United States Robert S. Kerr EPA/600/R-92/002
Environmental Protection Environmental Research Laboratory January 1992
Agency Ada, OK 74820
Research and Development
&EPA General Methods for
Remedial Operation
Performance Evaluations
-------
General Methods for
Remedial Operations
Performance Evaluations
Project Officer
Randall R. Ross
Extramural Activities and Assistance Division
Robert S. Kerr Environmental Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Ada, Oklahoma 74820
-------
Notice
The information in this document has been funded wholly or in part by the United States Environmental Protection Agency
under Cooperative Agreement No. CR-812808 to the National Center for Ground-Water Research. It has been subjected to
the Agency's peer review and administrative review, and it has been approved for publication as an EPA document. The
material presented in this document was extracted and edited by R.R. Ross and J. W. Keeley from a report entitled "Remedial
Operations Performance Evaluation Methodologies," written by Dr. J.F. Keely for EPA.
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
did not involve environmentally related measurements and did not involve a Quality Assurance Project Plan. Mention of
trade names or commercial products does not constitute endorsement or recommendation for use.
-------
Foreword
EPA is charged by Congress to protect the nation's land, air and water systems. Under a mandate of national environmental
laws focused on air and water quality, solid waste management and the control of toxic substances, pesticides, noise and
radiation, the Agency strives to formulate and implement actions which lead to a compatible balance between human
activities and the ability of natural systems to support and nurture life.
The Robert S. Kerr Environmental Research Laboratory is the Agency's center of expertise for investigation of the oil and
subsurface environment. Personnel at the Laboratory are responsible for management of research programs to: (a) determine
the fate, transport and transformation rates of pollutants in the soil, the unsaturated and saturated zones of the subsurface
environment; (b) define the processes to be used in characterizing the soil and subsurface environment as a receptor of
pollutants; (c) develop techniques for predicting the effect of pollutants on ground water, soil, and indigenous organisms; and
(d) define and demonstrate the applicability and limitations of using natural processes, indigenous to soil and subsurface
environment, for the protection of this resource.
The pump-and-treat process, whereby contaminated ground water is pumped to the surface for treatment, is one of the most
common ground-water remediation technologies used at hazardous waste sites. In fact, pump and treat is a necessary
component of practically all ground water remediation systems, whether for restoration or for plume containment. However,
recent research has identified complex chemical and physical interactions between contaminants and the subsurface media
which may impose limitations on the extraction part of the process. This has raised questions about the effectiveness of
ground water restoration systems and methods for evaluating that effectiveness. This report was developed to summarize the
technical considerations involved in evaluating the performance of ground-water remediation activities.
Clinton W. Hall
Director
Robert S. Kerr Environmental Research Laboratory
in
-------
IV
-------
Table of Contents
FIGURES iv
TABLES iv
SUMMARY v
I. INTRODUCTION 1
Source Characteristics 1
Hydrogeologic Complexities 2
Man-made Complexities 3
Pump-and-Treat Limitations 3
Innovations in Pump-and-Treat 5
II. COMPONENTS OF PERFORMANCE EVALUATIONS 5
Performance Evaluation Strategies 5
1. Locations for Plume Monitoring 7
2. Monitoring Criteria for Performance Evaluations 8
3. Strategies for Monitoring Locations and Criteria 9
4. Measures of Operational Effectiveness 10
5. Measures of Operational Efficiency 11
6. Strategies for Determination of Success/Closure 12
Data Collection Considerations 14
1. Purposes of Data Collection 14
2. Relevant Scale(s) of the Problem 15
3. Actual Scale of the Measurements 15
4. Data Quality vs. Quantity 17
Hydrogeologic Data 17
1. Topographic and Geographic Data 17
2. Geomorphologic and Geologic Data 17
3. Flow Rates 17
4. Hydraulic Parameters 18
5. Fluid-Behavior Data 18
6. Fluid Levels and Pressures 19
Chemical and Geochemical Data 20
1. Natural Ground-Water Chemistry Data 20
2. Soil Chemistry Data 20
3. Contaminant Reaction Data 20
Biological Data 20
Quality Assurance and Quality Control Data 21
III. METHOD SAND PROTOCOLS 21
Performance Evaluation Methods 21
1. Computational Methods 21
2. Statistical Methods 22
3. Graphical Methods 23
4. Theoretical 25
General Protocols for Performance Evaluations 26
1. General Protocol for Selection of Monitoring Criteria and Reporting Requirements 26
2. General Protocol for Violations Reporting and Responses 28
3. General Protocols for Design and Operation Modifications 29
REFERENCES 30
-------
List of Figures
Figure 1. Pump-and-treat limitations may result in the rebound of
contaminant concentrations after the termination of pumping 3
Figure 2. Flow velocities developed during pumping may limit
desorption of contaminants and prevent removal of maximum concentrations 4
Figure 3. Site characterization efforts should be conducted throughout
the RI/FS process to allow data refinement during assessment and remediation 6
Figure 4. Possible monitoring point locations in a ground-water
contamination plume undergoing pump-and-treat remediation 8
Figure 5. Concentration vs. time data plot and multiple regression line with
confidence bounds for judging persistence of the clean-up 13
Figure 6. The results of contouring a dataset by various mathematical
techniques maybe significantly different 24
Figure 7. Pattern diagram for selected VOCs 24
Figure 8. PERT Chart illustrating interdependency of tasks to produce outputs 25
Figure 9. Outline of key efforts needed to select and use monitoring and
reporting requirements 27
Figure 10. Key efforts needed to identify, verify, and correct inadequate
performance of the remediation 28
Figure 11. Key efforts needed to prepare for, implement, and
verify modifications of the remediation 29
List of Tables
1. Useful Statistical Methods for Performance Evaluations
Analysis of Variance (ANOVA) Techniques 23
VI
-------
Summary
This document was developed by an EPA-funded project to explain technical considerations and principles necessary to
evaluate the performance of ground-water contamination remediations at hazardous waste sites. This is neither a "cookbook,"
nor an encyclopedia of recommended field, laboratory, and data interpretation methods. Rather, this report presents and
discusses suggested generic principles for formulating site-specific performance evaluation strategies for ground-water
contamination remediations.
It is widely accepted that ground-water contamination problems cannot be adequately defined or addressed until the
governing physical, chemical, and biological processes which affect the fate and transport of contaminants are characterized
in detail. Recent research has led to a better understanding of these complex processes and how they control the movement of
contaminated ground water through the subsurface. This research has demonstrated that pump-and-treat remediations are far
more complicated than previously thought. Many of the complications result from the tortuosity of the ground-water
flowlines that are generated by the remediation wellfield and the re-distribution of contaminant pathways that occurs. An
immediate consequence of this re-distribution is that historical trends of contaminant concentrations at local monitoring wells
may no longer be useful for predictions and evaluations regarding the growth or reduction of the contaminant plume. Further,
the pattern of flow velocities and directions that resulted from the remediation wellfield may change substantially over time,
thus complicating attempts to evaluate the progress of the pump-and-treat remediation.
The increased ground-water flow velocities created by remediation wellfields may also introduce significant hydrodynamic
and chemical limitations to contaminant transport and subsequent remediation locally. Such limitations include
hydrodynamic isolation of portions of remediation wellfields, diffusively restricted movement of contaminants in low-
permeability subsurface materials, and failure to meet local chemical equilibrium requirements for maximal desorption and
liquid-liquid partitioning. For each well that is utilized to remove fluids from the subsurface, there is a stagnation zone
downgradient where contaminants do not receive active flushing while the well is pumping. Everywhere that water levels
have been lowered by the action of the remediation wellfield, there are sediments that may release additional contaminants
upon cessation of pumping and return of the water levels to normal levels. Throughout the sediments that overlie and
comprise the contaminated aquifer there may be inclusions of low-permeability sediments. Such inclusions may act as long-
term contaminant reservoirs which release contaminants much more slowly (often, by diffusion only) than the surrounding
sediments where the vast majority of the ground water flows. The presence of nonaqueous phase liquids (NAPLs) may
represent one of the most significant limitations to pump-and-treat remediations. These immiscible liquids may leave a trail
of residual contaminants, trapped by capillary forces, as they move through the unsaturated and saturated zones. This residual
contaminant mass may serve as a long term source of ground-water contamination. Proper site-specific characterizations of
these limitations are essential in evaluating the performance of pump-and-treat remediation systems.
Emphasis is given to the need to develop a monitoring strategy at each site which ensures that sampling locations and
schedules are meaningful, not only as early-warning alarm systems, but in measuring progress toward remediation goals and
to make adjustments to improve performance. Emphasis is also given to the use of data reduction, presentation, and
interpretation techniques that may be used in evaluating the performance of the remediation. These nominally include
ground-water flow and contaminant transport models, statistical analyses, graphical techniques, and theoretical relationships.
The optimal effectiveness (extent and uniformity of cleansing) and efficiency (minimization of costs and duration) of a
remediation can often be obtained by managing the pump-and-treat system in terms of flowrates and extraction locations in
response to reductions in contaminant mass in portions of the plume. Active management of this kind must be supported by
the continuous gathering of key site characterization data.
vn
-------
I. Introduction
injection wells as compared to the relatively simple natural
flowlines.
Ground-water pump-and-treat systems are the most
common remedial technology used to remediate and/or
contain ground-water contaminants at hazardous waste sites.
Pump-and-treat remedial actions involve the extraction of
contaminated ground water and subsequent treatment for
contaminant removal. This report presents a review of the
factors and subsurface processes which may affect the
performance of pump-and-treat remediations. Also
discussed are the basic components of performance
evaluations, and suggested methods and protocols for
evaluating performance. It is assumed that the reader is
familiar with the basic concepts of hydrogeology. For
additional information refer to U.S. EPA, 1989a, 1989b, and
Mercer and others, 1990.
Pump-and-treat remediations are planned or have been
initiated at many ground-water contamination sites across
the country. Regulatory responsibilities require that
adequate oversight of these remediations be made to answer
the question: What can be done to demonstrate whether or
not a remediation is generating the desired control?
Recently, other questions have evolved with the realization
that many pump-and-treat remediations are not functioning
as well as had been expected:
1. What can be done to determine whether or not the
remediation will meet its proposed timelines? and,
2. What can be done to determine whether or not the
remediation will stay within budget?
It may be possible to better answer these questions with the
aid of sophisticated data analysis tools, such as
computerized mathematical models of ground-water flow
and contaminant transport.
Computer models can be used to make predictions about
future performance, but the results are highly dependent on
the quantity and quality of field and laboratory data utilized.
This is also true of models used for ongoing performance
evaluations of pump-and-treat remediations. They must be
supported by detailed site information that is continually
updated. In most instances, an accurate performance
evaluation cannot be made simply by comparing data
obtained from monitoring wells during remediation to the
water quality data generated prior to the onset of
remediation. Historical trends of contaminant levels at local
monitoring wells are often not useful for comparisons with
data obtained during the operation of pump-and-treat
remediations. This is a consequence of the complex
ground-water flow patterns produced by extraction and
Detailed field investigations are required during remediation
to determine the locations of flowlines generated by a
pump-and-treat operation. It is likely that mathematical and
statistical models will be required to organize and analyze
this data. Ongoing site characterization efforts must include
examinations of the contaminant source characteristics, the
local and regional hydrogeologic complexities,
contaminant-specific behavior and plume dynamics.
Source Characteristics
There are many potential sources of ground-water
contamination at hazardous waste sites. These may be
categorized as either "point" or "nonpoint" sources,
depending on their geometry, and generally are engineered
structures with recognizable physical boundaries. Some
may release significant volumes of contaminants to the
environment during their operating history. Pits, ponds, and
lagoons, for example, constructed without lined bottoms or
sides are notorious for large fluid losses. Industrial and
municipal landfills are point sources that tend to generate
leachate by chemical reactions of the waste with infiltrating
precipitation. Such leachate very often contains high
concentrations of dissolved metals, inorganic salts, and
organic chemicals.
Historically, shallow injection wells have been used to
dispose of untreated stormwaters from city streets, parking
lots, highways, and airports and represent the most direct
avenue of contaminant introduction to ground water. Deep
injection wells continue to be used for the disposal of liquid
industrial waste. Point sources of ground-water
contamination also include leaks and spills at manufacturing
facilities such as refineries, pharmaceutical plants, fertilizer
plants, and other chemical production plants. Leaking
underground storage tanks and associated piping at gasoline
service stations and fuel depots also pose a serious point
source contamination problem. Some point sources, such as
abandoned wells, are very small and particularly difficult to
locate; whereas other point sources, such as mining
operations or municipal and industrial landspreading
operations are readily located, but their large size makes
identification of specific points of contaminant releases
difficult.
The more commonly recognized nonpoint sources of
ground-water contamination include agricultural
applications of fertilizers, herbicides, and pesticides. Other
traditional practices and activities that may result in
widespread releases of contaminants to underlying ground
water include the application (and storage) of salts to streets
and highways for winter de-icing, and the spraying of
-------
waste-oils on unpaved roads for dust suppression. Small
spills that typically occur during dispensing or refilling
operations at transfer points and storage locations often go
unnoticed or are ignored because of their minor size. The
aggregate effect of many small spills may be equivalent to a
non-point source of contaminant releases to the underlying
ground water. It is generally easier to identify, control, and
eliminate point sources than nonpoint sources. The control
of nonpoint sources is often limited to prevention or
restriction of the activities which generate the contaminant.
The mechanism by which sources introduce contaminants to
ground water has a profound effect on the duration and area!
extent of the resulting contamination. The duration of a
source release may affect the extent of contamination. For
example, a small spill at the surface may be attenuated over
a short distance while contaminants introduced below the
surface, such as leaking tanks and pipelines, will have much
less attenuation. The hydraulic impacts of such sources may
impart a three-dimensional character to the local ground-
water flow regime. Interactions with streams and other
surface water bodies may also impart three-dimensional
flow characteristics to contaminated ground water.
In ideal settings, the design of a pump-and-treat system
might be limited to the configuration of the extraction/
injection wellfield and the flowrates of the individual wells.
