the NUS data, while using the regional data
set to check that the general trend of simulated vs. observed values over the entire model
region was reasonable. Only the static water level data from wells that were surveyed to

+/- 1 meter of the DEM elevation were included in the regional data set.

Results

The steady state groundwater flow model was calibrated to minimize the RMSE
between the simulated and observed heads. This was achieved by adjusting the four
different hydraulic conductivity zones until a reasonable match was achieved. The final
conductivity values compare well with values determined by slug tests in the region
presented in Kehew et. al. (1996) (Table 2). The simulated heads are plotted relative to
the observed heads (Figure 17), which shows that the model reasonably predicts the
observed heads since values were within +/- 0.15m and the RMSE was 0.003 m. The
simulated regional heads also compared reasonably to the to the large database of heads
taken when the water wells were drilled, however the reliability of this data is subject to
question as discussed previously and a plot is not presented. A plot of the calibrated head

distribution for the top layer of the model is given in Figure 18

Discussion

The head map with superimposed flow vectors (Figure 19) shows that flow is

39

 

263

262.5 *'

Computed Head (m)

261 — ---------------- 1-0-. ------------- ----------------- ---------------- ---------------- a

 

262 ,_ ................ ............. :. ................. : ................. :, ................ ._

2615 _ ................ ....... . ......... ................. ................. ', .........

 

 

261.5

Observed Head (m)

40

Figure 17. Simulated vs. observed heads for the observation wells
in the vicinity of Schoolcraft.

 

>2

 
  

‘ I ll
l‘ l‘
ill
"il‘llllillliil'mi‘i1 i
illii il “in“ I
“ lillllm
i

    
  

   
  
   

fl H l llllll
"W llllllllllllllll "N i ii
i

\lllllllliil

“hi liliih

 
 

' “llllllll

lillliliiliililil y
Mil“ llllllll lllllllll ll

 

     

 

 

 

 

 

Heads (m) Boundary Key

I 272-276 I “\ ’ No Flow

m 268-272
264-268 N Constant Head
260'264 “WWW Constant Head
256-260 0 4
252456 E % General Head
248-252 km Drain

 

 

Figure 18. Head distribution of the calibrated regional model (view is of the top layer).

41

generally in a southeasterly direction, but localized variations in the flow pattern exist

near surface water features. Groundwater is discharging at streams and lakes, and the

wetlands are providing recharge to the system. The gradient of the water table is

relatively low, approximately 0.0012 near Schoolcraft. The shallow gradient was

expected since the topography surrounding Schoolcraft is very flat. Examination of

Figure 19 also shows that if plumes F and G were to continue to move downgradient,

they may eventually reach Barton Lake.

 

 

 

 

 

 

 

 

 

 

 

 

Table 2. Hydraulic conductivities of the various materials in the regional model.
Material Model K (cm/s) K (cm/s)l K (cm/s)2
coarse outwash 6.6 e-2 2.9 e-2 - 5.7 e-2 2.7 e-2
medium outwash 5.5 e-2 NA NA
moraine 3.0 e-4 NA NA
sandy till 1.0 e-5 NA NA
1. From Kehew et. al ., 1996
2. From Hyndman et. al ., 2000

 

 

Two minor difficulties arose during the model calibration process that affected the
model results. When the model was first constructed, the DEM elevations were assigned
as the elevations of the stream constant head boundaries. The simulated heads were
approximately 0.6 m (2ft) too high for this scenario. The conductivity values were
increased in order to lower the modeled heads, but even at unreasonable values (10 cm/s
in the coarse outwash), the heads were still too high. The model was reevaluated and it
was determined that streams in the model area were not 30 m wide, and therefore spatial
averaging of the DEM likely increased the elevations of cells where streams existed. The

stream constant head elevations were subsequently lowered by 0.6 m, which fit well with

42

 

>2

 
    
     

l'lll' 7
ll"

|
.. ‘ ..Il
. j .‘l ‘ I 1
w l ,. ll|||||
[/— . R .. | ~ - .Ill”ill A
r 4

{lll'illiilk

. |.
lilll '

 

 

 

 

 

 

 

Heads (m) Boundary Key

I 272-276 .. ,

.. I. 268-272 ’ ‘ N°F1°W

llllllll 264-268 Constant Head
260-264 llllllllllllllllll Constant Head
256-260 0 4
252—256 E E General Head
248'252 km Drain

 

 

Figure 19. Head distribution of the calibrated regional model with flow vectors to show
direction of groundwater movement predicted by the model (view is of the top layer).

43

the calibrated conductivities. Lowering the stream cell elevations by a smaller value
resulted in conductivities that were much higher than the observed values, and lowering
the elevations by a larger value resulted in conductivities that were somewhat lower than
the observed values. Also, the wetlands located to the southwest of Schoolcraft were
shown to play a significant role in controlling the head distribution. Initial model results
predicted gradients near the wetlands that were too steep, as if they were not providing
enough recharge. As a result, the conductance of the wetlands was increased to provide
more influence from the wetlands, and the predicted gradients were much more

reasonable.

44

CHAPTER III

THREE-DIMENSIONAL PLUME SCALE STEADY STATE GROUNDWATER
FLOW MODEL OF SCHOOLCRAFI‘ MICHIGAN

INTRODUCTION

A local plume scale flow model was developed to better represent the highly
heterogeneous outwash deposits containing plumes A-G but was linked to the natural
hydrologic boundaries through the regional model (Figure 20). The stages of the model
development were somewhat similar to that used in Chapter II. A conceptual model of
the heterogeneity in the aquifer was developed using numerous well logs. The conceptual
model was then implemented in MODFLOW using several modules of GMS. The
conductivity values for the various aquifer materials were adjusted until a reasonable

match was made between the simulated and observed data.

CONCEPTUAL MODEL
The conceptual model approach was used to develop the local model. First, the
model domain within the regional model was identified. The regional heads were then
evaluated to determine an appropriate method to apply the regionally derived boundaries.
Lastly, the aquifer materials were characterized so that they could be incorporated into

the model. The conceptual model was the basis for the numerical model.

Conceptual Boundaries
The regional geology that was outlined in Chapter II was examined to determine

the extent of the plume scale flow model. The three-layer regional model extended from

45

the ground surface to the till layer above the Coldwater Shale. The bottom two layers of
the regional model were not incorporated into the local model because the discontinuous
till unit that served as layer 2 of the regional model was continuous in Schoolcraft. The
till layer was a barrier to downward flow and thus was the bottom of the local model
(Figure 21A), while the ground served as the top of the model (Figure 21B).

The regional heads near Schoolcraft were evaluated to determine an appropriate
method to include the regional boundaries in the plume scale flow model. A common
method to identify a local model based on regional flow patterns is to define stream tubes
within the flow domain of the regional model. The stream tubes represent no flow
boundaries on the sides of the model that are parallel to flow. The upgradient and
downgradient sides of the model that are perpendicular to flow are then assigned as
constant head boundaries, where each typically coincides with an equipotential line. This
method was not easily implemented for Schoolcraft since the flow paths were irregular
due to the numerous wetlands near the village. Instead, a more appropriate method was to
define all boundaries of the local model as constant head. The process acted to embed the

local model within the regional model preserving the complex head distribution.

Conceptual Aquifer Material Types

Several well logs were examined and interpreted to identify the nature of the
outwash deposits that comprise the aquifer. A series of nine wells were drilled in a field
downgradient of the ARCO facility that are a part of a test grid to conduct in situ
experiments by MSU. The wells were drilled using a hollow stem auger and the core was

taken using a Waterloo continuous cohesionless sand sampler. The cores were visually

46

inspected to differentiate the various grain sizes of the outwash (Appendix A). In addition
to the nine MSU wells, over 100 well logs were previously collected during several
subsurface investigations by the NUS Corporation during the initial stages of the RI/FS
studies conducted in the late 1980’s (Figure 20). The NUS well logs are available in
Mayotte (1991) and show that the wells were drilled no deeper than a few centimeters
into the till layer that is approximately 27 meters below ground surface. The general trend
of the sediments was cycles of fining upward sequences of gravels and coarse sands to
medium and fine sands. This observation fits the previous interpretations that the deposits
were the result of a system of braided streams emanating from the meltwaters of the
Michigan Lobe of the Laurentide Ice Sheet (Kehew et. al., 1996). Distinct laterally
continuous layers on the scale of the local model were not identifiable because of the
nature of braided stream deposits. After examining the well logs, it was determined that
the deposits could be represented by five different material types, a qualitative

description of which is listed in Table 3.

 

Table 3. Qualitative description of the materials comprising the aquifer in the plume
scale model domain.

 

 

 

 

 

 

Material General Description
1 very fine to fine sands, subrounded, medium sphericity, well sorted
2 fine to medium sands, rounded, meduim sphericity, well sorted
3 coarse sands, subangular, medium sphericity, poorly sorted
4 medium to coarse sands, some fine gravel, subangular to subrounded, low
sphericity, poorly sorted
5 coarse sands, gravels, and pebbles, subangular to subrounded, medium

 

 

 

sphericity, poorly sorted

 

47

LEGEND

4 NUS Monitoring Well

I MSU Monitoring Well

0 500
SE

 

Figure 20. Location and orientation of the plume scale model in reference to the
regional model (cells are not shown as they are too small to be seen at the current
scale). Well locations are given for NUS and MSU wells where lithologic data were
available for aquifer characterization.

48

 

    

Elevation (m)
240 .5

240 .0
239 .5
239 .0

i-i 238 .5

— 238 .0

 

— 23?.5
23? .0

 

 

 

Elevation (m)
287.5
287.0
288.5
.; ,. 288.0
‘ 285.5
285.0
284.5
284.0
283.5
283.0
282.5

  

‘lllllllllllllllllilllillli ulliii ‘ I

lll hi‘lliii lllll!’

 

 

B

 

 

 

Figure 21. A) Contour map of the bottom of the plume scale flow model, which is also
the bottom of layer lin the regional model. B) Contour map of the ground elevations
in the plume scale flow model. The contour interval is 0.5 m.

49

IMPLEMENTATION OF THE CONCEPTUAL MODEL
The plumes scale steady state model was created for MODFLOW through GMS.
The complex stratigraphy was incorporated by using the Borehole, TIN, and Solid
modules of GMS. Telescopic grid refinement was performed as a regional to local
conversion in GMS to embed the local heads within the regional model. The local model

was calibrated with the same head data as the regional model.

Development of a 3-D Finite Difference Grid

A 3-D finite difference grid was created in GMS that represented the plume scale
flow model. Several factors were considered before the grid was built including the
extent of plume G, the location of the wetlands near the village, and the appropriate
vertical discretization to adequately incorporate the aquifer heterogeneity. After these
factors were considered, a 3-D finite difference grid was created and the physical
attributes are given in Table 4. The grid was rotated in the X/Y plane 47 degrees in a
counterclockwise direction to align its long axis roughly parallel to groundwater flow.
Also, it should be noted that the cells of the local model were 10 m by 10 m (in the
horizontal plane), which was scaled down from 30 m by 30 m in the regional model. The
location of the local model in reference to the regional model is shown in Figure 20 along

with the locations of boreholes that were used to conceptualize the aquifer materials.

 

Table 4. Physical characteristics of the plume scale groundwater flow model (all cells
were active).

 

 

 

 

 

 

 

 

 

 

Direction length (m) length (ft) # cells cell size (111) cell size (ft)
X 1200 3937 120 10 32.81
Y 2300 7546 230 10 32.81
Z variable variable 10 variable variable

 

50

 

Creation of Aquifer Solid Materials in GMS

Three-dimensional solid objects that represented the aquifer materials described
in Table 3 were used to incorporate the aquifer heterogeneity into the plume scale model.
The borehole data from the MSU (Appendix A) and the NUS (Mayotte, 1991) wells were
imported into GMS. Triangular interpolated networks (TIN ’s) were created from the
boreholes that represented the different aquifer materials. Solid objects were created from
the TIN ’s and were assigned hydraulic conductivities. The conductivity values that were
given to each material were obtained from slug tests conducted by the NUS Corporation
during the RI/FS of plumes A and F (the well screens were between 1-2 m long in each
well and were then correlated to the different aquifer materials). The slug test data were
interpreted using the method of Bouwer and Rice (1976) (Mayotte, 1991). The initial
value that was assigned was the average value determined for each material. The
resulting solids were the basis for the various layers of the MODF LOW finite difference
grid. Several cross-sections through the solids, two parallel and two perpendicular to
groundwater flow, are given in Figure 22 that represent the aquifer materials in the local
model. A more detailed description of the solid stratigraphy modeling is provided in

Appendix B.

Interpolation of Aquifer Solids to the MODFLOW Grid

The aquifer heterogeneity was applied to the MODF LOW grid using a mapping
function in GMS that interpolates layer elevations and assigns material properties to cells
in the MODFLOW grid based on the 3-D solids. Three mapping algorithms exist in GMS

and each could have potentially yielded different results. The three methods of

51

interpolation are referred to as Boundary Matching, Grid Overlay, and Grid Overlay with
Keq by GMS (BYU, 2000). The Boundary Matching algorithm creates layers that strictly
match the stratigraphic layers defined by the solids while also maintaining continuous
layers as per MODF LOW requirements. The major disadvantage of this interpolation
scheme is that thin layers are often created where stratigraphic units pinch out, which will
cause convergence problems in MODF LOW and subsequent transport simulations, thus
this method was not used. The Grid Overlay method creates layers that are relatively
uniform in thickness at a given X/Y location. The material properties are assigned based
on which solid encompasses the center of the grid cell. The Grid Overlay with Keq option
is identical to the Grid Overlay method in terms of how the layer elevations are assigned,
except that the conductivities in cells where multiple materials are present are calculated
differently. GMS uses an algorithm that computes equivalent horizontal and vertical
conductivities based on the volume of each material in the cell. Figure 23 provides a
visual representation of the three different interpolation methods for an arbitrary model
(BYU, 2000).

A final consideration that was given before the solid materials were mapped to the
grid was that it would be unnecessary to discretize the materials above the water table as
MODFLOW only describes the saturated portion of the aquifer. Inactive cells would
result if multiple layers existed above the water table, but the problem was easily
circumvented. Examination of the well logs revealed that the material above the water
table was a relatively homogeneous unit of fine sand. There probably was more
heterogeneity in this zone, but it was not identified in the well logs. Subsequently, the

solids above the water table were truncated so that GMS would not discretize the

52

 

 

 

 

 

 

 

 

 

 

 

 

 

I LEGEND I

D Material 1 — Very Fine to Fine Sand

- Material 2 — Fine to Medium Sands

- Material 3 — Coarse Sands

- Material 4 — Medium to Coarse Sands some Gravel
- Material 5 — Coarse Sands, Gravels, and Pebbles
————— Location of Cross-Sections

A NUS Monitoring Well

MSU Monitoring Well

Figure 22. Selected cross-sections through the 3-D GMS solids that represent the
aquifer materials in the plume scale model. The locations of the wells used to create
the solids are given on the inset. Image is presented in color (VB = 20X).

53

 

 

 

 

 

 

 

 

D

 

 

Figure 23. Graphical depiction of the results of the three GMS algorithms to map solid
material properties to a hypothetical MODFLOW model. A) An arbitrary set of solids.
B) Cross-section through A) using Boundary Overlay. C) Cross-section through A)
using Grid Overlay. D) Plan view of layer in A) showing the smoothing effects of Grid
Overlay with Kwq (modified from BYU, 2000). Image is presented in color.

54

materials above it. The DEM elevations were interpolated back to the local grid alter the
material properties and layer elevations were assigned from the solids to correct the
ground elevations. This action stretched out the top layer, while leaving vertical

discretization of the lower layers unchanged.

