United States
Environmental Protection
Agency
EPA 600/R-09/096 I September 2009 I www.epa.gov/ada
Impacts of DNAPL Source Treatment:
Experimental and Modeling
Assessment of the Benefits of
Partial DNAPL Source Removal
««: 20 SO 100 200 300 1000
Office of Research and Development
National Risk Management Research L
atorv, Ada, Oklahoma 74820
-------
-------
Impacts of DNAPL Source Treatment:
Experimental and Modeling
Assessment of the Benefits of
Partial DNAPL Source Removal
SERDP Project ER-1295
A. Lynn Wood, Principal Investigator
U.S. Environmental Protection Agency
Michael D. Annable
James W. Jawitz
University of Florida
Ronald W. Falta
Clemson University
Michael C. Brooks
Carl G. Enfield
U.S. Environmental Protection Agency
P. Suresh C. Rao
Purdue University
Mark N. Goltz
Air Force Institute of Technology
Office of Research and Development
National Risk Management Research Laboratory, Ada, Oklahoma 74820
-------
Notice
The work on which this report is based was supported by the U. S. Environmental
Protection Agency through its Office of Research and Development with funding
provided by the Strategic Environmental Research and Development Program
(SERDP), a collaborative effort involving the U.S. Department of Defense (DoD),
the U.S. Department of Energy (DOE), and the U.S. Environmental Protection
Agency (EPA). It has been subjected to EPA's peer and administrative review and
has been approved for publication as an EPA document. Mention of trade names or
commercial products does not constitute endorsement or recommendation for use.
All research projects making conclusions or recommendations based on
environmentally related measurements and funded by the Environmental Protection
Agency are required to participate in the Agency Quality Assurance Program. This
project was conducted under an approved Quality Assurance Project Plan. The
procedures specified in this plan were used without exception. Information on the
plan and documentation of the quality assurance activities and results are available
from the Principal Investigator.
-------
Foreword
The U.S. Environmental Protection Agency is charged by Congress with protecting the Nation's land, air, and water
resources. Under a mandate of national environmental laws, the Agency strives to formulate and implement actions
leading to a compatible balance between human activities and the ability of natural systems to support and nurture
life. To meet this mandate, EPA's research program is providing data and technical support for solving environmental
problems today and building a science knowledge base necessary to manage our ecological resources wisely,
understand how pollutants affect our health, and prevent or reduce environmental risks in the future.
The National Risk Management Research Laboratory (NRMRL) is the Agency's center for investigation of
technological and management approaches for preventing and reducing risks from pollution that threatens human
health and the environment. The focus of the Laboratory's research program is on methods and their cost-effectiveness
for prevention and control of pollution to air, land, water, and subsurface resources; protection of water quality in
public water systems; remediation of contaminated sites, sediments, and ground water; prevention and control of
indoor air pollution; and restoration of ecosystems. NRMRL collaborates with both public and private sector partners
to foster technologies that reduce the cost of compliance and to anticipate emerging problems. NRMRL's research
provides solutions to environmental problems by: developing and promoting technologies that protect and improve
the environment; advancing scientific and engineering information to support regulatory and policy decisions; and
providing the technical support and information transfer to ensure implementation of environmental regulations and
strategies at the national, state, and community levels.
This report has been prepared as part of the Laboratory's strategic long-term research plan. At many hazardous waste
sites contaminants reside in the subsurface as separate dense non-aqueous phase liquids (DNAPL). These DNAPL
serve as persistent sources of dissolved phase contamination and are a major impediment to successful and cost-
effective site cleanup. Although substantial progress has been made in the development of source zone remediation
technologies, total DNAPL removal remains problematic and the benefits of partial removal have been questioned.
The research reported here was undertaken to evaluate these benefits and to provide information and tools to
support the development of technically and economically responsible strategies for mitigating the impact of DNAPL
contaminated sites on human health and the environment.
Robert WTuls, Acting Director
Ground Water and Ecosystems Restoration Division
National Risk Management Research Laboratory
-------
-------
Contents
List of Figures vii
List of Tables xi
List of Acronyms xii
Acknowledgments xiii
Executive Summary xv
1.0 Objectives 1
2.0 Background 1
3.0 Materials and Methods 5
3.1 Data Quality 6
3.2 Field Studies 7
3.3 Laboratory Studies 7
3.4 Modeling Studies 8
4.0 Results and Accomplishments 9
4.1 Field Research 9
4. .1 Flux Measurement Methods 9
4. .2 Hill Air Force Base Operable Unit 2 12
4. .3 Fort Lewis East Gate Disposal Yard Site 17
4. .4 Former Sages Dry Cleaner Site, Jacksonville, Florida 22
4. .5 Borden CFB Site 30
4. .6 MIPT Uncertainty Analysis 34
4.2 Laboratory Research 45
4.2.1 Light Transmission Visualization Methods 45
4.2.2 Relationship between Source Mass and Source Strength 50
4.2.2.1 DNAPL source depletion: Linking architecture and flux response 50
4.2.2.2 Fluid and porous media property effects on dense non-aqueous phase
liquid migration and contaminant mass flux 54
4.2.2.3 Reactive tracer tests to predict dense non-aqueous phase liquid dissolution
dynamics in laboratory flow chambers 58
4.2.2.4 Laboratory investigation of flux reduction from dense non-aqueous phase
liquid (DNAPL) partial source zone remediation by enhanced dissolution ... 62
4.2.2.5 The impacts of partial remediation by sparging on down-gradient
DNAPL mass discharge 68
4.3 Theoretical and Modeling Research 73
4.3.1 Lagrangian Models 74
4.3.2 Coupled Source Zone and Plume Models 78
4.3.2.1 Analytical expressions of source flux and mass 80
4.3.2.2 Analytical approaches including plume remediation and natural attenuation. . 83
4.3.2.3 REMChlormodel 85
4.3.2.4 Cancer risk assessment 91
4.3.2.5 Analytical/numerical approach including plume remediation and natural
attenuation 92
-------
4.3.3 Model Assessments and Applications 96
4.3.3.1 Simulation of laboratory and field air sparging 96
4.3.3.2 Simulations of cosolvent flushing 102
4.3.3.3 Dissolved chemical discharge from fractured clay 110
4.3.3.4 Monte Carlo simulations of source behavior with remediation 113
4.3.3.5 Simplified contaminant source depletion models as analogs of
multiphase simulators 117
4.3.3.6 Temporal evolution of DNAPL source and contaminant flux
architecture 123
4.3.3.7 Predicting DNAPL dissolution using a simplified source depletion
model parameterized with partitioning tracers 127
5.0 Conclusions 131
6.0 References 133
7.0 Appendix A - List of Technical Publications 143
Peer-Reviewed Papers 143
Technical Reports 144
Conference Proceedings 145
Published Technical Abstracts 145
Book Chapters 149
-------
Figures
Figure ES-1. Impacts of Source Remediation: Conceptual Model xv
Figure 2.0-1. Plan view of DNAPL source and dissolved plume 2
Figure 2.0-2. Impacts of source remediation: conceptual model 3
Figure 2.0-3. Contaminant mass flux leaving the source zone 4
Figure 3.0-1. Integrated research approach 5
Figure 3.0-2. Project organizational chart 6
Figure 3.3-1. Coupled modeling approach used in the project 8
Figure 4.1-1. Plan view of Panel 5 area at Hill AFB 13
Figure 4.1-2. Hydrographs from Hill AFB OU2 wells U2-214 and U2-117, along with the
average clay elevation (dashed line) across the transect 14
Figure 4.1-3. Contaminant flux distribution from selected wells with the largest fluxes as
measured prior to remediation at Hill AFB 14
Figure 4.1-4. Average mass discharge (g/day) for each well at Hill AFB OU2, as measured by
PFMs for a) May 2002 (pre-remedial), b) June 2003 (first post-remedial), and
c) October 2004 (second post-remedial) tests; and as measured by MIPTs for
d) May 2002 (pre-remedial), e) June 2003 (first post-remedial), and f) October 2004
(second post-remedial) MIPTs 15
Figure 4.1-5. Plan view of the NA1 source area at the East Gate Disposal Yard site at Fort Lewis
and the downgradient flux well transect 18
Figure 4.1-6. Mass flux profiles measured in selected wells with the largest fluxes as measured
prior to remediation using the PFMs 20
Figure 4.1-7. Average mass discharge (g/day) for each well at Fort Lewis. Measured by a) PFMs
and b) MIPT for the pre-remedial tests; and by c) PFMs and d) MIPT for the
post-remedial tests. TCE is shown in black, and cis-DCE is shown in white 21
Figure 4.1-8. Multilevel sampler locations, flux well locations, and the general areas targeted
during the first and second ethanol flushing events at the Sages site 23
Figure 4.1-9. MLS source zone PCE/TCE and TCE/c-DCE ratios as a function of ethanol
concentration in the groundwater up to four years after the 1998 pilot test (Sages site). . .24
Figure 4.1-10. Sages source zone transect natural gradient concentrations for the 6-year period after
the August 1998 ethanol flood. Values are millimolar sums of chlorinated
ethenes (mM) 25
Figure 4.1-11. Sages source zone mass discharge (mmol/day) before and following the August 1998
ethanol flood 26
Figure 4.1-12. Sages downgradient concentration distribution for the 6-year period 27
Figure 4.1-13. Sages downgradient transect mass discharge in mmol/day for the six-year period
following the 1998 ethanol flushing event 28
Figure 4.1-14. Pre- and post-cosolvent flood MLS mass discharge measurements in the downgradient
transect at the Sages site 28
Figure 4.1-15. Pre- and post-cosolvent flood passive flux meter mass discharge measurements in
the down gradient well transect at the Sages site 29
-------
Figure 4.1-16. Pre- and post-cosolvent flood MIPT mass discharge measurements in the down
gradient transect at the Sages site 30
Figure 4.1-17. Borden controlled release DNAPL site instrumentation 31
Figure 4.1-18. PCE mass flux from passive flux meters installed in wells down-gradient of row 6
at the Borden CFB site 32
Figure 4.1-19. PCE mass flux from MIPT measurements at the Borden CFB site 32
Figure 4.1-20. Example breakthrough curve for the pre-remediation partitioning tracer test at the
Borden CFB site 33
Figure 4.1-21. Pre- and post-remedial mass flux distribution of PCE in the row 6 transect at the
Borden CFB site 34
Figure 4.1-22. Covariance of J plotted as a function of the covariance for C and q assuming C and q
are independent random variables 35
Figure 4.1-23. Standard deviation and the coefficient of variation for q as a function of uncertainty
in head measurements 37
Figure 4.1-24. Covariance of J plotted as a function of the covariance for C and the standard
deviation of head measurements 37
Figure 4.1-25. Layered (a) and block (b) macro-scale hydraulic conductivity patterns investigated
with numeric modeling simulations 39
Figure 4.1-26. Capture zones for an MIPT with seven concurrent pumping wells in an unconfined
aquifer with K = 10 m/day and q = 5 cm/day 40
Figure 4.1-27. Capture zones for an MIPT with seven concurrent pumping wells in an unconfined
aquifer with a log-normally distributed random K field 41
Figure 4.1-28. Comparison of capture zones with seven concurrent pumping wells for the
homogeneous case (a) and heterogeneous cases: realization 1 (b) and realization 2 (c). . .42
Figure 4.1-29. a) Capture zones during an MIPT consisting of seven concurrently pumping wells, and
b) the resulting concentration-time series for a mature contaminant plume centered at
y= 1.5m (midway between the third and fourth pumping wells), a maximum
concentration of 20 mg/L, and a plume width described by a standard deviation of
0.4m 43
Figure 4.1-30. Results from Monte Carlo simulations where plume position is treated as a uniform
random variable 44
Figure 4.2-1. Distribution of PCE in porous media as determined by LTV. 49
Figure 4.2-2. PCE volume calculated from image analysis versus actual PCE volume present in
the model 49
Figure4.2-3. Flux plane response to changes in NAPL architecture for experiment DCA-1 51
Figure 4.2-4. Comparison of the rate-limited streamtube, equilibrium streamtube, and effective
Damkohler approaches for modeling source depletion from experiments DCA-1,
TCE-1, and DCA-2 53
Figure 4.2-5. Tracings of observed distributions of DNAPL injected 10 cm above the flow chamber
bottom in 30/40 sand with a) PCE and untreated sand, b) PCE and 50% OTS treated
hydrophobic sand mixture, c) PCE and 100% hydrophobic sand, d) untreated sand and
DNAPLmixture of density 1.1 g/cm3 56
Figure 4.2-6. Fractional flux increase versus DNAPL loading for each interfacial tension value in
untreated 30/40 sand 57
Figure 4.2-7. Fractional flux increase versus PCE loading for varying density and media 57
Figure 4.2-8. PCE spills before flushing. Images from experiments 1 and 2 are reflected light
captured with standard digital camera and the NAPL is dyed red 60
-------
Figure 4.2-9. BTCs from surfactant flushing for Experiments I and II 61
Figure 4.2-10. BTCs from surfactant flushing for Experiment III 62
Figure 4.2-11. Displacement of resident water at different cosolvent injection volumes (PV) 64
Figure 4.2-12. BTCs for all Single Flush experiments with cosolvent mixtures 65
Figure 4.2-13. Aqueous based mass reduction ( RM ) versus flux reduction (Rj) of the single and
multiple-flushing experiments with the solubility scaled approach (SSA) path
estimation of the single-flushing experiment 66
Figure 4.2-14. Digital photographs of the 2-D flow chamber: (a) and (b) show the initial distribution
of PCE in Pack A and Pack B, respectively, and (c) and (d) show the sparging channels
during the sparging experiments for Pack A and Pack B, respectively 69
Figure 4.2-15. The initial PCE saturation distribution in the flow chamber and the saturation
distribution following each sparging event 71
Figure 4.2-16. Relative effluent mass discharge as a function of PCE mass removed for the two
experiments 72
Figure 4.3-1. Lagrangian conceptualization of DNAPL source zones 75
Figure 4.3-2. (a) Field dissolution data (from Jawitz et al., 1998) compared to equilibrium solution
with a bimodal travel time distribution and homogeneous NAPL content (b) Rj (RM)
based on model parameters 77
Figure 4.3-3. Measured and predicted remedial performance in a vertical circulation implementation
of surfactant extraction (Wood et al., 2005) 78
Figure 4.3-4. Solid lines are reduction in contaminant flux as a function of reduction in source zone
mass for a In i ={0.05, 0.2, 0.6, 1.0, 1.5, 2, 3} 78
Figure 4.3-5. Source zone dissolved concentrations as a function of source zone DNAPL mass
(from Falta et al., 2005a) 79
Figure 4.3-6. Power function representation of source mass/source discharge relationship
(Equation 4.3-6) 80
Figure 4.3-7. Conceptual model of source zone with time-dependent contaminant mass and
discharge 80
Figure 4.3-8. Source zone dissolved concentration with and without source remediation for F= 0.5
(from Falta et al., 2005a) 83
Figure 4.3-9. Source zone dissolved concentrations with and without source remediation for F = 2.0
(from Falta et al., 2005a) 83
Figure 4.3-10. Distance-time plot for advective transport with a single set of reaction rates 86
Figure 4.3-11. Distance-time plot for advective transport with multiple sets of reaction rates 87
Figure 4.3-12. Hypothetical design of an enhanced plume remediation scheme with an enhanced
reductive dechlorination zone for destruction of PCE and TCE and an enhanced
aerobic degradation zone for destruction of DCE and VC 88
Figure 4.3-13. Comparison of the REMChlor reactive streamtube solution with BIOCHLOR for a
two-zone case with steady state concentrations in the upstream zone (t = 20 years) 90
Figure 4.3-14. Steady-state TCE concentration at varying source mass reductions for a monitoring
well 800 meters downgradient of the source 95
Figure 4.3-15. Steady-state vinyl chloride (VC) concentrations at varying source mass reductions
for a monitoring well 800 meters downgradient of the source 95
Figure 4.3-16. A simulated sparge zone in a single medium multiphase flow model versus realistic
channeling that occurs in air sparging 97
-------
Figure 4.3-17. Schematic of connections in a dual media grid. Both media are globally connected
and share a local connection 98
Figure 4.3-18. Results from air sparging experiments (Heron et al. 2002) 99
Figure 4.3-19. Total experimental mass removal and simulated removal with TMVOC using a
dual-media approach withAu/d2 ratio from the laboratory-scale air sparging
simulation (75,000) and the adjusted AJd2 ratio (75) 101
Figure 4.3-20. Modeled and measured ternary phase diagram for the system of water-ethanol-PCE.. . .104
Figure 4.3-21. Plan view of model domain with the site map of well and MLS locations within the
approximated DNAPL source zone at Sages (adapted from Jawitz et al. 2000) 105
Figure 4.3-22. PCE effluent concentrations and ethanol effluent percentages for RW002, RW003,
RW004, RW005, RW006, RW007, and total fluid (heterogeneous case) 107
Figure 4.3-23. PCE effluent concentrations and ethanol effluent percentages for the calibrated
heterogeneous case 109
Figure 4.3-24. Conceptual view of a fractured clay aquitard Ill
Figure 4.3-25. Simulated TCE aqueous concentration in the fracture during the 20-year equilibrium
period for a 30 |im fracture with no TCE decay and a clay^c of 0.01 Ill
Figure 4.3-26. TCE concentration in the fracture effluent for the 30 |im and 100 |im fractures 112
Figure 4.3-27. TCE concentration in the fracture effluent for the 30 jam and 100 jam fractures 113
Figure 4.3-28. Two input probability distribution functions (without source remediation) 114
Figure 4.3-29. Probability simulation output of source concentration (without source remediation). . . .115
Figure 4.3-30. Three input probability distribution functions (with source remediation) 116
Figure 4.3-31. Probability simulation output of source concentration (with source remediation) 117
Figure 4.3-32. DNAPL spills for six representative cases 120
Figure 4.3-33. UTCHEM-generated dissolution profiles (symbols) and corresponding model fits for
simulation realizations 3-1 and 3-3 (tfm, = 3 ) 121
Figure 4.3-34. UTCHEM-generated mass reduction-flux reduction profiles (symbols) compared with DaM
(Set 1-3), ESM (Set 2-3), ESM (Set 3-l),and PLM (Set 3-3) fits (lines) to the dissolution
profile 122
Figure 4.3-35. Relationship between the relative reductions in the mean and standard deviations
of the contaminant flux 125
Figure 4.3-36. Locus of the Centroid of the DNAPL and contaminant flux distributions (at the source
control plane) in Domain 2 (white circles indicate the initial location of the centroid
and black stars denote their final location) 125
Figure 4.3-37. Spatial variance of the DNAPL and contaminant flux distributions in (a) Domain 1,
Case 1 (ISCO3D), and Domain 2, Cases 1 and 2 (T2VOC) 126
Figure 4.3-38. Non-partitioning and partitioning tracer data for simulation set 2-4 127
Figure 4.3-39. ESM-predicted and UTCHEM-generated dissolution profiles for simulation
sets 1-4, 2-4, and 3-2 128
Figure 4.3-40. ESM-predicted and UTCHEM-generated mass reduction-flux reduction profiles for
simulation set 1-4, 2-4, and 3-2 128
Figure 4.3-41. Comparison between the ESM and the GTP models for prediction of source
dissolution behavior 130
-------
Tables
Table ES-1. Summary of field demonstrations to assess DNAPL mass depletion flux reduction
relationships xvii
Table 3.1-1. Field study sites 7
Table 4.1-1. Summary of TCE and DCE mass discharge (g/day) and combined TCE and DCE
molar discharge (moles/day) for Hill AFB OU2 using PFM and MIPT results, as
well as estimates based on the Transect Method (TM) 16
Table 4.1-2. Summary of TCE and DCE mass discharge (g/day) and combined TCE and DCE
molar discharge (moles/day) for Fort Lewis EGDY site using PFM and MIPT results,
as well as estimates based on the Transect Method (TM) 22
Table 4.2-1. PCE distribution recovery, mean, and variance determined using the LTV models 47
Table 4.2-2. Summary of flow chamber experiments 48
Table 4.2-3. Summary of experimental conditions 50
Table 4.2-4. PCE/n-decane mixture densities, media sieve and grain sizes, and best fit p values
for mass loading/flux generation relationships measured in these media 55
Table 4.2-5. Summary of the four experiments, equipment and truncated moment analysis results. . . .59
Table 4.2-6. Fluid properties 63
Table 4.2-7. Numerical simulation input parameters 67
Table 4.2-8. Summary of PCE sparging test parameters 71
Table 4.2-9. Summary of PCE sparging test results 72
Table 4.3-1. Fraction of DNAPL source removal (X) required to reduce maximum plume length
by a specified amount using Equation (4.3-19) (from Falta et al., 2005a) 84
Table 4.3-2. California cancer risk slope factors for PCE and its degradation products
(OEHHA, 2006) 92
Table 4.3-3. Reaction zone degradation rates (Clement et al., 2000) 93
Table 4.3-4. Percent mass reduction necessary to achieve remediation goals 96
Table 4.3-5. Constant input parameters used in probabilistic simulation applications 115
Table 4.3-6. Attributes of the simulated DNAPL source zones used in the numerical simulations.. . .123
Table 4.3-7. Coefficient of variation (CV) of the Darcy flux (q ), contaminant flux (J), and the
DNAPL mass (^N) distribution 124
Table 4.3-8. Tracer-derived ESM parameters 129
-------
Acronyms
ADM Advection Dispersion Model
API American Petroleum Institute
BDI Borden DNAPL Injection
ETC Breakthrough Curve
CCD Charge Coupled Device
CDI Chronic Daily Intake
COV Coefficient of Variation
CP Control Plane
CV Coefficient of Variation
CVOC Chlorinated Volatile Organic
Compounds
DaM Damkohler number Model
DCE Dichloroethylene
DLL Dynamic-Link Library
DNAPL Dense Non-Aqueous Phase Liquid
EGDY East Gate Disposal Yard
EL Ethyl-Lactate
EPA Environmental Protection Agency
EPA/ORD Environmental Protection Agency/
Office of Research and Development
ESM Equilibrium Streamtube Model
GC/FID Gas Chromatograph / Flame lonization
Detector
GC/MS Gas Chromatograph/Mass
Spectrometer
GMS Groundwater Modeling System
GoldSim Probabilistic Simulation Software
GTP Ganglia To Pool ratio
IMPES Implicit in Pressure
IFT Interfacial Tension
IPT Integral Pumping Test
ISCO In-Situ Chemical Oxidation
LaSAR Lagrangian Stochastic Advective
Reactive
LTV Light Transmission Visualization
MCL Maximum Contaminant Level
MF Multiple Flushing
MIBK Methyl IsoButyl Ketone
MIPT Modified Integral Pumping Test
MLS Multi-Level Sampler
MODFLOW MODular three-dimensional finite-
difference ground-water FLOW model
NAPL Non-Aqueous Phase Liquid
ND Neutral Density
NDS Neutral Density Surfactant
NRMSD Normalized Root Mean Square
Deviation
OEHHA California Office of Environmental
Health Hazard Assessment
OU2 Operable Unit 2
PCE Tetrachloroethylene
PDF Probability Density Function
PFM Passive Flux Meter
PITT Partitioning Inter-well Tracer Test
PLM Power Law Model
PM Point measurement
RA Reagent Alcohol
REMChlor Remediation Evaluation Model for
Chlorinated Solvents
RMSE Root Mean Square Error
RT3D Three-dimensional, multispecies
reactive transport in groundwater
simulator
RW Recovery Well
SERDP Strategic Environmental Research and
Development Program
SF Single Flush
SI-GAC Silver-Impregnated Granular Activated
Carbon
S1F Cancer risk Slope Factor
SSLE Sum of Squares Logarithmic Error
T2VOC Numerical simulator for three-phase,
three-component, non-isothermal flow
of water, air and a volatile organic
compound in multidimensional
heterogeneous porous media
TCE Trichloroethylene
TE Water to gas Transfer Efficiency
TM Transect Method
TMVOC A numerical simulator for three-
phase non-isothermal flows of
multicomponent hydrocarbon mixtures
in saturated-unsaturated heterogeneous
media
TVD Total Variation Diminishing
USAGE U.S. Army Corps of Engineers
USGS United States Geological Survey
UTCHEM University of Texas Chemical
Composition Simulator
VC Vinyl Chloride
VOC Volatile Organic Compound
-------
Acknowledgments
The research upon which this report is based was a joint effort between the Strategic Environmental Research and
Development Program (SERDP, project ER-1295), EPA's National Risk Management Research Laboratory, Air
Force Institute of Technology (AFIT), Clemson University, University of Florida, and Purdue University. This report
presents the collective work of the research team. Contributing authors and their affiliations are listed below:
Contributing Author
Michael D. Annable
Mustafa M. Bob
Michael C. Brooks
Carl G. Enfield
Ronald W. Falta
Mark N. Goltz
James W. Jawitz
P. Suresh C. Rao
A. Lynn Wood
Affiliation
University of Florida, Department of Environmental Engineering Sciences, Gainesville,
FL32611
USEPA/ORD, National Risk Management Research Laboratory, Ada, OK 74820
USEPA/ORD, National Risk Management Research Laboratory, Ada, OK 74820
USEPA/ORD, National Risk Management Research Laboratory, Cincinnati, OH 45268
(retired)
Clemson University, Department of Environmental Engineering and Earth Sciences,
Clemson, SC 29634
US Air Force Institute of Technology, Wright-Patterson Air Force Base, OH 45433
University of Florida, Soil and Water Science Department, Gainesville, FL 32611
Purdue University, School of Engineering, West Lafayette, IN 47907
USEPA/ORD, National Risk Management Research Laboratory, Ada, OK 74820
Appreciation is extended to the following individuals who assisted with the many field, laboratory and modeling
studies that contributed to this project: A. Noman M. Ahsanuzzaman, Steve Acree, Nandita Basu, Sten Berglund,
Carol Bratt, Gordon Brown, Xaisong Chen, Jaehyun Cho, Pat Clark, Murray E. Close, Ravi Costa, Joseph J. Delfino,
Erica DiFilippo, George G. Demmy, Rob Earle, Felipe P. Espinoza, Dennis Fine, Mark Flintoft, Adrian Pure, John
Gierke, Erik Gonzales, Kirk Hatfield, John Hoggatt, Chuck Holbert, Junqi Huang, Ken Jewell, Andy Kaye, Robert
Knox, Tony Lee, Hailian Liang, Bob Lien, Liping Pang, Kira Lynch, Susan Mravik, Russell Neill, Emile Pitre, Jeff
Powell, Irene Poyer, Randall Ross, David Sabatini, Brad Scroggins, Randy Sillan, John Skender, Richard Smith, Mark
Stacy, Chris Totten, Darby VanAntwerp, Jianlin Wang, Mingyu Wang, and Alex McKown.
Critical and constructive reviews were provided by Murray Einarson (AMEC Geomatrix), Mark Kram (Groundswell
Technologies, Inc.), Scott Huling (USEPA/ORD National Risk Management Laboratory), John McCray (Colorado
School of Mines), Jon Atkinson (Air Force Center for Engineering and the Environment), and Michelle Simon
(USEPA/ORD National Risk Management Laboratory). Pat Bush (NCBA/SEE), Martha Williams and Kathy Tynsky
(SRA International) are hereby acknowledged for their assistance with technical editing and formatting.
-------
-------
Executive Summary
Once released into the subsurface environment, dense nonaqueous phase liquids (DNAPLs) serve as long term sources
of groundwater contamination, and are therefore a significant risk to water resources. Containment or removal, are
the two basic strategies that can be used in the management or restoration, respectively, of a DNAPL contaminated
site. Restoration can be pursued either through in-situ destruction or extraction, and a number of aggressive DNAPL
remediation techniques (e.g., alcohol or surfactant flushing; thermal treatment; air sparging; chemical oxidation) have
been developed for these purposes. Using the techniques, complete DNAPL removal may be possible at some sites, but
because of technology limitations, such as the lack of reliable characterization and assessment approaches, and financial
restrictions, partial mass depletion may be the most likely end result of aggressive source treatment at most DNAPL
contaminated sites. When it is not practical or economically feasible to achieve complete DNAPL mass depletion,
it must be determined if the aggregate benefits of partial DNAPL mass depletion are sufficient to reduce risks to an
acceptable level, and if the costs associated with this partial depletion are justified by the benefits received.
This report summarizes research conducted to address these issues with the primary objective of the research being
the development of a scientifically defensible approach for assessing the long-term environmental impacts (benefits)
of DNAPL removal from source zones. The technical approach used to achieve this objective was to (1) characterize
the relationships between DNAPL mass reduction, contaminant mass flux, and plume behavior, and (2) to use this
information to develop a strategy for assessing the benefits of DNAPL source remediation. As discussed in Section 3,
three lines of investigation were employed in our technical approach: field site demonstrations (Section 3.1), laboratory
experiments (Section 3.2), and numerical and analytical modeling (Section 3.3). Using a combination of these
applications allowed evaluation of varied hydrogeological settings and remediation scenarios. Specifically, field
demonstrations and laboratory experiments were used to investigate the relationships between aggressive source
treatment, mass removal, and flux response (Figure ES-1). Numerical and analytical modeling was likewise used
to investigate specific linkages between the components in Figure ES-1, as well as develop an overall framework
encompassing all conceptual model components.
The research project was primarily funded by the Strategic Environmental Research and Development Program, with
substantial in-kind contributions from the participating organizations. The research team consisted of members from the
Air Force Institute of Technology, Clemson University, the U.S. Environmental Protection Agency, the University of
Florida, and Purdue University.
The Project Conceptual Model and Technical
Approach
Sections 1 and 2 of the report discuss the project
background, motivation and objectives. Two spatial
domains of interest are defined: the source zone, and
the dissolved plume. The source zone is defined as the
groundwater region in which DNAPL is present, either as
randomly distributed sub-zones at residual saturations or
"pools" of DNAPL at higher saturations as accumulated
above capillary barriers. The term source-zone architecture
is used here to refer to both the form and function of the
DNAPL sub-zones as they impact the evolution of the
dissolved plume and the efficiency of remediation. The
shape and extent of the dissolved plume is determined
by mass flux [ML^T1] from the source-zone (which is
a function of the source architecture), by hydrogeologic
heterogeneity, and by natural attenuation processes
occurring within the plume zone.
r
Source
Remediation
Mass ^ ^
Reduction
1
System
Response
Flux
Response
J,
Plume
Response
^
Risk
Response
Figure ES-1. Impacts of Source Remediation:
Conceptual Model.
-------
The linkage between source zone remediation and risk reduction (i.e., benefits) is illustrated by the conceptual
model shown in Figure ES-1. The first component of the conceptual model is the implementation of an aggressive
remediation technology to achieve a rapid reduction in DNAPL mass. This in turn has an impact on the post-
remediation contaminant flux leaving the source zone in response to groundwater flow under natural or induced
gradients. Flux measurements are a critical link between changes in source-zone architecture and the resulting changes
in the contaminant plume. The third component (Plume Response) of the conceptual model is the impacts of source-
zone remediation on the dissolved plume. Because of their relatively large areal extent, dissolved contaminant plumes
typically serve as the dominant exposure pathway and are the main focus for regulatory compliance. Consequently,
changes in the contaminant plume due to source-zone treatment are important in an analysis of source-zone treatment
benefits. Another element of the conceptual model is the System Response, and it accounts for secondary, long-term
benefits in enhancing contaminant degradation within the source and plume due to certain source-zone treatment
technologies. Thus, aggressive source-zone remediation technologies can be designed not only to remove substantial
DNAPL mass, but also to promote contaminant attenuation in the plume.
The final component of the model is risk response. Risk is generally defined as the likelihood of the occurrence of
adverse health or ecological effects as the result of hazard exposure and is characterized by evaluating the extent and
effects of the exposure. In the case of groundwater contamination, risk is typically evaluated in terms of contaminant
concentrations in the groundwater. However, because the dissolved plume is often the dominant exposure pathway,
changes in groundwater concentrations in the source zone resulting from source treatment may not be the best metric
for assessing the risk response. Instead, a mass balance approach can be used to estimate the impact of source-zone
remediation on the dissolved plume. Thus, the endpoint for the risk assessment in our conceptual model is not based
on a reduction of the resident contaminant concentrations in ground water within the entire source zone. Rather, the
analysis is predicated on a reduction in the DNAPL content (total mass or average saturation), and, more importantly,
the alteration in source zone architecture as a result of remediation. This in turn affects the dissolved plume where
risk is defined in terms of two features of the dissolved plume: the spatial extent (defined by length or area or shape),
and the average contaminant concentration or mass. Our metrics for risk reduction are then based on whether (1)
the spatial extent of the existing dissolved plume is stable or decreases; (2) the total contaminant mass within the
plume is constant or diminishing; and (3) contaminant fluxes decrease at succeeding control planes along the plume.
Alternately, concentration-based risk responses at discrete receptors can be estimated. However, these estimates
require detailed site information that is difficult to obtain and seldom available, and will likely be accompanied by
greater uncertainty than mass balance-based estimates.
Field-scale Evaluation of Contaminant Flux Reduction
Changes in contaminant fluxes resulting from aggressive remediation of DNAPL source zones were investigated
at four sites: Hill AFB, Layton, Utah (Section 4.1.2); Fort Lewis Army Base, Tacoma, Washington (Section 4.1.3);
Former Sages Dry Cleaners, Jacksonville, Florida (Section 4.1.4); and Canadian Forces Base, Ontario, Canada
(Section 4.1.5). These sites were selected primarily because they represented a range of hydrogeologic conditions
and applied remedial treatments (see Table ES-1 for remedial treatments). Groundwater and contaminant fluxes were
measured before and after the source-zone treatment, and the measured contaminant fluxes were integrated across
the well transect to estimate contaminant mass discharge (MD; MT1) from the source zone. Results were primarily
used to investigate changes in flux due to DNAPL mass depletion at the field scale. Passive Flux Meters (PFM) and
a variation of the Integral Pumping Test (herein referred to as the modified IPT, or MIPT) were primarily used to
measure fluxes in wells installed along a transect down-gradient of the DNAPL source zones, and perpendicular to
the mean groundwater flow direction (see Section 4.1.1 for flux measurement methods). The MIPT was developed
as part of this study, and the theoretical development is presented in Section 4.1.1, and a method uncertainty
analysis is presented in Section 4.1.6. Uncertainty associated with passive flux meter measurements was completed
independently of this project, and results are discussed elsewhere.
-------
Table ES-1. Summary of field demonstrations to assess DNAPL mass depletion flux reduction relationships
Site
Primary
Contaminant
Treatment
Estimated Mass
Reduction
Flux Method
Pre-Remedial MD
(g/day)
Post-Remedial MD
(g/day)
MD Reduction
HillAFB
TCE
Surfactant
>60%
PFM
76
6.0
92%
MIPT
76
3.9
95%
Fort Lewis NA1
TCE
In Situ Heating
>68%
PFM
743
3.4
99.5%
MIPT
536
2.2
99.6%
CFB Borden
PCE
ISCO
Unknown
PFM
2.2
0.54
75%
MIPT
2.2
2.4
-9%
Sages Dry Cleaner
PCE/DCE
Cosolvent
>100%a
PFM
5.8
2.0
66%
MIPT
1.0
1.1
-10.0%
aThe estimated mass removed was greater than that initially estimated.
A summary of MD measurements is presented in Table ES-1. Pre-remediation MD ranged three orders of magnitude,
from values as low as 1.0 g/day at the Sages site based on MIPT results, to values as high as 743 g/day at the Fort Lewis
site based on PFM results. Significant reductions in MD were observed at Hill AFB and Fort Lewis following source
mass depletion. Results at these sites also indicate similarity between the different flux-measurement methods
employed. However, results from PFM and MIPT as deployed at the Former Sages Dry Cleaner site and CFB Borden
were not as similar. Differences between measurement results were noted, and while the PFM results indicated flux
reductions due to source treatment, the MIPT results did not (Table ES-1). One trait common to all sites was the
observation that the majority of the mass flux occurred over a fraction of the control plane area, which suggests flux
measurements may be useful to improve treatment efficiency by targeting treatment to locations of elevated flux.
Laboratory-scale Evaluation of Contaminant Flux Reduction
The laboratory component of the project (Section 4.2) focused on controlled fluid-displacement experiments conducted
in two-dimensional laboratory aquifer models. With respect to Figure ES-1, laboratory experiments were used to
investigate the interrelation between source treatment, mass depletion and flux reduction. Accounting for spatial
heterogeneity of both media and DNAPL saturation is critical to the study of source-zone removal. Two-dimensional
flow chamber experiments allowed for the emplacement of heterogeneous distributions of porous media and
contaminants, and more closely approached the three-dimensional complexity of the field-scale processes compared
to one-dimensional column studies. Experiments were conducted in glass-sided, two-dimensional flow chambers to
promote characterization through visualization techniques.
Innovative Techniques for Source-Zone Visualization. The architecture of the DNAPL source zone is an important
component in the relationship between DNAPL source mass depletion and flux reduction, and Light Transmission
Visualization (LTV) techniques were used to characterize the DNAPL architecture in the physical aquifer models
(Section 4.2.1). Three general types of systems were explored: undyed water/dyed DNAPL, dyed water/undyed
DNAPL, and undyed water/undyed DNAPL. Since light adsorption characteristics of dyed DNAPL may change as
DNAPL mass is removed through dissolution or volatilization, the latter two approaches offer an advantage over the
first. Moreover, the last system has the added advantage of eliminating concerns associated with dye induced changes
in interfacial tension properties. The LTV methods developed were expansions of earlier LTV methods and took into
account both absorption and refraction light theories. Results indicated that the LTV methods developed were capable
of measuring DNAPL content with minimal calibration procedures.
-------
Relationships between DNAPL source architecture and flux response. Section 4.2.2.1 summarizes laboratory
experiments conducted to investigate the relationship between DNAPL mass depletion and contaminant mass
flux in four heterogeneous model source zones. Flux information was coupled with image analysis to investigate
the relationship between flux and DNAPL architecture. Once emplaced, the DNAPL was removed by aqueous
dissolution, and results indicated that in the systems studied, the relationship between DNAPL mass reduction and
contaminant mass flux was primarily controlled by the DNAPL architecture. For three of the experiments, the
relationship between reductions in mass and flux for the integrated response of the whole flow cell was found to be
approximately 1:1. Responses from individual well segments however deviated from the 1:1 behavior. Results from
the fourth experiment and from a theoretical scenario were used to illustrate cases where mass depletion/flux reduction
would deviate significantly from the 1:1 behavior observed in the other three experiments. Two simplified models,
a streamrube model and an effective Damkohler model, were investigated for their ability to simulate experimental
dissolution dynamics, and both modeling approaches were able to adequately simulate the experimental results.
Another experimental study was conducted to investigate the benefits of partial removal of DNAPL source zones
using enhanced dissolution (Section 4.2.2.4). Four flushing agents with a range of densities were evaluated in eight
controlled laboratory experiments to examine the effects of displacement fluid property contrasts and associated
override and underride on contaminant flux reduction vs. mass reduction relationships. Source zone dissolution
experiments using either a single flushing event or multiple shorter-duration floods were used to deplete DNAPL
mass. All of the single-flushing experiments exhibited similar mass depletion-flux reduction relationships, indicating
that override and underride effects associated with cosolvent densities did not significantly affect the remediation
performance. However, for the flushing solutions with densities different from water, the mass depletion-flux
reduction relationship of the multiple injection experiments tended to be less desirable in the sense that there was
less flux reduction as mass was depleted. UTCHEM simulations supported the observations from the laboratory
experiments and demonstrated the capability of this model to predict mass depletion-flux reduction relationships for
non-uniformly distributed DNAPL sources.
The effects of fluid density and interfacial tension (IFT), and porous media grain size and wettability on contaminant
flux were evaluated in flow chambers in which DNAPL mass was incrementally injected (Section 4.2.2.2).
Contaminant mass flux was found to increase as grain size decreased, and increased slightly as the DNAPL
approached neutral buoyancy. DNAPL spatial distribution and contaminant mass flux were only minimally affected
by IFT and by intermediate-level wettability changes, but were dramatically affected by wettability reversal. The
relationship between DNAPL loading and flux generation became more linear as grain size decreased and density
contrast between fluids decreased. These results imply that capillary flow characteristics of the porous media and fluid
properties will control mass flux generation from source zones.
Reactive tracer tests to predict DNAPL flux responses. Section 4.2.2.3 summarizes reactive tracer tests that were
conducted to evaluate the relationship between contaminant mass depletion and flux reduction. The reduction in
contaminant flux resulting from partial mass removal was obtained from continuous and pulsed cosolvent and
surfactant flushing dissolution tests in laboratory flow chambers packed with heterogeneous porous media. Using
the streamtube concept, a Lagrangian analytical solution was used to predict contaminant dissolution. The analytical
solution was independently parameterized using nonreactive and reactive tracer tests, and the predicted dissolution
was compared to the observed data. Reactive travel time variance, derived from this combination of nonreactive and
reactive tracers, was used to predict the mass depletion/flux reduction relationship. Predictions based on the tracer
tests closely matched measured dissolution data, suggesting that tracers can be used to characterize the DNAPL spatial
distribution, which controls the dissolution behavior.
Modeling Evaluations of DNAPL Source-Zone Remediation
Numerical and analytical models, based on both Eulerian and Lagrangian frameworks, were used to investigate
linkages between the conceptual model components illustrated in Figure ES-1. Numerical models with an Eulerian
framework were used in this project to investigate DNAPL dissolution from source zones, DNAPL mass depletion
due to aggressive treatment techniques, temporal characteristics of mass flux crossing the source control plane,
DNAPL mass depletion/flux reduction relationships, and plume responses to flux reduction brought about by DNAPL
mass depletion. Lagrangian-based analytical models were used to investigate DNAPL dissolution from source
-------
zones and DNAPL mass depletion/flux reduction relationships. Moreover, research was conducted on the field-scale
parameterization of Lagrangian models using tracer tests, and on the development of a holistic source-plume model
using a semi-analytical approach within an Eulerian framework.
Numerical models. DNAPL remediation by air sparging at both laboratory and field scales was examined using
TMVOC (Section 4.3.3.1). Results successfully matched mass removal data for both laboratory and field-scale
experiments. Moreover, a dual-domain mass transfer technique accurately predicted concentration tailing following
pulsed sparging, but simulations using this technique did not match concentration profiles observed during sparging.
Mass transfer rates used to simulate the laboratory experiments were much larger than those used to simulate the field
experiment, suggesting that mass transfer rates measured at a bench scale may not be applicable to field-scale processes.
A drawback noted to this modeling approach for air sparging was the relatively large amount of site parameter data
needed. However, this study showed a moderate amount of site information can still produce reasonable results if data
are available for model calibration.
Section 4.3.3.2. presents simulations using UTCHEM of DNAPL remediation associated with the first cosolvent
flooding pilot test conducted at the former Sages Dry Cleaner site in Jacksonville, Florida. The field simulations were
carried out in three steps. At the first level, a simple uncalibrated two-layered permeability model was developed, which
was capable of roughly reproducing the production well concentrations of alcohol, but not of PCE. At the second level,
a more refined permeability distribution was used, and this model was able to more accurately simulate the recovery of
injected alcohol from the production wells, but was still unable to accurately predict the PCE removal. The final model
used the refined permeability distribution of the second model, but also included a calibration process where observed
PCE breakthrough curves from extraction wells were used to adjust the initial DNAPL distribution. The results of these
simulations suggest that the effectiveness of DNAPL source zone remediation was mainly affected by characteristics of
the spatial heterogeneity of porous media and the variable (and unknown) DNAPL distribution.
Numerical models were also used to investigate changes in mass flux due to DNAPL mass depletion (Section 4.3.3.6).
Using numerical models ISCO3D and T2VOC, changes in the distribution of DNAPL mass within a heterogeneous
aquifer during dissolution, and the resulting changes in the contaminant flux distribution at a source control plane were
investigated. As DNAPL mass was depleted, the mean and standard deviation of the flux distribution at the source
control plane decreased proportionally (i.e., constant coefficient of variation). However, as DNAPL mass was depleted,
the mean of the DNAPL mass distribution decreased faster than the standard deviation (i.e., increasing coefficient of
variation). Also, the spatial moments of the contaminant flux distribution remained essentially constant over time,
suggesting that - for DNAPL source zones cleaned up through in-situ flushing - the contaminant flux distribution
remains stable throughout the DNAPL depletion process while the flux magnitude gradually decreases over time. These
observations appeared to be independent of the permeability heterogeneity, correlation structure, and the DNAPL
dissolution model, for the cases studied. Analysis of flux data from the Hill AFB site also revealed temporal invariance
of the contaminant flux distribution.
Section 4.3.3.3 summarizes a modeling study using T2VOC, conducted to explore the matrix diffusion process in
a fractured aquitard, and assesses the potential impact to an underlying potable water aquifer. It was assumed that
a DNAPL spill has occurred on top of a fractured clay aquitard, and that a small, finite amount of the DNAPL had
penetrated the fractures, but no DNAPL had reached the aquifer itself. Results indicated that DNAPL disappeared
from the fractures relatively quickly due to forward diffusion processes (i.e., from the fracture to the clay). However,
in the absence of degradation in the clay matrix, back diffusion from the clay to the fractures served as a contaminant
source for hundreds to thousands of years based on the size of the fractures. Moderate degradation in the clay resulted
in reduced time scales for the back diffusion source, ranging from tens to hundreds of years. It should be noted that an
estimate of the actual impact of this source to the underlying aquifer and potential downgradient receptors can only be
made in a mass discharge framework. For example, the impact may be negligible if the contaminant mass discharge
from the fractured clay aquitard is small relative to groundwater flow in the aquifer.
Also investigated was the combined use of numerical and analytical models (Section 4.3.2.5). In this approach, an
analytical power law model (PLM) was used to describe the relationship between DNAPL source mass depletion and
flux reduction. This was used as a boundary condition to a numeric transport model (MODFLOW/RT3D) to investigate
the plume response to source-zone DNAPL mass depletion. The model was based on a previous biodegradation
modeling study conducted at a Dover AFB site. Results were used to investigate the extent of source mass depletion
required to reach maximum contaminant level (MCL) goals for TCE and vinyl chloride at the receptor located 800 m
-------
downgradient of the source. When the exponent of the power function (F) was set to 2.0, approximately 92% of the
mass in the source zone needed to be removed to reduce vinyl chloride concentrations to the MCL at the receptor.
However, as F was decreased, the required mass removal to reach MCL goals increased.
Analytical Models in a Lagrangian framework. The use of Lagrangian models to simulate DNAPL source
dissolution, and the flux response to changes in source mass under dissolution processes is discussed in Section 4.3.1.
Lagrangian models describe the source zone as a collection of non-interacting streamtubes, with hydrodynamic and
DNAPL heterogeneity represented by the variations in the travel time and trajectory-integrated DNAPL saturation
within the streamtubes. The properties of the hydrodynamic field and the DNAPL saturation distribution are described
by the mean and standard deviation of the non-reactive travel time and the trajectory integrated DNAPL content, which
are estimated by performing non-reactive and reactive tracer tests. Two techniques were investigated for estimating
the characteristics of the travel time and the trajectory integrated DNAPL saturation distribution based on tracer test
results: the method of moments where measured moments from tracer breakthrough curves are equated with derived
moment equations to calculate the required parameters, and an inverse modeling approach where tracer breakthrough
curves were fitted to a transport model and the parameters estimated by a genetic optimization algorithm.
A DNAPL source depletion model was also formulated based on the Lagrangian concept of the reactive travel time,
defined as the time required to remove DNAPL from a streamtube under a given set of flushing conditions. This
modeling approach was used to estimate contaminant elution curves from pilot-scale in-situ flushing tests, and results
compared well to observed elution behavior. Parameters for the models were derived from tracer tests conducted prior
to flushing. Furthermore, this approach was used to evaluate the relationship between contaminant mass depletion and
contaminant flux reduction. The combined effects of aquifer and NAPL heterogeneities were shown to be captured in
a single parameter, reactive travel time variability, which was determined to be the most important factor controlling
the mass depletion/flux reduction relationship. It was shown that as heterogeneity in aquifer properties and NAPL
spatial distribution increased, less mass depletion was required to achieve a given flux reduction, although the overall
source longevity increased.
Section 4.3.3.7 summarizes additional investigation into the use of non-reactive and reactive tracer tests to
parameterize the streamtube model. The spill and subsequent dissolution of DNAPLs was simulated in two-
dimensional domains having different hydraulic conductivity distributions using the UTCHEM numerical model.
Non-reactive and reactive tracers were used to characterize the Lagrangian properties (travel time and trajectory-
integrated DNAPL content statistics) of the DNAPL source zones, which were in turn shown to be sufficient for
accurate prediction of source dissolution behavior using the streamtube model over the relatively broad range of
hydraulic conductivity distributions tested. The results were found to be relatively insensitive to travel time variability,
suggesting that dissolution could be accurately predicted even if the travel time variance was only coarsely estimated.
Estimation of the streamtube model parameters was also demonstrated using an approximate technique based on
Eulerian data, as may be required in the absence of tracer data.
Analytical Models in an Eulerian framework. In addition to the Lagrangian based approach, three other analytical
DNAPL source depletion models based on an Eulerian framework were evaluated for the prediction of long-term
effects of source depletion under natural gradient flow (Section 4.3.3.5). The spill and subsequent dissolution of
DNAPLs was simulated in domains having different hydraulic conductivity distributions using the UTCHEM model.
The four analytical models (i.e., the equilibrium streamtube model [ESM], the advection dispersion model [ADM], the
PLM, and the Damkohler number model [DaM]) were then fit to the dissolution profiles obtained from the UTCHEM
simulations. All four models, though very different in their conceptualization, included parameters that described the
mean DNAPL mass and the joint variability in the velocity and DNAPL distributions. The variability parameter was
observed to be strongly correlated with the variance of the log hydraulic conductivity field in the ESM and ADM, but
weakly correlated in the PLM and DaM. All four models were able to emulate the characteristics of the dissolution
profiles generated from the complex numerical simulator, but in general the PLM consisting of a single fitting
parameter did not match the UTCHEM results as well as the other models consisting of two fitting parameters.
Nonetheless, the PLM provides a convenient means to develop a screening level model useful in the evaluation
of DNAPL remediation. Analytical solutions were developed for approximating the time-dependent contaminant
discharge from DNAPL source zones undergoing dissolution and other decay processes (Section 4.3.2.1). The source
functions assumed the PLM described the relationship between source mass and contaminant discharge, and accounted
for partial DNAPL source remediation (depletion) at any time after the initial DNAPL release. The empirical exponent
(F) of the PLM is a function of the flow field heterogeneity, DNAPL architecture, and the correlation between them.
The source function was used as a time-dependent boundary condition in an idealized chemical transport model to
-------
develop leading order approximations of the plume response to DNAPL source removal. Moreover, the source function
was also used as a time-dependent flux-boundary condition in a semi-analytical contaminant transport model (Section
4.3.2.2). The semi-analytical transport model considered advection, retardation, three-dimensional dispersion, and
sequential first-order decay/production of several species. A separate solution was used to compute the time-dependent
mass of each contaminant in the plume. This approach allowed for flexible simulation of natural attenuation or
remediation efforts that enhance plume degradation. The plume remediation effort may be applied to limited regions
of the plume, for limited durations, and it may have different chemical effects on different contaminant species in
the decay chain. A series of examples using different values of F illustrated the benefits of partial DNAPL source
remediation for various site conditions. In general, when F >1, relatively large short-term reductions in the plume
concentrations and mass occur, but the source longevity is not strongly affected. Conversely, when F<1, the short-
term reductions in the plume concentrations and mass are smaller, but the source longevity can be greatly reduced. In
addition, the model was used in a Monte Carlo analysis to investigate the impact of uncertainty in site conditions on
remedial performance (Section 4.3.3.4).
It is often difficult at contaminated sites to decide whether remediation efforts should be focused on the contaminant
source, the dissolved plume, or on both zones. The decision process at these sites is hampered by a lack of quantitative
tools for comparing remediation alternatives. Consequently, the source function model coupled with the semi-analytical
model as described in the previous paragraph is viewed as a screening-level mass balance approach to assess the
transient effects of simultaneous DNAPL source and plume remediation. A graphical user interface was developed
for the model, and the complete package, referred to as REMChlor, is publicly available on the internet. Many of the
regulated groundwater contaminants and their degradation products are considered to be known or probable carcinogens
by the U.S. EPA, and cancer risk from exposure to carcinogens was also incorporated into the model. This modeling
approach essentially captures all of the conceptual model elements shown in Figure ES-1, and provides a screening
level protocol with which to assess benefits obtained from DNAPL source remediation.
-------
-------
1.0
Objectives
There is currently no consensus in the academic, technical, and regulatory communities on the ecological
or environmental impacts of dense non-aqueous phase liquid (DNAPL) source-zone treatment. The cost
of source-zone treatment is high, and the anticipated benefits need to be understood before significant
resources are committed to source-zone removal. The primary objective of this project was to develop a
scientifically defensible approach for assessing the long-term environmental benefits of DNAPL removal
from source zones. The technical focus of the project was to: (1) characterize the relationship between
DNAPL mass reduction and contaminant mass flux (used as a measure of source "strength"), and (2) to
use this information to develop a strategy for assessing the benefits of DNAPL source remediation. Project
objectives were accomplished through a comprehensive approach that included coordinated and integrated
field, laboratory, and modeling studies. This approach allowed the evaluation of several hydrogeologic
settings with varied contaminant and remediation scenarios. Specific objectives of the project were:
• Characterize the functional relationship between DNAPL architecture, mass removal and
contaminant mass flux in laboratory aquifer models under well-defined conditions,
• Assess the response to DNAPL mass removal through mass flux and plume behavior at several
field sites,
• Conduct numerical simulations to describe the relationship between DNAPL removal, mass flux
and subsequent plume response for the laboratory and field sites, and
• Compile statistics on the general relationship between partial DNAPL removal and contaminant
flux reduction for simulations of several hydrogeologic templates of actual field sites
2.0
Background
Through field-scale research funded by SERDP and other agencies, it has been demonstrated that a high
percentage (-70 to 90+%) of the NAPL mass can be removed from source zones by implementing several
aggressive in-situ technologies, such as alcohol or surfactant flushing or steam flooding (Wood and Enfield,
1999). These studies and others (Lowe et al., 1999) have shown that while a significant fraction of the
NAPL mass can be efficiently removed in a short period, the efficiency of NAPL mass extraction often
decays exponentially with increasing mass removed. Thus, it may be neither practical nor economically
feasible to remove all of the NAPL mass from the source zone. With finite resources, one needs to know the
maximum amount of NAPL mass that can be left in the source zone while being protective of human health
and the environment.