Unfortunately, there are few natural settings that appear to
be ideal at the scale of interest in pump-and-treat
remediations. Karst and fractured rock aquifers frequently
exhibit highly complex flow and contaminant transport
behaviors, as do many interbedded sand-and-gravel aquifers
(e.g., diffusively restricted flow in silt and clay strata)
(Barker and others, 1988; Feenstra and others, 1984;
Spayed, 1985; White, 1988). The general physiographic
setting may also be highly restrictive to pump-and-treat
remediation, such as in wetlands and wildlife sanctuaries.
The properties and size of the contaminant source and
resulting plume are also key features of the remedial action
setting. The amount of contaminant that may be readily
removed by pump-and-treat remediation in a reasonable
amount of time may be only a small portion of what is in the
subsurface (Hossain and Corapcioglu, 1988; Hunt and
others, 1988a and 1988b; Keely and others, 1986; Stover,
1989; Testa and Paczkowski, 1989).
Hydrogeologic Complexities
The geologic processes that shape the surface of the earth
are so varied it is easy to understand why most aquifers do
not exist as a single horizontal stratum of uniform
composition and thickness. As a result, the hydraulic impact
of extraction wells is often distorted, and if this occurs in the
zone of remediation, the system may not be as effective as
anticipated and additional pumping might be required. By
examining some of the more common hydrogeologic
complexities, it may be possible to anticipate these
additional requirements and account for them in the system
design.
Some of the most complex hydrogeologic settings for
pump-and-treat remediations are those that involve the
contamination of zones of low-permeability. These
formations can be categorized as active aquifers, despite
very low yields and high drawdowns. Remediations in
these types of settings may be further complicated by the
presence of limited occurrences of high permeability
sediments, such as sand and gravel, or fractures.
Pump-and-treat remediation in low-permeability formations
may be limited because yields are too low to allow
continuous pumping. Extraction wells may be designed to
operate on demand or, in some cases, it may be possible to
maintain continuous pumping by providing recharge over
the capture zone of each well.
Minor pump-and-treat problems can also be found in high-
permeability formations. Flow net complications can result
from the minor occurrence of low-permeability strata. For
example, in formations such as alluvial floodplains and
glacial outwash deposits, it is common to find thin clay and
silt lenses extending tens to hundreds of feet in the direction
of flow.
In some geologic settings, most of the ground-water flow
occurs through fractures in low-permeability rock
formations (Barker and others, 1988; Feenstra and others,
1984; Spayed, 1985). Contaminant movement through
fractures may be primarily controlled by advective forces.
However, flow through the bulk matrix of many rocks is so
slow that contaminant movement may be dominated by
molecular diffusion.
Extensively fractured rock may exhibit hydraulic behavior
similar to porous media. In fractured rock aquifers
characterized by a very high degree of fracturing, many of
the equations developed for porous media may be applied
with reasonable confidence for the design of remediation
wellfields (Bear and Verruijt, 1987; de Marsily, 1986;
Mercer and Faust, 1981). When only a few fractures
dominate flow, the relationship of flowrate to drawdown
may be highly nonlinear and unpredictable. Equations have
been developed which allow prediction of drawdowns for
situations where a single horizontal fracture or a single
vertical fracture passes through a pumping well, but there
are no genetically useful equations for multi-borehole or
multi-fracture situations. The usual design approach for
systems dominated by discrete fractures is to try to position
each extraction well such that it intersects as many fractures
-------
as possible. In such cases it will probably be necessary to
conduct several pilot tests of extraction wells to develop
site-specific relationships of flowrate to drawdown. Such
empirical relationships may be highly nonlinear as a result
of changes in the fracture aperture during pumping due to
nonuniform drainage along the fracture face and the
siphoning effects of irregular fractures. In settings where
the flowrate-drawdown relationship at one extraction well
differs markedly from that of another, the wellfield design
strategy must default to well-by-well installation and
testing. Each extraction well should be tested singly, and in
concert with existing extraction wells to determine the
short-term effectiveness in providing the desired gradient
control. Supplementary extraction wells may be necessary
where additional control is needed.
Similar considerations and measures may be required for
hazardous waste sites located in karst terrains. Two flow
regimes, open-channel and pipe flow, occur in karst settings
that rarely occur elsewhere in the subsurface (White, 1988).
Because of these complexities, advection and dispersion of
contaminants through karst aquifers are not adequately
describable by Darcy's Law and other porous media
concepts.
It has been shown that streams and springs control the flow
through karst terrain. Although dye tracers have proven
useful in studying fluid movement in karst aquifers, such
studies have not yielded relationships that can be transferred
from one site to another. While it may seem reasonable to
locate extraction wells on the basis of intersecting karst
streams, one should not expect a consistent response as the
drawdown and flowrate may vary substantially over short
periods of time.
Man-made Complexities
It is not uncommon for the construction of extraction wells
to be hampered by buried tanks and utility lines, streets,
sewers, and water and gas pipes. Railroads, canals, and
power transmission lines can present similar location
difficulties as well as add safety hazards. This lack of
access can create major difficulties for pump-and-treat
remediations, not only from the standpoint of well
locations, but the required piping network as well.
Seasonally operated pumps and on-demand pumping may
produce unanticipated hydraulic impacts locally. If a pump-
and-treat remedy is designed without the knowledge of local
pumping wells, it may not overcome the hydraulic impact
sufficiently to maintain effective containment of
contaminants (Keely, 1984; Keely and Tsang, 1983; Ward
and others, 1987). Potential failure of the remediation by
this avenue is unlikely to be recognized unless
hydrodynamic monitoring is instituted for both horizontal
and vertical gradient control.
Drilling restrictions may be in effect for other purposes such
as water conservation, water rights, or to minimize upconing
of saline water. Fortunately, many institutional controls are
negotiable, particularly if the potential impacts of operation
of a pump-and-treat remediation can be shown to be
controllable or of only a temporary nature.
Pump-and-Treat Limitations
Conventional pump-and-treat remediations are based on the
operation of extraction and injection wells. The level of
contamination measured at monitoring wells may be greatly
reduced in a moderate period of time, but low levels of
contamination may persist indefinitely. The contaminant
mass removed may decline over time and gradually
approach a residual level in the latter stages (Figure 1). At
that point, large volumes of water are treated to remove
small amounts of contaminants. Depending on contaminant
residuals within the aquifer, this may cause a remediation to
be continued indefinitely. An increase in ground-water
contaminant levels may follow if the remediation is
discontinued prior to removal of all residual contaminants.
Several contaminant transport processes are potentially
responsible for the contaminant rebound effects and may
include:
(1) diffusion of contaminants in low-permeability
sediments,
(2) hydrodynamic isolation within wellfields,
(3) desorption of contaminants from sediments, and
(4) partitioning of immiscible fluids into ground water.
Time—>•
Figure 1. Pump-and-treat limitations may result in the rebound of
contaminant concentrations after the termination of
pumping (after Keely, 1989).
-------
Local variations in ground-water flow results from the
heterogeneous nature of interlayered high-permeability and
low-permeability sediments (Anderson, 1979; Guvenand
Molz, 1986; Keely, 1984; Keely and others, 1986; Matheron
and de Marsily, 1980; Molz and others, 1986 and 1988;
Osiensky and others, 1984; Satlin and Bedient, 1988).
When operating a remediation wellfield, these variations in
flow may result in relatively rapid cleansing of the higher-
permeability sediments, which may conduct most ground-
water flow. Contaminants are removed from the lower-
permeability sediments very slowly primarily by diffusion.
The specific rate at which this diffusive release occurs is
dependent on the concentration gradient within and external
to the low-permeability sediments. As the higher-
permeability sediments are remediated, the concentration
gradient drawing contaminants from the lower-permeability
sediments is at its greatest and is exhausted only when the
chemical concentrations are at equilibrium.
Fine-grained sediments generally have lower permeabilities
and orders-of-magnitude greater surface areas than coarse-
grained sediments (Bittonand Gerba, 1984; Corey, 1977;
Hillel, 1982; Keely and others, 1986; Mackay and others,
1986; Piwoni and Banerjee, 1989). Much greater amounts
of contaminants may accumulate in low-permeability
sediments by adsorption, ion-exchange, or other surface
chemical processes as compared with contaminant
accumulations in a like volume of high-permeability
sediments (Bouchard and others, 1989; Enfield and
Bengtsson, 1988). The occurrence of low-permeability
materials in higher-permeability formations may be a major
limitation to pump-and-treat remediations. Diffusive
transport of contaminants from the low-permeability stratum
to the surrounding high-permeability formation may dictate
the time necessary for complete remediation.
If substantial occurrences of low-permeability materials are
adequately characterized it may be possible to target these
selectively during the pump-and-treat remediation for
excavation, in-situ soil washing, hydrofracturing, or other
remedial methods. However, if low-permeability materials
occur as heterogeneously distributed microstrata that defy
physical removal or in-situ treatment, conventional pump-
and-treat technologies may be implemented for the
containment of ground-water contaminants, rather than
remediation.
The operation of any wellfield results in the formation of
hydrodynamically isolated stagnation zones downgradient
of extraction wells and upgradient of injection wells. Mass
transport into or out of the isolated zone occurs mainly by
diffusion. Remedial action wells located within the bounds
of a contaminant plume may have stagnation zones which
are not effectively remediated. Flowline patterns must be
altered by changing the locations of pumping wells, or by
altering the balance of flowrates among pumping wells to
remediate the stagnation zones.
Adsorption and desorption of organic chemicals by
sediments are not always instantaneous processes as
hundreds of hours may be required for certain compounds to
reach chemical equilibrium (Bahr, 1989; Bouchard and
others, 1988; Brusseau and others, 1989; Curtis and others,
1986; Goltz and Roberts, 1986 and 1988; Lee and others,
1988; Mackay and others, 1986; Miller and others, 1985;
Nkedi-Kizza and others, 1985; Piwoni and Banerjee, 1989;
Valocchi, 1988; Woodburn and others, 1986). When a
pump-and-treat remediation is implemented, ground-water
flow velocities will be significantly increased and it may not
be possible for passing ground water to achieve or maintain
chemical equilibrium with the contaminants (Figure 2).
Consequently, large volumes of mildly contaminated water
may be produced over a long period of time.
Advection
Organic Carbon or
Mineral Oxide Surface
cs
a
'3 fi
§ o
2
+j
S
o
o
Eauil. Cone.
Slow Desorption
Initial Rapid
Desorption
Time
Figure 2. Flow velocities developed during pumping may limit
desorption of contaminants and prevent removal of
maximum concentrations (after Keely, 1989).
-------
Immiscible, or nonaqueous phase liquids (NAPL), may
impose significant limitations on pump-and-treat
remediations. NAPLs are characterized as either light
(LNAPL) or dense (DNAPL) with respect to their density
relative to that of water. LNAPLs, often referred to as
"floaters," may accumulate at the top of the saturated zone
while DNAPLs, often referred to as "sinkers," are capable
of moving vertically through the unsaturated zone and
portions of the saturated zone.
Both light and dense immiscible liquids are subject to
capillary forces of sediments which may retain or trap
portions of the liquid in pore spaces (Huling and Weaver,
1991: Faust, 1985; Hinchee and Reisinger, 1987; Hunt and
others, 1988aand 1988b; Stover, 1989; Taylor, 1987). This
may be particularly troublesome when an extraction well is
utilized when a continuous phase (free product) LNAPL is
present as both the water and free product flow into the cone
of depression. The cone of depression may contain a
significant volume of trapped residual contaminant below
the water-table, which will serve as a continuous
contaminant source long after the extraction well stops
operating. The extent of contamination generated by the
residual contaminant source may exceed that generated by
the LNAPL pool prior to the onset of pumping.
Contaminant plumes composed of components of an
immiscible liquid are generated by the dissolution of the
NAPL into adjacent ground water, according to liquid-liquid
partitioning principles (Hunt and others, 1988a and 1988b;
Stover, 1989; Stumm and Morgan, 1981). Generally, the
lower the solubility, the slower the dissolution rate (Stumm
and Morgan, 1981; Weast, 1985). In many instances
involving pump-and-treat remediations, the contact time
between ground water and immiscible fluids may be too
short to allow maximum concentrations to develop.
Analogous to the limitation posed by slow chemical
equilibrium for some adsorptive contaminants, pumping-
induced limitation of liquid partitioning will generate large
volumes of mildly contaminated water under continuous
operation of the remediation wellfield.
There is evidence that DNAPLs are present at many
hazardous waste sites but they may largely go undetected
due to the numerous variables influencing their transport
and fate. As a result, the presence of DNAPLs may be a
limiting factor to ground-water pump-and-treat
remediations.
Innovations in Pump-and-Treat
To minimize the potential for poor contaminant removal
rates, innovative pumping strategies may be required.
These involve reconfiguring the remediation wellfield,
adopting unconventional pumping schedules, or integrating
pumping with other remediation technologies such as
subsurface barrier walls.
One innovation in pump-and-treat remediations is pulsed
pumping. Pulsed operation of hydraulic systems is the
cycling of extraction or injection wells in active and resting
phases. The resting phase of a pulsed-pumping operation
may allow sufficient time for contaminants to diffuse out of
low-permeability zones into adjacent high-permeability
zones and allow sorbed contaminants and NAPL residuals
sufficient time to reach equilibrium. Subsequent pumping
then removes maximum contaminants in a minimum
volume. Pulsed pumping will also bring zones of stagnation
into active flowpaths.
Other methods to enhance the efficiency of ground-water
remediations currently being pursued include the use of
surfactants, biorestoration, fracture enhancement, and
physical containment.
II. Components of
Performance Evaluations
Performance Evaluation Strategies
Performance evaluations of remedial actions may be
conducted at hazardous waste sites for a variety of reasons.
One of the most common reasons is to satisfy the
monitoring requirements established by negotiated or court
settlements, thus ensuring that the desired controls on
ground-water flow and contaminant transport are
maintained. Another reason may be to determine if the
remedial action objectives can be accomplished using the
selected remedial action alternative(s). The strategy
adopted for conducting a performance evaluation must focus
on the kinds of data that will be collected and how those
data will be presented for interpretation and decision
making purposes. It is best to start building a strategy,
therefore, by examining the data collection studies that have
been done in the past at a site.