Boundary Assignment

The model boundaries that were previously identified were applied to the model.
The calibrated head distribution from the regional model was interpolated as the starting
heads of the local model using the Inverse Distance Weighted Algorithm in GMS. The
four sides of the model were then assigned as constant head boundaries. The top of the
model was assigned as a recharge boundary with a value of 23.7 cm/yr that was used in

the regional model.

MODEL CALIBRATION

Results

The model was calibrated by adjusting the conductivities of the five different
aquifer materials. Table 5 provides information on the conductivity data available to
calibrate the local model. Listed in the table for each of the five material types are the
number of values, the average K, the high K, the low K, and the final K for each material
used in the model. The conductivities were constrained between the high and low
conductivity values for the respective material during the calibration process. Also,
during calibration it was noted that heads were mounding in the interior portions of the

model, which was fixed by lowering the recharge to 12 cm/yr. The resulting calibrated

55

local model heads match the heads from the regional model (Figure 24). The simulated
heads were plotted with respect to the observed heads for the regional and local models,
close inspection reveals a reasonable match (Figure 25). Also, the root mean square error
(RMSE) of the regional calibration was 0.003 m, while the RMSE for the local

calibration was 0.006 m, which is also a reasonable match.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 5. Hydraulic conductivity data (K) for the different aquifer materials described in
Table 3 (observed data from Mayotte, 1991).
Material # values ligh K (cm/s) Low K (cm/s) Avg. K (cm/s) Model K (cm/s)
1 3 4.00E-02 4.50E-03 9.30E-03 2.60E-02
2 6 5 .80E-02 6.30E-03 4.10E-02 3 .50E-02
3 5 9.00E-02 4.00E-02 5.50E-02 6.00E-02
4 1 1 7.50E-02 3.50E-02 4.60E-02 8.80E-02
5 5 1.10E-01 5.20E-02 7.10E-02 1.00E-01
Discussion

The calibrated head distribution was extremely important since all transport
modeling was completed using the plume scale model. Early simulations predicted the
observed head values at the monitoring wells rather well, but the head pattern did not
adequately match the regional distribution; the heads were too high in central area of the
model. Initially, it was thought that the starting heads were interpolated incorrectly from
the regional model. Inspection of contours from the regional model overlain upon the
plume scale model showed that the boundaries were interpolated correctly. The next
logical step was to increase hydraulic conductivity. The conductivity values were
increased in all of the aquifer materials to a value equal to the highest observed values in

Table 5, but the problem persisted. The aquifer materials were then assigned the

56

0 500
SE

meters

 

 

Head (m)

2628-2632
2624-2628
2620-2624
1' '1 261.6-262.0
“ 2612-2616
260.8-261.2
2604-2608

 

 

 

 

Figure 24. Head distribution of the top layer of the calibrated plume scale model,
which shows flow is in a southeasterly direction. The local heads are shown in
grayscale, while the regional heads are superimposed in black. Inspection between the
two distributions reveals a reasonable match.

57

 

263 I I I I I I I I I I I I T I I I I I f I I I I I

262.5 e ---------------- ----------------- ----------------- ---------------- a

l
u
v
i a
I . '
I ‘ ‘
u
‘ s
I . '
.
:26:Z ._ ................ r ............. ....' .............................. . ..................
a
u
l , '
' s
. . I.
l— ' I
. ‘ . '
n . .
9
' I
u
l ‘ I
,
. . ‘
. .
n
I u
u
I I
-

Computed Head (m)

261.5 — """""""" """"""" """"""""" """"""""" """""""" ‘

 

261 __ _______________ €09 ____________ __________ 9 Regional Model _________
_ Q; 3 0 Local Model

 

 

 

 

Observed Head (m)

Figure 25. Plot of simulated vs. observed heads for the plume scale
and regional scale model.

58

conductivity value of the layer 1 outwash from the regional model and simulated as
homogeneous, but again the problem persisted. The recharge was reduced by 50%, and
the head distribution of the plume scale model matched the regional distribution much
more reasonably. If the problem had not been fixed, the subsequent transport simulations
would have predicted irregular plume shapes that were spreading in a direction transverse

to flow much more than what was observed.

59

CHAPTER IV
THREE-DIMEN SION AL REACTIVE TRANSPORT MODEL OF PLUME G,
SCHOOLCRAFT, MICHIGAN
INTRODUCTION
The transport and decay processes of plume G compounds emanating from

ARCO industries in Schoolcraft, M1 were evaluated using a numerical model. The model
was developed by analyzing past investigations of the ARCO site, reviewing literature
pertaining to degradation of the source chemicals, and evaluating geochemical data to
evaluate potential transformation pathways. The transport, sorption, and decay of the
plume was represented using an RT3D model, which was built in GMS within the plume
scale flow model described in Chapter HI. The model was calibrated using a new
approach that compares calculated ratios of compounds to their daughter byproduct
(termed mother/daughter ratios) for observed field data and model simulated results. The
degradation rates were adjusted during the calibration process but were constrained
between observed rates presented in Suarez and Rifai (1999). The model reasonably
predicted mother/daughter ratios at a monitoring well transect downgradient of the
ARCO facility. The calibrated degradation rates were low in comparison with the range
presented in Suarez and Rifai (1999), but are on the same order of magnitude as those

observed at the Dover site as reported by Clement et. al. (2000).

ARCO SITE BACKGROUND

The source of plume G was ARCO Industries Corporation (ARCO), located on

East Eliza Street in Schoolcraft, Michigan. ARCO was an automotive parts manufacturer

6O

that used chlorinated solvents to degrease parts in the production process and to clean the
plant floor. The facility opened in 1953, but ARCO did not assume operations until 1967.
Extensive sampling conducted in the mid 1980’s by ARCO led to a detailed description
of the extent of contamination on the company’s property. Several remediation systems
were implemented in the past to mitigate the damage caused to the environment. A
review of the site investigations and remedial activities was used to develop a conceptual

reactive transport model.

Source Areas

Five regions on the ARCO property were identified as possible sources of
contamination based on visible inspection of the property and through discussions with
plant employees (Everett, 1990) (Figure 26). Area I was a material and waste-handling
site that was used throughout plant operations. Area H was an unlined pond that was used
to hold drainage waste from the plant from 1953 until approximately 1965. Area III was a
material and waste-handling site that was used throughout the facility’s operation. The
South Seepage ponds were unlined pits used for handling wastewater from the plant after
the Area H pond was closed in 1965. The Plasticizer Loading Area (PLA) was used for
transferring bulk plasticizers from trucks into the building. The vadose zone in these five
areas was sampled to identify and quantify the extent of contamination.

Vadose zone soil sampling in the five suspected source areas was conducted
numerous times from 1986-1991. Samples were taken mainly in the top 3 meters of
sediments, but in some cases, samples from depths of up to 6 meters were obtained and

analyzed. Area I was the most affected site where sediments were contaminated with

61

 

 

 

 

 

 

 

 

 

 

 

l 8 ii
l 'l'C
l lr\
SOUTH SEEPAGE POND Ir j!
' — 0 40 K
1 meters

 

 

 

 

Figure 26. Location of the five suspected source areas on the ARCO property
(modified from Everett, 1990).

62

tetrachloroethene (PCE) at up to 300,000 rig/kg, trichloroethene (TCE) at up to 1,200,000
jug/kg, and trichloroethane (TCA) at up to 14,000 pig/kg. Area H was not as heavily
contaminated as Areal, but concentrations of PCB and TCE were 200 [Lg/kg in some
portions of the old pond, while no TCA was detected. Area IH sediments were somewhat
more contaminated than Area H with PCB and TCE concentrations at about 20,000 rig/kg
and TCA concentrations at about 15000 [lg/kg. The South Seepage Pond contained a
bottom sludge material that ranged in thickness from 0 to 2 m. Levels of PCB and TCE in
the sludge were about 100 rig/kg, while TCA concentrations were about 20 rig/kg. A soil
gas survey conducted in December of 1989 at the PLA indicated that chlorinated organics
were present, but Everett (1990) stated that they were at very low levels. As a result, no
further sampling was done in the PLA and it was not considered a significant source of

groundwater contamination.

Previous Remediation Systems

Area I was the most heavily contaminated sector of the ARCO property and
ARCO agreed to remediate the vadose zone sediments of Area I. A three-phase test was
conducted from June of 1987 to September of 1988 to test the applicability of a soil vapor
extraction (SVE) system at Areal. The Phase I system was composed of one injection
and one extraction well while the Phase H and 1H systems included the Phase I wells, but
also incorporated another set of injection and extraction wells. The three SVE tests
removed an estimated 110 kg of chlorinated organics; approximately 70 kg were PCB, 28
kg were TCE, 8 kg were 1,2-dichloroethene (1,2-DCE), and 4 kg were TCA. Although

the test showed promising results, a full-scale system was never implemented for reasons

63

that were never stated.

Contaminated sludge was excavated from the South Seepage pond as the SVE
system was being tested at Area 1. Approximately 115 m3 of contaminated sludge were
removed from the bottom of the northeastern section of the pond in June of 1988.
Another 64 m3 of sludge were removed from the remaining contaminated portions of the
pond in May of 1989. The sludge materials were disposed off site during both excavation
operations. The entire pond was back-filled with clean sand to match the existing grade
after the excavations were completed (Everett, 1990).

The remedial activities at ARCO shifted to development of a remediation system
to treat the contaminated groundwater in the early 1990’s. The Michigan Department of
Environmental Quality (MDEQ) initially decided to remediate plume G with a pump and
treat system. The proposed system was composed of three purge wells, two of which
already existed and were located roughly on the longitudinal centerline of plume F (PW-1
and PW-2), and one well that was installed (PW-3) (Figure 27). The stripping tower was
located downgradient of ARCO and discharge was controlled by an NPDES permit that
did not allow release of chromium or arsenic to the receiving water. The problem with the
system was that PW-l and PW-2 would capture arsenic and hexavalent chromium from
plume F in addition to the chlorinated organics from plume G if they were activated. PW-
3 was located far downgradient of plume F and about 300 meters in front of plume G (at
its 1990 location), therefore chromium and arsenic were not an immediate problem.
Subsequently, PW-l and PW-2 could not be used until an appropriate pre-treatment
strategy was implemented to remove arsenic and chromium from the captured water

before it was treated and discharged at the stripping tower. The pretreatment system was

64

A PW—l
I PW—2
° PW-3 m

 

Figure 27. Location of the three purge wells that were part of the initial pump and
treat system to remediate plumes F and G (modified from Mayotte, 1991).

65

never developed and PW-l and PW-2 never went online. However, PW-3 was activated

on July 13, 1992 at a rate of 1090 m3/day and is still in operation.

DEGRADATION PROCESSES

The suspected initial contaminants of plume G were PCE, TCE, and TCA, as
determined from past site investigations (Everett, 1990). Degradation byproducts of the
source contaminants, such as cis-l,2-dichloroethene (cis-1,2—DCE), trans-1,2-
dichloroethene (trans-1,2-DCE), 1,1-dichloroethene (1,1-DCE), vinyl chloride (VC), and
1,1-dichloroethane (1,1-DCA), were present downgradient of the source regions. The
presence of the daughter by-products implies that contaminant degradation processes had
occurred at the site. The literature was reviewed to understand the degradation pathways
of the plume G contaminants in order to develop a model to mathematically describe the
transformation processes.

Contaminants are transformed through biotic or abiotic proceses in natural
systems. Bouwer et al. (1981) presented the first evidence of biodegradation of
chlorinated solvents. Since that research, many studies (including Vogel and McCarty,
1985; Freedman and Gosset, 1989; Semprini et. al., 1995; de Bruin et. al., 1992; Bradley
and Chapelle, 1998) have examined processes that degrade chlorinated ethenes/ethanes in
groundwater environments. A conceptual model for the transformation of PCB, TCE, and
TCA that was the result of the aforementioned research is depicted in Figure 28. One of
the main concerns addressed in Michigan State University’s bioremediation proposal to
the Michigan Department of Environmental Quality was the abundance of VC at plume

G (the MCL is 2 ppb). Review of the conceptual degradation pathways (Figure 28) shows

66

 

CCIZCCIZ

 

 

 

 

 

 

 

 

 

 

 

 

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

PCE
CClzCHCl CH3CC13
TCE TCA
5}} El 18351;}; <-- -_,. BC] '+ productsl
)5
CClZCH2 CHClCHCl CH3CHC12
1,2-DCE 1,1-DCE 1,1-DCA
-'i'cEi'I§£c§£1{{e1;’-+- -->E-'2'Ei'l'prs&uetsi
CHClCH2 CH3CH2C1
VC CA
i'éi';’§;£,211.‘c1;‘5<--
t r"
CHZCH2 CH3COOH
Ethene Ethanol
Anaerobic Pathway {1.9116131515511051 ..... )Abiotic Pathway

 

 

 

 

Figure 28. Biotic and abiotic transformations of PCB, TCE, and TCA under aerobic and
anaerobic conditions (modified from McCarty, 1997; and Bedient et. al., 1999). All
transformations are biotic unless indicated with a dotted arrow, which is the abiotic
pathway. Heavy arrow indicates preferential reaction.

67

that VC can be formed by degradation of PCE/TCE or TCA. Therefore, literature that
addressed transformation processes that lead to the production of vinyl chloride were

reviewed.

Abiotic Degradation Processes

Abiotic transformation processes are less studied in groundwater systems because
TCA is the only major chlorinated solvent that can be degraded at reasonable rate in this
manner (McCarty, 1997). TCA can chemically transform into 1,1-DCE or ethanol (acetic
acid) (McCarty, 1997). Several researchers determined in laboratory experiments that the
preferential abiotic reaction of TCA was ethanol formation (Haag and Mill, 1988; Cline
and Delfino, 1989). Both studies determined that approximately 79% of TCA
transformed to ethanol and 21% transformed to 1,1-DCE. Although only one fifth of the
degraded product is 1,1-DCE, the process is important because the MCL of 1,1-DCE is 7
rig/L and it can subsequently degrade to VC (McCarty, 1997). McCarty (1997) reported
that when 1,1-DCE is found in groundwater that it is probably the result of the chemical

transformation of TCA.

Biotic Degradation Processes

Contaminants are biodegraded directly or indirectly through the
oxidation/reduction reaction of microbial respiration. The basic idea is that a coupled
redox reaction occurs where electrons are passed from an electron donor through a
complex series of biochemical reactions to an electron acceptor in order for the organism

to obtain energy (Chapelle, 1993). The electron donor in microbial reactions is often

68

referred to as the primary substrate and in aquifers is usually the organic carbon that is
present, such as humic or fulvic acid (Chapelle, 1993). The common electron acceptors in
groundwater systems in order of decreasing energy yield are dissolved oxygen (DO),
NO3', Fe(HI), Mn(IV), S047”, and CO; (Azadpour-Keeley et. al., 2001). The primary
substrate is oxidized during metabolism because it loses electrons, while the electron
acceptor is reduced because it gains electrons. A side product of metabolism is that
enzymes and/or cofactors can be produced and are released into the surrounding medium
(Chapelle, 1993). Microbial respiration and enzyme/cofactor production are an integral
part of biodegradation.

Microbial degradation mechanisms are divided into aerobic, anaerobic, and
cometabolic processes (McCarty, 1997). Aerobic and anaerobic simply refer to whether
the process occurs in the presence (aerobic) or absence (anaerobic) of oxygen.
Cometabolic reactions occur when a contaminant is degraded by an enzyme or cofactor
produced in microbial metabolism. Such processes have been referred to as fortuitous
because the organism that produced the enzyme/cofactor derives no energetic benefits
from the reaction (McCarty, 1997). Cometabolic reactions can occur in aerobic or

anaerobic environments.