A reduction in the DNAPL source zone mass is expected to provide the following explicit benefits: a
reduction in human health risks and adverse ecological impacts; a decrease in source and plume longevity;
and lower life-cycle cost of site cleanup (Sale and McWhorter, 2001; EPA, 2003). However, there is
disagreement in the technical and regulatory communities as to whether the aggregate benefits of partial
DNAPL mass removal are sufficient to reduce risks to an acceptable level, and if the costs associated with
partial DNAPL mass removal are justified by the benefits received. Furthermore, there is no consensus on
appropriate risk metrics that should be used to evaluate the derived benefits from partial mass removal from
DNAPL source zones, especially if ecological receptors are included in such evaluations. These types of
considerations provided the motivation for this project, with the following primary research question to be
answered: What are the impacts of partial NAPL source removal in heterogeneous aquifers with respect to
-------
the extent of migration and the longevity of the contaminant plume compared to no source removal? Thus,
the goal was to provide a scientific analysis and decision framework for exposure assessment as an element
of risk-cost-benefit analysis.
Two spatial domains of interest were defined: the source zone, and the dissolved plume (Figure 2.0-1). The
source zone is defined as the groundwater region in which DNAPL is present as a separate phase, either as
randomly distributed sub-zones at residual saturations or "pools" of accumulation above confining units
(Mackay and Cherry, 1989; Feenstra et al., 1996). The DNAPL sub-zones are usually thin, discontinuous
DNAPL accumulations where subtle changes in either texture or permeability occur (e.g., Schwille, 1988;
Feenstra et al., 1996). Thus, a significant volume fraction of the source zone may contain no DNAPL.
Following Sale and McWhorter (2001), the term source-zone architecture is used to collectively refer
to the geometry (shapes, sizes, inter-connections), spatial distribution, and DNAPL content of the sub-
zones. Thus, the term architecture implies both the form and function of the sub-zones as they impact
efficiency of remediation, and the evolution of the dissolved plume. At a given site, the DNAPL source-
zone architecture depends on the following primary factors: (1) the nature of the DNAPL spill scenario
(i.e., locations, rate, volume); (2) the spatial structure of the hydrogeologic properties of the aquifer (i.e,
capillary retention and hydraulic conductivity functions); and (3) the DNAPL fluid properties (i.e., density,
viscosity, wettability). Small DNAPL source zones can generate extensive dissolved plumes (Mackay and
Cherry, 1989). The shape and extent of the dissolved plume are determined by several factors including:
source-zone architecture; hydrogeologic heterogeneity; time since NAPL release; groundwater velocity;
and the magnitude and coupling of several natural attenuation processes (e.g., retardation, dispersion,
reductive dechlorination) occurring within the plume zone. Thus, information regarding the spatial extent
and architecture of the existing (pre-remediation) source zone and the plume characteristics (spatial extent,
concentration distribution, and attenuation processes) will facilitate the design of effective site remediation
strategies.
Dissolved
Plurne
Figure 2.0-1. Plan view of DNAPL source and dissolved plume.
The linkage between source zone remediation and potential system response benefits are considered
according to the conceptual model shown in Figure 2.0-2. The first component of this model is the
implementation of an aggressive remediation technology to achieve a rapid reduction in DNAPL mass
within the source zone, via either in-situ extraction or destruction. The endpoint for our risk assessment
is not based on a reduction of the resident contaminant concentrations in ground water within the entire
source zone. Rather, the analysis is predicated on a reduction in the DNAPL content (total mass or average
saturation), and, more importantly, the alteration in source zone architecture as a result of remediation. The
nature of the expected changes in the source-zone architecture depends on the processes underlying the
remediation technology and how it is implemented at a site.
-------
Source Remediation
Mass Reduction
System Response
f(HS,DAz..-)
Flux Response
Plume Response
^
r
Risk Response
HS = hydrodynamic structure
DA = DNAPL architecture
T = Remedial technology
Figure 2.0-2. Impacts of source remediation: conceptual model.
The second model component relates post-remediation conditions (e.g., modified hydrodynamic contact
with the remaining DNAPL mass) to contaminant flux leaving the source zone in response to groundwater
flow under induced gradients (as in pump-and-treat) or natural gradients. Contaminant flux is defined as
the rate of contaminant mass per unit area leaving the source zone across a control plane orthogonal to
the mean groundwater flow, and may vary widely over the control plane (Figure 2.0-3; Guilbeault, et al.,
2005). Contaminant mass discharge, MD, is the total contaminant mass per unit time from the source and
is a direct measure of the source strength. Feenstra et al. (1996) discussed the use of contaminant flux as a
metric for risk assessment at DNAPL sites. Einarson and Mackay (2001) used this concept more recently to
evaluate the impacts of source zones on down-gradient water-supply wells. They provide a simple approach
for estimating likely contaminant concentrations in down-gradient supply wells given the contaminant flux
from a source within the capture zone, and demonstrate the method's utility at several sites.
The third model component is an analysis of the impacts of source-zone remediation on the dissolved
plume. Risk is defined in terms of two features of the dissolved plume: the spatial extent (defined by
length or area or shape), and the average contaminant concentration. The metrics for risk reduction are
then based on whether (1) the spatial extent of the existing dissolved plume is stable or decreases; (2) the
total contaminant mass within the plume is constant or diminishing; and (3) contaminant fluxes decrease
at succeeding control planes along the plume. The linkage between source-zone and plume assessments
is that to achieve a stable or shrinking plume, the contaminant fluxes leaving the source zone must be
equal to or less than the intrinsic potential within the plume zone to attenuate the contaminants of interest.
In this project, existing protocols were used, as well as, newly developed approaches to estimate the
effects of contaminant mass-flux reductions from partial source removal on changes in plume geometry
and concentration. Source remediation technologies often have secondary long-term impacts within
the source zone and the dissolved plume (System Response in Figure 2.0-2). For example, long-term
monitoring data collected subsequent to in-situ ethanol flushing of the DNAPL source zone at the Sages
site (Jacksonville, FL) indicate that the residual ethanol in the source zone serves as an electron donor in
promoting reductive dechlorination of PCE (Mravik et al., 2000). Field tests conducted at the Hill and
Dover AFB sites show that microbial activity was increased after in-situ alcohol or surfactant flushing.
Thus, aggressive source-zone remediation technologies can be designed not only to remove substantial
DNAPL mass, but also to promote contaminant attenuation in the plume.
-------
The final component of the model is risk response. It is recognized that risk assessment methodologies
should be used to evaluate the benefits accrued from partial DNAPL mass removal in the source zone. It
is possible to estimate the probabilities of contaminant concentrations exceeding some critical value (e.g.,
MCL) in groundwater to evaluate risks to human health, but similar analyses for ecosystem impacts are
much more complex, in part because there are multiple receptors that have different sensitivities to type,
magnitude and duration of exposure to contaminants (Suter et al., 1993). Here, we focus on exposure
assessment as a component of more comprehensive risk assessment.
Control
Plane
Contain i nant
Flux(j)
B'
Most contaminated
Least contaminated
Source Zone
B
Figure 2.0-3. Contaminant mass flux leaving the source zone.
Remedial objectives and site conditions are primary determinants of the need for source treatment or control
at a site. As illustrated in Figure 2.0-1, points of compliance are often located in the dissolved groundwater
plume and the groundwater concentrations at these points (or more generally, compliance planes
considering the entire cross section across the plume) are functions of contaminant discharge from the
source (source strength) and the attenuation capacity within the plume. The integrated behavior of DNAPL
sources and plumes must be evaluated when designing a remediation system (Chapelle and Bradley, 1998;
Chapelle et al., 2004; Widdowson et al., 2005; Basu et al., 2006). The objective of flux-based remedial
design is to reduce contaminant discharge at the down-gradient compliance plane to achieve appropriate
risk reduction (Basu et al., 2006). Thus, the development of a site remediation strategy involves defining
a target receptor loading based on the remedial objective similar to approaches used in surface water
discharge permitting (Kovalic, 1987). For example, if the water at the receptor is to be used for drinking,
the target receptor loading, MDr would be (Einarson and Mackay, 2001):
MDr=(MCL)q0Aw
(2.0-1)
where q0 is the groundwater or Darcy flux and AW is a representative area that is a function of the capture
zone of the supply well. The target loading can be achieved by (1) decreasing qQ, (2) decreasing the source
strength through source treatment, (3) increasing plume degradation rates through plume treatment, or (4) a
combination of source and plume treatment. Source strength can be reduced by removing contaminant
mass from the source. It is very difficult to characterize the magnitude and distribution of the source mass,
but it is easier to characterize the source strength (Basu et al., 2006). It has been shown that source mass
can be related to source flux or source strength (Rao et al., 2001; Rao and Jawitz, 2003; Parker and Park,
2004; Jawitz et al., 2005; Falta et al., 2005a, b; Enfield et al., 2005; Wood et al., 2005) using a power
function of the form:
M
5,0
(2.0-2)
where, MD and MD 0 are the present and initial source strength (mass discharge from the source), Mg
and Mg 0 are present and initial source masses, and F is an empirical constant that is a function of the
heterogeneity of the flow field and the NAPL architecture (Rao et al., 2001; Jawitz et al., 2005; Falta et al.,
2005a,b).
-------
3.0
Materials and Methods
The fundamental approach for this research was an integration of field, laboratory and modeling studies
(Figure 3.0-1). Field data provided the foundation for the project, while laboratory investigations were
used for in-depth evaluation of important processes, and modeling provided the framework for assessment
and application of study results. Laboratory tests were designed to supplement field studies and focused
primarily on the relationships between hydrodynamic structure, DNAPL architecture and contaminant mass
discharge.
2-D Lab Experiments |
(Flushing, Sparging)
•/ DNAPL mass removal
V DNAPL architecture
^ Mass flux
Field Demonstrations
(Dover AFB, Hill AFB, Sages,
Borden CFB, Ft. Lewis)
S DNAPL mass reduction
•S Mass flux response
Plume response
Mathematical Modeling
(Coupled Source & Plume)
S Model testing (lab & field)
•/ Monte Carlo simulations
-------
A.LynnWood,TPOC
Project Management
M. Annable, Co-PI
Field/Lab Investigations
M. Annable, C. Enfield
Groundwater & Contaminant
Flux
A. L. Wood
Soil Coring &
Hydrodynamics
J. Jawitz, A. L. Wood
2-D Flow Experiments
Predominately
SERDP funds
R. Falta, Co-PI
Numerical Investigations
R. Falta
3-D Source Zone Modeling
C. Enfield, S. Rao, J. Jawitz
Lagrangian Modeling
M. Goltz
Plume Modeling
C. Enfield, S. Rao
User Liaison, Technical
Outputs&Project Integration
Predominately Non
SERDP funds
Figure 3.0-2. Project organizational chart.
3.1 Data Quality
The following quality assurance project plans (QAPPs) were prepared to encompass the laboratory,
field, and numerical investigations that comprised the research project described in this report: Impacts
of DNAPL Source Zone Treatment: Laboratory and Field Assessment of the Benefits of Partial Source
Removal; Modeling Assessment of the Benefits of Partial DNAPL Source Removal; and Benefits of Partial
Source Removal: An Experimental Assessment. These plans detail data quality assurance objectives,
training, documentation and records, sampling, chain of custody, analytical methods requirements, quality
control, instrumentation, calibration and frequency, inspection, data management, assessment and oversight,
and data validation and usability. The QAPPs were reviewed and approved by EPA's quality management
staff.
The data contained in this report were collected in accordance with provisions described in the quality
assurance project plans. The quality of the data has been evaluated and determined to be sufficient for
the intended applications unless otherwise noted in the report. A variety of experimental and analytical
procedures were used in this project. These procedures are briefly described in Section 4 of this report.
Additional information can be found in the peer-reviewed project publications that are cited throughout
Section 4.
-------
3.2
Field Studies
DNAPL source treatment effectiveness was evaluated at four sites: Hill Air Force Base Operable
Unit 2, Layton, Utah; Fort Lewis East Gate Disposal Yard Site (EGDY), Tacoma, Washington; Borden
Forces Base, Borden, Ontario, Canada; and Sages Dry Cleaner, Jacksonville, Florida (see Table 3.1-1).
Contaminant discharge from the DNAPL source zone (source strength) was measured at each site before
and after DNAPL source treatment. These data were collected using two innovative techniques: passive
flux meters (PFM) and a modified version of the integrated pumping technique (MIPT). In addition,
multiple-level sampler data were collected at the Sages site and these data along with independent estimates
of groundwater flow velocities provided a third estimate of contaminant discharge. At each site, source
strength was measured along a control plane established by installing a transect of wells immediately
downgradient of the source and approximately perpendicular to the mean groundwater flow direction as
determined from hydraulic head data.
The field sites represented a range of hydrogeological conditions and remediation technologies. Surfactant
flooding was used to remove DNAPL mass at the Hill AFB site; thermal resistive heating was the
technology of choice at the Fort Lewis EGDY; source mass reduction was accomplished by flushing the
Sages site with ethanol; and in-situ oxidation was selected at the Borden site. Details of site conditions,
contaminant properties and measurement techniques are provided in the project manuscripts that are
identified and summarized in Section 4 of this report.
Table 3.1-1. Field study sites.
Field Site
Hill AFB OU2
Fort Lewis
Borden
Sages
Location
Hill AFB, Utah
Tacoma, Washington
Canadian Forces Base,
Ontario, Canada
Sages Dry Cleaner,
Jacksonville, Florida
Primary
Contaminants
TCE
TCE
PCE
PCE, DCE
Initial Mass
Estimate
1360 liters
2600 to 8200
liters
5 liters
19 liters
Treatment Technology
Surfactant-Enhanced Aquifer
Remediation
Resistive Heating
In-situ Chemical Oxidation
(permanganate)
Cosolvent Flood
(ethanol)
3.3 Laboratory Studies
Laboratory experiments focused on controlled fluid-displacement experiments conducted in two-
dimensional model heterogeneous aquifers. The goals of the experiments were to: (1) quantify the reduction
in contaminant flux due to partial removal of DNAPL source zones, and (2) to provide high quality data
sets for validating the deterministic process models and the stochastic stream tube models. The work
included the development and application of methods for characterizing the source-zone architecture
using partitioning tracer tests, and the application of innovative visualization techniques to observe source
distribution and alterations.
Two types of extraction technologies were evaluated: in-situ flushing (surfactant or cosolvent), and
volatilization (air sparging). DNAPL was emplaced within a range of model hydrogeologic settings and
down-gradient contaminant concentrations and mass fluxes were measured both before and after source
depletion. Experiments were conducted in two-dimensional flow chambers with tempered-glass side
walls to facilitate the evaluation of media structure, contaminant distribution, fluid displacement and
DNAPL removal dynamics using visualization techniques. Details of the flow chambers, measurement and
visualization techniques, and experimental procedures are described in Section 4.
-------
3.4 Modeling Studies
The basic modeling approach was to couple source-zone models with reactive solute transport models to
estimate the benefits of contaminant source remediation (Figure 3.3-1). Three basic source zone modeling
approaches were used for the project.
DNAPL Source Zone
Remediation Models
(Numerical, Streamtube,
Power Function)
Groundwater Plume Models with Full
Biogeochemical Reactions
(RT3D, Biochlor, REMChlor)
Couple Models
At the Edge of the
Source Zone to
Provide
Contaminant
Flux to Plume
Models
Figure 3.3-1. Coupled modeling approach used in the project.
The first was to apply existing deterministic codes, such as T2VOC and UTCHEM, to the DNAPL source
zone to simulate the remediation processes during remediation and to predict the temporal and spatial
distribution of contaminant fluxes leaving the source zone. The second approach utilized the Lagrangian
Stochastic Advective-Reactive (LaSAR) streamtube technique to estimate mass flux across a control plane.
With this method, the complexities of the medium are represented by the statistics and spatial correlation
structure for the major parameters and the contaminant flux at the control plane is the ensemble average
of all the stream tubes that contribute to the sampled flow. In the third approach, contaminant discharge
from the source was estimated by a power function that quantifies the relationship between source mass and
mass discharge. In this third approach a power function was used to describe the changes in source strength
as a function of source mass reduction. Output from each of these approaches provided input boundary
fluxes for dissolved-plume transport simulations. Modeling approaches and applications are described in
Section 4.
-------
4.0
Results and Accomplishments
Results of the research conducted through this project are described in a variety of products. A complete
list of these products is presented in Appendix A. The following sections of the report provide an overview
of the detailed information contained in these products.
4.1 Field Research
A primary objective of this project was to evaluate changes in contaminant flux and discharge resulting
from source zone treatment. An extensive search was conducted to locate field sites with delineated
DNAPL source zone and where aggressive source treatment was planned during the first two years of
the project. Four field sites were selected: 1) Hill Air Force Base, UT, 2) Ft. Lewis, WA, 3) Sages Dry
Cleaner site in Jacksonville, FL, and 4) Canadian Forces Base, Borden, ON, Canada. At each of these sites,
contaminant flux and discharge were measured down-gradient of the source zone both before and after
source remediation. Flux was measured using both passive flux meters and integral pump tests. Results of
field-based experiments conducted as part of this project are described in detail in the following reports and
journal articles:
Brooks, M.C., A.L. Wood, M.D. Annable, K. Hatfield, J. Cho, C. Holbert, P.S.C. Rao, C.G. Enfield,
K. Lynch, and R.E. Smith (2008). Changes in contaminant mass discharge from DNAPL source
mass depletion: evaluation at two field sites. Journal Contaminant Hydrology 102(1-2): 140-153,
doi:10.1016/j.jconhyd.2008.05.008.
Brown, G. (2006). Using Multilevel Samplers to Assess Ethanol Flushing and Enhanced Bioremediation at
Former Sages Drycleaners. M.S. Thesis, University of Florida.
Goltz, M.N., J. Huang, M.E. Close, M.J. Flintoft, and L.Pang (2008). Use of tandem circulation wells to
measure hydraulic conductivity without groundwater extraction. Journal of Contaminant Hydrology
100(3-4): 127-136.
4.1.1 Flux Measurement Methods
Passive Flux Meters (PFM)
General experimental methods for deployment and recovery of PFMs are described by Annable et al. (2005)
and Basu et al. (2006). Briefly, the approach to the measurement of groundwater and contaminant fluxes
involves deployment of a permeable, sorbent pack (i.e., PFM) in the wells located along a transect of wells
that are screened across the aquifer depth of interest. The PFM sorbent (e.g., silver-impregnated granular
activated carbon [SI-GAC]) is pre-loaded with known amounts of tracers and these tracers are displaced
from the sorbent at rates proportional to groundwater flux and the tracer retardation on the sorbent. The
PFM sorbent material is selected to capture the target contaminants dissolved in the groundwater flowing
through the device during the designated deployment period. Because these measurements are made under
natural gradient groundwater flow conditions, this approach is referred to as "passive" in contrast to the
MIPT method (see below) that requires pumping.
The PFMs may be packed on site and are constructed to match the saturated thickness in each well, and
multiple PFMs may be deployed as needed in wells to cover well-screen intervals. Each PFM sock can be
divided into segments separated by Norprene rubber washers to prevent vertical water flow in the PFM.
A sample of the sorbent is collected for analysis of the initial tracer concentrations (analytical details are
provided in Annable et al. (2005)). The completed PFM is inserted into the well to the desired screen
interval.
-------
Estimations of groundwater flux based on depletion of alcohol tracers, and contaminant fluxes based on
contaminant accumulation, have been described by Hatfield et al. (2004) and Annable et al. (2005). After
a specified period of exposure to groundwater flow, the PFM is removed from the well; the sorbent is
carefully extracted to quantify the mass of contaminants intercepted and the residual masses of all resident
tracers. Groundwater fluxes (seepage velocities) are calculated from the tracer mass depletion and sorbed
contaminant mass is used to directly estimate contaminant fluxes. Retardation factors for tracer depletion
from the PFM sorbent were reported by Annable et al. (2005). The depth profile of contaminant fluxes from
each well is averaged to determine J for each well. These averages are calculated by weighting the local
contaminant flux values (J) by the length of the vertical interval (b) as follows:
(4.1-D
where n-vert is the number of intervals in the well screen, b. is the length of each interval [L], and J. is
the local contaminant flux [ML^T"1]. The J values are integrated over the width of the control plane to
determine a transect-wide contaminant mass discharge, M D (g/day):
\
(4-1-2)
where n-well is the number of wells and A is the area [L2] represented by the vertical sampling interval (b
times the horizontal spacing between wells). Because this calculation is predicated on the assumption that
contaminant fluxes measured in the wells are representative of the entire well spacing, reducing this spacing
should improve mass discharge estimates. In addition, techniques used to interpolate between wells can
impact these estimates.
Integral Pumping Test (IPT)
The IPT technique to measure J and MD is primarily based on the measurement of the contaminant
concentration-time series, C( t ), in the effluent of multiple pumping wells aligned perpendicular to the
prevailing direction of groundwater flow. While the use of C( t ) data for groundwater investigations was
introduced in the early 1980's (Keely, 1982; Keely and Wolf, 1983), estimation of Jar\dMD using the
IPT method was first described by Schwarz et al. (1998), and IPT field applications were described by
Bockelmann et al. (2001, 2003) and Bauer et al. (2004). Under steady-state conditions, the mass discharge
entering the capture zone of the pumping wells is equal to the product of the pumping rate and pumping
well effluent concentration (Holder et al., 1998; Einarson and Mackay, 2001). However, achieving this
steady-state condition may not be a practical goal for short-term pumping tests. Alternatively, the transient
concentration-time series from a pumping well may be used to estimate the average spatial concentration in
the capture zone of the well (e.g., Bayer-Raich et al., 2004, 2006), which is combined with an estimate of
the groundwater flux to give the contaminant flux.
For the MIPTs conducted as part of this study, all wells were pumped concurrently rather than sequentially
as conducted in previous deployments of the IPT method. Wells were pumped concurrently to minimize
the disturbance to the contaminant distribution that would result from pumping each well sequentially, and
to avoid the uncertainty associated with double counting mass located between adjacent wells. However,
the disadvantage to concurrent pumping is both the development of stagnation points between wells, and
a more complex flow field which precludes the use of the Bayer-Raich et al. (2004) analytical solution.
With respect to the former disadvantage, it is nonetheless possible to sample a large fraction of the space
between wells, so that the general goal of the IPT (i.e., a measured response integrated over a large area)
is maintained. With respect to the latter disadvantage, the average spatial concentration in the capture
zone of the well was estimated using the average concentration of the concentration-time series. The mass
discharge was then obtained from the product of the contaminant flux and the cross-sectional flow area
associated with the wells in the transect.
-------
Using an average of the concentration-time series is, however, only an approximation that should be
evaluated on a case by case basis. For transient pumping tests, the only statement that can be made without
simplifying assumptions is that the mass recovered in the pumping well effluent over a given duration
is equal to the mass in the capture zone of the well corresponding to the pumping duration and pumping
rate. Beyond this, simplifying assumptions are needed if the concentration-time series from the pumping
well effluent is to be used to estimate an average spatial concentration crossing the control plane around
the pumping well. Following Bayer-Raich et al (2004), two degrees of spatial freedom are eliminated
by assuming the concentration in the pumping well effluent is depth-averaged, and that changes in
concentration over the length of the capture zone parallel to the direction of groundwater can be neglected.
Therefore, concentration is reduced to a one-dimensional function that varies only in the spatial dimension
perpendicular to groundwater flow. A closed form analytical solution was presented in Bayer-Raich et al
(2004) for the condition of small dimensionless pumping times (see Bear and Jacob (1965) for a definition
of dimensionless pumping time), and uses the analytical travel-time solution provided by Bear and Jacob
(1965) for a single pumping well in a uniform flow field. This solution was extended to the case of an
aquifer with linear retardation by Bayer-Raich et al. (2006). Since there is no general solution to travel time
in a uniform flow field with multiple pumping wells, however, a numerical technique would be required to
calculate the travel time distribution for each well and subsequently estimate the average one-dimensional
concentration. However, if the concentration-time series is uniform or varies mildly, then the spatial
concentration in the capture zone is uniform (or approximately so), and an average of the concentration-
time series provides a reasonable estimate of the average one-dimensional concentration distribution at the
control plane. With a few exceptions, the concentration-time series information obtained during the MIPTs
conducted in this project were relatively uniform (for example, see the discussion in Brooks et al 2008 for
the Hill AFB and Fort Lewis sites). When the concentration-time series changes significantly with time,
then an average is less accurate. Work is currently being conducted to quantify the inaccuracy of a straight
average when the concentration-time series varies significantly with time.
The IPT method was further modified to allow the estimation of groundwater fluxes based on the hydraulic
information collected during the MIPT The basis for this new approach was the assumption that the
aquifer in the vicinity of the well transect could be described as one with homogeneous and isotropic
hydraulic conductivity, and a uniform saturated thickness. Consequently, the hydraulic head (h; L) can be
described using superposition of uniform flow and sink terms (Bear, 1979):
(4.1-3)
where; B is the saturated thickness [L]; T is the aquifer transmissivity [I/T1], equal to the product of
hydraulic conductivity K [ LT-1] and B; and Q is the pumping rate [ L3T-1] for a well located at xw,yw.
Assuming the origin of the coordinate system coincides with a pumping well of interest and that the well
transect is aligned perpendicular to the groundwater flow direction, the difference in head between the
pumping well and a down-gradient monitoring well is:
o
where the summation term accounts for concurrent pumping from n other wells in the transect, and r
and r are the distances from the observation well to the Ith pumping well and from the pumping well
of interest to the other pumping wells, respectively. Note that r for the pumping well of interest itself
refers to the well radius. Equation (4.1-4) is linear with a slope of (4;tT)"1 and an intercept of-qBAxT'1.
By measuring the difference in hydraulic head under a series of pumping rates, the hydraulic conductivity
and the groundwater flux can be estimated using linear regression techniques and an a priori estimate of
the saturated thickness. Equation (4.1-4) is based on the assumption that the ambient flow direction is
perpendicular to the well transect. If this condition is not met, equation (4.1-3) can be modified to account
for non-perpendicular flow (Brooks, et al, 2008).
-------
For this project, each of the MIPTs was conducted immediately following the retrieval of the PFMs from
the wells, thus allowing for a comparison of measured fluxes with minimum time lag. During the MIPT,
flow rates and water levels are measured periodically at the wells and the combined flow is recorded as it
accumulates in a temporary storage tank. Water levels are also monitored in nearby observation wells. The
pumping rates are selected to bracket the calculated specific discharge rate across the control plane area,
while still yielding a sufficiently large volume of pumped water (and hence sampling space surrounding the
well) over the duration of the test.
Groundwater samples from the effluent of each pumping well are collected for contaminant analysis
approximately once every three to six hours throughout each MIPT. Once collected, the samples are
immediately placed in coolers with frozen ice packs, and shipped overnight to the laboratory for refrigerated
storage and analysis.
Transect Method
The "Transect Method" (TM), as described in API (2003)2_involves estimating contaminant discharge
per well as the product of the average groundwater flux (~q , where the double over bar indicates a spatial
average over a region larger than the well), the well cross sectional area (^4), and the flux-averaged
contaminant concentration in the well (C , where the subscript J indicates a flux-averaged value, and
the over bar indicates a spatial scale associated with the well). In turn, q is estimated as the product of
the hydraulic gradient (7) and .saturated hydraulic conductivity (f£). Measured hydraulic gradients and
previously reported values of f£ were used to estimate ~q whereas contaminant concentrations measured
during the MIPTs were used to estimate C Note that the first value of the concentration-time series
measured during the MIPTs most closely represents the concentration measured by traditional groundwater
sampling (e.g., bailing or pumping), and was therefore used in this approach.
4.1.2 Hill Air Force Base Operable Unit 2
From 1967 to 1975, spent degreasing solvents, primarily TCE, were disposed into two unlined disposal pits
dug into the sandy surface soil located along the northeast boundary of Hill AFB, near Layton, UT As a
result, the DNAPL contaminated the underlying alluvium (the Provo Formation), predominantly as a mobile
phase, pooled in topographic depressions on top of a thick clay aquiclude (the Alpine Formation), but also
as an immobile or "residual" phase retained as ganglia by capillary forces in the alluvium's pore spaces.
Under the Installation Restoration Program that began at Hill AFB in 1981, significant work has been
completed to characterize, remove, and contain the site contamination; the site is now called Operational
Unit number 2 (OU2).
In 1996, a containment wall (Figure 4.1-1) was constructed around the known source prior to initiation of
several actions for enhanced DNAPL mass extraction within the contained zone. Additional DNAPL was
discovered outside of the containment well in a depression in the clay surface (see Figure 4.1-1) in 1997
during efforts to quantify the mass of DNAPL remaining in the formation. This area, referred to as the
Panel 5 area, was the focus of the present study (Brooks et al., 2008). The paleo-channel aquifer at OU2
consists of the shallow, permeable Provo Formation underlain by the thick, clay and silt Alpine Formation.
The hydrogeology and DNAPL distribution within the containment wall have been described by Meinardus
et al. (2002).
-------
1870580 1970600 18796180 1*70*40 1870660 1S7W-BO 18707W 1«7073O
2»J2OO
JU2-149
JU2-151
AU2-153
AU2-155
JU2-157
1870580 MMMB 1870620 187O440 1870660 1S706W 1S707M 1070730
Figure 4.1-1. Plan view of Panel 5 area at Hill AFB. The thick black line in the lower left corner
represents the containment wall installed around OU2. The triangular symbols represent
wells used for mass flux measurements. Contour lines denote the elevation of the top of the
Alpine Formation (Brooks et al., 2008).
To evaluate groundwater and contaminant fluxes, a control-plane transect often monitoring wells (wells
U2-116 and U2-148 through U2-157) was installed in an approximate north-south direction, down-gradient
from Panel 5 and just up-gradient from a natural spillway above the Weber River Valley (Figure 4.1-1). In
October of 2001, with remediation of the DNAPL zone outside of the containment wall planned for the
following summer, pumping used to contain the source area was discontinued. This allowed groundwater
to flow over the sill area and through the transect of flux wells. A hydrograph from wells U2-214 and
U2-117 is shown in Figure 4.1-2, and indicates the water elevation relative to the average clay elevation
across the sill, as well as the water elevation during the remediation and flux measurement events.
In May 2002, groundwater and contaminant fluxes were measured, prior to the surfactant flood, using both
innovative techniques described above. The PFM method allowed characterization of depth variations in
local groundwater and contaminant fluxes along the well screen, while the MIPT provided flux values that
were depth-averaged over the well-screen interval. The mass discharge estimates from PFM and MIPT
were also compared with those from the Transect Method. The PFM measurements were performed prior
to disturbing the groundwater flow field with the pumping tests. Surfactant-enhanced aquifer remediation
was performed by URS and INTERA using the wells shown in Figure 4.1-1 (wells to the northeast of
the containment wall) during June and July 2002 (URS and INTERA, 2003). Based on the results from
a partitioning tracer test conducted in Panel 5, the pre-remedial estimate of DNAPL mass was 2,180 kg
(360 gallons). Estimates of the DNAPL mass removed by the surfactant flood ranged from 1,340 kg
(221 gallons), based on liquid-phase separation in the effluent treatment system, to 2,250 kg (371 gallons),
based on effluent TCE breakthrough curves from the extraction wells.
-------
1422 -
1421 -
1420
1419
Jul-01
Jan-02
Jul-02
Jan-03
Jul-03
Jan-04
Jul-04
Date
Additional, post-remedial 1424 i
activities continued until February
2003, after which pumping
stopped and groundwater levels 1423"
increased, allowing groundwater
to flow over the spillway (see
Figure 4.1-2). Post-remedial
characterizations indicated some
DNAPL remained in the Panel 5
source zone. Post-remedial flux
measurements were conducted
at the transect in June 2003
and again in October 2004.
Groundwater levels in the transect
wells during the first post-remedial
flux measurements were similar
to those during pre-remedial
measurements (Figure 4.1-2).
Prior to the second, post-
remediation flux measurement,
however, groundwater levels rose
approximately 107 cm (3.5 feet)
above those during the first two
measurements, presumably in
response to above-average storm events that produced twice the amount of rainfall between December 2003
and February 2004 compared to the same period in the previous two years
Contaminant fluxes were measured before and twice after the surfactant flood by collecting 63, 48 and
106 SI-GAC samples from PFM deployments during each of the events. The TCE contaminant flux profiles
from selected wells with the largest fluxes measured prior to remediation are provided in Figure 4.1-3.
The contaminant flux profiles generally indicate that the higher contaminant flux zones were located near
the aquifer clay interface prior to remediation. The primary contaminant mass flux is located between
wells 152 and 149, with minimal flux outside this region. Following the surfactant flood, TCE levels were
reduced significantly with only minimal local hot spots remaining in the transect (further details can be
found in Brooks et al 2008).
Figure 4.1-2. Hydrographs from Hill AFB OU2 wells U2-214
and U2-117, along with the average clay elevation
(dashed line) across the transect. Also shown are
selected events relative to the remedial and flux
measurement activities.
1422.7"
Ik
g 1421.7
1
W
i
;
-",
N
x^
) J--
s
^*
y
i
\
>
>
1
;
V
;
^ i
04 8 12 16 04 8 12 16 0 4 8 12 16 0 4 8 12 16 0 4 8 12 16
Mass Flux (g/m2/day)
Figure 4.1-3. Contaminant flux distribution from selected wells with the largest fluxes as measured prior
to remediation at Hill AFB (Brooks et al., 2008).
-------
The depth profiles of contaminant fluxes from each well were averaged to determine a mean contaminant
flux value for each well (Equation 4.1-1). These depth-averaged values are plotted in Figure 4.1-4 and can
be later compared to the average contaminant flux for each well estimate based on the MIPTs. The data
can also be integrated over the width of the control plane to determine a transect-wide contaminant mass
discharge MD (g/day):
n—well f n—vert
= 22M.,)
(4.1-5)
where n-well is the number of wells and A. is the area represented by the vertical sampling interval, b.,
and the horizontal spacing between wells. TCE mass discharge before source treatment was 76 g/day
based on both PFM and MIPT methods (Table 4.1-1). Measurements one and two years after the source
remediation ranged from 6.0 to 2.4 g/day (first and second measurements, respectively) based on the
PFM approach, and 3.9 and 7.1 g/day based the MIPT method. Mass discharge, calculated using the TM
approach, was 78, 7.2, and 1.3 g/day, for the pre-remedial, first and second post-remedial tests, respectively.
These discharge measurement techniques provided comparable results, and show significant reduction
(>90%) in TCE contaminant mass fluxes as a result of source treatment.
24
20
16-
12-
(a)
fill]
I
eb
154 152 150 148 116 149 151 153 155 157
(b)
154 152 150 148 116 149 151 153 155 157
2.5
2.0
1.5
1.04-
0.5-1-
0.0
(c)
154 152 150 148 116 149 151 153 155 157
24
20
16
12
(d)
I
1
I
I
1.8
1.5
1.2
0.9
0.6
0.3
0.0
154 152 150 148 116 149 151 153 155 157
(e)
154 152 150 148 116 149 151 153 155 157
154 152 150 148 116 149 151 153 155 157
Figure 4.1-4. Average mass discharge (g/day) for each well at Hill AFB OU2, as measured by PFMs for
a) May 2002 (pre-remedial), b) June 2003 (first post-remedial), and c) October 2004 (second
post-remedial) tests; and as measured by MIPTs for d) May 2002 (pre-remedial), June 2003
(first post-remedial), and f) October 2004 (second post-remedial). TCE is shown in black,
cis-DCE is shown in white, and MIBK is shown in grey. Note the change in scale on the
y-axis to accommodate the data. In figures c) and f), the MIBK mass discharge is plotted on
the second y-axis.
-------
Table 4.1-1. Summary of TCE and DCE mass discharge (g/day) and combined TCE and DCE molar
discharge (moles/day) for Hill AFB OU2 using PFM and MIPT results, as well as estimates
based on the Transect Method (TM).
Discharge
TCE
(g/day)
DCE1
(g/day)
TCE & DCE
(mole/day)
Method
PFM
MIPT
TM
PFM
MIPT
TM
PFM
MIPT
TM
Pre-Remediation
Test
May 2002
76
76
78
a
b
b
0.58
0.58
0.59
First Post-
Remediation Test
Jun 2003
6.0
3.9
7.2
3.0
2.0
3.8
0.077
0.051
0.094
Second Post-
Remediation Test
Oct 2004
2.4
7.1
1.3
5.8
20
4.2
0.078
0.26
0.054
'Based on results from selected samples, DCE concentrations were an order of magnitude less than TCE concentrations during tests prior to source
treatment. Therefore, DCE was not analyzed during the pre-remediation flux tests. More information is provided in Brooks et al. (2008).
a - DCE concentrations were below the level of quantification.
b — DCE was not included in all sample analysis.
Under steady-state conditions, the contaminant mass discharge [M/T] entering the capture zone of a
well during the MIPT is the product of the pumping rate [L3/T] and contaminant concentration [M/L3].
Achieving such steady-state conditions most likely did not occur within the test durations, consequently, the
contaminant mass discharge was calculated as the product of the contaminant flux and the cross-sectional
area assigned to each well. The cross-sectional area for each well was based on the well spacing (305 cm)
and the depth between the water table and the clay interface. The contaminant flux was calculated as
the product of the Darcy flux (the transect-wide average Darcy flux based on the MIPT hydraulic data
as described in Section 4.1.1) and the average concentration of the concentration-time series. The mass
discharge measured for each well in each test is shown in Figure 4.1-4. The mass discharge estimates based
on the MIPT tests are similar to those measured by the PFM method with exceptions noted for the DCE
mass discharge and combined TCE/DCE molar discharge measured during the second post-remediation test
(Table 4.1-1). The difference in estimates is primarily attributed to the difference in estimated Darcy flux
for the two methods, which differed by a factor of 2.
Results from the post-remediation tests indicate a significant increase in the concentration of DCE relative
to the pre-remediation DCE concentrations (Figure 4.1-4). Since DCE was not present in significant
concentrations prior to remediation, these results indicate that the surfactant used for in-situ flushing may
have contributed to the reductive dechlorination of TCE, as suggested by the presence of Methyl Isobutyl
Ketone (MIBK) in the second post-remedial test (Figure 4.1-4). The contaminant flux of TCE and DCE
can be combined on a molar basis as shown in Table 4.1-1. The total pre-remediation molar flux was
0.58 to 0.59 moles/day, and with the exception of the second post-remediation results from the MIPT, the
post-remediation total mass discharge (molar basis) TCE and DCE ranged from 0.051 to 0.094 moles/day.
Based on these measures, total molar discharge reductions ranged from 84 to 91%. If groundwater flow
continues as it has in the past, the contaminant discharge will likely continue to decline for the next several
years before new, quasi steady-state conditions are achieved. At this site the wells were relatively close
together (-300 cm spacing). Concentrations of TCE observed at the well remained relatively constant for
all of the pumping rates used. This suggests that it was not possible using the MIPT method to delineate
spatial variability within the well spacing using the flow steps implemented during these measurements.
Considering the well spacing, finer delineation may not be necessary.
-------
During remedial activities within Panel 5 at the OU2 site, the surfactant flood resulted in estimated DNAPL
mass removal ranging from 1,340 kg (221 gallons) to 2,250 kg (371 gallons) (URS and INTERA, 2003).
Based on the estimated DNAPL mass prior to the remedial activity [2,180 kg (360 gallons) based on a
partitioning tracer test], 60% or more of the DNAPL was removed. Free phase DNAPL has since been
detected in wells within Panel 5, indicating some DNAPL was not removed. Based on the PFM and MIPT
contaminant flux measurements reported here, TCE flux reductions ranged from 92% to 97%, based on the
PFM and MIPT measurements. Consequently, at this site partial DNAPL mass removal from the source
did result in a reduction in the contaminant mass flux. It should be noted that, as shown by groundwater
elevation data in Figure 4.1-2, groundwater flowed over the spillway for approximately four months
prior to the first post-remediation flux measurements. This is insufficient time to establish steady-state
contaminant transport conditions from the source area to the well transect. However, in light of the active
pumping activity that occurred in the Panel 5 source area, the flux measurements cannot be viewed as those
originating from a stagnant pool of water in contact with the DNAPL. Moreover, the first post-remediation
measurements were conducted under very similar conditions to the pre-remediation measurements with
respect to the duration in which groundwater flowed through the transect prior to the flux measurements.
4.1.3 Fort Lewis East Gate Disposal Yard Site
The EGDY site, located on the Fort Lewis Military Reservation near Tacoma, Washington, was used
from 1946 to 1960 as a disposal area for drums of used solvents and oils that were placed in excavated
trenches, and is the source for a large chlorinated solvent plume (predominantly TCE) which extends to
the northwest for approximately 2.5 miles (USAGE, 2002). The plume was discovered in the mid 1980's,
and the primary source was identified as the EGDY site. The operation of a pump-and-treat system was
started in 1995 for hydraulic control purposes, and drum excavation activities at the EGDY site were
conducted between late 2000 and mid 2001. In 2001, thermal treatment was selected as a source removal
method, and subsequent characterization work conducted in 2001 and 2002 identified three main areas of
DNAPL contamination within the EGDY site. These areas were referred to as NAPL Areas 1, 2, and 3
(NA1, NA2, and NA3, respectively). The focus of work reported here was on the performance assessment
associated with remedial activity conducted in NA1 (Brooks et al., 2008). While the source zone contained
other petroleum-related contaminants, our focus herein is limited to an assessment of the changes in mass
discharge of chlorinated solvents.
The surficial, unconfined aquifer at the site is composed of the Vashon Recessional Outwash/Steilacoom
gravel unit (consisting of loose, well-graded sandy, cobbly gravel or gravelly sand). In the immediate
vicinity of NA1, this unit is underlain by Vashon Till (consisting of loose to dense silty, sandy gravel with
some clay), which is considered to be a generally continuous intermediate aquitard. This layer in turn is
underlain by more Vashon Recessional Outwash/Steilacoom gravel or Vashon Advance Outwash (loose
sandy gravel to gravelly sand with cobbles). The contamination is predominantly located in the surficial
unconfined aquifer, although contamination has migrated into the aquifer below the Vashon Till layer
(USAGE, 2002). Regional groundwater flow is directed to the northwest towards American Lake, as
indicated by the orientation of the TCE contaminant plume.
A control-plane transect often monitoring wells (LC-201 through LC-207, LC-211, LC-212 and LC-213)
was installed in the surficial aquifer immediately downgradient of the source area (Figure 4.1-5) to
measure groundwater and contaminant discharge from NA1. In November of 2003 the groundwater and
contaminant discharge were measured prior to thermal treatment using PFMs and MIPTs. In-situ resistive
heating thermal treatment was conducted by Thermal Remediation Services, Inc in NA1 from December
2003 until August 2004 (Beyke and Fleming, 2005; TRS, 2005; Powell et al., 2007). To ensure pre- and
post-remedial mass discharge measurements were comparable, the post-remedial measurements were
delayed until the groundwater temperatures had returned to ambient levels. Moreover, treatment of NA2
was initiated shortly after NA1 treatment, consequently NA1 post remedial discharge measurements were
delayed until the completion of NA2 treatment (which occurred in August, 2005) because of the proximity
of the flux well transect to NA2. Consequently, post-remedial discharge measurements for NA1 were
collected in June 2006.
-------
Shallow transect wells LC-201 though LC-207 (Figure 4.1-5) were installed at a nominal spacing of 6.1 m
(20 feet) between August 23 and 25, 2003 using a rotosonic drilling method. Additional information on the
design and installation of these wells are presented by Brooks et al., (2008). Initial samples collected from
these wells indicated relatively high concentrations in LC-207 (data not shown), consequently the transect
was extended to the north by three more wells (LC-211, LC-212, and LC-213). These additional wells were
installed between October 16 and 17, 2003 using an air rotary drilling method, and similar construction
materials as used for wells LC-201 through LC-207. All wells were screened from above the water table
to the top of the till unit at a depth of approximately 10.1 m (33 ft). The wells were installed in a 15.2-cm
diameter borehole, and were constructed using 5.08-cm diameter stainless steel casings, 2.44 m in length,
and 5.08-cm diameter stainless steel well screens, 7.62 m in length. A 12/20 sand filter pack was installed
in the annular space around the well screens, on top of which was placed a bentonite seal, and the remaining
space to land surface was filled with grout.
MW-BO!
,9A
A
HCW-14
^LC-213
* LC-212
4LC-211
4 LC-207
4 LC-206
4LC-205
4LC-204
.A MW-D14
LC-203
4 LC-202
A LC-201
• i
Boundary
of treated
area for
NAPL
Area 1
0 10 20
Approximate scale (m)
Figure 4.1-5. Plan view of the NA1 source area at the East Gate Disposal Yard site at Fort Lewis and
the downgradient flux well transect. The diamonds represent flux wells and the triangles
represent hydraulic monitoring points (Brooks et al., 2008).
The PFMs were deployed for a period of three days: October 22, 2003 through October 25, 2003 for
pre-remedial measurements, and June 2, 2006 to June 5, 2006 for post-remedial measurements. After
the designated deployment period, the PFMs were retrieved and sampled using similar methods to those
described in Annable et al. (2005). Note that the PFM results reported below were not corrected for
convergence associated with flow around the well screen (Hatfield et al., 2004; Klammer et al., 2007). The
convergence factor (defined as the ratio of flux inside the PFM to that in the aquifer) can be estimated based
on hydraulic conductivity estimates of the aquifer (Ka ), well-screen filter pack material (K acl), well
screen (K ) and PFM sorbent used (KDm ). Often these properties are difficult to estimate or measure
v screen' ^ PFM' " "
because fine materials are removed in the developed zone near the well, some of which may become
trapped in the well screen, reducing its hydraulic conductivity. Best estimates of these properties for the
NA1 site are K .. = 9,100 cm/d;/C, = 43,200 cm/d; K = 3,540 cm/d and Kpn= 33,000 cm/d.
aquifer y 7 jitter pack. y 7 screen 7 rrM 7
This produces an estimated convergence factor of 1.03. The PFM data presented here are uncorrected (i.e.,
-------
the flux as measured inside the PFM is presented), and given the relative insignificance of 3%, it is assumed
that the flux inside the PFM and that in the aquifer are the same. Also, it is noted that the convergence
factor has no impact on the comparison of pre- to post-remediation PFM results.
The MIPTs were conducted as soon as possible following the retrieval of the PFMs from the wells to
allow a comparison of measured fluxes with a minimum time lag. The pre-remedial MIPT was conducted
between November 3, 2003 and November 7, 2003, and the post-remedial MIPT was conducted between
June 12, 2006 and June 17, 2006. Flow rates and water levels were measured periodically at each of the
wells and the combined flow was recorded as it accumulated in a temporary storage tank. Water levels
during the pre-remedial shallow transect MIPT were also monitored in one well (D14) near well LC-203,
and one down-gradient well (B09); and water levels in two additional down-gradient wells (HCW-5 and
HCW-14) were also monitored during the shallow post-remedial MIPT.
For the pre-remedial MIPT, each well was pumped using a single peristaltic pump with dual heads. Once
the MIPT was underway, it became evident however that the peristaltic pumps were inadequate to induce
sufficient hydraulic stress based on the limited amount of drawdown observed in the pumping wells. Based
on this experience, the pumping equipment used during the post-remedial MIPT included a centrifugal
pump that provided for higher pumping rates.
Groundwater samples from the effluent of each pumping well were collected for contaminant analysis
approximately once every four hours throughout each MIPT. Once collected, the samples were
immediately placed in coolers with frozen ice packs, and the coolers were shipped overnight to the
laboratory for refrigerated storage and analysis using a gas chromatograph/flame ionization detector
(GC/FID). For the post-remedial test, initial results indicated that the contaminant concentrations were
below method detection limits. Consequently, the samples were re-analyzed using a hexane extraction
process to concentrate the contaminant in the samples prior to GC/FID analysis.
Based on 204 samples collected from the shallow flux well transect during the pre-remedial PFM
deployment, the minimum, maximum, mean, and spatial standard deviation groundwater flux values were
0.8 cm/day, 68.8 cm/day, 26.9 cm/day, and 19.3 cm/day, respectively. Note that PFM data are integrated
over the vertical sampling interval of approximately 30 cm. Based on 166 samples collected from the
shallow flux well transect during the post-remedial PFM deployment, the minimum, maximum, mean,
and spatial standard deviation groundwater flux values were 0.3 cm/day, 70.4 cm/day, 16.1 cm/day, and
11.7 cm/day, respectively. The mean post-remedial groundwater flux was less than the pre-remedial
groundwater flux, apparently the result of changes in the local hydraulic gradients.
As discussed above, the pumping rates used during the pre-remedial MIPT were not large enough to induce
sufficient changes in hydraulic head to estimate the groundwater flux directly from the test results, an issue
that was compounded by interference from a series of precipitation events that occurred prior to and during
the MIPT. As a result, the pre-remedial groundwater flux was estimated by scaling the Darcy flux estimates
obtained during the post-remedial MIPT using the ratio of the pre- to post-remedial hydraulic gradients as
measured in surrounding wells. Furthermore, an analysis of the hydraulic gradient in the vicinity of the
transect indicated considerable temporal variation in the flow direction, and it was considered necessary to
explicitly account for the flow direction relative to the transect.
The hydraulic gradient in the vicinity of the transect was estimated using hydraulic head measurements
obtained from the US Army Corps of Engineers. For the pre-remedial test, an average of measurements
collected on November 4 and November 20, 2003 were used, providing an average hydraulic gradient
magnitude of 0.0037, directed at an azimuth of 294 degrees. For the post-remedial test, the average
of measurements collected on June 6 and June 12, 2006 were used, which gave an average hydraulic
gradient of 0.0021 directed at an azimuth of 219 degrees (see Brooks et al [2008] for a discussion of the
groundwater flow direction and its impact on mass discharge estimates crossing the source control plane).
While November 4, 2003 and June 12, 2006 corresponded to the start of the MIPTs, the rates used at the
beginning of these tests were considered insufficient to affect the surrounding water elevations (i.e., the
maximum drawdown in the wells was only 2 cm over the first six hours of the test, prior to or during which
the surrounding water level measurements were collected).
-------
The transect-wide average post-remedial groundwater flux based on the MIPT data was estimated
at 18.0 cm/day. The pre-remedial groundwater flux was estimated by multiplying the post-remedial
groundwater flux by the ratio of the pre-remedial to post-remedial hydraulic gradients. The pre-remedial
transect-wide groundwater flux estimate was therefore 31.7 cm/day.
Contaminant flux measurements based on the accumulation of contaminant in PFMs were completed
at each of the 204 pre-remedial sampling locations and 166 post-remedial sampling locations. The
contaminant flux profiles from selected wells with the largest fluxes measured prior to remediation
are provided in Figure 4.1-6. The depth profiles of contaminant flux from each well were averaged to
determine a mean contaminant flux value for each well, which was then multiplied by the cross-sectional
area associated with that well (equal to the product of the saturated thickness and well spacing) to provide
a mass discharge estimate per well (Figure 4.1-7). The data were also integrated over the width of the
control plane to determine a transect-wide contaminant mass discharge M D (g/day). The estimated total
pre-remediation TCE contaminant mass discharge was 743 g/day, while the post-remediation value was
3.4 g/day (Table 4.1-2). This corresponds to a contaminant mass discharge reduction of approximately
99.5%. The pre and post remedial DCE mass discharge was 155 g/day and 5.7 g/day, respectively,
corresponding to a mass discharge reduction of 96.4%.
c)LC-211
1
]
]
i
i^j
]
]
i
i ,
/
• —
<
r\
\
^
^
»
d
i
%
/
f
I
I
]
]
}
]
}
)L(
LI
>21
2
74
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Mass Flux (g/m2/day)
Figure 4.1-6. Mass flux profiles measured in selected wells with the largest fluxes as measured prior
to remediation using the PFMs. The solid diamond represents TCE, and the open square
represents DCE (Brooks et al, 2008).
-------
(a)
300
200
100
0
_Q
LC- LC- LC- LC- LC- LC- LC- LC- LC- LC-
•8
s o»
gj» 300-
^ 200
100
A -
201 202 203 204 205 206 207 211 212 213
I
B^
•
1
In
n
LC- LC- LC- LC- LC- LC- LC- LC- LC- LC-
201 202 203 204 205 206 207 211 212 213
(d)
I.D-
.0
n ^
n n-
—
|
|
|
u
u
LC- LC- LC- LC- LC- LC- LC- LC- LC- LC-
201 202 203 204 205 206 207 211 212 213
LC- LC- LC- LC- LC- LC- LC- LC- LC- LC-
201 202 203 204 205 206 207 211 212 213
Figure 4.1-7. Average mass discharge (g/day) for each well at Fort Lewis. Measured by a) PFMs and
b) MIPT for the pre-remedial tests; and by c) PFMs and d) MIPT for the post-remedial tests.