The phases of work conducted in site investigations follow a
pattern of increasing detail and focus (Figure 3). This can
be viewed as a continuous data collection program that may
be extended into the remediation phase to benefit oversight
monitoring and operational management purposes alike. In
the Superfund site investigation process, the preliminary
assessment phase involves collection of readily obtainable
information which may include technical data for a
screening-level assessment of the potential need for further
investigation. Where the preliminary assessment indicates a
need for additional characterization, the next phase of work
-------
• Preliminary Assessment
•Site Inspection
Scoring of
RI/FS
Remedial Investigation
•Site
Characterization
• Treatability
Investigation
• Development/Screening
of Alternatives
• Detailed Analysis
of Alternatives
Feasibility Study
• Remedy Selection
' Record of Decision
• Remedial Design
' Remedial Action
Figure 3. Site characterization efforts should be conducted throughout the RI/FS process to allow data refinement during assessment and
remediation.
most often consists of site visits for visual appraisal and
limited sampling.
If these additional efforts indicate the need for further
characterization, then a preliminary assessment/site
investigation (PA/SI) is performed and the data used to
score the site using the Hazardous Ranking System. This
score determines whether the site will be listed on the NPL
and, consequently, whether Superfund resources can be
utilized. If so, a remedial investigation and feasibility study
(RI/FS) are undertaken. The remedial investigation is a
focused field and laboratory study of the distribution of
contaminants and potential transport and exposure
pathways. The feasibility study is a detailed examination of
the remedial action alternatives (U.S. EPA, 1988a).
The data collected during the RI/FS, including treatability
studies if warranted, should be such that the remediation
selection decision is supported and that pertinent parameters
can be determined including pumping rates, the number and
location of extraction wells, and other remediation features.
In some cases, however, additional information may be
required and collected during predesign. Experience has
shown that it may also be useful to conduct periodic
operational studies to support active management of a
remediation system (U.S. EPA, 1989a). Such studies may
consider the effects of changing flow rates, adding or
deleting wells, altering the treatment train or other changes
in the remedial action.
Strategies for performance evaluations should view data
gathering activities during the remediation as an extension
of the historical data gathering and characterization of the
site. For the sake of continuity and the ability to conduct
meaningful comparisons at future dates, the performance
evaluation program should (i) expand the collection of key
hydrodynamic and chemical data; and (ii) continue and
expand the site characterization efforts to refine the
databases regarding the hydrogeology and transport
processes, to the full extent that this is needed for both
oversight monitoring and operational management purposes.
Traditionally, the key controls on the form and quality of the
technical data obtained during remediations have been the
monitoring criteria that are selected and the monitoring
point locations at which those criteria are to be applied.
This traditional collection of ground-water samples for
chemical analyses and measurements of water level
elevations has not been adequate for detailed performance
evaluations. There is a need for ongoing improvement of
the site characterization database. This database may be
updated by periodic or opportunistic testing, which may
include analysis of chemical extracts of aquifer sediment
samples collected from new boreholes, and aquifer testing,
geophysical and geological logging of new wells. Such
information would allow refinement of ground-water
velocity and flowpath estimates under various pumping
conditions.
The collection of aquifer sediment samples for identification
and quantification of contaminant mass reserves is an
example of a technical activity that should be routine for
pre-operational site characterization studies, but which is
rarely done during operation of remedial actions. This is
-------
unfortunate because it may provide one of the few direct
means of evaluating the general effectiveness of the
remedial action (e.g., the magnitude of contaminant mass
reduction obtained during a given time period). However,
due to the inherent heterogeneous nature of the subsurface
this may only provide general indication of a remediation's
effectiveness. Together with operational data, information
from the analyses of aquifer samples can be used to estimate
the future rate of contaminant mass reduction under the
existing wellfield configuration and pumping schedule.
With the same information, it should also be possible to
estimate the best performance of the remediation, which is
defined here as the lowest contaminant concentration profile
maintainable. However, this assumes that the contaminant
sources have been accurately delineated and that no
immiscible contaminant sources are present. Continuous
characterization of the site during remediation will provide
the basis for determining the optimal wellfield configuration
and pumping schedule for the next forecast period, and for
projecting the best performance during that period. One of
the best reasons for active remediation management is based
on economics. Pumping the minimum volume at the
highest concentrations and in the minimum time frame is the
prime avenue to cost minimization.
Note, however, that even if an initially optimized
remediation is operated continuously without any changes in
wellfield configuration or pumping schedule, occasional re-
estimation of its best performance may be necessary, due to
technological advances and because estimates of the
contaminant masses in reserve and their effective rates of
depletion may change with time. As a consequence, there is
a need for continuous collection of routine monitoring and
operational data during remediation. Operational data is
critical for system optimization, and must be obtained
regularly to achieve and maintain the most efficient and
effective remediation. Important data include:
(i) flow rates of extraction and injection
wells;
(ii) water-level elevations in all wells and piezometers;
and
(iii) the concentrations of contaminants in monitoring
wells, extraction and injection wells, and the
influent and effluent of each treatment process unit.
1. Locations for Plume Monitoring
It has long been recognized that it is important to provide
continuous information on what is happening downgradient
of the plume boundary, because this is the means by which
potential receptors may be warned of the advance of the
plume. It is also important to monitor inside the plume
during remediation. Such actions may make it possible to
determine which parts of the plume are being remediated
effectively and efficiently. This facilitates management of
the remediation wellfield for best performance. For
example, this may be accomplished by reducing flowrates
from extraction wells producing water from relatively clean
zones and increasing the flowrates from wells producing
water from highly contaminated zones. By contrast, the
exclusive use of oversight monitoring points downgradient
of the plume boundary does not allow an understanding of
plume behavior during remediation, except to indicate when
contaminants have migrated beyond the zone of remedial
action.
There are many kinds of monitoring points which are
useful for performance evaluations. Natural water quality
or background sampling locations are the most widely
used monitoring points and are positioned outside of the
plume boundary (Figure 4). The location is chosen such
that:
(i) it is neither in the plume nor in adjacent areas that
may be affected by the remediation;
(ii) if downgradient, it is in an uncontaminated
portion of the aquifer through which the plume
would migrate if the remediation failed; and
(iii) its location minimizes the possibility of detecting
other potential sources of contamination (e.g.,
relevant to the target site only).
Data gathered from background monitoring points located
upgradient or cross-gradient of a contaminant plume may
also serve to indicate when contaminants have migrated
beyond the zone of remedial action. This latter might occur
when injection wells are used to return treated water to the
aquifer at the upgradient edge of the plume and
simultaneously have the unintended effect of forcing a
portion of the contaminant plume outward. The upgradient
and cross-gradient locations of background monitoring
points may also be useful in providing a continuous check
on the regional water quality, thus providing early warning
of previously undetected contaminants flowing into the area.
Public water supply wells located downgradient of a plume
are a second kind of beyond-plume monitoring point. These
wells were originally drilled in locations suitable for water
supply purposes and were never intended to serve as
monitoring wells. However, they may be sampled to assure
the quality of ground water that is delivered to consumers
using monitoring criteria that are related to specific
contaminants associated with a site and maintenance of
existing quality, or other health-based standards.
-------
Gradient-control is a third type of compliance which is
being used with increasing frequency. It is established
specifically for determinations of the directions and magni-
tudes of hydraulic gradients. Such a monitoring "point" is
comprised of a cluster of small diameter wells that have
very short screened intervals (e.g., piezometers), and is
usually located just outside the perimeter of the plume.
Water level measurements obtained from wells that are
screened over comparable stratigraphic intervals are used to
prepare detailed contour maps from which the directions
and magnitudes of local hydraulic gradients can be approxi-
mated. An evaluation of the vertical gradients can be made
in a similar fashion by comparing water level measurements
from adjacent wells screened at different depths. It is
usually just as important to determine the vertical hydraulic
gradients as the horizontal hydraulic gradients, because a
remediation wellfield may control only the uppermost
portions of a contaminant plume (e.g., when remediation
wells are too shallow or have insufficient flow rates).
Less often utilized than any of the foregoing are within-
plume monitoring points. Most of these are installed during
the site investigation phase, prior to the remediation. These
wells are needed to monitor the progress of remediation.
These can be subdivided into on-site plume monitoring
points, located within the property boundary of the facility
that contains the source of the contaminant plume, and off-
site plume monitoring points that are located beyond the
facility boundary.
2. Monitoring Criteria for Performance Evaluations
It is convenient to identify four categories of monitoring and
performance evaluation criteria: chemical, hydrodynamic,
treatment efficiency, and administrative control.
Although several federal statutes impose limitations on
chemical concentrations in fresh water, EPA's 1985 National
Contingency Plan (NCP) revision and the Superfund
Amendments and Reauthorization Act (SARA) of 1986
provide the concept of applicable or relevant and
appropriate requirements (ARARs) that derive from any and
all federal statutes. ARARs also effectively include
regional, state, and local statutes, because EPA is required to
consider these (which a state may actively encourage by
withholding its concurrence signature in the site settlement).
Chemical monitoring criteria may be based on a
combination of:
(i) Maximum Contaminant Levels (MCLs) that
have been promulgated for drinking water
supplies under the Safe Drinking Water Act
of 1974, as amended by the SDWA
Amendments of 1986;
(ii) Alternate Concentration Limits (ACLs) that
are health/risk-driven and technology-driven
are site-specific to ground-water corrective
action plans (CAPs) under the Resource
Municipal
Wellfield
Potential
Receptors
Hospital
Backup Wei I
© Extraction Wells
fa Gradient Control Points
® Injection Wells
(M) Monitoring Wells
A Off-site Background
Quality Wells
o
o
. (M) (M)
(MJ ;
Bldg
2
Bldg
1
Sou
— — ^
Property Boundary
Ground Water
Flow Direction
Figure 4. Possible monitoring point locations in a ground-water contamination plume undergoing pump-and-treat remediation.
-------
Conservation and Recovery Act of 1976, as
amended by the Hazardous Waste
Amendments Act of 1984;
(iii) State Water Quality Standards that derive
from the Clean Water Act of 1972 and other
statutes that protect aquatic wildlife and their
habitat; and
(iv) county, municipal, or regional restrictions on
the quality of water that may be discharged to
local lands, or other requirements for the
return of pumped water in water-short areas.
The actual numerical levels of chemical oversight
monitoring criteria may be affected by detection limits, and
natural water quality. For example, the acceptable
concentration level for a given contaminant in health-based
terms may actually be much lower than can be physically
determined with routine instrumentation, or the natural
levels of certain contaminants (e.g., arsenic or selenium) are
greater than ARARs.
Hydrodynamic monitoring criteria include:
(i) prevention of infiltration through the unsaturated
zone;
(ii) maintenance of an inward hydraulic gradient at the
boundary of a plume of ground-water
contamination; and
(iii) providing minimum flows in streams or wetlands.
Vertical profiles of moisture content may be used to
determine whether there is effective prevention of
infiltration through the unsaturated zone (e.g., below a site
cap). The requirement that an inward hydraulic gradient be
maintained at the periphery of a contaminant plume
undergoing pump-and-treat remediation is imposed to
ensure that no portion of the plume migrates away from the
zone of remedial action.
Evaluations of the hydraulic gradients exerted by
remediation wellfields require many measurements. To
assess this performance adequately, the hydraulic gradients
must be measured accurately in three dimensions between
each pair of adjacent pumping or injection wells. Velocity
divides may occur in the horizontal plane near active wells
and may develop in the vertical dimension if the well
hydraulically influences only the upper portion of the
aquifer. The ground-water velocity is essentially zero at the
stagnation point which forms downgradient of pumping
wells.
The performance standards that are applied to the treatment
plant portion of a pump-and-treat remediation are most
often phrased in terms of treatment efficiency criteria.
Although simple comparisons of total organic chemical
loading in the influent and the effluent of the treatment train
give some idea as to overall treatment efficiency, it is
important to make a detailed examination of the influent and
effluent to each treatment unit.
In some cases, individual compounds present limitations to
the treatment process because of effluent quality standards
for discharging to a local stream (e.g., National Pollutant
Discharge Elimination System (NPDES) permit
requirements). In other cases, after some time of pumping
VOC-laden water for treatment solely by air-stripping, non-
volatiles may begin to arrive in the influent and require the
addition of carbon filtration to the treatment train. In still
other cases, there is a need for examination of key locations
within the treatment process to troubleshoot the effects of
certain contaminants on treatment units. Iron and calcium,
for example, may precipitate in an air stripper and
dramatically decrease throughput.
Administrative control monitoring criteria may be codified
governmental rules and regulations (e.g., for fire safety and
electrical hazards), but they also include:
(i) effective implementation/enforcement of drilling
bans and other access-limiting administrative
orders;
(ii) proof of maintenance of site security; and
(iii) reporting requirements, such as frequency and
character of operational monitoring reports.
Combinations of chemical, hydrodynamic, treatment
efficiency and administrative control monitoring criteria are
generally stipulated at monitoring points. The exact
combination for a specific monitoring point depends on its
location relative to the source of contamination and the
potential receptors.
3. Strategies for Monitoring Locations and Criteria
Each kind of monitoring point has a specific and distinct
role to play in evaluating the progress of a remediation. In
choosing chemical monitoring criteria, it is essential to
recognize the interdependency of the criteria at different
monitoring locations.
The monitoring point locations and criteria should be
selected initially on the basis of a detailed site
characterization, from which transport pathways prior to
remediation are identifiable, and from which the probable
transport pathways during remediation may be predicted. It
may be useful to complete pre-design studies and pilot
-------
studies of remediations prior to selection of certain
monitoring point locations and criteria to allow consider-
ation of the effects of system operation on flow and
transport pathways.
Specifically, monitoring point locations and criteria should
be keyed not only to the no-action transport pathways, but
also to the probable transport pathways during routine
operation and under partial operational failures. Generally,
complete failure may be presumed to be the same as no-
action, though this may not be true if the pump-and-treat
remediation is augmented by other technology such as
subsurface barrier walls. Since the chemical composition
and flow pattern of the zone of remedial action may change
substantially as the remediation progresses, the monitoring
point locations and criteria may also need to change.
Assuming that a remediation wellfield is operated with
constant system configuration and pumping schedule, it will
still generate flow velocities that are highly variable
throughout the zone of remedial action. In part, this is due
to the heterogeneous nature of the subsurface sediments
(e.g., hydraulic conductivity may vary considerably over
short distances). This may also occur because each
pumping well causes local velocity variations.