Biotic Degradation Process: Occurrence Under Aerobic Conditions

Aerobic processes are more energetically favorable to microorganisms than
anaerobic processes, but are often not dominant transformation mechanisms in plumes of
chlorinated hydrocarbons. PCE, TCE, and TCA do not serve as electron donors or

acceptors in aerobic groundwater systems (McCarty, 1997). TCE can be cometabolically

69

degraded in aerobic environments, but PCE cannot (McCarty, 1997). Significant
cometabolic degradation of TCE was only found to occur when large amounts of
dissolved oxygen (DO) and electron donors (methane or phenol) were present (McCarty
and Semprini, 1994). DCE and VC have the potential to be either directly oxidized or
degraded cometabolically in aerobic environments with recent field evidence provided by
Bradley and Chapelle (1998). TCA can be cometabolically degraded in aerobic
environments, but the process is slow, while DCA can be oxidized in an aerobic system

(Bedient et. al. , 1999).

Biotic Degradation Processes: Occurrence Under Anaerobic Conditions

The two degradation processes in anaerobic environments are reductive
dechlorination and cometabolism. Reductive dechlorination occurs when the contaminant
substitutes for either NO3', Fe(HI), Mn(IV), 8042', or C02 as the terminal electron
acceptor (Wilson et al., 1997). For example, PCE can accept electrons produced during
the metabolism of an oxidizable carbon source (electron donor) thereby releasing a
chloride ion to become TCE (Azadpour-Keeley et. al., 2001). PCE and TCE can be
completely dechlorinated to ethene through the sequential reduction reaction presented in
Figure 28 in an anaerobic environment (Wilson, 1997). Also, PCB and TCE can be co-
metabolically degraded in anaerobic environments (McCarty and Semprini, 1994). TCA
can be reductively dechlorinated to DCA, which can be filrther dechlorinated to
chloroethane (CA). CA is relatively recalcitrant to biological processes, but can be

abiotically converted to ethanol (McCarty and Semprini, 1994).

70

Biotic Degradation Processes: Rate Sensitivity to Redox Conditions

The redox conditions of the groundwater system control the relative rates of
degradation. Reaction rates are much higher for reductive dechlorination processes when
strongly reducing conditions exist, such as methanogenesis, than when weaker reducing
conditions exist, such as denitrification. Also, it has been observed that anaerobic
transformation rates are higher for the more chlorinated compounds (i.e. PCE, TCE, and
TCA) than the less chlorinated ones (DCE, DCA, and VC), because they are already
relatively reduced (McCarty and Semprini, 1994). Lastly, aerobic transformations of the
less chlorinated species (DCE, DCA, and VC) are much faster than the anaerobic

pathways (Chapelle, 1997).

CURRENT METHODS TO ESTIMATE FIELD SCALE DEGRADATION RATES
The transport model of plume G required degradation rates as input parameters;
therefore; a literature review was completed to determine the best method to estimate the
field scale degradation rates of plume G compounds. The review revealed three methods
that are currently used. Wiedemeier et. al. (1996) provides a detailed description of two
of the methods, which are tracer normalization and a method derived by Buscheck and
Alcantar (1995). The third method involves adjusting degradation rates in a numerical
model to match field data and was demonstrated recently by Clement et. al. (2000). A

description of the methods is provided along with the advantages/disadvantages of each.

Tracer Normalization

Wiedemeier et. al. (1996) presented a method to estimate field scale degradation

71

rates between two points along a flow path, which was referred to as tracer normalization.
Recalcitrant organic or inorganic compounds associated with the contaminant plume are
used as tracers to quantify the non-destructive attenuation mechanisms. Two tracers that
are commonly used are trimethylbenzene (TMB) isomers, which are only present if fuel
hydrocarbons are a co-contaminant, and chloride. The first step is to normalize the

contaminant concentration at a downgradient point using the following equation:

CB corr : CB[T—A]
. TB

where C 3w, = corrected contaminant concentration at point B downgradient of A

CB = measured contaminant concentration at point B

TA = tracer concentration at a point A upgradient of B

TB = tracer concentration at point B downgradient of A
The corrected value is the anticipated concentration that would result if dispersion and
dilution had not occurred between points A and B. The corrected value theoretically
reflects only contaminant degradation. The first order decay rate between the two points

can be calculated using the following equation:

C

_ -N

B,corr
where C3,“,rr = normalized contaminant concentration at downgradient point B
CA = contaminant concentration at upgradient point A
h = first-order biological decay rate [I/T]

t = travel time from point A to point B for the given contaminant

The advantages of tracer normalization are that it is relatively straightforward, the

calculations are easy and take minimal time, and the results are readily available to

72

compare with other study sites. However, several drawbacks exist to the tracer
normalization method. The primary contaminants and their daughter byproducts are
usually the only compounds quantified during a sampling event. As a result, tracer data
are often not available to perform the normalizations. Even if tracer data are available, the
method assumes that the two points where the contaminant concentrations were measured
lie along a flow path and that the travel time between the two points is known. Unless the
flow regime has been studied extensively and subsequent sampling points sited in a

manner to incorporate flow paths, the results would be questionable.

Semi-Graphical Method of Buscheck and Alcantar (1995)

Buscheck and Alcantar (1995) derived a semi—graphical method to estimate field
scale degradation rates and Wiedemeier et. al. (1996) indicated that the method is an
alternative to the tracer normalization method. First, a In-linear plot is constructed with
contaminant concentration plotted relative to distance along a flow path. Once the plot is

created, the first order decay rate is calculated from the following equation:

2
h: V“ [1140(1)] —1
4a,r v"

where h = first-order biodegradation rate constant

 

vC = retarded contaminant velocity in the x direction
01,, = longitudinal dispersion

k/vx = slope of the line formed by making a ln-linear plot of contaminant
concentration relative to distance along a flow path

The semi-graphical method requires less data than the tracer normalization method, it is

also simple and straightforward, and the results can be readily compared to published

73

rates in order to compare the site to other research areas. There are some disadvantages to
the semi-graphical method. First, the equation was derived for a steady state plume.
Determining that a plume is at steady state requires a large data set collected through time
to ensure that steady state conditions exist, which is expensive and the data are often not
available. In addition, the equation only applies to l-D systems and is therefore limited

for complex 3-D flow systems.

Numerical Modeling

Field scale degradation rates can be estimated by calibrating a numerical model to
field data. Clement et. al. (2000) provides an example of this methodology through
research conducted at the Dover Air Force Base in Delaware. The authors created a 2-D
flow and reactive transport model based on site-specific geologic, hydrogeologic, and
geochemical data. Mesocosms were used to determine laboratory rates of degradation,
which served as a starting point during their model calibration. The degradation rates
were bounded during the calibration process by the literature values presented in
Wiedemeier et. al. (1996). The objective function of the process was to match calculated
observed contaminant masses to model predicted contaminant masses. The model
successfully predicted contaminant distributions and reproduced contaminant masses
adequately. An advantage of the method is that it considers processes controlling the
whole plume, not just those occurring along a 1-D flow path, and therefore can be applied
to 3-D field problems. As with the other methods, some drawbacks exist. Creating a site-
specific model is labor intensive and requires much more data than the two previously

described methods. Also, it is difficult to match observed contaminant masses when there

74

is almost always limited knowledge of the location and amount of all the contaminant

inputs to the system.

Method to Estimate Rates at Plume G

A numerical modeling method was used to estimate field scale degradation
rates of plume G compounds. Ideally, the other two methods could be used to calculate
rates to constrain the model calibration and to better understand the relative rates of
degradation. However, no tracer data were available for plume G, thus the tracer
normalization method could not be used. Also, sampling point locations in plume G were
not conducive to the semi-graphical method. As a result, the transport model was solely

used to estimate field scale degradation rates of plume G compounds.

CONCEPTUAL MODEL
The conceptual model approach was used to develop the plume G transport
model. The literature review of the past site investigations provided a framework for
model stress periods. Geochemical data were evaluated in order to determine the
predominant degradation reactions in the aquifer and to identify possible redox zones. A
method to calibrate the model was then developed. The calibration procedure is a new
approach that was developed for plume G, but may prove useful at other sites where

source compounds have transformed.

Model Stress Periods

Multiple stress periods were needed to adequately simulate plume G. A timeline

75

of events at ARCO is presented below to summarize activities that potentially affected

numerical simulations (Everett, 1990).

1953 — Rubber and plastic auto parts manufacturing facility on Eliza Street in
Schoolcraft, Michigan opens and presumably, improper storage and disposal of
source chemicals begins at Areas 1, H, and ID.

1965 — Wastewater drainage was diverted from the Area H pond to the South
Seepage Pond.

1967 — Arco Industries Corporation assumes ownership and operation of the
manufacturing facility on Eliza Street in Schoolcraft.

1985 — The Michigan Department of Natural Resources identifies ARCO as a
source of contamination of the groundwater downgradient of the site property.

June 1987 to September 1988 — The Area I soil vapor extraction system tests
remove 110 kg of chlorinated solvents from the vadose zone.

1988 — Production of plastic/rubber parts and use of solvents is discontinued at
ARCO.

June 1988 — Approximately 115 m3 of contaminated sludge were excavated from
the northeastern section of the South Seepage pond.

May 1989 — Approximately 64 m3 of contaminated sludge were excavated from
the southeastern section of the South Seepage pond.

July 13, 1992 — PW-3 began extraction of groundwater from the aquifer to an off-
site air stripping tower. The well discharges at 1090 m3/day and is still in
operation.

This timeline reveals that site usage has varied throughout the facility’s existence,

therefore recharge rates and source concentrations were different at each area, which

required multiple stress periods for the transport model.

The four source areas were determined to have elevated levels of recharge based

on knowledge of past on-site activities. Areas 1 and IH were used as general clean-up

areas where objects were washed with a variety of solvents and water, thereby increasing

recharge. The duration of elevated recharge at these sites were assumed to exist from the

76

plant’s startup in 1953, until operations effectively ceased in 1988. The Area H pond and
the South Seepage pond held pools of standing wastewater and were subsequently areas
of elevated recharge that were higher than the recharge at Areas I and 1H. Area H was
only used from 1953-1965, and the elevated recharge was assumed to occur during that
period. The South Seepage pond was assumed to be an area of elevated recharge from its
opening in 1965 until the first excavation in June of 1988, when the pond sediments had
fully dried. The recharge at the source areas acted to deliver contaminants to the
groundwater as water infiltrated and replenished the aquifer.

Estimating the source term in a contaminant transport model is a difficult task;
therefore, several assumptions were made to deal with the source problem. First, the
degree of contamination was found to be variable from area to area, but the concentration
at a given site was assumed constant. Information provided in the description of the site
background indicated that the degree of contamination varied from highest to lowest in
the following order: Areal, Area HI, Area H, and the South Seepage pond. Second, Areas
1, H, and HI were assumed to be source areas from the time the facility opened in 1953
until the end of the model simulation. Justification for this assumption lies in the fact that
significant contamination was detected at all of the sites during investigations in the late
1980’s and early 1990’s and that no significant source removal operations were enacted
at any of the three sites in the past. Even though the SVE system removed 110 kg of
chlorinated solvents from the vadose zone in Areal, it was conservatively estimated that
at least 10,000 kg of PCB and 13,000 kg of TCE were present within the top 2 m of
sediments at Areal. The removal of 110 kg of chlorinated organics was essentially

minimal and was not expected to decrease the source term. The South Seepage Pond was

77

assumed to be a source from its opening in 1965 until the first excavation in June of
1988. The unexcavated portion of the pond was considered a source from 1965 until May
of 1989 when the final remnants of contaminated pond sludge were removed. The final
excavation was assumed to have completely removed the South Seepage pond as a source
since subsequent sampling of the soils below and adjacent to the pond indicated that little
or no residual contamination remained (Everett, 1990).

Analysis of the aforementioned site history revealed that since the ARCO plant
opened in 1953, there were four periods where recharge rates and/or source locations
varied and one period when the heads changed due to drawdown from a pumping well. A
summary of the five periods is presented in Table 6. Note that the table indicates whether
recharge was elevated (yes/no) and if the area was a source (yes/no) during that specific
period. Also, only the southern half of the South Seepage pond was a source in the third
stress period since the northern half of the pond was excavated in June of 1988. The
specific quantitative values for source recharge rates and concentrations were assigned
later, when the model was created. The stress period that starts on 5/5/1992 was the result
of the activation of PW-3, which influenced the heads near the downgradient end of the
plume. The resulting head distribution from PW-3 can be incorporated into a second
plume scale model in which the heads were derived from the regional model that
included PW-3 and the initial concentrations of plume G species were derived from the
first transport model that ended on July 13, 1992. It should be noted that not all five of
the stress periods were incorporated in the RT3D model due to the limited observed data
that were available past 1989 (this is explained later in the thesis in the Conceptual

Calibration section).

78

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 6. Summary of conceptual stress periods for plume G transport models.
Duration Area I Area II Area 111 SSP
Period (start/stop) recharg source recharge source recharg source recharge source
1 ““1953 yes yes yes yes yes yes no no
1/1/1965
2 In“ 965 yes yes no yes yes yes yes yes
6/27/1988
3 £25132: no yes no yes no yes no 1/2*
4 5/8/1989 no yes no yes no yes no no
5/5/1992
5/5/1992
5** no yes no yes no yes no no
present
recharge - yes = elevated recharge, no = plume scale model recharge (12 cm/year)
* Only the southern half of the pond was a source during this stress period
** PW-3 was in operation this time period

 

Conceptual Reactions of Interest

The literature review of the degradation processes revealed that the reactions are
complex and are controlled by the redox state of the aquifer; therefore, geochemical data
were examined to determine the probable degradation pathways of plume G compounds.
Eight data sets were available for interpretation of processes. Seven data sets were
collected by ARCO consultants from 1988-1996; however, five of the datasets only
contained results from 5-6 wells, and thus were not very helpfill in determining potential
process on the plume scale. Also, the ARCO data sets only contained results for the
chlorinated compounds present in plume G; no redox analyses were performed. The
eighth data set consisted of samples collected and analyzed by the Civil and
Environmental Engineering Laboratory at MSU from nine MSU monitoring wells in
plume G. Redox analyses were performed during the MSU sampling event and were used

to identify possible redox conditions in the aquifer.

79

Analyses of samples from the MSU sample grid were conducted by the Civil and
Environmental Engineering Laboratories at MSU in February of 2001. The results
showed that elevated levels of nitrate existed until a depth of about 13 meters below
ground surface (mbgs) where the levels subsequently decreased with increasing depth
until the bottom of the aquifer at a depth of about 26-27 mbgs (Figure 29). Chappelle
(1993) indicated that if there is a source of nitrate in a system, it will accumulate in an
aerobic environment and will be reduced and decrease in concentration in an anaerobic
system. The results indicate that the aquifer at the site of the MSU sample grid is aerobic
to a depth of approximately 13 mbgs and anaerobic to the base of the aquifer at
approximately 26-27 mbgs. The anaerobic zone can be further split into a zone of
denitrification from 13 mbgs to 19 mbgs where nitrate decreases with depth until it
becomes almost depleted, which indicates that the aquifer may be in a more reducing
state from 19 mbgs to the clay layer at about 27 mbgs. No other redox data specific to
plume G were available and the redox conditions throughout the plume were assumed to
be the same as those delineated at the MSU sample grid. Subsequently, both the aerobic
and anaerobic pathways of degradation depicted in Figure 28 were included in the model
of plume G reactions.