TCE is shown in black, and cis-DCE is shown in white. Note the change in scale on the
y-axis from the pre- to the post-remedial results to accommodate the data.
Using the MIPT, the total mass discharge across the transect was estimated as the sum of the contaminant
mass discharge for each well, calculated as the product of the contaminant flux and the cross-sectional
area assigned to each well. The cross-sectional area for each well was based on the well spacing (average
spacing of 6.4 m) and the depth between the water table and the bottom of the well screen (approximately
7.6 m). The contaminant flux was calculated as the product of the groundwater flux and the average
concentration measured in the pumping well effluents. For these calculations, sampling results reported as
"not detected" were assigned a value of zero. The mass discharge measured for each well in each test is
shown in Figure 4.1-7. The total TCE mass loading for each test and the DCE mass loading for the pre-
remedial test is presented in Table 4.1-2, along with the combined TCE and DCE molar loading for each
test.
The pre-remedial transect-wide TCE mass discharge estimates based on the MIPT test (536 g/day) was less
than the estimate by the PFM method (743 g/day), while the DCE mass discharge estimate by the MIPT
method (257 g/day) was higher than that estimated by the PFM method (155 g/day). For the post-remedial
test, the TCE mass discharge by the MIPT method (2.2 g/day) was likewise less than the PFM estimate
(3.4 g/day), but the DCE mass discharge by the PFM (5.7 g/day) was higher than the MIPT method
(0 g/day, DCE was not detected in the pumping well effluent samples).
A mean vertical gradient in the Vashon aquifer was estimated to be directed upward at 0.019 near NAPL
Areas 1 and 2 (USAGE, 2002). Vertical gradients may create ambient flow and mixing in wells that
are screened across heterogeneous contaminant distributions. However, once PFMs are deployed in
the wells, the sorbent material and impermeable spacers minimize vertical flow in the well. Moreover,
while deployed the PFM is measuring a temporally integrated response to groundwater that flows into the
well, which predominantly occurs from up gradient, not down gradient. Any biases in the distribution of
contaminants created by ambient flow in the well will be predominantly reflected down gradient of the
well, not up gradient. Alternatively, if it is assumed that the PFMs or an aqueous sample (in the case of
the MIPT for example), is impacted by ambient flow in the well, the critical issue is the potential dilution
of concentration below detection limits as a result of vertical flow in the well. Otherwise, dilution of
concentration (but still detectable) across the well depth should not affect the mass discharge estimates
since the depth over which the dilution occurred is explicitly included in the estimate of the control plane
area. Using an estimated pre-remedial detection limit of 0.1 mg/L, a pre-remedial groundwater flux of
30 cm/day, and a control plane area of-418 m2, then a rough estimate of the sensitivity of the pre-remedial
mass discharge estimate is ~13 g/day, which is an order of magnitude lower than the pre-remedial mass
discharge estimates. Likewise, using an estimated post-remedial detection limit of 0.005 mg/L, a post-
-------
remedial groundwater flux of 16 cm/day, and a control plane area of-418 m2, then the estimated sensitivity
of the post-remedial measurements is ~0.3 g/day, which is again an order of magnitude lower than the post-
remedial mass discharge estimates. Consequently, it seems unlikely that dilution due to vertical flow would
significantly change our conclusions about mass discharge reduction at Fort Lewis.
During remedial activities at NA1, the thermal treatment resulted in an estimated TCE mass removal
of 2,576 kg (425 gallons) (Beyke and Fleming, 2005). Based on the estimated DNAPL mass prior to
the remedial activity [3,800 kg (627 gallons) based on boring data], 68% of the TCE was removed.
According to PFM and MIPT contaminant flux measurements, TCE flux reductions were 99% or higher.
Consequently, at this site partial DNAPL mass removal from the source did result in a significant reduction
in the contaminant mass flux.
Table 4.1-2. Summary of TCE and DCE mass discharge (g/day) and combined TCE and DCE molar
discharge (moles/day) for Fort Lewis EGDY site using PFM and MIPT results, as well as
estimates based on the Transect Method (TM).
Discharge
TCE
(g/day)
DCE
(g/day)
TCE & DCE
(mole/day)
Method
PFM
MIPT
TM
PFM
MIPT
TM
PFM
MIPT
TM
Pre-Remediation
Test
Oct 2003
743
536
688
155
257
288
7.3
6.7
8.2
Post-Remediation
Test
Jun 2006
3.4
2.2
2.8
5.7
0
0
0.085
0.017
0.021
Percent
Reduction in
Mass or Molar
Discharge
99.5%
99.6%
99.6%
96.4%
100%
100%
98.8%
99.8%
99.7%
4.1.4 Former Sages Dry Cleaner Site, Jacksonville, Florida
The former Sages Dry Cleaner site in Jacksonville, Florida, was operational from 1968 to 1973 and from
1979 to 1989, and the primary subsurface contaminant at the site is PCE. Although the PCE release history
is unknown and PCE was not stored on site, PCE is presumed to have been released through operational
activities associated with the dry cleaner business. Based on soil cores, the site-specific hydrogeology
consists primarily of fine-grained sands with a discontinuous clay layer at approximately 10.7 m (35 ft)
below ground surface (bgs). The water table ranges from approximately 2 to 2.6 m (6.5 to 8.5 feet) bgs
with a natural hydraulic gradient of 0.0025 to 0.006. The groundwater flow direction is generally westward
toward a drainage canal located immediately adjacent to the western boundary of the site. The canal
generally serves as a sink for groundwater discharge, except during large storm events when the canal water
elevation rises, resulting in a reversal of the hydraulic gradient. Contamination is limited to the surficial
aquifer system, and DNAPL characterization activity indicated the presence of DNAPL predominantly at
approximately 9.1 m (30 ft) bgs.
-------
A pilot-scale field test of in-situ ethanol flushing was conducted at the site in August 1998 (Jawitz et al.,
2000). The target area for this first flood is shown in Figure 4.1-8. The initial mass of PCE in the target
area was estimated at 68 L of PCE based on a pre-remedial partitioning tracer test. The ethanol flushing
demonstration removed 43 L based on mass recovered from the extraction wells, resulting in 62% removal
effectiveness. This performance estimate is supported by soil core results (65% removal) and a post-
remedial partitioning tracer test (63% removal). Extensive monitoring of contaminant concentrations and
geochemical parameters within the source zone and downgradient plume using a conventional monitoring
well network were conducted for more than 3 years following the pilot-scale ethanol flood (Mravik et al.,
2003). This monitoring confirmed enhanced in-situ reductive dechlorination resulting from the residual
ethanol that remained in the subsurface.
A second ethanol flood to remove additional PCE was conducted in 2004 (see Figure 4.1-8 for the target
area of the second flood, located from 7.6 to 10.7 m [25 to 35 ft] bgs). This provided the opportunity to
evaluate the impact of source depletion on contaminant mass flux. A total of 13 injection and recovery
wells were installed to conduct the second flood. Soil samples from 7.6 to 10.7 m (25 to 35 ft) bgs were
collected from cores obtained during well installation, and results were used to estimate that 11 L of PCE
were present in the target zone (LFR, 2005). Ethanol flushing of the second target area during June and
July, 2004 resulted in the removal of approximately 21 L of PCE, the majority of which came from one of
three recovery wells (RW 202, the eastern most recovery well located between MLS 1 and MLS2 as shown
in Figure 4.1-8).
At the Sages site, bundled MLS wells were employed to add a vertical component to remediation and
monitoring studies for three dimensional analyses. A month prior to the August 1998 ethanol flood,
MLS 1-7 were installed in the source zone, each coupled to a respective recovery well (RW). Moreover,
MLS locations 8 through 12 were installed at the same time to monitor the downgradient plume. MLS
locations 13 and 14 were installed in mid 2004, and were used to augment the existing MLS monitoring
network. During sampling, each MLS well tube was pumped and purged to 40 mL before the sample was
collected. The samples were analyzed by gas chromatography for ethanol and volatile organic chemicals
(VOCs). Figure 4.1-8 shows the MLS and flux well locations, as well as the general areas targeted during
the first and second ethanol flooding events.
2193180
- 2193160
D)
2193140
2193120
Groundwater flow direction
Second Flood Target Area
S3
466460 466480 466500 466520
Easting (feet)
466540
466560
Figure 4.1-8. Multilevel sampler locations, flux well locations, and the general areas targeted during the
first and second ethanol flushing events at the Sages site. Units shown are the Florida State
Plane Coordinate System.
-------
The MLS network was sampled semi-annually for the six-year period following the 1998 ethanol flood pilot
test. The MLSs located orthogonal to the mean groundwater flow in the source zone (i.e., MLS-1, MLS-3,
and MLS-7) are referred to as the source zone transect. The MLSs perpendicular to groundwater flow ten
meters downgradient from the source zone (i.e., MLS-9, MLS-10, and MLS-11 prior to 2004, and those
MLSs plus MLS-13 and MLS-14 afterwards) are designated the downgradient transect. The final MLS
sampling event used in this study took place just prior to the second, full scale flushing event in June 2004.
Long term MLS monitoring was performed to assess the changes in concentration throughout the site
over the stated time period. Spatial patterns of the dissolved plume emanating from the source zone are
discussed as well as mass discharge estimates across the transects. The concentrations are reported as the
molar sum concentration due to the initiation of microbial reductive degradation of residual PCE evidenced
by increases in PCE biodegradation daughter products.
The mean molar concentration sum of each transect was calculated and then MD was determined from the
product of the site estimated Darcy flux and the cross sectional area of each transect:
= Cam*q*AK
(4.1-6)
The site characterization estimates for hydraulic conductivity (6 m/day) and hydraulic gradient (0.0025)
were used to estimate the site Darcy flux (q ) of 0.015 m/day. The mean mass flux (7) was calculated as
the product of Csum and 1 . The mass discharge was determined from the product of the mass flux and the
cross sectional area of the transect (A^ using the modified version of Guilbeault et al. (2005) and Kubert
and Finkel (2006) in Equation 4.1-6.
Mravik et al., (2003) reported increases of acetate as ethanol concentrations decreased due to the microbial
oxidation of ethanol. Furthermore, ethanol is non-partitioning and completely miscible, thus it will be
slowly carried away by the groundwater. As ethanol was removed from the source zone by natural gradient
flow and microbial degradation, daughter products of PCE dechlorination were detected.
The field test demonstrated the threshold toxicity of ethanol on the microbial population that was observed
by Mravik et al. (2003) in laboratory tests with Sages core material. They reported laboratory column
biodegradation of PCE when the ethanol concentration was less than 1%. In Figure 4.1.9, the ethanol and
daughter/parent concentration ratios are plotted for post remedial MLS sampling six months to four years
after the pilot test. In the deeper lower flow zones, less of the remedial fluids were recovered during the
test. The highest concentrations of residual ethanol were found in the deepest MLS samples immediately
after the flood (up to 18%). Once the ethanol concentration dropped below 1%, increases in TCE and
c-DCE concentrations were observed in groundwater within the transect.
-1C
tf)
O 14-
"n 12 -
DJ ^
g 1°
(0
1 6
£ 4 -
O)
3 21
CO |
o 0L_
O
O
oo 00
I • o
m&vfLft-nl
n°
^ OH
m n N^
B
O
• TCE/PCE
ODCE/TCE
o
ly^ pTIi-i i-i
. _ «
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Ethanol (%)
Figure 4.1-9. MLS source zone PCE/TCE and TCE/c-DCE ratios as a function of ethanol concentration in
the groundwater up to four years after the 1998 pilot test (Sages site).
-------
Source Zone Transect Concentrations. The source zone concentrations were monitored for six years after
the flushing pilot test at the MLS wells. The concentrations of PCE, TCE, DCE, and VC were monitored at
the source zone transect (MLS-1, MLS-3, and MLS-7).
Aside from a slight post-remedial PCE flux rebound, the site demonstrated little variability in the mean
molar sum in each MLS for the first year after the flood. A slight increase in molar concentrations was
detected at the source zone transect in January 1999, 5 months after the flood. By May 2000, the mean
concentration molar sum began to increase significantly. In 2002, the concentrations increased markedly
during the period of greatest biological activity, and then began to decrease in 2003. By June 2004, source
zone molar sum concentrations decreased further to a range of 0.03-0.5 mM, lower than the initial post
flushing concentration range of 0.19-0.98 mM in August 1998. Several plausible explanations for the
significant decrease in source zone concentrations include, source mass depletion, aging of the source
zone architecture, and decrease of microbial activity due to loss of ethanol substrate. Surfer™ was used to
create contour plots of the source zone transect molar sum of chlorinated species (Figure 4.1-10). Standard
kriging was applied to interpolate between measured data points. The black plus (+) marks indicate the
location of each MLS sampling point.
MLS-7
MLS-1
MLS-7
MLS-1
MLS-3
07-7999
03-2000
MLS-1
MLS-3
03-2001
10-2002
1.8 mM
1.2 mM
0.6 mM
0.0 mM
03-2003
06-2004
Figure 4.1-10. Sages source zone transect natural gradient concentrations for the 6-year period after the
August 1998 ethanol flood. Values are millimolar sums of chlorinated ethenes (mM).
Transect is MLS-7 to MLS-3, 1.49 m across, 8.08 m to 9.91 m bgs.
-------
Source Zone Transect Mass Discharge. As aqueous PCE discharge decreased in 1999 and 2000, DCE
mass flux rate increased as a result of biological dehalorespiration (Figure 4.1-11). This trend continued
until 2003. In the last two years of the study period (2003 and 2004) all discharges decreased except vinyl
chloride (VC), which was not detected until 2003 and 2004. Maximum and minimum mass discharge
values determined in the source zone transect ranged from 1.9 to 31.1 mmol/day (0.3 to 5.2 g/day) for
PCE, 0 to 2.3 mmol/day (0 to 0.3 g/day) for TCE, 0 to 36.9 mmol/day (0 to 3.6 g/day) for DCE, and 0 to
0.9 mmol/day (0 to 0.06 g/day) for VC. The site experienced a post-remedial rebound of mass discharge
of PCE in the first six months after the pilot flushing test. This is likely due to the residual ethanol
concentrations enhancing PCE solubility until natural gradient flow removed the majority of the ethanol.
However, by the end of the first post-remedial year PCE discharge began to decrease and continued to
decrease until the end of the study period. Figure 4.1-11 shows the source zone transect mass discharge
over the six-year period after the 1998 ethanol flood.
o
0
E>
n
Q
w
eo n
5°
40
Time
Figure 4.1-11. Sages source zone mass discharge (mmol/day) before and following the August 1998
ethanol flood.
After the initial post remedial rebound, up to 31.1 mmol/day (5.2 g/day), aqueous PCE discharge decreased
to below 20 mmol/day (3.3 g/day), eighteen months after the flood event, in the March 2000 sampling
event. During most of the post-remedial sampling period, the site demonstrated an increase in the total
chlorinated ethene mass discharge compared to the decreasing PCE discharge. This is largely due to the
formation of PCE daughter products down the biodegradation pathway which are much more soluble and
have lower sorption to solids and thus can be in solution at higher total molar concentrations (Chu et al.,
2004). By 2002, DCE discharge was the dominant species of the molar sum discharge indicating increased
biodegradation of PCE and TCE. All chlorinated ethane mass discharges decreased in 2003 and 2004,
years five and six respectively, except for the detection of low levels of VC. A paired t-test of the data
demonstrated that the molar sum of concentrations at the end of study period, June 2004, were statistically
different from the initial molar sum in September 1998. The total discharge reduction over the last two
years of the study period may in part be due to source depletion, but is most likely a result of biosubstrate
losses as ethanol was removed by groundwater flow and microbial use.
Downgradient Transect Concentrations. In addition to monitoring the source zone, the MLS transect
installed downgradient of the source zone was sampled as well. The results of the downgradient control
plane analysis are summarized in the Surfer™ contourplots in Figure 4.1-12. The transect consisted of
MLS-9, MLS-10, and MLS-11 for the time period from 1998-2003. Two additional samplers (MLS-13 and
MLS-14) were installed in 2004 in preparation for the second ethanol flood. The molar sum concentration
values in the downgradient transect are slightly lower than that of the source zone transect due to
dispersion, and degradation in the 10 m of aquifer between the two control planes, 1.1 mM maximum to
1.8 mM maximum respectively. Looking at the mean MLS well results, the downgradient transect recorded
-------
relatively stable behavior for the first two post remedial-years (1999 and 2000) while recording a slight
increase in the center of the transect in 2000. Throughout MLS monitoring at this transect, the higher
concentrations were detected in the center of the transect at MLS-10, and in the upper region of MLS-9.
In most of the transect, increases in concentrations were observed until peaking in 2002 at 1.08 mM
(Figure 4.1-12). After 2002, reductions in aqueous chlorinated concentrations were observed until the end
of the sampling period in 2004, concluding at 0.33 mM. These changes will be further examined through
mass discharge evaluation in the next section.
MLS-9
MLS-11 MLS-9
MLS-10
MLS-11
MLS-9
03-2001
MLS-10
10-2002 I
MLS-11 MLS-9 MLS-13 MLS-10 MLS-14 MLS-11
-12
10-2003 06-2004
Figure 4.1-12. Sages downgradient concentration distribution for the 6-year period.
Downgradient Transect Mass Discharge. At the second control plane, chlorinated species and molar
sum mass discharge were also calculated. These MLS wells were not installed prior to the pilot flood, so
only an initial post remedial assessment is provided here. Figure 4.1-13 shows the chlorinated species
discharge across the transect for the six-year timeframe of this project. The downgradient transect recorded
some initial site flux increases in the first six months, likely site rebound, until decreases in PCE and the
associated increases in DCE were observed at the end of the first post-remedial year in 1999. The PCE
discharge began decreasing the most during 2001 and 2002, and accordingly the DCE and molar sum
spiked markedly in this time frame. In 2003, all mass discharges began decreasing until reaching their
lowest levels at the cessation of sampling in 2004.
The magnitudes of the individual chlorinated ethene mass discharges determined in the downgradient MLS
transect ranged from 1.5 to 68.4 mmol/day (0.2 to 11.3 g/day) for PCE, 0 to 58.1 mmol/day (0 to 7.6 g/day)
forTCE,Oto 13 3.2 mmol/day (0 to 12.9 g/day) for DCE, and 0 to 1.5 mmol/day (0 to 0.09 g/day) for VC.
-------
The patterns are comparable to the source
zone transect, however the magnitude
of the mass discharge is much greater.
Although the concentrations were slightly
lower in the downgradient transect, the
cross-sectional area is much greater in
the downgradient transect, and thus the
mass discharge is much larger. In the last
two years of the study period, the mass
discharge decreased significantly, but
the values were not statistically different
from the initial post-remedial discharge.
The observed decrease is due in part
to the biodegradation of the residual
source mass, but more likely due to the
exhaustion of the residual ethanol as a
food source for dechlorinating microbes.
Another reason for the higher final flux
in 2004 was the small additional PCE
source mass discovered outside the pilot
treatment zone. This region, just west of
the pilot flood area, was treated during the
full scale remediation in July 2004.
H A D LJ = U ^-
"9 ,•• •<&•; r-&---r O- r-O""t
Time
Figure 4.1-13. Sages downgradient transect mass discharge
in mmol/day for the six-year period following
the 1998 ethanol flushing event.
Mass Flux Based on Downgradient MLS Sampling. Pre- and post-flood mass flux was determined from
MLS sampling in the downgradient control plane for the second, full scale ethanol flood in July 2004.
The results are summarized in Figure 4.1-14. The second remediation was highly successful in removing
the residual PCE, as it was below detection limit in the post-remedial samples. However, there must be
some remaining PCE in the subsurface materials due to the presence and increases in PCE biodegradation
daughter products recorded after the 2004 flood. With ethanol serving as an electron donor, indigenous
microbes dechlorinated the residual PCE. PCE was present in aqueous form in the original source zone in
2007, but at a maximum of 0.08 mM. This concentration is likely easily dechlorinated in the 10 meters of
aquifer between the source zone and the downgradient transect.
Sages June 2004
n 9^
n 90
">;
CO
^
"3) o 15-
V
E>
CO
"5 0.10
CO
b
CO
CO
CO o 05-
• PCE
• TCE
D c-DCE
• VC
L
nfl „
9 13 10 14 11
MLS Well
0.40 ,
0.35.
_ 0.30.
>•
CO
1 °-25
p 0.20.
CO
1 0.15.
b
-------
Mass Flux based on Passive Flux Meter Measurements. Mass flux was measured at the downgradient
transect using PFMs deployed in the 5 flux wells (FW09, FW10, FW11, FW13, and FW14 as shown in
Figure 4.1-8). These wells were screened over a 3.05 m (10 ft) interval (approximately 7.62 to 10.67 m [25
to 35 ft] bgs) consistent with the MLS sampling network. This measurement was conducted in June, 2004
just prior to the second cosolvent flood. The results from this PFM deployment are summarized in Figure
4.1-15. The integrated mass discharge at the down gradient transect was 59.8 mmol/day (5.8 g/day) DCE.
Other degradation products were present at significantly lower mass loads (Vinyl chloride = 7.5 mmol/day
[0.47 g/day]).
The post-remediation PFM flux measurement was conducted in March, 2007. The procedures used
for this measurement replicated the pre-remedial measurement. The post-remediation PFM flux data
are also provided in Figure 4.1-15. The integrated mass discharge at the down gradient transect was
20.6 mmol/day (2.0 g/day) DCE. Other degradation products were present at significantly lower mass loads
(Vinyl chloride = 0.64 mmol/day [0.04 g/day]). Based on these data, there was a 66% decline in DCE mass
discharge at the down gradient transect.
2.5
2
1
ro 1.5
TJ
CO
•i
•£: 0.5
0
Sages June 2004
1 .
-i
L
• PCE
• TCE
D DCE
D VC
i
-
FW9 FW13 FW10 FW 14 FW11
2.5
2
CO
;g
•D
CO
— ' i
f 0.5
0
Sages March 2007
i-i
• PCE
D TCE
D DCE
D VC
n _
FW9 FW13 FW10 FW 14 FW11
Figure 4.1-15. Pre- and post-cosolvent flood passive flux meter mass discharge measurements in the down
gradient well transect at the Sages site.
Mass Flux based on Modified Integral Pump Test Measurements. Both pre- and post-remedial PFM
deployments were followed by MIPTs. The pre-remedial MIPT was conducted for approximately 2.2
days, and pumping rates ranged from approximately 10 to 600 mL/min per flux well. A total of 13 samples
were collected from the effluent of each pumping well over the duration of the test, and once collected, the
samples were temporally stored on site in coolers until being transported to the University of Florida for
GC/FID analysis. Analysis of hydraulic head data collected following PFM retrieval but prior to the start
of the MIPT indicated the groundwater flow was directed away from the canal as opposed to the typical
direction as shown in Figure 4.1-8. Furthermore, subsequent analysis of the hydraulic data collected
during the pre-remedial MIPT suggested groundwater was flowing generally to the east rather than the
west. Storm events have been noted to raise the water level in a canal located immediately to the west of
the site sufficiently to temporarily reverse the groundwater flow direction. It is speculated that this may
have occurred prior to the MIPT, which would account for both the head data collected prior to the start of
the MIPT, and the results based on the MIPT hydraulic data. Consequently, the hydraulic data collected
during the pre-remedial MIPT was only used to estimate the average hydraulic conductivity in the vicinity
of the transect (1.7 m/day), and the hydraulic gradient magnitude (0.0035) and direction based on historic
data was used to estimate the groundwater flux (0.6 cm/day). The compound with the highest average
concentration from all wells was DCE, followed by PCE, TCE, and VC. Figure 4.1-16 summarizes the
mass discharge per well based on the MIPT data. The total DCE mass discharge based on the MIPT results
was 10.3 mmol/day (1.0 g/day). This value is in general agreement with the final DCE mass discharge
-------
based on the MLS results (Figure 4.1-14), but less than the PFM results. In general however, the mass
discharge spatial pattern based on the PFM and MIPT results agree, with each suggesting a decreasing
trend moving from the southern to the northern end of the transect. The difference in magnitude between
the PFM and MIPT results can predominantly be attributed to the difference in groundwater flux estimates.
The average groundwater flux based on the PFM results was 2.6 cm/day, and the average groundwater flux
based on the MIPT results was 0.6 cm/day. These groundwater flux estimates differ by a factor of 4.5,
which accounts for most of the difference in DCE mass flux estimates.
The post-remedial MIPT was generally conducted in a similar manner to the pre-remedial test, with the
exception that the test was conducted for a longer duration (3.7 days) in an attempt to improve the measured
hydraulic response and consequently the groundwater flux estimate. Pumping rates ranged from 15 mL/min
to 50 mL/min, and 16 samples were collected from each pumping well. Analysis was completed on every
other sample however to minimize the laboratory workload, and inspection of the resulting concentration-
time series indicated that analysis of the remaining samples was not warranted. The compound with the
highest average concentration from all wells was DCE, followed by TCE and VC. No PCE was detected.
The average groundwater flux based on the MIPT hydraulic data was 0.6 cm/day (directed in the westward
direction). The total DCE mass discharge was 1.1 g/day based on the MIPT results, indicating no change
in DCE mass discharge as a result of the second cosolvent flood. As with the pre-remedial test, the mass
discharge based on the MIPT results was less than the mass discharge estimates based on the PFM results,
while the general spatial distribution effluxes was similar.
0.5 -I
0.4
•>;
CO
5
Jo,.
E>
CO
|
b 0.2
%
CO
0.1
0.0 •
Sages June 2004
r
;
12
DPCE
• TCE
DVC
iflJL
FW09 FW13 FW10 FW14 FW11
Sages March 2007
>»
CO
S
^no
Discharge
3 C
J C
(O
(/>
co
5
0 1 -
r
;
-1 •
-|
DPCE
• TCE
DDCE
• VC
FW09 FW13 FW10 FW14 FW11
Figure 4.1-16. Pre- and post-cosolvent flood MIPT mass discharge measurements in the down gradient
transect at the Sages site.
4.1.5 Borden CFB Site
The effectiveness of DNAPL source zone treatment with In-Situ Chemical Oxidation (ISCO) was evaluated
using down gradient mass flux and discharge measurements. The source area consisted of DNAPL
that remained in the aquifer six years after a controlled release of 50 L of a mixture of PCE, TCE and
chloroform. This release is referred to as the Borden DNAPL Injection (BDI). The remaining mass was
primarily PCE with an estimated total volume of 5 L. The study was conducted in the unconsolidated
Borden sand aquifer with a saturated thickness of about 3.2 m. The aquifer is underlain by a regional silt
and clay unit. The water table varies between 0.5 and 1.5 m below ground surface. The mass flux and
discharge leaving this source area was quantified using three methods, multilevel samplers, passive flux
meters and integral pump tests. Measurements were conducted shortly before the source area was assessed
-------
using a partitioning interwell tracer test (PITT). Following the PITT the DNAPL source area was treated
with permanganate. Injections of permanganate were conducted over a two month period delivering a
load of oxidant to the source area. The objective of the treatment was to produce partial mass removal
of the DNAPL in the source area. After a period of approximately 1.3 years the mass flux and discharge
was again measured using the same protocols. This interval between treatment and post-treatment flux
measurements was deemed adequate to assess flux response at a distance of 2 m down gradient of the
source area given the Darcy velocity estimate of 4 cm/day.
The BDI site was instrumented heavily with multilevel samplers as part of an ongoing study to assess
plume dynamics (Figure 4.1-17). The Row-6 set of multilevel samplers was selected for mass flux
measurements. The initial water quality data indicated that the PCE plume was located between MLS
6-2 and MLS 6-7. Thus, fully screened wells were installed one meter down gradient of each of these
multilevel samplers.
00
.3
4,903,080-
4,903,070—
4,903,060 —
4,903,050-
>8 Multi-level
Sampler Rows
.5
..•;.-•;..-.4 ••
'. "••" "!• :.';-;:::* DNAPL Source Area
2''
589,270
589,280
Easting (m)
Figure 4.1-17. Borden controlled release DNAPL site instrumentation.
The resulting 8 wells were used to measure mass flux using the PFM and MIPT methods in May of 2004.
The PFM data from this deployment is presented in Figure 4.1-18. It became apparent that the center of
the plume had shifted to the southwest (to the left when viewed from source area in Figure 4.1-17). Based
on this observation, an additional five wells were installed at one meter spacing to extend the transect
for plume capture. In July of 2004 a second flux assessment was conducted using the new wells and a
subset of the wells from the previous deployment. Not anticipated was the fact that the plume had shifted
position back into the original set of wells selected for flux assessment. This change in flow direction was
apparently the result of a transition from wet spring conditions to a lower water table position observed
in the summer. This is supported by the measured groundwater elevations (0.5 m higher in May) and the
measured Darcy velocities from the PFMs (4.4 cm/day in May and 2.8 cm/day in July). The higher water
table in the spring tends to discharge to a ravine causing a significant ground water flow direction shift
locally. Thus, it could not be confirmed that either PFM deployment sampled the entire plume width.
Groundwater samples collected from all 12 wells just prior to the July 2004 flux tests were used to assist in
estimating contaminant mass flux along the entire transect. These data are also provided in Figure 4.1-18.
The ground water data indicates that the width of the PCE plume is similar to the width of the domain
measured by the PFM deployments in July 2004 and that the concentration declines rapidly on the northeast
side of the plume. Based on this information the mass integration was conducted using only the July data.
Integrating the local PFM-based PCE mass flux values vertically within each well and horizontally across
the transect the mass discharge of PCE was 2.2 g/day (TCE = 0.023 g/day).
-------
200
400
600
800
1000
Distance from FW6-12 (cm)
Figure 4.1-18. PCE mass flux from passive flux meters installed in wells down-gradient of row 6 at the
Borden CFB site.
The results of the pre-remedial MIPT are provided in Figure 4. 1-19. The values presented are average mass
flux per well. The general distribution of the pre-remedial flux across the transect is consistent with that
observed during the PFM deployment. The integrated pre-remedial PCE discharge across the transect was
2.2 g/day using data from both sampling events (higher value used when two values were available), which
matches the value based on the PFMs.
O«_
ft 4-
Oa_
ft *) -
01 _
r— Jviay-u«nrTBj r— jui-ut I±TBJ -ex- \nnay-vo t±-osi»
=*5^
t
/
/
/
I
jf. —
*•" ^
1
1
1
I
1
/
1
/
i
i
/
s
^^f
A
ft
/ \
/ \
/ \
/ \
J>"
s
/
\
\
\
\ t
\ I
\ /
v
x
J\
— "^\
\
*«;;
*>.
/ ^
/
/
/
\
\
\
\
s^
w
^y
^x
s.
'"w
— ^ —
Figure 4.1-19. PCE mass flux from MIPT measurements at the Borden CFB site.
Following the pre-remedial PFM and MIPT tests, a PITT test was conducted in August 2004. The PITT
used four injection wells and four extraction wells that encompassed the DNAPL source zone extending
down to the Row 6 multilevel samplers. An example breakthrough curve from the PITT is provided in
Figure 4.1-20. The retardation of the partitioning tracers 2-octanol and 3,5,5-trimethyl-l-hexanol are
evident. Based on the measured tracer retardation and system flow rates, the estimated volume of DNAPL
in the source area was about 7 L which is somewhat higher than the 5 L estimated by plume mass balance.
With the previous mass discharge estimate of 2.2 g/day and the PITT-based estimate of 7 L of PCE
(approximately 12 Kg) a simple zero order assumption for mass dissolution (constant discharge) indicates a
period of 14 years to remove the remaining mass.
-------
Well 2
o
(J
o
0.01
0.001
0.0001
0.00001
Time (days)
Figure 4.1-20. Example breakthrough curve for the pre-remediation partitioning tracer test at the Borden
CFB site.
Following the PITT the site was treated with permanganate to oxidize the PCE in the source zone. The
treatment was not intended to remove the entire source zone mass but rather some fraction of the mass to
assess changes in mass flux and discharge. The partial treatment took place over several months following
the PITT.
The post-remedial PFM and MIPT deployments were conducted in May 2006. During these deployments
all 12 flux wells were evaluated using PFM while 8 of the wells were pumped during the MIPT. The
average Darcy flux for this measurement was 4.0 cm/day based on the PFMs thus groundwater flow
was similar to the first deployment conducted. The groundwater level was also comparable. The PFM
derived well mass fluxes are presented in Figure 4.1-18. The maximum flux is located at well FW6-9
and is dominant at the single well. The distribution of post-remedial fluxes is likely similar to the
distribution during the first PFM deployment that entirely missed the peak mass flux. The integrated
PCE mass discharge across the transect was 0.54 g/day. This represents a 75% decline in PCE discharge.
However, the post-remedial MIPT mass discharge estimate is substantially larger than the PFM estimate
(Figure 4.1-19). The pre- and post-remedial MIPT measurements are similar. The post PCE discharge was
2.4 g/day slightly higher than the pre-remedial measurements. The differences between the MIPT and PFM
data for the post measurement could be due to the interwell distribution of mass flux within the transect.
The MIPT interrogates a larger volume of the formation and is more likely to identify interwell "hot spots"
and thus provide a more representative measure of mass discharge.
It is estimated that the PITT removed only about 1% of the DNAPL mass and that the natural gradient
dissolution would have removed approximately 8%. Thus, any significant reduction in mass discharge
could be attributed to the ISCO. The quantity of mass removed would account for approximately 20 years
of natural gradient dissolution assuming a first order (exponential decay) model for mass dissolution.
The PFM data can also be used to evaluate the spatial distribution of mass flux from the source zone.
Two contour plots are presented (Figures 4.1-21 of the mass flux distribution along the transect with the
perspective looking down gradient from the source area. Both the pre- and post-remediation distributions
are fairly narrowly defined "hot spots". The pre-remediation mass flux encompasses three wells but is
in a narrow vertical zone approximately 0.5 meters thick (Figure 4. l-21a). Following the ISCO a "hot
spot" still remains but is lower in strength and smaller in size. This type of flux distribution data could be
beneficial for targeting remedial efforts within the source zone.
-------
a)
1
b)
1
—
—
040
L
r
L
\
^
E
\
L
^
^
L^
j
. O4-1
1 "" PCE Flux
JJ *-" (ma/m2/d)
E
gSLgg a 2000-2400
[Qj " • 1600-2000
oi G I^UU-TOUU
S6 Elevation 0800-1200
• 400-800
QC
b5 n 0-400
CA
00
CO
CM
2 11 10 9 8 7 6 5
Flux Well ID
PCE Flux
I T
1
•w
Is
L
m
m
OJ U 1000-120
57 a 600 800
S6 D 400-600
.S5 EleVati°n .200^00
3>| • 0-200
S3
32
prp Plnv^n/m2/Haw>
2 11 10 9 8 7 6 5.5 5 4.5 4 32
Flux Well ID
0
Figure 4.1-21. Pre- and post-remedial mass flux distribution of PCE in the row 6 transect at the Borden
CFB site.
4.1.6 MIPT Uncertainty Analysis
To date, work published on the uncertainty of IPX measurements includes Jarsjo et al. (2005); Zeru and
Schafer (2005); Jarsjo and Bayer-Raich (2008), and Goltz et al. (2009). With the exception of Goltz et
al (2009), what has been published has been developed for the IPX method as originally introduced (i.e.,
sequentially pumping each well at a constant rate). Goltz et al (2009) investigated the accuracy of two
methods (tandem circulation wells and MIPX) to measure contaminant flux in an artificial aquifer. For the
tests conducted, MIPX results underestimated the true flux by 43% to 70%. Xhe contaminant (chloride)
concentrations used in the tests were relatively uniform in both space and time, therefore it can be assumed
that the error associated with the MIPX results was due to the estimate of groundwater flux rather than the
interpretation of the concentration-time series. Possible explanations offered for the MIPX errors noted
were well screen loss, partially-screened wells (the wells used had alternating screen and blank sections), or
a failure to obtain horizontal steady-state flow. As a comparison, the tandem circulation well method was
more accurate, underestimating mass flux by only 2% to 16%.
-------
Additional investigations were conducted independently to evaluate the uncertainty associated with the
modified IPX (MIPT) as described in Section 4.1.1. Uncertainty in flux measurements based on MIPTs may
originate from spatial heterogeneity, a failure to meet test assumptions, or from measurement uncertainty.
Assuming the former are met, the latter may be quantified using error propagation techniques and estimates
of uncertainty in measurements of contaminant concentrations, water elevations, well locations, and flow
rates. However, uncertainty in MIPT results due to measurement uncertainty (or measurement error) will
most likely be less than uncertainty associated with a failure to meet test assumptions. Presented here
are selected results of a systematic study to investigate the impact of both types of uncertainty on mass
discharge measurements using the MIPT. The impact of measurement errors and modeling assumptions
on estimates of the Darcy flux are first investigated, followed by an assessment of modeling assumptions
related to the interpretation of the concentration-time series information.
Contaminant flux based on the MIPT is estimated from
J = Cq
(4.1-7)
where J is the flux (ML^T1), C is the concentration (ML"3) estimated from the concentration time-
series, and q is the groundwater flux (LT'1) estimated from the hydraulic data during pumping. As a
first approximation, the assumption is made that the uncertainty in the measurements of C and q are
independent, and therefore the uncertainty in J as expressed by the variance, is
(4.1-8)
Equation (4.1-8) can be rearranged to express the coefficient of variation of J, COV[J], as a function of
the COV for q and C as Figure 4. 1-22 illustrates. Two lines are shown in the graph, the blue line with
diamonds represents the case where COV[g] = COV[C], and the purple line with squares represents the
case where either COV[q] * COV[C]=0 or COV[C] * COV[ 0 but COVfo] < COV[C] or COV[C] < COVfe], COV[J] will fall between the two
lines. For our analysis here the simplifying
assumption has been made that C and q are
independent, but it is noted that they may be
correlated in certain circumstances, perhaps for
example in the early stages after a spill when
the high concentrations may be correlated to the
high groundwater flux regions, or in an aged
system after the DNAPL has been depleted,
where the high concentrations may be correlated
with the low conductivity regions.
Uncertainty due to Measurement Uncertainty.
An analysis of the uncertainty in q as
estimated from the hydraulic data will first be
presented. Given the use of linear regression
to fit a line described by Equation 4. 1-4 to the
measured data, a Monte Carlo analysis was
used to propagate measurement error in the
independent and dependent variables. Of the
measurements used to estimate groundwater
flux from the MIPT (i.e., pumping rates,
saturated thicknesses, inter-well distances, and
water levels), water levels were considered
200%
160%
120%
8
O 80%
40%
v[q] # covra,
20%
40% 60% 80%
COV[q] and COV[C]
100%
Figure 4.1-22. Covariance of ./plotted as a function of
the covariance for C and q assuming C
and q are independent random variables.
The blue line with diamonds represents
the case where COV[
-------
the most difficult to accurately and precisely measure since changes in head over relatively short distances
are desired. Consequently, the analysis presented here is focused on the sensitivity of groundwater flux
estimates based on the MIPT method to variability in water level measurements. The results will in
turn be used with Equation (4. 1-8) to estimate the sensitivity in flux estimates to groundwater flux and
concentration variability.
The uncertainty associated with the change in head from a pumping well in the transect to a down gradient
monitoring well is
where <7^A[ ., is the variance in the head difference, <7^[ofa] is the variance in the head at the observation
well, and (Th[i] is the variance of the head at the /* pumping well. If it is assumed that the variance in any
given head measurement is independent of location and independent from one another, then the variance in
the head difference is 2crh . Head measurements during pumping tests are typically collected using pressure
transducers and data logging devices, in which case the head is calculated from the elevation of the pressure
transducer plus the height of water above the pressure transducer. Assuming error in the latter can be
neglected due to the fact that it is an average of a number of readings once steady state has been achieved,
<7^ = (7^D , where (7^, is the variance of the pressure transducer elevation. The pressure transducer
elevation is typically calculated as the top of casing elevation minus the sum of the distance to water below
the top of casing and the height of the water above the pressure transducer. If measurement error in each of
these is assumed independent, then
And if the variance in each is assumed equal, then 0*^, =3
| 50% COV) in the groundwater flux estimate.
In practice, this can be viewed as the result of attempting to measure changes in groundwater elevation
over relatively short distances, and highlights the need for accurate and precise head measurements. As the
distance between the flux well transect and the down gradient observation points increases, the uncertainty
due to head measurement error will decrease (i.e, compare results for Dx = 3 m versus Dx = 6 m in
Figure 4. 1-23). However, moving the observation point further down gradient to minimize the uncertainty
due to head measurements should be balanced against increasing the test duration, the volume of water
pumped, and increasing the area over which an average is estimated.
-------
100
10
5
9 1
w '
0.1-
— •-
-n>
f=
R
m
MC&A
HP**
Hi
,,,
,««
r*
: --*•
11 w»*
«•
0.0001 0.001 0.01 0.1
Error in head measurement (m)
1000.00%
100.00%
^ 10.00%
CT
1.00%-
0.10%
0.0
.rv
3 m
MHRA
-»- Dx=6 m, MC&A
i
4
001
1*
r"
^r-
ir**
ft "
t
z
1 p
^
-'•^
ii; =
= = ;
!
0.001 0.01 0.1
Error in head measurement (m)
Figure 4.1-23. Standard deviation and the coefficient of variation for q as a function of uncertainty in head
measurements.
Shown in Figure 4.1-24 are COV[J] as a function of head measurement uncertainty, expressed as
STD[
-------
the domain size was sufficiently large to approximate an aquifer of infinite areal extent. A control plane
transect consisting of seven pumping wells was located along the y axis. The wells in the transect were
spaced 3 meters apart with the center well located at the origin. Three monitoring wells were located along
the x axis, and were also spaced 3 m apart. A variable grid size was used, and the grid size in the vicinity
of the wells was 0.08 m by 0.08 m. Thus the grid spacing was comparable to that of a typical two-inch
monitoring well (i.e., 4 cm radius in the model versus ~2.5 cm radius for a typical monitoring well).
Two initial transient simulations under confined conditions were conducted, the first with a single
pumping well at the origin and the second with all seven wells of the transect pumping. Each of these
two simulations consisted of five stress periods with increasing pumping rates, Q = {1, 2.5, 5, 7.5, and
10 m3/day}, and each stress period lasted 24 hours. Results from both simulations were analyzed using
Equation 4.1-4. The resulting estimates for K and q were compared to the specified values of 10 m/day
and 5 cm/day, respectively. For the first simulation with a single pumping well at the origin, the estimated
values of K and q were 8.72 m/day and 4.45 cm/day, yielding percent errors of 12.8% and 10.9%,
respectively. These errors are attributed to the fact that the numerical model predicts an average head over
the area of the entire cell, while the analytical model predicts the head at a radius of 0.04 m. If the heads
in the cells of the pumping well and the adjacent down gradient cell are averaged, and the result used as
the head of the pumping well in Equation 4.1-2, then the estimated K and q (and associated errors) are
10.11 m/day (1.1%) and 5.09 cm/day (1.9%), respectively. Forthe sake of comparison, in the remaining
analysis the head in the pumping well will be estimated as the average of the heads in the cell corresponding
to the pumping well and the cell immediately down gradient. In the second simulation with all pumping
wells active, the estimated K and q were 10.09 m/day (0.9 % error) and 5.02 cm/day (0.3 % error).
Two large-scale hydraulic conductivity heterogeneous patterns were investigated. The first pattern
consisted of a layered hydraulic conductivity system consisting of an aquifer with three layers as depicted in
Figure 4.1-25(a). Each layer was 3.33 m thick, and the geometric average hydraulic conductivity was equal
to 10 m/day, and the variation in conductivity between the layers was an order of magnitude. Transient
simulations with five stress periods were conducted with all seven pumping wells active. Each stress
period was 24 hours, and the pumping rates for each well were 1,2.5,5,7.5, and 10 m3/day. The hydraulic
conductivity of the top, middle, and bottom layers was 100, 10, and 1 m/day, respectively. For this layered
system, an equivalent hydraulic conductivity is calculated using
K =yML 41.n
where K is the average hydraulic conductivity, B and K. are the thickness and hydraulic conductivity of
the Ith layer, and #T is the total thickness. Application of Equation (4.1-11) to the layered system results in
an average hydraulic conductivity of 37 m/day, and with the hydraulic gradient of 0.005 imposed by the
boundary conditions, the average groundwater flux is 18.5 cm/day. Results from the simulation provided
estimates of K= 21.6 m/day and q = 10.7 cm/day, respectively. The percent error between these values
and the average values calculated above (i.e., K= 37 m/day and q = 18.5 cm/day) were 41.6% and 42.0%,
respectively, which suggests that estimation of groundwater flux based on the MIPT approach is sensitive to
the vertical distribution of hydraulic conductivity.
The second heterogeneous pattern investigated consisted of the blocks as shown in Figure 4.1-25(b). The
confined aquifer thickness was set equal to 10m. The center block was 40 m wide, and the other two
blocks extended over the remaining modeling domain. The hydraulic conductivity of the center block was
10 m/day, and that of the blocks on each side was 1 and 100 m/day, resulting in a geometric average K of
10 m/day and a variation in hydraulic conductivity between blocks equal to an order of magnitude. All
seven pumping wells and monitoring wells were located in the center section where K= 10 m/day and
q = 5 cm/day. The distance to the changes in K from the pumping wells on the ends of the transect was
llm. The estimated K and q were 10.1 m/day and 5 .0 cm/day, respectively, and the percent error relative to
the true values of the center block was 1.2% and 0.6%, respectively.
-------
b)
Figure 4.1-25. Layered (a) and block (b) macro-scale hydraulic conductivity patterns investigated with
numeric modeling simulations.
Unconfined Simulations. Several simulations conducted for the confined aquifer were repeated using
unconfined conditions to investigate how this deviation from the assumed conditions might impact the
results. The constant head up-gradient boundary condition was set to a value of 10 m and the down
gradient boundary condition was 5 m. The Boussinesq equation (Bear, 1979) under steady-state, no
recharge conditions (i.e., d/dx(h dh/dx) = 0) was used to estimate the pre-pumping saturated thickness
and groundwater flux at the control plane. The pre-pumping saturated thickness under these conditions is
independent of hydraulic conductivity, and the value at the flux control plane is 7.79 m. For the first case,
a single well at the origin was pumped over five stress periods (each of 24-hour duration) with increasing
pumping rates, Q = {1, 2.5, 5, 7.5, and lOmVday}. The unconfined aquifer consisted of a single layer of
hydraulic conductivity equal to 10 m/day, the groundwater flux at the transect was therefore 4.816 cm/day.
The estimated values of K and q based on the application of the MIPT to the model simulation results were
9.98 m/day and 4.93 cm/day, respectively; and the percent error between these values and the true values
was 0.2% and -2.3%, respectively. A second identical simulation was conducted with the only difference
being that all seven pumping wells were pumped using the same conditions previously applied to the well at
the origin. The estimated values of K and q based on the application of the MIPT to this model simulation
were 9.68 m/day and 4.98 cm/day, respectively. The percent error between these values and the true values
was 3.2% and -3.4%, respectively.
The layered case (Figure 4.1-25a) was also repeated as unconfined simulations. Transient simulations
with five stress periods of 24-hour duration were conducted with all seven pumping wells active.
The pumping rates for each well were 1, 2.5, 5, 7.5, and 10 m3/day. For the first layered pattern, the
hydraulic conductivity of the top, middle, and bottom layers was 1,10, and 100 m/day, respectively.
Equation (4.1-11) was used to calculate an equivalent hydraulic conductivity for the aquifer as a whole.
However, since this was an unconfined simulation, the top layer was only partially saturated, and the
saturated thickness was estimated at 1.12m (i.e., 7.79 m minus 2/3 * 10 m). The resulting equivalent
hydraulic conductivity was 47.3 m/day, and the groundwater flux based on this value of K was 22.8 cm/day.
Based on the application of the MIPT method to the simulation results, the estimated hydraulic conductivity
and groundwater flux were 46.6 m/day and 23.0 cm/day, resulting in errors of 1.4% and -1.2%, respectively.
A second, unconfined layered system was also simulated, the only difference being the hydraulic
conductivity of the top, middle, and bottom layers was 100, 10, and 1 m/day, respectively. The equivalent
K from this distribution and saturated thickness was 19.1 m/day, and the resulting groundwater flux was
9.2 cm/day. The estimated K and q based on the application of the MIPT to the simulation results were
25.1 m/day and 12.3 cm/day, respectively; and the resulting errors were -31.7% and -33.7%. These errors
were larger than those associated with the previous simulation, and are more similar to the errors noted for
-------
the three-layered, confined simulations. These results therefore suggest that the MIPT approach may be
sensitive to the actual vertical distribution in hydraulic conductivity, rather than just vertical variation itself.
In addition to the large-scale heterogeneous patterns shown in Figure 4.1-25, a random hydraulic
conductivity field with no correlation structure was used with a geometric mean hydraulic conductivity
of 10 m/day, and log-transformed standard deviation of 1. Two simulations were conducted: one with the
center well at the origin pumping only and another with all seven wells pumping. The pumping rates and
durations applied in each case were the same as those previously used. These simulations were used to
assess how well an average hydraulic conductivity and groundwater flux value could be estimated where
the scale of pumping was much larger than the scale of the hydraulic conductivity heterogeneity. For the
case with a single pumping well only, the estimated K and q based on the application of the MIPT to the
simulation results were 5.0 m/day (49.9% error) and 2.8 cm/day (40.7% error), respectively. The errors
noted were based on the mean value of the hydraulic conductivity. For the case with all wells pumping,
the estimated K and q based on the application of the MIPT to the simulation results were 6.0 m/day
(39.9% error) and 3.5 cm/day (26.5% error), respectively. As expected, the results with all seven wells
pumping were better at predicting the average hydraulic conductivity and groundwater flux compared to the
results with a single pumping well.
Comparison of Ideal and Non-ideal Capture Zones. Shown in Figure 4.1-26 are the capture zones
associated with a MIPT consisting of 7 concurrently pumping wells conducted in a homogeneous,
unconfined aquifer. As described earlier, the hydraulic conductivity is 10 m/day, the regional groundwater
flux is 5 cm/day directed in the positive x axis, and the initial saturated thickness at the well transect is
7.8 m. At early times, the capture zones for all wells appear symmetric about the pumping well. However,
at later times while the capture zone for the center well remains symmetric, the capture zones for the other
wells become increasingly skewed away from the origin due to concurrent pumping. Moreover, at later
times the capture zone for the two pumping wells on the ends of the transect draw water predominantly
from the regions beyond the ends of the transect. While this complicates the use of the concentration-time
series information to estimate the spatial distribution of the contaminant, it does help ensure the transect
captures the full width of the contaminant plume. Furthermore, with the exception of the two wells on the
ends of the transect, the well spacing is a good approximation to the width of the capture zone at later times.
a)
b)
Figure 4.1-26. Capture zones for an MIPT with seven concurrent pumping wells in an unconfined aquifer
with K = 10 m/day and q = 5 cm/day. Shown from left to right are the capture zones
associated with the five pumping-rate steps of a) 1 m3/day, b) 2.5 m3/day, c) 5 m3/day,
d) 7.5 m3/day, e) 10 m3/day. Each step lasted 24 hours.
-------
Shown in Figure 4.1-27 are the capture zones again associated with an MIPT consisting of 7 concurrently
pumping wells conducted in an unconfined aquifer with regional groundwater flux of 5 cm/day (in the
positive x-axis direction) and an initial saturated thickness at the well transect of 7.8 m. However, in this
simulation the hydraulic conductivity is heterogeneous, and assumed to be an uncorrelated log-normally
distributed random space function with a mean of 10 m/day and a log-transformed standard deviation of 1.
Figure 4.1-28 shows the capture zone at day 5 for a second hydraulic conductivity realization from the same
statistical distribution, and results from the homogeneous case and the first realization are also shown for
comparison. Differences in capture zones for the heterogeneous case relative to the homogeneous case are
noted by the second day (i.e., compare Figures 4.1-26b with 4.1-27b). As expected, differences are noted
in the spatial distribution of the capture zones between the two realizations (compare Figure 4.1-28b with
4. l-28c). Based on the figures shown, it appears that the width of the capture zone associated with the three
center wells can still be approximated by the well spacing. However, the capture zones for the two wells on
each end are larger in width than the well spacing.
a)
b)
Figure 4.1-27. Capture zones for an MIPT with seven concurrent pumping wells in an unconfined aquifer
with a log-normally distributed random AT field. Shown from left to right are the capture
zones associated with five pumping-rate steps of a) 1 m3/day, b) 2.5 m3/day, c) 5 m3/day,
d) 7.5 m3/day, e) 10 m3/day. Each step lasted 24 hours.
-------
a)
b)
Figure 4.1-28. Comparison of capture zones with seven concurrent pumping wells for the homogeneous
case (a) and heterogeneous cases: realization 1 (b) and realization 2 (c).
Uncertainty due to Concentration-Time Series Interpretations. For the MIPTs conducted at the four field
sites in this SERDP project, all wells were pumped concurrently rather than sequentially as conducted
in the original IPX method. Wells were pumped concurrently in our tests to minimize the disturbance
to the contaminant distribution that would result from pumping each well sequentially, and to avoid the
uncertainty associated with double counting mass located between adjacent wells. However, disadvantages
to concurrent pumping include the development of stagnation points between wells, and a more complex
flow field, which prevents the use of the Bayer-Raich et al. (2004) analytical solution. With respect to the
first disadvantage, it is nonetheless possible to sample a large fraction of the space between wells, so that
the general goal of the IPX (i.e., a measured response integrated over a large area) is maintained. With
respect to the second disadvantage, the average spatial concentration in the capture zone of the well was
estimated using the average concentration of the concentration-time series. Xhe mass discharge was then
obtained from the product of the contaminant flux and the cross-sectional flow area associated with the well
transect.