There may be a need to relocate or add extraction or
injection wells to better remediate portions of the interior
core of the contaminant plume. These kinds of system
modifications underscore the need to anticipate changes in
chemical patterns at monitoring locations, by taking into
account the detailed flowlines generated by the wellfield,
and the changes in contaminant concentrations that might
result from variations in the influences of transport
processes along those flowlines.
One way to organize studies of possible monitoring point
locations is by mapping flowlines generated by remediation
wellfields in sufficient detail to characterize the capture
zones generated by extraction wells and mounding zones
produced by injection wells. For each remediation well, the
location of a stagnation point and the maximum width of the
capture zone should be identified. A different flowline plot
may be required for each change in key variables (e.g., well
flowrates and locations, natural hydraulic gradient, etc.).
Provided that the estimates of key parameters are reasonably
accurate, such flowline illustrations should form a sound
starting point for selections of monitoring point locations. It
is important to ensure that external stresses (e.g., local
wells) are accounted for when mapping flowlines.
4. Measures of Operational Effectiveness
At the heart of a remediation performance evaluation is the
need to evaluate its operational effectiveness, which may be
measured in a number of ways. One may view it as the
general degree of hydrodynamic control exerted and/or the
general degree of contamination clean-up achieved. To
quantify the general degree of hydrodynamic control
exerted, one may compute the mathematical average of
several estimates of the horizontal and vertical gradients
along the outer bounds of the contaminant plume. The
average value of the hydraulic gradients should be inward,
toward the core of the plume. To quantify the general
degree of contamination clean-up achieved, one may
compute the average value of the total contaminant loads
(e.g., total VOCs) at all monitoring locations and compare
this computed average with previous values.
The total contaminant mass present at any monitoring point
may be estimated by analyses of chemical extracts of
subsurface sediment samples. The physicochemical basis of
this estimation procedure is that chemical extraction
techniques recover all of the contaminant mass present in
the sediment sample, regardless of phase. Obviously, the
sampling technique, spatial coverage of sampling efforts,
and extraction efficiencies are quality control considerations
that must be considered for this technique to generate
reliable estimates. Under ideal hydrogeologic conditions
(e.g., homogeneous and isotropic sediments) the estimated
values should trend downward over time if the remediation
progresses satisfactorily.
However, there may be significant limitations to such
contaminant mass estimation methods. The degree of
heterogeneity of subsurface sediments will strongly
influence estimate reliability. Slight variations in vertical
and horizontal hydraulic conductivity may significantly
affect the migration of both aqueous phase and immiscible
phase contaminants. Furthermore, it is generally accepted
that the adsorptive capacity of sediments for organic
contaminants increases with decreasing grain size. As a
result, contaminant mass estimations using sediment
extraction techniques may only provide general indications
of the effectiveness of a remediation. As the degree of
hydrogeologic complexity increases, the likelihood of
obtaining replicable sampling and analytical results
decreases.
The total contaminant mass present at a monitoring location
has often been estimated from contaminant concentrations
in ground-water samples. This approach employs
adsorption isotherms to infer the concentration of
contaminant adsorbed by subsurface sediments from the
measured contaminant concentrations in ground-water
samples. Unfortunately, this approach may seriously
underestimate the mass present, because the assumptions on
which the method is based may be easily violated.
The total contaminant mass present at monitoring locations
has also been represented by the direct sum of the
10
-------
concentration values of the contaminants found in ground-
water samples. However, this is a near-meaningless
approach for many pump-and-treat remediations, primarily
because remediation wellfields cause increased flow
velocities locally and induce invasion of the plume by
uncontaminated water beneath and adjacent to the plume.
These and other factors serve to depress contaminant
concentrations in the ground-water, at least until pumping
ceases and chemical equilibrium can be re-established.
The general degree of hydrodynamic control or clean-up, as
represented by averaged values of hydraulic gradients or
total contaminant mass, is an informative and useful
measure of operational effectiveness, but it does not address
the variations in degree of control or clean-up. Therefore, a
second way one may view the operational effectiveness of a
pump-and-treat remediation is with regard to the spatial
uniformity of the control exerted and/or the clean-up level
achieved. In this context, hydrodynamic control should be
evident along the periphery of the plume at each location
where hydraulic gradients are determined from water-level
elevation measurements. Likewise, the total contaminant
load at each within-plume monitoring point should not vary
significantly from the average.
Operational effectiveness may also be viewed in terms of
the persistence of the desired effects of the remediation. Is
hydrodynamic control of a contaminant plume maintained
continuously, regardless of seasonal variations in the
recharge and flow rates of the regional system? Do
concentration profiles remain stable or are they affected by
residual contaminants as the water table rises and falls
throughout the year? Are concentration profiles affected by
increased infiltration due to aging, settling and slumping of
materials beneath an on-site cap? Any of the foregoing may
cause the desired effects of a pump-and-treat remediation to
be diminished over time. Persistence of desired effects is a
very real concern when evaluating operational effectiveness.
5. Measures of Operational Efficiency
While operational effectiveness may be the primary focus of
regulatory entities, operational efficiency is more often the
focus of those who bear the financial responsibility for the
remediation. There are a number of ways to measure
operational efficiency, and some are germane to both
regulatory and fiscal interests.
The minimization of total costs, while reaching and
maintaining the remediation targets, is the most
economically-oriented view of operational efficiency. Total
costs can be tracked on a quarterly or annual basis to
provide a relatively simple picture but, the operation of the
system may affect this picture positively or negatively. For
example, reduced downtime and stabilized operations
usually follow the initial shakedown period of operation;
automation of some manual operations may be possible as
experience is gained with the system or as technology
advances. The total costs per year may decline continuously
as the remediation progresses. However, the costs may also
be erratic, because of changes in the specific pattern of
contaminant removal. For example, portions of the plume
may not respond to remedial actions as predicted,
particularly if any previously undelineated sources are
present (e.g., NAPLs). Such sources may serve as long-
term reservoirs and continue to release contaminants for a
very long time (U.S. EPA, 1989a, Mercer and others, 1990
and Huling and Weaver, 1991).
Unfortunately, the tracking of total costs also can be a
misleading measure of operational efficiency unless costs
are normalized for the effects of inflation, interest rates, and
depreciation of recoverable assets. Very real trends of
improved operational efficiency may be completely hidden
until costs are normalized. This can be a confusing task
because such factors may vary substantially from region to
region, and from small to large organizations.
A less complicated means of evaluating the operational
efficiency of a pump-and-treat remediation is to look for
maximization of contaminant removal per unit volume of
pumped and treated ground water. This approach focuses
on the mechanical efficiency, from which bottom-line
economics may be inferred without regard to market forces.
It is generally acknowledged that costs are inversely
proportional to the contaminant levels of the treated water
(e.g., it normally costs less on a per pound basis to remove
contaminants from a concentrated wastestream than from a
dilute wastestream). Hence, to evaluate the operational
efficiency by this approach, the concentration levels of the
contaminants in the pumped waters must be tracked. Under
idealized conditions the management of a remediation
system should be such that the most contaminated
wastestream is continuously produced. This concentration
would gradually decrease as the aqueous phase
contaminants are depleted from the subsurface. However,
the presence of residual contaminants adsorbed to aquifer
materials and/or immiscible contaminants (e.g., NAPLs)
may result in elevated contaminant concentrations for very
long periods of time.
One consequence of maximizing the wastestream
concentrations is that the minimum volume of contaminated
water may be produced over the lifetime of the remediation.
This is of interest particularly to regulators where ground
water is in short supply. The total volume of pumped and
treated ground water may therefore be a measure of
operational efficiency. Naturally, there are economic
attractions to minimizing the total volume of pumped and
treated ground water. It is generally accepted that costs are
11
-------
directly proportional to the amount of material handled for a
given contaminant concentration level. It is also clear that the
electrical and maintenance costs of extraction and injection
well pumps depend directly on the volume of ground water
handled.
A concurrent benefit of minimizing the total volume of
pumped and treated ground water is that the time for
completion of the remediation may be minimized. This is of
interest to both regulators and the financial backers of
remediations, since they share sensitivity to public perceptions
of progress at the site. The time for completion of major
phases of the remediation may also be considered a measure
of operational efficiency. The obvious regulatory benefits of
minimization of time for completion of remediation are the
abbreviated costs of oversight efforts, and enhanced credibility
with the public. The economic benefits to the financial
backers of remediations may include reduced labor and
operational costs, and increased value of salvageable assets.
6. Strategies for Determination of Success/Closure
Measures of the operational effectiveness and efficiency of
a pump-and-treat remediation can be used to monitor the
progress of the remediation. Endpoints must be selected to
bring the remediation to termination. Such endpoints may
involve absolute or relative measures of success and may
include statistical considerations.
Absolute measures of the desirability of terminating the
operational phase of a pump-and-treat remediation can be as
simple as the removal of a specific mass of contaminants, such
as recovery of the total estimated volume of free product in
gasoline-contaminated sediments. Conversely, absolute
measures may be such things as achieving specific
contaminant concentration levels (e.g., MCLs). The former is
much less likely to be accepted by regulators than the latter,
primarily because it is often extremely difficult to accurately
estimate the total mass of contaminant present in the
subsurface. Consequently, pre-operational decisions to
remove only a specific mass or volume run the risk of
terminating the remediation without significant reduction of
risk to human health and the environment. Absolute measures
are more likely to meet opposition by those responsible for the
operation of the remediation because of the difficulties
associated with ensuring clean-up to specific levels. A similar
concern may be expressed by the regulators, since it is difficult
to foresee what technological innovations may develop during
the years of remedial operations. Fortunately, absolute
measures need not be chosen as stand-alone endpoints;
contingencies for demonstrated limitations or technological
innovations can be integrated into the phrasing of absolute
measures.
Relative measures of the success of remedial operations
specify that specific percentage reductions of contaminant
concentrations be achieved. These have not been used as
stand-alone criteria for termination of remediations, but are
often implemented as requirements such that the treatment
process operates continuously at a known efficiency of
contaminant removal. There is merit to considering relative
measures for a contaminant plume, since specific percentage
reductions of contaminant concentrations can be translated
into risk reductions to human health and the environment.
For all practical purposes, however, percentage reductions
that are applied to known pre-operational contaminant
concentration levels are equivalent to absolute measures
since the resulting concentrations are readily computed.
One major incentive for phrasing clean-up targets as
specified percentage reductions from pre-operational
concentration levels is the public relations benefits. Most
individuals can readily grasp the idea that 99.99% of the
contaminants were removed, whereas far fewer are schooled
in what specific concentration levels may mean.
Regardless of whether absolute or relative measures of the
success of a pump-and-treat remediation are employed, it is
clear that a single round of water-level measurements or
samples will not suffice to justify termination of a pump-
and-treat remediation. At a minimum, successive sampling
on a monthly or quarterly basis is needed. The results are
examined to determine whether or not the remediation has
truly reached the specified targets. Such examinations of
the datasets inevitably reveal that the data values vary from
one sampling/measurement event to the next. While it is
possible to view these data value variations rigidly, such that
each value from each sampling/measurement event must be
at or below the target concentration level, this approach may
be unduly restrictive and may not be justified statistically.
The evaluation of pre-termination datasets with statistical
measures entails plotting the contaminant concentration
values versus time and fitting a multiple/nonlinear
regression line to the plotted values (Figure 5). Two kinds
of outcomes of the regression line fitting are acceptable
initial indications that the remediation may be ready for
termination. The first being that the contaminant
concentrations have leveled off and the regression line has a
slope equal to zero at an average contaminant concentration
value that is at or below the target concentration. The
second being that the regression line has a negative slope
and concentrations are below the target concentration.
The slope of a fitted regression line is an insufficient
statistical measure by itself, since data values inevitably
vary about this line. A Student's t-test should be performed
to ascertain that the slope of the fitted regression line is not
greater than zero at a 95% level of confidence. Quality
control considerations also indicate that the vast majority of
the data values should fall within a 95% confidence interval
12
-------
200
1981
1985
1987
1989
Figure 5. Concentration vs. time data plot and multiple regression line
with confidence bounds for judging persistence of the
clean-up.
about the regression line. Regulators are concerned only
with the upper 95% confidence interval boundary for most
parameters.
Following termination of the remediation, the sampling/
measurement events should be continued to assure
persistence of the effects of the remediation. A declining
frequency of sampling/measurement events is most
appropriate for this approach. For example, monthly
sampling events for the first three or four months
immediately following termination may be adequate. This
should be followed by quarterly sampling events until the
end of the second post-operational year, and annual
sampling events until the end of the post-operational
regulatory period. Data obtained during these sampling
events should be used to update regression plots. Re-
activation of the remedial actions and/or re-evaluation of the
appropriateness of the remedy may be indicated when the
slope is no longer statistically at or below zero, or when two
or more successive data values are above the upper 95%
confidence interval bound of the regression line.
It may seem desirable to base the decisions to terminate or
re-activate a pump-and-treat remediation on examination of
a single regression plot that represents all monitoring point
locations. Technically, it may be possible to justify the
grouping of results from monitoring point locations of a
certain kind. For example, on-site monitoring point
locations may fall into this category. At downgradient
background monitoring point locations, however, grouping
of results may have undesirable sociopolitical consequences.
The idea that one potential receptor is better protected than
another is generally unacceptable—although it may be
acceptable in some cases if the natures of the potential
receptors differ substantially (such as a water-supply
wellfield versus an ecological habitat).
In those cases where grouping of like locations is
determined to be acceptable, there arises the question of
how to do the grouping: should individual regression plots
be made and their slopes averaged, or should a single
regression line be fitted to a dataset containing the values
from all of the monitoring point locations belonging to the
group? The former option is equivalent to the latter option
only when the population distribution forms of the datasets
from the individual wells have comparable statistical
behavior (e.g., variance and skewness) and are wholly
independent of each other.