Examination of Figure 28 shows that there are two sources of 1,1-DCE in an
aquifer if TCA and TCE are present. Most TCE degrades to the cis— isomer of 1,2-DCE,
with only minor amounts transformed to the trans- isomer of 1,2-DCE, and even less to
1,1-DCE (Chappelle, 1993). McCarty (1997) indicated that if TCA were present in an
aquifer, the predominant transformation pathway would be the biotic route and that the

abiotic transformation to 1,1-DCE would be negligible. If abiotic transformation of TCA

8O

Nitrate (ppm)

 

0 10 20 30 4O 50 6O 70 80
0 I T I I I I ITI WTI T r I I I T T I I T I I I I f7 T—r T T I_T I I T I I I I
5 l ............................................ i, ......................................... 1
I; _l

Depth Below Ground Surface (m)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

_ . -
'6
” 1
15 ._.. ............. . ......................... ........................................... _.
_ {O -
l . -
r : i
20 ,._............... , ............................................. .2
>— . z E —l
l o -
o? - ‘ - - . 7
25 I Lgl l I 1 Ill 1 1 Lil 1 L411 1 I l IJ_LI I l I I 14 J 1_L11
.1 N
‘1 O
5 A
A ‘ ‘
A A ‘

 

 

 

 

 

 

 

 

D
D...
D
D

 

‘ NUS Monitoring Well
I MSU Monitoring Well

 

 

 

 

 

 

Figure 29. Nitrate concentrations at monitoring well MSU-5 from February, 2001.
Note that nitrate starts to decrease at about 13 mbgs until about 19 mbgs, where it
remains constant. Results suggest that the aquifer is aerobic from the water table to 13
mbgs, anaerobic denitrifying from 13 mbgs to 19 mbgs, and somewhat more anaerobic
from 19 mbgs to the clay at about 26-27 mbgs.

81

was negligible, then the TCA pathway to VC could essentially be ignored. However,
observed concentrations of 1,1-DCE throughout the entire plume ranged from 0-360 ppb.
Although some 1,1-DCE forms as the result of the degradation of TCE, McCarty (1997)
indicated that if 1,1-DCE were present in an aquifer, it was most likely the result of the
abiotic transformation of TCA. As a result, the TCA degradation reaction was a
potentially significant source to vinyl chloride production and was included in the model
of plume G reactions. All of the conceptual reactions depicted in Figure 28 were assumed
to occur at plume G, except that 1,1-DCE was assumed to only form from the abiotic

transformation of TCA.

Conceptual Method of “Calibration ”

Calibration of a model typically involves adjusting model parameters until a
minima in an objective function is achieved, such as sum of residuals between measured
and calculated concentrations. Transport models are often calibrated by predicting the
center of mass of each species in the plume, or by predicting observed concentrations at
monitoring points within the plume (Bedient et. al, 1999). However, contaminant plumes
are often poorly characterized by observed data; plume G was no exception. Limited data
were available for the plume G study; subsequently, traditional methods of calibration
were not employed. Observed data were analyzed at a downgradient well transect. The
model was calibrated by comparing the simulated results to the observed results at the
monitoring well transect.

Several key problems existed with the plume G data set that limited comparisons

of simulated and observed concentrations. First, samples were not analyzed for chloride.

82

If chloride concentrations were known, then the mass of each source could be better
estimated through methods presented by Wiedemeier et. al. (1996). Second, there were
only two data sets (May, 1988 and June, 1989) where sufficient data existed throughout
the plume area, but a consistent set of wells were not sampled during the two different
sample events making temporal comparisons of the simulation results difficult.
Subsequently, the model was considered a preliminary model and was compared to the
June 1988 data set, which meant that only stress periods 1 and 2 of Table 6 were
incorporated into the model.

A transect of wells perpendicular to plume G was located approximately 650
meters downgradient of ARCO where observation data for June of 1988 were available
(Figure 30). The transect was about 450 meters wide and consisted of 18 wells. The wells
were not true multilevel wells in a single borehole, but a series of 3 wells within a few
meters of each other in the horizontal plane where each well was screened at a different
elevation. TCE concentrations at the transect are plotted in Figure 31. Note that there is a
significant difference in concentrations, even at points that are somewhat close to each
other, such as MW-44A and MW-44B where concentration changes from 580 ppb to
3700 ppb with only 3 meters difference in the vertical distance between the points.
Possible explanations for the differences are the variable redox conditions of the aquifer
and/or preferential flow paths in the aquifer. The geologic heterogeneity of the aquifer
that was represented in the plume scale flow model did not include small gravel lenses or
other small preferential flow paths. As a result, the plume scale flow model was not
expected to predict the point observed values at the transect, but it was expected to

predict the relative mass of each species in three conceptual contamination zones defined

83

by the redox conditions at the transect.

The model was used to predict the vertical distribution of contamination in three
conceptual contamination zones at the monitoring well transect depicted in Figure 31.
First, the transect was divided into three zones based on the previously delineated redox
conditions. The shallow zone corresponded to layers 1-4 (0 to 13 mbgs) of the plume
scale flow model and was aerobic. The middle contamination zone correlated to layers 5-
7 (13 to 19 mbgs) of the plume scale flow model and was anaerobic (denitrifying). The
deep contamination zone correlated to layers 8-10 (19 to approximately 27 mbgs) of the
plume scale model grid and was somewhat more anaerobic than the middle zone based on
Figure 29. The observed mass of each contaminant (PCE, TCE, 1,2-DCE, 1,1-DCE,
TCA, DCA, and VC) in the aqueous phase was estimated using Theissen polygons in the
vertical plane and assuming an aquifer thickness equal to the diameter of the monitoring
wells (10 cm) and a porosity of 0.375 (Hoard, 2002). The calculated observed masses for
each species are summarized in Table 7 and were the basis for comparisons with the
simulated results.

The model results would need to be analyzed to determine the simulated mass of
each compound in the shallow, middle, and deep contamination zones at the monitoring
well transect for comparison with the estimated observed masses. The simulated mass in
each zone was not calculated with Theissen polygons because simulated data were
available for each grid cell at the transect, therefore data did not need to be interpolated
with the polygons. The aqueous mass of each species was calculated for each cell using
the cell area in the vertical plane multiplied by the width of the wells (10 cm), the

porosity, and the aqueous concentration that was simulated for the given cell. The

84

calculated masses for the cells in the three zones were summed to yield simulated masses

for the three contamination zones at the same location where observed masses were

estimated.

 

mbgs, 1.07x105 L) zones for June of 1988.

Table 7. Estimated total aqueous mass of observed compounds at well transect in
shallow (0-13 mbgs, 1.44x105 L), middle (13-19 mbgs, 1.36x105 L), and deep (19-27

 

 

 

 

 

 

 

 

 

 

Zone PCE (g) TCE(g) 1,2-DCE (g) TCA (g) DCA (g) 1,1-chg) veg;
Shallow 2.37 68.15 13.93 14.32 0.18 0.91 0.00
Middle 4.53 93.22 15.70 16.75 0.33 0.94 0.00
Deep 9.02 95.54 37.71 16.41 4.98 2.97 8.32
Totals 15.92 256.91 67.34 47.48 5.49 4.82 8.32

 

 

 

The calculated simulated and observed masses for the downgradient transect were

used to check the source zone concentrations and then to adjust the model degradation

rates to calibrate the model. Initially, simulated masses were compared with the estimated

observed masses in Table 7 to ensure reasonable source concentrations. However, since

there was so much uncertainty in the source term, the objective function of the calibration

was to match the ratio of the mass of each species to the mass of its daughter byproduct

in each zone (i.e. PCE/TCE for the shallow zone = 2.37/68.15 = 0.0348). The ratios are

independent of the absolute mass of each compound. Even if the simulated mass is not

equal to the observed mass, the model can represent plume G processes if the

mother/daughter ratios that are calculated from the simulated masses reasonably match

the calculated ratios from the observed data.

85

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

{L N
" A
_ A
A
L
/'
A ‘ A
A
A
A ‘ ‘
‘ A
A NUS Monitoring Well 0 500
I MSU Monitoring Well meters

 

 

Figure 30. Location of downgradient monitoring well transect used to identify vertical
zones of contamination. The numbers correlate to the NUS number assigned to the
wells during the remedial investigation studies in the late 1980’s (Mayotte, 1991).

86

 

TCE (ppb)

 

Jh

Middle — Anaerobic

Jll\

Depth Below Ground Surface (m)

[It-up .\ll;1l:lllhlt‘
(more reducing)

 

Jllli

300
Width (m)

Figure 31. Plot of TCE concentrations at a well transect perpendicular to groundwater

flow approximately 650 meters downgradient of the ARCO facility. The shaded

regions are the shallow, middle, and deep generalized zones of contamination. The
view is looking ugradient towards ARCO. Image is presented in color.

87

IMPLEMENTATION OF THE CONCEPTUAL MODEL
The transport model was implemented as an RT3D model through GMS using the
flow model created in Chapter H1. The first step was to develop a reactions package to
mathematically represent the conceptual degradation reactions presented in Figure 28.
The model parameters required by the equations were then assigned; among which were
the porosity, dispersivity, retardation factors, contaminant yield factors, and initial
degradation rates. The source boundaries were then assigned and the model was

calibrated according to the method outlined in the Conceptual Calibration section.

Creation of an RT 3D Reactions Package

The 3-D advection-dispersion equation with first order decay was solved using
RT3D (Clement, 1997). RT3D is a modular computer code for simulating reactive
transport in three-dimensional groundwater systems. The code uses a reaction operator-
split numerical strategy to separate the components of the transport equations so that the
coupled partial differential equations that describe reactive transport can be assembled as
a set of ordinary differential equations and solved separately from the advection-
dispersion portion of the equation. The equations are coded according to an RT3D
specified format into a reactions package that is compiled as a dynamic link library file
and used by RT3D to solve for the concentrations of the user-defined species (Clement,
1997)

The coupled set of ordinary differential equations equations representing the

conceptual degradation model presented in Figure 28 are as follows (note that these

88

equations represent only the degradation reactions; advection and dispersion are also

simulated in the model, but not with the equations below):

d[PCE] _ - KPCE[PCE]
dt RPCE

 

leCE] : YTCl-L‘PCE K PCE [PCE] ' KTCE1[TCE] ' KTCEz [TCE]
dt R TCE

d[1,2 - DCE]

: er-DClarcrz KTCE1[TCE] ‘ Kl.2-DCE1[1’2 ' DCE] ' Kl,2-DCE2 [1’2 ' DCE]
dt

Rl,2-DCE

leCA] : ' KTCA-A [TCA](A) ‘ KTCA-Bl [TCA](B) ’ KTCA-BZ [TCA](B)
dt RTCA

d[191' DCA] = (B)Yl.l-DCA/TCAKTCA-Bl[TCA] - I<l,l-DCA1[1’1 - DCA] - I(l.l-DCAZ[1’1 - DCA]
dt

I{Ll-DCA

d[191' DCE] : (0'21)(A)Yl.l-DCE’I'CAKTCA-A[TCA] - Kl.l-DCE1[1’1 - DCE] - I<l.l-DCE2[1’1 - DCE]
dt

R1,1-DCE

 

d[VC] = YVC‘!’1.2-DCEK1.2-DCE1[1’2 " DCE] + chm-DCEKLr-Dcal [1 ,1 ' DCE] " KVC1[VC] ' ch2 [VC]
dt va

d[CA] : YCA/l.l-DCAKl.l-DCAI [1’1 - DCA] - KCA [CA]
dt RCA

d[Ethene] __ YEthene/V C K VC 1 [VC] - K Ethene] [Ethene] - K EtheneZ [Ethene]
dt

R Ethene

diEthanOI] = (0° 79)(A)YElhanol'TC A KTCA - A [TCA] + YEthanor'rC A KCA [CA] " K Ethanol [Ethanol]
dt

R Ethanol

89

(1 Cl
[dt I = {YCl/‘PCEKPCE [PCE] + YCUTCEKTCEI[TCE] + YCl/TCEKTCEZ [TCE](3) +

~<

Clue-DCEKLz-ocslllaz ‘ DCE] + YClI’1.2-DCEK1.2-DCEZ[1’2 ‘ DCEKZ) +
Cl/TCA KTCA-Bl [TCA](B) + YCI/TCAKTCA-BZ [TCAI(B)(3) +
CWCAKTCA-A[TCA](A)(O°21) + YCVI‘CAKTCA-A[TCA](A)(O'79)(3) +
Cl/l.l-DCAK1.l-DCA1[1’1 " DCA] + YCl/1.1-DCAKl,l-DCA2[1’1 ' DCAKZ) +
Cl/l,l-DCEKl.l-DCE1[191' DCE] + YCl/1,1-DCEK1.1-DCE2[1’1 ' DCEKZ) +
YCl/‘VCKVCI [VC] + chchvcz [VC] + YCl/CAKCA [CA]} ch1

~<><-<

r<

where [PCE], [TCE], [1,2-DCE], [TCA], [1,1-DCA], [1,1-DCE], [VC], [CA], [Ethene],
[Ethanol], and [C1] are contaminant concentrations [ppb]; Kpcg, Kresr, Kl,2-DCEl, KTCA-31,
K1,1-DCA1, Kl,l-DCE1, Kym, KEthener, KEthanol are the first order anaerobic biodegradation rates
[day"]; Krcaz, Kl,2-DCE2, KTCA-BZ,K1,1-DCA2, K1,1-DCE2,KVC2,I(Ethene2 are the fifSt order
aerobic biodegradation rates [day-1]; KTCM, and KCA, are the first order abiotic
degradation rates [day"]; and RPCE, RTCE, RTCA, Rl,2-DCE, Rl,1-DCA, Rl,l-DCE, va. RCA.
Rgthene, REthanol are contaminant retardation factors. The yield factors (Y) are dependent on
the stoichiometry of the reaction and the molecular weights of the compounds. An
example calculation for the anaerobic degradation of PCE is that one mole of PCE
degrades to become one mole of TCE, or YTCEjPCE = molecular weight of TCE /
molecular weight of PCE (131.4 g/mol/ 165.8 g/mol = 0.79). The abiotic (A) and biotic
(B) factors account for the two major paths that TCA is degraded and were adjusted
during model calibration since no values were reported in the literature (note A+B=1).
The 0.79 and 0.21 factors account for the observation that in the abiotic pathway of TCA
degradation, 79% of TCA degrades to ethanol and 21% degrades to 1,1-DCE. Note again
that it was assumed that the sole source of 1,1-DCE in the system was the abiotic

degradation of TCA based on McCarty (1997). Although the equations include

90

chlorethane (CA), ethene, ethanol, and chloride, there were no observed data for these
compounds to compare to the RT3D simulated values. These components were included

in the reactions package for future modeling exercises of plume G.

Assignment of Model Parameters
The transport model required many input parameters, which were applied

to the model in GMS. The advection/dispersion parameters were assimilated from field
studies in the area and from the literature. The source concentrations were estimated from
results of the initial site investigations of the area (Everett, 1990). The source zone
infiltration rates were calculated from methods presented in the literature. Initial
degradation rates were applied to the model based on results presented in Suarez and
Rifai (1999).

The advection and dispersion properties of the transport model were assigned in
GMS. The porosity value that was used was 0.375, which was an average value
determined from sieve analyses at plume A (Hoard, 2002). The retardation factors were
estimated by a method presented in Bedient et. al (1999) (Table 8). It is typically useful
to have site-specific retardation factors determined by batch reactions with material
obtained from the site, but since the model of plume G was one of the first studies
completed, retardation factors were not available. The longitudinal dispersivity that was
applied to the model was 3 m based on an overall travel path of 1340 m using a
relationship presented in Xu and Eckstein (1995). The ratio of horizontal to longitudinal
dispersivity that was used was 0.1 (Aziz et. al., 2000) and the value of vertical to

longitudinal dispersivity was 0.05 (ASTM, 1995). The porosity, retardation factors, and

91

dispersivity values remained constant during model calibration (Table 8).

 

Table 8. Summary of transport model parameters which remained constant
throughout model calibration.