An analysis of the potential uncertainty in mass discharge measurements due to the interpretation of the
concentration-time series information as conducted for the MIPX was completed using a Monte Carlo
analysis. Xhe analysis consisted of two main parts: first the calculation of particles captured at a series of
times associated with a particular MIPX application; and the second, the calculation of the concentration-
time series by superimposing the distribution of particles captured with the contaminant plume distribution.
Similar to the analysis presented by Bayer-Raich et al. (2004), the analysis conducted here assumed: (1) a
two-dimensional (in the x-y plane), homogeneous and isotropic aquifer with uniform flow (directed in the
positive x-axis); and (2) a mature plume such that variations in concentration across the captured zone
in the x direction were assumed negligible, consequently concentration varied only in the y-direction.
Furthermore, for the analysis presented here, the concentration distribution along the y-axis was described
using a normal probability density function, and advection was assumed to be the predominant transport
process, and dispersive transport was neglected.
A series of particles (n = 360) were released around each well, and a reverse particle tracking algorithm
was used to defined the loci of particles captured at a given time. Xhe reverse particle tracking algorithm
was based on fourth-order, Runge-Kutta numerical integration with an adaptive time-step. Input variables
for this calculation consisted of groundwater flux, aquifer thickness, porosity, number of wells (and well
locations), pumping rates, and pumping durations. Xhe concentration at each location was then calculated
using the normal PDF (probability density function) with an assumed mean and standard deviation, and the
average concentration was then calculated for all particles at a given time. Xhe result was a concentration-
-------
time series for each well. The code was validated by comparing the true average concentration to that
estimated by applying the Bayer-Raich et al. (2004) analytical solution to the concentration-time series as
produced by the code when a single pumping well was specified. Uncertainty in mass discharge estimates
was then investigated by conducting a Monte Carlo analysis in which the mean and standard deviation of
the normal PDF (which described the contaminant plume) were uniform random variables.
Results from a single realization are shown in Figure 4.1-29 for a case consisting of seven concurrently
pumping wells spaced 3 m apart along the y axis. Pumping rates were Q = {1, 2.5, 5, 7.5, and 10} m3/day,
with each rate applied for a single day. The uniform groundwater flux was directed in the positive x-axis at
a rate of 5 cm/day, the aquifer thickness was 10 m, and the porosity was 0.15. Figure 4.1-29a) shows the
resulting capture zones at the end of each pumping step. The contaminant plume was centered at y = 1.5 m
(midway between the third and fourth pumping well) with a maximum concentration of 20 mg/L, and a
plume width defined using a standard deviation of 0.4 m (see lines in Figure 4.1-29a, which represent ±3a).
Figure 4. l-29b shows the concentration-time series resulting from this MIPT For this particular case, the
plume is centered midway between wells 3 and 4, which represents a more difficult case for MIPTs with
concurrent pumping wells; and the plume is relatively narrow, so that concentration is only detected in
wells 3 and 4. It is interesting to note in this case that while concentration continues to increase across the
entire test duration in well 4, concentrations reach a peak around day 4 for well 3, and then start to decline.
This can be explained by inspection of the capture zones for well 3 as shown in Figure 4. l-29a), which
reveal that an increasing proportion of clean water is drawn into this well as the test progresses due to the
effects of concurrent pumping. For this case the true mass discharge across the transect was 10 g/day, and
the measured mass discharge based on the average concentration across the entire concentration-time series
was 4.4 g/day, resulting in a percent error of 56%. Given the relative change in concentration with time
(as shown in Figure 4. l-29b), another approach that could have been used is to average concentration only
over the late-time portion of the concentration-time series (i.e., last day), in which case the measured mass
discharge is 9.0 g/day, with a resulting percent error of only 10%.
a) 15
-10
-15
-Well 6
-Well?
2 3
Time (days)
Figure 4.1-29. a) Capture zones during an MIPT consisting of seven concurrently pumping wells, and
b) the resulting concentration-time series for a mature contaminant plume centered
at y = 1.5m (midway between the third and fourth pumping wells), a maximum
concentration of 20 mg/L, and a plume width described by a standard deviation of 0.4 m.
-------
Using the same aquifer parameters and MIPT details as specified in the preceding paragraph, Monte
Carlo simulations with 1000 iterations were conducted in which the plume center-line varied according
to a uniform random variable, with limits of-9 < y < 9 m. Three separate Monte Carlo simulations
were completed, one each with plume standard deviations of 0.4, 0.8, and 1.2. Results are shown in
Figure 4.1-30, where the average percent error (i.e., the average of the error from each realization, defined
as [true MD - MD]/true MD) is plotted as a function of the plume standard deviation. The bars in the figure
indicate plus and minus one standard deviation of the calculated errors. As expected, the error decreases as
the plume width increases across the transect, resulting in a more homogeneous concentration distribution.
Also shown in Figure 4.1-30 is the percent error based on a point measurement technique (referred to as
PM in Figure 4.1-30). The mass discharge per well using a point measurement technique was calculated
as the product of the groundwater flux, cross-sectional area (aquifer thickness times well spacing), and the
concentration calculated at the well. The range in errors associated with the point measurement technique
is larger than the MIPT technique, especially for narrow plumes. It is also noted that while the average
error for the point measurement technique is zero, there is an apparent bias of approximately 8% in the
average error for the MIPT technique. Further modeling simulations (results not shown) indicated that this
is predominantly due to approximating the capture zone width using the well spacing, and using the average
of the concentration-time series as an estimate of the average spatial concentration in the well capture zone.
150%
100%
i_ 50%
liJ
4—1
0)
Q_
0%
-50%
-100%
-150%
0.4 0.8 1.2
Plume Standard Deviation (m)
Figure 4.1-30. Results from Monte Carlo simulations where plume position is treated as a uniform random
variable. Plotted in the y-axis is the average percent error in measured mass discharge
relative to the true mass discharge for MIPT and point measurement techniques (PM).
In summary, modeling simulations were used to investigate the uncertainty associated with the
measurement of mass discharge using the MIPT approach. Uncertainty in measurements may originate
from either a failure to meet test assumptions or from measurement uncertainty. For the conditions
investigated, uncertainty in Darcy flux was generally 40% or less, but was as high as 100% in some cases.
Using this upper value, the resulting uncertainty in mass flux ranged from approximately 120% to 160%
when the uncertainty in the average concentration ranged from 10% to 50%, respectively. Likewise,
uncertainty in mass discharge estimates resulting from uncertainty in the interpretation of the concentration-
time series information was also investigated. For those conditions investigated, the uncertainty in mass
discharge estimates ranged from approximately 30% to 3% for narrow to broad plumes relative to the well
spacing. However, results also suggested a positive bias in the average error of the MIPT measurements.
-------
4.2 Laboratory Research
The field studies described above provide important data concerning the response to DNAPL source
treatment; however, the difficulty and expense of quantifying DNAPL mass and architecture prevent an in-
depth field assessment of the factors that influence this response. Thus, a series of laboratory experiments
were designed to examine these factors. Bench-top experiments were conducted in two-dimensional
flow chambers that served as aquifer models. The flow chambers were packed with porous media in
both homogeneous and heterogeneous configurations and DNAPL was released at various locations.
Relationships between contaminant architecture and mass flux were evaluated for dissolution by aqueous,
surfactant, and cosolvent solutions and for mass depletion through air sparging. Light transmission
visualization (LTV) methods were used to map DNAPL distribution in the flow chambers. Laboratory
experiments are described in detail in the following peer-reviewed journal articles:
Light Transmission Visualization Methods:
Wang, H., X. Chen, and J.W. Jawitz (2008). Locally-calibrated light transmission visualization methods
to quantify nonaqueous phase liquid dissolution dynamics in porous media. Journal of Contaminant
Hydrology 102(l-2):28-38, doi:10.1016/j.jconhyd.2008.05.003.
Bob, M.M., M.C. Brooks, S.C. Mravik, and A.L. Wood (2008). A modified light transmission visualization
method for DNAPL saturation measurements in 2-D models. Advances in Water Resources 31(5):727-
742.
Relationship between Source Mass and Source Strength:
Pure, A.D., J.W. Jawitz, and M.D. Annable (2006). DNAPL source depletion: linking architecture and flux
response. Journal of Contaminant Hydrology 85(3-4): 118-140, doi: 10.1016/j.jconhyd.2006.01.002.
Totten, C.T., M.D. Annable, J.W. Jawitz, and J.J. Delfino (2007). Fluid and porous media property effects
on dense non-aqueous phase liquid migration and contaminant mass flux. Environmental Science and
Technology 41(5): 1622-1627.
Chen, X., and J.W. Jawitz (2008). Reactive tracer tests to predict DNAPL dissolution dynamics in
laboratory flow chambers. Environmental Science and Technology 42(14):5285-5291.
Kaye, A.J., J. Cho, N.B. Basu, X. Chen, M.D. Annable, and J.W. Jawitz (2008). Laboratory
investigation of flux reduction from dense non-aqueous phase liquid (DNAPL) partial source zone
remediation by enhanced dissolution. Journal of Contaminant Hydrology 102(1-2): 17-28, doi:
10.1016/j.jconhyd.2008.01.006 .
Bob, M., M.C. Brooks, S.C. Mravik, and A.L. Wood (2009). The impacts of partial remediation by
sparging on down-gradient DNAPL mass discharge. (In preparation).
4.2.1 Light Transmission Visualization Methods
A limited number of techniques can be used to accurately measure DNAPL and water content in laboratory
physical models (Darnault et al., 1998). Partitioning tracers, for example, have been shown to give an
accurate estimate of a known volume of chlorinated solvents emplaced in a sand-packed column (Jin et al.,
1995; Wilson and Mackay, 1995). A limitation of this method, however, is that the saturation obtained is
typically a bulk saturation estimate and information is not obtained on the detailed distribution of saturation
across the sand column. Other, non-invasive laboratory techniques that have been used to quantify fluid
content in laboratory experiments include X-ray measurements (Liu et al., 1993; Rimmer et al., 1998;
Tidwell and Glass, 1994), gamma ray radiation (Ostrom et al., 1998; Hopmans and Dane, 1986) and light
transmission visualization (LTV; Hoa, 1981; Niemet and Selker, 2001; Tidwell and Glass, 1994; Darnault
et al., 1998). Because of the limitations associated with the use of X-ray and gamma ray radiation (e.g. slow
-------
measurements, hazard of working with high energy sources, high cost), the LTV technique has recently
gained popularity (Weisbrod et al., 2003).
In this project, LTV methods were used to quantify DNAPL saturations in two-dimensional (2-D), two fluid
phase systems (Wang et al., 2008; Bob et al., 2008). The methods reported here are expansions of earlier
LTV methods and take into account both absorption and refraction light theories. Based on these methods,
DNAPL and water saturations can rapidly be obtained point wise across sand-packed 2-D flow chambers
without the need to develop calibration curves. The methods were applied to measure, for the first time,
undyed DNAPL saturation in small 2-D chambers.
The extent to which light is transmitted through the flow chamber is a function of adsorption by the porous
media and resident fluids and refraction at fluid/fluid and media/fluid interfaces. These adsorptive and
interfacial processes can be quantified by applying Beer's and Fresnel's laws to obtain:
/ = c/mn
-------
The new multiple wavelength method (Model 4) is expected to be universally applicable to any translucent
porous media containing two immiscible fluids (e.g., water-air, water-NAPL). Results from the sand-
water-PCE system evaluated here showed that Model 3 (Model C of Niemet and Selker, 2001) and Model 4
(multiple wavelength model) were able to accurately quantify PCE dissolution dynamics during surfactant
flushing. The average mass recoveries from these two imaging methods were about 90% during seven
cycles of surfactant flushing that sequentially reduced the average NAPL saturation from 2.6xlO~2 to
7.5xlO"4. Results from the three experiments (I, II, and III) are given in Table 4.2-1. Initial PCE saturation
distributions are presented here for all three experiments. In Experiment III, a 2% solution of dyed
Tween 80 with the same dye concentration as the resident water was displaced through the flow chamber to
rapidly dissolve the NAPL. After 1 to 3 pore volumes (PVs) of Tween injection (applied PVs for each cycle
are given in Table 4.2-1), the surfactant solution was displaced with 4 PVs of dyed water and images were
collected to quantify the PCE distribution. This cycle of surfactant was repeated seven times until 97% of
the PCE was flushed from the chamber. The PCE mass recoveries were calculated by comparing the known
PCE mass to the masses obtained from LTV estimation from the four models.
Method 2:
The saturations of both dyed and undyed DNAPL were measured in a water/oil system without dying
the water phase (Bob et al, 2007). To our knowledge, this is the first attempt to use light transmission
visualization to quantify DNAPL in multiphase systems without the use of dyes. Different known amounts
of PCE were added to a small silica sand-packed 2-D chamber that was initially saturated with water and
the amount of PCE in the chamber after each addition was calculated from image analyses. In another set
of experiments, known amounts of PCE were released into a larger 2-D model and air sparging experiments
were carried out to remove different fractions of the PCE mass from the chamber. A GC/MS instrument that
was connected to the chamber was utilized to quantify the PCE in the vapor phase leaving the chamber. In
addition, a carbon column was also connected to the chamber to trap all PCE removed from the model and
was analyzed to determine the mass of PCE removed.
Table 4.2-1. PCE distribution recovery, mean, and variance determined using the LTV models.
Experiment-flushing
cycle
PVsa
Mass Recovery (%)'
% of PCE removed
Model 1 Model 2 Model 3 Model 4
I
III-O
m-1
III-2
III-3
III-4
III-5
III-6
III-7
Average
Standard deviation
0.8
1.2
2.2
2.2
2.3
1.6
1.0
86.2
84.6
85.5
78.1
69.7
69.7
66.9
87.2
71.1
119.6
81.9
15.4
74.6
69.6
69.9
63.8
57.0
57.0
54.7
71.4
58.3
97.8
67.4
12.8
98.2
94.9
96.2
87.9
78.4
78.4
75.2
98.0
80.0
134.3
92.2
17.3
99.3
99.5
101.0
94.5
96.1
84.1
79.6
94.6
77.6
104.8
93.1
9.4
0.0
11.0
29.9
51.1
69.2
82.9
90.1
97.1
" Pore volumes of2%Tween 80 applied in each cycle
b Ratio of PCE volume estimated using LTV models and determined from effluent-based mass balance.
The accuracy of the method was evaluated based on mass balance calculations. For the first set of
experiments, the amounts of PCE calculated from image analyses were compared to the actual amounts
-------
added and for the second set of experiments image analyses results were compared to mass balance
estimates derived from effluent stream monitoring.
The LTV experimental system consisted of a 2-D flow chamber mounted on the front of a light source and
a digital camera. Flow chambers were constructed of aluminum frames and 0.635 cm or 1.27 cm thick
tempered glass plates and were packed with Accusand silica sands (Unimin, Le Sueur, MN). Drainage ports
in the bottom of the internal aluminum frame and injection ports attached to the top allowed inflow and
outflow of fluids. The large chambers used in the air sparging experiments had additional ports across the
top of the chamber to direct the effluent vapors through an in-line GC/MS and into an activated carbon trap.
Nominal internal dimensions of the small and large chambers were 15.24 cm by 15.24 cm by 1.4 cm and
48.26 cm by 48.26 cm by 1.4 cm, respectively. Flow chambers were packed with a range of sand fractions
and media heterogeneities.
Spatial distribution of light transmitted through the flow chamber was determined by collecting digital
images. The images were collected using a thermoelectrically air-cooled charge coupled device (CCD)
camera (Princeton Instruments model number 7486-002). The shuttered, monochromatic camera had a
16-bit dynamic range (65585 grey levels) and 512 by 512 spatial resolution. Images were collected using
either a 600 nm or a 543 nm center-wavelength, 10 nm band-pass filter (Melles Griot, Rochester, NY).
A summary of flow chamber experiments is presented in Table 4.2-2. PCE distribution within a flow
chamber was ascertained by subtracting images taken with PCE present in the chamber from images taken
when the chamber was fully water-saturated. Examples of these differential images for Pack A with 2 ml
PCE and Pack B with 4 ml PCE are shown in Figure 4.2-1. Since light adsorption by undyed PCE and
water is negligible at 600 nm, the contrast in transmitted light shown here is refraction at the water-oil
interfaces. DNAPL saturation patterns estimated from these images are also shown in Figure 4.2-1
The correlation between PCE volumes calculated from image analysis and actual PCE volumes within the
flow chambers is shown in Figure 4.2-2. The visualization technique provided good estimates of PCE mass
for both undyed and dyed PCE.
Table 4.2-2. Summary of flow chamber experiments.
Experimental
Code
Pack A
PackB
PackC
PackD
PackE
PackF
PackG
Dyed PCE
No
No
Yes
Yes
Yes
Yes
Yes
Chamber
Size
Small
Small
Small
Small
Large
Large
Large
Incremental
addition/removal
Addition
Addition
Addition
Addition
Removal
Removal
Removal
Imaging Wavelength
(nm)
600
600
600
543
600
600
543
-------
40 60 80 100 120 140
X-Axis (mm) (a)
,§
-------
4.2.2 Relationship between Source Mass and Source Strength
Given the limitation of remedial technologies to meet acceptable water quality standards, particularly with
respect to DNAPLs, the benefits of conducting costly remedial measures are unclear (EPA, 2003; NRC,
2005). A number of source-treatment benefits have been suggested, and one such benefit is a reduction
in contaminant discharge from the DNAPL source area (Soga et al., 2002; Rao et al., 2002; Soga et al.,
2004). If the flux which emanates from the source area is reduced by source treatment, conceptually the
contaminant plume may respond in a manner which reduces the contaminant risk, such as a reduction
in total plume mass or a reduction in plume spatial extent. The flux response (Figure 2.0-2) to source
treatment is critical to understanding the benefits of source treatment in terms of plume response.
4.2.2.1 DNAPL source depletion: Linking architecture and flux response
The relationship between DNAPL mass reduction and contaminant mass flux was investigated
experimentally in four model source zones (Pure et al., 2006). The flow cell design for the experiments
featured a segmented extraction well that allowed for analysis of spatially resolved flux information. The
well segments were numbered sequentially from the bottom of the flow chamber upwards (i.e. the bottom
segment is referred to throughout as port 1 where the top port is referred to as port 7). This flux information
was coupled with image analysis of the NAPL spatial distribution to investigate the relationship between
flux and the upgradient NAPL architecture. Three experiments were conducted with 1,2-dichloroethane
(DCA) and one with trichloroethene (TCE). DCA was selected for its high aqueous solubility so that the
experiments could be conducted in a reasonable time frame. A summary of the experimental conditions can
be found in Table 4.2-3. For brevity the four experiments will hereafter be referred to as DCA-1, DCA-2,
DCA-3, and TCE-1.
A simple image analysis technique was utilized to provide semi-quantitative assessment of NAPL
architecture. Reflective digital images of the flow chamber were taken at each sampling acquisition. Note
that images were taken from only one side of the flow cell but compared favorably (in terms of similar
NAPL distributions) with the side of the flow cell that was not imaged. The images were overlain with a
finely discretized grid (grid cell size approximately 0.2 cm2) for image processing. A binary map of the
NAPL distribution was developed based on the presence (in which case the grid cell was filled black) or
absence (in which case the grid cell was filled white) of NAPL in each individual grid cell. The black and
white image was imported into Mathcad using an image processing function that returns matrix containing
distinct numeric values for white (no NAPL) and black (NAPL) grid cells.
Table 4.2-3. Summary of experimental conditions.
Name
Volume DNAPL injected (ml)
Injections
Flow chamber design
Packing
Pore water velocity (cm/hr)
Experiment 1
DCA-1
10
Single
Segmented
Discrete layers
22.9
Experiment 2
TCE-1
10
Single
Segmented
Discrete layers
18.7
Experiment 3
DCA-2
12
Double
Segmented
Continuous
layering
18.2
Experiment 4
DCA-3
10
Single
In-line
Fine gradation
17.3
Trajectory integrated NAPL content, S, is defined as S = (SNr\) I9 (where §N is the trajectory average
NAPL saturation, r| is the porosity [L3L~3], 9 is the water content [L3 L~3]). Trajectory integrated NAPL
contents were estimated by assuming horizontal flow trajectories, such that a trajectory could be represented
by one row in the grid. Note that the assumption of horizontal flow trajectories is not a requirement of the
streamtube model, but rather a simplification that was necessary in order to use image analysis to estimate
trajectory integrated NAPL contents. A simple algorithm was used to count the number of elements in
-------
each row of the matrix that had a numeric value corresponding to the positive identification of NAPL.
This process returned an array containing the number of NAPL containing grid cells in each trajectory. By
assuming a uniform saturation for each grid cell, the mass fraction in each trajectory was estimated as the
number of NAPL-containing grid cells in each row divided by the total number of NAPL-containing grid
cells in the domain.
Mass removal via aqueous dissolution as a function of time was determined by integrating the dissolution
breakthrough curve. Mass balances for all experiments were in the range of 80-90%. Possible sources of
mass balance errors include volatilization during sampling, inaccuracies in the measurement of flow, and
residual NAPL remaining in the flow cell upon completion of the experiment. Fractional mass reduction
(i.e., mass removal compared to initial mass) was evaluated without attempting to correct or add back in the
various experimental mass losses.
An example of the flux plane response to temporal changes in the NAPL architecture is shown in
Figures 4.2-3. The term 'flux plane response' is used to refer to the distribution effluxes at the down-
gradient segmented extraction well. The bar graphs plot the fraction of the total initial flux discharged
from individual extraction well segments. Results are shown for six selected time steps. Also reported for
each time step is the fraction of the total initial mass remaining (FR) and the fraction of the total initial
flux (Fj), where FR is defined as the total NAPL mass in the system at a given time step divided by the
total initial NAPL mass in the system at the beginning of the experiment. Likewise Fj is defined as the
total flux out of all well segments at a given time step divided by the total initial flux at the beginning of
the experiment. The segmented flux data is intended to be coupled with information about the up-gradient
NAPL distribution to investigate the linkages between NAPL architecture and flux.
Figure 4.2-3. Flux plane response to changes in NAPL architecture for experiment DCA-1. Note that FR is
the mass fraction ((mass at time ^/(initial mass att = 0)) and F} is the flux fraction ((flux at
time 0/(initial flux at t = 0)). Figures a through f show NAPL distribution and vertical flux
distributions (bar graphs) at progressively longer dissolution times (Pure et al., 2006).
-------
For discussion purposes it is useful to consider each well segment (or port) as corresponding to a bundle of
streamtubes with a given distribution of trajectory integrated NAPL contents. It is also useful to recall that,
based on the image analysis procedure and the definition of S, an elongated NAPL morphology in the mean
flow direction translates to a larger estimated trajectory integrated NAPL content, (e.g., a NAPL pool would
have a much larger estimated value of S than a vertical finger).
Figure 4.2-3 indicates that reductions in flux were realized for partial reductions in mass as certain well
segments showed flux decreases at earlier dissolution times than other well segments. The well segments
that continued to produce mass at later dissolution times were predominantly associated with streamtube
bundles characterized by larger S values (NAPL morphologies that were elongated in the mean flow
direction). The well segments that were depleted of mass at earlier dissolution times were predominantly
associated with streamtube bundles characterized by smaller S values (NAPL morphologies that were
'less elongated' in the mean flow direction). In summary, qualitative linkages between the flux plane
response and upgradient changes in NAPL architecture indicated the important role the Lagrangian NAPL
architecture played in governing dissolution dynamics in the systems studied.
Mathematical model simulations were used to further investigate the importance of Lagrangian NAPL
architecture in controlling dissolution dynamics. Two simplified models, a stochastic-advective
(streamtube) model and a power function model were used for this investigation. An analytical solution for
flux-averaged concentration using the rate-limited streamtube model was presented by Jawitz et al, 2005
(summarized in Section 4.3.3.7):
1-
exp
V.fe
-01
Pw
^Yf^ 1 K \ / 1 —1— f^Yf^
t/A-U 1 n/.kJ-tr- \ 1 t/A-U
*Afc-0
PAT
1
(4.2-2)
where t is the travel time or residence time along a streamtube, tR is total runtime, C [M L3] is the
contaminant concentration in a streamtube, Cg [M L3] is the contaminant solubility, n is the number of
streamtubes, pN is the NAPL density, kt is a mass transfer rate coefficient that is linearly related to S, the
trajectory integrated NAPL content, ar\dfc is the fraction of streamtubes that are contaminated.
If equilibrium mass transfer is assumed, Eq. (4.2-2) reduces to:
M
c.
(4.2-3)
where H is the Heaviside step function, R is the retardation factor, and i is the reactive travel time which
represents the combined effects of aquifer hydrodynamics and NAPL spatial distribution on NAPL
dissolution (Cvetkovic et al., 1998; Jawitz et al., 2003, 2005).
The power function model used in this study relates source zone effluent concentrations to reductions
in NAPL mass. The model was presented by Parker and Park (2004) and is based on the concept of an
effective Damkohler number (Da= KeffL I q , where Key \_T ' J is the effective mass transfer coefficient, L is
the source zone length in the mean flow direction, and ^[Lr"1] is the average Darcy flux for the source
zone). For brevity this model will be referred to as the Da model. The concept of the Da model is to relate
the field-scale mass transfer rate coefficient to the average groundwater velocity and to the temporally
changing (decreasing) global NAPL mass (Parker and Park, 2004):
-------
Cr
(4.2-4)
(4.2-5)
where C} out [ML"3] is the flux averaged concentration exiting the source zone, ^S[LT1] is the average
saturated hydraulic conductivity of the source zone, Ms(t) is the NAPL mass in the source zone at time t,
Msg is the initial NAPL mass in the source zone, and KO [T1], Pp and (32 are empirical fitting parameters.
If p\= 1, the flux-averaged concentration exiting the source zone is independent of the mean groundwater
velocity and the Da model is similar to the equilibrium streamtube model (Eq. 4.2-3).
The fraction of the total initial NAPL mass contained in each grid row (considered here as streamtubes) was
estimated from image analysis of experiments DCA-1, TCE-1, and DCA-2. The fraction contaminated was
converted to values of S using porosity and water content and inserted into Eqs. (4.2-2) and (4.2-3). The
data were modeled under the assumption that the dissolution dynamics were primarily controlled by the
S distribution. A nonreactive travel time, t, equal to the mean nonreactive travel time for the entire system
was assumed for each streamtube (measured mean travel times for the three experiments were 1.2 hr for
DCA-1, 1.7 hr for TCE-1, and 1.3 hr for DCA-2). The superposition of n solutions, with n being the number
of trajectories from the image analysis procedure, was then used to develop the plots in Figure 4.2-4.
DCA-1
Rut limlw* SBWBHitoe
fc m S J/tll
ie. OJ2
DCA-1
tqulllbflum
ft m O.a
TCI-I
liquHIMun MninKubt
OtA-J
[quHibnum SEi»«au
Ir - fl JS
* *•*
Dissolution Time (hrs)
Figure 4.2-4. Comparison of the rate-limited streamtube, equilibrium streamtube, and effective
Damkohler approaches for modeling source depletion from experiments DCA-1, TCE-1,
and DCA-2 (Pure et al., 2006).
-------
The nonequilibrium model required fitting one parameter (kt) and the Da model required fitting 2
parameters (P15 (32). Two approaches were available for estimating^. In the first approach, image analysis
was used to determine the fraction of streamtubes that contain NAPL. The second approach used the
initial flux-averaged concentration (Cg). For example, if the initial flux averaged concentration was 50%
of solubility,^ = 0.50 could be assumed. For the equilibrium streamtube model, estimating fc , based on
the number of trajectories containing NAPL, resulted in an overestimation of Cg. Explanations for the
overestimation of Cg when using an^ obtained from image analysis include velocities high enough to
induce rate-limiting effects in the smaller trajectories and incomplete span of the NAPL across the thickness
of the flow chamber in certain locations, which would result in bypass flow that was not captured in the
two-dimensional image analysis technique. After estimating fc from Cg, and using NAPL distribution
information obtained from the image analysis, the equilibrium streamtube models matched the observed
dissolution behavior closely (Figure 4.2-4), which supports the argument that the NAPL architecture
was the primary factor controlling dissolution behavior in the systems studied. Additional processes not
explicitly accounted for in the model, such as relative permeability, transverse dispersivity, and velocity
variability are thought to exert a secondary influence on the dissolution behavior of the systems studied.
Although KO in the Da model is treated in Parker and Park (2004) as a fitting parameter, we suggest that
the physical meaning of KO is similar to fc and is primarily a measure of the fraction of streamtubes
intersecting the contaminated portion of the source zone. Thus, for this application KO was estimated by
fitting to C0. As shown in Figure 4.2-4, the Da model with this estimated KO and (3j = (32 =1 also provided a
good fit to the data.
In summary, results indicate that in the systems studied, the relationship between DNAPL mass reduction
and contaminant mass flux was primarily controlled by the NAPL architecture. A specific definition of
NAPL architecture was employed where the source zone is resolved into a collection of streamtubes with
spatial variability in NAPL saturation along each streamtube integrated and transformed into an effective
NAPL content for each streamtube. The distribution of NAPL contents among the streamtubes (NAPL
architecture) controlled dissolution dynamics. Two simplified models, a streamtube model and an effective
Damkohler number model, were investigated for their ability to simulate dissolution dynamics.
4.2.2.2 Fluid and porous media property effects on dense non-aqueous phase liquid migration
and contaminant mass flux
The effects of fluid and porous media properties on DNAPL migration and associated contaminant mass
flux generation were evaluated in laboratory experiments (Totten et al., 2007). Relationships between
DNAPL mass and solute mass flux were generated by measuring steady-state mass flux following step-
wise injection of PCE into flow chambers packed with homogeneous porous media. The effects of fluid
properties including density and interfacial tension (IFT), and media properties including grain size and
wettability were evaluated by varying the density contrast and interfacial tension properties between PCE
and water, and by varying the porous media mean grain diameter and wettability characteristics. Pertinent
fluid and media properties for the experimental systems investigated are shown in Table 4.2-4. The design
of the flow chamber was similar to that described by Jawitz et al. (1998), with dimensions of 14 cm by 28
cm and a pore volume of approximately 200 cm3. Down-gradient mass flux following each incremental
PCE injection was measured under steady-state water flow and after stabilization of the down-gradient
PCE concentration. At each injection step, the red-dyed PCE distribution was traced from the side of the
chamber to maintain a qualitative record of the PCE geometry. The experimentally measured contaminant
concentrations were compared to predicted values assuming both equilibrium and rate-limited dissolution.
A grid was overlain on the flow domain forming horizontal streamtubes, and the amount of DNAPL
contacted in a given streamtube (or grid row) was tabulated in a manner similar to that described by Pure et
al. (2006), and as summarized in the previous section. The equilibrium solute concentration was estimated
as simply the product of the PCE solubility limit and the fraction of the streamtubes exposed to PCE. Rate-
limited dissolution was accounted for using the modified Sherwood number (Sh') (Powers et al., 1992).
-------
Table 4.2-4. PCE/n-decane mixture densities, media sieve and grain sizes, and best fit p values for mass
loading/flux generation relationships measured in these media.
DNAPL density (g/cm3)
1.6
1.6
1.6
1.6
1.6
1.4
1.4
1.1
1.1
1.0
US Standard
Sieve No.
20/30
30/40
40/50
40/60
50/70
30/40
40/50
30/40
40/50
30/40
mean grain size (mm)
0.68
0.48
0.35
0.32
0.23
0.48
0.35
0.48
0.35
0.48
P(R2)
0.41 (0.97)
0.54 (0.81)
0.93 (0.97)
0.26 (0.91)
0.43 (0.88)
0.66 (0.96)
0.83 (0.99)
0.88 (0.97)
1.0 (0.99)
0.92 (0.98)
PCE colored with Oil-red-O dye (< 1 x 10~4 M) was injected into the lateral center of the chamber 6.5 cm
above the bottom and approximately 4 cm below the top of the sand layer. To evaluate the influence of oil/
water wettability, PCE/water interfacial tension was modified by adding 0.0025, 0.005, 0.01, 0.025, 0.05
and 0.1 percent by volume of the surfactant Span 80 to PCE. Span 80 (sorbitan monooleate) was selected
because of its low hydrophile/lipophile balance (HLB) value of 4.3. A low HLB surfactant was required to
limit partitioning of the surfactant into the aqueous phase. The effect of surfactant concentration on PCE/
water IFT was measured using a tensiometer (Tensiomat model 21, Fisher Scientific: accuracy ±0.25%).
DNAPL density was modified by using mixtures of PCE and n-decane (p = 0.73 g/cm3). The low aqueous
solubility of n-decane (0.003 mg/L) ensured limited partitioning from the non-aqueous to the aqueous
phase. The IFT of the 50/50 mole fraction of PCE and n-decane mixture (52 dynes/cm) was found to be
similar to that of PCE/water (47 dynes/cm), thus density was the primary property being varied in these
experiments.
Experiments were conducted with DNAPL/water IFT values of 3, 13, and 47 dynes/cm (Span 80
concentrations of 0.05, 0.025, and 0 volume percent) in 30/40 sand packs. In this range of Span 80
concentrations, PCE solubility increased minimally (approximately 6%). The general pattern of DNAPL
migration after initial release at the injection point was rapid migration to the base of the flow chamber
followed by lateral spreading, producing a zone of residually trapped DNAPL between the injection point
and the flow chamber base, and a pool across the base (Figure 4.2-5a). After each incremental release,
the DNAPL distribution extent was traced and the numbers provided on the figure panels represent the
cumulative number of 0.5 cm3 PCE injections. Based on observations of the PCE distribution in the flow
chamber, the extent of DNAPL lateral spreading increased slightly as IFT decreased. Pooling began by the
second injection in each experiment, indicating that DNAPL vertical migration in the 30/40 sand was not
significantly altered by reducing the IFT, and was therefore controlled by fluid density contrast.
-------
a)
H=10cm
L=18cm
/ 5
5 \
b)
c)
d)
Figure 4.2-5. Tracings of observed distributions of DNAPL injected 10 cm above the flow chamber
bottom in 30/40 sand with a) PCE and untreated sand, b) PCE and 50% OTS treated
hydrophobic sand mixture, c) PCE and 100% hydrophobic sand, d) untreated sand and
DNAPL mixture of density 1.1 g/cm3 (Totten et al, 2007).
The mass flux generation as a function of PCE loading for these experiments is shown in Figure 4.2-6.
As the IFT decreased, the effluent relative concentrations increased slightly. The maximum relative
concentration appeared to be bounded by the injection point location since DNAPL did not migrate above
the injection point. Because the media above the injection point (located at 0.6 of the aquifer height) was
uncontaminated, the effluent relative concentration would not be expected to exceed this value (assuming
DNAPL did not significantly alter the aqueous flow field).
The mass flux generation data were fit with a simple mass reduction, RM, flux reduction ^, power
relationship Rj = R^p resulting in Pp = {0.54, 0.50, 0.67} with R2 = {0.80, 0.95, 0.98} for the 47, 13 and
3 dyne/cm systems, respectively. The entire data set was fit in Figure 4.2-6 with an average Pp = 0.5
(R2=0.82). These results indicate that over the range of IFTs tested, relatively minor changes ( p =0.5-0.67)
were observed in the characteristics of the mass flux generation with PCE loading. Interfacial tensions
lower than those tested here may be required to affect migration or prevent entrapment, which could
significantly reduce mass flux.
-------
0.60
0.50
• Untreated (47 dynes/cm)
• 13 dynes/cm
A 3 dynes/cm
Best Fit, Beta=0.5
0.00
0.00
0.20
0.40 0.60
PCE/PCEmax
0.80
1.00
Figure 4.2-6. Fractional flux increase versus DNAPL loading for each interfacial tension value in
untreated 30/40 sand (Totten et al., 2007).
Fluid density, modified using different ratios of PCE and n-decane, indicated that as density difference
decreased, vertical migration was reduced and lateral migration was enhanced. This trend is illustrated
in Figure 4.2-5a (pure PCE with p =1.6 g/cm3) and Figure 4.2-5d ( PCE and n-decane mixture with
p = 1.1 g/cm3). In experiments with PCE, pooling usually began by the second injection (1.0 cm3),
indicating complete vertical migration from the injection port to the bottom of the chamber. The lower
density experiments required as many as nine injections (4.5 cm3) to cause full migration from the injection
port to the bottom of the experimental chamber and in the case of the p = 1.0 g/cm3 experiment, no pooling
occurred, indicating no full downward migration.
The PCE relative concentrations at final PCE loading were similar for all experiments, with the lower
density tests generally exhibiting slightly higher values (Figure 4.2-7). These results are likely due to the
observed migration of DNAPL above and below the injection port (Figure 4.2-5d), providing a larger cross
sectional area exposure to flow.
0.60
0.50
0.40
i
50.30
0.20
0.10
• 30/401.65 g/cm3
• 30/401.4
A 30/401.1
D 30/401.0
X 40/501.65
• 40/501.4
+ 40/501.1
Fit, Beta = 0.43
— • Fit, Beta = 0.85
0.00
0.00
0.20
0.40 0.60
PCE/PCEma»
0.80
1.00
Figure 4.2-7. Fractional flux increase versus PCE loading for varying density and media (Totten et al.,
2007).
-------
The Pp values that were fit for each DNAPL density experiment are provided in Table 4.2-4. The mass
loading/flux generation relationship became more linear as the density contrast decreased. The trend was
particularly evident in the 30/40 sand. The shape of these curves, coupled with the DNAPL tracings,
indicate that reduced vertical migration following each 0.5 cm3 injection altered the mass flux generation
relationship. Rapid downward migration, typically for higher density contrasts (i.e., pure PCE) resulted
in rapid mass flux increases at low DNAPL loading when compared to the slower downward migration
associated with lower fluid density contrasts. How mass becomes distributed as ganglia and pools, and how
uniformly distributed those domains are, define the initial system dissolution characteristics.
The mass load and mass flux relationships measured in the laboratory experiments were compared to
predictions from equilibrium and non-equilibrium dissolution models based on DNAPL contact lengths
and Sherwood number correlations applied within 0.5 cm horizontal streamtubes through the flow domain.
For the 20/30 sand, substantial differences were observed between the equilibrium and non-equilibrium
predictions. The short length of the source zone in the groundwater flow direction may not have allowed
sufficient contact time for equilibrium to have been achieved. In contrast, the longer source zones generated
in the 100% hydrophobic sand (Figure 4.2-5c) were likely sufficient to achieve equilibrium. However,
in both cases the measured relative concentrations were lower than those predicted, even under non-
equilibrium conditions.
The model observations suggest that flow by-passing likely occurred in both experiments with the
hydrophobic sand results comparing more favorably than the untreated sand. The hydrophobic media likely
had less by-passing due to the more-uniform distribution of the PCE resulting from capillary wicking. The
simple model applied here does capture the general shape of the flux generation mass loading relationships,
suggesting that the distribution observed on the chamber wall reflects the general distribution of the DNAPL
in the chamber. Application of the mass transfer correlations developed for homogeneously contaminated
systems to more complex flow and contaminant distributions may be problematic given the lack of
knowledge of water flow paths and DNAPL distributions within those flow paths. The predicted mass flux
generation characteristics could be improved with laboratory characterization techniques that account for
the DNAPL distribution across the thickness of the flow chamber (e.g., light transmission visualization
techniques). While this may provide greater understanding in laboratory systems, field scale assessment
may have to rely on tracer approaches to characterize these relationships.
4.2.2.3 Reactive tracer tests to predict dense non-aqueous phase liquid dissolution dynamics in
laboratory flow chambers.
Reactive (i.e., partitioning) tracer tests were conducted to evaluate the relationship between contaminant
mass reduction, RM, and flux reduction, Rj in laboratory experiments with porous media contaminated
with a DNAPL (Chen and Jawitz, 2008). The reduction in groundwater contaminant flux resulting from
partial mass removal was obtained from cosolvent and surfactant flushing dissolution tests in laboratory
flow chambers packed with heterogeneous porous media. Using the concept of streamtubes, a Lagrangian
analytical solution was applied to study the contaminant dissolution. Analytical solution parameters
related to aquifer hydrodynamic heterogeneities were determined from a nonreactive tracer, while those
related to DNAPL spatial distribution heterogeneity were obtained from a reactive tracer. The parameter of
reactive travel time variability was derived from this combination of tracers and can be used to predict the
relationship between RM and Rj.
Tetrachloroethylene (PCE) was selected as the DNAPL in this experiment. Dissolution experiments were
conducted with one cosolvent flushing test (50% ethanol/water) and three surfactant flushing tests (2%
Tween-80/water) to evaluate the effects of contaminant mass depletion on flux reduction. The four sets
of experiments were conducted in two-dimensional flow chambers, as summarized in Table 4.2-5. The
flushing solutions were selected based on their relatively high PCE solubilization capacity (~104 mg/L
for both cosolvent and surfactant solutions) and moderate PCE-water interfacial tension (-2.4 dyn/cm
for cosolvent solution and -5.0 dyn/cm for surfactant solution respectively), which promoted PCE
solubilization with minimal risk of mobilization.
Two flow chambers, similar to the design of that described by Jawitz et al. (1998), were employed in this
study. The PCE was injected into the upper portion of the domain at 0.5 mL/min by a syringe pump.
-------
Following redistribution of the injected DNAPL, approximately 10 pore volumes (PVs) of de-ionized
water were displaced continuously through the flow chamber and flux-averaged PCE concentrations were
measured in the effluent. Tracer tests were then conducted with methanol as the non-partitioning tracer, and
2,4-dimethyl-3pentanol (2,4DMP) n-hexanol, and 2-octanol as partitioning tracers. Approximately 9 PVs
of the cosolvent solution (experiment I) and 14 PVs surfactant solution (experiment II) were continuously
pumped through the chamber until either 97% of the PCE was removed or the effluent concentration was
lower than the gas chromatograph (GC) detection limit. For experiments III and IV, the remedial solutions
were pumped for 1 to 2 PVs and then changed back to water flushing until the PCE effluent concentration
was constant, and then another 1 to 2 PVs of remedial solution were injected, again followed by water
flushing. This pulsed flushing was repeated for 4 to 7 cycles. The PCE mass removal as a function of time
was determined by integrating the dissolution breakthrough curves. PCE recovery rates for all experiments
were in the range of 80-97%.
Table 4.2-5. Summary of the four experiments, equipment and truncated moment analysis results.
remediation solution
flow rate (mL/min)
flushing
Pore volume (mL)
PCE injected (mL)
Experiment I
cosolvent
5.5
single
950
11.95
Experiment II
surfactant
3.2
single
310
11.58
Experiment III
surfactant
7.5
pulse
1200
8.00
Experiment IV
surfactant
3.2
pulse
305
9.12
Under equilibrium conditions, the flux-averaged concentration, C}, exiting the source zone can be
expressed as a function of flushing duration t (Jawitz et al., 2005):
fQ,t
-------
Experiment I
Experiment II
Figure 4.2-8. PCE spills before flushing. Images from experiments 1 and 2 are reflected light captured
with standard digital camera and the NAPL is dyed red. Images from experiments III and
IV are light transmission images captured with a CoolSnap 16-bit digital black-and-white
camera (Chen and Jawitz, 2008).
-------
Experiments I and II were conducted by continuous flushing by cosolvent and surfactant respectively. The
measured dissolution breakthrough curves (BTCs) for PCE for these continuous flushing experiments are
shown in Figure 4.2-9. The BTCs were predicted from the equilibrium dissolution equation 4.2-6 using the
parameters obtained from the tracer tests. In the equilibrium streamtube model, the DNAPL architecture is
the primary factor controlling the dissolution behavior, and other processes such as transverse dispersivity
and relative permeability changes during dissolution are assumed to be negligible. The average relative
errors between the model predictions and measured data for experiments I and II were 0.12 and 0.13
respectively.
Experiment I
experimental
streamtube model
P.V.
I
(I.S
0.6-1
0.4
0.2-1
0
Experiment II
experimental
streamtube model
ill
P.V.
15
20
Figure 4.2-9. BTCs from surfactant flushing for Experiments I and II (Chen and Jawitz, 2008.
For the pulsed flushing experiments III and IV, the concentration of the injected surfactant solution C(f) in
equation 4.2-6 changed with time. It is difficult to simultaneously measure concentration of the surfactant
and PCE by the inline GC. Assuming negligible partitioning of the surfactant into the PCE, the surfactant
breakthrough curve would exhibit the same shape as the non-partitioning tracer.
The concentration of the surfactant may be expressed as:
l=!«/l
(4.2-7)
where |o,]ni and otnt were attained from non-partitioning tracer tests, and tg is the pulse duration of multiple
flushing (PV).
-------
Several studies have indicated that the equilibrium solubility of PCE exhibits a linear relationship with
aqueous surfactant concentration (Edwards et al., 1991; Pennell et al., 1993; and Johnson and John, 1999).
Batch solubility tests showed that PCE solubility increased linearly for surfactant (Tween 80) concentrations
between 1000 and 40,000 mg/L. The observed linear relationship was:
Cs=0.227Cf+\60
(4.2-8)
The concentration of PCE was predicted dynamically using equations 4.2-7 and 4.2-8 (results from experiment
III are shown in Figure 4.2-10). The average relative errors for experiments III and IV were 0.12 and 0.33
respectively.
Kxpctimi'iit III
JMIU -
3
^ 2000-
•a 1500 -
A
h
fl
S 1000-
o
o
U 5UO
eu
M .
1
, ,
c (
1
l*^|
,
^_ J
I
£
1
0 t\|HI IIIH llt.ll
- strcainliihi1 model
L
V I
^
l d
w. L
10
2(1
.10
4(1
5(1
P.V.
Figure 4.2-10. BTCs from surfactant flushing for Experiment III (Chen and Jawitz, 2008).
In summary, non-partitioning and partitioning tracers were applied to predict the DNAPL dissolution
dynamics and mass reduction and flux reduction relationships using an equilibrium streamtube model. A
single measurable parameter, reactive travel time variance, was applied to describe the overall system
heterogeneity including both media heterogeneity and DNAPL architecture. Increasing the reactive travel
time variance generated increased breakthrough curves spreading and shorter flushing duration to reduce
DNAPL mass flux, which in turn leads to more favorable relationships between mass reduction and flux
reduction.
4.2.2.4 Laboratory investigation of flux reduction from dense non-aqueous phase liquid (DNAPL)
partial source zone remediation by enhanced dissolution
This study investigated the benefits of partial source zone removal using enhanced dissolution (Kaye et al.,
2008). The benefits were assessed by characterizing the relationship between reductions in DNAPL mass
and the corresponding reduction in contaminant mass discharge in several laboratory scale experiments. In
these experiments, the effects of fluid override and underride associated with cosolvent flushing on the mass
reduction (RM) vs. flux reduction (Rj) relationship were investigated. Mass reduction and flux reduction
are defined as I--TTL and U—j-\, where M is DNAPL mass at time t, M is the initial DNAPL mass,
V ™s,o/ V Jaj
Jt is contaminant mass flux at time t, and J() is the initial contaminant flux. Experiments were conducted
using a single flushing event to remove most of the DNAPL from the system or in multiple shorter duration
floods to determine the path of the RM and Rj relationship.
-------
Three unique cosolvent mixtures and one surfactant mixture were each tested in a two-dimensional
heterogeneous aquifer model. Accounting for spatial heterogeneity of both media and DNAPL saturation
is critical to the study of source zone removal. Two-dimensional flow chamber experiments allow for the
emplacement of heterogeneous distributions of porous media. Tetrachloroethylene (PCE) was used as a
model compound because of its ubiquity as an environmental contaminant and its relatively low aqueous
solubility compared to other DNAPL contaminants. A lower-solubility DNAPL was preferred for enhanced
dissolution laboratory experiments to minimize source zone depletion during water flow periods.
Eight solubilization experiments were conducted in the two-dimensional flow chambers packed with
porous media and contaminated by PCE. As in previous studies, Accusand (Unimin Minnesota Corp.), a
translucent porous media, was used for all of the experiments (Pure et al., 2006). The domain was packed
with 20/30 sand as the main media with lenses of 40/50 sand throughout the domain. It is estimated that
90% of the flow field was in the 20/30 sand and 10% was in the 40/50.
The cosolvents used in the experiments were reagent alcohol (RA) (90.5% Ethanol, 4.5% Methanol, 5.0%
Iso Propyl Alcohol) and ethyl-lactate (EL). Three different cosolvent mixtures (50% RA/50% Distilled
Ionized (DI) water; 40% EL/60% DI water; and 18% RA/26% EL/56% DI water) and one surfactant
mixture (2% Tween-80 by weight in DI water) were used in both single- and multiple-flushing experiments
(eight experiments total). The cosolvent mixture percentages presented here are volume based. The
mixtures will be referred to as 50% RA, 40% EL, Neutral Density (ND) and Surfactant (NDS) respectively.
The physical and chemical properties of the mixtures are presented in Table 4.2-6. The mixtures were
chosen because they have similar PCE solubilities, but significantly different densities.
Table 4.2-6. Fluid properties.
Density, p
Viscosity
PCE Solubility,
(g/cm3)
' (cp),
C^mg/L)
Water
1.000
0.895b
150b
50% RA
0.89
2.386a-b
10,003
40% EL
1
2
8
.014
.238
,881
ND mixture
1
2
8
.001
.369
,850
NDS
1.
0
10,
00
.9
000
1 Van Valkenberg, 1999
» Lide, D.R. and Frederikse, H.P.R., 1996
In this study two modes of injection were employed. In the single-flushing experiments (SF), each mixture
was continuously flushed through the contaminated flow chamber to achieve more than 90% DNAPL mass
removal. The Darcy velocity was approximately 1.3 m/d. Steady-state aqueous concentrations and mass
fluxes were measured both before and after the source zone removal. After aqueous flushing, the remaining
mass was recovered by enhanced dissolution to complete a mass balance. After all of the contaminant mass
had been removed from the chamber, the flow characteristics were determined using a dye tracer. If the
characteristics were favorable (i.e., no dead zones) PCE was injected into the media and a new experiment
was started. If they were unfavorable, the chamber was taken apart, cleaned, and repacked.
In the multiple-flushing experiments (MF), the duration of each flushing episode was limited so that
only a portion of the DNAPL mass was removed. The spatial distribution of the remaining DNAPL and
corresponding integrated down-gradient contaminant mass flux were measured under steady-state aqueous
conditions. Then, additional flushing agent was injected to further reduce the contaminant mass, and
again, the steady-state aqueous contaminant mass flux was measured. This process was repeated until
approximately 90% of the total mass was removed or the initial mass flux was reduced by more than 90%.
The goal was to have either four or five consecutive injections that would achieve approximately 20%,
40%, 60%, 80%, and above 90% total contaminant mass removal and to measure the aqueous steady-
state contaminant mass flux at each step. After the initial mass or flux was reduced by more than 90%, the
remaining mass was removed and quantified to complete a mass balance.
-------
Figures 4.2-11A B & C show the behavior of the cosolvent mixtures displacing the resident water in the
chamber. These figures were created by tracing the dyed fluid front (shown as green lines in the figures).
Dyed fluid fronts and DNAPL distributions were delineated by collecting digital images during the
experiments. To improve resolution, a light transmission visualization (LTV) approach was used, where a
light source was mounted on the back side of the flow chamber.
Figure 4.2-11. Displacement of resident water at different cosolvent injection volumes (PV). A) neutral
density cosolvent, B) 40% EL, and C) 50% RA. The flow was from left to right at a Darcy
velocity of approximately 1.3 m/d (Kaye et al., 2008).
A relatively vertical displacement front was observed for the ND solvent while significant underride and
override were observed during the 40% EL and 50% RA floods, respectively. The mixing zone of the
displacing front of all the mixtures was approximately 2 cm wide. The 50% RA reached the bottom of the
-------
chamber after approximately 1.5 PVs. However, based on the effluent cosolvent concentrations, it was
estimated that it did not completely displace the water at the bottom of the chamber until approximately
2 PVs had been injected. The 40% EL never completely displaced the resident water during the
experiment; water remained in the top effluent corner of the chamber after almost 7 PVs of cosolvent
had been injected. However, the 40% EL front reached all of the PCE contaminated areas, both residual
and pooled, after only about 1.2 PVs. The angle from the horizontal of the 50% RA mixture override
approached zero with continued cosolvent injection, while the angle from the horizontal made by the 40%
EL mixture underride of the water stayed nearly constant from approximately 0.5 to 7 PVs.
Breakthrough curves are shown in Figure 4.2-12 for PCE and the cosolvents for the single injection flushing
experiments. Visual images captured during the RA flood indicate that after injection of approximately
2.7 PVs of the cosolvent, most of the PCE mass that had been removed was from the residual portions
above the lenses. Also at that point, the cosolvent had just reached the bottom of the chamber because of
override (Figure 4.2-llc), and there was little change in the PCE mass in the lower portions of the chamber.
However, at 3.3 PVs, the cosolvent had removed much of the residual above the bottom pool. Toward the
end of the continuous injection, only pools of PCE remained above the top lens and at the bottom of the
chamber. The cosolvent injected between 6.1 and 8.6 PVs removed most of the PCE above the top lens, but
relatively little from the bottom pool. The lack of removal of the pool at the bottom of the chamber is due to
a combination of the density override and the higher saturation of PCE in the pools.
1.2 -i
RA-SF:50%RABTC
EL-SF:40%ELBTC
ND-SF: ND cosolvent BTC
RA-SF: PCE BTC
EL-SF: PCE BTC
ND-SF: PCE BTC
- RA-SF: UTCHEM
Figure 4.2-12. BTCs for all Single Flush experiments with cosolvent mixtures (Kaye et al., 2008).
Flux reduction that was based on the aqueous flushing data collected before and after the single-flush
experiments yielded only a single point on &RM vs. Rj plot (Figure 4.2-13). The multiple-flushing
experiments were designed to capture more data on this curve by conducting sequential single floods, and
resulted in 4 or 5 additional points on the Rj (RM) graph for each experiment. The Rj was based on the
steady-state aqueous PCE concentrations measured after each flushing compared to the initial steady-state
concentration.