If the slopes of individual regression plots are averaged
from datasets of different monitoring points, the computed
95% confidence interval will be largely independent and
insensitive to actual variations of the data in each of the
datasets. If a regression line is fitted to a super-dataset that
contains data from all monitoring locations, the computed
95% confidence interval will be dependent on the actual
variations of the data in all contributing datasets. The
confidence interval computed from the super-dataset may be
larger than the confidence interval computed by averaging
the slopes of individual regression plots. As a result,
regulators may prefer the former option (smaller 95%
confidence interval), and responsible parties may prefer the
latter option (larger 95% confidence interval). Neither is
more or less correct statistically.
The foregoing discussion presumes that a decision has been
made at each monitoring point regarding whether or not to
prepare regression plots of contaminant concentration levels
versus time for each contaminant, for groups of
contaminants, or for the total contaminant load. In
addressing this point, the methods for computing values of
95% confidence intervals may be considered in a manner
that is parallel to the foregoing discussion. One may either
compute the 95% confidence interval of measurements of
each contaminant at a given monitoring point location and
then average these results or, combine several (perhaps all)
data values into a single dataset and then compute its 95%
confidence interval. These computations must be based on
normalized data values (e.g., the results of dividing the raw
data values of each contaminant by its average value) or the
results will be biased toward the contaminant datasets with
the highest concentration levels. Again, regulators may
prefer the former option (a smaller 95% confidence interval
results), and responsible parties may prefer the latter option
(a larger 95% confidence interval results). The grouping of
the datasets from several contaminants at a given
monitoring point location may be desirable only if justified
13
-------
by similarities in contaminant transport behaviors and the
health risks posed by the contaminants to be grouped.
Data Collection Considerations
Efforts made to characterize ground-water contamination
problems for remediation have traditionally focussed on the
quantification of contaminant levels rather than the potential
pathways and mechanisms that affect transport (Bear and
Verruijt, 1987; Boutwell and others, 1985; Faust and others,
1981; Keely and others, 1986; Konikow, 1986; Osiensky
and others, 1984), however, this bias is diminishing because
of a growing recognition of the limited usefulness of
chemical samplings of monitoring wells. It is clear that
periodic sampling can produce successive snapshots of the
distribution of contaminants at a site, but it is equally clear
that one can rarely expect to infer key contaminant transport
features and parameters from such snapshots. An approach
to site characterization is emerging wherein greater
emphasis is given to detailed characterization of transport
pathways and mechanisms, while retaining a commitment to
definition of the full extents and compositions of
contaminant plumes (Bouchard and others, 1989 and 1988;
Curtis and others, 1986; de Marsily, 1986; Goltz and
Roberts, 1986; Jorgensen and others, 1982; Mackay and
others, 1986; Major and others, 1988; Matheron and de
Marsily, 1980; Molz and others, 1986 and 1988; Scanlon,
1989; Usonoff and Guzman-Guzman, 1989).
1. Purposes of Data Collection
The objectives of collecting data are determined in large
part by the intended use (Boutwell and others, 1985; Davis,
1986; Huyakorn, 1984; Krabbenhoft and Anderson, 1986;
Loftis and others, 1987; Mercer and others, 1983; Ophori
and Toth, 1989). If data are to be used to assess the general
potential for contaminant releases and transport at a newly
discovered site, the objectives will be limited to problem
screening efforts for the time being. Based on the outcome
of the screening process, further data collection events may
be planned. This is an iterative process, which usually calls
for increasingly more detailed data collection which
culminates in a comprehensive characterization of the
physical extent and chemical makeup of the contamination
problem. Traditionally, the identification of transport
pathways has been limited to an understanding of the natural
flow system in such studies.
An understanding of contaminant migration pathways is
essential to the selection and design of an appropriate
remedy. Unfortunately, the objectives of past site
characterization studies have rarely included sufficiently
detailed testing for hydraulic parameters and have often
failed to identify key/limiting hydrogeologic processes or
features. As a result, the remedy selection process has often
proceeded without the benefit of information vital to the
proper evaluation of the success of various remedial actions.
Instead, such details were often addressed during the design
phase following the remedy selection, if at all.
With the benefit of hindsight and an acknowledgement of
the growing number of reported ineffective remediations, it
is clear that the site screening and site characterization
studies that precede remedy selection must include remedial
design parameters as primary data collection objectives.
This means that detailed hydraulic testing, bench treatability
studies, and pilot operations of individual extraction wells
must begin during the site characterization studies.
Fortunately, this has been recently acknowledged and since
1989 the EPA Regional Offices have supported this
approach (Cannon, 1989).
During the process of remedial design, it is essential to
formulate data objectives for the operational, closure, and
post-closure phases of the site work. While nominally
needed for regulatory compliance, data should also be
tailored to provide key information for active management
of the remediation (e.g., individual flows and chemical loads
of all extraction and injection wells, detailed water-level
measurements, treatment unit efficiencies). Data collections
for regulatory compliance purposes may include samplings
from upgradient and downgradient public supply wells,
private wells, and industrial supply wells, and from
monitoring wells placed close to particularly sensitive
ecological habitats.
The objectives of data collections for regulatory compliance
purposes extend beyond system effectiveness and efficiency
as operational considerations, to the driving force behind all
of the site efforts—protection of human health and the
environment. In regulatory terms, the effectiveness of the
remediation matters because of the potential that each
exposure to contaminants poses added, cumulative risk to
society. Likewise, the efficiency of the remediation matters
because the potential duration of exposure is limited
ultimately by the duration of the remediation. Efficiency
also matters because of the limited funds and human
resources that are available to conduct such work.
2. Relevant Scale(s) of the Problem
Perhaps the most subjective influence on data collection
strategies relates to the size of the problem. History
suggests that the larger the size of a contamination problem,
the more likely the problem will be addressed at a minimal
level in screening and site characterization efforts.
From a statistical viewpoint, the scale of the problem can be
addressed by breaking it down into definable regions
(Davis, 1986; Huntsberger and Billingsley, 1977;
14
-------
Mendenhall, 1968; Steinhorst and Williams, 1985; Taylor,
1987; Ward and others, 1987). For example, the nominal
plume may encompass an area of a few hundred acres and
an average depth of a few tens of feet. A relatively dense
monitoring network is most appropriate within the plume
with each monitoring location consisting of a single well or
a cluster of monitoring wells. Each monitoring well is thus
constructed to sample a different specific zone and allow
vertical and horizontal delineation of the contaminant
plume.
By contrast, the areas that may be impacted if remedial
actions are not implemented may be many hundreds or
thousands of feet downgradient from the contaminant
source, and may encompass hundreds or thousands of acres.
The major data collection objective within the zone between
the contaminant plume and potential receptors is to provide
an early warning of failure of the remediation to contain the
plume. Consequently, it is important to have an
understanding of where the plume may migrate if it escapes
the remedial action zone. These data needs may decrease
with increasing distances down-gradient from the plume.
The monitoring well and hydrogeologic testing locations
should be skewed toward a dense pattern immediately
downgradient of the plume and a sparse pattern farther
downgradient.
It is important to understand the connection between the
local hydrogeologic setting and the regional flow system in
studies of subsurface contaminant transport (Mackay and
others, 1986; Osiensky and others, 1984; Scanlon, 1989;
Ward and others, 1987). It is often necessary to thoroughly
investigate the regional trends in water levels to formulate
appropriate boundary conditions and water balances for a
model of the local flow system.
3. Actual Scale of the Measurements
There are many ways to estimate the hydraulic conductivity
of subsurface materials (Cedergren, 1989; Corey, 1977;
Davis and DeWeist, 1966; Freeze and Cherry, 1979; Mercer
and others, 1982) and it is not uncommon for the various
methods to yield differing results when performed at the
same location. There are numerous reasons for this, but one
of the major factors relates to the scale of the measurements
(Anderson, 1979; Mercer and others, 1982). For example, it
has been shown that laboratory measurements of the
hydraulic conductivities of clay and silt samples may be as
much as two to four orders of magnitude lower than the
values determined by field tests and that such
determinations tend to have a much greater variability
(Anderson, 1979; Klotz and others, 1980; Mercer and
others, 1982; Molz and others, 1986a and 1988). Clay and
silt samples are often compacted during sampling operations
and laboratory analysis. Additionally, the sample volume
may be too small to intercept the macropores and other
preferential flow paths that conduct most of the flow
through the stratum.
Depending on the magnitude of hydraulic stress created, the
hydraulic conductivity values obtained from field tests may
vary considerably at the same location. Small-volume tests,
such as slug tests, affect only a few cubic feet of formation
materials surrounding the well screen. As a result, it is not
unusual for the hydraulic conductivity values obtained from
slug tests of closely spaced wells to differ by several orders
of magnitude. The results are highly sensitive to minor
variations in well construction details, especially filter pack,
screen position and grouting. The testing locations are
effectively isolated despite their proximity to one another.
This is underscored by the stochastic character of such
measurements. They often appear to be random and fully
independent in settings where trends in physiographic
features are clearly evident (Dagan, 1982a; de Marsily,
1986; Gelhar, 1986; Vomvoris and Gelhar, 1986).
Conversely, the hydraulic conductivity values obtained from
large-volume field methods, such as aquifer pumping tests
and two-well tracer tests, may affect thousands to millions
of cubic feet of subsurface materials. These field methods
often demonstrate considerable dependency and
autocorrelation. This is understandable given the potential
for physical overlap of the effective areas of testing with
large volume tests, but such character is observed even
when it is clear that overlaps do not occur. One reason is
that autocorrelation is endemic to the occurrence of
physiographic trends and the hydraulic forces that help to
shape strata. Dipping strata impose directional control on
natural flows, as do streambed sediments that are laid and
reworked into lenses parallel to the course of a river (Blatt
and others, 1980; Compton, 1962; Davis and DeWeist,
1966; Fetter, 1988; Freeze and Cherry, 1979; Heath, 1983;
Osiensky and others, 1984; Todd, 1980). These act as
intrinsic biases to flow direction, something that can be
quantified in terms of anisotropy. The large-scale continuity
of physiographic trends makes it highly probable that
hydraulic conductivity values determined by large-volume
field methods will autocorrelate.
While large-scale continuity is a hallmark of physiographic
features, it is by no means a hallmark of contaminant
distributions. Consequently, the questions of how to
interpolate between measured locations and how to
extrapolate contaminant trends in time emerge as a major
issue. All too often, these questions have been
misinterpreted to mean that sampling locations and
frequencies are the only concerns that merit attention. In
point of fact, however, the representativeness of each
sample is of equal or greater concern. An examination of
possible strategies for characterizing the reliability of
15
-------
chemical measurements is warranted, and begins with an
examination of sampling techniques.
Bailers or other low-volume sampling devices are
commonly used to obtain samples from monitoring wells for
organic chemical analyses. Discharging wells such as the
extraction wells in a remediation wellfield may also be
sampled, but for such wells the contaminant concentration
levels are often a function of the duration of pumping. This
fact may be exploited by the collection of successive
samples as pumping begins and continues (e.g., chemical
time-series sampling) with each successive sample
representing ground water from a part of the aquifer farther
from the extraction well. (Keely, 1982). The overall effect
is similar to increasing the number of point-sampled
monitoring wells. The volume of the sample is important,
because point samples are tacitly presumed to represent the
average chemical quality of fairly large volumes of the
aquifer from which they were withdrawn; such as occurs in
numerical modeling efforts, where grid blocks of several
thousand cubic feet are assigned average concentration
values.
The use of chemical time-series sampling is quite similar to
the use of tracer tests to gain insights into preferential paths
of contaminant movement (Keely, 1982). Ground-water
tracers are defined as matter or energy transported by water
which will give information concerning the direction and
velocity of ground water and potential contaminants. Since
the time of release and the input concentration of intentional
tracers are known but the inputs are unknown for
contaminant studies, the interpretation of data from
chemical time-series sampling might initially seem to be
severely hampered. However, it is the pattern of relative
concentrations over time which is actually examined in
either case. The absolute values are neither required nor are
they readily obtainable under all but the very best controlled
field tests in relatively homogenous settings. There are
often too many avenues for tracer losses (sorption, ion-
exchange, biodegradation, etc.) to allow for rigorous mass
balances over reasonable distances.
Regardless of whether chemical time-series sampling results
in a pattern of chemical arrival that show definite trends or
are completely random, the data collected are useful to the
investigator in terms of the statistical uncertainties of
contaminant concentrations in the plume being defined.
Such data can be helpful in deciding when a model has been
adequately calibrated and when it is predicting ground-water
flow or contaminant transport with reasonable accuracy.
They can minimize the amount of adjusting of key
hydrogeologic parameters that often accompanies efforts to
calibrate a model to field data. Chemical time-series
sampling results may be used directly in modeling efforts to
generate estimates of uncertainties in the model results.
However, care must be taken to understand the
autocorrelation of such data, because successive samples are
not completely independent since they represent waters that
have passed through the zones occupied by previous
samples. The ground-water contaminant concentrations will
be affected by interactions with sediments and residual
contaminants present in those zones.
The simplest approach to predicting the arrival of
contaminants at a pumping well from a selected distance
involves the relationship of the total volume of water
pumped to the distance traveled. Assuming radial flow to a
well, the distance waters travel during specific times are
indicated by concentric circles. Obviously, samples
collected after several hours of pumping represent an
average water quality of a larger volume of the aquifer than
samples collected only a few minutes after the onset of
pumping.
4. Data Quality vs. Quantity
Regardless of the data objectives, there eventually comes a
point in planning where one faces a trade-off between
quality and quantity. In ground-water investigations, the
descent from high to poor quality data is easily made. It is
extremely difficult to plan for data of moderate quality—
sophisticated field methods are either performed properly or
they are not. The economics of bad data collections are
painfully obvious. Not only can the magnitude and extent
of the contamination problem be wholly misunderstood, but
the selected remedy can be so off target as to actually
worsen the problem. However, there are many field
methods which produce good relative information that may
prove valuable to site characterization. The usefulness of
such methods and their results should not be
underestimated.
Hydrogeologic Data
1. Topographic and Geographic Data
Topographic and geographic data are most commonly
available as elevation contour maps. These are available
from the U.S. Geological Survey (USGS), the U.S. Forest
Service, the U.S. Army Corps of Engineers, and various
State agencies. At the local/site scale, subdivision plats and
city planning maps provide contour spacings as small as
0.25 feet. Aerial photographs of many cities are available
from several government agencies at five to ten year
intervals starting in the late 1940's. Remote sensing images
are also available from various government agencies. Such
maps and photos are useful for gaining an understanding of
the physiographical trends and structural controls on
ground-water movement.