 

RPCE 2.50 YTCWE 0.792 YCWCE 0.270
RTCE 1-60 Y1 ,Z-DCE/TCE 0-738 Yer/1.2-003 0-366
Rm),-E 1.22 Y,,,.DCA,TCA 0.742 Y0,TCA 0.266
RTCA 1.66 10,..meTCA 0.727 Yew.DCE 0.366
R1 ,l-DCA 1-18 YEthanol/TCA 0450 Your ,l-DCA 0-358
Rum-E 1.31 vawcE 0.645 YCVCA 0.550
RVC 1.09 YVW.DCE 0.645 00* 3 m
RCA ### YCM,,-DCA 0.652 002011" 0.1
REthene #44 1{E,,,,,,,.,VC 0.449 over“ 0.05
REthanol ### YEthanol/CA 0-931 11"" 0-3 75
RC. 1.00 YCWCE 0.214

 

 

 

Rspccies - Retardation factors calculated fiom Bedient et. a1. (1999)
## - Observed data were not available; retardation factors not needed

 

 

Yspecimpecies - Yield factors calculated from molecular weights

* - Longitudinal dispersivity calculated from Xu and Eckstein (1995)
** - Transversezlongitudinal dispersivity from Aziz et. al. (1999)
*** - Vertical:longitudinal dispersivity from ASTM (1995)

**** - porosity from Hoard (2002)

 

 

 

 

 

 

 

The model was divided into stress periods 1 and 2 that were outlined in Table 6 in
order to assign the various recharge and source values to the model. A recharge rate of
0.065 m/day was estimated for the ponds (Area H and the South Seepage Pond) using a
method presented by Youngs et. al. (1996). The source recharge rate for Areas I and H]
was assumed to be equal to half of the value for the ponds (0.0325 m/day). An average
soil concentration of each of the source species was determined from results presented in
Everett (1990). Distribution coefficients determined from Bedient et. al. (1999) were

used to estimate the aqueous concentration, assuming equilibrium partitioning, at the

92

contaminated source soils. The aqueous concentrations were assigned as recharge
boundaries in the RT3D model (Table 9). The source recharge rates and source

concentrations remained constant during the model calibration process.

 

 

 

 

 

 

 

 

 

 

Table 9. Estimated source concentrations for the model source areas based on site
investigation results assimilated in Everett (1990).
Source Contaminant PCE TCE TCA
80"“ Am Kd (L/kg)* 0.48 0.13 0.15
Area I Soil Conc.(ug/kg) 70000 70000 14000
Aqueous Conc. (pg/L) 145833 538462 93333
Area H Soil Conc.(ug/kg) 150 210 0
Aqueous Conc. (jig/L) 313 1615 0
Area HI Soil Conc.(ug/kg) 20000 20000 15000
Aqueous Conc. (pg/L) 41667 153846 100000
South Seepage Soil Conc.(1lg/kg) 100 100 20
Pond Aqueous Conc. (lg/L) 208 769 133
* Estimated by method presented in Bedient et. al. (1999)

 

The first order degradation rates were the last parameters assigned to the transport
model. The model includes aerobic and anaerobic biodegradation rates, as well as abiotic
decay rates. However, it was determined previously that the top four layers of the model
existed under aerobic conditions, while the bottom six layers were anaerobic. Therefore
aerobic rates (TCE2, TCA-B2, 1,2-DCE2, 1,1-DCE2, DCA2, VC2) were zero in
anaerobic zones, and the anaerobic rates (TCEl, TCA-Bl, 1,2-DCE1, 1,1-DCE1, DCAl,
VCl) were zero in the aerobic zones. PCE does not degrade under aerobic conditions and
therefore the decay rate was zero for PCE in layers 1-4 (the shallow aerobic zone). The
abiotic TCA rate was determined by several authors to be solely a function of
temperature; therefore, the TCA abiotic rate was taken from McCarty (1997) based on a

temperature of 15°C and remained constant in all 10 layers. The initial values that were

93

assigned for all other rates were the mean value determined by a literature search that was
completed by Suarez and Rifai (1999). The authors compiled a list of first order

degradation rates for chlorinated ethenes/ethanes and results are presented in Table 10.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 10. Summary of first order decay rates for chlorinated
compounds as comfled by Suarez and Rifai (1999).
Species Low (day") Higflday") Mean (day-1)
PCE 0 0.41 0.051
TCEl 0 3.13 0.086
TCE2 0 1.65 0.346
1,2-DCE1 0 0.2 0.004
1,2-DCE2 0 1.96 0.476
TCA-A* 0.000375 0.000397 0.000383
TCA-Bl 0 2.33 0.355
TCA-B2 0 1.18 0.079
DCAl 0 0.044 0.036
DCA2 0 0.131 0.048
1,1-DCE1 0 0.27 0.019
1,1-DCE2 0 1.15 0.458
VCl 0 0.52 0.153
VC2 0.043 0.12 0.55
* values summarized by McCarty (1997) for a temperature of 15°C
1 = anaerobic rate , 2 = aerobic rate

 

MODEL CALIBRATION
The biodegradation rates were adjusted during model calibration, but were
constrained between literature values presented in Suarez and Rifai (1999) and adjusted
based on knowledge of relative rates of degradation according to the redox conditions.
Chappelle (1997) indicated that the aerobic degradation rates for the less chlorinated
species (DCE, DCA, VC) were higher than the anaerobic degradation rates, and this was

a constraint on rate adjustment. Also, the aerobic degradation rates should increase from

94

TCE to DCE to VC and from TCA to DCA (Chappelle, 1997), and this was also a
constraint on the parameter adjustment. Lastly, anaerobic rates increase as the system
becomes more reducing, therefore the anaerobic rates in the deep zone (layers 8-10) were
constrained to be higher than the anaerobic rates in the middle zone (layers 5-7). These

constraints helped to focus the calibration process.

Results

The simulated mass for each species in the three conceptual contamination zones
is listed in Table 11, while the mother/daughter ratios are in Table 12. Note that there was
no mass of vinyl chloride in the shallow and middle zone, but a calculated ratio was
needed to evaluate the model. As a result, the detection limit of 1 ppb was used to
calculate a mass for the shallow and middle zones so that ratios could be calculated
without dividing by zero. The model reasonably predicts the total mass of each species at
the monitoring well transect, except PCE concentrations are somewhat low. The general
trend in the vertical distribution of the masses is consistent with the observed trend as
well. However, there is too much 1,2-DCE, DCA, 1,1-DCE, and VC in the middle zone.
The mother/daughter ratios of the middle zone help to depict this problem (Table 12).
The residuals of the ratios in the middle zone are all too low, indicating that the mass in
the middle zone is too high.

The calibrated rates used in the model simulations are given in Table 13 along
with the model determined abiotic (A) and biotic (B) TCA factors. The rates are within
the range of observed values tabulated by Suarez and Riafi (1999). The model rates are

somewhat low compared to the observed ranges listed in Table 13. Explanation for lower

95

rates is that the tabulated values were taken from a variety of redox conditions and some

were taken from laboratory studies, both of which may be different from the Schoolcraft

site. However, the model rates are similar (10'3 to 10'5 day") to the field scale rates

determined by Clement et. al. (2000) for the Dover site. The abiotic (A) and biotic (B)

factors account for the relative amount of TCA that is transformed biotically or

abiotically; it is a mass balance term and therefore A+B=1. The model determined A and

B factors were 0.4 and 0.6 respectively. If an abiotic factor less than 0.4 (which also

corresponded to a biotic factor that was greater than 0.6) was used, the 1,1-DCE mass

was too low and the DCA mass was too high.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 11. Observed and simulated masses for the three conceptual contamination zones
outlined in the Conceptual Calibration section.
OBSERVED MASS

PCE TCE 1,2-DCE TCA DCA 1,1-DCE VC
shallow 2.37 68.15 13.93 14.32 0.18 0.91 0.00
middle 4.53 93.22 15.70 16.75 0.33 0.94 0.00
deep 9.02 95.54 37.71 16.41 4.98 2.97 8.32
total 15.92 256.91 67.34 47.48 5.49 4.82 8.32

SIMULATED MASS

PCE TCE 1,2-DCE TCA DCA 1,1-DCE VC
shallow 0.49 66.46 8.26 l 1.56 0.16 0.95 0.06
middle 1.43 93.95 26.15 17.26 2.30 2.16 6.61
deep 1.95 98.38 37.80 17.32 3.24 2.50 16.97
total 3.87 258.79 72.21 46.14 5.70 5.61 23.63

RESID UALS (SIM ULA TED - OBSER VED)

PCE T CE 1,2-DCE T CA DCA 1,1-DCE V C
shallow -1. 88 -1. 69 -5. 67 -2. 76 -0. 02 0.03 0. 06
middle -3.11 0. 73 10. 45 0.50 1.97 1.23 6. 61
deep -7. 06 2.84 0.09 0.92 -1. 75 -0.4 7 8. 65
total -12. 05 1.88 4.87 -1.34 0.20 0. 79 15.31

 

96

 

 

Table 12. Observed and simulated contaminant mother/daugher ratios for the three
conceptual contamination zones outlined in the Conceptual Calibration section.

 

OBSERVED MOTHER/DAUGHTER RATIOS

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

PCE/TCE TCE/Dela" DCE’WC TCA/DCA TCA/DCE* DCE*/VC
shallow 0.035 4.893 103.165 80.070 15.666 6.771
middle 0.049 5.938 112.141 51.193 17.916 6.680
deep 0.094 2.534 4.532 3.291 5.526 0.357

SIMULATED MOTHER/DAUGHTER RATIOS

PCE/TCE TCE/DCE“ 13013th TCA/DCA TCA/DCE“ DCE*/VC
shallow 0.007 8.049 140.987 72.539 12.214 16.164
middle 0.015 3.592 3.959 7.502 7.975 0.328
deep 0.020 2.603 2.228 5.353 6.925 0.147

RESID UALS (SIMULA TED - OBSER VED)

PCE/TCE TCE/DCE “ DCE#/VC TCA/DCA TCA/DCE“ DCE*/VC
shallow -0027 3.156 37.822 -7531 -3452 9.393
middle -0. 033 -2345 -108.182 -43.691 -9941 -6.352
deep -0075 0.069 -2304 2.062 1.399 -0209

 

 

 

 

 

 

 

# - TCE/DCE and DCE/V C were calculated using the 1,2-DCE isomer
* - TCA/DCE and DCE/V C were calculated using the 1,1-DCE isomer

 

 

 

Table 13. Summary of calibrated parameters used in the RT3D model.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

First Oder Description Shallow Aerobic Middle Anerobic Deep Anaerobic
Rate Zone (day-l) Zone (day-1) Zone (day-1)
PCE anerobic l .0E-05 1 .0E-04
TCEl anaerobic 1.35E-04 2.50E-04
TCE2 aerobic 1 .1 5E-04
1,2-DCE1 anaerobic 1.00E-04 4.00E-04
1,2-DCE2 aerobic 1 .00E-03
TCA-A abiotic 3.83E-04 3.83E—04 3.83 E-04
TCA-B 1 anaerobic 2.50E-04 6.00E-04
TCA-B2 aerobic 1 .00E-04
DCAl anaerobic 9.49E-04 9.50E-04
DCA2 aerobic 6.10E-03
11DCE1 anaerobic 1.00E-04 4.00E-04
1 1DCE2 aerobic 9.50E-04
VCl anaerobic 9.99E-05 1 .00E-04
VC2 aerobic 4.30E-02

A=O.4 B= 0.6

 

97

 

 

Model Sensitivity to Degradation Rates

The RT3D model is a numerical approximation to a complex problem and the
solution may not be unique. The range of observed degradation rates for the species
present in plume G is large (Table 10). Simply calibrating the model by matching the
mother/daughter ratios could lead to an erroneous solution if other factors such as plume
extent were not examined. Preliminary sensitivity simulations helped to identify a range
of probable degradation rates in which the observed plume extent was appropriately
predicted. Sensitivity simulations were then run to determine if the modeled rates were
the optimal rates for the model.

The final rates used in the model were on the order of 10'2 to 10'5 day'1 with most
rates on the order of 10‘4 day'l. The range was determined from multiple preliminary
simulations, which served to Show model sensitivities to the degradation rate. The
preliminary run used the mean degradation rates listed in Table 10, but these rates were
much too high (the source species were degraded too fast and thus the plume had not
moved far enough downgradient to observed positions). The subsequent runs were
completed by decreasing all of the biodegradation rates by a factor of 10, 100, 1000, etc.
The results showed that the degradation rates of PCB, TCE, and TCA were the most
important parameters because they controlled the production of the daughter byproducts.
Using rates for the source species that were less than the final rates (10'6 day'l instead of
10" day‘l) resulted in too much mass in the system, which subsequently moved too far
downgradient of observed positions. The opposite was true if the rates for the sources
species were higher than the final rates (10'2 day1 instead of 10‘4 day'l). The preliminary

simulations provided useful information by narrowing the range of possible parameters.

98

Multiple simulations were run to evaluate the model sensitivity to changes in
degradation rates. The final model rates listed in Table 13 were assumed to be the base
case rates. The process involved changing the degradation rate of one species in one of
the conceptual contamination zones (shallow, middle, and deep), while holding all other
rates at the base case. As stated previously, the plume extent was sensitive to the PCE,
TCE, and TCA rates, therefore, these rates were changed most from the base case values.
The objective function was the root mean squared error (RMSE) of the mother/daughter
ratios. Note that there were essentially two different sources in the system (PCE/TCE and
TCA), which led to somewhat different degradation byproducts. Therefore, the objective
function was calculated differently for the PCE/T CE pathway versus the TCA pathway.
The RMSE for the PCE/TCE path was calculated based on the PCE/TCE, TCE/1,2-DCE,
and the 1,2-DCE/V C ratios, while the RMSE for the TCA path was calculated based on
the TCA/DCA, TCA/1,1-DCE, and the 1,1-DCEN C ratios.

The results of the sensitivity simulations are presented in Figures 32-39. Figures
32-35 depict model changes in parameters that affected the PCE/TCE pathway, while
Figures 36-39 depict model changes in parameters that affected the TCA pathway. The
plots show that the modeled rates are at a minimum in the chosen objective function.
Some plots (Figures 35 and 39) Show dashed lines that represent boundaries beyond
which rates were not adjusted. Figures 35A and 39A show a boundary at 0.043 day",
which is the literature minimum given in Table 10. Figures 35B and 39B are bounded by
the constraint that the anaerobic rates in the middle zone cannot be higher than the deep
zone (the deep zone is more reducing which favors stronger transformation rates). The

plots also show that in general, the systems are most sensitive to decreases in the rates.

99

40

39

38

RMSE

37

36

35

40

39

38

RMSE

37

36

35

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

PCE Degradation Rate (day‘)

T ' T T t Y * ' l * * _
L l I
r. ......................................................................... .. .......................... .................. j
* o O o a
: Q . 1
Z Z
: , :
i— i l i L i l 1 l l r r i l L a r l l l .4
0 0.00002 0.00004 0.00006 0.00008 0.0001 0.00012
PCE Degradation Rate (day‘)
* f T * r r r r f r r r _.
l; ............... ' ............... g ................................. ..................................... B ,L
>- O . . -<
C 5 = 9 Z
l : . .
l. ...................................................................................................................... _1
C 1
e G -
._ ..................................................................................................................... —i
P . i . l l . l i . . 1 l l l i . . i l l l A
0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014

Figure 32. Sensitivity plots of model error (RMSE) of the mother/daughter ratios
(PCE/TCE path) in response to varying degradation rates (model rate is circled). A)
The anaerobic PCE rate in the middle zone. B) The anaerobic PCE rate in the deep
zone (note, PCE is not currently known to degrade under aerobic conditions, therefore
there is no aerobic PCE rate in the shallow zone).