The entire Rj(RM) path was also estimated from the SF experiments by approximating equivalent
concentrations during aqueous flushing with a normalized concentration, C as follows:
CN=C(t}-
c
(4.2-9)
-------
Where C ft) is the effluent PCE concentration at each time, Cw is the aqueous PCE solubility, and Csfft)
is the PCE solubility in the flushing solution at each time. The latter term is a different function for each
flushing agent. The estimated Rj(RM) paths of the SF experiments (designated SSA in Figure 4.2-13) all
followed the same favorable trend just above the 1:1 line.
Several conclusions can be drawn from these results: 1) the experimental DNAPL architectures that were
carefully constructed to be consistent between experiments resulted in similar Rj(RM) behavior, 2) partial
source zone remediation using enhanced dissolution in these similarly packed systems resulted in flux
reductions approximately proportional to the mass reductions, and 3) the override and underride associated
with a single injection of cosolvent did not greatly influence the Rj(RM) relationship.
The RA-MF and EL-MF experiments resulted in less-favorable Rj(RM) compared to the SF experiments
with the same fluids (Figure 4.2-13). The repeating override and underride associated with multiple
injections of these flushing agents was the primary difference between the SF and MF experiments. The
NDS-MF data support this observation as the Rj(RM) data (not shown) generally follow the SF data from
all the flushing agents. Because of the stable displacement of NDC, multiple injections of this flushing
agent were also expected to closely approximate the SF experiment Rj(RM) paths. However, the NDC-MF
data matched the data from neither the SF nor the other MF experiments. This is likely explained by the
different initial NAPL distribution in this experiment.
All Cosolvent Flushing Experiments
RM vs. Rj
0.1 0.2 0.3 0.4 0.5 0.6 0.7
Fractional Mass Reduction
0.8 0.9
* 50% RA-SF
D 40% EL-SF SSA
AND-MF
• 40% EL-SF
A ND-SF SSA
X SF-MF
A ND-SF
O 50% RA-MF
SF-SF
» 50% RA-SF SSA
D 40% EL-MF
Figure 4.2-13. Aqueous based mass reduction ( RM ) versus flux reduction ( Rj) of the single and multiple-
flushing experiments with the solubility scaled approach (SSA) path estimation of the
single-flushing experiment (Kaye et al., 2008).
-------
The RMVS. ^relationships of all of the single-flushing experiments were nearly identical, and in all of
these experiments the initial DNAPL distributions were similar. Together, these observations indicate
that the override and underride effects associated with the 50% RA and the 40% EL during miscible fluid
displacement did not significantly affect the remediation performance of the agents during the single-
flushing experiments. The Rj (RM) relationship of multiple injection experiments with both 50% RA and
40% EL were less favorable in the sense that there was less Rj for a given RM than in the single injection
experiments with the same flushing agent; however, the initial DNAPL architectures varied somewhat in
these experiments. To eliminate the effects of different initial NAPL distributions, UTCHEM simulations
were conducted for four cases (RA-SF, RA-MF, ND-SF and ND-MF) with a single initial NAPL
distribution.
The ability of UTCHEM to capture the dissolution and density effects was demonstrated by simulating
the RA-SF case. The DNAPL and media distribution input data for the model were created using the
images from the light transmission measurements. The images were overlain with a finely discretized
grid and saturation values were assigned based on the color intensity, such that the total mass of DNAPL
in the domain was equal to the known mass injected. The details of the simulation domain and the fluid
characteristics are presented in Table 4.2-7. The bimodal curve data presented by Hayden et al (1999)
were used to estimate a bimodal curve height of 0.83. The numerical simulations used the same boundary
conditions and flows as the RA-SF experiment.
The simulated ethanol and PCE breakthrough curves were in good agreement with observed laboratory
data and provided confidence in the ability of the numerical simulator to capture the enhanced dissolution
process and the density effects. The same DNAPL distribution was then used to simulate the RA-MF,
NDC-SF, and NDC-MF cases. The results indicated that the Rj (RM) relationship was affected only slightly
by the displacement fluid density differences associated with a single injection of cosolvent, and somewhat
more so by the repeating override and underride associated with multiple injections but the latter effect was
not as significant as that suggested by the experimental data.
Table 4.2-7. Numerical simulation input parameters.
Parameter
Value
Units
Domain Characteristics
Grid Spacing: Ar
Grid Spacing: Ay
Grid Spacing: Az
Longitudinal Dispersivity1
Transverse Dispersivity1
Porosity
NAPL Saturation
0.005
0.015
0.005
0.0017
0.00017
0.383
0.07-0.25
m
m
m
m
m
Fluid Characteristics: Density
Density of water
Density of PCE
Density of ethanol
1.0
1.625
0.785
g/cm3
g/cm3
g/cm3
Density of water-ethanol mixtures: linear function of density of water and ethanol
-------
Parameter
Value
Units
Fluid Characteristics: Viscosity
Dynamic viscosity of water
Dynamic viscosity of PCE
0.895
0.866
cp
cp
Dynamic viscosity of water-ethanol and water-ND solvent mixture:
c , where/, is the cosolvent fraction2
Fluid Characteristics: Solubility Parameters3
Aqueous Solubility of PCE
Maximum height of bimodal curve
Oil concentration at Plait point
150
0.8306
0.8554
mg/L
v/v
v/v
Fluid Characteristics: Interfacial tension parameters"'
Interfacial tension:
/"IW
is water concentration in NAPL phase (v/v),
nw is the NAPL concentration in the water phase, and
*°
P is the cosolvent concentration in the cosolvent poor phase
Interfacial tension between water and PCE in absence of cosolvent,
IFT0
Exponent describing effect of cosolvent, kLp
Mutual solubility term in absence of cosolvent, XQ
40
1.98
-3.82
dynes/cm
1 Dispersivity values were estimated by fitting non-reactive tracer breakthrough curves
2 Coefficients determined by fitting the viscosity profile measured in the laboratory
3 UTCHEM assumes that the bimodal curve is symmetrical
"Li andFu (1992) and Liang andFalta (2008)
4.2.2.5 The impacts of partial remediation by sparging on down-gradient DNAPL mass discharge
This study focused on assessing the impacts of DNAPL partial remediation by sparging on down-gradient
mass discharge using two-dimensional (2-D) flow chambers packed with silica sands (Bob et al., 2009).
The DNAPL source zone was created by adding known amounts of PCE to the chamber in two different
experiments, and incremental removal of PCE mass from source zone was achieved by sparging. Following
each sparging experiment, the chamber was flushed with water for a sufficient time to re-establish a steady
state condition and effluent aqueous samples were collected and analyzed for dissolved PCE concentration,
which was used to calculate the contaminant mass discharge. Image analysis based on LTV was employed
to measure PCE saturation distribution in the chamber before and after each remediation step.
The flow chamber used in this study had nominal internal dimensions of 48.26 cm by 48.26 cm by 1.40 cm,
and was constructed using a square aluminum frame (1.27 cm X 1.27 cm) and 1.27 cm thick tempered glass
plates. Influent and effluent conditions were designed to essentially represent fully screened wells across the
~48 cm height of the chamber. A 6-way sampling valve was connected to the effluent line to minimize PCE
-------
volatile losses during aqueous sample collection. In addition to the ports used to inject the PCE, the top
section of the frame had four additional ports with 100-micron frits for capture of vapor phase PCE during
sparging. The stainless steel tubes extending from each of these four ports were all merged into one effluent
stainless steel tube line that was connected to a GC/MS instrument that was used to monitor vapor-phase
PCE concentration (Fine et al, 2008). Using a tee split, this effluent line was also connected to a carbon
column that captured all PCE vapor. The carbon was analyzed by extraction as a secondary check for PCE
mass balance. The bottom frame had three equally-spaced ports that were used for sparging.
The chamber was packed dry with Accusand silica sands (Unimin, Le Sueur, MN). The whole model was
packed with 20/30 sand to achieve a homogeneous packing for the first experiment (referred to as Pack A).
For the second experiment (Pack B), a mixture of 20/30 and 40/60 sand at a ratio of 3:1 was used to achieve
a homogeneous pack with a broader sand size distribution.
In both experiments, the chamber was first flushed with carbon dioxide to facilitate subsequent water
saturation, then fully saturated with de-aired water and finally HPLC grade (>99.9% purity) PCE (Sigma-
Aldrich Corporate, MO) dyed with Oil-Red-O (Fisher Scientific, Fair Lawn, NJ) at a concentration of
0.05 g/L was released into the chamber using a glass syringe fitted with a 20-gauge stainless steel needle.
The low concentration of the dye used here is not expected to significantly alter relevant physical properties
of PCE (Taylor et al., 2001). Digital images collected after PCE redistribution in the chamber are shown
in Figure 4.2-14, as well as gas channels that formed during sparging experiments. Note that the mode of
PCE addition was different in the two experiments. In the first experiment, 8.0 ml of the dyed PCE were
added to the chamber through the top center port with the needle inserted to about 3.0 cm below the water
table level. In the second experiment, PCE lateral spreading was enhanced by releasing PCE through all
three top ports (3.0 ml through the center port and 2.5 ml through each of the other two ports). As before,
the needle was inserted to about 3.0 cm below the water table. In both experiments, the PCE was allowed
to redistribute for 72 hours and at least two LTV images were collected for each pack prior to starting the
sparging experiments.
Figure 4.2-14. Digital photographs of the 2-D flow chamber: (a) and (b) show the initial distribution of
PCE in Pack A and Pack B, respectively, and (c) and (d) show the sparging channels during
the sparging experiments for Pack A and Pack B, respectively.
-------
Two sparging events were conducted in the first experiment (Pack A) and four sparging events were
conducted in the second experiment (Pack B). In each sparging event, a certain fraction of the PCE, as
monitored by the GC/MS instrument was removed after which an aqueous flux test was conducted. The
mass discharge in the presence of reduced PCE mass was then calculated and results were compared to the
mass discharge in the presence of the initial PCE mass. It should be noted that CO2, rather than air, was
used in sparging to facilitate the aqueous re-saturation of the chamber. While sparging efficiency by CO2
could be different than that by air, the performance of sparging as a remediation technique was not a major
concern in this study but, as stated earlier, the focus was on the mass discharge emanating from the DNAPL
source as a response to DNAPL mass reduction. The PCE saturation distribution in the chamber before and
following each sparging test was measured using the modified LTV method developed by Bob et al. (2008).
The initial PCE saturation distribution in the chamber as calculated from image analyses, as well as the
PCE saturation distribution following each sparging event, is shown in Figure 4.2-15 for both experiments.
In this study, changes in mass flux were assessed with respect to two features of DNAPL architecture: the
nature of the local scale PCE saturation and the PCE spatial distribution. With respect to the nature of the
PCE saturation, an assessment was made based on whether the PCE was at or below residual saturation, in
which case the PCE was considered to exist as ganglia in the pore space. Otherwise, if the PCE saturation
was above the residual saturation, the PCE was considered to exist in the pore space as pools. The 12%
and 15% saturation values that defined the residual saturation for pack A and B, respectively, were obtained
gravimetrically from PCE residual saturation measurements conducted in a small 2-D flow chamber. The
initial PCE saturation distributions indicated that PCE was present in the chamber as ganglia at or below
residual saturation (<12% for Pack A and < 15% for Pack B) as well as pools of high saturation (>12% or
15%).
The PCE saturation distribution following the sparging events in these experiments showed that sparging
can remove PCE from pooled areas as well as from areas of residual PCE saturation (see Figure 4.2-15).
A summary of experimental parameters for Pack A and Pack B is given in Table 4.2-8 and experimental
results are summarized in Table 4.2-9. The PCE mass removed was obtained by averaging the results of the
granular activated carbon (GAC) analysis and the GC/MS. PCE mass calculations obtained from saturation
values measured by the LTV method agreed quite well with GAC and GC/MS results. A detailed discussion
of the LTV results and the errors associated with LTV analyses for these experiments are presented
elsewhere (Bob et al., 2008). The on-line GC/MS instrument provided the advantage of monitoring how
much PCE mass was removed during a given sparging event (Fine et al., 2008). Consequently, the sparging
was ceased when a specified fraction of the PCE mass was removed. Total PCE mass fraction removal
from Pack A and Pack B was 55% and 78%, respectively (Table 4.2-9). While these results, as well as
others (Bass et al., 2000, Waduge et al., 2004) show that sparging can be an effective NAPL remediation
technology, of particular interest in this study was the aqueous effluent contaminant mass discharge
resulting from partial removal of the DNAPL. To assess this, the PCE mass fraction removed following
each sparging event in both experiments was plotted as a function of relative mass discharge, and these data
are shown in Figure 4.2-16. The steady state flux-averaged effluent PCE concentration was used to calculate
the mass discharge following each sparging test, and this mass discharge was normalized to the initial mass
discharge (i.e. prior to conducting the first sparging test in each experiment) to obtain the relative mass
discharge.
Figure 4.2-16 shows that the removal of 55% of PCE mass in Pack A experiment resulted in 55% reduction
in mass discharge but a similar mass removal in Pack B experiment resulted in very small reduction in
mass discharge. Both packs were characterized by a high initial ganglia percentage of PCE distribution, but
while the ganglia fraction of Pack A essentially remained the same as PCE mass was removed, the ganglia
percentage in Pack B increased as PCE was removed (Table 4-2-9). This difference in mass discharge
response between these two experiments highlights the importance of DNAPL architecture in controlling
the down gradient mass discharge.
-------
450
360
2&0
1
ISO
so
A
Pack A
A
I.
>
• ••
V
4SO C50 ' ISO 250 '350 '45G"
' .1 . >if im
450
- 350
250
I .»
90
Q
V
•
Pack B
&
***
fr
060 160250350460050150250
•450
380
I
•3 2SO
ISO
0
oua
•t\ -
'
01
350
350
250
i:*
BO
0
•*
L
0,3
0?
01
450 050150250350450
• * *
0,C4
OD?
.
0 50 150 250 350 450
050 150 250 3SQ 450
\-'i
Figure 4.2-15. The initial PCE saturation distribution in the flow chamber and the saturation distribution
following each sparging event. Two sparging tests were conducted in Pack A and four
sparging tests were conducted in Pack B.
Table 4.2-8. Summary of PCE sparging test parameters.
Parameter
Porous media
Porosity, pore volume (ml)
Volume of PCE injected (ml)
% Ganglia
Overall PCE saturation (%)
Initial flux-averaged steady-state
concentration (mg/L)
Flux test aqueous flow rate (ml/min)
Initial mass discharge (mg/min)
Pack A
20/30 Accusand
0.311 (1026)
8.0
85
0.78
61.3 ±7.0
3.01 ±0.07
0.189 ±0.023
PackB
Mixture of 20/30 and 40/60
in a 3 : 1 ratio
0.296(931)
8.0
74
0.86
55.4 ±16.2
3. 13 ±0.05
0.171 ±0.050
-------
Table 4.2-9. Summary of PCE sparging test results.
Parameter
Sparging rate (LPM) and
duration (hours)
Total PCE removed (ml, %)
PCE present as ganglia (%)
Effluent flux-averaged
concentration (mg/L)
Relative mass discharge
WK,0)
First
Sparging Event
Pack A
0.200 (5)
3.0(37.5)
91
41.7 ±
11.9
0.66 ±
0.03
PackB
0.200 (5)
1.9(23.8)
94
66.2 ±
14.8
1.21 ±
0.45
Second
Sparging Event
Pack A
0.200 (4)
4.4(55.0)
88
25. 8 ±5.7
0.40 ±
0.10
PackB
0.200 (3)
3.7(46.3)
93
51.8±8.0
0.95 ±
0.31
Third
Sparging
Event
PackB
0.200 (9)
5.1(63.8)
96
48.3 ±9.1
0.89 ±
0.31
Fourth
Sparging
Event
PackB
0.200 (9)
6.2 (77.5)
100
2.3 ±0.3
0.04 ±
0.01
As more PCE mass is removed in Pack B, the mass discharge decreased significantly to 4% of the initial
value indicating a 96% reduction (Figure 4.2-16). This mass reduction/mass discharge relationship where
mass discharge reduction is higher than mass reduction is quite favorable. It is worth noting that the 96%
reduction in mass discharge achieved in this experiment was associated with a PCE flux-averaged effluent
concentration of 2.3 mg/L (see Table 4.2-9), which is significantly higher than the MCL for PCE. However,
the mass discharge at this point (0.007 mg/min) was significantly smaller than the initial mass discharge.
Consequently, beneficial changes in the associated plume at a real site are expected, such as a reduction in
the spatial extent of the plume or the total contaminant mass in the plume. The extent to which the plume
would change depends on both the magnitude of flux reduction and the natural attenuation capacity of the
down gradient aquifer formation.
1.8
1.6
1.4
1.2
0.8
0.6
S> 0.4
CO
OL
0.2
0.0
Pack A
Pack B
0
80
100
20 40 60
Mass Removed (%)
Figure 4.2-16. Relative effluent mass discharge as a function of PCE mass removed for the two
experiments.
-------
4.3 Theoretical and Modeling Research
There are three broad groups of models that have been used for the description of contaminant transport
at DNAPL sites: (1) numerical models, (2) Lagrangian, or streamtube based models, and (3) analytical
models. While the numerical models are based on the most detailed description of process behavior at the
pore scale, they are computationally intensive and require information on spatially variable parameters like
local permeability and DNAPL saturations that are difficult to characterize with reasonable accuracy using
the available technologies. Though these models are generally not parameterizable at the field scale, they
help in improving our understanding of the behavior of DNAPL sources and plumes.
At the other end of the spectrum are the empirical or semi-empirical analytical models that assume
homogeneity of parameters and are based on several simplifying assumptions. It has been suggested that
even though the "local" scale processes behind DNAPL dissolution are complex, the temporal evolution of
contaminant concentrations at the source control plane can be reasonably described by these simple models.
Though site-specific parameterization of these models are not yet that well established, the models, due to
their computational simplicity and fast run times, serve as good tools to (1) run optimization routines for
partial source and plume treatment, and (2) perform uncertainty assessments.
The Lagrangian models are the only "practical" techniques currently available for predicting site specific
mass removal mass flux relationships. These models are based on estimating integrated source zone
parameters using tracer tests and then using the parameters for prediction of source flux and down-gradient
impacts. The ability to characterize source zones in terms of only a handful of trajectory-integrated
parameters, in contrast to the potentially very large number of spatially distributed parameters required
for traditional Eulerian approaches, has increased the popularity of these models over the last few years.
However, tracer tests are very expensive and not yet affordable at most field sites.
The benefits of partial source zone remediation were evaluated from fundamental physical principles using
each of these modeling approaches. The results of these analyses are described in detail in the following
peer-reviewed journal articles and reports:
Lagrangian Models
Jawitz, J.W., A.D. Pure, G.G. Demmy, S. Berglund, and P.S.C. Rao (2005). Groundwater contaminant
flux reduction resulting from nonaqueous phase liquid mass reduction. Water Resources Research
41(10):W10408.1-15, doi:10.1029/2004WR003825.
Wood, A.L., C.G. Enfield, F. Espinoza, M. Annable, M.C. Brooks, P.S.C. Rao, D. Sabatini and R. Knox
(2005). Design of aquifer remediation systems: (2) Estimating site-specific performance and benefits
of partial source removal. Journal of Contaminant Hydrology 81(1-4): 148-166.
Coupled Source Zone and Plume Models
Falta, R.W., P.S.C. Rao, and N. Basu (2005a). Assessing the impacts of partial mass depletion in DNAPL
source zones: I. Analytical modeling of source strength functions and plume response. Journal of
Contaminant Hydrology 78(4):259-280.
Falta, R.W., N. Basu, and P.S.C. Rao (2005b). Assessing the impacts of partial mass depletion in DNAPL
source zones: II. Coupling source strength functions to plume evolution. Journal of Contaminant
Hydrology 79(1-2): 45-66.
Falta, R.W. (2008). Methodology for comparing source and plume remediation alternatives. Ground Water
46(2):272-285.
Falta, R.W., M.B. Stacy, A.N.M Ahsanuzzaman, M. Wang, and R. Earle (2007). REMChlor Remediation
Evaluation Model for Chlorinated Solvents User's Manual Version 1.0. http://www.epa.gov/ada/
csmos/models/remchlor.html
Wagner, D.E., J. Huang, J.L. Heiderscheidt, and M.N. Goltz. (2009). Modeling study to quantify the
benefits of groundwater contaminant source remediation. (In preparation).
-------
Model Assessments and Applications
Basu, N.B., A.D. Pure, and J.W. Jawitz (2008b). Simplified contaminant source depletion models as
analogs of multiphase simulators. Journal of Contaminant Hydrology 97(3-4): 87-99, doi: 10.1016/j.
jconhyd.2008.01.001.
Basu, N.B., A.D. Pure, and J.W. Jawitz (2008c). Predicting DNAPL dissolution using a simplified source
depletion model parameterized with partitioning tracers. Water Resources Research 44(7):W07414.1-
13, doi:10.1029/2007WR006008.
Basu, N.B., P.S.C. Rao, R.W. Falta, M.D. Annable, J.W. Jawitz, and K. Hatfield (2008a). Temporal
evolution of DNAPL source and contaminant flux distribution: impacts of source mass depletion.
Journal of Contaminant Hydrology 95 (3-4): 93-109.
Falta, R.W. (2003). Modeling sub-grid block scale DNAPL pool dissolution using a dual domain approach.
Water Resources Research 39(12):SBH 18.1-8.
Falta, R.W. (2005). Dissolved Chemical Discharge from Fractured Clay Aquitards Contaminated by
DNAPLs, In: Dynamics of Fluids in Fractured Rock, B. Faybishenko, P.A. Witherspoon, and J. Gale,
eds., Geophysical Monograph 162, American Geophysical. Union.
Falta, R.W. (2009). Monte Carlo simulations of source behavior with remediation. (In preparation).
Liang, H., and R.W. Falta (2007). Modeling field-scale cosolvent flooding for DNAPL source zone
remediation. Journal of Contaminant Hydrology 96(1-4): 1-16.
Van Antwerp, D., R.W. Falta, and J.S. Gierke (2008). Numerical simulation of field-scale contaminant
mass transfer during air sparging. Vadose Zone Journal 7:294-304.
4.3.1 Lagrangian Models
Lagrangian models (Jawitz et al., 2005; Enfield et al., 2005; Wood et al., 2005; Basu et al., 2008b)
describe the source zone as a collection of non-interacting streamtubes with hydrodynamic and DNAPL
heterogeneity represented by the variations in the travel time t and trajectory-integrated DNAPL saturations
within the streamtubes (Figure 4.3-1). These models are computationally less intensive and require fewer
parameters than three-dimensional numerical models, but describe the spatial variability of source zones
unlike one-dimensional flow and transport codes that assume homogeneity. The Lagrangian models assume
that flow is stable and steady, advection is the dominant transport process, NAPL is immobile, and NAPL
saturations are low such that dissolution does not significantly alter the properties of the flow field. These
assumptions do not hold under conditions of displacement instabilities induced by buoyant or viscous forces
(Jawitz et al., 1998), transient flow due to time-varying boundary conditions such as seasonal shifts in
hydraulic gradients (Rivett and Feenstra, 2005), alteration of flow field due to relative permeability changes
during dissolution (Geller and Hunt, 1993), and nonadvective transport of solutes with different effective
diffusion coefficients (Cirpka and Kitanidis, 2000). The Lagrangian models described here were developed
for dissolution-based remediation activities. There are two steps in the prediction of source depletion using
Lagrangian techniques: (1) Estimation of the statistical properties of the hydrodynamic and the DNAPL
distribution, and (2) Prediction of DNAPL dissolution using the estimated properties.
-------
Mass
flux
Figure 4.3-1. Lagrangian conceptualization of DNAPL source zones.
Estimation of the statistical properties of the hydrodynamic and DNAPL distribution
In the Lagarangian model, the properties of the hydrodynamic field and the DNAPL saturation distribution
are described by the mean and standard deviation of the non-reactive travel time, t, and the trajectory
integrated DNAPL content, S respectively.
It is important to distinguish the Lagrangian parameter S that represents an integral measure between
the injection and monitoring planes from the local DNAPL content S. The NAPL content is defined as
S =SN I (1 - SN), where SN is NAPL saturation. The mean and standard deviation of the t and S are estimated
by performing non-partitioning and partitioning tracer tests under the same initial and boundary conditions
as the dissolution experiment. The travel time and the NAPL contents are assumed to be lognormally
distributed, and correlated (yT § = correlation coefficient) random variables. Moreover, the flow field can be
composed of multiple hydrologic units and can thus, be described by multiple lognormal distributions.
Two techniques were investigated for estimating the characteristics of the travel time and the DNAPL
saturation distribution: (1) the method of moments (Jawitz et al., 2005; Basu et al., 2008b) where measured
moments from tracer breakthrough curves are equated with derived moment equations to calculate the
required parameters, and (2) the inverse modeling approach (Enfield et al., 2005; Wood et al., 2005) where
tracer breakthrough curves are fitted to a transport model and the parameters estimated by optimization.
Method of moments: The moments of the t and S (m\, m\, mf, m%) distributions are estimated from the
moments of the non-partitioning and partitioning tracer breakthrough curves (m"p, m^p, mp, mp) using the
following equations (Jawitz et al., 2003)
m«P=m
-------
t
(4.3-ld)
.
~ /
—
~
3 A
(4.3-le)
where, O^is the NAPL-water partitioning coefficient, t0 is the solute pulse injection duration,
(j) = (/ - Sn J / (l - Sn J, f is the fraction of the domain that is contaminated, Sn is the average NAPL
saturation for the entire spatial domain, and the correlation term y = exp (p -
-------
where Mn[M] is the NAPL mass in the individual streamtube, Q is the flow rate [L3/T], p^ is the NAPL
density, v is the velocity in the streamtube [L/T], L is the length of the streamtube, A is the cross sectional
area of the streamtube [L2], t is the residence time or the travel time and N is the ratio of the NAPL density
to the solubility ( p^ ICS).
For an arbitrary flushing duration L, all streamtubes with reactive travel times less than L, are clean,
and contribute zero flux, while all streamtubes with T > Lare still contaminated. Under the equilibrium
assumption, all streamtubes with T > Lore at solubility, and the NAPL concentration for a given streamtube
/' can be expressed as a function of the flushing duration T as:
C,(tf) =
CwforQ
-------
EW-5
Concentration
ProjConc
o Accumulative mass
(davs)
Figure 4.3-3. Measured and predicted remedial performance in a vertical circulation implementation of
surfactant extraction (Wood et al., 2005).
0 0.2 0.4 0.6 0.8 1
fractional mass reduction, Rm
Figure 4.3-4. Solid lines are reduction in contaminant flux as a function of reduction in source zone mass
for a In i ={0.05, 0.2, 0.6, 1.0, 1.5, 2, 3}. Dashed lines are tie lines for flushing durations
tf= {0.8, 2,4,12} pore volumes (PV). Note thatN = 58 and ^lnS= 0.05 (Jawitz et al., 2005).
4.3.2 Coupled Source Zone and Plume Models
Despite the sophistication of the three-dimensional and Lagrangian models, they are difficult to use at most
field sites owing to financial constraints which limit the ability to obtain the necessary input data. Thus,
analytical models of source behavior (Zhu and Sykes, 2004; Parker and Park, 2004; Falta et al., 2005a;
Christ et al., 2006; Basu et al., 2008b) can be important decision making tools for risk assessment and
remedial design. In these models, the source zone is represented by an effective homogeneous domain with
the flux-averaged concentration at the source control plane (Cj(t)) described as a function of the mean
-------
advective flow (<7 , LT1) and the mean DNAPL mass (Ms (t)) in the domain. The source mass/source
concentration relationships are coupled with simple mass balance equations to describe the time varying
Cj(t) as a function of the initial mass (Ms 0) and the initial flux averaged concentration ( C,- 0).
The power function model (Zhu and Sykes, 2004; Falta et al., 2005a) is based on the assumption that the
source mass/source discharge relationships shown in Figure 4.3-5 can be approximated by a simple power
function relationship (Rao et al., 2001; Rao and Jawitz, 2003; Parker and Park, 2004; Zhu and Sykes, 2004;
Falta et al., 2005a):
c,
Ms(t)
-ir
(4.3-6)
where, the exponent, F, determines the source discharge response to changing source mass, and generally
represents site specific conditions, such as flow field heterogeneity, DNAPL distribution, and the correlation
between the two. If F = 1, there is a 1:1 relationship (Figure 4.3-6). Values of F less than one produce
C vs M curves that fall in the upper half of the graph (above the 1:1 line), while values of F greater than
one produce C vs M curves that fall in the lower half of the graph. Field and laboratory data suggest that
a F value of one is reasonable in some cases, but theoretical analyses indicate that a range of F values are
possible, depending mainly on the correlation of the contamination distribution to the permeability field.
DiFilipo and Brusseau (2008) analyzed data from a number of field studies and found that the majority of
source remediations could be characterized using values of F between 0.5 and 2.0, though several studies
had higher and lower values.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
M/Mo
A Dover AFBPCE
release (U. of Florida)
O Dover AFBPCE
release (Clemson U.)
-3-D simulation,
negative correlation
wlthk
-3-D simulation,
positive correlation
wlthk
-Transient simulation
of DNAPL pool
scenario
- Equation (2) with an
exponent of 0.5
-K- Equation (2) with an
exponent of 2.0
Figure 4.3-5. Source zone dissolved concentrations as a function of source zone DNAPL mass (Falta et
al., 2005a).
-------
r=o
M/M,
Figure 4.3-6. Power function representation of source mass/source discharge relationship
(Equation 4.3-6).
4.3.2.1 Analytical expressions of source flux and mass
The power function relationship (Equation 4.3-6) can be used to describe the flux-averaged concentration at
the source control plane as a function of the initial conditions using a simple mass balance technique. The
contaminant discharge from a source zone is the product of the flowrate of water passing through the source
zone, and the average concentration of contaminant in that water (Figure 4.3-7). Source discharge thus has
units of mass per time, and is not to be confused with mass flux, which is a discharge divided by an area. If
water flows through the source at a rate of Q (t), and if the mass in the source zone is also subject to some
form of chemical or biological first order decay, then a mass balance on the source gives:
dt
•=e(oc,(o-v^
(4.3-7)
where Ms is the mass remaining in the source zone with time, Cj(t) is the time-dependent source dissolved
concentration (flow averaged), and hs is the source decay rate by processes other than dissolution. Water
flow through the source may be due to infiltration (above the water table) or groundwater flow (below the
water table).
DNAPL source
Groundwater flow, Vd zone /
Dissolved plume
cin=o
^^^ ^
Source
MASS, M(t)
w
Cout= Cs(t)
Figure 4.3-7. Conceptual model of source zone with time-dependent contaminant mass and discharge.
-------
If the water flow rate through the source zone in Equation (4.3-7) is assumed to be constant, the power
function (Equation 4.3-6) can be substituted to get:
dt
(4.3-8)
This equation is nonlinear for F values other than zero and one, but it can be linearized using Bernoulli's
transformation, and solved to get (Falta et al., 2005a):
Ms(t} = \-f^L + \Ml-T0 +f^f pr~1)V <4-3-9)
L ' V ' / J
Using Equation (4.3-6), this leads to the time-dependent source discharge function:
(4.3-10)
Similar expressions can be derived for the case ofks= 0 (Parker and Park, 2004; Zhu and Sykes, 2004).
A very important special case of Equation (4.3-8) occurs when F = 1 and X5= 0. In that case, the
differential equation is linear and may be integrated to get a simple exponential decay solution (Newell
et al., 1996; Parker and Park, 2004; Zhu and Sykes, 2004):
Ms(t)=Ms_0e s'° (4.3-11)
and QCj^
Cj,s(t) = Cj,oe S'° (4.3-12)
Therefore, when F=l, both the source mass and the source discharge will decline exponentially with time.
If A,s=0, then the apparent source decay rate due to dissolution is QCJO /Ms^0, giving a source half-life
of .693Mol(QCo) (Newell andAdamson, 2005). This type of source behavior has been observed in the
field at many chlorinated solvent sites (Newell andAdamson, 2005; McGuire et al., 2006; Newell et al.,
2006), as well as at sites contaminated by petroleum hydrocarbons (Newell et al., 2002). The widely used
EPA BIOCHLOR (Aziz et al., 2002) and BIOSCREEN (Newell et al., 1996) analytic models for natural
attenuation include exponentially decaying source terms.
An important characteristic of source zones with F greater than or equal to one, is that theoretically the
source is never completely depleted, and the source discharge is always greater than zero, even at large
times. In simple terms, this happens because the rate of discharge from the source drops as fast or faster
than the rate of mass depletion of the source. This type of response might be expected from a site where the
DNAPL distribution was correlated with low permeability. In this case, most of the contaminant mass is
trapped in low permeability zones and removing the small amount in high permeability zones will have a
large effect on contaminant discharge. However, this type of source behavior also tends to lead to extensive
tailing with time because the source is never completely depleted by dissolution.
When F<1, the source has a finite life, and the source discharge eventually is equal to zero. This type of
behavior could occur when DNAPL is preferentially located in high permeability zones. In this case a
given fractional reduction in source mass would produce a smaller fractional reduction in discharge. This
type of source behavior leads to relatively rapid removal of the source by dissolution with little tailing.
-------
Another useful special case occurs when F=0.5. This leads to a source concentration that declines as a
linear function of time (Falta et al., 2005a; Newell and Adamson, 2005):
OC2
~ U "-t (4.3-13)
and the source completely disappears at a time of
2M
s-°
QCj.0
(4.3-14)
The simplest model of source behavior is one in which F =0, which leads to a constant source discharge
(concentration) until the source is fully depleted. This is also known as a "step function" model, and the
source mass declines at a constant rate with respect to time.
The source model (Equations 4.3-9 and 4.3-10) represents source depletion by the natural process of
dissolution and perhaps some other form of chemical or biological decay. This model can easily be
modified to account for aggressive source remediation activities that remove a substantial fraction of the
source mass over a short period of time (Falta et al., 2005a). If a source remediation effort (such as alcohol
or surfactant flooding, chemical oxidation, thermal treatment, air sparging) begins at a time of tj, and ends at
a time of t2, during which a fraction, X of the source mass is removed, the functions can be simply rescaled.
Then the source mass and concentration following remediation are given by:
\-oc ( oc ^
MM) = \ J'2+\ M'"r+ ^ J'2 leF-W-v }• (4.3-15)
•S v / i n y-r 0,2 i n y-r
(4.3-17)
V / "J,l I / A <•> -I O\
where MS1 is the source mass at tf, andMS2 is the source mass at t2 . The change in source discharge
following remediation varies as the fraction of mass remaining (l-X) raised to the power F . Therefore
if F =1, a linear reduction of source discharge is expected; iff =2, the discharge will drop as the square
of the mass fraction remaining, while iff =0.5, the discharge will drop as the square root of the mass
fraction remaining. Examples of this type of source behavior with and without remediation are shown
in Figures 4.3-8 and 4.3-9, for a case where the initial source mass is 1620 kg, with an initial source
concentration of 100 mg/1, and a water flow rate of 600 m3/yr.
-------
100000
g 10000
1000
100
-•-no remediation,
gamma = 0.5
-•-remove 90%
after 20 years,
gamma = 0.5
-O remove 90% at
time zero,
gamma = 0.5
20 40 60 80
Time since DNAPL release, years
100
Figure 4.3-8. Source zone dissolved concentration with and without source remediation for F= 0.5 (Falta
etal.,2005a).
100000
10000
1000
100
-no remediation,
gamma = 2.0
-remove 90%
after 20 years,
gamma = 2.0
-remove 90% at
time zero,
gamma = 2.0
20 40 60 80
Time since DNAPL release, years
100
Figure 4.3-9. Source zone dissolved concentrations with and without source remediation for F = 2.0
(Falta et al, 2005a).
4.3.2.2 Analytical approaches including plume remediation and natural attenuation
Experience with natural attenuation as a remedy for plume management has shown that mathematical
models can play an important role in the selection of remedial alternatives (Wiedemeier et al., 1999; NRC,
2000; Alvarez and Illman, 2006). In many cases, screening level simulations performed with analytical
models such as BIOCHLOR (Aziz et al., 2000) or BIOSCREEN (Newell et al., 1996) are effective
for demonstrating the applicability of natural attenuation. A recent study of 45 CVOC sites found that
mathematical models were used at 60 percent of the sites, and that BIOCHLOR was the most frequently
used model (McGuire et al., 2004). BIOCHLOR is based on a modified version of the Domenico (1987)
analytical approximation for advective transport with three-dimensional dispersion. BIOCHLOR uses an
-------
analytical transformation developed by Sun et al (1999a, b) to model the chain decay of chlorinated solvents
such as PCE (PCE-»TCE-»DCE-»VC-»ethene) or TCA (TCA-»DCA-»CA-»ethane) assuming first
order reactions. The source term can be a constant concentration, or it can decay exponentially with time.
BIOCHLOR, however, does not simulate source or plume remediation. These types of analytical models are
also commonly used to estimate exposures for risk assessment.
An analytical approach that accounts for source remediation using the simplified source functions described
above was described in Falta et al. (2005a). This model simulates simple one-dimensional advective
transport of a single species with first order decay. The model was used to make several generalizations
about the relationship between source remediation and plume response. One of the most significant
observations was that the maximum extent of a plume is a logarithmic function of the amount of source
remediation. Considering the special case where F is equal to one, and assuming that source remediation
takes place immediately following a DNAPL release, the maximum length of a plume was shown to be
(4.3-19)
In
where XR is the maximum plume length with source remediation, xm is the maximum plume length without
source remediation, X is the fraction of DNAPL source removed, C0 is the initial source concentration, and
Cm is the regulatory limit for the chemical (the concentration that defines the edge of the plume). As shown
in Table 4.3-1, when the plume is defined by a very low concentration relative to the source concentration, it
becomes very difficult to "shrink" a plume by source remediation alone. On the other hand, for F =1, source
remediation efforts will reduce the future mass, concentrations, and cancer risk in a plume by an amount
equal to the fraction of mass removed.
Table 4.3-1. Fraction of DNAPL source removal (X) required to reduce maximum plume length by a
specified amount using Equation (4.3-19) (from Falta et al., 2005a).
Relative maximum plume length,
x
NR
cm
C0
w-1
io-2
w-3
io-4
w-5
0.8
.37
.60
.75
.84
.900
0.5
.68
.900
.968
.990
.997
0.3
.80
.960
.992
.998
.9997
0.2
.84
.975
.996
.9994
.9999
0.1
.87
.984
.998
.9997
.99997
A more realistic plume model coupled to source remediation was developed in Falta et al. (2005b). This
semi-analytical model uses the same source function, but the plume model simulates natural attenuation
by advection with three-dimensional dispersion, and it includes first-order parent-daughter chain decay
using the method of Sun et al. (1999a, b). This model was used to study the effects of immediate and
delayed source remediation on the plume response. One major conclusion from this study was that source
remediation was much more effective if it was conducted soon after a DNAPL release. If the source
remediation is delayed, much of the contaminant mass is already in the plume, and source remediation will
have little or no effect on the leading edge of the plume, and some type of plume remediation is probably
required.
A major limitation of these models (Biochlor, Falta et al. (2005b)) is that they cannot simulate enhanced
plume remediation, where the biogeochemical conditions are manipulated in order to increase rates of
-------
contaminant destruction. Destruction of CVOCs in dissolved plumes can occur under natural conditions by
biodegradation processes including reductive dechlorination, aerobic oxidation, anaerobic oxidation, and
aerobic co-metabolism (Weidemeier et al, 1999; NRC 2000; Alvarez and Illman, 2006). It is now fairly
common to engineer in-situ biodegradation systems for enhancing one or more of these processes in order
to allow the plume to attenuate in a shorter distance, or to reduce plume concentrations in locations that are
detached from the source. These enhancements usually involve addition of an electron donor (hydrogen,
lactate, molasses or a hydrogen releasing compound) for enhancing anaerobic processes, or an electron
acceptor (oxygen, air, H2O2 , or an oxygen releasing compound) for enhancing aerobic processes (Chapelle
et al., 2003; Alvarez and Illman, 2006). In other cases, reactive barriers or basic hydraulic control through
pump-and-treat are used to manage the dissolved plume. In order to simulate these processes, models
need to be able to describe spatially and temporally varying reaction rates. Of the models described so far,
only the BIOCHLOR model allows for two spatial zones to be denned in which the solute decay rates are
different, but this is only valid if the solute concentrations in the upstream zone are at steady-state, which
implies a constant source concentration in time.
Because of the above limitation, a new analytical plume model that can simulate these types of variable
plume reactions, along with source remediation was developed. This model is called REMChlor for
Remediation Evaluation Model for Chlorinated solvents (Falta, 2008). The key difference in REMChlor
and other plume models is that the chemical reaction parameters (rates, yield coefficients) can now be
arbitrary functions of both time and distance from the source. A graphical user's interface for the model was
developed by the Center for Subsurface Modeling Support (CSMoS). The REMChlor model is available
at http://www.epa.gov/ada/csmos/models/remchlor.html. The REMChlor analytical approach assumes
a constant groundwater pore velocity of v in the x-direction, with longitudinal, transverse, and vertical
dispersion. The solute can be retarded by adsorption, but the different solutes involved in coupled reactions
must have the same retardation factor. These assumptions are similar to those used in previous natural
attenuation plume models such as BIOCHLOR (Aziz et al., 2000; 2002), BIOSCREEN (Newell et al.,
1996), LNAST (Huntley and Beckett, 2002), and the model by Falta et al. (2005b).
4.3.2.3 REMChlor model
The governing equation for the dissolved concentration of each contaminant species in the plume, C, is:
nac ac a2c a2c a2c , , ,^^
R — = -v -- \-a v — -+a v — -+a v — - + rxn(x,t) (4.3-20)
dt a* a*2 y dy2 dz2 ^ '
Where v is pore velocity; ot^. a and otz are the longitudinal, transverse, and vertical dispersivities,
respectively; R is the retardation coefficient, and rxn(x,t) represents the rate of generation (+) or destruction
(-) of the species due to chemical or biological reactions that are spatially and temporally variable. This
plume model is coupled with the source zone mass balance given by Equation (4.3-7), using the power
function relationship for the flux-averaged concentration and mass remaining in the source (Q vs Ms)
relationship [Equation (4.3-6)]. A specified flux boundary condition at x=Q ensures that the rate of
discharge from the source zone is exactly equal to the rate at which contaminants enter the plume (see van
Genuchten and Alves (1982)). The mass flux entering the plume is specified as:
A
(43.2I)
where A is the area over which the contaminant flux enters the groundwater flow system, Q(t ) is rate of
groundwater flow, and r\ is the porosity. Outside of this area, the mass flux is zero. For sources that are
located below the water table, A would be the cross-sectional area of the source zone perpendicular to the
groundwater flow. For sources located above the water table, A would be the cross-sectional area at the
top of the water table perpendicular to flow that was required to accommodate the infiltration rate from
the source. Falta et al. (2005b) solved Equations (4.3-20) and (4.3-21) analytically for the case of first
order decay reactions with constant and uniform decay rates, using a Laplace transform method, combined
-------
with Domenico's (1987) approximation for transverse and vertical dispersion. Analytical solution of
Equation (4.3-20) with variable plume reaction rates by this method would be much more difficult. Instead,
a different approach is taken where the solute advection and reactions are decoupled from the longitudinal
dispersion using a simple streamtube technique. Scale-dependent longitudinal dispersion is accounted
for by considering a collection of streamtubes with a normally distributed pore velocity. Transverse and
vertical dispersion are then simulated using Domenico's (1987) approximation.
The reactive plume model is based on a simple one-dimensional streamtube that is characterized by a
constant pore velocity and solute retardation factor. Since there is only advection taking place in the
streamtube, the flux boundary condition at the edge of the source zone simplifies to:
- (4.3-22)
If the source is located below the water table, and Q=fyvA, then the flux boundary condition is just the time-
dependent source concentration,
= C,(0 (4.3-23)
where Cj (t) could be calculated, for example, by Equations (4.3-10) and (4.3-16).
One-dimensional advective transport of a solute can be represented graphically on a distance-time plot
(Figure 4.3-10). Here, the time axis corresponds to the time since the contaminant was first released to the
groundwater system, while the distance axis is the distance downstream from the source.
The advective front moves at a constant velocity of v IR , so that at any location, x, the front passes by at a
time of t =Rx Iv. At any time, the front is located at x=vt IR, and the solute concentration is always zero
below this line (ahead of the front). In the absence of any plume degradation process, the concentration at
any location behind the advective front can be determined from the time of solute release from the source,
* release • ^or a distance from the source, x, the travel time is ttrmel=Rx Iv . Therefore, if the total time is t, the
parcel of water found at that location (x,t) was released from the source at a time of:
(4.3-24)
and the concentration at that (x,t) point would be:
(4.3-25)
Time
Distance from source
Figure 4.3-10. Distance-time plot for advective transport with a single set of reaction rates (Falta, 2008).
-------
Plume reactions can easily be included in this advective streamtube model. As a parcel of solute is
translated downstream, it is not subject to any mixing processes, so it is conceptually equivalent to a batch
reaction that starts at time t'=Q with an intitial condition of C (tmlease) x=0 and reacts for a period of time
equal to the travel time to position x, t '=Rx Iv . As an example, if the solute reaction was first order decay
in the aqueous phase with a decay rate coefficient of k, then the equivalent batch reaction is:
=~kc Wlth
(4.3-26)
Then at location (x,t) behind the front, the solute concentration would be:
(4.3-27)
This result is exactly the same as the Laplace transform solution to Equations (4.3-20) and (4.3-21) with
zero dispersion (Falta et al., 2005a). More complicated coupled reactions can be considered using this same
method, but a fundamental limitation is that all of the reacting solutes must move at the same velocity in the
groundwater, so they must be assumed to have a single retardation factor, R.
The analysis can be extended to the case of time and distance dependent reaction rates by dividing the time-
distance domain into distinct zones (Figure 4.3-11). Here, nine zones have been chosen to approximately
represent conditions downgradient from a contaminant source over the life of a plume. The first time zone
after the spill, t < tl, could represent a period following the contaminant release where no manipulation
of the plume has yet been attempted, a period of natural attenuation. The second time zone after the spill,
t\< t<(2 could represent a temporary period of active plume remediation (enhanced attenuation). The final
time zone, t > t2, could be used to represent long term conditions in the plume after manipulation of the
plume ended (another period of natural attenuation).
time
Distance from source
Figure 4.3-11. Distance-time plot for advective transport with multiple sets of reaction rates (Falta, 2008).
Distance from the source is similarly divided into zones so that near the source, for x
-------
remediation scheme could consist of the addition of electron donor in the first 400 m of the plume to
increase the rate of reductive dechlorination of PCE and its daughter product, trichloroethylene (TCE).
Because the daughter products cis-l,2-dicloroethylene (DCE) and vinyl chloride (VC) do not degrade as
readily by reductive dechlorination, but they can degrade aerobically, a different reaction zone could be
created from 400 to 700 m, where aerobic degradation was stimulated (Chapelle et al., 2003; Alvarez and
Illman, 2006). Downgradient of this zone, conditions might revert back to their natural state.
Each of these space-
time zones can have
a different decay rate
for each chemical
species
time 2025
2005
1975
Natural
attenuation
Reductive
dechlorination
Natural
attenuation
Natural
attenuation
I Aerobic
I degradation
Natural
attenuation
Natural
attenuation
Natural
attenuation
Natural
attenuation
400 700
Distance from source, m
Figure 4.3-12. Hypothetical design of an enhanced plume remediation scheme with an enhanced reductive
dechlorination zone for destruction of PCE and TCE and an enhanced aerobic degradation
zone for destruction of DCE and VC. All other zones revert to natural background conditions
(Falta, 2008).
The analytical solution for multiple reaction zones is developed using the residence time in each zone to
develop the batch reaction solution for that zone. The initial conditions for the batch reaction are the final
conditions from the previously encountered reaction zone. The residence times in each reaction zone are
calculated using straightforward logic. For the example shown in Figure 4.3-11, the solutes that are present
at location (x ,t) left the source at a time, tmlease that was before tl, so they initially encounter reaction
zone (I). The residence time in zone (I) is then t(r= tr trelease . The solutes next enter zone (II), where
they remain until they cross xl , at a time of trelease + Rx^ I v . Therefore, the residence time in zone (II)
is t(II)= tmlease + Rx^ I v-tl . The solutes next enter zone (V), where they remain until t2, so the residence
time in zone (V) is
tabulated.
'= '-
lease
/ v . In this way, the residence times for each reaction zone are
In general, solutes can pass through any of the nine reaction zones, so a total of nine reaction zone residence
times are computed. For any given value of (x ,t), the advective path leading to that location will cross at
most five zones, so several of the zone residence times are zero. The analytical solution is constructed by
sequentially performing the batch reactions in each zone that is encountered, starting with a concentration of
^ ( Release )Uo • With the zone numbering scheme used in Figure 4.3-11, the numerical value of the reaction
zone always increases with increasing travel distance.
Going back to the example of a single solute undergoing first order decay in the aqueous phase, a set of nine
reaction rates are defined (£I- £IX). The solute concentration at (x, t) is then:
_o exp-
(4.3-28)
-------
A problem of significant practical interest in groundwater remediation involves simultaneous first order
parent-daughter decay/production reactions. Considering a four component system, the relevant batch
reaction equations for species A,B,C,D in zones (n) are:
^—4")___j( £ 7.C. —»C n(0) = C r (43-29)
- = yBA(n]kA(n}CA(n) -kB{n)CB(n) I-C. -> CB(n) (0) = CBM) (4.3-30)
C/f
]-= k C -k C I-C-^CD()(°) = CD( i) (4.3-32)
where yij(n) are the yield coefficients for each parent-daughter reaction. These can also depend on distance
and time if the nature of a reaction changes so that it no longer produces the same proportion of daughter
product from decay of the parent (for example during aerobic degradation of DCE, little VC is produced).
Equations (4.3-29 and 4.3-30) are written for reaction zone (n), and the reactions proceed for a period equal
to the residence time, t{n}, with initial conditions (LC.) that are the concentrations from end of the previous
reaction zone. The starting conditions for the first reaction zone are C.(0) = Cj(t - Rx I v) x=0.
Following methods used in chemical reactor design (see, for example, Chen (1983)), the coupled reaction
equations can be solved by Laplace transform methods to yield:
(4.3-33)
CB(n) = CA(n-l)f2 ^A(n) A(«) ^BA(n) >*(„) + CB(n-l)fl (^B(n)^(n)) (4.3-34)
' t(n]
(4.3-36)
where the A,.(H) =ki(,/R, and
e-« (4.3-37)
e*~\ (4.3-38)
-------
/3 (A ,^2,^3, y-il, ^32 , 0 = y^y-l
~-u
(4.3-39)
(4.3-40)
(4-4) (^-4)
These batch solutions are widely known as the Bateman equations (Bateman, 1910), originally developed
to describe the decay and production of radioactive species. Here, the equations have been modified
to account for variable daughter yield coefficients. These batch solutions are also equivalent to those
developed by Sun et al. (1999) in their analytical transformations.
The REMChlor streamtube plume model using Equations (4.3-33)-(4.3-40) is compared with the two-
zone BIOCHLOR model in Figure 4.3-13. This case assumes a constant source concentration of PCE of
1 mg/1, with a pore velocity of 100 m/yr, a retardation factor of 2, and no dispersion. Referring back to
Figure 4.3-11, this test case only uses zones (I) and (IV), with steady-state conditions in zone (I). This is
a special case for the new solution, which can handle fully transient conditions using all nine zones in space
and time.
Over the first 500 m, PCE and TCE are allowed to degrade with first order rate constants of 0.693/yr (an
aqueous half life of one year), but DCE and VC do not degrade at all in this zone. Beyond 500 m, the
DCE and VC degrade with a rate constant of 0.693/yr, but the PCE and TCE do not degrade at all. The
simulation time of 20 years is sufficiently long for steady-state conditions to be present in the first zone,
a requirement of the BIOCHLOR model. The two solutions produce an identical result. In the first zone,
PCE degrades, producing TCE, which degrades into DCE. The process stalls at DCE in this zone, and the
concentration of VC is zero. Beyond 500 m, the concentrations of PCE and TCE are constant, while the
DCE degrades to form VC, which also degrades.
500 1000
Distance from Source, m
Figure 4.3-13. Comparison of the REMChlor reactive streamtube solution with BIOCHLOR for a two-zone
case with steady state concentrations in the upstream zone (t = 20 years). The solid lines are
computed using the REMChlor solution while the symbols are calculated using BIOCHLOR
(Falta, 2008).
-------
4.3.2.4 Cancer risk assessment
Many of the regulated groundwater contaminants and their degradation products are considered to be
known or probable carcinogens by the US EPA. Cancer risk from exposure to carcinogens is quantified
using the chronic daily intake (CDI) of the carcinogen (mg/kg-day) and a cancer risk slope factor (SIF)
that has units of risk per mg/kg-day. The CDI is a dose rate averaged over a human lifespan, regardless of
the exposure period (US EPA, 1989). The maximum exposure period, tex is usually limited to 30 years. The
risk is calculated as:
Risk = 1 - exp (-CD/ x SIF) (4.3-41)
which for small risks is equivalent to (US EPA, 1989)
Risk = CDI X SIF (4.3-42)
The total carcinogenic risk, RiskT, from exposure to multiple carcinogens (for example PCE, TCE, and VC)
is calculated as the sum of the individual risks (US EPA, 1989)
RiskT =^Riskt (4.3-43)
A major exposure route for contaminated groundwater is water from wells in the dissolved plume area. The
contaminants contained in the water may be ingested directly, in drinking water, and if they are volatile,
they may be inhaled as the contaminant partitions from the water into the air in the house (McKone, 1987).