16
-------
It is essential to measure all depths to water against a
common datum and most topographic maps are not accurate
enough for this purpose. Differences of only a few feet may
be comparable to the seasonal changes in water levels at
many sites. The elevations of the tops of the monitoring
wells must be surveyed to an accuracy of better than 0.1
feet. Otherwise, the potential for misinterpretation of the
hydraulic gradient and direction of ground-water flow may
be significant. Surveyed elevations should be checked
periodically because of the potential for slumping and
settlement of capped waste and compaction of pumped
strata.
2. Geomorphologic and Geologic Data
Maps which describe the geomorphologic character of large
tracts of land are available from the USGS and State
geologic agencies. Specifically, it is common for basin
boundaries and structural features to be illustrated on maps
that cover an entire state or physiographic province. Fault
lines and fracture trace orientations are often detailed on
such maps, but these may have little direct applicability at
the local or site scale. Although geologic maps which
describe regional stratigraphic and lithologic features are
available, it is essential to have an experienced geologist
detail the local setting. Principally, the geologist will rely
on inspection of local exposures, drill cores, and
excavations. Geophysical techniques, particularly the
borehole methods, can be of great assistance to the geologist
in evaluating the nature and extent of specific strata, as well
as the general water quality.
3. Flow Rates
Rainfall events, streambed losses, and snowmelt provide the
majority of natural recharge to the subsurface. Precipitation
records are kept by the National Oceanic and Atmospheric
Administration. Estimates of effective recharge are made
by the U.S. Department of Agriculture for crop purposes,
and by the USGS and State geological agencies for water
supply purposes. However, these estimates are generally
based on basinwide studies and do not provide much detail
at the site level. Aquifers also may be recharged purposely
by infiltration basins and injection wells. These artificial
recharge mechanisms are manmade and directly measurable.
Soil scientists have devised a number of methods for
measuring the movement of water through the unsaturated
zone. However, most of these methods are intended
primarily for use in relatively shallow surface soils. The
most traditional methods include tensiometers, gypsum
blocks and collection pans. State-of-the-art methods for
determining soil moisture content include the use of neutron
probes, pressure transducers, and time-domain-
reflectometry (TDR). These sophisticated methods are
subject to interference and produce data which must be
properly interpreted to quantify fluid movement.
The flow rates of springs can be measured with ordinary
surface-water weirs or flumes. Likewise, baseflow
recession curves can be used to estimate the discharge from
an aquifer to streams. Aquifer discharges can also be
approximated by using estimates of the hydraulic
conductivity and hydraulic gradient in Darcy's Law. The
cross-sectional area through which the aquifer is discharging
is also needed for this analysis. Errors in the cross-sectional
area are unlikely to be more than two- or three-fold. This is
negligible given the possible order(s)-of-magnitude
uncertainty that may be associated with hydraulic
conductivity estimates. Ground-water flownets have been
used traditionally to estimate aquifer flows, with reasonable
success at the basin level, but not necessarily at the local/site
level. However, flow nets can be useful for determining
ground-water flow direction, making time of travel
estimates and assessing ground-water vulnerability (U.S.
EPA, 1989b).
4. Hydraulic Parameters
There are several methods of estimating the hydraulic
conductivity of an aquifer. Hydraulic conductivity values
obtained from laboratory permeameter methods represent
point estimates only. Unconsolidated sands and gravels may
become disordered and compressed during extraction from
the field and emplacement in the permeameter columns,
which may significantly alter their hydraulic properties.
Clays may be extracted from the field with less disturbance,
but they often undergo compression during emplacement
into permeameter columns. Additionally, high hydraulic
gradients are required to induce flow through clay samples
in laboratory permeameter studies. From practical
experience, it is easy to see that laboratory estimates of the
hydraulic conductivities of aquifer materials may be
variable.
It may be difficult to take a sample that accurately
represents the amount of secondary permeability present in
the clay resulting from macropores or other preferential
flow paths. This lack of representativeness usually accounts
for the common observation that many clay units
determined to be impermeable from laboratory tests are
often characterized as relatively permeable units by field
tests. This takes on great importance in many settings
because estimates of the potential effectiveness of a
remediation often depend heavily on the properties of low
permeability strata.
Numerous small-volume field tests for determining
hydraulic conductivity have been described in the literature
(Bouwer and Rice, 1976; Papadopulos and others, 1973).
17
-------
Most methods involve displacement of a fixed volume (slug)
of water in a piezometer and relating the observed hydraulic
response to the properties of the aquifer in the immediate
vicinity of the piezometer. Storage coefficients obtained from
slug tests may not be representative of the aquifer due to the
small volume of aquifer being tested. This is important
because both the transmissivity and the storage coefficient are
needed for the well hydraulics computations used to size and
position wells for remedial actions.
Recent developments in electromagnetic and thermal borehole
flowmeter technology allow the determination of horizontal
hydraulic conductivity values as a function of vertical position
within an aquifer (Young and Waldrop, 1989; Hess and Paillet,
1989). However, the hydraulic conductivity, transmissivity
and storage coefficient values obtained at most hazardous
waste sites are estimated using various mathematical analyses
of aquifer pumping test data. The average saturated thickness
of that portion of the aquifer affected by the pump test must be
determined from drilling records or estimated by geophysical
methods. If the formation is very fine-grained, the capillary
fringe above the water table may be large enough to comprise
a non-negligible portion of the effective saturated thickness.
In such cases, continuous coring may be extremely valuable
in defining the magnitude of the capillary rise locally. If the
formation is coarse-grained, the capillary fringe will comprise
a negligible portion of the effective saturated thickness.
In addition to the hydraulic conductivity, the porosity and
hydraulic gradient must be estimated in order to compute the
average rate of movement of the ground water. Tracer tests
may be used in the field to estimate the effective porosity, but
this is possible only if the hydraulic conductivity has been
defined with very high accuracy. In practical terms, it may not
be possible to achieve such accuracy because the variations in
hydraulic conductivity about the mean value at a given site
may have a range of two or more orders of magnitude. By
contrast, effective porosity values for most aquifers lie
between ten and forty percent and may be estimated within a
factor of two.
Tracer tests can be conducted by injection and subsequent
withdrawal of a tracer solution in a single well through an
isolated injection zone. The extreme sensitivity of single-well
tracer tests to well construction details and the small volume
of aquifer affected limits the applicability of this method for
site characterization studies. However, two- well tracer tests
have been devised to avoid most of these problems. Two basic
kinds of two-well tests include natural gradient and forced
gradient tracer tests. Natural gradient tracer tests are
conducted releasing a small volume of concentrated tracer
solution in an upgradient well and monitoring the concen-
tration of the tracer over time in the downgradient well.
Forced gradient tracer tests involve injecting a dilute tracer
solution in one well, while the other well withdraws ground
water at an identical rate. While both natural gradient and
forced gradient tracer tests can ostensibly yield the same
information, they differ in their sensitivity to dispersive forces,
due to the radically different velocities that are involved.
Natural gradient tracer tests tend to be biased toward large
dispersion coefficients and forced gradient tracer tests tend to
be biased toward smaller dispersion coefficients. The
distinction is irrelevant for most pump-and-treat remediations,
however, because they generate such strongly advective flows
that dispersive forces are usually negligible by comparison.
Forced gradient tracer tests are intrinsically suited to provide
appropriate estimates of hydraulic parameters for pump-and-
treat remediations, because of the obvious similarities in
hydraulic and hydrodynamic behavior.
5. Fluid-Behavior Data
Several important fluid-behavioral properties include
miscibility, density, solubility, and viscosity. Miscibility is the
property that describes the ability of one fluid to dissolve into
or completely commingle with another fluid. Miscible fluids
do not exhibit stratification or separation of distinct phases.
Familiar examples of organic solvents that are miscible in
water include ethanol, ethylene glycol, and acetone.
Immiscible fluids, or NAPLs, exhibit separation of fluids at a
distinct interface. Familiar examples of immiscible fluids are
gasoline, creosote, trichloroethylene (TCE) and ether.
Dense immiscible fluids (DNAPLs) exhibit fingering during
advancement, as well as loss of residuals by snap-off in small
pore-throats and trapping in dead-end pores. Because of the
existence of a wetting phase (water) on subsurface grain
surfaces prior to percolation of spills or subsurface releases
from pipelines and tanks, the penetration of TCE and other
DNAPLs into intergranular pores is highly sensitive to a
critical effective pore size and the acceptable surface tension
that results at the interface between the two fluids for that pore
size. This description presumes that the water film on the
grains is at residual levels after complete gravity drainage; if
water is present beyond that needed to satisfy the hygroscopic
force of the wetting tension, it will be present as a pendular
ring at the bottom face of the grain, and may coalesce with
heavy films from other grains and thereby gain sufficient mass
so that the acceleration of gravity overcomes the total wetting
tension and downward flow occurs. Because of the orders-of-
magnitude range of variability of effective pore sizes in most
settings, dense immiscible fluids such as TCE penetrate non-
uniformly.
6. Fluid Levels and Pressures
Some of the most important measurements taken at ground-
water contamination sites are water level elevations as
hydrogeologic features and hydraulic parameters can often
18
-------
be estimated with such data. Many sites are located in
highly active hydrologic and hydraulic settings where large
daily water level fluctuations are possible, due to the
influence of local streams and industrial and municipal
wells, but hourly variations may occur also. Such variations
are not uniform regionally or locally, and tend to die out
logarithmically away from the source of the disturbance. It
is essential to collect consistent data, so that differences in
water level elevations recorded at various wells represent
the spatial distribution of water levels and not temporal
variations.
Wells of comparable construction (e.g., screen length) must
be used. Small diameter wells with short screens are the
most suitable for providing detailed information on
horizontal and vertical gradients. Clusters of such wells are
effective and efficient for delineating horizontal and vertical
gradients.
In terms of hydrostatic position, it is incorrect to group
water level measurements from wells whose screened
intervals occur at greatly differing elevations. This is
especially important at sites with significant vertical
hydraulic gradients. Otherwise, contour maps drawn from
the resulting data will present a picture that is an indefinable
mixture of the horizontal and vertical components of
hydraulic gradient. It may be important to measure water
pressure rather than water level elevations, especially where
there is the possibility of the contaminant plume being more
dense than the native formation water (e.g., plumes with
total dissolved solids in excess of 1,000 parts per million).
Water supply wells, irrigation wells, and industrial supply
wells are usually constructed with lengthy screens and often
without annular seals above the filter pack. Consequently,
measurements of their static water level elevations are
integrated averages of vertical head differences over a large
section of the aquifer. Such measurements should not be
combined with measurements from observation wells or
monitoring wells to produce contour maps.
Water level measurements obtained from water supply wells
or other production wells during operation are greatly
affected by the pressure drop that occurs as ground water
passes through the entrance openings of the wellscreen. The
magnitude of the effect varies from well to well, over time,
and as a function of the flow rate. It is very difficult to
accurately estimate water level elevations immediately
outside wellscreens.
There are numerous methods for measuring water-level
elevations. The most common are chalked steel measuring
tapes, electrical probes and pressure transducers. Chalked
tapes have been used for decades because of their simplicity,
reliability, and accuracy (about 0.01 ft). The method is
extremely slow and most suitable for static measurements.
Electric water level probes usually come with crimped foot
or meter markers, but may become unreliable due to the
tendency of the wire to stretch over time. A steel measuring
tape should be used to measure the length of electrical wire
lowered into the well when practical.
Pressure transducers utilize very sensitive strain gauges and
amplification circuitry to measure changes in hydrostatic
pressure, which may be converted to water level elevations.
Most pressure transducers incorporate an internal
atmospheric venting tube that prevents pressure
measurements from being affected by barometric
fluctuations. Pressure transducer measurements are easily
automated, so that no re-positioning need be done to acquire
successive measurements, and extremely rapid
measurements (greater than one per second) are possible.
The acquired measurements are stored in data loggers that
often allow real-time analyses of the data.
Chemical and Geochemical Data
1. Natural Ground-Water Chemistry Data
The goal of most aquifer remediations is to return ground
water to its beneficial uses where practicable; therefore, it is
necessary to characterize the chemical and biological
composition of nearby uncontaminated wells to determine
natural ground-water quality. Samples from
uncontaminated wells should be obtained with the same
rigor and care given to samples collected from suspected
zones of contamination. At a minimum, samples should be
analyzed for major cations and anions to allow the
computation of cation/anion balances. Descriptive graphical
analyses methods utilize such data, and are time-proven
methods of characterizing natural water quality (Cheng,
1988; Freeze and Cherry, 1979; Loftis and others, 1987;
Novak and Eckstein, 1988; Ophori and Toth, 1989; Rogers,
1989; Scanlon, 1989).
2. Soil Chemistry Data
For complete characterization and transport evaluation,
native soils must be collected and analyzed for major ion
chemistry (Corey, 1977; Hillel, 1982; Mercado, 1985;
Mercer and others, 1983; Osiensky and others, 1988;
Rogers, 1989). In addition, the organic carbon content of
the soil should be determined because of the significant role
it plays in the retardation of organic contaminants
(Bouchard and others, 1989; Lee and others, 1988; Mackay
and others, 1986; Nkedi-Kizza and others, 1985, 1987 and
1989; Piwoni and Banerjee, 1989).
Soils and subsoils can be easily sampled from the sidewalk
of excavated pits, but deeper sediments must be obtained
19
-------
from cores obtained during drilling operations. Traditional
split-spoon and Shelby tube samplers may be used to obtain
core materials for use in chemical analyses, lithologic and
stratigraphic mapping, and laboratory batch studies. However,
special coring devices and modified laboratory glove boxes
purged with an inert gas may be used to retrieve cores used in
fate and transport evaluations, and microbiological studies
(Nkedi-Kizza and others, 1989; Scarf and others, 1981;
SiegristandMcCarty, 1987).
3. Contaminant Reaction Data
Contaminated soils are generally sampled similarly to
uncontaminated soils. However, air-tight sample containers
are necessary when sampling volatile organic contaminants.