100

37.5

35

50

47.5

37.5

35

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

>- T T I T T Y T T T T T- -<
I : é a
a. ..... A .3
I o 3
E o 3
I o I
i: ....................................................................................... :
:0 _ . . . . 2
: . j I 2 I j
...... 9 _
E 9 g r 2 a s 4
: . = = s :
l l . 1 l . . 1' l l A I l . . i l l A i . . l 7

0 0.0002 0.0004 0.0006 0.0008 0.001 0 0012

TCE-2 Degradation Rate (day‘)

AT 1 r r r VVVVV r n -i
:. ................................................................................ 3
I; ....... 9. ......... .................. .3
l- ; § 3
L. ................ 3 .......... .0 .......... _ .............................................................. g ................ 9.;
l— : . j .4
Z S o 2
E3 ........................................................................ L
'— 1 I a i L i . l i r l lllll 1 ..... 1 1 ML 1 i l . i r i i d

0 0.00006 0.00012 0.00018 0.00024 0.0003 0.00036

TCE-l Degradation Rate (day'l)

__ f 4 a r r * ' r r * r * * ‘ 4
i- i -l
3? ............................................................................................................. C ..1
i 1
I 2
: o . :
: 3
E 0 . 3
~ 1 ° 0 : ‘
LQ. ............................. § ................. j
~ . i . . i . . j . . i . i . i . . .

0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012

TCE-l Degradation Rate (day‘)

101

Figure 33. Sensitivity plots of model error (RMSE) of the mother/daughter ratios

(PCE/TCE path) in response to varying degradation rates (model rate is circled). A)
The aerobic TCE rate in the shallow zone. B) The anaerobic TCE rate in the middle
zone. C) The anaerobic TCE rate in the deep zone.

95

85

75

65

RMSE

55

45

35

42

41

40

39

RMSE

38

37

36

95

85

75

65

RMSE

55

45

35

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

’- 1 * fl * f f * fl * * I 7 fl _.
2 f A :
E 3 5
3......W ........................................................................................................ _:
C i
»— . -1
:9 ............. j .............. 9 .................................................................... ; .................. _:
: + 3 I
: Q 1 1 i 1 944 1 1 1 i 1 1 . 1 1 1 . 1 1 1 Z
0 0.004 0.008 0.012 0.016 0.02 0.024
1,2-DCE-2 Degradation Rate (day")
L # 1 f IV 1 r r l * * _
1 f 3 o 1
r .......................... I. ................................ g
3 .......................... I ............................. , ................................................ B -_:.
E I ‘ ° 3
E 1 3
E ........................................................................................................................ _:
. . . 1 m1 . 1 . 1 . 1 . . . 1 1 . . i 1 . . i
0 0.00007 0.00014 0.00021 0.00028 0.00035 0.00042
1,2-DCE-1 Degradation Rate (day“)
__ r T * 1 r r fi fl * 1 v r 1 _4
L— ................. ; ................................................................................................. .1
:_ ....... ‘ ......... ...................................................................................................... _:
g ..................................................................................................... z:
E 1 3
: : . Q 0 . . O . :1
b 1 1 1 1 1 1 1 Q 1 1 1 1 1 1 A l 1 1 1 1 1 1 1 d
0 0.0002 0.0004 0.0006 0.0008 0.001 0. 012

1,2-DCE-l Degradation Rate (day‘l)

Figure 34. Sensitivity plots of model error (RMSE) of the mother/daughter ratios
(PCE/TCE path) in response to varying degradation rates (model rate is circled). A)
The aerobic 1,2-DCE rate in the shallow zone. B) The anaerobic 1,2-DCE rate in the
middle zone (dashed line is where 1,2-DCE rate cannot be less than the VC rate in this
zone). C) The anaerobic 1,2-DCE rate in the deep zone.

102

 

 

 

 

 

 

 

 

115 . . ~ 1 . - . 1 1 1 . - .
>- ' :r :4
F : i 9 d
105 L11‘..‘_' ................................................................. .. ___:
+— - ' -<
+- ; : . . 1 A _y
l‘ ' ‘ : : t j
95 _l ................ . .................. _
I g g t , , 3
85 _'_ _l. ....................................... _1
m : ? 3 . : , :
CD " 3 :I r : : -1
7S ;......._--......_..: .................... :..._............-...in..-“.......-..._...:....1..-.-.....-._...:.. ................ _1
— f :| : : : 1
>— ‘ . ' , L :‘4
65 ;.,...-_....,11--.E....._........_....1......-......1-.....4:....-1....-...1..-1.1..1..........--.1..1.; .................. _.1
: E :l f O f : 3
55 L .................. ............... - .................... .................. .5
_ 1 I : : ; .1
e 1 - : : : n
45 ; ................. i ..................... ................... ; ..................... : ..................... .................. _-:
: 3 :é E : : 3
35 * 1 1 1 i 1 1 R1 1 1 1 i 1 1 1 1 1 1 1 1 1 1 1
0 0.02 0.04 0.06 0.08 0.1 0.12

VC-2 Degradation Rate (day‘l)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

45 . . Y 1 . v . I f . Y .I v r v 1 . v v i
I i a 1 2
43 ._ ..... 1 .............. -.1 ..................... L ..... B 1-_(
LL} 4] ;.. ................ 1 .................... ; .................... .I .................. _:.
U) P i f : I I ‘
E 39 ; ....... , ......... .................... i. ......... . ......... ................. r, ................... I .................. S
“ E E i E ' 1
t a s 5 s 1
' : z : d
37 1...; ............ . ........ ............... . ..... ..................... ; ..................................... _1
35 l‘ 1 n 1 i 1 1 1 1 1 1 1 1 1 1 1 i 1 1 1 '1 1 1 1 A
0 0.00002 0.00004 0.00006 0.00008 0.0001 0.00012
VC-l Degradation Rate (day")
45 ‘ i * a I *fi 1 1 r T 1 r r
_ I :
43 ._.' ........... . ......... ;-...-.._.-.-....-_11.: ....... ...............’ .......... . .......... L ..... ——4
: l E 1 2 C :
11.1 1. ; ................ I ......................................... . __________________ j
m '- | - -1
5 ~ : = :
39 P ................. I ......... . ......... 3 ............................................................ z
37 L ....................................................................................................................... L
C l , _ , . Z
35 1 1 i 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 J 1 1 1
0 0.0001 0.0002 0.0003 0.0004 0.0005 0.0006

VC-l Degradation Rate (day‘)

Figure 35. Sensitivity plots of model error (RMSE) of the mother/daughter ratios
(PCE/TCE path) in response to varying degradation rates (model rate is circled). A)
The aerobic VC rate in the shallow zone (dashed line is literature value boundary). B)
The anaerobic VC rate in the middle zone (dashed line indicates VC cannot be higher
than deep zone). C) The anaerobic VC rate in the deep zone (dashed line represents
boundary where the VC rate in the deep zone cannot be lower than the middle zone).

103

 

25 -.3.T.-.-,..Tfi....,..-.,..-.,fi---

 

23

 

 

 

2T

RMSE
O
O
O

 

 

1 1 ‘
1 v ' ‘
1 . 1 1 1 .
>- . . . 1
A A A A A A

11.1 111 11 1

0.0002 0.0003 0.0004 0.0005 0.0006 0.0007
TCA-B2 Degradation Rate (day")

 

 

 

 

 

 

25.13-fi
}- -<
" . ‘ i 1 i - -‘
. . 1 . . .
»— 1 . , . 1 1 -<
: I 1 - ' ,
23 P ....................... , .......................................... ._ .................................... _1
>- ; ' .' 1 l l 4
r- . - 1 1 .1
; . : . O 1
"' . - ~ 1 1 - —<
‘ I i . I I
21 __.1.11.......1.11.........1......1...........-.-...1 ...... . ....... ...3”......_......-..:.1..............;. .............. _.
>— . 1 - = A .4

RMSE
O

 

 

*1

 

l— . . . . . .
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 l 1 1 1 1 l 1 1 1 1 1
0 0.0001 0.0002 0.0003 0.0004 0.0005 0.0006 0.0007

TCA-Bl Degradation Rate (day‘)

 

25 T T T Ta T T ' ' T ' ' ' T ' T ' Ta 7 T ' TV

.4

~ : u - 1 I "‘
1 ~ 1 1 1 u

r— . . 1 - 1 .1
. 1 1 u ' U
, 1 1 1 ‘ ‘

,_._.. .......... 1 .................................................. ., .......................................... _.
. . 1 . 1 4
‘ 1 1 1 I

’- ' . 1 . 1 I —4
1 . , 1 1 1

‘ ‘ 1 . 1 _‘

-1

 

 

 

 

. . . 1 .
>- ' ‘ 1 1 . .

21 .__ .................... . ......................... . ....... ;..... ....-.1...:..............-..; ............... ._1
. < ‘ 1 1 ‘
. - . 1 . 1

RMSE

 

 

g4 11

'5 1i1111.11111111
0.0002 0.0004 0.0006 0.0008 0.00]

 

1 i . 1 1 i . 1 1
0.0012 0.0014 0.0016

TCA-Bl Degradation Rate (day“)

Figure 36. Sensitivity plots of model error (RMSE) of the mother/daughter ratios

(TCA path) in response to varying degradation rates (model rate is circled). A) The
aerobic TCA rate in the shallow zone. B) The anaerobic TCA rate in the middle zone.
C) The anaerobic TCA rate in the deep zone.

104

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

30 1— ‘ 1 r T fir fir * fir * # r T V
E 5 2
275 E ...... .9 ......... ?' ......................................................................... A La
1 S 1
: a j
25 ,— .................. . ............................................................................................ _-:
m C 3 a
CD 2 ' I
E 225 C“ .................................................................................................................. j
i o :
20 L. ........................................ _‘1
" : 1
175 :_W ............................ T .......... ........................................ . ................. L
i s 6) ° 5 3
15 P 1 g . i m L - 1 1 1 1 i m 1 . 1 1 1 1 D
0 0.002 0.004 0.006 0.008 0.01 0.012
DCA-2 Degradation Rate (day")
30 ’- fi 1 r * T T T T 1 1 T ' Y Y ..
t j l 1
27.5 :_ ............................................................................................... I .......... 1 B k:
— 1
25 :_ ............................................................................................. l.... ................. .3
a : . I : r
225 L ....... 9. ..................................................................................... 1....- .................. _Z
5 C ° :
20 :_ ............................................................ ................................... I ................ _:
: 0 | 1 :
175 L .......................................... ........................................................... .2
15 ’ 1 1 l . i . 1 . i 4 1# i . . 1 1 1 1 "
0 0 .0002 0.0004 0.0006 0.0008 0.001 0.00 l 2
DCA-l Degradation Rate (day")
30 . - - - 1 . . . - 1 - - . . . . . . . I . . . - I . fi- 2
Z s a i g E 1
27.5 E— ------------------ f. .................... g .................... ...... C J
25 E ................. ,. .. ................. .1.
m _ 1
m p 2
E 225 To" ......................................................................................................................... _.
1.9 1
20 7...... ..................................................................................................................... L
17.5 E_ .......... O. .......... 0. ........................................................................ f. ................. _:
._ J . O ‘
: Q : 3
l5 ” . z n i 1 1 1 1 1 1 1 1 1 . 1 . 1 1 . . 1 1 1 . A . "
0 0.001 0.002 0.003 0.004 0.005 0.006

DCA-1 Degradation Rate (day")

Figure 37. Sensitivity plots of model error (RMSE) of the mother/daughter ratios
(TCA path) in response to varying degradation rates (model rate is circled). A) The
aerobic DCA rate in the shallow zone. B) The anaerobic DCA rate in the middle zone
(dashed line indicates the rate in the middle zone cannot be higher than in the deep
zone). C) The anaerobic DCA rate in the deep zone.

105

 

40

35

3O

RMSE

25

20

16.5

RMSE

RMSE

Figure 38 Sensitivity plots of model error (RMSE) of the mother/daughter ratios (TCA
path) in response to varying degradation rates (model rate is circled). A) The aerobic
1,1-DCE rate in the shallow zone. B) The anaerobic 1,1-DCE rate in the middle zone
(dashed line indicates the 1,1-DCE rate cannot be lower than the VC rate in this zone).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

"""" I ' ‘ ' ' ' T *—' Y I z T f .1
L .................................... g ........................................................... _. A .0;
:0 i i
:. ................................... - ........................................................... : ................. _J
: 0 - V , _ ’ ___________________ é ................. :

: 0 é :

M 1 1 1 G '1 1 1 1 1 1 lllll 1 1 1 A 1 1 1 i ..... d
0 0 0012 0.0024 0.0030 0.0048 0.006 0.0072

11DCE-2 Degradation Rate (day")

' ' ' ' T ' fl ' r r V? I * fl 1 V _4

.

........................ I B .....1

., ... . I ...................................................................... xi

I , o 2

i _____________ _ _______________ _ ............................................................................ .

I 1

1

E . . 1 1' 1 I. 1 1 . 1 1 1 1 1 . 1 1 1 . i . 1 #2
0 0.00007 0.000 14 0.00021 0.00028 0.00035 0.00042

1 1DCE-l Degradation Rate (day‘l)

_ r fi m z r r T f r f * r fi' _‘

: . $ ‘ :

l- : E -4

:- .................. 1....................: ............................................................. . .................. __:

: ° 6 ° ° :

:_ .................. ' ....................... ' .................................................................................. _.‘1

l- : '1

C ; j

g. ............ ......................................................................................... i

I E ' g

’— 1 1 1 i + 1 1 1 1 1 1 1 1 1 1 1 1 1 1 l 1 1 1 "l
0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012

llDCE-l Degradation Rate (day")

C) The anaerobic 1,1-DCE rate the deep zone.

106

 

zzvafiYVfivvvvuvv 1111111

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

21 E. .............................. I..-........... ...... A .1:
1 5 5 :I 5 5 o 3
20 L ............... .................. 1 .......... , .......... j. .................. ................ f ................... j
I z 3 ;| s s :
Lu ’ 3
Cl) 19 Ll .......................................................... .:
E E g 3 :
18 :. .................. I ..................................................... .:
E il 5
I7 :. ................... 1 ...................................... ................................ ... ........................ .5
16 1:. .................. .. é ....... ..., ................. .3
15 : 1 i 1 1 1 1 1 1 1 il 1 1 1 1 1 1 - l 1 - 1 :
0 0.014 0.028 0.042 0.050 0.07 0084
V02 Degradation Rate (day")
22 . . . r r . . I . . . . . . . . . . . - _
2] P! ................................................................................................... I ...... B .:
E l 3
20 T ..................................................................................................... _ .................. 1
: o I 2
531 19 :. ...................................................................................................... I. ................. .:
E 18 a .............................................. 1 ................................................. 1 ................. .3
1 1 1 1
17 E. ......................................................................................................................... .
16 E. ........................................................................................................................ T:
15 : 1 1 1. 1 1 1 1 1 1 1 1 1 1 l 1 1 - ’ 1 1 1 Z
0 0.00002 0.00004 0.00006 0.00008 0.0001 0.00012
VC-l Degradation Rate (day")
27- ' ' ‘ ‘. ' " ‘ F ' ' ‘ I ' ' * 1 ' ' ' 1 ' ' - -
3 E E F E 1
2| T ........ I ........... ...... C .:
I E I 3 s 2 5 :
20 “.----- ..... .................... g. .................... g. .................. _:
E E I i g g g :
m 19 - ................. .3
m t I j
E 18 T ........................... I ............................................................................................ .:
17 f. ............................ I. ............................................................................................. .g
E - o o o 9 3
16 E. ..................................................................................................................... .j
15 t 1 1 1 I1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 :
0 0.00007 0.00014 0.00021 0.00028 0.00035 0.00042

G
VC-l Degradation Rate (day'l)

Figure 39. Sensitivity plots of model error (RMSE) of the mother/daughter ratios
(TCA path) in response to varying degradation rates (model rate is circled). A) The
aerobic VC rate in the shallow zone (dashed line is literature value boundary). B) The
anaerobic VC rate in the middle zone (dashed line indicates VC cannot be higher than
deep zone). C) The anaerobic VC rate in the deep zone (dashed line represents
boundary where the VC rate in the deep zone cannot be lower than the middle zone).