The cancer risk from inhalation is often as large or larger than the risk from ingestion (McKone, 1987;
Williams etal, 2004).
The CDI for both ingestion (CDIG) and inhalation (CDIH) is computed using the well water concentration
in mg/L averaged over the maximum exposure period,
^weU(t )dt (4.3-44)
lex max(CU-(el)
The lower limit in the integral restricts the exposure period to a maximum of tex . The upper limit, t, is
the time since the contaminant release occurred. Once the average tap water concentration is known, the
ingestion and inhalation cancer risks can be calculated using standard methods (see, for example, Maxwell
et al. (1998); McKone (1987); Williams et al. (2004).
The cancer risk from water ingestion assumes a daily water intake of WD (L/d), and a body mass of Mh (kg).
Then using the standard life expectancy, tlife, the CDIa is
W vt \
(4.3-45)
Typical values for WD,Mb, and tufe are 2 L/d , 70 kg, and 70 years, respectively. The ingestion risk is then:
Risk0 = 1 - exp (-CD/o X SIFG) (4.3-46)
where SIFG is the oral slope factor for the carcinogen.
Calculation of the inhalation exposure requires estimation of indoor air concentrations of the carcinogen
that result from water use in the house. The standard approach is to separately consider the shower stall, the
bathroom, and the remainder of the house (McKone, 1987). An empirical water to gas transfer efficiency
(TE) is used to relate the contaminant mass flowrate in water passing through parts of the house (shower,
-------
bathroom, house) to the rate of mass loading in the indoor air. Based on extensive data sets collected for
radon gas exposure, the TE for a volatile chemical is computed from the radon data with a correction for
the different Henry's constant and aqueous and gaseous diffusion rates. Of these parameters, the TE is most
sensitive to the aqueous diffusion rate, because that tends to limit the rate of mass transfer into indoor air
(McKone, 1987). Typical values for TE are in the range of 0.3 to 0.9, depending mainly on the nature of the
water use (for example a shower versus a dishwasher).
The CDIH is computed separately for the three main compartments (shower, bathroom, house) using the
average indoor air concentration of the carcinogen, Ca in units of mg/m3. For compartment kc, this
concentration is calculated from the average water concentraton using the tap water use rate W^ (L/hr), the
TEk, and the air exchange rate, VRk (m3/hr):
_ ( W, x TE, \
(4.3-47)
The CDIH for each compartment (CDIk ) depends on the daily exposure time in the compartment,
ETk (hr/d), and the inhalation rate, HR (m3/hr):
(4.3-48)
The risk from inhalation sums over the compartments
RiskH = 1 - exp
where SIFH is the inhalation slope factor.
(4.3-49)
The lifetime excess cancer risk slope factors vary widely among different chemicals, and they are often
revised or withdrawn. Conflicting values of SIF can be found in different sources in many instances.
Table 4.3-2 lists current (as of February, 2006) recommended inhalation and oral slope factors for PCE,
TCE, and VC from the California Office of Environmental Health Hazard Assessment (OEHHA, 2006).
Table 4.3-2. California cancer risk slope factors for PCE and its degradation products (OEHHA, 2006).
Chemical
Tetrachloroethylene (PCE)
Trichloroethylene (TCE)
Cis-l,2-Dichloroethylene (DCE)
Vinyl chloride (VC)
Inhalation Slope Factor
(mg/kg-day)-1
0.021
0.007
-not a carcinogen
.270
Oral Slope Factor
(mg/kg-day)-1
0.540
0.013
-not a carcinogen
0.270
4.3.2.5 Analytical/numerical approach including plume remediation and natural attenuation
An alternative approach to the analytical REMChlor Model described above is to couple the source mass/
source discharge relationship (Equation 4.3-6) to a numerical code that simulates advective/dispersive/
reactive dissolved solute transport (Equation 4.3-20). In this study, Equation 4.3-20 is utilized within the
framework of the Department of Defense Groundwater Modeling System (GMS) version 5.1. GMS is
employed to run the MODFLOW2000 and Reactive Transport in 3-Dimensions (RT3D) computer codes
which employ the fate and transport model.
-------
GMS uses the MODFLOW2000 computer code to model the three-dimensional steady state and transient
movement of groundwater (USGS, 2000). This is accomplished using a model based on the following
partial differential equation (from USGS, 2000):
_a/7 .......
" = ^ (4-3-50)
Where K^^K^, and A^ are hydraulic conductivities along the x, y, and z-axis respectively [LT1] ,
h = Potentiometric head [L] , gv is volumetric flux per unit volume [T1] , SS is specific storage of the
porous material [L"1] , and t is time [T] .
In this study, steady-state flow was assumed, so the potentiometric head did not change with time. In order
to apply this model to determine the steady-state head field, constant head boundaries were specified. Once
the boundary conditions were set, the model was run. The model uses the finite -difference method to solve
the steady-state version of Equation 4.3-50 and determine the head at each cell (USGS, 2000). The heads
calculated in MODFLOW are then used in the Darcy equation, along with the cell hydraulic conductivities,
to determine the groundwater velocities at each cell, which are then input to the fate and transport model.
The reaction term in equation 4.3-20 (rxri) is a source/sink term that is used to represent the biodegradation
of the chlorinated solvents. A number of models have been used to simulate chlorinated solvent
biodegradation kinetics (e.g. first-order, Monod, dual-Monod). For this study, the effects of source
zone remediation are simulated for a contaminated site at Dover AFB, DE, that was originally modeled
by Clement et al. (2000). Following Clement et al. (2000), a first-order model was used to simulate
biodegradation kinetics. Clement et al. (2000) found that the contaminant degradation at the Dover AFB
site could be accurately represented by first-order kinetics in four distinct reaction zones: anaerobic zone-1,
anaerobic zone-2, transition zone, and aerobic zone. The kinetic parameters used in this study are the ones
used by Clement et al. (2000) and presented in Table 4.3-3 for each of the four chlorinated ethenes (PCE,
TCE, DCE, and VC).
Table 4.3-3. Reaction zone degradation rates (Clement et al., 2000).
Calibrated degradation rate constants (day4)
First-order rate
constant
Kp (anaerobic)
KIl (anaerobic)
KI2 (aerobic)
KDl (anaerobic)
KD2 (aerobic)
Kvl (anaerobic)
KV2 (aerobic)
Km (anaerobic)
KE2 (aerobic)
Associated
contaminant
PCE
TCE
TCE
DCE
DCE
VC
VC
ETH
ETH
Anaerobic
zone-1
3.2xlO-4
9.0xlO-4
0.0
8.45xlO-4
0.0
8.0xlO-3
0.0
2.4xlO-2
0.0
Anaerobic
zone-2
4.0xlO-4
4.5xlO-4
0.0
6.5xlO-4
0.0
4.0xlO-3
0.0
1.2xlO-2
0.0
Transition zone
l.OxlO-4
1.125xlO-4
0.4xlO-5
1.625xlO-4
1.6xlO-3
l.OxlO-3
0.8xlO-3
0.3xlO-2
0.4xlO-2
Aerobic zone
0.0
0.0
l.OxlO-5
0.0
4.0xlO-3
0.0
2.0xlO-3
0.0
l.OxlO-2
-------
RT3D uses the following versions of Equation 4.3-20 to describe the fate and transport of the dissolved
chlorinated compounds (from Clement, 1997):
<4'3-51)
[PCE] - ^ [TCE] - "*" irc£]
(43-52)
(", (DCE])+ Ymn , cpD1, (pn , q>£1 are first-order anaerobic degradation rates for PCE, TCE, DCE,
VC, and ethene, respectively [ T1 ], cpra , cpD2, cp^ , cp£2 are first-order aerobic degradation rates for TCE,
DCE, VC, and ethene, respectively [ T1 ], and YT/P, YD/T, YV/D, YE/V are chlorinated compound yield under
anaerobic reductive dechlorination conditions (0.79, 0.74, 0.64, and 0.45 , respectively) [ MM"1 ].
The chlorinated compound yield coefficients (YB/A) are used to track mass. For example, YT/P = 0.79
indicates that the anaerobic reduction of 1 gram of chlorinated compound A (PCE in this instance) leads
to the production of 0.79 gram of chlorinated compound B (TCE in this case). In Equations 4.3-51
through 4.3-55, sorption is assumed negligible. The concentration of PCE in the source zone was used as
a boundary condition in the model. The PCE concentration was determined by the Zhu and Sykes (2004)
power function, while the concentrations of TCE, DCE, VC, and ethene at the boundary were set to zero.
All initial concentrations of chlorinated solutes throughout the aquifer were set to zero. The model was
verified by reproducing the results obtained by Clement et al. (2000).
The relationships between source remediation and contaminant concentrations at downgradient receptors
were estimated by combining the Zhu and Sykes (2004) power function model and the fate and transport
model in RT3D. Figures 4.3-14 and 4.3-15 use the model developed in this study and the Dover AFB
site conditions (Clement et al., 2000) to show the TCE and vinyl chloride (VC) concentration reductions,
respectively, achieved at a receptor 800 meters downgradient of the source zone versus the mass reduction
in the source zone.
-------
102
10°
1"-'
0 10 20 30 40 50 60 70 80 90 100
Source tfeEE Reduction (%)
Figure 4.3-14. Steady-state TCE concentration at varying source mass reductions for a monitoring well 800
meters downgradient of the source.
These figures also show the respective maximum contaminant levels (MCLs), often used as the remediation
goal, of the contaminants. The initial or maximum concentration for the PCE concentration in the source
zone used in this model is 120 mg/L. Also, the percentage of source mass reduction is equal to the portion
of mass removed from the source zone, defined as:
% Source Mass Reduction =
M
*100%
5,0
(4.3-56)
20 30 40 50 60 70 80 90 100
Figure 4.3-15. Steady-state vinyl chloride (VC) concentrations at varying source mass reductions for a
monitoring well 800 meters downgradient of the source.
By using Figures 4.3-14 and 4.3-15, we can estimate how much source mass removal is required to achieve
MCLs at a downgradient receptor. For instance, if F = 2.0, approximately 98% of the mass in the source
zone would need to be removed to reduce TCE concentrations downgradient to MCL. Note, however, that
-------
with 98% source-mass removal, VC concentrations at the well would be well below the MCL. Table 4.3-4
shows the approximate source mass reductions needed to attain the MCLs for the different contaminants.
For the Dover AFB site, the source mass reductions necessary to accomplish the remediation goals for VC
are consistently less than what is needed to achieve TCE goals and DCE goals are more easily attained
than VC goals. Note also the inverse relationship between F and the fractional mass reduction required to
achieve remedial goals.
Table 4.3-4. Percent mass reduction necessary to achieve remediation goals.
Compound
TCE
DCE
VC
MCL
(l-ig/L)
5
70
2
F
0.1
100.0
100.0
100.0
0.5
100.0
100.0
100.0
1.0
99.9
98.6
99.4
2.0
97.8
88.3
92.3
10.0
53.4
34.9
40.1
4.3.3 Model Assessments and Applications
4.3.3.1 Simulation of laboratory and field air sparging
Comprehensive data from laboratory and field scale air sparging experiments were examined using
TMVOC, a multicomponent compositional multiphase flow simulator (Van Antwerp et al., 2008; Pruess
and Battistelli, 2002). Many studies have used column data to examine mass removal, but it is important to
assess how accurately processes at that scale represent larger field scale processes. The laboratory and field
air sparging studies simulated in this work were chosen because they have comprehensive data sets and they
were designed to complement each other.
The controlled two-dimensional laboratory sparging test was simulated first. The capabilities of the local
equilibrium and dual-domain mass transfer approaches for representing the observed contaminant removal
rates were assessed. Using the understanding gained from simulating the controlled two-dimensional
experiment, the field experiment was simulated, and the applicability of the laboratory-scale mass transfer
coefficient to a field scale system was evaluated.
In a traditional numerical flow simulation, each element represents a single continuum. Within each
element local equilibrium is normally assumed, in which case there is no accounting for local mass transfer
limitations that may occur due to fingering of gas channels. Multiphase simulations of sparging patterns
show the gradational change in averaged gas saturation outward from the injection location during air
sparging (Figure 4.3-16). A more detailed representation may be that the air channel density is greater
near the injection location and the air channel density decreases with distance from the sparge well. Air
sparging studies have shown that channeling leaves some pore water relatively unaffected by the sparged
gas (Ji et al., 1993; Elder and Benson, 1999). Contamination (NAPL or dissolved) that is isolated from
the gas channels is not removed unless it diffuses through the water, into the channels. In general, kinetic
interphase mass transfer needs to be incorporated into a multiphase flow and transport model to accurately
simulate contaminant removal by air sparging, because it is not practical to resolve the individual gas
channels in a field-scale simulation.
-------
Model representation of sparse *oneya& saturations
Realistic channel ing.
where channel densi
increases near the
injection point
Figure 4.3-16. A simulated sparge zone in a single medium multiphase flow model versus realistic
channeling that occurs in air sparging. A modeled gas saturation distribution under constant
mass injection of air is on the left. Illustrations of pore scale channeling that occurs in a real
system are shown on the right (Van Antwerp et al., 2008).
A dual-domain approach (Falta, 2000) was used to incorporate kinetic interphase mass transfer into the
TMVOC multiphase flow code. In this approach, the 3-dimensional grid is divided into two volume
fractions: one that represents coarser grained fractions of the porous media where the gas channels are
created, and one that represents finer grained materials that remain nearly saturated (Figure 4.3-17). Due
to capillary forces (air entry pressure), gas will tend to preferentially flow through the coarser material
when both are initially water saturated. The properties of these two volumes can be chosen so that the
composite media behaves like a normal porous media. The two volume fractions are locally connected one
dimensionally, so that fluids and contaminant components can move between the two domains according to
pressure and concentration gradients. In the present work, the fine-grained fraction tends to contain mostly
water, so the mass transfer to the coarser grained fraction is limited mainly by aqueous diffusion.
-------
Fine-grained, high Sw
almost no gas flow
Aqueous and gas
phase diffusive
transport of
contaminant
- Coarse-grained media
- Fine-grained matrix
\
Coarse-grained, high Sg,
high gas velocity
Figure 4.3-17. Schematic of connections in a dual media grid. Both media are globally connected and
share a local connection. Also shown is conceptualization of dual-media processes in a grid
block (Van Antwerp et al., 2008.
Air sparging was studied in a 1-m long by 0.5-m high by 4-cm thick laboratory apparatus to investigate
removal rates and mobility issues of a DNAPL, PCE, during sparging in a controlled two-dimensional (2-D)
configuration (Heron et al., 2002). Reported system configuration, soil properties, and sparging conditions
were used as the simulation input. Chemical input parameters were obtained from published properties.
The sand-water-air characteristic/permeability parameters were selected to produce a sparge zone of similar
size and shape to photographs of the initial sparging period based on preliminary two-phase (air/water)
simulations. No soil permeability data were available for the medium sand used in the experiment, so it was
assumed to have an intrinsic permeability between 10-100 darcies, which is a typical permeability range for
a medium sand. Using the initial injection rate and selected capillary parameters, the intrinsic permeability
was adjusted to produce a sparge zone of similar size and shape to that appearing in photographs of the
initial sparging period. These two-phase simulations showed that an intrinsic permeability of 30 darcies
produced a gas plume most similar to the pictures. Model results were fit to observations of capillary
rise and the formation of the sparge zone. Contaminant injection and redistribution were simulated using
TMVOC along with the air sparging performance for varying injection rates, as reported in the study.
Simulated PCE vapor concentrations and cumulative mass removed were compared to the experimental
data.
A DNAPL source was created in the apparatus by rapidly injecting 100 ml of liquid PCE near the top of
the box. Lowering and raising the water level produced a vertically smeared zone of PCE. Experimentally
measured air sparging results from Heron et al. (2002) are summarized in Figure 4.3-18. The top part of
Figure 4.3-18 shows the measured offgas PCE concentration (open circles) during the experiment, while
the bottom part shows the cumulative mass of PCE removed (diamonds). The sparging rate was varied and
included three 8-hour long pulses of air and three 1-hr pulses.
For multiphase flow problems, such as air sparging, the effects of multiple phases in the pore spaces must
be considered. Generally, the relative permeabilities are assumed to be a function of the saturation in the
-------
respective phase. More details on the capillary function and relative permeability formulation are reported
in Van Antwerp et al., (2008).
In this modeling analysis of mass removal during air sparging, the two-dimensional grid consisted of
1,624 elements (28 rows high, 58 columns long) to accurately resolve the initial distribution of PCE. The
element dimension perpendicular to the 2-D plane of the experiment was equal to the thickness of the
apparatus. The exterior of the grid is a "no-flow" boundary. The initial water saturation and pressure
distribution are assigned by assuming gravity-capillary equilibrium.
A qualitative comparison of color photographs of the initial distribution of NAPL PCE, that was dyed-red,
to simulations of the NAPL redistribution patterns after injection were used to obtain the residual NAPL
saturation, since it was not measured. The residual NAPL saturation, used in the three-phase relative
permeability function, was adjusted until the simulated spilled PCE distribution approximately matched the
distribution inferred from the pictures. Air injection in the experiment occurred at a point approximately
12-cm above the top of the silt layer. An appropriately located element was chosen to inject air at a constant
rate. TMVOC allows for tabular input of time-variable injection rates, which were obtained from the
experiment.
o Effluent PCE concentration
TMVOC - Dual Domain
—TMVOC - Single Domain
Elapsed Time (days)
Added PCE mass
8 hr pulses
1 hr pulses
-•-PCE mass removed
^—TMVOC - Single Domain
TMVOC - Dual Domain
Alpha: 17/8.4
20 30 40 50 60
Days after start of sparging
70
80
Figure 4.3-18. Results from air sparging experiments (Heron et al. 2002). The top figure (a) shows
simulated effluent concentrations using single- and dual-domain approach with fine grid
(28 x 58) along with the experimental data from the 2-D air sparging test. In the bottom
figure (b), total mass removed as predicted by a dual-domain numerical simulation using
a fine grid along with experimental mass removed from the 2-D test. The alpha ratio
is representative of the capillary contrast between the coarse and fine-grained media
(Van Antwerp et al., 2008).
-------
Simulations with a single-domain local equilibrium model configuration over-predicted the contaminant
mass removal rate (Figure 4.3-18). Initial concentrations predicted by the model are similar to the
experimental data; however, the single-domain model is unable to resolve the rapid decrease in
concentrations after one day of sparging. The over prediction of the mass removal rates for the single-
domain model is even more pronounced for the cumulative mass removal calculations. Within the first
4 days the single-domain model predicts that 120 g of PCE were removed followed by the removal of
an additional 20 g. In the experiment, only 90 g of PCE were actually removed. The differences in the
cumulative mass removal between single domain and the experimental measurements (Figure 4.3-18)
became smaller after 20 days because in the single domain simulation nearly all of the mass of PCE in the
sparge zone had been removed, while in the dual-domain simulation, there was still PCE present in the
sparge zone beyond 20 days.
Properties from the single medium simulation were adjusted for the two domains of the dual-domain
simulation following the logic presented in Falta (2000). It was assumed that gas channels would form
in approximately 20% of the sand during air sparging. A volume fraction of 20% of the original element
was assigned properties of a coarse-grained media and the remaining volume fraction (80%) was assigned
properties of a fine-grained media to account for zones of stagnant pore water. The volume fraction in
which gas channels form cannot be readily measured, so it was a calibration variable. After adjusting
the contrast in dual-domain capillary properties (a'gv/), the effluent concentrations from the dual-domain
simulation were very similar to the laboratory data (Figure 4.3-18). As a result, the cumulative mass
removal as a function of time also closely matched the experimental data.
An equivalent mass transfer value of 4.85 x 10~5 s"1 was calculated for this experiment, which is in the range
reported by Braida and Ong (1998). This value is also close to that calculated for column studies (Falta,
2000), having the same ratio of interfacial area (Au) to interfacial distance (d2) . Since this laboratory-
scale study was designed to complement the field-scale study, this An I d2 ratio was used initially for the
field-scale simulation.
The next step was to use TMVOC to simulate a controlled air sparging field test. Both single- and dual-
domain approaches were used in an attempt to model the observed removal rates. Only the first 57 days of
operation, which included 5 operational changes (with 3 configurations and 1 shutdown), were simulated.
During the second part of the test, which was not simulated, water was re-circulated in the test cell.
Simulating this condition would require some code modifications to TMVOC.
This air sparging demonstration was conducted in a test cell at the Dover National Test Site (DNTS),
Dover Air Force Base, Dover, Delaware. The cell was nominally 4.6-m long by 3.0-m wide by 12-m deep
(15-ft by 10-ft by 40-ft). Lateral boundaries were formed by sheet piles driven through the formation and
into a clay layer 12.2-m below the ground surface. This clay layer created a lower confining boundary.
Investigators from Michigan Technological University performed the air sparging experiment in 1999
(Gierke et al., 2002). Constraints on the initial contaminant mass, injection/extraction flows, and the
comprehensive data collection make this air sparging study well suited for testing models.
The simulation model for this site assumes homogeneous geologic conditions, however non-reactive tracer
tests performed specifically in the field test cell (Falta et al., 2003) indicated zones of varying permeability,
with reported horizontal permeabilities ranging from 1 x 10~n m2 to 8 x 10~13 m2 (Applied Research
Associates, 1996). A homogeneous model is certainly an oversimplification of the geology, but assuming
a homogeneous soil system allows for the calibration aspects to be performed more straightforwardly.
Furthermore, detailed knowledge of local heterogeneities is usually not available at real field sites and was
not at this site. Although the laboratory system described previously was modeled as isotropic, the field site
was assumed to have vertical anisotropy.
Since the injection and extraction gas flow rates were known and fixed in the simulation, the intrinsic
permeability was adjusted using a preliminary multiphase flow simulation to produce air pressures in
the wells similar to the measured values reported from the field. The modeled pressures were within 5%
of observed pressures (25% of the observed vacuum pressure) when using an intrinsic permeability of
2 x 10"11 m2 in the horizontal (x ,y ) direction and 1 x 10~n m2 in the vertical (z) direction.
-------
The sparging system injected air into the lowest 1-foot interval of the well screen (11.9-m to 12.2-m bgs).
A point mass injection was used to simulate the injection of 80% relative humidity air. Measured flow rates
from the experiment were specified as the model input. Three different configurations were modeled during
the simulated 57 days. During the experiment, combined flows through the extraction vents (33 scfm)
were kept at approximately twice the rate of injection to prevent the sparge well vapors from emerging
through the cell surface. These wells were also modeled using a specified mass rate of extraction based on
the measured rates. The extraction wells were screened from 6.1-m to 12.2-m bgs, however, production
of air occurs only from the screened interval above the saturated zone. Initially, this interval was 4.6-m but
shortened as the water table mounded upward as the sparging zone developed (around 0.4-m, depending on
injection/extraction well configuration).
A DNAPL source was created by injecting 107 kg of PCE into the test cell (Gierke, et al. 2002). The
goal of the PCE release was to create a heterogeneously distributed, residual-NAPL contaminated zone
in the lower-most 1.5 m) of the coastal sand deposits above the underlying clay layer. PCE was injected
at a depth of 10.7 m bgs through 3-mm diameter injection ports distributed across the cell. Details of
the resulting DNAPL structure are unknown so the PCE was emplaced in the system based on reported
injection locations and volumes.
A traditional single-domain, local equilibrium model was tested based on the parameters and sinks and
source described. As happened for the laboratory experiment, the field-scale, single-domain simulation
over-predicted mass removal (Figure 4.3-19). The over-prediction is because subgridblock-scale mass
transfer limitations are not accounted for in a single-domain approach. Simulated PCE removal (red line)
for the first few days matches the experimentally derived mass removal rate fairly well, but after about 5
days the simulation results diverge from the actual measurements. The simulation predicted the total mass
removal over 57 days to be approximately 65 kg, whereas the actual mass removal was 39 kg. Differences
between the predicted and experimental results are likely due, in part, to the initial contaminant distribution
and soil heterogeneities, and local diffusive limitations.
Properties from the single medium simulation were adjusted for the dual domains in the same manner as in
the laboratory experiment simulation. The coarse grained domain was again set equal to 20% of the total
volume. Again, it is unknown whether precisely 20% of the sand volume would form gas channels, but the
estimate is reasonable based on Falta (2000) and the observed increase in the water table elevation. The
dual-domain properties were based upon the single medium soil characteristics pertaining to the coastal
sediments at the field site. A dual-domain model was run initially using the equivalent mass transfer rate
(Au Id2) ratio determined for the laboratory experiment (75,000), but the model still over-predicted the
PCE removal rate as shown in Figure 4.3-19 (yellow line).
o Experiment Data
TMVOC - Single Domain
TMVOC - Dual Domain, A/d: 75000
TMVOC - Dual Domain, A/d: 75
21 28 35
Elapsed Time (Days)
42
49
56
Figure 4.3-19. Total experimental mass removal and simulated removal with TMVOC using a dual-media
approach with A12/d2 ratio from the laboratory-scale air sparging simulation (75,000) and the
adjusted A12/d2 ratio (75) (Van Antwerp et al., 2008).
-------
The single-domain, local-equilibrium model predicts that almost all of the contaminant within the sparge
zone was removed after approximately 15 days. A three-dimensional view of the DNAPL and gas
saturation distribution at 15 days shows that while the two simulation approaches predict similar overall
gas flow patterns, the dual-domain model results in a slower removal of DNAPL from the sparge zone. The
pooled PCE NAPL in the lower portion of the simulation remains unaffected throughout the course of both
simulations.
By reducing the ratio of An Id2 to 75 (equivalent to increasing d2 and /or decreasing An ) to reduce the
effective mass transfer coefficient, a better match of the total mass removal was obtained (Figure 4.3-19,
green line). Calibration of the model to the laboratory data depended on the contrast in gas saturations
between the two domains, however, in calibration of the field-scale model the gas saturation contrast had a
smaller effect than the Au Id2 ratio (Van Antwerp, 2006). The simulated mass removal using a dual-domain
approach was similar to the actual data for the first flow configuration. At the first flow configuration
change, occurring at 15 days, the model predicts a more abrupt increase in the mass removal rate than
was observed. The simulated mass removed from this time forward is approximately 5 kg higher than the
observed. Flow configuration 2, which begins at day 15, involves ceasing air injection into one well and
doubling the flowrate at the other injection location.
In summary, a multi-phase numerical simulator was adapted to account for mass transfer limitations through
a dual-domain approach. Using TMVOC as a single medium model for contaminant removal scenarios
over-predicted the mass removal rates due to a local equilibrium assumption in each element. The dual-
domain-model based simulation successfully matched data from the mass removal of a dense non-aqueous
phase liquid for laboratory and field-scale air sparging experiments.
Although moisture characteristics of the sand used in the laboratory were unknown, the constraints of the
known capillary rise, photographs of sparge zone width response to injection rate, and initial contaminant
distribution provided enough information for reasonable estimation of soil parameters through model
calibration. Contaminant mass removed during operational changes in air injection was accurately resolved
in the dual-domain simulations, but not in single-domain simulations assuming local equilibrium. During
pulsed operation in the two-dimensional experiment, the dual-domain model could accurately predict the
time for concentrations to tail off after a pulse of air, although exact matches of concentrations during the
pulses could not be obtained.
The A12 /d2 ratio that was fitted to the laboratory data, (which is analogous to fitting a mass transfer
coefficient to a system) was 1000 times larger for the field-scale air sparging demonstration. This indicates
that mass transfer rates measured at a bench scale may not be applicable to field scale processes. One
explanation for the difference in mass transfer coefficients may be the formation of a relatively uniform
sparge zone in the two-dimensional box that is directly in contact with the NAPL PCE. In the field system
there are many more heterogeneities affecting the sparge zone formation, and larger zones may be excluded
from gas flow.
One drawback of this multiphase dual-domain approach to modeling mass removal in air sparging is the
relatively large amount of site parameter data needed. However, as this study shows, a moderate amount of
site information can still produce reasonable results if the data exists for model calibration at the field scale.
Calibration of the model is important and using laboratory derived parameters may not be accurate.
4.3.3.2 Simulations of cosolvent flushing
This section presents UTCHEM simulations of a full scale field cosolvent flooding operation at a site with
hydraulic control (Liang and Falta, 2007; Jawitz et al., 2000). The field effort was conducted at the former
Sages Dry Cleaner site in Jacksonville, Florida. The site was contaminated by PCE with an estimated
DNAPL volume of 68 liters. Approximately 34,000 liters of 95% ethanol were injected at the site during
the flooding period of 3 days, followed by a 5-day water flood. The remediation effort was terminated after
8 days, and it resulted in the removal of approximately 42 L of PCE. It is believed that the remaining PCE
was located in zones with limited hydrodynamic accessibility (Jawitz et al., 2000).
The numerical simulator used in this work, UTCHEM (version 9.3) was obtained from the University of
Texas (Delshad et al., 1996). UTCHEM is a three dimensional, multicomponent, multiphase, compositional
-------
surfactant flooding simulator that has been used both for environmental applications as well as enhanced
oil recovery (Brown et al., 1994; Camilleri et al, 1987; Delshad et al., 1996; Freeze, et al., 1994; Liu et al.,
1994; Pope and Nelson, 1978). UTCHEM uses the finite difference method with temporal discretization
that is implicit in pressure and explicit in concentration (IMPES-like). UTCHEM uses a total variation
diminishing (TVD) scheme that is approximately third-order in space to minimize numerical dispersion and
grid-orientation effects (Datta Gupta et al., 1991; Harten, 1983; Liu et al., 1994).
To simulate cosolvent flooding and the resulting transport processes, the equilibrium phase partitioning
behavior between DNAPL, water and cosolvent is of central importance. A ternary phase diagram can be
used to describe the equilibrium phase partitioning behavior for surfactant flooding as well as for cosolvent
flooding. A general discussion of cosolvent flooding performance based on ternary phase behavior has
been presented elsewhere (Falta, 1998). In UTCHEM, the phase behavior model is based on Hand
representations of the ternary phase system (Hand, 1930).
The phase behavior of the ethanol/PCE/water ternary system has been extensively studied (Hayden et al.,
1999; Lunn, 1998, Lunn and Kueper, 1997; Van Valkenburg, 1999). As shown in Figure 4.3-20, the ethanol/
PCE/water system exhibits the typical Type II (-) phase behavior with negative tie-line slopes. This tie line
behavior indicates that the ethanol preferentially partitions into the aqueous phase rather than the DNAPL
phase. An aqueous phase with a high ethanol concentration can dissolve a large amount of DNAPL,
increasing the solubility by several orders of magnitude. Thus the main recovery mechanism for this phase
environment is enhanced dissolution, although separate phase DNAPL mobilization is also possible at high
injected alcohol concentrations (Falta, 1998).
The UTCHEM surfactant phase behavior package was fit to the ethanol/PCE/water phase behavior data of
Hayden et al. (1999). The comparison of the UTCHEM representation and the experimental measurements
are shown in Figure 4.3-20. It is important to recognize that the simplified Hand model used in UTCHEM
may not exactly represent all parts of the ternary phase behavior. For example, in our fit of the Hayden
et. al. (1999) data, we focused primarily on matching the height of the binodal curve, and the slopes of
the tie lines. Because the actual curve is asymmetrical, our best fit does not perfectly match the aqueous
phase composition leg of the binodal curve, so the calculated DNAPL solubilities are only approximate. It
is possible to more precisely fit the aqueous phase composition leg of the binodal curve, but this tends to
result in a binodal curve that is substantially higher than the laboratory data indicate.
From the phase diagram, the maximum concentration of ethanol to form a two-phase region is 0.725
(volume fraction). The PCE concentration at the plait point is 0.854 (volume fraction). These are important
input parameters used to describe the phase behavior in the numerical model. As shown in Figure 4.3-20 the
UTCHEM phase behavior package can reasonably account for the phase behavior of the water/ethanol/PCE
system.
-------
Ethanol
100A0
Plait Point
Water 100 90 so ?o so 50
40
30 20
10
- -o - - Measured (Hayden et al., 1999)
- Model (UTCHEM)
Figure 4.3-20. Modeled and measured ternary phase diagram for the system of water-ethanol-PCE (Liang
and Falta, 2008).
The in-situ cosolvent flooding effort simulated here was conducted at the former Sages Dry Cleaner site
in Jacksonville, Florida. The description of site background and field test operation is based on the work
reported by Sillan et al. (1999) and Jawitz et al. (2000). The DNAPL source zone was characterized
both before and after alcohol flooding using soil cores, groundwater samples, and interwell partitioning
tracer tests (Jawitz et al., 2000). The PCE source area in plan view was approximately 7.3 m by 2.7 m
(Figure 4.3-21) (Jawitz et al., 2000). The total estimated volume of PCE was 68 liters (109 kg) at the
beginning of the test, equivalent to an overall average saturation of 0.004. Nine wells were used in this
ethanol flooding field test, three injection wells (IWs) in the center and six recovery/hydraulic containment
wells (RWs) along the perimeter of the source area (Figure 4.3-21). Approximately 34,000 liters (2 PVs) of
95% ethanol were delivered to the three injection wells during the flooding period of 3 days, followed by a
5-day water flood. The field test was terminated after 8 days, and resulted in about 42 liters (67 kg) of PCE
removal.
The model domain was taken as a 24.7 m x 13.8 m x 10.7 m region discretized into 12,500 (25 x 20 x 25)
cells with fine spacing in the treatment zone and coarse spacing outside the source zone (Figure 4.3-21).
Constant atmospheric pressure was maintained at the top boundary. The bottom was a no flow boundary
associated with a clay confining unit. The left and the right sides were constant head boundaries that match
the regional gradient, with regional flow toward the west under a natural hydraulic gradient of 0.0025. The
other two sides were no flow boundaries that are parallel to the regional flow direction.
Based on the positions of the well packers, the upper zone and the lower zone of each injection well
were modeled as two separate wells. A total of 12 simulation wells including 6 extraction wells were
used. The injection and extraction flow rates were based on the field data. The initial DNAPL saturation
distribution was interpolated from soil core data and the estimated initial total DNAPL mass. Measured soil
concentrations of PCE (mass fraction) at different depths were used to calculate DNAPL saturation.
-------
a
extent of
zone: 1.5
flow direction
nated horizontal
DNAPL source
mX2.7m ^
•s
\
X
-N
t
~RW
L
"
2
IW001 IW8°2
24.7m
ra
m
07
H •••****
h
- E
/
^
Injection Well (IW)
Recovery Well (RW)
t.7m
Water table
Figure 4.3-21. Plan view of model domain with the site map of well and MLS locations within the
approximated DNAPL source zone at Sages (adapted from Jawitz et al. 2000) (Liang and
Falta, 2008).
Three cases were considered for the Sages simulations. First, a basic model was built as a two-unit system
with a simple permeability structure (referred to as the base case). To gain a better approximation to
the actual permeability distribution in the field, hydraulic conductivities estimated with a borehole flow
meter were used to define a layered case (referred to as the heterogeneous case). Adjustment of zones of
horizontal permeability based on the field test well flow rates was also included in this heterogeneous
case. Finally, the heterogeneous flow model was calibrated by adjusting initial DNAPL distribution using
extraction well PCE breakthrough curves (referred to as the calibrated heterogeneous case).
-------
The basic model was built as a two-unit system with a simple hydraulic conductivity field. The average
hydraulic conductivity values from slug tests of 6.10 m/day for the upper unit (0-9 m bgs) and 3.05 m/day
for the lower unit (9-10.7 m bgs) were used. An anisotropy ratio of 0.1 for vertical hydraulic conductivity
to horizontal hydraulic conductivity was assumed. The overall porosity was set to 0.3. The initial DNAPL
saturation distribution was interpolated based on soil core data, and distributed so that the initial total
DNAPL mass was 109 kg.
Base case
Combining the flow weighted average of the six recovery wells yielded the total effluent concentrations.
For this simple base case, the model overestimated total ethanol recovery at early times and underestimated
PCE concentrations. However, this was a completely uncalibrated model with a very simple permeability
structure.
From the simulation results of the basic model, the underestimation of ethanol effluent percentages from
RW003 and RW004 suggest that the area around these two wells may be less permeable. Because of the
lower permeability, RW003 and RW004 would extract a smaller amount of water from outside the target
zone, thereby yielding the higher ethanol effluent percentages. On the other hand, the over predicted ethanol
effluent percentages from RW006 and RW007 suggest that the formation in the vicinity of these two wells
could be more permeable. Thus RW006 and RW007 would extract much more water from outside the
pattern, diluting the ethanol, and yielding the lower ethanol effluent percentages. The fact that the flow
rates for RW006 and RW007 were about 2-3 times greater than those for RW003 and RW004 (with the
same drawdown in the well) supports this hypothesis.
Heterogeneous Case
To gain a better approximation to the actual permeability that is present in the field, hydraulic conductivities
estimated with a borehole flow meter (Mravik et al., 2003) were used to condition this case. For each
model layer, the harmonic mean calculated from field measurements was used to calculate the vertical
hydraulic conductivity, and the weighted average was used to calculate the horizontal conductivity. If
only one measurement was available for a model layer, the measurement was used for horizontal value,
and an anisotropy factor of 0.1 was assumed. Both the field flow rates and the base case simulation results
suggested that the area around RW006 and RW007 might be more permeable than the area around RW003
and RW004. Thus, the adjustment of horizontal permeability to make a higher permeability zone around
RW006 and RW007 and a lower permeability zone around RW003 and RW004 was included in the
heterogeneous case. The same initial DNAPL saturation distribution was used as in the base case.
The comparisons of simulated and field test results for PCE effluent concentrations and ethanol percentages
are shown in Figure 4.3-22 for recovery wells RW002, RW003, RW004, RW005, RW006, RW007, and
the total extraction fluids. For the total extraction fluids (Figure 4.3-22), the predicted overall ethanol
percentages show good agreement with the field data, but the total PCE concentrations were underestimated
by about a factor of two or three.
Compared to the base case, the layered heterogeneous case did a much better job of predicting the ethanol
percentages, but still underestimated the PCE concentrations and the mass removal. Nevertheless, this
heterogeneous model still has not been calibrated, but it used more information from the experimental site
to develop a more accurate flow model.
-------
g 900
1
2, |
I
Time (day)
0 affnr- i 1 1 j==i—. ' "rm-aaa 0
012345678
Time (day)
U
g
345
Time (day)
a 600
O
w
U 300
1
—
J
r n
j
..
^/Cd
r
r i
L J
fehnqgp
+
— ©-
%¥tof^
PCE
-Mod PCE
-Ethanol
• Mod Ethanol
i
1
1
1
1
1
1
L
1
1
k 1
20 1
,200 -__T . r__
a
U
§
RW007 |
Ttoe(day)
o mfnmi^—i 1——i i i iT TTI o
012345678
Time (day)
U
too
e(day)
Figure 4.3-22. PCE effluent concentrations and ethanol effluent percentages for RW002, RW003, RW004,
RW005, RW006, RW007, and total fluid (heterogeneous case) (Liang and Falta, 2008).
-------
Calibrated Heterogeneous Case
Uncertainty in the initial DNAPL saturation distribution estimated from soil core data could account for
the underestimation of the PCE effluent concentrations in the heterogeneous case. Since the soil core data
only give PCE concentrations at a few discrete locations, it is not possible to know the details of the actual
PCE distribution prior to remediation. Therefore, for this calibrated case, the initial DNAPL distribution
was adjusted based on the PCE mass represented by breakthrough curves from the extraction wells. The
same amount of total initial PCE mass (109 kg) was maintained for this adjusted initial DNAPL saturation
distribution, and the individual soil sample concentrations are still honored. The adjusted initial DNAPL
saturation distribution was used in the calibrated heterogeneous model, which used the same permeability
structure as the heterogeneous case.
Only a finite number of initial DNAPL saturation distributions were tested in this calibration effort, and
we acknowledge that it is possible that a better calibration could be achieved with more realizations.
Furthermore, it is possible that several somewhat different initial DNAPL distributions might result in a
similar overall match of the field data.
Comparisons of simulated and field test results for PCE effluent concentrations and ethanol percentages are
shown in Figures 4.3-23 for recovery wells RW002, RW003, RW004, RW005, RW006, RW007, and the
total extraction fluids. For the total extraction fluids (Figure 4.3-23), the predicted total ethanol percentages
show good agreement with the field data, and the total PCE concentrations are close to the field data.
Compared to the results of the uncalibrated heterogeneous case, this calibrated model with a new adjusted
initial DNAPL saturation distribution showed a substantially better prediction of the PCE effluent
concentrations and PCE mass recovery. Unfortunately, the lack of practical field methods for characterizing
DNAPL distributions makes precise well-by-well prediction of contaminant effluent concentrations difficult
or impossible.
The results of a root mean squared error (RMSE) analysis for the three cases also shows this stepwise
improvement of the model predictions. Compared to the base case, the RMSE on ethanol percentages
for the more refined permeability model is reduced significantly from 7.4% to 3.83%. The final model,
which calibrated the initial PCE distribution, shows a decrease of the RMSE on PCE concentrations from
171.9mg/Lto72.08mg/L.
-------
40 g
10 Ed
12345678
Time (day)
u
Ed
30
s
23456
Time (day)
S 600
U
H 300
234567
Time (day)
U
lid
U
RW007 I
10 5
01234567
Time (day)
Figure 4.3-23. PCE effluent concentrations and ethanol effluent percentages for the calibrated
heterogeneous case (Liang and Falta, 2008.
-------
4.3.3.3 Dissolved chemical discharge from fractured clay
Recent experimental and theoretical studies have demonstrated that DNAPLs present in fractures in low
permeability fractured porous media can dissolve and diffuse from the fractures into the matrix (Parker
et al., 1994; 1997; Ross and Lu, 1999; Slough et al., 1999; Esposito and Thomson, 1999; O'Hara et al.,
2000; Reynolds and Kueper, 2001; 2002; 2004; Parker et al., 2004). This relatively rapid mass transfer
into the porous matrix occurs because of the sharp concentration gradients that exist after a DNAPL enters
a fracture. Once the matrix has been contaminated, the contaminant can slowly diffuse back into water
flowing through the fractures, leading to a long-term source of groundwater contamination. It is possible
that after all of the DNAPL has been removed from a contaminated site, this dissolved mass trapped in the
low permeability matrix could continue to act as a groundwater source term, perhaps for hundreds of years
(Parker et al., 1997; Esposito and Thomson, 1999; Reynolds and Kueper, 2002).
Reynolds and Kueper (2002) provide some compelling analyses of the possible long-term behavior of
DNAPL chemicals in fractured clays. They considered a case where 5 ml of a DNAPL were spilled into
a 30 |im aperture fracture in a 3 m thick clay aquitard. Following the DNAPL release, they simulated the
immediate flushing of the fracture with clean water under a gradient of 0.01. Because of the very large
surface area and small volume of the fracture, the DNAPL only persists in the fracture for a few hours or
days, and essentially all of it diffuses into the clay matrix, despite the clean water flush of the fracture.
However, the water exiting the fracture following the clean water flush is contaminated, and a significant
result of their analysis is that the chemical concentration leaving the aquitard remains above regulatory
limits for 1,000 years or more under this scenario.
The purpose of this study was to explore the matrix diffusion process further, and in particular to assess
the likely impact to a potable water aquifer that might underlie a contaminated aquitard (Falta, 2005). The
primary scenario that will be considered here is shown in Figure 4.3-24. It is assumed that a DNAPL spill
has occurred on top of a fractured clay aquitard, and that a small amount of the DNAPL has penetrated the
fractures, but that no DNAPL has reached the aquifer itself. Under these conditions, it is well known that
the DNAPL will disappear relatively quickly owing to matrix diffusion effects (Parker et al., 1994; 1997;
Reynolds and Kueper, 2002). If a downward gradient is imposed across the aquitard (caused, for example,
by groundwater pumping), then contaminated water leaving the aquitard will serve as a long-term source of
groundwater contamination.
Single fracture simulations were used to examine the sensitivity of the contaminated fractured clay system
to fracture flow rates and contaminant decay in the clay matrix. These simulations were performed
using the T2VOC compositional multiphase flow simulator, developed at Lawrence Berkeley National
Laboratory (Falta et al., 1995). The simulation grid is similar to that used by Reynolds and Kueper (2002);
it explicitly discretizes a single fracture and the clay, with a very fine grid spacing (-100 |o,m) in the clay
near the fracture. A total of 2,700 gridblocks are used in the simulation. The fracture aperture was either
30 or 100 |im, the clay was 3 m thick, with the same properties as in the Reynolds and Kueper case: a
very low permeability (1.0 x 10~17 m2), ahigh fraction of organic carbon (0.01), and amoderate porosity
(0.3). The DNAPL is given the properties of TCE, the aqueous diffusion coefficient was set equal to
1.0 x 10"9 m2 /s , and the tortuosity is taken to be 0.67 following Millington and Quirk (1961). The fracture
apertures of 30 and 100 |im used in this study fall within the 1 to 100 |im aperture range typically reported
for shallow clay deposits (Parker et al., 1994).
-------
Fractures
Figure 4.3-24. Conceptual view of a fractured clay aquitard. While the aquitard has a low effective
hydraulic conductivity, water can flow through the fractures into the aquifer below.
Contamination in the clay aquitard can then contaminate the upper part of the aquifer as
shown (Falta, 2005).
To focus mainly on the long-term behavior of this TCE source, the DNAPL emplacement process is
simplified somewhat compared to Reynolds and Kueper (2002). Instead of simulating a 5 ml DNAPL
spill, the initial condition in these simulations is a DNAPL residual saturation of 10% or 3% throughout the
fracture, depending on its aperture. This corresponds to an initial TCE mass of 13.56 g for either fracture.
Because of symmetry, only half of the fracture/clay is simulated, and a no-flow outer boundary at 1.45 m
corresponds to an equivalent fracture spacing of 2.9 m. The simulations each consist of two parts. In part
one, the 13.56 g of TCE DNAPL is placed in the fracture with no water flow and allowed to equilibrate for
a period of 20 years. Interestingly, the DNAPL phase disappears from the fracture in a few hours in this
scenario, as a result of aqueous phase diffusion into the clay matrix. This is somewhat faster than Reynolds
and Kueper (2002) simulated, but their DNAPL saturations in the upper part of the fracture were higher
because of their DNAPL release scenario. As Reynolds and Kueper (2002) have pointed out, behavior of
DNAPL in the fracture is very sensitive to the precise initial and system conditions.
Figure 4.3-25 shows the aqueous concentration of TCE in the fracture over the 20-year equilibration period.
Once the DNAPL phase disappears, the concentration in the fracture shows a logarithmic decline down to a
value of about 2270 |o,g/L, a reduction of a factor of 440, due to diffusion into the matrix. All of the initial
TCE mass remains in the volume. After 20 years, a strong concentration gradient still exists in the clay, and
significant dissolved concentrations have reached the model boundary at 1.45 m from the fracture.
TCE aqueous concentration in fracture
with no water flow through fracture
10000000
1000000
100000
10000
1000
0.00001 0.0001 0.001
0.01 0.1
time, years
Figure 4.3-25. Simulated TCE aqueous concentration in the fracture during the 20-year equilibrium period
for a 30 |im fracture with no TCE decay and a c\ayfoc of 0.01 (Falta, 2005).
-------
Part two of the simulation models clean water flushing of the fracture under a vertical gradient of 0.01. For
30 jam, the equivalent parallel plate permeability is 7.5 x 10~n m2, so the imposed gradient results in a Darcy
velocity of 0.64 m/day, or a water flow rate of 19.2 mL/day. At 100 years, the concentration gradients in
the clay have become smaller, but they are still present. For the 30 |im fracture case, after 100 years of
flushing, the matrix and fracture effluent concentration (Figure 4.3-26) remains fairly high, and only about
7% of the TCE has been removed from the system. Even after 2,851 years (the end of the simulation),
for the 30 |im fracture, high concentrations of the TCE remain in the system, and the fracture effluent
concentration is still about 240 |o,g/L (Figure 4.3-26).
As noted by Reynolds and Kueper (2002), the contaminant removal by fracture flushing occurs in this case
on a time scale measured in the thousands of years. However it should be recognized that this scenario did
not consider any type of chemical or biological degradation of the TCE. It is not known at what rate, or by
what mechanisms chlorinated solvents might degrade in a low permeability matrix, but it is plausible that
some degradation could take place. To illustrate the possible importance of TCE degradation, a second
simulation was run, using identical parameters, but assuming that the TCE can degrade in the aqueous
phase, with a half life of 10 years. This half life is substantially longer than what is often observed in
dissolved TCE plumes at field sites (Weidemeyer et al, 1999), and the production of daughter products is
neglected. This simulation also uses a lower fraction of organic carbon in the clay compared to the previous
simulation, 0.002 instead of 0.01. The same initial TCE mass is present in both cases. The fracture effluent
from this case with TCE degradation is shown as the curve with grey diamonds in Figure 4.3-27. While the
early time behavior is similar to the simulation with no degradation, a large effect is seen beginning at about
10 years. The fracture effluent concentration drops below 1 |lg/L in less than 300 years in this case.
10000 -E
1000
100
10-
O numerical 2-D, 30 urn
analytical, 30 um, gamma=2
D numerical 2-D, 100 um
—— analytical, 100 um, gamma=2
10 100 1000
time, years
10000
Figure 4.3-26. TCE concentration in the fracture effluent for the 30 |im and 100 |im fractures. There is no
decay of TCE in the clay matrix and the/oc = 0.01 (Falta, 2005).
The amount of flushing that takes place in a 30 |im fracture is relatively small because of the small area of
the fracture. As the fracture aperture increases, the flow rate through the fracture increases approximately
as the cube of the aperture. Thus, a 100 |im fracture (which represents the largest aperture typically found
in a fractured clay) would conduct 37 times more water than a 30 jam fracture under the same gradient. A
pair of simulations was performed using a 100 jam fracture aperture. The other system parameters were
maintained at their previous values, and the total TCE mass initially in the fracture was the same as in the
previous cases. Following the 20-year equilibration period, the fracture was flushed at a Darcy velocity
of 7.1 m/day, or a flow rate of 710 mL/day, corresponding to a hydraulic gradient of 0.01. The simulated
fracture effluent is shown in Figure 4.3-26 for a case with no TCE degradation (white squares), and in
Figure 4.3-27 for a case with TCE degradation (white squares). Here, the larger rate of flushing leads to a
-------
more rapid depletion of the TCE mass in the clay, and the curves begin to drop after about 1 year. Although
the decline in fracture concentration is more rapid in this case than it was for the 30 jam case, the time
scale for achieving regulatory limits in the fracture is still on the order of 1,000 years for the case without
any TCE degradation. As would be expected, including TCE degradation speeds up this process, and the
fracture concentration drops below 1 ng/L at about 150 years.
10000
1000
numerical 2-D, 30 urn
— analytical, 30 um, gamma=2
D numerical 2-D, 100 um
analytical, 100 um, gamma=2
100
100
time, years
1000
10000
Figure 4.3-27. TCE concentration in the fracture effluent for the 30 jam and 100 jam fractures. The
dissolved TCE has an aqueous phase half-life of 10 years in the clay matrix and the
foc = 0.002 (Falta, 2005).
4.3.3.4 Monte Carlo simulations of source behavior with remediation
Addressing the benefit of DNAPL source remediation effort on plume response is a difficult challenge due
to the intrinsic complexities and uncertainties in the process and performance of these remediation efforts
(Falta et al., 2005a, Falta, 2007). Uncertainties and variability in parameters arise from hydrological and
biogeochemical properties (e.g., hydraulic conductivity), from the effects of remediation (fraction of source
removal, effect of source removal on discharge, rate of natural or enhanced attenuation in plume), and from
the site condition and history (size and timing of contaminant releases, discharge to groundwater).
Several three-dimensional multiphase numerical models of source zone behavior that account for these
parameters have been developed (see Falta et al., 1992, 1995; Delshad et al., 1996). These codes have been
useful for improving the understanding of the physical and chemical processes that control the contaminant
fate, transport, and removal in the source zone. These models use the deterministic approach, taking a
single value for each parameter (perhaps in each gridblock), leading to a single prediction of the system
response. Typically, these are selected to be "best estimates" or sometimes "worst case values". In reality,
the hydrogeologic, geochemical, and process parameters used in a model are either variable, uncertain, or
both variable and uncertain. The deterministic model does not reflect the nature of overall uncertainty in a
simulation. The best approach to capture this inherent uncertainty is probabilistic simulation (e.g., using the
Monte Carlo method), in which the uncertain parameters are represented by probability density functions
(PDFs), and the result itself is also represented by a probability distribution.
Predicting the effect of the source remediation on plume behavior also has been limited by the lack of
tools that explicitly link source and plume remediation (Falta et al., 2005a, Falta, 2008). A new analytical
model, called REMChlor, coupling source and plume remediation, was developed to evaluate the impact of
source and plume remediation at a more generic and strategic level (Falta, 2008). This transport model is
not specific to any remediation technology. The contaminant source remediation is simulated as a fractional
-------
removal at a future time; plume remediation is modeled considering decay rates of parent and daughter
compounds that are variable in space and time (Falta, 2008).
In the present work, the DNAPL source function from the new analytical transport code was linked to a
commercial probabilistic simulation software (GoldSim, 2007). Probabilistic simulations were conducted to
evaluate the influence of the uncertainty in parameters on the effectiveness of source remediation in terms
of the source discharge.
Among several commercial packages, GoldSim software (GoldSim, 2007) was chosen to perform
probabilistic simulations. GoldSim is widely used in the nuclear industry for conducting performance
assessment calculations. The analytical FORTRAN code for the DNAPL source function, REMChlor, was
linked to GoldSim through a dynamic link library (DLL) approach. The probabilistic simulation model
was developed by solving Equations (4.3-6) and (4.3-7) analytically (see Falta et al, 2005a), and using that
analytical solution as the simulation model in GoldSim.