Laboratory studies of contaminant transport behavior often
utilize soil columns packed with subsurface sediments (Bitton
andGerba, 1984; SiegristandMcCarty, 1987). Columns
packed with uncontaminated soil samples are used to
determine retardation coefficients for specific contaminants.
Column studies may also be used to obtain estimates of the
number of pore volumes of water that must be flushed through
the material to reduce contaminant concentrations to specific
levels.
Batch, or flask studies may be utilized to determine site
specific distribution coefficients for contaminants and
sediments. These data may be used to predict the effects of
retardation on the transport of contaminants through the
subsurface. The fraction of organic carbon (foe) present in
subsurface materials will strongly influence the mobility of
many contaminants.
Biological Data
Until recently, it was presumed that the subsurface was largely
devoid of bacteria. This was shown to be untrue in the early
1980's, when various microbial populations were found in
subsurface sediment samples (Scarf and others, 1981; Bitton
and Gerba, 1984). Since then, researchers have attempted to
identify which organisms might be responsible for the
degradation of specific organic chemical contaminants.
Presently, it is generally accepted that no single strain of
microbe is responsible; rather, consortiums of bacteria engage
in degradation activities concomitantly.
The population dynamics of subsurface biota depend heavily
on limiting nutrients (Bitton and Gerba, 1984; Bouwer and
Wright, 1988; Major and others, 1988; Molz and others, 1986;
Tinsley, 1979). The carbon atom skeleton of an organic
chemical contaminant represents a fuel source to biota, but
they cannot utilize it unless they are able to acquire sufficient
supporting nutrients, such as nitrogen and phosphate (Molz
and others, 1986; Roberts and others, 1989; Tinsley, 1979).
Threshold concentration effects have also been observed, so
that it is possible to have an adequate supply of nutrients but
still generate little degradation because of the low
concentration level of the fuel source, the contaminant
(Borden and others, 1989; Flathman and others, 1989; Jensen
and others, 1988; Roberts and others, 1989; Spain and others,
1989; Srinivasan and Mercer, 1988). Typically, total dissolved
carbon loads of less than 100 parts per billion represent an
inadequate source of energy to stimulate bioactivity.
Contaminants may be present in certain portions of a plume at
concentration levels that are toxic to the indigenous biota.
However, if the biota are able to survive and function
minimally, they may adapt to the contaminant and begin
degradation. Hence, the rate of transformation of specific
contaminants may change over time. Great interest has been
taken in the possibility of stimulating bio-activity by the
addition of nutrients and/or fuel sources (e.g., methanol or
ethanol) to subsurface sediments. Both laboratory and field
experiments are under way at a number of locations across the
nation to investigate this possibility (Bouwer and Wright,
1988; Major and others, 1988; Roberts and others, 1989; Spain
and others, 1989; Thomas and Ward, 1989). The laboratory
experiments are often conducted in flow columns.
Quality Assurance and Quality Control Data
There are a great number of texts available that describe
quality assurance and quality control methods (e.g., Davis,
1986; Huntsberger and Billingsley, 1977; Mendenhall, 1968;
SAS Institute, 1988c and 1989b). Quality control methods
presume that the data possess specific statistical properties and
yield results that must be narrowly interpreted. The
repeatability of a result from a measurement of any parameter
is termed the precision of the method of measurement. The
accuracy of a measurement of any parameter is the degree of
closeness of its value to the actual/real value. Accuracy and
precision do not adequately describe the appropriateness of a
measurement for its intended use. The degree of
representativeness is by far an overriding consideration. For
example, measurements of the concentration of a contaminant
in ground water removed from a monitoring well may be
obtained with extreme care, such that one concludes that they
are probably highly accurate, but there is no guarantee that
such measured values represent the contaminant levels 10 feet,
100 feet, or more away.
III. Methods and Protocols
Performance Evaluation Methods
The proper selection and use of computational, statistical,
graphical and other methods for performance evaluations of
pump-and-treat remediations are as important as the
20
-------
strategies being pursued. This is because the inferences that
are drawn and the perceptions that are generated depend to a
great extent on the manner in which the data are processed
and presented. For detailed discussions of the methods
applicable to ground-water contamination data the reader is
referred to publications on methods of numerical analyses
(Bear and Verruijt, 1987; Burden and others, 1981; Chau,
1988; de Marsily, 1986; Faust and others, 1981; Gerald and
Wheatley, 1984; Javendel and others, 1984; Mercer and
Faust, 1981; Press and others, 1986; Remson and others,
1971; van der Heidje and others, 1985; Walton, 1985 and
1987; Wang and Anderson, 1982; Yeh and Han, 1989),
statistical analyses (Asbeck and Haimes, 1984; Davis, 1986;
Hallenbeck and Cunningham, 1986; Huntsberger and
Billingsley, 1977; Mendenhall, 1968; SAS Institute, 1988a,
1988c, 1988d, and 1989b; Taylor, 1987; Keely, 1989;
Vomvoris and Gelhar, 1986; Vouk and others, 1985),
graphical construction techniques and analyses (Cedergren,
1989; Cheng, 1988; CIS World, 1989; SAS Institute, 1989a
and 1989b; Tufte, 1983; van derHeijde and Srinivasan,
1983; Watson, 1984), and theoretical relationships (Bitton
and Gerba, 1984; Corey, 1977; Hillel, 1982; Krabbenhoft
and Anderson, 1986; Kueper and Frind, 1988; Stumm and
Morgan, 1981; White, 1988). Although space limitations
prohibit detailed discussions of the foregoing methods,
some general precautions are cited which may be valuable
to the reader in selecting and applying methods of data
presentation and analysis.
1. Computational Methods
Methods for the computation of key indicators of perform-
ance range from simple arithmetic to highly sophisticated
transport models incorporating chemical, physicochemical,
and biological reaction rates. The use of a particular method
should be justified by the conditions to which it is being
applied. Such justifications should be supported by the
appropriate quantity and quality of requisite data.
Subsurface contaminant transport models incorporate a
number of theoretical assumptions about certain natural
processes governing the transport and fate of contaminants.
However, these assumptions may be violated at many sites.
In order to make solutions tractable, simplifications are
made in applications of theory to practical problems. A
common simplification is the assumption that flow towards
a well is horizontal. This may then allow the application of
a two-dimensional model, rather than a more data intensive
three-dimensional model, to many sites. Two-dimensional
models typically do not represent complexities of three-
dimensional flow regimes. Most pump-and-treat
remediations use partially penetrating wells, which induce
significant vertical flow, whereas the two-dimensional
models implicitly assume that the remediation wells are
screened throughout the saturated thickness of the aquifer.
The majority of these models assume that the density and
viscosity of the fluid(s) are equal to that of pure water. It is
known that fluid density differences may cause three-
dimensional flow (e.g., highTDS leachate from landfills).
However, if the limitations of the models are realized and
accounted for, they may serve as useful tools to predict
ground-water flow and contaminant transport relatively
accurately at many sites.
It is essential to have appropriate field and laboratory
determinations of natural process parameters and variables.
These may be used to determine the validity and usefulness
of each modeling attempt. Errors arising from inadequate
data are not addressed properly by mathematical tests such
as sensitivity analyses, or by the application of stochastic
techniques for estimating uncertainty. Such tests and
stochastic simulations assume that the underlying concep-
tual basis of the model is correct (Andersen and others,
1984; Bear and Verruijt, 1987; Dagan, 1982b; El-Kadi,
1988; Konikow, 1986; Krabbenhoft and Anderson, 1986;
Smith and Freeze, 1979aand 1979b). A conceptual model
should be modified only if the data justify the change. The
high degree of hydrogeological, chemical, and microbio-
logical complexity typically present in field situations
requires site-specific characterization of various natural
processes by detailed field and laboratory investigations.
It is also important that both the mathematics that describe
the models and the input parameters used by the models be
subjected to rigorous quality control measures. Otherwise,
results from field applications of models may be
qualitatively and quantitatively incorrect. If done properly,
mathematical modeling may be used to organize vast
amounts of disparate data into a sensible framework that
may provide realistic appraisals of the performance of
ground-water remediations.
Some of the more traditional appraisals conducted with
mathematical models include:
(i) evaluation of changes in size and relative
concentrations of conservative solute plumes, using
flow models;
(ii) evaluation of changes in size and relative
concentration of reactive solute plumes, using flow
models;
(iii) evaluation of changes in size and average
concentration of conservative plumes, using
contaminant transport models; and
(iv) evaluation of changes in size and average
concentration of reactive plumes, using
contaminant transport models.
21
-------
Models may also be used to evaluate the effects of changes
in basic design and operating parameters, so that the most
effective and efficient remediation can be attained.
Geochemical speciation models and reaction rate models
can likewise be employed to improve the interpretation of
data that might otherwise be appraised by simple cation/
anion balances. The outputs of speciation and reaction rate
models may yield useful information concerning the release
rates of contaminants, treatment needs, and disposal options.
Quality control procedures have been developed for the
creation, documentation, and intercomparison of models
(Huyakorn, 1984; Mercer and Faust, 1981; Mercer and
others, 1982; Remson and others, 1971; van der Heijde and
others, 1985). These include:
(i) functionality testing of subroutines and linkages
between subroutines and the main driver of a
program;
(ii) double key entry of all test data, spell-checking,
automatic syntax checking, and debugging;
(iii) rate-of-convergence and other stability checks;
comparison of model outputs with known solutions
to analytical equations; and
(iv) partial validation by blind prediction of controlled
experiments.
The development of quality control procedures for
applications of models to field problems has not progressed
beyond the bounds of individual professional judgment.
2. Statistical Methods
The potential power of statistical methods has been tapped
infrequently in ground-water contamination investigations,
aside from their use in quality assurance protocols. There
are other important uses of statistical methods, however, as
shown in Table 1. While data interpretation and
presentation methods vary widely, most site documents lack
statistical evaluations. Many site reports present
inappropriate simplifications of datasets, such as grouping
or averaging broad categories of data, without regard to the
statistical validity of those simplifications.
Laboratory scientists use analysis of variance (ANOVA) and
covariance (ANCOVA) to analyze the sources of error in
sample analyses. Certain errors, such as those associated
with operator error, may be identified and segregated from
the irreducible errors associated with instrumental
limitations. It may then be possible to institute changes that
minimize and control such errors within specific confidence
Table I. Useful Statistical Methods for Performance Evaluations
Analysis of Variance (ANOVA) Techniques
ANOVA techniques may be used to segregate errors due to chemical
analyses from those errors that are due to sampling procedures and from
the intrinsic variability of the contaminant concentrations at each sampling
point.
Correlation Coefficients
Correlation coefficients can be used to provide justification for lumping
various chemicals together (e.g., total VOCs), or for using a single
chemical as a class representative, or to link sources by similar chemical
behavior.
Regression Equations
Regression equations may be used to predict contaminant loads based on
historical records and supplemental data, and may be used to test cause-
and-effect hypothesis about sources and contaminant release rates.
Surface Trend Analysis Techniques
Surface trend analysis techniques may be used to identify recurring and
intermittent (e.g., seasonal) trends in contour maps or ground-water
elevations and contaminant distributions, which may be extrapolated to
source locations or future plume trajectories.
bounds (Davis, 1986; Huntsberger and Billingsley, 1977;
SAS Institute, 1988c and 1989b; Taylor, 1987)
Multiple or linear regression plots may be used to examine
time-wise or space-wise trends in contaminant
concentrations and water-level elevations. When used in
conjunction with Student's t-test and Fischer's F-test,
regression equations can test hypotheses about cause-and-
effect relationships (Huntsberger and Billingsley, 1977; SAS
Institute, 1988c, 1988dand 1989b). Correlation coefficients
may be computed for one variable versus another to study
the association between them, and to justify grouping,
averaging, or selection of class/category representatives.
Datasets that contain too few observations or wildly
fluctuating values may be analyzed with non-parametric
methods to gain some sense of average or expected value
and associated uncertainty.
The general linear model (GLM), auto-regressive moving
average (ARMA), and auto-regressive integrated moving
average (ARIMA) techniques may provide some additional
information about the direction in which the mean is
trending and its stability (SAS Institute, 1988c and 1988d).
Geostatistical methods, such as kriging, cokriging, and
conditional simulations, offer similar capabilities for two-
and three-dimensional evaluations. The expected values at
key locations and the uncertainties associated with them
22
-------
may be estimated at specified confidence levels (Steinhorst
and Williams, 1985; Usunoff and Guzman-Guzman, 1989).
Cluster analyses and principal component analyses may also
be used to examine two- and three-dimensional datasets to
identify dominant or recurring patterns.
3. Graphical Methods
Graphical methods of data presentation and analysis have
been used heavily in ground-water flow and geochemical
evaluations (Cedergren, 1989; Freeze and Cherry, 1979;
Javendel and others, 1984; Keely, 1984; Keely andTsang,
1983; Watson, 1984; Cheng, 1988; Rogers, 1989; Stumm
and Morgan, 1981; Tinsley, 1979).
Flowline plots offer much information because velocity
computations are typically provided as a vehicle for
obtaining the flowlines. Most flowline plotting software
allows the user to create a flowline plot on which are
superimposed plots of injection fronts or arrival fronts for
specific durations of pumping. From analysis of such plots,
it is possible to estimate the number of pore volumes of
ground water that will be removed over a set period of time
of constant pumping, at different locations in the
contaminant plume.
Contour maps of contaminant concentrations or water-level
elevations are probably the most common graphics prepared
in site studies. When contour maps are prepared by hand,
the practitioner's judgment is automatically applied to
shaping and smoothing contours. Beyond purely
mathematical decisions regarding the interpolation (e.g.,
linear, geometric, lognormal) of values at locations that lie
between measured points, there is the need to account for
the behavior of real physical boundaries (e.g., streams,
valley walls and bedrock).
When contour maps are prepared for identical data sets by
different automated mathematical routines dissimilar
outcomes are possible (Figure 6). The perceptions that may
be generated by examination of contour maps that have been
prepared by different automated mathematical interpolation
schemes can vary substantially because both the shapes of
the contours and the areas contained by specific contours
may vary considerably. It is clear that there are contouring
techniques, such as kriging, that allow measured data points
to be predicted precisely. However, even these are subject
to map-edge effects and often have difficulty with
boundaries. The shapes of the contours that are generated
by kriging depend on the kind of model that the data are
postulated to follow (e.g., spherical, parabolic). Moreover,
by varying the density of the grid of data or the number of
contours used to produce a contour map, major differences
in the locations and bounded areas of the contours may
occur.