107

Discussion

The simulated mass of PCB was lower than the estimated mass in each zone. The
reason is that PCE is retarded more than the other compounds, and at the time the
observed measurements were taken, the PCB plume was just arriving at the monitoring
well transect. The retardation could be lowered to allow more mass to reach the transect,
but the modeled PCE retardation factor was at the low end of the range of observed
values in sandy aquifers. Also, increasing hydraulic conductivities caused the other
species to travel too far downgradient. The higher observed mass may be due to small
areas of preferential flow along which PCE traveled and reached the transect before the
majority of the PCB plume arrived.

Examination of the observed masses shows several trends that can be explained
by the suspected redox conditions. The 1,2-DCE, DCA, and VC values are greatest in the
deep zone. This can be attributed to two factors: recharge has forced the majority of the
mass to the middle and deep zones and also that the degradation rates of TCE and TCA
are greater in the deep anaerobic zone, which results in the production of more mass of
the daughter products in the deep zone. Vinyl chloride is completely absent from the
shallow zone, which may be due to the high aerobic degradation rate of VC in the
shallow zone, which removes the VC from the system. VC is also absent from the middle
zone, which may be the result of no VC production in this region because DCE does not
readily degrade under denitrifying conditions (Chappelle, 1997). However, the middle
zone is the problem area of the model, and using a very low degradation rate (105/day or
lower) resulted in significantly much more mass in the middle zone. The problem may be

that the conductivities in these zones are too high, therefore too much mass is moving

108

through these zones.

The nature of numerical model results is that the solutions are non unique.
However, estimating the degradation rates of plume G with the RT3D model was useful.
The preliminary sensitivity simulations narrowed the range of reasonable rates to around
10'3 to 10‘4 day'l. Although a combination of other rates in this range may produce the
same results, at the very minimum it has been shown that the natural degradation rates are
rather low. This indicates that while natural attenuation is occurring at the site, it seems
unlikely that it would be a viable treatment strategy for the entire plume since the

degradation rates are so low.

SIGNIFICANCE AND APPLICATIONS

Novel Aspects of the Model

Several aspects of the RT3D model were significant. The plume G reactions
package is similar to the pre-defined sequential aerobic/anaerobic PCE/TCE
dechlorination model predefined in RT3D (Clement, 1997) that was used at the Dover
site by Clement et al. (2000). However, the model differs from Clement et. al. (2000)
because it addresses TCA degradation in the same system as PCE/TCE degradation. The
model also simultaneously describes abiotic and biotic transformation of TCA, which
does not appear to have been examined in the modeling literature. Perhaps the most
important aspect of the work was the calibration approach. Using the ratio of parent
compound/daughter degradation byproduct as the comparison to calibrate the model was
a new approach to reactive transport model calibration. The ratio approach to reactive

transport calibration could be applied to any situation where source contaminants have

109

degraded to daughter byproducts. The approach focuses research on understanding other
aspects of the system, such as degradation process, instead of concentrating on the source

term problem.

Applications of the Research to the Site

The RT3D model that was created for plume G has several applications for
research at the site. The model can be used to estimate remediation times for the
compounds if monitored natural attenuation was the only strategy employed. Also, if an
in situ system such as the plume A biocurtain is developed for plume G, then the natural
attenuation rates could be coupled with a model that describes biostimulation or
bioaugmentation to estimate a remediation time for the in-situ system. Both of the
aforementioned applications could become part of a cost benefit analysis to compare the
different remediation techniques available at the site, such as pump and treat, monitored

natural attenuation, and in situ bioremediation.

RECOMMENDATIONS FOR FUTURE WORK
The results for the RT3D model were based on several assumptions and further
research would help to determine if the assumptions were valid. Reducing the uncertainty
associated with the advection and dispersion parameters of the model is an important task
that could be easily addressed. Collection, analysis, and interpretation of a plume scale
data set that includes redox analyses would be more costly and require more time, but
may serve to reduce significant uncertainty associated with redox interpretations and

subsequently the degradation rates. Lab analyses of degradation processes at the site

110

would serve to further support the rate estimations. Lastly, using an automated
optimization scheme would quantify the model’s sensitivity to perturbations in the
degradations rates and allow assessment of the uniqueness of an optimal solution.

The hydraulic conductivity of the aquifer was represented by five different
materials in the flow model, yet the retardation factors and porosity used in the reactive
transport model did not reflect this heterogeneity. The porosity and sorption
characteristics of the different materials were expected to be different. The porosity of the
different materials could be determined from sieve analyses of the MSU 1-9 cores, while
the sorption properties could be determined through batch reactions. The results are
significant because the porosity affects the average linear velocity calculated by the
advection equation and the sorption characteristics control the retardation factors of each
species. The effect of one or both factors may significantly reduce or increase the travel
time for solutes in the system, which may subsequently change the degradation rates
estimated during the calibration.

A plume scale geochemical dataset that includes redox data in addition to analysis
for all compounds included in the RT3D reaction package equations is needed to further
constrain the model. Redox analyses could be used to check the interpretations made in
this thesis and to further delineate any other redox zones that may be present, such as a
more reducing environment near the source as is typically observed at contaminant
plumes (Azadpour-Keeley, et. al. 2001). Tracer normalization calculations to determine
site-specific degradation rates could be performed using chloride data, which could be
used to check the results of this work. The dataset could be used to estimate the total

mass of each contaminant in the system, which provides another constraint on the

111

contaminant source term of the model. Lastly, collection of the plume scale geochemical
dataset at multiple time periods (with about a year or so between each sampling event to
see significant change in contaminant concentrations) would allow temporal comparisons
of the model to observed data, which would serve to further reduce the uncertainty
associated with the model results.

Laboratory analyses of aquifer materials from the site would help support (or
negate) a major assumption of the model. It was assumed that 1,1-DCE was formed
solely as the result of abiotic degradation of TCA based on McCarty (1997). Lab analyses
could determine if 1,1-DCE was being produced by abiotic TCA transformation, or if
TCE degradation produced the significant amount of 1,1-DCE observed at the site. A side
product of the tests would be degradation rates that could be compared to the field
estimated rates. The analyses are significant because VC is a major concern at the site,
and the bioremediation system that is designed would need to incorporate the TCA
source of 1,1—DCE and subsequently VC if lab tests confirmed McCarty (1997).

Automated calibration of the model using an optimization scheme would be
beneficial to the study. An optimization code would greatly reduce the time needed to
calibrate the model. The trial and error method was used, which was very time
consuming. The optimization run would also quantify that the estimated degradation rates
resulted in a minima in the objective function and could provide sensitivity parameters

for the model inputs.

112

APPENDICES

113

APPENDIX A. WELL LOG DESCRIPTIONS FOR MSU WELLS 1-9

 

Well Log for MSU-1
Ground Elevation: 264.9 m

Well drilled with a Hollow Stem Auger and aquifer sampled
with Waterloo Cohesionless Sand Sampler

 

 

 

Top Depth (milottom General Description
0 9.1 NO SAMPLE

9.1 12.2 very fine to fine sand

12.2 13.4 medium to coarse sand

13.4 13.7 NO SAMPLE

13.7 14.6 fine to medium sand

14.6 16.2 medium to coarse sand

16.2 16.5 NO SAMPLE

16.5 17.1 medium to coarse sand with some gravel

17.1 18.0 coarse sand with some gravel

18.0 18.3 fine sand

18.3 19.8 NO SAMPLE

19.8 25.0 coarse sand and gravel

25.0 25.3 clay

 

 

 

Well Log for MSU-2
Ground Elevation: 265.1 m

Well drilled with a Hollow Stem Auger and aquifer sampled
with Waterloo Cohesionless Sand Sampler

 

 

 

Top Depth (ml)30 t tom General Description
0 9.1 NO SAMPLE
9.1 12.2 very fine to fine sand
12.2 12.8 fine to medium sand
12.8 13.7 medium to coarse sand with some gravel
13.7 16.2 fine to medium sand
16.2 16.8 NO SAMPLE
16.8 17.7 medium sand
17.7 18.0 NO SAMPLE
18.0 24.7 coarse sand and gravel
24.7 25.0 clay

 

 

 

 

Well Log for MSU-3
Ground Elevation: 265 m

Well drilled with a Hollow Stem Auger and aquifer sampled
with Waterloo Cohesionless Sand Sampler

 

 

 

 

 

Top Depth (milottom General Description
0 9.1 NO SAMPLE
9.1 12.5 very fine to fine sand
12.5 13.1 medium to coarse sand with some gravel
13.1 13.4 NO SAMPLE
13.4 14.6 fine sand
14.6 18.3 medium to coarse sand with some gravel
18.3 25.0 coarse sand and gravel
25.0 25.3 clay

 

 

114

 

 

 

 

 

 

 

 

 

Well Log for MSU-4 Well drilled with a Hollow Stem Auger and aquifer sampled
Ground Elevation: 264.9 m with Waterloo Cohesionless Sand Sampler
Top Del“! ("1:30 t tom General Description
0 9.1 NO SAMPLE
9.1 12.2 very fine to fine sand
12.2 12.8 NO SAMPLE
12.8 14.6 medium to coarse sand
14.6 14.9 NO SAMPLE
14.9 15.2 fine to medium sand
15.2 18.0 medium to coarse sand
18.0 18.3 NO SAMPLE
18.3 23.8 medium to coarse sand with some gravel
23.8 24.7 coarse sand and gravel
24.7 25.3 clay

 

 

Well Log for MSU-5
Ground Elevation: 265 m

Well drilled with a Hollow Stem Auger and aquifer sampled

with Waterloo Cohesionless Sand Sampler

 

 

 

 

Top Depth (mEottom General Description
0 9.1 NO SAMPLE
9.1 13.7 very fine to fine sand
13.7 14.6 medium sand
14.6 18.3 medium to coarse sand with some gravel
18.3 19.8 NO SAMPLE
19.8 20.1 fine to medium gravel
20.1 20.7 medium to coarse sand with some gravel
20.7 21.3 NO SAMPLE
21.3 25.3 coarse sand and gravel
25.3 25.6 NO SAMPLE
25.6 25.9 Clay

 

 

Well Log for MSU-6
Ground Elevation: 265.1 m

Well drilled with a Hollow Stem Auger and aquifer sampled

with Waterloo Cohesionless Sand Sampler

 

 

 

 

 

 

Top Depth (ml)30ttom General Description
0 9.1 NO SAMPLE
9.1 12.2 fine sand
12.2 12.5 NO SAMPLE
12.5 13.1 medium to coarse sand with some gravel
13.1 13.7 medium sand
13.7 17.4 medium to coarse sand with some gravel
17.4 18.3 fine to medium sand with some gravel
18.3 22.3 medium to coarse sand with some gravel
22.3 25.0 coarse sand and gravel
25.0 25.3 clay

 

115

 

 

Well Log for MSU-7
Ground Elevation: 265.1 m

Well drilled with a Hollow Stem Auger and aquifer sampled

with Waterloo Cohesionless Sand Sampler

 

 

 

 

 

 

Top Depth (mEottom General Description
0 9.1 NO SAMPLE

9.1 12.2 fine sand

12.2 13.1 medium to coarse sand

13.1 13.4 NO SAMPLE

13.4 13.7 fine sand

13.7 16.2 medium to coarse sand with some gravel
16.2 16.8 medium sand

16.8 18.6 NO SAMPLE

18.6 19.5 medium to coarse sand

19.5 19.8 NO SAMPLE

19.8 24.7 coarse sand and gravel

24.7 25.3 medium to coarse sand

25.3 25.6 fine to medium gravel

25.6 25.9 clay

 

 

Well Log for MSU-8
Ground Elevation: 264.9 m

Well drilled with a Hollow Stem Auger and aquifer sampled

with Waterloo Cohesionless Sand Sampler

 

 

 

 

Top Depth ("1:30 t tom General Description
0 9.1 NO SAMPLE
9.1 12.2 very fine to fine sand
12.2 13.1 coarse sand with some gravel
13.1 13.4 NO SAMPLE
13.4 13.7 fine to medium sand
13.7 16.8 medium to coarse sand
16.8 18.3 medium sand
18.3 27.1 coarse sand and gravel
27.1 27.4 gravel
27.4 27.7 clay

 

 

Well Log for MSU-9
Ground Elevation: 265 m

Well drilled with a Hollow Stem Auger and aquifer sampled

with Waterloo Cohesionless Sand Sampler

 

 

 

 

 

 

Top Depth (mgiottom General Description
0 9.1 NO SAMPLE
9.1 12.2 very fine to fine sand
12.2 13.1 medium sand
13.1 13.7 medium sand with some gravel
13.7 14.6 medium to coarse sand
14.6 18.3 medium to coarse sand with some gravel
18.3 22.9 NO SAMPLE
22.9 23.5 coarse sand and gravel
23.5 24.4 medium sand with some gravel

 

116

 

 

APPENDIX B

3-D SOLID STRATIGRAPHY MODELING IN GMS

INTRODUCTION

Appendix B was included to more fiilly explain the development of the 3-D solid
objects in GMS. The methodology is explained by describing the steps used to create one
of the 3-D solid objects. The solid that will be explained is depicted in Figure B-l, which
shows the maximum extent in plan view and the 3-D object in relation to the cross
section presented in Figure 22. The example is a solid that represents a portion of the
aquifer that is composed of material 2, or fine to medium sands. The general progression
started by creating boreholes from well log data, creating Triangular Interpolated

Networks (TINS) from the Boreholes, and then creating Solids from the TINS.

GMS BOREHOLE DATA ANALYSIS

The well logs were digitized into a specific file format that was required by GMS
before any work started. The process involved tabulating the X,Y,Z coordinates of the
materials in a given borehole and listing them in descending order (please refer to the
GMS manual (BYU, 2000) for further description of the Borehole file format). The
borehole file was then imported into GMS. A plan view map showing the locations of the
boreholes used is presented in Figure B-2. Note that there are several “dummy” wells
located along the perimeter of the local model area. These wells were created and
included so that irregular solid geometries would not be encountered along the margins of

the grid. The lithology and thickness of the materials in the “dummy” wells were simply

117

determined from the NUS or MSU wells closest to them. The aquifer materials near

plume G were based on observed data, thus the dummy wells had no effect on results.

TRIANGULAR INTERPOLATED NETWORKS

The first step required to make the TINS from the boreholes was to define a
boundary to contain the extrapolated TIN S. A boundary was created that was slightly
larger than the local model grid (Figure B-2). Two TINS were needed to make a given
solid; one TIN was needed to represent the top surface of the material and one TIN was
needed to represent the bottom surface of the material. For the example, the top tin was
created by selecting the contact of material 1 and material 2 in the appropriate boreholes
and then using the CONTACTS to TIN command of GMS. The TIN that is created
extends outward from the selected boreholes until it intersects the boundary polygon and
it is truncated midway between adjacent boreholes where no contact is selected. The
lower TIN was created in a similar manner, except that the nodes interior to the model
along the edge were moved to make it slightly larger than the top TIN (the reason is
explained later in the CREATION OF 3-D SOLID OBJECTS section). The top and
bottom TINS are shown in Figures B-3A and B-3B respectively with the boreholes used
to create them. The interior edge nodes of the top TIN were then snapped to the lower

TIN, which pinched out the material at the edge (Figure B-3C).

CREATION OF 3-D SOLID OBJECTS

The top and bottom TINS for the example were selected and the FILL BETWEEN

TINS to SOLIDS command was used to create the solid. GMS creates the solid by filling

118

downward from the top TIN until it intersects the lower TIN. If the lower TIN is smaller
than the top TIN, GMS will extrapolate far below the bottom TIN, creating an erroneous
solid. This was the sole reason for modifying the bottom TIN to make it slightly larger
than the top TIN. The solid was then complete, and was assigned appropriate material
properties as described in chapter HI (Figure B-3D). All of the other solids were created

in the exact same manner as just described.