The solution of Equation (4.3-7) with the power function (Equation 4.3-6) can be used to predict the time-
dependent depletion of the source zone mass by dissolution. The time-dependent mass is then used in
Equation (4.3-6) to calculate the time-dependent source discharge. This model can simulate a wide range of
source responses to mass loss, and it can include effects of remediation (Falta et al., 2005a).
The input variables for the current probabilistic DNAPL source model include the initial source DNAPL
mass (Mso), the initial average source concentration (CJO), the exponent of the power function ( F ,
gamma) relating source strength to source mass, the source decay rate by processes in addition to
dissolution (ks), the water flow rate through the source zone (Q), and the mass fraction (X) of the source
removed by a source remediation effort. All of these input variables are uncertain, but the largest uncertainty
is likely to be in the power function exponent, F , in the initial DNAPL source mass, Mso, and in the
fraction of source removed by remediation, X.
PDF of Mo
PDF of Gamma
0.0015
0.0010
0.0005
0.0000
500 1000 1500
Mo (kg)
2000
Figure 4.3-28. Two input probability distribution functions (without source remediation).
Two examples are given to illustrate the probabilistic DNAPL model. The first example is a probabilistic
simulation using the DNAPL source function model, Equations (4.3-6) and (4.3-7) without any remediation
effort (only natural rates of dissolution) for a hypothetical DNAPL site. The power function exponent
was assumed to have a log-normal distribution with a mean value of 1, and a range between 0.5 and 2 for
most values (Figure 4.3-28). The initial DNAPL source mass was treated as a stochastic variable, using a
triangular PDF with a most likely value of 1000 kg, a minimum value of 500 kg, and a maximum value of
2000 kg (Figure 4.3-28). All of the other variables were held constant in this simulation (Table 4.3-5).
-------
Table 4.3-5. Constant input parameters used in probabilistic simulation applications.
Input Parameter
Value
Averaged Initial Concentration
Darcy Velocity
X-section Area
Porosity
Source Decay Rate
Time until Remediation
Period of Remediation
100 mg/L
20 m/yr
30m2
0.33
0 1/yr
31 yr
0.5yr
One thousand Monte Carlo deterministic runs were used to generate the time-dependent statistical results.
Unlike a conventional deterministic simulation that only gives a single prediction of system behavior, the
probabilistic simulation output shows the mean and ranges of possible system behavior (the time-dependent
source concentration and mass discharge here) responding to the stochastic input variables (F ar\dMso).
The overall source concentration (Figure 4.3-29) tends to drop gradually due to the natural dissolution,
faster in the first 40 years and slower after that. Note that with an initial value of 100 mg/L, the source
concentration after 100 years natural dissolution still remains at a mean value of about 1.2 mg/L.
Mass discharge leaving the source zone is linearly related to the source concentration so that it shows a
similar trend as the source concentration. The initial mass discharge is 60 kg/yr and it gradually decreases
to about a mean value of 0.8 kg/yr after 100 years of natural dissolution. This high source concentration and
mass discharge may still result in a large dissolved plume zone.
Concentration vs. Time
1.0e02
1.0e-01
20 30
40 50 60
Time (yr)
70 80 90 100
Upper Bound
25% Percentile
Mean
95% Percentile
5% Percentile
Median
75% Percentile
Lower Bound
Figure 4.3-29. Probability simulation output of source concentration (without source remediation).
-------
The second example is a probabilistic simulation using the DNAPL source function model with source
remediation (see Falta et al., 2005a). Source remediation was assumed to be conducted at the 3 lstyear after
DNAPL release and lasted for a half year. In addition to two stochastic variables (Mso and F) in the first
example, the fraction of removed source mass, X, was also considered as a stochastic variable, using a
triangular PDF with a most likely value of 80%, a minimum value of 60%, and a maximum value of 95%
(Figure 4.3-30). All of the other variables were held constant as in the first example, and 1000 Monte Carlo
deterministic runs were used to generate the time-dependent statistical results.
PDF of Mo
PDF of Gamma
0.0015
0.0010
0.0005
0.0000
2.0
1.5
1.0
0.5
0.0
500
1000 1500
Mo (kg)
2000
0.1
1
Gamma
10
PDF Of X
0.6 0.7 0.8 0.9
Fraction of mass removed
Figure 4.3-30. Three input probability distribution functions (with source remediation).
Again, the probabilistic simulation outputs show the mean and ranges of possible time-dependent source
concentration and mass discharge responding to three stochastic input variables (F ,MSQ andX). The
overall source concentration range shows a similar gradual decrease in the first 31 years due to the natural
dissolution (Figure 4.3-31). As the result of a half-year source remediation effort started at the 31st year, a
sharp drop of concentration occurred in a short period of time. Following a half-year remediation effort,
the mean value of the source concentration drops to 1.2 mg/L at the 53rd year and around 0.3 mg/L at the
100th year, respectively. The simulation result also indicates a decrease of uncertainty that is reflected by a
narrower band of source concentration range.
Mass discharge leaving the source zone shows a similar trend as the source concentration. The mass
discharge gradually drops in the first 31 years due to the natural dissolution process. A sharp drop of mass
discharge then occurs in a short period of time after the source remediation effort. Following the half year
remediation effort, the mean value of the mass discharge decreases to 0.8 kg/yr at the 53rd year and less than
0.2 kg/yr at the 100th year, respectively. Compared to the first example, the second example would result in
a smaller dissolved plume zone due to the source remediation effort. The probabilistic simulation outputs
for both examples show that the uncertainty in source concentration and discharge is increasing with
increasing time.
-------
Concentration vs. Time
1.0e02
1.0e01
c
o
0)
o
o
o
1.0eOO
1.0e-01
10 20 30 40 50 60
Time (yr)
70 80 90 100
Upper Bound
25% Percentile
Mean
95% Percentile
5% Percentile
Median
75% Percentile
Lower Bound
Figure 4.3-31. Probability simulation output of source concentration (with source remediation).
In summary, the GoldSim model is capable of performing the probabilistic simulation using the DNAPL
source function model. It took less than a minute of computer time on a personal computer to run 1000
Monte Carlo realizations using the DNAPL source function model linked to GoldSim through a DLL
approach.
Although only three stochastic input variables have been tested at this point, it is evident that this model
can account for more uncertain variables, such as groundwater velocity, retardation factor, and plume decay
rates. By using the full source and plume function model and considering all the uncertainties occurring
during process and performance of DNAPL remediation efforts, probabilistic simulation can better
predict the impact and benefit of source and plume remediation and form the basis for decisions regarding
remediation alternatives.
4.3.3.5 Simplified contaminant source depletion models as analogs of multiphase simulators
Four simplified DNAPL source depletion models recently introduced in the literature are evaluated for
the prediction of long-term effects of source depletion under natural gradient flow (Basu et al., 2008b).
These models are simple in form (a power function equation is an example) but are shown here to serve
as mathematical analogs to complex multiphase flow and transport simulators. The spill and subsequent
-------
dissolution of DNAPLs was simulated in domains having different hydrologic characteristics (variance of
the log conductivity field = 0.2, 1 and 3) using the multiphase flow and transport simulator UTCHEM. The
dissolution profiles were fitted using four analytical models: the equilibrium streamtube model (ESM), the
advection dispersion model (ADM), the power law model (PLM) and the Damkohler number model (DaM).
All four models, though very different in their conceptualization, include two basic parameters that describe
the mean DNAPL mass and the joint variability in the velocity and DNAPL distributions.
Three of the four source depletion models (ESM, ADM and PLM) can be expressed in a generalized form
where the flux-averaged concentration exiting the source zone, Cj [ M/L3 ], can be related to a source depletion
term, SD:
C (V )
-^-LL = l-SD(Vp) (4.3-57)
where Cs is the solubility of the DNAPL in groundwater [M/L3], V is the number of pore volumes of natural
flowing groundwater that have traversed the source zone, and the contaminated fraction, fcfl , is the fraction
of the streamlines initially containing DNAPL. The three models differ only in the formulation of SD. The
DaM has a different form but simplifies to the PLM under specific conditions, as described in detail below.
The equilibrium streamtube model (ESM) is based on a Lagrangian approach where the DNAPL source
zone is conceptualized as a collection of non-interacting streamtubes, with hydrodynamic and DNAPL
heterogeneity represented by the variation of the travel time and DNAPL saturation among the streamtubes.
Using the concept of reactive travel time, SD(V) is equivalent to the cumulative distribution function of T
(Jawitz et al., 2005):
(4.3-58)
where |o,lni and alni are the mean and standard deviation of the log transformed variable In i .
Equation (4.3-58) is appealing in that it is a compact analytical expression that can be employed in a
spreadsheet type application to efficiently explore key system parameters that are frequently identified
as playing a central role in the natural gradient depletion of a DNAPL source zone, including: the total
DNAPL mass and the mean travel time (directly related to the first moment of S and t), the variability of
the Lagrangian DNAPL architecture (related to the second moment of S) and the variability of the velocity
field (second moment of t). Further details of the ESM approach are presented in Jawitz et al. (2005).
Here, we will use |o,lni and ahi as fitting parameters to simulate the source depletion profile. Physically,
|a,hT quantifies the mean of the combined hydrodynamic field and DNAPL architecture, while alni quantifies
the variability of the two distributions (Jawitz et al., 2005).
Jury and Roth (1990), among others, have demonstrated the mathematical similarity of the advection
dispersion equation and the lognormal travel time distribution. An alternative to representing SD(V ) as
the cumulative distribution function of the lognormal variable T would then be the dimensionless analytical
solution for the advection dispersion equation (continuous injection; injection and detection in flux) given
by (Kreft and Zuber, 1978):
-exp(P)erfc
(4.3-59)
where P is the Peclet number and R is the retardation coefficient. The Peclet number is used here in a
manner analogous to alni in the ESM, functioning as a variability index to describe the combined effects
of hydrodynamics and DNAPL architecture. The retardation factor, R , is used to describe the total mass
of DNAPL in the source zone in a manner similar to |o,ln T . Note that R represents the DNAPL mass in the
contaminated zone and can be converted to a total-domain retardation using fc, as described in Jawitz et al.,
2003.
-------
Power law source depletion models have recently been presented by Zhu and Sykes (2004) and Falta et al.
(2005a). The concept behind the PLM is to relate Cf to changes in DNAPL mass though an empirical
variability index. Such models can be generalized as:
A/f (17 \\r
(4.3-60)
M
S,o
where M(V ) is the DNAPL mass in the source zone after V pore volumes of flow, Msfl is the initial
DNAPL mass in the source zone, and PP is an empirical index that is a function of the variability of both
the DNAPL distribution and the permeability field. Combining equations (4.3-57) and (4.3-60), Cj can be
expressed as:
' Ms(Vp)
(4.3-61)
The parameter PP is similar to the variability parameter ah T in the ESM and increases with increasing
domain heterogeneity (Falta et al., 2005a).
Parker and Park (2004) presented a simplified model for estimating DNAPL source zone depletion based on
the concept of an effective Damkohler number Da:
Cj(Vp) ' -- "- (4.3-62)
p' _ i
c.
Here, Da = KeffLs I q where K „. [T1] is the field-scale effective mass transfer coefficient, Ls is the source
zone length in the mean flow direction, and q [LT1] is the average Darcy flux for the source zone. The form
of this equation is slightly different from the previous three models and thus Cj (V ) cannot be expressed in
terms of SD(V). The effective mass transfer coefficient (K „), is related to changes in the DNAPL mass and
the average groundwater velocity (Parker and Park, 2004):
~\/f(~\7 \\ 2
(4.3-63)
where KS [LT1] is the average saturated hydraulic conductivity of the source zone, and KO [T1] p:, and P2
are empirical fitting parameters. The parameter P: describes rate limited dissolution that is neglected in the
other models. For all values of P: except 1 the flux-averaged concentration exiting the source zone is a
function of the Darcy flux, indicating that changes in contact time between the advecting groundwater and
the DNAPL phase result in rate limiting effects. The "mass depletion exponent" P2 is related to the
groundwater velocity distribution, the DNAPL geometry, and the correlation between the two (Parker and
Park, 2004).
The ESM, ADM, PLM, and DaM were evaluated for their ability to serve as mathematical analogs to
simulations conducted using the multiphase flow and transport simulator UTCHEM (Delshad et al., 1996)
over a range of hydraulic conductivity field variances ( &\n£ ). The 2-D model domain was 30m long
(r-direction) and 10m deep (z-direction) with 60 cells along the x-direction and 125 cells along the
z-direction. The parameters used in the UTCHEM simulations were selected to represent typical values,
consistent with recent numerical simulations of multiphase flow and transport. After the DNAPL spill, the
dissolution process was initiated through a pair of fully-screened injection and extraction wells at the left
and right boundaries. The dissolution kinetics of the DNAPL were evaluated under both equilibrium and
rate-limiting conditions. The effect of non-equilibrium dissolution was implemented using a mass transfer
coefficient described by the power law model of Miller et al. (1990).
DNAPL distributions are presented in Figure 4.3-32 for spills in two of the four realizations from each of
the three simulation sets. Low variability in the permeability field (&\n£ = 0.2) led to relatively uniform
-------
DNAPL architecture with the potential of greater pooling at the underlying capillary barrier (no flow
boundary in the numerical simulations). As the variability of the permeability field increased, the total
penetration depth decreased, leading to a DNAPL architecture that was more dominated by 'pools'.
Another interesting feature is the difference in DNAPL architectures in different realizations of random
fields with the same geostatistical parameters. These differences are more pronounced for domains with
greater heterogeneity (Figure 4.3-32). DNAPL spill simulations are very sensitive to local permeability
contrasts and thus, very different spill scenarios may sometimes be created in domains having the same
<•
The UTCHEM dissolution profiles were fitted using Equations (4.3-57) and (4.3-58) for the ESM, (4.3-57)
and (4.3-59) for the ADM, (4.3-61) for the PLM, and (4.3-62) and (4.3-63) for the DaM. The fraction
contaminated fc in the ESM, ADM and PLM is the ratio between the initial flux-averaged concentration
Cj (V ) and the aqueous solubility of the DNAPL Cs. The rate-limiting parameter P: in the DaM was
assumed to be equal to 1 since local-scale mass transfer was not found to significantly affect the dissolution
profiles. Fitting parameter values were obtained by fitting the UTCHEM simulation results to the four
source depletion models using the normalized root mean square deviation (NRMSD, Jawitz et al. 2003) as
the objective function for regression.
All four models (except the power function model for simulation set 1) provided reasonable fits (based on
visual inspection and low NRMSD values) of the UTCHEM-generated dissolution data. For illustration,
UTCHEM-generated dissolution profiles are compared to ESM, PLM and DaM fits in Figure 4.3-33. Since
ADM fits were very similar to the ESM fits, they have not been shown in the figure. The good fits suggest
that these simplified source depletion models can describe the dissolution behavior from DNAPL spills
generated by sophisticated numerical simulators such as UTCHEM.
1-1
Set 1-3
ff^ -0.1*
SetJ-l
•
S« 3-J
CT:. -3
£ g
o o
Figure 4.3-32. DNAPL spills for six representative cases. Note that x < 7 m and x > 23 m are cropped from
the figures to focus on the DNAPL (Basu et al., 2008b).
-------
1 -I
0.8 -
0.6 -
0.4 -
0.2 -
Set3-1:UTCHEM
Set 3-1: ESM fit; NRMSD = 0.013
Set 3-1: PLM fit; NRMSD = 0.023
Set 3-1: DaM fit; NRMSD = 0.013
Set 3-3: UTCHEM
Set 3-3: ESM fit; NRMSD = 0.007
Set 3-3: PLM fit; NRMSD = 0.02
Set 3-3: DaM fit; NRMSD = 0.017
20
Pore Volumes
Figure 4.3-33. UTCHEM-generated dissolution profiles (symbols) and corresponding model fits for
simulation realizations 3-1 and 3-3 (
-------
• set 1-3, variance (In k) - 0.2
• set 2-3, variance (In k) = 1
+ set 3-1, variance (In k) = 3
x set 3-3, variance (In k) = 3
0
0.2
0.4 0.6
Mass Reduction, /?„
0.8
Figure 4.3-34. UTCHEM-generated mass reduction-flux reduction profiles (symbols) compared with DaM
(Set 1-3), ESM (Set 2-3), ESM (Set 3-l),and PLM (Set 3-3) fits (lines) to the dissolution
profile (Basuetal, 2008b).
One of the objectives of this study was to compare the four source depletion models in terms of their
ability to describe UTCHEM-generated dissolution profiles. Three of the four models (ESM, ADM and
DaM) resulted in similar fits. The single-parameter PLM failed to capture the dissolution behavior as well
as the other three models, especially for the low -heterogeneity cases. In addition to the fitting parameters,
the PLM and DaM also require information on the breakthrough curve that is integrated to calculate
Ms (V)
. This information is not required in the ESM and ADM.
The physical interpretations of the fitted parameters can be broadly summarized as parameters that describe
the characteristics of the travel time and the DNAPL saturation distribution. In the ESM, the parameter a\^
describes the variability in the travel time and the DNAPL saturation distribution, while the parameter
HlnT describes the mean values. In the ADM, P describes the variability in the travel time and the DNAPL
saturation distributions, while R describes the DNAPL mass in the contaminated zone. In PLM and DaM,
the parameters PP and P2 describe the variability of the DNAPL saturation distribution. The parameter KO in
the DaM is analogous tofc in the other three models. The DaM has a third parameter P: that accounts for
rate limited dissolution that is neglected in the other models presented in this paper. Note that a
modification of the ESM was proposed by Jawitz et al. (2005) that accounted for rate-limited effects.
The challenge now lies in estimation of the model parameters at the field scale. Parker and Park (2004)
proposed that the models can be calibrated to measured source zone mass flux over time. This requires
detailed flux characterization over the entire source depletion history, which is impossible at most field sites.
Models calibrated for shorter time spans will not be able to capture the entire dissolution behavior. In any
case, a priori estimation of the model parameters is preferred to requiring dependence on the measured
dissolution profile.
Christ et al. (2006) proposed the use of the ganglia-to-pool (GTP) ratio for calibrating the DaM by
establishing an empirical relationship between GTP and the mass transfer coefficient. However, techniques
for determining the initial GTP ratio at the field scale have not yet been established. Jawitz et al. (2003,
2005) proposed the use of non-reactive and reactive tracers to describe aquifer heterogeneity and DNAPL
architecture in terms of ESM parameters that could be used to forecast remedial performance. These
authors presented theoretical arguments and limited field evidence in support of this approach, and Pure
et al. (2006) presented laboratory data that also supported the ESM. However comprehensive validation of
the ability of tracer-determined ESM parameters to predict remedial performance remains for future work.
-------
4.3.3.6 Temporal evolution ofDNAPL source and contaminant flux architecture
The objective of this study was to investigate the impact ofDNAPL source mass depletion on the alteration
of the contaminant flux [J(y,z,t)] distribution at the control plane (CP ) (Basu et al., 2008a). Contaminant
mass discharge (MD ; local flux values integrated over the source CP) and flux distribution act as the
key linkages for understanding the relationship between source mass depletion and plume dynamics
(EPA, 2003). Simulation results from two separate numerical studies and CP flux monitoring data from
a source depletion field study at Hill Air Force Base, Utah (Brooks et al., 2008) were used for the present
analysis. Model outputs were examined to determine the temporal evolution of the population statistics
(mean, variance, coefficient of variation) and the spatial structure (centroid, first and second moments
and variograms) of source mass and contaminant fluxes. The associated changes in trajectory-integrated
DNAPL content (S) (Jawitz et al., 2005) were also investigated to link source depletion to flux changes.
Groundwater flow and contaminant transport were simulated in three-dimensional, heterogeneous, random,
spatially correlated permeability z fields with emplaced DNAPL sources. Two different sets of simulations
were analyzed for this study: Set 1 was done using the numerical code ISCO3D (Zhang and Schwartz,
2000), while Set 2 was done using the code T2VOC (Falta et al., 1995). These two sets of simulations were
also different in terms of flow-domain configurations, mass-transfer assumptions, and boundary conditions.
Attributes of the simulated DNAPL source zones are presented in Table 4.3-6. The objective behind
presenting such different scenarios was to demonstrate that the resulting observations were not a function of
the assumptions of the domain characteristics or the flow and transport model used.
Table 4.3-6. Attributes of the simulated DNAPL source zones used in the numerical simulations.
Set 1: • Mean intrinsic permeability Case
Domain 1 u.e = 9.7 x io~13 m2 1
• Domain dimensions 3 m x 0.49 m x 0.98 m
(nx = 43, ny = 7; nz = 14)
• Correlation lengths ^ = 0.7 m, ^ = 0.07 m,
^ = 0.07 m
• Darcy flux q = 5.4 cm/day Case
2
Case
3
>*N and e positively correlated
>N) = ln(8) + 26.2, (e in m2; SNmax = 0.15)
• DNAPL massM= 2.36 kg
•<*L=2.45
• SN and e positively correlated
ln(5*N) = ln(e) + 23.6, (e in m2; SNmax = 0.15;
• DNAPL mass M= 2.36 kg
• Binary distribution of 5*N (SN = 0.15 and 0.5)
• DNAPL mass M= 12.3 kg
Set 2: • Mean intrinsic permeability Case
Domain 2 u.6 = 2 * 10~12 m2 1
• Domain dimensions 30 m x 30 m x 15 m
(nx = ny = 20, nz = 25)
• Correlation lengths ^ = 30 m, E, = 10 m,
• Darcy flux q = 0.6 cm/day Case
• Sv and e positively correlated
ln(5-N ) = ln(e) + 23.2, (e in m2; SNmax = 0.17)
• DNAPL mass M= 1950 kg
SN and e negatively correlated
= - ln(8) - 3 1.7, (e in m2 ; SNmax = 0.09)
DNAPLmassM=826kg
-------
Dissolution of DNAPL source mass during groundwater flow was simulated using ISCO3D (Domain 1) and
T2VOC (Domain 2) codes. Simulated, time-varying DNAPL content and contaminant flux distributions
at the outflow control plane during mass depletion were analyzed to study the effects of source depletion
on contaminant flux (J) distribution. In addition, passive flux meter (PFM) data were used to evaluate the
impact of source mass depletion on the evolution ofCP flux distribution at operable unit 2 (OU2), Hill Air
Force Base, Utah. These PFM data had horizontal and vertical resolutions of 3 m and ~25 cm, respectively.
The initial values for the mean, standard deviation and the coefficient of variation (CV) of the contaminant
flux distribution (J), and the DNAPL distribution (S^ ) within the domain for the simulated cases are
presented in Table 4.3-7. The higher value of(CVs ) relative to CVj is attributed to the sparse distribution
of the DNAPL within the source zone, with certain regions having no DNAPL and other regions at residual
saturation (see Feenstra and Cherry, 1996; Meinardus et al., 2002). An evaluation of population statistics
for J and S^ distributions as a function of increasing source mass removal showed that the CVof the S^
distribution increased with time, while that of the J distribution oscillated but remained relatively constant.
The relatively constant CVj at the source CP is an interesting feature and was thus investigated for all five
cases (Table 4.3-6). Since the initial CVj for the domains are different, we normalized the data by plotting
the relative reductions in the mean contaminant flux (R^ ) and the flux standard deviation (Ra):
= ^-^_
A
(4.3-65)
Where jo,, and a. are the initial mean and standard deviation of the distribution, while \it and a^ are the
mean and standard deviation of the distribution at time t. The results from simulations for all cases are
plotted in Figure 4.3-35. All data sets indicate a 1: 1 relationship between R and Ra , indicating a constant
coefficient of variation. For the cases we examined, the relationship thus appears to be independent of the
hydrodynamic heterogeneity as well as the correlation structure between the DNAPL and the permeability
field.
Consistent with the numerical simulations, the TCE flux data from the Hill AFB field study (Brooks et al.,
2008), also suggest that source mass depletion led to a proportional reduction in the mean and standard
deviation of the J distribution, and thus, a temporally constant CVj (Figure 4.3-35). However, it should be
noted that to claim temporal stability based only on measurements at three points in time is difficult. Field
studies are needed to characterize the complete evolution with time of the contaminant flux distribution at
DNAPL source CPs.
Table 4.3-7. Coefficient of variation (CV) of the Darcy flux (q ), contaminant flux (J), and the DNAPL
mass (S^) distribution.
Distribution properties Domain I (ISCO3D) Domain II (T2VOC)
Case 1 Case 2 Case 3 Case 1 Case 2
Initial CV
i
Initial CVj
Initial CV~
0.6
0.7
7.0
1.7
1.7
6.4
0.6
0.7
6.4
1.7
4.6
8.6
1.7
4.8
11.8
-------
J -I
E
J«H
5
E 0.4
•f.
a
- l>4
J|
v, Dsuula I
• D«nial «.jii» ;
IK.iiLiiu : i*. .1 .<• L'
DulliHn 2 l(.nt I)
I Hill \m. IUl,
1.1 0.4 Oft
R»diKifouiB HID Mriin
Figure 4.3-35. Relationship between the relative reductions in the mean and standard deviations of the
contaminant flux (Basu et al., 2008a).
The evolution of the source distribution and the contaminant flux distribution at source CP were also
investigated by examining the changes in spatial moments of SN and J. The first normalized moments
define location of the centroid, while the second centralized moments indicate the spread of the distribution
about the centroid. Physically, these two parameters give an indication of the mean location of the 'hotspot'
and its spatial extent. For the trajectory of the centroid, we focused on the T2VOC simulations instead
of ISCO3D because (1) of the larger spatial dimensions of the flow domain, and (2) alternative DNAPL
saturation distributions available based on positive or negative permeability (e ) vs. SN correlation.
The loci of the centroid of the contaminant flux distributions are shown alongside the projection of the loci
of the DNAPL distributions at the source CP (Figure 4.3-36). The Darcy flux (q ) distribution at the same
CP is superimposed on this plot to enable a comparison between centroids ofJ and up-gradient DNAPL
mass. As DNAPL mass is depleted, the centroid of the DNAPL distribution shifts from the high e to the
low-s zones because of the extended timelines required for DNAPL mass removal from those regions
(Figure 4.3-36). For the positively correlated case, there was a larger amount of DNAPL mass at the
bottom of the domain and thus, the centroid moves in that direction, while for the negatively correlated
case, there was more DNAPL mass at the top causing the centroid to shift upwards.
Flux (positive correlation)
Flux (negative correlation)
Sn (Positive Correlation)
Sn (Negative Correlation)
•—r^o
darcy velocity (m/s)
2.00E-06
9.28E-07
4.31 E-07
2.00E-07
9.28E-08
4.31 E-08
2.00E-08
12
13
14
15
16
Figure 4.3-36. Locus of the Centroid of the DNAPL and contaminant flux distributions (at the source
control plane) in Domain 2 (white circles indicate the initial location of the centroid and
black stars denote their final location). Note that: (I) only the section of the domain
below the water table is shown in the figure; (2) the initial location of the centroid of the
contaminant flux distribution is the same in the positive and negative correlation cases
(Basu et al., 2008a).
-------
The second centralized spatial moments about the center of mass (a^., a2 , <52zz) are a measure of the spread
of the SN distribution along the x, y and z axes, respectively. Three example cases, Case 1 in Domain 1
(Figure 4.3-37a) and Cases 2 and 3 in Domain 2 (Figure 4.3-37b), are presented to show the temporal
evolution of the spreads of the DNAPL and the contaminant flux distributions. The other simulations
exhibited similar behavior (data not shown). These results support the previous hypotheses regarding
the approximately time-invariant nature of the contaminant flux distribution. The spreads of the flux
distribution were less variable in time than those of the DNAPL distribution, indicating again the greater
stability of the J distribution (Figure 4.3-37b).
(a)
0.5 -i
E
Q 0.3 •
40 60
Mass Depletion (%)
s|gma_xx mosm\« correlation)
sigma_yy (positi\« correlation)
sigma_zz (positive correlation)
sigma_xx (negati\« correlation)
sigma_yy megati\« correlation)
sigma_zz (negati\« correlation)
40 60
Mass Depletion (% )
80
100
o
£
1
Q
X
4 •
£ 2
'o
20
-sigma_yy (positive correlation)
-sigma_zz (positive correlation)
-sigma_yy(negative correlation)
-sigma_zz (negative correlation)
40 60 80
Mass Depletion (%)
100
Figure 4.3-37. Spatial variance of the DNAPL and contaminant flux distributions in (a) Domain 1, Case 1
(ISCO3D), and (b) Domain 2, Cases 1 and 2 (T2VOC). Note that the spreads of the flux
distributions have been plotted after flushing about one pore volume through the domain
such that the fluxes at the exit control plane are reflective of the source conditions (Basu et
al., 2008a).
This study shows that both the mean (\a ) and the standard deviation (o ) of the contaminant flux distribution
at the source CP decreases with source mass reduction. Moreover, \a and o decrease proportionally,
resulting in a coefficient of variation that appears to be essentially constant in time. The spatial moments
of the contaminant flux distribution appear to be also essentially constant in time, indicating that - for
DNAPL source zones cleaned up through in-situ flushing - the contaminant flux distribution remains stable
throughout the DNAPL mass depletion process while the flux magnitude gradually fades away with time.
It should be noted, however, that these conclusions are preliminary. Rigorous investigations that
account for spill scenarios, different remedial alternatives and flow configurations are required prior to
generalization and subsequent application of these results. More importantly, field studies that characterize
the temporal evolution of the source and flux distribution need to be conducted.
-------
4.3.3.7 Predicting DNAPL dissolution using a simplified source depletion model parameterized
with partitioning tracers
Simulations of non-partitioning and partitioning tracers were used to parameterize the equilibrium
streamtube model (ESM) that predicts the dissolution dynamics of DNAPLs as a function of the
Lagrangian properties of DNAPL source zones (Basu et al., 2008c). Lagrangian, or streamtube-based,
approaches characterize source zones with as few as two trajectory-integrated parameters, in contrast to
the potentially thousands of parameters required to describe the point-by-point variability in permeability
and DNAPL in traditional Eulerian modeling approaches. The spill and subsequent dissolution of
DNAPLs was simulated in two-dimensional domains having different hydrologic characteristics using the
multiphase flow and transport simulator UTCHEM. Simulations were conducted with log conductivity
variances (7lnE = {0.2, 1, 3} to evaluate a range of field scenarios. A set of simulations comprised of
four realizations was run for each value of crtae. Non-partitioning and partitioning tracers were used to
characterize the Lagrangian properties (non-reactive travel time (t) and trajectory-integrated DNAPL
content (S) statistics) of DNAPL source zones, which were in turn shown to be sufficient for accurate
prediction of source dissolution behavior using the ESM throughout the relatively broad range of hydraulic
conductivity variances tested here.
Tracer BTCs were found to be closely matched by lognormal distributions generated using |o,ln t and aln t
determined from the tracer moments (Figure 4.3-38.), suggesting that the assumption of log-normality for
the non-reactive and reactive travel time distributions is reasonable.
o
o
• non partitioning tracer
A partitioning tracer
01234
Pore Vblumes
Figure 4.3-38. Non-partitioning and partitioning tracer data for simulation set 2-4. Symbols represent
UTCHEM output while solid lines represent lognormal fits (Basu et al., 2008c).
The reactive travel times (distribution parameters, determined from t and S distributions (T = NtS, where
TV is the ratio of the DNAPL density to aqueous solubility)), were used to predict the DNAPL dissolution
profile using the ESM. The UTCHEM-generated dissolution profile is shown along with the tracer-based
ESM predicted profile for an example realization from each of the three simulation sets in Figure 4.3-39.
In these simulations, t = 0 is defined as the time at which the contaminant peak arrives at the exit control
plane (x = 30 m). The ESM accurately predicted the dissolution dynamics at the source control plane with
average normalized root mean square deviation (NRMSD) values of 0.022, 0.025, and 0.027 for (7fne = 0.2,
1, and 3, respectively.
-------
0.8 -
0.6 -
y
o
• Sim 1-4 (0^=0.2, oln,= 0.55)
A Sim2-4(oL = 1,olnT=1.03)
Sim3-2(ofn/t = 3, oln,= 1.16)
0.4 -
0.2 -
0 -
Pore Volumes
Figure 4.3-39. ESM-predicted and UTCHEM-generated dissolution profiles for simulation sets 1-4, 2-4,
and 3-2. Symbols represent UTCHEM output while solid lines represent predictions (Basu
et al, 2008c).
The predicted dissolution profiles were integrated to examine the relationship between mass reduction RM
and flux reduction Rj (Figure 4.3-40). Flux reduction is defined as 1 - Q (t)/Cj (t =0 ) and mass reduction
is similarly defined as 1 - Ms (t) IMS (t =0 ) where Q (t) is the flux-averaged concentration at the exit
control plane at time t, and Ms (t) is DNAPL mass in the source zone at time t. The predicted mass
reduction-flux reduction profiles also matched the UTCHEM-simulated behavior well, with NRMSD values
varying between 0.034 (a^E = 0.2) and 0.054 (o^e = 3).
1 -i
0.8 -
-2 0.6 -
o
0.4 -
0.2 -
Sim 1-4 (0 = 0.2,01, =0.55)
Sim 2-4 (ofnk=1,0|nt= 1-03)
Sim3-2(0?n/t=3,0|n,
0.2 0.4 0.6
Mass Reduction, Rm
I
0.8
Figure 4.3-40. ESM-predicted and UTCHEM-generated mass reduction-flux reduction profiles for
simulation sets 1-4, 2-4, and 3-2. Symbols represent UTCHEM output while solid lines
represent predictions (Basu et al., 2008c).
Following Christ et al. (2006), the ganglia-to-pool ratio (GTP) was defined as the ratio of the DNAPL mass
in cells with DNAPL saturation less than 15%, to the mass in cells with saturation greater than 15%. Based
on the GTP ratio, the following up-scaled mass transfer model was proposed (Christ et al., 2006) for the
prediction of flux-weighted concentration CAT) from DNAPL source zones:
-------
c
(4.3-66)
where PG= 1.5 GTP"0-26 and Cs is the aqueous solubility. The GTP correlation coefficient PG was developed
from simulations with GTP limited between 0.5 and 24. This model was used to predict the dissolution
BTCs and the Rj (RM) relationships for the simulations presented here; two example cases are presented
in Figure 4.3-41. The model predicts a temporally constant aqueous flux for GTP equal to and was
thus incapable of representing the dissolution dynamics for simulations 1-2 and 1-3 (Table 4.3-8). Within
simulations where the GTP was within the prescribed range (simulation sets 2 and 3), the model proved to
be a good predictor of dissolution dynamics in some cases (e.g., simulation 3-2, Figure 4.3-41); however
for other cases (e.g., simulation 2-4, Figure 4.3-41) it was a relatively poor predictor of dissolution.
Table 4.3-8. Tracer-derived ESM parameters.
Simul-
ation
1-1
1-2
1-3
1-4
2-1
2-2
2-3
2-4
3-1
3-2
3-3
3-4
Domain properties
Mm*
-6.85
-6.92
-6.92
-6.94
-5.62
-5.62
-5.62
-5.62
-5.62
-5.62
-5.62
-5.62
°L
0.19
0.19
0.18
0.24
1
1
1
1
o
J
3
o
6
o
6
SN
0.00419
0.0033
0.0033
0.00446
0.00393
0.00446
0.00357
0.00446
0.00447
0.00445
0.00446
0.00449
Non-reactive
tracer
parameters
Uta,
-0.01
-0.01
-0.01
-0.01
-0.06
-0.06
-0.08
-0.05
-0.15
-0.19
-0.20
-0.17
°ln;
0.17
0.15
0.17
0.15
0.33
0.34
0.40
0.31
0.54
0.62
0.63
0.58
Partitioning
tracer
parameters
hn.
-5.58
-5.90
-5.91
-5.55
-5.95
-5.61
-5.74
-5.82
-5.31
-4.82
-5.77
-5.40
GlnS
0.47
0.61
0.65
0.55
1.02
0.87
0.77
0.98
0.36
0.98
1.44
0.95
Reactive travel
time
distribution
moments
hnx
1.39
1.09
1.06
1.44
0.98
1.32
1.17
1.12
1.54
1.97
1.02
1.43
°ln.
0.50
0.63
0.67
0.55
1.07
0.93
0.87
1.03
0.65
1.16
1.57
1.11
Ganglia
to pool
ratio
oo
oo
oo
oo
31.8
29.6
35.5
27.6
4.1
2.3
1.1
2.7
S,r = average NAPL saturation
-------
u
1 -
0.1
UTCHEM
simulation
streamtube model
-
GTP model
(GTP = 10)
I I I
0 0.2 0.4 0.6
V.f. U.*t U.O
Mass Reduction, R,
0.8 1
I
I
4-1
2
v
u
§
u
10000 -|
1000 -
i
100 i
10 i
1
0.1
Sim 3 - 2
UTCHEM
D simulation
streamtube model
GTP model
(GTP = 1.5)
I I I
0 0.2 0.4 0.6
v.e. u.t u.o <
Mass Reduction, R,
0.8 1
\l.f. U.t U.O
Mass Reduction, R,
0.8 1
0.4 0.6 0.8 1
U.£ \l.t U.O
Mass Reduction, Rm
Figure 4.3-41. Comparison between the ESM and the GTP models for prediction of source dissolution
behavior (Basu et al., 2008c).
Since the GTP ratio accounts for the population distribution of the NAPL but not its spatial structure, it is
an incomplete descriptor of the DNAPL architecture. The Lagrangian description of the t and S distributions
characterizes both the NAPL population distribution and spatial structure. No definite correlation was
observed between GTP and ^^ for the simulations presented here (Table 4.3-8). Finally, it is noted that it
has not been determined how the GTP ratio would be measured at field sites (Christ et al., 2006).
In summary, the results were found to be relatively insensitive to travel time variability, suggesting that
dissolution could be accurately predicted even if the travel time variance was only coarsely estimated.
Estimation of the ESM parameters was also demonstrated using an approximate technique based on
Eulerian data in the absence of tracer data; however determining the minimum amount of such data required
remains for future work. Finally, the streamtube model was shown to be a more unique predictor of
dissolution behavior than approaches based on the ganglia-to-pool model for source zone characterization.
-------
5.0
Conclusions
The principal goal of this project was the development of tools by which to evaluate the impacts of partial
DNAPL source removal. The research was based on the examination of contaminant flux responses
to source zone treatment at DNAPL-contaminated sites. These field studies were supplemented with
in-depth laboratory studies of the factors that control the relationship between DNAPL mass depletion
and contaminant mass flux. Results from field and laboratory investigations were used to evaluate the
performance of existing source remediation models and to develop new source and coupled source-plume
simulators.
Results of the field, laboratory, and modeling investigations show that although response to DNAPL
source treatment is dependent on a number of site specific factors (e.g., site hydrogeology, contaminant
composition and architecture, remedial technology), significant contaminant flux reductions can be
achieved at many sites by partial DNAPL removal. Modeling results suggest that as heterogeneity in
aquifer properties and NAPL spatial distribution increases, less mass reduction is required to achieve
a given flux reduction, although the overall source longevity may increase. In addition, early source
treatment enhances risk reduction by decreasing contaminant discharge to the plume.
Contaminant mass flux and mass discharge data were collected before and after source treatment at four
field sites. These data provide the first comprehensive field assessment of mass flux/discharge response
to DNAPL source treatment. Three flux measurement techniques (transect method, passive flux meters
and modified integral pump tests) were used. The MIPT employed in this study was based on published
techniques and included concurrent pumping across a transect in conjunction with hydraulic head
monitoring to provide a direct estimate of groundwater flux.
Mass discharge reductions of approximately 90% and 99% were observed at the Hill Air Force Base and
Fort Lewis sites, respectively. Passive flux meter and MIPT test data exhibited comparable results for these
sites. Less dramatic mass discharge reductions were seen at the Borden and Sages sites and significant
differences were observed between measurement methods. Because the support volumes for these
measurement techniques can be very different, heterogeneities in contaminant distribution and hydraulic
properties could be responsible for these differences.
Relationships between contaminant architecture and mass flux were evaluated in the laboratory using
2-dimensional flow chambers. Contaminants were injected into the flow chambers to create DNAPL source
zones and subsequent DNAPL mass depletion was accomplished by dissolution by aqueous, surfactant or
cosolvent solutions, or by gas-phase sparging. Source strength function (relationship between DNAPL
mass and contaminant mass flux) was observed to be primarily controlled by the DNAPL architecture
which can be characterized by the trajectory integrated NAPL content distribution. Reactive and non-
reactive tracer tests were used to estimate Lagrangian parameters (travel time and trajectory-integrated
NAPL content statistics) required for predicting source dissolution behavior. These estimated parameters
were used in the equilibrium streamtube model to accurately predict source dissolution.
The benefits of partial source zone remediation were evaluated from fundamental physical principles
using numerical, Lagrangian and analytical approaches. Numerical models are based on the most detailed
description of process behavior at the pore scale, but they are computationally intensive and require
information on spatially variable parameters that are difficult to characterize with reasonable accuracy using
available technologies. Though it may be difficult to calibrate these models to specific field sites, they
nonetheless help improve our understanding of behavior of DNAPL sources and plumes from a process
standpoint. Numerical models were used in this study to simulate laboratory and field source treatment
experiments and demonstrations. Both laboratory and pilot-scale field tests of air sparging for DNAPL
removal were simulated using TMVOC, a multi-component compositional multiphase flow simulator;
-------
and a field pilot study of cosolvent flushing was simulated using UTCHEM, also a multi-component
compositional multiphase flow simulator.
The Lagrangian models offer the only practical techniques currently available for predicting site specific
mass removal/mass flux relationships. These models describe the source zone as a collection of non-
interacting streamtubes with hydrodynamic and DNAPL heterogeneity represented by variations in travel
time and trajectory-integrated DNAPL saturations within the streamtubes. Lagrangian models were used
to estimate contaminant elution curves from laboratory and field pilot-scale in-situ flushing tests. Results
suggest that this modeling approach can predict dissolution behavior using reactive and non-reactive travel
time distributions to estimate hydrodynamic variability and NAPL distribution heterogeneity. However,
as currently deployed, tracer tests that are used to characterize these travel time distributions are very
expensive and may not be affordable at most field sites.
Because numerical and Lagrangian models are difficult to parameterize, analytical models of source
behavior can be important decision making tools for risk assessment and remedial design. In these models,
the source is represented by an effective homogeneous domain with the flux-averaged concentration at the
source control plane described as a function of the mean advective flow and the mean DNAPL mass in the
domain.
Predicting the effect of the source remediation on plume behavior has been limited by the lack of tools that
explicitly link source and plume remediation. To address this problem, coupled source and plume models
are presented. One of these models, REMChlor, was developed to evaluate the impact of source and plume
remediation at a more generic and strategic level. This screening-level mass balance approach is not specific
to any remediation technology. The contaminant source model is based on a power function relationship
between source mass and source discharge, and it can consider partial source remediation at any time after
the initial contaminant release. The source model serves as a time-dependent mass flux boundary condition
to a new analytical plume model. The plume model simulates first-order sequential decay and production of
several species, and the decay rates and parent/daughter yield coefficients are variable functions of time and
distance.
-------
6.0
References
Alvarez, P.J.J., and W.A. Illman (2006). Bioremediation and Natural Attenuation. Wiley-Interscience.
Annable, M.D., K. Hatfield, J. Cho, H. Klammler, B.L. Parker, J.A. Cherry, and P.S.C Rao (2005).
Field-scale evaluation of the passive flux meter for simultaneous measurement of groundwater and
contaminant fluxes. Environmental Science and Technology 39(18):7194-7201.
API (2003). Groundwater Remediation Strategies Tool. Washington, DC, 80 pp.
Applied Research Associates (1996). Groundwater Remediation Field Laboratory. Hydrogeology
Characterization and Site Development. Dover AFB, Delaware,. 1:192.
Aziz, C.E., C.J. Newell, and J.R Gonzales (2002). BIOCHLOR Natural Attenuation Decision Support
System Version 2.2, User's Manual Addendum, from http://www.epa.gov/ada/download/models/
biochlor22.pdf
Aziz, C.E., C.J. Newell, J.R. Gonzales, P. Haas, T. Prabhakar, and Y. Sun (2000). BIOCHLOR Natural
Attenuation Decision Support System. User's Manual, Version 1.0, National Risk Management
Research Laboratory, U.S. EPA, Ada, OK.
Bass, D.H., N.A. Hastings, and R.A. Brown (2000). Performance of air sparging systems: a review of case
studies. Journal of Hazardous Materials 72(2-3): 101-119.
Basu, N.B., P.S.C. Rao, I. C. Poyer, M.D. Annable, and K. Hatfield (2006). Flux-based assessment at a
manufacturing site contaminated with trichloroethylene. Journal of Contaminant Hydrology 86(1-
2): 105-127.
Basu, N.B., P.S.C. Rao, R.W. Falta, M.D. Annable, J.W. Jawitz, and K. Hatfield (2008a). Temporal
evolution of DNAPL source and contaminant flux distribution: impacts of source mass depletion.
Journal of Contaminant Hydrology 95 (3-4): 93-109.
Basu, N.B., A.D. Pure, and J.W. Jawitz (2008b). Simplified contaminant source depletion models as
analogs of multiphase simulators. Journal of Contaminant Hydrology 97(3-4):87-99, doi: 10.1016/j.
jconhyd.2008.01.001.
Basu, N.B., A.D. Pure, and J.W. Jawitz (2008c). Predicting DNAPL dissolution using a simplified source
depletion model parameterized with partitioning tracers. Water Resources Research 44(7):W07414.1-
13, doi:10.1029/2007WR006008.
Bateman, H. (1910). The solution of a system of differential equations occurring in the theory of radio-
active transformations. Proceedings of the Cambridge Philosophical Society 15:423-527.
Bauer, S., M. Bayer-Raich, T. Holder, C. Kolesar, D. Muller, and T. Ptak (2004). Quantification of
groundwater contamination in an urban area using integral pump tests. Journal of Contaminant
Hydrology 75(3-4): 183-213.
Bayer-Raich, M., J. Jarsjo, R. Liedl, T. Ptak, and G. Teutsch (2004). Average contaminant concentration and
mass flow in aquifers from time-dependent pumping well data: analytical framework. Water Resources
Research 40(8):W08303.1-10.
Bayer-Raich, M., J. Jarsjo, R. Liedl, T. Ptak, and G. Teutsch (2006). Integral pumping test analysis of
linearly sorbed groundwater contaminants using multiple wells: inferring mass flows and natural
attenuation rates. Water Resources Research 42, W08411. doi: 10.1029/2005WR004244.
Bear, J. (1979). Hydraulics of Groundwater. McGraw Hill Book Company, New York.
Bear, J., and M. Jacobs (1965). On the movement of water injected into aquifers. Journal of Hydrology
3:37-57.
Beyke, G. and D. Fleming (2005). In situ thermal remediation of DNAPL and LNAPL using electrical
resistance heating. Remediation Journal 15(3):5-22.
-------
Bob, M.M., M.C. Brooks, S.C. Mravik, and A.L. Wood (2008). A modified light transmission visualization
method for DNAPL saturation measurements in 2-D models. Advances in Water Resources 31(5): 727-
742.
Bob, M.M., M.C. Brooks, S.C. Mravik, and A.L. Wood (2009). The impacts of partial remediation by
sparging on down-gradient DNAPL mass discharge. (In preparation).
Bockelmann, A., T. Ptak, and G. Teutsch (2001). An analytical quantification of mass fluxes and natural
attenuation rate constants at a former gasworks site. Journal of Contaminant Hydrology 53(3-4):429-
453.
Bockelmann, A., D. Zamfirescu, T. Ptak, P.Grathwohl, and G. Teutsch (2003). Quantification of mass
fluxes and natural attenuation rates at an industrial site with a limited monitoring network: a case study.
Journal of Contaminant Hydrology 60(l-2):97-121.
Braida, W.J. and S.K. Ong (1998). Air Sparging: Air-water mass transfer coefficients. Water Resource
Research 34(12):3245-3254.
Broholm, K., S. Feenstra, S., and J.A. Cherry (1999). Solvent release into a sandy aquifer. 1. Overview of
source distribution and dissolution behavior. Environmental Science and Technology 33(5):681-690.
Brooks, M.C., A.L. Wood, M.D. Annable, K. Hatfield, J. Cho, C. Holbert, P.S.C. Rao, C.G. Enfield, K.
Lynch, and R.E. Smith (2008). Changes in contaminant mass discharge from DNAPL source mass
depletion: Evaluation at two field sites. Journal Contaminant Hydrology 102(1-2): 140-153.
Brown, C.L., G.A. Pope, L.M. Abriola, and K. Sepehrnoori (1994). Simulation of surfactant-enhanced
aquifer remediation. Water Resources Research 30(ll):2959-2977.
Brown, G. (2006). Using Multilevel Samplers to Assess Ethanol Flushing and Enhanced Bioremediation at
Former Sages Drycleaners. M.S. Thesis, University of Florida.
Camilleri, D., A. Fil, G.A. Pope, B.A. Rouse, and K. Sepehrnoori (1987). Improvements in physical-
property models used in micellar/polymer flooding. SPE Reservoir Engineering 2(7):433-440.
Chapelle, F.H. and P.M. Bradley (1998). Selecting remediation goals by assessing the natural attenuation
capacity of groundwater systems. Bioremediation Journal 2(3-4):227-238.
Chapelle, F.H., B.B. Looney, and C. Sink (2004). A mass balance approach to monitored natural attenuation.
In: Remediation of Chlorinated and Re calcitrant Compounds, A.R. Gavaskar and A.S.C. Chen (eds).
Chapelle, F.H., J.S. Widdowson, J.S. Bruner, E. Mendez, and C.C. Casey (2003). Methodology for
Estimating Times of Remediation Associated with Monitored Natural Attenuation, U.S. Dept. of the
Interior, U.S. Geological Survey; U.S. Geological Survey, Water-Resources Investigations Report 03-
4057.
Chen, N.H. (1983), Process Reactor Design, Allyn and Bacon, Inc., Boston, MA.
Christ, J.A., C.A. Ramsburg, K.D. Pennell, and L.M. Abriola (2006). Estimating mass discharge from dense
nonaqueous phase liquid source zones using upscaled mass transfer coefficients: an evaluation using
multiphase numerical simulations. Water Resources Research 42(11):W11420.1-13.
Chu, M., PK. Kitanidis and PL. McCarty (2004). Possible factors controlling the effectiveness of
bioenhanced dissolution of non-aqueous phase tetrachloroethene. Advances in Water Resources
27(6):601-615.
Cirpka, O.A. and P.K. Kitanidis (2000). Impact of biomass-decay terms on the simulation of pulsed
bioremediation. Ground Water 38(2):254-263.
Clement, T.P. (1997). RT3D - A Modular Computer Code for Simulating Reactive Multi-Species Transport
in 3-Dimensional Groundwater Systems. Pacific Northwest National Laboratory, Richland, WA, 59 pp.
Clement, T.P, C.D. Johnson, Y. Sun, G.M. Klecka, and C. Bartlett (2000). Natural attenuation of chlorinated
ethene compounds: model development and field-scale application at the Dover site. Journal of
Contaminant Hydrology 42(2): 113 -140.
-------
Conrad, S.H., R.J. Glass, and W.J. Peplinski (2002). Bench-scale visualization of DNAPL remediation
processes in analog heterogeneous aquifers: surfactant floods and in situ oxidation using permanganate.
Journal of Contaminant Hydrology 58(1-2): 13-49.
Cvetkovic, V., G. Dagan, and H. Cheng (1998). Contaminant transport in aquifers with spatially
variable hydraulic and sorption properties. Proceedings of the Royal Society of London, A. Math
454(1976):2173-2207.
Darnault C.J., J.A. Throop, D.A. DiCarlo, A. Rimmer, T.S. Steenhuis, and J-Y. Parlange (1998).
Visualization by light transmission of oil and water contents in transient two-phase flow fields. Journal
of Contaminant Hydrology 31(3-4):337-348.
Datta-Gupta, A., G.A. Pope, K. Sepehrnoori, and M.J. King (1991). High resolution monotonic schemes for
reservoir fluid flow simulation. In Situ 15(3):289-317.
Delshad, M., G.A. Pope, and K. Sepehrnoori (1996). A compositional simulator for modeling surfactant
enhanced aquifer remediation, 1. Formulation. Journal of Contaminant Hydrology 23(4):303-327.
DiFilippo, E.L. and Brusseau, M.L. (2008). Relationship between mass-flux reduction and source-zone
mass removal: analysis of field data. Journal of Contaminant Hydrology 98(l-2):22-35.
Domenico, PA. (1987). Analytical model for multidimensional transport of a decaying contaminant species.
Journal of Hydrology 91(1-2): 49-58.
Edwards, D.A., R.G. Luthy, and Z. Liu (1991). Solubilization of polycyclic aromatic hydrocarbons in
micellar nonionic surfactant solutions. Environmental Science and Technology 25(1): 127-133.
Einarson, M.D., and D.M. Mackay (2001). Predicting impacts of groundwater contamination.
Environmental Science and Technology 35(3):66A-73A.
Elder, C., and C. Benson (1999). Air channel formation, size, spacing, and tortuosity during air sparging.
Ground Water Monitoring and Remediation 19(3): 171-181.
Enfield, C.G., A.L. Wood, P.P. Espinoza, M.C. Brooks, M. Annable, and P.S.C. Rao (2005). Design of
aquifer remediation systems: (1) Describing hydraulic structure and NAPL architecture using tracers.
Journal of Contaminant Hydrology 81(1-4): 125-147.
Esposito, S.J. and N.R. Thomson (1999). Two-phase flow and transport in a single fracture-porous medium
system. Journal of Contaminant Hydrology 37(3-4):319-341.
Falta, R.W., K. Pruess, I. Javandel, and P.A. Witherspoon, (1992). Numerical modeling of steam injection
for the removal of nonaqueous phase liquids from the subsurface, 1. Numerical formulation.
Water Resources Research, 28(2): 451-465.
Falta, R.W. (1998). Using phase diagrams to predict the performance of cosolvent floods for NAPL
remediation. Ground Water Monitoring and Remediation 18(3):94-102.
Falta, R.W. (2000). Numerical modeling of kinetic interphase mass transfer during air sparging using a
dual-media approach. Water Resources Research 36(12):3391-3400.
Falta, R.W. (2003). Modeling sub-grid-block-scale dense nonaqueous phase liquid (DNAPL) pool
dissolution using a dual-domain approach. Water Resources Research 39(12):SBH 18.1-8.
Falta, R.W. (2005). Dissolved Chemical Discharge from Fractured Clay Aquitards Contaminated by
DNAPLs, In: Dynamics of Fluids in Fractured Rock, B. Faybishenko, P.A. Witherspoon, and J. Gale,
eds., Geophysical Monograph 162, American Geophysical Union.
Falta, R.W. (2009). Monte Carlo simulations of source behavior with remediation. (In preparation).
Falta, R.W., C. Lee, S. Brame, J. Coates, and J. Myers (2003). Field Evaluation of Cosolvent-Enhanced In-
Situ Remediation Cooperative Agreement CR-821992-01-0, Clemson University. 86 pp.
Falta, R.W., M.B. Stacy, A.N.M Ahsanuzzaman, M. Wang, and R Earle (2007). REMChlor Remediation
Evaluation Model for Chlorinated Solvents User's Manual Version 1.0. http://www.epa.gov/ada/csmos/
models/remchlor.html
Falta, R.W. (2008). Methodology for comparing source and plume remediation alternatives. Ground Water
46(2):272-285.
-------
Falta, R.W., N. Basu, and P.S.C. Rao (2005b). Assessing impacts of partial mass depletion in DNAPL
source zones: II. Coupling source strength functions to plume evolution. Journal of Contaminant
Hydrology 79(l-2):45-66.
Falta, R.W., K. Pruess, S. Finsterle, and A. Battistelli (1995). T2VOC User's Guide. University of
California, Berkeley, CA, Lawrence Berkeley Laboratory, Earth Sciences Division, Report LBL-36400.
Falta, R.W., P.S.C. Rao, and N. Basu (2005a). Assessing the impacts of partial mass depletion in DNAPL
source zones: I. Analytical modeling of source strength functions and plume response. Journal of
Contaminant Hydrology 78(4):259-280.
Feenstra, S., J.A. Cherry, and B.L. Parker (1996). Conceptual models for the behavior of dense non-aqueous
phase liquids (DNAPLS) in the subsurface. In: Dense Chlorinated Solvents and Other DNAPLs in
Groundwater: History, Behavior, and Remediation, pp. 53-88.
Fine, D., M. Brooks, M. Bob, S. Mravik and L. Wood (2008). Continuous determination of high-vapor-
phase concentrations of tetrachloroethylene using on-line mass spectrometry. Journal of Analytical
Chemistry 80(4): 1328-1335
Freeze, G.A., J.C. Fountain, and G.A. Pope (1994). Modeling the surfactant-enhanced remediation of
PCE at the Borden Test Site using the UTCHEM compositional simulator. Toxic Substances and the
Hydrologic Sciences, pp. 339-345.
Pure, A.D., J.W. Jawitz, and M.D. Annable (2006). DNAPL source depletion: linking architecture and flux
response. Journal of Contaminant Hydrology 85(3-4): 118-140.
Geller, J.T., and J.R. Hunt (1993). Mass transfer from nonaqueous phase organic liquids in water saturated
porous media. Water Resources Research 29(4):833-845.
Gierke, J.S., D.A. Taege, D.L. Perram, and N.J. Hutzler (2002). Design Guidance for Air Sparging of Dense
Nonaqueous Phase Liquids (Tetrachloroethylene, PCE) in Unconsolidated Geological Environments
(Coastal Plain Aquifer): Based on the Performance Measured in a Field Test Conducted at the
Groundwater Remediation Field Laboratory (GRFL), Dover National Test Site (DNTS), Dover Air
Force Base (DAFB), Delaware.
GoldSim (2007). From http ://www.goldsim.com/
Goltz, M.N., J. Huang, M.E. Close, M.J. Flintoft, and L. Pang (2008). Use of tandem circulation wells to
measure hydraulic conductivity without groundwater extraction. Journal of Contaminant Hydrology
100(3-4): 127-136.
Goltz, M.N., M.E. Close, H. Yoon, J. Huang, M.J. Flintoft, S. Kim, and C. Enfield (2009). Validation of
two innovative methods to measure contaminant mass flux in groundwater. Journal of Contaminant
Hydrology 106(1 -2) :51-61.
Guilbeault, M.A., B.L. Parker, and J.A. Cherry (2005). Mass and flux distributions from DNAPL zones in
sandy aquifers. Ground Water 43(l):70-86.
Hand, D.B. (1930). The distribution of a consolute liquid between two immiscible liquids. Journal of
Physical Chemistry 34:1961.
Harten, A. (1983). High resolution schemes for hyperbolic conservation lows. SIAMReview 25(1): 1085-
1090.
Hatfield, K., M.D. Annable, S. Kuhn, P.S.C. Rao, and T Campbell (2002). Anew method for quantifying
contaminant flux at hazardous waste sites. In: Groundwater Quality: Natural and Enhanced
Restoration of Groundwater Pollution. Groundwater Quality 2001 Conference, Sheffield, UK, IAHS.
Hatfield, K., M.D. Annable, J. Cho, P.S.C. Rao, and H. Klammler (2004). A direct passive method for
measuring water and contaminant fluxes in porous media. Journal of Contaminant Hydrology 75(3-
4):155-181.
Hayden, N.J., J. Diebold, and G. Noyes (1999). Phase behavior of chlorinated solvent + water + alcohol
mixtures with application to alcohol flushing. Journal of Chemical and Engineering Data 44(5): 1085-
1090.
-------
Heron, G., J.S. Gierke, B. Faulkner, S. Mravik, L. Wood, and C.G. Enfield (2002). Pulsed air sparging
in aquifers contaminated with dense nonaqueous phase liquids. Ground Water Monitoring and
Remediation 22(4): 73-82.
Hoa, M. (1981). A new method allowing the measurement of rapid variations of the water content in sandy
porous media. Water Resources Research 17(l):41-48.
Holder, T.H., G. Teutsch, T. Ptak, and R. Schwarz (1998). Anew approach for source zone characterization:
the Neckar Valley study. Groundwater quality: remediation and protection. Proceedings of the GQ'98
Conference, vol. 250. IAHS Publication, Tubingen, Germany, pp.49-55.
Hopmans J.W., and J.H. Dane (1986). Calibration of a dual-energy gamma radiation system for multiple
point measurements in a soil. Water Resources Research 22(7): 1009-1114.
Huntley, D. and G.D. Beckett (2002). Persistence of LNAPL sources: relationship between risk reduction
and LNAPL recovery. Journal of Contaminant Hydrology 5 9(1-2): 3-26.
Jarsjo, J. and M. Bayer-Raich (2008). Estimating plume degradation rates in aquifers: effect of
propagating measurement and methodological errors. Water Resources Research 44:W02501,
doi:10.1029/2006WR005568.
Jarsjo, J., M. Bayer-Raich, and T. Ptak (2005). Monitoring groundwater contamination and delineating
source zones at industrial sites: uncertainty analyses using integral pumping tests. Journal of
Contaminant Hydrology 79(3-4): 107-134.
Jawitz, J.W., M.D. Annable, and P.S.C. Rao (1998). Miscible fluid displacement stability in unconfined
porous media: two-dimensional flow experiments and simulations. Journal of Contaminant Hydrology
31(3-4):211-230.
Jawitz, J.W., MD. Annable, G.G. Demmy, and P.S.C. Rao (2003). Estimating non-aqueous phase liquid
spatial variability using partitioning tracer higher temporal moments. Water Resources Research
39(7):SBH 7.1-19.
Jawitz, J.W., A.D. Pure, G.G. Demmy, S, Berglund (2005). Groundwater contaminant flux
reduction resulting from nonaqueous phase liquid mass reduction. Water Resources Research
41(10):W10408.1-15.
Jawitz, J.W., RK. Sillan, M.D. Annable, P.S.C. Rao, and K. Warner (2000). In-situ alcohol flushing of a
DNAPL source zone at a dry cleaner site. Environmental Science and Technology 34(17):3722-3729.
Ji, W., A. Dahmani, D.P Ahlfeld, J.D. Lin and E. Hill (1993). Laboratory study of air sparging: air flow
visualization. Ground Water Monitoring and Remediation 13(4): 115-126.
Jin, M., M. Delshad, V. Dwarakanath, D.C. McKinney, G.A. Pope, K. Sepehrnoori, C.E. Tilburg, and
R.E. Jackson (1995). Partitioning tracer test for detection, estimation, and remediation performance
assessment of subsurface nonaqueous phase liquids. Water Resources Research 31(5): 1201-1212.
Johnson, W.P and W.W. John (1999). PCE solubilization and mobilization by commercial humic acid.
Journal of Contaminant Hydrology 35(4):343-362.
Jury, W.A., and K. Roth (1990). Transfer Functions and Solute Movement Through Soil: Theory and
Application. Birkhauser, Boston, MA.
Keely, J.F. (1982). Chemical time-series sampling. Ground Water Monitoring Review 2:29-38.
Keely, J.F., and F. Wolf (1983). Field applications of chemical time-series sampling. Ground Water
Monitoring Review 3:26-33.
Klammer, H., K. Hatfield, M.D. Annable, E. Agyei, B. Parker, J.A. Cherry, and P.S.C. Rao (2007). General
analytical treatment of the flow field relevant to the interpretation of passive fluxmeter measurements.
Water Resources Research 43(4):W04407.
Kovalic, J.M. (1987). The Clean Water Act of 1987, 2nd edition, Water Pollution Control Federation,
Alexandria, VA.
Kreft A, and A. Zuber (1978). On the physical meaning of the dispersion equation and its solutions for
different initial and boundary conditions. Chemical Engineering Science 33:1471-1480.
-------
Kubert, M., and M. Finkel (2006). Contaminant mass discharge estimation in groundwater based on
multi-level point measurements: a numerical evaluation of expected errors. Journal of Contaminant
Hydrology 84(l-2):55-80.
Lerner, D.N. and G. Teutsch (1995). Recommendations for level-determined sampling in wells. Journal of
Hydrology 171(3-4):355-377.
LFR (2005). Phase I Injection Completion Report Former Sages Dry Cleaners. Prepared for Florida
Department of Environmental Protection, Tallahassee, Florida.
Li, B. and J. Fu (1992). Interfacial tensions of two-liquid-phase ternary systems. Journal of Chemical and
Engineering Data 37(2): 172-174.
Liang, H., and R.W. Falta (2007). Modeling field-scale cosolvent flooding for DNAPL source zone
remediation. Journal of Contaminant Hydrology 96(1-4): 1-16.
Liu, J., M. Delshad, G.A. Pope and K. Sepehrnoori (1994). Application of higher order flux-limited methods
in compositional simulations. Journal of Transport in Porous Media 16(1): 1-29.
Liu Y., B. Bierck, J.S. Selker, T.S. Steenhuuis, and J-Y. Parlange (1993). High intensity x-ray and
tensiometer measurements in rapidly changing preferential flow fields. Soil Science Society of America
Journal 57:1188-1192.
Lowe, D.F., C.L. Oubre, C.H. Ward, and T.J. Simpkin (1999). Surfactants and Cosolvents for NAPL
Remediation: A Technology Practices Manual. Lewis Publishers.
Lunn, S.R.D. (1998). Alcohol Flooding of Porous Media Contaminated with Hazardous Immiscible Liquids.
Ph.D. Dissertation, Civil Engineering, Queen's University, Kingston, Ontario, Canada.
Lunn, S.R.D. and B.H. Kueper (1997). Removal of pooled dense, non-aqueous phase liquid from saturated
porous media using upward gradient alcohol floods. Water Resource Research 33(10):2207-2220.
MacKay, D.M. and J.A. Cherry (1989). Groundwater contamination: pump-and-treat remediation.
Environmental Science and Technology 23 (6): 630- 636.
Maxwell, R.M., S.D. Pelmulder, A.F.B. Tompson, andW.E. Kastenberg (1998). On the development of a
new methodology for groundwater-driven health risk assessment Water Resources Research 34(4): 833-
847.
McGuire, T.M., J.M. McDade, and C.J. Newell, (2006). Performance of DNAPL source depletion
technologies at 59 chlorinated solvent-impacted sites. Ground Water Monitoring and Remediation
26(l):73-84.
McGuire, T.M., C.J. Newell, B.B. Looney, K.M. Vangelas, and C.H. Sink (2004). Historical analysis
of monitored natural attenuation: a survey of 191 chlorinated solvent sites and 45 solvent plumes.
Remediation Journal 15 (1): 99-112.
McKone, T.E. (1987). Human exposure to volatile organic compounds in household tap water: the indoor
inhalation pathway. Environmental Science and Technology 21(12): 1194-1201.
Meinardus, H.W., V. Dwarakanath, J. Ewing, G.J. Hirasaki, R.E. Jackson, M. Jin, J.S. Ginn, J.T. Londergan,
C.A. Miller, and G.A. Pope (2002). Performance assessment of NAPL remediation in heterogeneous
alluvium. Journal of Contaminant Hydrology 54(3-4): 173-193.
Miller, C.T., M.M. Poirier-McNeill, and A.S. Mayer (1990). Dissolution of trapped nonaqueous phase
liquids: mass transfer characteristics. Water Resources Research 26:2783-2796.
Millington, R.J., and J.P Quirk (1961). Permeability of porous solids. Transactions of the Faraday Society
57:1200-1207.
Mravik, S.C., G.W. Sewell, and A.L. Wood (2000). Cosolvent-based source remediation approaches.
In: Remediation of Chlorinated and Recalcitrant Compounds: Physical and Thermal Technologies,
pp. 269-276.
Mravik, S.C., R.K. Sillan, A.L. Wood, and G.W. Sewell (2003). Field evaluation of the solvent extraction
residual biotreatment technology. Environmental Science and Technology 37(21):5040-5049.
-------
Newell, C., R.K. McLeod, and J.R. Gonzales (1996). BIOSCREENNatural Attenuation Decision Support
System User's Manual, Version 1.3, U.S. Environmental Protection Agency, National Risk Management
Research Laboratory.
Newell, C.J., H.S. Rifai, J.T. Wilson, J.A. Connor, J.J. Aziz, and M.P. Suarez (2002). Calculation and
Use of First-Order Rate Constants for Monitored Natural Attenuation Studies, U.S. EPA Remedial
Technology Fact Sheet. U. S. Environmental Protection Agency.
Newell, C.J. and D.T. Adamson (2005). Planning-level source decay models to evaluate impact of source
depletion on remediation time frame. Remediation Journal 15(4):27-47.
Newell, C.J., I. Cowie, T.M. McGuire, and W. McNab (2006). Multi-year temporal changes in chlorinated
solvent concentrations at 23 MNA sites. Journal of Environmental Engineering 132(6):653-663.
Niemet M.R., and J.S. Selker (2001). A new method for quantification of liquid saturation in 2-D
translucent porous media systems using light transmission. Advances in Water Resources 24(6):651-
666.
NRC (2000). Natural Attenuation for Groundwater Remediation, National Research Council, National
Academy Press, Washington, DC, 274 pp.
NRC (2005). Contaminants in the Subsurface Source Zone Assessment and Remediation. Committee on
Source Removal of Contaminants in the Subsurface. The National Academies Press. Washington, DC,
372pp.
O'Hara, S.K., B.L. Parker, P.R. Jorgensen, and J.A. Cherry (2000). Trichloroethene DNAPLflow and mass
distribution in naturally fractured clay: evidence of aperture variability. Water Resources Research
36(1):135-147.
OEHHA (2006). Cancer Potency Information, California Office of Environmental Health Hazard
Assessment, http://www.oehha.ca.gov/risk/ChemicalDB/index.asp.
Ostrom M., C. Hofstee, H. Dane, and R.J. Lenhard (1998). Single-source gamma radiation procedures for
improved calibration measurements in porous media. Soil Science 163(8):646-656.
Parker, B.L., J.A. Cherry, and S.W. Chapman (2004). Field study of TCE diffusion profiles below DNAPL
to assess aquitard integrity. Journal of Contaminant Hydrology 74(1-4): 197-230.
Parker, B.L., R.W. Gillham, and J.A. Cherry (1994). Diffusive disappearance of immiscible-phase organic
liquids in fractured geologic media. Ground Water 32(5):805-820.
Parker, B.L., D.B. McWhorter, and J.A. Cherry (1997). Diffusive loss of non-aqueous phase organic
solvents from idealized fracture networks in geologic media. Ground Water 35(6): 1077-1088.
Parker, J.C. and E. Park (2004). Modeling field-scale dense nonaqueous phase liquid dissolution kinetics in
heterogeneous aquifers. Water Resources Research 40(W05109): 1-12.
Pennell, K.D., L.M. Abriola, and W.J.Weber, Jr. (1993). Surfactant enhanced solubilization of residual
dodecane in soil columns. 2. Mathematical modeling. Environmental Science and Technology
27(12):2341-2351.
Pickens, J.F., J.A. Cherry, G.E. Grisak, W.F. Merrittand, and B.A. Risto (1978). A multilevel device for
ground-water sampling and piezometric monitoring. Ground Water 16(5):322-327.
Pope, G.A., and R.C. Nelson (1978). A chemical flooding compositional simulator. Society of Petroleum
Engineers Journal 18(5):339-354.
Powell, T, G. Smith, J. Sturza, K. Lynch, and M. Truex (2007). New advancements for in situ treatment
using electrical resistance heating. Remediation Journal 17(2):51-70.
Powers, S.E., L.M. Abriola, and W.J. Weber (1992). An experimental investigation of nonaqueous phase
liquid dissolution in saturated subsurface systems: steady-state mass-transfer rates. Water Resources
Research 28(10):2691-2705.
Pruess, K. and A. Battistelli (2002). TMVOC, a Numerical Simulator for Three-Phase Non-Isothermal
Flows of Multicomponent Hydrocarbon Mixtures in Saturated-UnsaturatedHeterogeneous Media,
Lawrence Berkeley National Laboratory, Berkeley, CA, Report LBNL-49375.
-------
Rao, P.S.C., M.D. Annable, and H. Kim (2000). NAPL source zone characterization and remediation
technology performance assessment: recent developments and applications of tracer techniques. Journal
of Contaminant Hydrology 45(l-2):63-78.
Rao, P.S.C., J.W. Jawitz, C.G. Enfield, R.W. Falta Jr, M.D. Annable, and A.L. Wood (2002). Technology
integration for contaminated site remediation: cleanup goals and performance criteria. In: Groundwater
Quality: Natural and Enhanced Restoration of Groundwater Pollution. Proceedings of the Groundwater
Quality 2001 Conference, Sheffield, UK, June 2001, pp. 571-578.
Rao, P.S.C. and J.W. Jawitz (2003). Comment on "Steady-state mass transfer from single-component dense
non-aqueous phase liquids in uniform flow fields" by T.C. Sale and D.B. McWhorter. Water Resources
Research 39(3): 1068.
Reynolds, D.A. and B.H. Kueper (2001). Multiphase flow and transport in fractured clay/sand sequences.
Journal of Contaminant Hydrology 51(l-2):41-62.
Reynolds, D.A. and B.H. Kueper (2002). Numerical examination of the factors controlling DNAPL
migration through a single fracture. Ground Water 40(4):368-377.
Reynolds, D.A. and B. H. Kueper (2004). Multiphase flow and transport through fractured heterogeneous
porous media. Journal of Contaminant Hydrology 71(l-4):89-110.
Rimmer A., D.A. DiCarlo, T.S. Steenhuis, B. Bierck, D. Durnford and J-Y. Parlange (1998). Rapid fluid
content measurement method for fingered flow in an oil-water-sand system using synchrotron x-rays.
Journal of Contaminant Hydrology 31(3-4):315-334.
Rivett, M.O. and S. Feenstra (2005). Dissolution of an emplaced source of DNAPL in a natural aquifer
setting. Environmental Science and Technology 39(2):447-455.
Rivett, M.O., S. Feenstra, and J.A. Cherry (2001). A controlled field experiment on groundwater
contamination by a multicomponent DNAPL: creation of the emplaced-source and overview of
dissolved plume development. Journal of Contaminant Hydrology 49(1-2): 111-149.
Ross, B. and N. Lu (1999). Dynamics of DNAPL penetration into fractured porous media. Ground Water
37(1):140-147.
Sale, T.C., and D.B. McWhorter (2001). Steady state mass transfer from single-component dense
nonaqueous phase liquids in uniform flow fields. Water Resources Research 37(2):393-404.
Sanchez-Vila, X., P.M. Meier, and J. Carrera (1999). Pumping tests in heterogeneous aquifers: an analytical
study of what can be obtained from their interpretation using Jacob's method. Water Resources
Research 35(4):943-952.
Schwarz, R., T. Ptak, T. Holder, and G. Teutsch (1998). Groundwater risk assessment at contaminated
sites: a new investigation approach. In: Groundwater Quality: Remediation and Protection. GQ'98
Conference, Tubingen, Germany, IAHS Publication No. 250.
Schwille, J.F. (1988). Dense Chlorinated Solvents in Porous and Fractured Media, Model Experiments.
Lewis Publishers, Chelsea, MI.
Sillan, R.K., M.D.Annable, and P.S.C. Rao (1997). In-situ Cosolvent Flushing for Enhanced Solubilization
of a Complex NAPL: Comprehensive Field-Scale Evaluation, Draft Final Report. University of Florida,
Gainesville, FL.
Sillan, R.K., M.D. Annable, and P.S.C. Rao (1999). Evaluation ofln-Situ DNAPL Remediation and
Innovative Site Characterization Techniques. Florida Center for Solid and Hazardous Waste
Management, Gainesville, FL.
Slough, K.J., E.A. Sudicky, PA. Forsyth (1999). Numerical simulation of multiphase flow and phase
partitioning in discretely fractured geologic media. Journal of Contaminant Hydrology 40(2): 107-136.
Soga, K., J. Page, and N. Gowers (2002). NAPL source zone remediation - mass flux approach. In:
Environmental Geotechnics. Fourth International Congress on Environmental Geotechnics (4th ICEG),
Rio de Janeiro, Brazil, 2:1069-1081.
Soga, K., J.W. Page, and T.H. Illangasekare (2004). A review of NAPL source zone remediation efficiency
and the mass flux approach. Journal of Hazardous Materials 110(1-3): 13-27.
-------
Sun, Y, J.N. Petersen, and T.P. Clement. (1999a). Analytical solutions for multiple species reactive
transport in multiple dimensions. Journal of Contaminant Hydrology 35(4):429-440.
Sun, Y, J.N. Petersen, T.P. Clement, and R.S. Skeen (1999b). Development of analytical solutions for
multi-species transport with serial and parallel reactions. Water Resources Research 35(1): 185-190.
Suter, G.W. (1993). Ecological Risk Assessment. Lewis Publishers, Chelsea, MI, 643 pp.
Thermal Remediation Services, Inc. (2005). Final NAPL Area 1 Completion Report. In Situ Thermal
Remediation (Electrical Resistance Heating) East Gate Disposal Yard. Ft. Lewis, WA.
Tidwell V.C., and R.G. Glass (1994). X-ray and visible light transmission for laboratory measurement of
two dimensional saturation fields in thin-slab systems. Water Resources Research 30(ll):2873-2882.
Totten, C.T., M.D. Annable, J.W. Jawitz, and J.J. Delfino (2007). Fluid and porous media property effects
on dense non-aqueous phase liquid migration and contaminant mass flux. Environmental Science and
Technology 41(5): 1622-1627.
U.S. ACE (2002). Field Investigation Report, Phase II Remediation Investigation. East Gate Disposal Yard,
Fort Lewis, WA, DSERTS No. FTLE-67.
U.S. EPA (2003). In-situ Thermal Treatment of Chlorinated Solvents: Fundamentals and Field
Applications, U.S. Environmental Protection Agency, Office of Solid Waste and Emergency Response,
Office of Superfund Remediation, Technology Innovation Office, Washington DC. EPA 542-R-04-010.
http://www.epa.gov/tio/download/remed/epa542r04010.pdf
U.S. EPA. (1989). Risk Assessment Guidance for Superfund, Volume 1, Part A. Office of Emergency and
Remedial Response, Washington, D.C. EPA/540/1-89/002.
http://www.epa.gov/oswer/riskassessment/ragsa/index.htm
U.S.G.S. (2000).MODFLOW-2000, The U.S. Geological Survey Modular Ground-Water Model, User
Guide to Modularization Concepts and the Ground-Water Flow Process. U.S.G.S. Open File Report
00-92, 121 pp.
URS Corporation (URS) and INTERA (2003). Final Panel 1 and Panel 5 Surfactant Enhanced Aquifer
Remediation (SEAR) Report, Operable Unit 2, Hill Air Force Base, Utah. Prepared for the Air Force
Center for Environmental Excellence and the Environmental Management Division, Hill Air Force
Base, Utah.
VanAntwerp, D. (2006). Using a Dual-Media Approach to Model Air Sparging Mass Transfer, Clemson
University.
VanAntwerp, D., R.W. Falta, and J.S. Gierke (2008). Numerical simulation of field scale contaminant mass
transfer during air sparging. Vadose Zone Journal 7:294-304.
van Genuchten, M.T and W.J. Alves (1982). Analytical Solutions of the One-Dimensional Convective-
Dispersive Solute Transport Equation, U.S. Department of Agriculture, Technical Bulletin No. 1661,
151pp.
van Valkenburg, M.E. (1999). Solubilization and Mobilization ofPerchloroethylene by Cosolvents in
Porous Media. Ph.D. Dissertation, Environmental Engineering Sciences, University of Florida.
Gainesville, FL.
Waduge, W.A.; Soga, K. and Kawabata, J. (2004). Effect of DNAPL entrapment conditions on air sparging
remediation efficiency. Journal of Hazardous Materials 110(1-3): 173-183
Wagner, D.E., J. Huang, J.L. Heiderscheidt, and M.N. Goltz. (2009). Modeling study to quantify the
benefits of groundwater contaminant source remediation. (In preparation).
Wang, H., X. Chen, and J.W. Jawitz (2008). Locally calibrated light transmission visualization methods to
quantify nonaqueous phase liquid mass in porous media. Journal of Contaminant Hydrology 102(1-
2):29-38.
Weidemeier, T.D., H.S. Rifai, C.J. Newell, and J.T.Wilson (1999). Natural Attenuation of Fuels and
Chlorinated Solvents in the Subsurface. John Wiley and Sons, Inc., New York, NY, 632 pp.
-------
Weidemeier, T.H., M.S. Swanson, D.E. Moutoux, E.K. Gordon, J.T. Wilson, B.H. Wilson, D.H. Kampbell,
P.E. Haas, R.N. Miller, J.E. Hansen, and F.H. Chapelle (1998). Technical Protocol for Evaluating
Natural Attenuation of Chlorinated Solvents in Ground Water. EPA 600/R-98/128, 248 pp.
Weisbrod N., M. Niemet, and J.S. Selker (2003). Light transmission technique for the evaluation
of colloidal transport and dynamics in porous media. Environmental Science and Technology
37(16):3694-3700.
Widdowson, M.A., S. Shearer, R.G. Andersen, and J.T. Novak (2005). Remediation of polycyclic aromatic
hydrocarbon compounds in groundwater using poplar trees. Environmental Science and Technology
39(6):1598-1605.
Williams, P.R.D., L. Benton, and PJ. Sheehan (2004). The risk of MTBE relative to other VOCs in public
drinking water in California. Risk Analysis 24(3):621-634.
Wilson R.D., and M.D. Mackay (1995). Direct detection of residual non-aqueous phase liquid in the
saturated zone using SF6 as a partitioning tracer. Environmental Science and Technology 29(5): 1255-
1258.
Wood, A.L., and C.G. Enfield (1999). In Situ Enhanced Source Removal EPA/600/C-99/002.
Wood, A.L., C.G. Enfield, P.P. Espinoza, M. Annable, M.C. Brooks, P.S.C. Rao, D. Sabatini, and R. Knox
(2005). Design of aquifer remediation systems: (2) Estimating site-specific performance and benefits of
partial source removal. Journal of Contaminant Hydrology 81(1-4): 148-166.
Zeru, A., and G. Schafer (2005). Analysis of groundwater contamination using concentration-time series
recorded during an integral pumping test: bias introduced by strong concentration gradients within the
plume. Journal of Contaminant Hydrology 81(1-4): 106-124.
Zhang, H., and F.W. Schwartz (2000). Simulating the in situ oxidative treatment of chlorinated compounds
by potassium permanganate. Water Resources Research 36(10):3031-3042.
Zhang, Y.K., LA. Khan, X.H. Chen and R.F. Spalding (2006). Transport and degradation of ethanol in
groundwater. Journal of Contaminant Hydrology 82(3): 183-194.
Zhu, L, and J. F. Sykes (2004). Simple screening models of NAPL dissolution in the subsurface. Journal of
Contaminant Hydrology 72( 1 -4): 245 -25 8.
-------
7.0
Appendix A
List of Technical Publications
Peer-Reviewed Papers:
Basu, N.B., A.D. Pure, and J.W. Jawitz (2008), Simplified contaminant source depletion models as
analogs of multiphase simulators, Journal of Contaminant Hydrology 97(3-4):87-99 doi: 10.1016/j.
jconhyd.2008.01.001.
Basu, N.B., A.D. Pure, and J.W. Jawitz (2008). Predicting DNAPL dissolution using a simplified source
depletion model parameterized with partitioning tracers, Water Resources Research, 44:W07414,
doi: 10.1029/2007WR006008.
Basu, N.B., P.S.C. Rao, R.W. Falta, M.D. Annable, J.W. Jawitz, and K. Hatfield (2008). Temporal evolution
of DNAPL source and contaminant flux distribution: impacts of source mass depletion, Journal of
Contaminant Hydrology, 95 (3 -4): 93 -109.
Bob, M., M. Brooks, L. Wood, T. Lee, and C. Enfield (2008). A modified light transmission visualization
method for DNAPL saturation measurements in 2-D models. Advances in Water Resources Journal
31(5):727-742.
Brooks, M.C., A.L. Wood, M.D. Annable, K. Hatfield, J. Cho, C. Holbert, P.S.C. Rao, C.G. Enfield,
K. Lynch, and R.E. Smith. (2008). Changes in contaminant mass discharge from DNAPL source
mass depletion: evaluation at two field sites. Journal Contaminant Hydrology 102(1-2): 140-153,
doi:10.1016/j.jconhyd.2008.05.008.
Chen, X., and J.W. Jawitz, (2008). Reactive tracer tests to predict DNAPL dissolution dynamics in
laboratory flow chambers. Environmental Science & Technology 42(14):5285-5291.
Cho, J., M.D. Annable, and P.S.C. Rao (2003). Residual alcohol influence on NAPL saturation estimates
based on partitioning tracers. Environmental Science & Technology, 37(8): 1639-1644.
Falta, R.W. (2003). A dual domain approach for modeling subgridblock scale DNAPL pool dissolution.
Water Resources Research 39(12):SBH 18-1-8.
Falta, R.W. (2008), Methodology for comparing source and plume remediation alternatives. Ground Water
46(2):272-285.
Falta, R.W., N. Basu and P.S.C. Rao (2005). Assessing the impacts of partial mass depletion in DNAPL
source zones: II. Coupling source strength functions to plume evolution. Journal of Contaminant
Hydrology 79(1-2): 45-66.
Falta, R.W., P.S.C. Rao and N. Basu (2005). Assessing the impacts of partial mass depletion in DNAPL
source zones: I. Analytical modeling of source strength functions and plume response. Journal of
Contaminant Hydrology 78(4):259-280.
Fine, D., M. Brooks, M. Bob, S. Mravik and L. Wood (2008). Continuous determination of high-vapor-
phase concentrations of tetrachloroethylene using on-line mass spectrometry. Journal of Analytical
Chemistry 80(4): 1328-1335.
Pure, A.D., J.W. Jawitz, and M.D. Annable (2006). DNAPL source depletion: linking architecture and flux
response. Journal of Contaminant Hydrology 85(3-4): 118-140.
Goltz, M.N., M.E. Close, H. Yoon, J. Huang, M.J. Flintoft, S.J. Kim, and C.G. Enfield (2009). Validation
of two innovative methods to measure contaminant mass flux in groundwater, Journal of Contaminant
Hydrology 106(1-2): 51-61.
-------
Goltz, M.N., S. Kim, H. Yoon, and J. Park (2007). Review of groundwater contaminant mass flux
measurement. Environmental Engineering Research 12(4):176-193.
Goltz, M.N.; J. Huang, M.E. Close, M.J. Flintoft, and L. Pang (2008). Use of tandem circulation wells to
measure hydraulic conductivity without groundwater extraction. Journal of Contaminant Hydrology
100(3-4): 127-136.
Jawitz, J.W., A.D. Pure, G.G. Demmy, S. Berglund, and P.S.C. Rao (2005). Groundwater contaminant
flux reduction resulting from nonaqueous phase liquid mass reduction. Water Resources Research.
41(10):W10408.1-15.
Kaye, A.J., J. Cho, N.B. Basu, X. Chen, M.D. Annable, and J.W. Jawitz (2008). Laboratory investigation
of flux reduction from dense non-aqueous phase liquid (DNAPL) partial source zone remediation
by enhanced dissolution. Journal of Contaminant Hydrology 102(1-2): 17-28, doi: 10.1016/j.
jconhyd.2008.01.006.
Liang, H., and R.W. Falta (2008). modeling field-scale cosolvent flooding for DNAPL source zone
remediation. Journal of Contaminant Hydrology, 96(1-4):1-16.
Rao, P.S.C. and J.W. Jawitz (2003). Comment on "Steady state mass transfer from single-component dense
nonaqueous phase liquids in uniform flow fields" by T.C. Sale and D.B. McWhorter. Water Resources
Research 39(3): 1068.
Totten, C.T., M.D. Annable, J.W. Jawitz, and J.J. Delfino (2006). Fluid and porous media property effects
on dense non-aqueous phase liquid migration and contaminant mass flux, Environmental Science &
Technology 41(5): 1622-1627.
VanAntwerp, D., R.W. Falta, and J.S. Gierke (2008). Numerical simulation of field scale contaminant mass
transfer during air sparging, Vadose Zone Journal, 7:294-304.
Wang, H., X. Chen, and J.W. Jawitz (2008). Locally calibrated light transmission visualization methods
to quantify nonaqueous phase liquid dissolution dynamics in porous media. Journal of Contaminant
Hydrology 102(1-2)29-38, doi:10.1016/j.jconhyd.2008.05.003.
Wood, A.L., C.G. Enfield, F. Espinoza, M. Annable, M.C. Brooks, P.S.C. Rao, D. Sabatini and R Knox
(2005). Design of aquifer remediation systems: (2) Estimating site-specific performance and benefits of
partial source removal Journal of Contaminant Hydrology 81(1-4): 148-166.
Wood, R.C., J. Huang and M.N. Goltz (2006). Modeling chlorinated solvent bioremediation using
hydrogen release compound (HRC). Remediation Journal 10(3): 129-141.
Zhang, C., H. Yoon, C.J. Werth, A.J. Valocchi, M.B. Basu, and J.W. Jawitz (2008). Evaluation of simplified
mass transfer models to simulate the impacts of source zone architecture on nonaqueous phase liquid
dissolution in heterogeneous porous media. Journal of Contaminant Hydrology 102(l-2):49-60,
doi:10.1016/j.jconhyd.2008.05.007.
Zhang, R. A.L. Wood, C.G. Enfield and S-W. Jeong (2003). Stochastical analysis of surfactant-enhanced
remediation of denser-than-water nonaqueous phase liquid (DNAPL)-contaminated soils. Journal of
Environmental Quality 32(3):957-965.
Technical Reports:
Basu, N. (2006). Flux Based Site Assessment and Design of an Integrated Remediation System. Ph.D.
Dissertation, Purdue University.
Brown, G. (2006). Using Multilevel Samplers to Assess Ethanol Flushing and Enhanced Bioremediation at
Former Sages Drycleaners. M.S. Thesis, University of Florida.
Bulsara, N. (2004/ The Occurrence of the Lead Scavengers Ethylene Dibromide and 1,2-Dichloroethane in
Groundwater from Leaking Gasoline Underground Storage Tanks with a Case Study in South Carolina.
M.S. Thesis, Clemson University.
Falta, R.W. and M.B. Stacy (2008). REMChlor - Remediation Evaluation Model for Chlorinated Solvents.
http://www.epa.gov/ada/csmos/models.html.
-------
Pure, A.D. (2005J. Relationships Between Mass and Flux in DNAPL Source Zones. Ph.D. Dissertation.
University of Florida.
Kavanaugh, M. and P.S.C. Rao, eds. (2003). The DNAPL Remediation Challenge: Is There a Case for
Source Depletion? EPA/600/R-03/143.
Kaye, A. (2006). Laboratory Investigation of Dense Non-Aqueous Phase Liquid (DNAPL) Partial Source
Zone Remediation Using Cosolvents. M.S. Thesis, University of Florida.
Kim, S.J. (2005). Validation of an Innovative Groundwater Contaminant Flux Measurement Method. M.S.
Thesis, AFIT/GES/ENV/05M-02, School of Engineering and Management, Air Force Institute of
Technology.
Lee, J. (2006). Development and Assessment of Surfactant Modified Sorbents to Measure Inorganic and
Organic Contaminant Fluxes Using Passive Flux Meter. Ph.D. Dissertation, Purdue University.
Liang, H. (2005). Multiphase Flow Modeling of Field Scale Cosolvent Flooding for DNAPL Remediation.
M.S. Thesis, University of Clemson.
Totten, C. (2006). Effect of Porous Media and Fluid Properties on Dense Non-Aqueous Phase Liquid
Migration and Dilution Mass Flux. PhD Dissertation, University of Florida.
VanAntwerp, D. (2006). Using a Dual-Media Approach to Model Air Sparging Mass Transfer. M.S. thesis,
Clemson University.
Wagner, D.E. (2006). Modeling Study to Quantify the Benefits of Groundwater Contaminant Source
Remediation. M.S. Thesis, AFIT/GES/ENV/06M-07, School of Engineering and Management, Air
Force Institute of Technology.
Wheeldon, J.G. (2008). An Evaluation of Current Mass Flux Measurement Practices and Implementation
Guide. M.S. Thesis, AFIT/GEM/ENV/08M-23, School of Engineering and Management, Air Force
Institute of Technology.
Wood, R.C. (2005). Modeling Application of Hydrogen Release Compound to Effect In Situ Bioremediation
of Chlorinated Solvent-Contaminated Groundwater. M.S. Thesis, AFIT/GEM/ENV/05M-14, School of
Engineering and Management, Air Force Institute of Technology.
Yoon, H. (2006). Validation of Methods to Measure Mass Flux of a Groundwater Contaminant. M.S.
Thesis, AFIT/GES/ENV/06M-08, School of Engineering and Management, Air Force Institute of
Technology.
Conference Proceedings:
Brown, G.H., M.D. Annable, J. Cho, H. Kim (2005). Surfactant enhanced air sparging in Borden sand.
Bringing Groundwater Quality Research to the Watershed Scale (Proceedings ofGQ2004, the 4th
International Groundwater Quality Conference, held at Waterloo, Canada, July 2004). IAHS Publ. 297,
pp. 418-425.
Falta, R.W. (2003). Modeling subgridblock scale DNAPL pool dissolution using a dual domain approach.
In: Proceedings TOUGH Symposium, Lawrence Berkeley National Laboratory, Berkeley, CA. May 12-
14.
Published Technical Abstracts:
Annable, M.D. (2003). Flux based site characterization and remedial performance assessment. SERDP
Annual Technical Symposium, Washington, DC, December 2-4.
-------
Annable, M.D., M.C. Brooks, C.G. Enfield, R.W. Falta, M.N. Goltz, J.W. Jawitz, P.S.C. Rao, and
A.L. Wood (2006). Field and laboratory evaluation of DNAPL remedial performance. Partners in
Environmental Technology Technical Symposium and Workshop, Washington, DC, November 28-30.
Basu, N.B., A.D. Pure, and J.W. Jawitz (2006). Prediction of down-gradient impacts of DNAPL source
depletion using tracer techniques. Fall Meeting of the American Geophysical Union, EOS Transactions
AGU, 87(52), Fall Supplement, Abstract H23G-04, San Francisco, CA, December 11-15.
Bob, M., M.C. Brooks, T.R. Lee, C.G. Enfield, and A.L. Wood (2004). Saturation measurements of
immiscible fluids in 2-D static systems: validation by light transmission visualization, American
Geophysical Union Fall Meeting, San Francisco, CA, December 13-17.
Bob, M.M, A.L.Wood and T.R. Lee (2004). The impact of partial DNAPL source zone remediation. Fourth
International Conference on Remediation of Chlorinated and Recalcitrant Compounds, Monterey, CA,
May 24-27.
Bob, M.M, M.C. Brooks, S.C. Mravik and A.L. Wood (2007). An experimental assessment of the
impacts of partial DNAPL source zone depletion using sparging as a remediation technology. Second
International Conference on DNAPL: Characterization and Remediation, Niagara Falls, NY, September
24-28.
Bob, M.M, M.C. Brooks, S.C. Mravik and A.L. Wood (2007). A modified light transmission visualization
method for DNAPL saturation measurements in 2-D models. The 2007 AGU Fall Meeting, San
Francisco, CA, December 10-14.
Brooks, M.C., C.G. Enfield, M.D. Annable, and A.L. Wood (2004). Field measurements of contaminant flux
by integral pumping tests, American Geophysical Union Fall Meeting, San Francisco, CA, December
13-17.
Brooks, M.C., C.G. Enfield, M.D. Annable, and A.L. Wood (2004). Field measurements of pre- and post-
remedial contaminant flux by integral pumping tests, SERDP/ESTCP Partners in Environmental
Technology Symposium, Washington DC, November 30-December 2.
Brooks, M.C. and A.L. Wood (2005). Measurement and use of contaminant flux as an assessment tool for
DNAPL remedial performance. NIEHS Web Seminar, August 10.
Brown, G.H., M.D. Annable, and H. Kim (2004). Design of a surfactant enhanced air sparging field test.
Groundwater Quality 2004, 4th International Conference, Waterloo, July 19-22.
Chen, X., H. Wang, and J.W. Jawitz (2006). Comparison of tracers and image analysis for characterizing
DNAPL spatial distribution and the effects of mass removal on contaminant flux reduction. Fall
Meeting of the American Geophysical Union, EOS Transactions AGU, 87(52), Fall Supplement,
Abstract HI 1C-1264, San Francisco, CA, December 11-15.
Espinoza, F., A.L. Wood, M.C. Brooks, and C.G. Enfield (2004). Using tracers to describe NAPL
heterogeneity, American Geophysical Union Fall Meeting, San Francisco, CA, December 13-17.
Falta, R. (2004). Contaminant discharge from fractured clays contaminated with DNAPL. Proceedings of
the Second International Symposium, Dynamics of Fluids in Fractured Rock, Berkeley, CA, February
10-12.
Falta, R.W. (2004). The relationship between partial contaminant source zone remediation and groundwater
plume attenuation. Spring Meeting of the American Geophysical Union, abstract in EOS Transactions,
AGU, 85(17), Montreal, Canada, May 17-21.
Falta, R.W. (2005). Assessing the benefit of radiological source remediation efforts in terms of groundwater
plume attenuation. Poster presented at the 50th Annual Meeting of the Health Physics Society, American
Conference of Radiological Safety, Spokane, WA, July 10-14.
Falta, R.W. (2005). Impact of DNAPL source remediation on dissolved plumes. Invited talk presented at
the Theis Conference, National Ground Water Association, Sedona, AZ January 14-17.
Falta, R.W. (2005). The relationship between partial DNAPL source zone remediation and groundwater
plume attenuation. Presented at the 6th Environmental Technology Symposium and Workshop, U.S.
Army Environmental Center, Portland, OR, March 14-16.
-------
Falta, R.W. (2006). Impacts of DNAPL source zone treatment: experimental and modeling assessment of
benefits of partial source removal. SERDP/ESTCP DNAPL Source Zone Workshop, Baltimore, MD,
March 7-9.
Falta, R.W. (2006). REMChlor - Remediation evaluation model for chlorinated solvent sites. 18th Annual
National Tanks Conference, U.S. Environmental Protection Agency, Memphis, TN, March 20-22.
Falta, R.W., P.S.C. Rao, N. Basu, A.L. Wood, C. Enfield, M. Annable, J. Jawitz and M. Goltz (2004).
Analytical assessment of the impacts of partial mass depletion in DNAPL source zones. SERDP/
ESTCP Partners in Environmental Technology Symposium, Washington, DC, November 29-December
2.
Pure, A.D., J.W. Jawitz, and M.D. Annable (2004). Designing flushing-based remediation systems for
maximum reduction in contaminant mass discharge, EOS Transactions, AGU, 85(46), Fall Meeting
Supplement, Abstract H42A-05. December 13-16.
Pure, A.D., J.W. Jawitz, and M.D. Annable (2004). Relationship between NAPL mass reduction and
contaminant source strength: laboratory experiments. Groundwater Quality 2004, 4th International
Conference, Waterloo, July 19-22.
Pure, A.D., J.W. Jawitz and M.D. Annable (2003). NAPL mass reduction effects on contaminant source
strength: laboratory experiments. American Geophysical Union Fall Meeting, San Francisco, CA,
December 8-11.
Goltz, M. (2003). Impact of source treatment. AFIT Remediation Project Manager Briefing. Wright-
Patterson AFB, April.
Goltz, M.N. (2007). Groundwater contaminant mass flux measurement methods, Stanford Hydrogeology
Seminar Series, Stanford, CA, November 7.
Goltz, M.N. (2007). Modeling and measuring the benefits of a groundwater contaminant source
remediation, Yonsei University, Seoul, Korea, June 21.
Goltz, M.N. (2007). Modeling and measuring the benefits of remediating a groundwater contamination
source. Hanyang University, Seoul, Korea, June 20.
Goltz, M.N., J. Huang, D.E. Wagner, and J.L. Heiderscheidt (2006). Modeling the benefits of groundwater
contaminant source remediation. 2006 Western Pacific Geophysics Meeting, Beijing, China, July 24-27.
Goltz, M.N., J. Huang, D.E. Wagner, and J.L. Heiderscheidt (2006). Modeling the benefits of groundwater
contaminant source remediation. Partners in Environmental Technology Technical Symposium and
Workshop, Washington, DC, November 28-30.
Jawitz, J., M. Annable, M. Brooks, C. Enfield, R. Falta, M. Goltz, S. Rao, and L. Wood (2005). Impacts
of DNAPL source treatment on contaminant mass flux. SERDP Annual Technical Symposium,
Washington, DC, November 29-December 1.
Jawitz, J.W. (2003). Benefits of partial removal of NAPL source zones and measurement of solute fluxes.
United States Department of Agriculture Western Regional Project W-82 Regional Project Meeting,
Tucson, AZ, January 9-10.
Jawitz, J.W. (2003). The relationship between mass reduction and flux reduction for NAPL source zones.
225th American Chemical Society National Meeting, Division of Environmental Chemistry, New
Orleans, LA, March 23-28.
Jawitz, J.W. (2004). Effects of contaminated site age on dissolution dynamics. EOS Transactions, AGU,
85(46), Fall Meeting Supplement, Abstract H31A-0369, December 13-16.
Jawitz, J.W. (2004). Groundwater contaminant flux reduction resulting from contaminant mass reduction.
Gordon Research Conference: Modeling Flow and Transport in Permeable Porous Media, The Queens
College, Oxford, UK, July 11-16.
Jawitz, J.W. (2004). Groundwater contaminant flux reduction resulting from contaminant mass reduction.
Groundwater Quality 2004, 4th International Conference, Waterloo, July 19-22.
Jawitz, J.W. (2006). DNAPL source zone characterization and remedial performance assessment, U.S. EPA
National Remedial Project Manager Training Conference, New Orleans, LA, June 19-22.
-------
Jawitz, J.W., N. Basu, and X. Chen (2007). Prediction of down-gradient impacts of DNAPL source
depletion using tracer techniques: Laboratory and modeling validation. EOS Transactions. AGU,
88(23), Joint Assembly Supplement, Abstract H22C-03, Acapulco, Mexico, May 22-25.
Jawitz, J.W., N. Basu, X. Chen, and A. Pure (2006). Tracer techniques for characterization of aquifer
and contaminant heterogeneity, and remedial performance prediction in a Lagrangian framework.
Geological Society of America Annual Meeting and Exposition, Philadelphia, PA, October 22-25.
Jawitz, J.W., X. Chen, and N.B. Basu (2007). Reactive tracer techniques for predicting DNAPL mass
depletion and flux reduction. Groundwater Quality 2007: 5th International Conference, Fremantle,
Australia, December 2-7.
Jawitz, J.W. and P.S.C. Rao (2003). Influence of aquifer heterogeneity on the mass reduction/flux reduction
relationship for NAPL source zones. Joint Meeting of the European Geophysical Union and the
American Geophysical Union, Abstract EAE03-A-07882, HS5-1MO2O-005, Nice, France, April 7-11.
Rao, P.S.C. (2003). DNAPL source remediation: what is the case for partial source depletion? NRC
Committee Meeting, Irvine, CA, August 19.
Totten, C.T., M.D. Annable, J.W. Jawitz, and J.J. Delfino (2004). Influence of media and fluid properties
on NAPL residual geometry and contaminant mass flux, EOS Transactions. AGU, 85(46), Fall Meeting
Supplement, Abstract H42A-03, December 13-16.
VanAntwerp, D., and R.W. Falta (2006). Modeling kinetic interphase mass transfer during field scale air
sparging operations with a dual domain approach. TOUGH Symposium 2006, Berkeley, CA, May 15-
17.
Wang, H., X. Chen, and J.W. Jawitz (2006). Visualization methods to quantify DNAPL dynamics in
chemical remediation. Fall Meeting of the American Geophysical Union, EOS Transactions AGU,
87(52), Fall Supplement, Abstract H11C-1267, San Francisco, CA, December 11-15.
Wood, A.L. (2003). DNAPL source remediation: assessment of state of the science and practice. EPA
Science Council, Ada, OK, October 21.
Wood, A.L. (2003). Impacts of DNAPL source remediation on source strength. SERDP Annual Technical
Symposium, Washington, DC, December 2-4.
Wood, A.L. (2003). Need and benefits of source control. ORD/HSRC Meeting. Cincinnati, OH, August 26.
Wood, A.L. (2005). DNAPL source remediation research. DOE Performance Monitoring Workshop, Butte,
MT, June 21-23.
Wood, A.L., M.D. Annable, J.W. Jawitz, C.G. Enfield, RW. Falta, M.N. Goltz and P.S.C. Rao (2004).
Impact of DNAPL source treatment on contaminant mass flux. Third International Conference on
Remediation of Chlorinated and Recalcitrant Compounds. Monterey, CA, May 24-27.
Wood, A.L. and C.G. Enfield (2003). Performance monitoring and forecasting. U.S. EPA Technical Support
Project Semi-Annual Meeting, Seattle, WA, April 21-25.
Wood, A.L. and M.C. Brooks (2005). DNAPL source remediation research. Ground Water Forum Meeting,
Phoenix, AZ, May 24.
Wood, A.L. and M.C. Brooks (2007). Flux-based site management. U.S. EPA Ground Water Forum, Las
Vegas,NV, Novembers.
Wood, R.C., J. Huang, C.A. Bleckmann, and M.N. Goltz (2005). Modeling in situ bioremediation of
chlorinated solvent-contaminated groundwater using HRC(r). Eighth International In Situ and On-site
Bioremediation Symposium, Baltimore MD, June 6-9.
Yoon, H., M.E. Close, J. Huang, M.C. Brooks, A.L. Wood, J. Bright, and M.N. Goltz (2006). Validation of
mass flux measurement methods in an artificial aquifer. Fifth International Conference on Remediation
of Chlorinated and Recalcitrant Compounds, Monterey, CA, May 22-25.
-------
Book Chapters:
Annable M.D. (2008). Mass flux as a remedial performance metric at NAPL contaminated sites. In:
Methods and Techniques for Cleaning-up Contaminated Sites. M.D. Annable, M. Teodorescu, P.
Hlavinek and L. Diels (eds). NATO Science for Peace and Security Series-C, pp. 177-186.
Brooks, M.C. and A.L. Wood (2006). The measurement and use of contaminant flux for performance
assessment of DNAPL remediation. In: Remediation of Hazardous Waste in the Subsurface: Bridging
Flask and Field. C.J. Clark II and A.S. Lindner (eds.). ACS Symposium Series 940. American
Chemical Society, Washington, DC, pp. 211-220.
Falta, R.W. (2005). Dissolved chemical discharge from fractured clay aquitards contaminated by DNAPLS.
In: Dynamics of Fluids in Fractured Rock, B. Faybeshenko (ed). Geophysical Monograph Series
Volume 162, American Geophysical Union.
-------
-------
Recycled/Recyclable
Printed with vegetable-based ink on
paper that contains a minimum of
50% post-consumer fiber content
processed chlorine free
-------
-------
m
o
o
I
a
o
I
3
(D
I
(D
m
x
(D
I
0)
a
2.
(£2
8
(D
3
(D
DO
(D
9L
O
0)
O
§
(D
(D
O
0)
------- |