119

 

 

 

 

 

 

 

 

 

 

 

 

l MATERIAL KEY I

- Material 1 — Very Fine to Fine Sand

- Material 2 — Fine to Medium Sands

- Material 3 — Coarse Sands

- Material 4 — Medium to Coarse Sands some Gravel
-

Material 5 — Coarse Sands, Gravels, and Pebbles

 

 

Figure B-l. Location of the solid for which an explanation of 3-D solid construction in
GMS is presented in Appendix B. The inset shows the areal extent of the solid with
respect to the local model grid and in reference to the cross-sections. Image is
presented in color (V E = 20X).

120

 

 

 

 

 

 

 

 

 

 

 

 

 

 

\
1.;

 

 

 

 

 

 

 

 

 

 

 

 

 

 

A
A
A ‘ ‘
A A
A
A “
, e‘ \
\
‘ ‘ ‘ \.

 

 

 

 

A NUS Monitoring Well

I MSU Monitoring Well
. “Dummy” Well

 

LEGEND

 

 

E2] Local Model Grid
1:] TIN Boundary Polygon

O 800
5:

meters

 

Figure B-2. Location of the wells used to create the TINS and solids in GMS.

121

 

 

 

 

 

 

 

D

l MATERIAL KEY I

Material 1 — Very Fine to Fine Sand

 

Material 2 — Fine to Medium Sands
Material 3 — Coarse Sands
Material 4 — Medium to Coarse Sands some Gravel

Material 5 — Coarse Sands, Gravels, and Pebbles

 

 

 

Figure B-3. Generalized steps of solid object construction from a set of borehole objects
in GMS. A) Creation of a TIN representing the top of a solid. B) Creation of a TIN
representing the bottom of a solid. C) Top TIN with interior nodes snapped to the bottom
TIN. D) Resulting solid created by filling between C) and B). Image is presented in color.

122

REFERENCES

123

REFERENCES

American Society for Testing and Materials (ASTM). 1995. Standard Guide for Risk-
Based Corrective Action Applied at Petroleum Release Sites. ASTM E-1739-95,
Philadelphia, PA.

Anderson, MP, and W.W. Woessner. 1992. Applied Groundwater Modeling: Simulation
of F low and Advective Transport. Academic Press Inc., San Diego, CA.

Azadpour—Keeley, A., J .W. Keeley, H.H. Russel, and G.W. Sewell. 2001. Monitored
Natural Attenuation of Contaminants in the Subsurface: Processes. Ground Water
Monitoring and Remediation, XXI(2): 97-107.

Aziz, C.E., C.J. Newell, J.R. Gonzales, P. Haas, T.P. Clement, Y. Sun, and D.G. Jewett.
2000. Biochlor: Natural Attenuation Decision Support System: User’s Manual.
USEPA, Cincinnati, OH.

Barrese, PG. 1991. A Hydrogeochemical Characterization of a Glacial-Drift Aquifer
System in Southwest Michigan. Master’s Thesis, Western Michigan University,
Kalamazoo, MI.

Bear, J ., M.S. Beljin, and RR. Ross. 1992. Fundamentals of Groundwater Modeling.
Ground Water Issue. Report EPA/540/S-92/005. Ada, OK, US. EPA.

Bedient, P.B., H.S. Rifai, C.J. Newell. 1999. Ground Water Contamination Transport
and Remediation, 2"d Edition. Prentice Hall, Upper Saddle River, NJ.

Boutt, D.F., D.W. Hyndman, B.C. Pijinowski, and D.T. Long. 2001. Identifying Potential
Land Use-Derived Solute Sources to Stream Baseflow Using Ground Water
Models and GIS. Ground Water, 39(1): 24-34.

Bouwer, H., and RC. Rice. 1976. A Slug Test For Determining Hydraulic Conductivity
of Unconfined Aquifers with Completely or Partially Penetrating Wells. Water
Resources Research, 12(3): 423-428.

Bouwer, E.J., B.E. Rittrnann, P.L. McCarty. 1981. Anaerobic Degradation of
Halogenated 1- and 2- Carbon Organic Compounds. Environmental Science and
Technology, 15(5): 596-599.

Bradley, P.M., and EH. Chappelle. 1998. Effect of Contaminant Concentration on

Aerobic Microbial Mineralization of DCE and VC in Stream-Bed Sediments.
Environmental Science and Technology, 32(5): 553-557.

124

Brigham Young University. 2000. Department of Defense Groundwater Modeling System
(GMS): Version 3.1 User '3 Manual. Salt Lake City, UT.

Buscheck, T.E., and GM. Alcantar. 1995. Regression Techniques and Analytical
Solutions to Demonstrate Intrinsic Bioremediation. Proceedings, 1995 Battelle
International Conference on In Situ and On Site Bioremediation. Batelle Press,
Columbus, OH.

Buxton, H., and TE. Reilly. 1986. A Technique for Analysis of Ground- Water Systems at
Regional and Subregional Scales Applied to Long Island, New York. US.
Geological Survey Water-Supply Paper 2310: 129-142.

Chapelle, F .H. 1993. Ground- Water Microbiology and Geochemistry. John Wiley and
Sons, Upper Saddle River, NJ.

Chapelle, RH. 1997. Identifying Redox Conditions that Favor the Natural Attenuation of
Chlorinated Ethenes in Contaminated Ground-Water Systems. Proceedings,
Symposium on Natural Attenuation of Chlorinated Organics in Ground Water,
Dallas, Texas, September 11-13, 1996: 19-22. Report EPA/540/R-97/504, US.
EPA, Washington, DC.

Clement, TR 1997. RT 3D - A Modular Computer Code for Simulating Reactive Multi-
Species Transport in 3-Dimensional Groundwater Aquifers. Battelle Pacific
Northwest National Laboratory Report, PNNL-l 1720.

Clement, T. P., C.D. Johnson, Y. Sun, G.M. Klecka, and C. Bartlett. 2000. Natural
Attenuation of Chlorinated Ethene Compounds: Model Development and F ield-
Scale Application at the Dover Site. Journal of Contaminant Hydrology, 42: 113-
140.

Cline, P.V., and J .J . Delfino. 1989. Effect of Subsurface Sediment on Hydrolysis of
Haloalkanes and Epoxides. In Biohazards of Drinking Water Treatment. R.A.
Larson editor, Lewis Publishers, Chelsea, MI.

de Bruin,W.P., M.J.J. Kotterman, M.A. Posthumus, D. Schraa, and A.J.B. Zehnder. 1992.
Complete Biological Reductive Transformation of Tetrachloroethylene to Ethene.

Applied Environmental Microbiology, 58(6): 1996-2000.

Dorr, J .A., and DP. Eschman. 1970. The Geology of Michigan. University of Michigan
Press, Ann Arbor, MI.

Driscoll, F.G. 1986. Groundwater and Wells 2’“! Edition. U.S. Filter/Johnson Screens, St.
Paul, MN.

ESRI. 1998. ArcView. E.S.R.I., Redlands, CA.

125

Everett, E. 1990. Soil Remediation Plan, Contaminated Soil Areas, Arco Industries
Corporation Schoolcraft, Michigan. Hunter/Keck, Inc., Williamston, MI.

Fetter, CW. 1993. Contaminant Hydrogeology. Prentice Hall, Upper Saddle River, NJ.

Finkbeiner, K. 1994. Subsurface Glacial Geology of the Area Between the T ekonsha and
Kalamazoo Moraines, Kalamazoo County, Michigan. Master’s Thesis, Western
Michigan University, Kalamazoo, MI,

Forstat, D.W., and Sorenson. 1982. Bedrock Topography of Kalamazoo County. County
Geologic Map Series, Michigan Geological Survey Division, Lansing, MI.

Freedman, D.L., and J .M. Gosset. 1989. Biological Reductive Dechlorination of
Tetrachloroethylene and Trichloroethylene to Ethylene Under Methanogenic
Conditions. Applied Environmental Microbiology, 55: 2144-2151.

Freeze, R.A., and J .A. Cherry. 1979. Groundwater. Prentice Hall, Englewood Cliffs, NJ.

Haag, W.R., and T. Mill. 1988. Transformation Kinetics of 1,1,1-Trichloroethane to the
Stable Product 1,1-Dichloroethene. Environmental Science and Technology, 22:
658-663.

Hoard, C. J. 2002. The Value of Hydraulic Conductivity Data in Groundwater Flow and
Transport Models of Schoolcraft Plume A. Master’s Thesis, Michigan State
University, East Lansing, MI.

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.

Ibrahim, A. 1970. The Application of Gravity Method to Mapping Bedrock on Kalamazoo
County, Michigan. P.h.D. Dissertation, Michigan State University, East Lansing,
MI.

Jackson, RE, and V. Dwarakanath. 1999. Chlorinated Degreasing Solvents: Physical-
Chemical Properties Affecting Aquifer Contamination and Remediation. Ground
Water Monitoring and Remediation 19(4): 102-1 10.

Kehew, A.E., W.T. Straw, W.K. Steinmann, P.G. Barrese, G. Passarella, and W. Peng.
1996. Groundwater-Quality and Flow in a Shallow Glaciofluvial Aquifer
Impacted by Agricultural Contamination. Ground Water 34(3): 491-500.

Leverett, F., and RB. Taylor. 1915. The Pleistocene of Indiana and Michigan and the
History of the Great Lakes. US. Geological Survey, Monologue LIB.

126

Lillienthal, RT. 1978. Stratigraphic Cross Sections of the Michigan Basin. Michigan
Geological Survey Division, Report Investigation 19.

Martin, H. 1957. Outline of the Geologic History of Kalamazoo County. Michigan
Geological Survey Division, Open File Report.

Mayotte, T.J. 1991. Village of Schoolcraft Kalamazoo County, Michigan Final Remedial
Investigation and Feasibility Study of Remedial Alternatives, Plume A and Plume
F. Halliburton NUS Environmental Corporation Project Number 4M84.

McCarty, PL. and L. Semprini. 1994. Ground-Water Treatment for Chlorinated Ethenes.
In Handbook of Bioremediation, R. N. Noriss editor, Lewis Publishers, Boca
Raton, FL.

McCarty, PL. 1997. Biotic and Abiotic Transformations of Chlorinated Solvents in
Ground Water. Proceedings, Symposium on Natural Attenuation of Chlorinated
Organics in Ground Water, Dallas, Texas, September 11-13, 1996: 9-11. Report
EPA/540/R-97/504, US. EPA, Washington, DC.

McCarty, P.L., M.N. Goltz, G.D. Hopkins, M. E. Dolan, J .P. Allan, B.T. Kawakami, and
T.J. Carrothers. 1998. Full-Scale Evaluation of In Situ Cometabolic Degradation
of Trichloroethylene in Groundwater Through Toluene Injection. Environmental
Science and T echnolog», 32(1): 88-100.

McDonald, 6., and A.W. Harbaugh. 1988. A Modular Three-Dimensional F inite-
Diflerence Groundwater Flow Model. US. Geological Survey Techniques of
Water-Resources Investigations, Book 6, chapter A1, Reston, VA.

Michigan Resource Inventory System (MiRIS). 2000. Groundwater Database. Michigan
Department of Environmental Quality, Lansing, MI.

Miller, RT, and CI. Voss. 1987. Finite-Difference Grid for a Doublet Well in an
Anisotropic Aquifer. Ground Water, 24(4): 490-496.

Milstein, R.J. 1987. Bedrock Geology of Michigan. State of Michigan, Department of
Natural Resources, Geological Survey, Lansing, MI.

Monaghan, G.W., and G.J. Larson. 1982. Surficial Geology of Kalamazoo County.
County Geologic Map Series, Michigan Geologic Survey, Lansing, MI.

Monaghan, G.W., and G.J. Larson. 1986. Late Wisconsinan Drift Stratigraphy of the
Saginaw Ice Lobe in South-Central Michigan. GSA Bulletin, 97: 324-328.

Monaghan, G.W., G.J. Larson, and GD. Gephatt. 1986. Late Wisconsinan Drift

Stratigraphy of the Lake Michigan Lobe in Southwestern Michigan. GSA Bulletin,
97: 329-334.

127

Passero, RN. 1981. Hydrogeologic Investigation for the Kalamazoo County, KL Avenue
Sanitary Landfill. Wilkins and Wheaton Testing Laboratory.

Reilly, TE. 2001. System and Boundary Conceptualization in Ground- Water Flow
Simulation. Techniques of Water-Resources Investigations of the US. Geological
Survey, Book 3, Applications of Hydraulics, Chapter B8, Reston, VA.

Rheaume, SJ. 1990. Geohydrology and Water Quality of Kalamazoo County, Michigan,
1986-88. Water-Resources Investigations Report 90-4028, Lansing, MI.

Semprini, L., P.K. Kitanidis, D.H. Kampbell, and J.T. Wilson. 1995. Anaerobic
Transformation of Chlorinated Aliphatic Hydrocarbons in a Sand Aquifer Based
on Spatial Chemical Distributions. Water Resources Research, 31(4): 1051-1062.

Steinmann, W.K. 1994. The Geological and Hydrogeological Characterization of the
Schoolcraft Aquifer, Schoolcraft, Michigan. Master’s Thesis, Western Michigan
University, Kalamazoo, MI.

Suarez, M.P., and HS. Rifai. 1999. Biodegradation Rates for Fules Hydrocarbons and
Chlorinated Solvents in Groundwater. Bioremediation Journal, 3(4): 337-362.

US. Environmental Protection Agency. 1997. Use of Monitoring Natural Attenuation at
Superfund, RCRA Corrective Action, and Underground Storage Tank Sites. Office
of Solid Waste and Emergency Response Directive 9200.4-17.

Vogel, T.M., and PL. McCarty. 1985. Biotransfonnation of Tetrachloroehtylene to
Trichloroethylene, Dichloroethylene, Vinyl Chloride, and Carbon Dioxide under
Methanogenic Conditions. Applied and Environmental Microbiology, 49(5):
1080-1083.

Vogel, T.M., C.S. Criddle, and PL. McCarty. 1987. Transformation of Halogentated
Aliphatic Compounds. Environmental Science and Technology 21(8): 722-736.

Ward, D.S., D.R. Buss, J .W. Mercer, and SS. Hughes. 1987. Evaluation of a
Groundwater Corrective Action at the Chem-Dyne Hazardous Waste Site Using a

Telescopic Mesh refinement Modeling Approach. Water Resources Research
23(4): 603-617.

Wiedemeier, T.H., M.A. Swanson, D.E. Moutoux, E.K. Gordon, J .T. Wilson, B.H.
Wilson, D.H. Kampbel, J. Hansen, and P. Haas. 1996. Technical Protocol for
Evaluating Natural Attenuation of Chlorinated Solvents in Groundwater. Air

Force Center for Environmental Excellence, Technology Transfer Division,
Brooks AFB, San Antonio, Texas.

Wilson, J .T., D.H. Kampbell, and J .W. Weaver. 1997. Environmental Chemistry and the
Kinetics of Biotransformation of Chlorinated Organic Compounds in Ground

128

Water. Proceedings, Symposium on Natural Attenuation of Chlorinated Organics
in Ground Water, Dallas, Texas, September 11-13, 1996: 133-136. Report
EPA/540/R-97/504, US. EPA, Washington, DC.

Xu, M., and Y. Eckstein. 1995. Use of Weighted Least-Squares Method in Evaluation of
the Relationship Between Dispersivity and Scale. Ground Water, 33(6): 905-908.

Youngs, E.G., G. Spoor, and GR. Goodall. 1996. Infiltration from Surface Ponds into

Soils Overlying a Very Permeable Substratum. Journal of Hydrology, 186: 327-
334.

129