20500
20000 -
19500 -
19000 -
18500
i—^r 1 1 I 1 1 r
18500 19000 19500 20000 20500
Inverse Distance Squared
20500
20000 -
19500 -
19000 -
18500
18500
19000 19500
20000
20500
Kriging (Universal)
Figure 6. The results of contouring a dataset by various mathematical
techniques may be significantly different.
Uni-directional plots, such as contaminant concentration
levels versus distance between source areas and potential
receptors, may also be used in performance evaluations.
These are constructed by plotting contaminant concentration
as a function of distance along a specific flowline. If the
flowline patterns shift due to the influence of pumping
wells, transport extrapolations will no longer be valid.
Extrapolations must be validated by proof of duration of
23
-------
transport to allow sufficient time for contaminants to
traverse the length of the flowline in a single orientation.
Time plots of contaminant concentration levels on a
monthly or quarterly basis may be prepared for specific
monitoring locations, but caution must be used to validate
decreasing or increasing trends of concentration levels.
Changes in contaminant concentration levels may occur as a
result of any combination of (i) source release
characteristics, (ii) depletion of the contaminants, and (iii)
flowline pattern shifts.
Stiff diagrams are multivariate plots of the milli-equivalence
values of the major cations and anions of water samples that
produce readily recognizable patterns of the chemical
composition of ground-water. The major ionic composition
of contaminated ground water may differ significantly from
the natural ground-water quality in adjacent areas, often
resulting in different Stiff diagram patterns. Water quality
specialists use the diagrams to differentiate zones of varying
water quality, such as leachate plumes emanating from
landfills and the upconing of saline water due to pumping.
Geochemical prospectors also use this and similar graphical
techniques to identify waters associated with mineral
deposits.
Recently, multivariate plots similar to Stiff diagrams have
been adapted to display organic chemical contaminants
(Lesage and Lapcevic, 1990). For example, a compound of
interest such as trichloroethene may be evaluated in terms of
its contribution to the total organic contaminant load, or
against other specific contaminants, so that some
differentiation of source contributions to the overall plume
can be obtained (Figure 7).
parts per billion
Tetrachloro
Trichloro
1,2-Dichloro -
1,1-Dichloro
Mono Chloro
ETHANES
ETHENES
Figure 7. Pattern diagram for selected VOCs.
Aside from the plotting of discrete quantitative values of
chemical concentrations or water-level elevations, plots of
qualitative information are useful. Soil survey maps,
drillers' logs, geologic cross-sections, and fence diagrams of
sediment sequences are common qualitative methods of
presenting data. All are subject to major errors when data
points are sparse or inappropriately sampled; most tend to
be used in a manner that understates the true uncertainties
involved. For example, it may not be appropriate to detail
the possible linkages of strata denoted on the geologic log of
one location with that of another when the strata are known
to occur as remnants of stream channels and reworked
floodplain deposits. Inspection of raw datasets by those
whose judgment has been shaped by many hours at local
sites can provide valuable input for decisions on how
strongly to associate occurrences of similar sediments.
In addition to quantitative and qualitative plots and maps,
certain charts and illustrations can be prepared for use in
management of the remedial activities. Program evaluation
and review technique (PERT) charts may be used to map the
dependencies of each task on those that precede it, and can
be configured to facilitate analysis of the operations by the
critical pathway method (CPM), in which the chain of core
tasks that must be completed to lead to a specific output can
be identified and protected during modifications to the list
of tasks (Figure 8). Although this may seem trivial when a
single individual is charting tasks for a single project, it is
clearly more complex when coordinating the tasks of many
individuals or projects.
Other hierarchal action diagrams, such as sample selection
and routing protocols, are useful tools for improvement of
the efficiency and effectiveness of pump-and-treat
remediations. Database management systems, particularly
those that operate under a geographical information system
(GIS) are also extremely valuable for analysis and
presentation of hydrogeologic and ground-water
contamination data.
4. Theoretical
General indicators of physicochemical reaction potential
include the cation-exchange-capacity (CEC), alkalinity, and
acidity of subsurface samples. Theoretical relationships link
CEC, alkalinity, and acidity to the reserve of reactants, and
consequently to the potential products. Theoretical
equations have also been developed and tested for various
wastewater treatment unit designs.
Batch sorption isotherms and static microcosms can be used
to generate estimates of limiting reaction rates of specific
contaminants. Information on the dynamics of such
reactions is obtained from flow-through column studies
utilizing labelled tracers. It is most common to evaluate
24
-------
Water Table I
Measurements 1
Aquitard Head 1
Measurements 1
Artesian Head \
Measurements \
Excavation &
Drilling Logs
Surface
Geophysics
Aerial
Photographs
Contour Maps of Static Water Levels
andPotentiometric Surfaces
Local (Hydro)Geology Cross-Sections & Maps
Used to Select Observation Well Locations
Pumping & Tracer Tests for
Aquifer Hydraulic Properties
Drilling of Observation Wells &
Collection of Sediment Samples
Maps of Aquifer Properties &
Equilibrium Pumping Contours
Lab Evaluation of Sediment
Porosity & Moisture Retention
HYDROGEOLOGIC FLOW MODEL
Figure 8. PERT Chart illustrating interdependency of tasks to produce outputs.
data from such experiments by comparison to an equation
that has been derived from theory, as opposed to the use of
empirical relationships.
The probable reaction rates and products of some
compounds may be predicted by quantitative structure-
activity relationships (QSAR). These relationships typically
utilize information such as the physical and symmetry
properties of the molecule to draw analogies to the behavior
of other molecules.
General Protocols for Performance
Evaluations
The protocols presented here are generalized and should be
refined on a site-specific basis. The flowcharts that are
provided are neither exhaustive nor exclusive, but should
serve as reasonable skeletons on which to build site-specific
charts of key activities. The latter can be accomplished
most fruitfully by utilization of software that create and
allow interactive modification and automated updating of
PERT charts.
1. General Protocol for Selection of Monitoring
Criteria and Reporting Requirements
(1) Begin the selection of performance evaluation and
monitoring criteria by reviewing the data
collection history at the site to determine the
potential need for adoption of specific spatial or
temporal patterns of sampling to ensure continuity
with previous efforts.
(2) Pre-operational programs for sample collection
and analyses, quality control, and data interpre-
tation will be required between the completion of
the RI/FS and effective start-up of the remediation.
These will be needed to provide current
information regarding the threat to human health
and the environment, and the size and orientation
of the zone of remedial action that should be
established to prevent loss of portions of the
contaminant plume to downgradient areas. Select
monitoring criteria appropriate to these data needs.
(3) Bench treatability studies and pilot plant
demonstrations may be called for by design
engineers prior to commencement of construction
of the final remedy.
(4) Consider conducting studies at flow rates
equivalent to those expected for different locations
within the proposed remediation wellfield.
(5) Obtain samples of subsurface sediments for
chemical extraction analyses of contaminant
25
-------
residuals and for key transport properties such as
total organic carbon during opportunities presented
by pre-construction activities (e.g., excavations,
trenching, drilling). Evaluate the potential for
NAPLs at the site.
(6) Overlay plots of all sampling events and sample
types to identify locations of inadequate sampling
and to set a schedule to fill such vacancies. Select
performance evaluation criteria appropriate to these
activities.
(7) Prepare brief reports of sample collection and
analyses as data becomes available, rather than
wait for all items to be synthesized into a major
report. This information can affect ongoing
preparations for commencement of operation.
(8) Modify the frequency and locations for sampling
during operation, as necessary. Specifically focus
on characterization of the flowlines that will result
from operation of the remediation wellfield, and on
tracking the transport of contaminants along the
flowlines. Anticipate flowline shifts due to partial
shutdowns or operational failures.
(9) Provide adequate coverage of monitoring wells and
piezometers beyond the horizontal and vertical
bounds of the plume, such that transgressions of
flowlines originating in the plume can be tracked to
ensure eventual capture by an extraction well, and
such that the contribution of uncontaminated
ground water to the total volume pumped can be
examined.
(10) Make provisions for measurements germane to
active management of the operation, such as flow
rates and contaminant concentration levels for (i)
individual extraction wells, (ii) treatment plant
influent, (iii) key treatment units, and (iv) the plant
effluent.
(11) Reduce the frequency of chemical sampling (but
not water-level elevation measurements) if stable
trends of contaminant concentration levels are
apparent, and provide contingencies for increased
sampling in response to those periods when the
data are highly erratic.
(12) Continue filling gaps in the site characterization
database with opportunistic and periodic testing.
(13) Evaluate the rate of depletion of the residual
contamination by collection of sediment samples
for chemical extraction and analysis. This
operation should be accompanied by a statistical
evaluation of contaminant distribution at each
sampling point. Chart the estimated mass of
contaminants removed, both total and per location;
compare these estimates with the mass of
contaminants removed by the treatment system.
(14) Plot and examine water-level elevation contours to
determine the adequacy of hydrodynamic control
and to identify stagnation zones for subsequent
action.
(15) As the concentration-time plots begin to flatten out,
try varying the flowrates of a few extraction wells
at a time to determine the potential for
improvement with wellfield scheduling or
configuration changes.
(16) When all potential improvements have been
attempted and exhausted, prepare for termination
of the remediation by increasing the sampling
frequency to build a statistically significant
database for plotting regressions with 95%
confidence bounds so that success can be judged.
(17) Throughout the operational phase of the
remediation, utilize monitoring data as soon as they
become available, for active management
purposes; prepare comprehensive annual reports
for regulatory oversight purposes.
(18) If the regression plots indicate that the target
contaminant concentration levels have been
reached, propose termination of the remediation.
(19) If permission to begin termination is granted by the
regulators, continue the sampling program at the
same (high) frequency for up to two years to test
the need to re-activate and/or re-evaluate the
remediation. Prepare brief reports quarterly for
management and regulatory purposes during this
period.
(20) Subsequently reduce the frequencies of collecting
and reporting on samples and measurements to an
annual basis to monitor the persistence of the
effects of the remediation; continue at that level
until the end of the regulatory period.
2. General Protocol for Violations Reporting and
Responses
(1) Examine the data collected during each sampling
event for internal consistency prior to accepting the
accuracy or relevancy of anomalous values. If a
26
-------
Good
Condition
Detailed Inspection of
Remediation Wellfield
Develop Baseline Chemical
and Hydraulic Datasets
Poor Condition
Perform All Repairs
and Maintenance
Control
—i OK
Evaluate Effectiveness
of Hydraulic Control
Out of
Control
No Action
Needed
Evaluate Mass
Recovery Rates
Poor or
Declining
Evaluate Potential Effectiveness
of Alternate Wellfield Configurations
Reconfigure Remediation Wellfield
\ Wells Not
i Needed
Evaluate Resulting Flowline Pattern
•with Respect to Monitoring Well Locations
Additional Monitoring
Wells Needed
Install Additional Monitoring Wells
Figure 9. Outline of key efforts needed to select and use monitoring and reporting requirements.
single monitoring point indicates an unacceptable
level of a particular contaminant, check the
concentration levels of other contaminants at the
same location to see if they have approximately the
same relationship (ratio) to the suspect contaminant
as has been shown to occur generally for the
plume; if not, reject the suspect value as internally
inconsistent.
(2) If general agreement seems to be indicated, or
cannot be judged due to the small number of
contaminants associated with a particular plume
(e.g., predominantly TCE and nothing else), then
compare the suspect value with the concentration
levels found in nearby wells; if poor agreement is
indicated, reject considering the suspect value as
locally inconsistent.
(3) If the suspect value is not rejected due to internal
inconsistency or due to local inconsistency, retain
the value in the dataset as valid and immediately
resample or remeasure at that location and at two
or more adjacent locations (for control). Evaluate
the resulting data in the same fashion as the initial
data.
(4) Regardless of the status (rejected or retained) of the
first such indication at a specific location, consider
successive indications to confirm the loss of
adequate control of the zone of remedial action.
(5) Study the wellfield scheduling to determine at
which wells the flowrates may be increased to
provide the degree of control needed.
27
-------
(6) Expect that flowline patterns will shift with the
implementation of new pumping schedules, and
plan to evaluate the significance of the shift in
terms of the need to reposition sampling/
measurement points.
(7) Verify the operational viability (e.g., minimum
hydrodynamic control) of the modified wellfield by
increasing the frequency of sampling/measurement
for a period of three to six months.
(8) Utilize information obtained during the increased
sampling/measurement period to assess the impacts
of the correctional action in terms of changes in the
rate of depletion of the contaminant residuals.
3. General Protocols for Design and Operation
Modifications
(1) Modifications of system design or operation may
be called for either as a result of checking the
(2)
(3)
remediation for readiness for termination or as a
result of violations. In either case, if a major
modification is anticipated, it is prudent to perform
a detailed inspection and analysis of the system
first. All piping, valves, and protective shielding
should be in good condition; leaks, indications of
corrosion or excessive wear, and poor electrical
grounds should be repaired. All gauges and
electronic monitors should be in calibration and
should provide appropriate responses to tests; if
not, the deficiencies should be remedied.
A few sampling/measurement events should then
be undertaken to provide a baseline for comparison
to data to be collected after the modification, and to
provide immediate data for development of the
modification.
Development of the modification should proceed
with the targeted hydraulic control and clean-up
levels still firmly in mind. If the remediation has
Examine Monitoring Data for Anomalous
Contaminant Levels & Hydraulic Gradients
No
Anomalies
Anomalies Found
Examine the Dataset Containing Anomalous
Value (s), for Internal (Point) Consistency .. . .
^Inconsistent
Consistent
Examine the Dataset Containing Anomalous
Value (s), for Local (Area) Consistency
Consistent
Retain Anomalous Data
Re-sample & Remeasurefor Confirmation
Confirmed
Design and Implement Modifications to
Flow Rates and/or Wellfield Configuration
i Inconsistent
Not
\Confirmed
Remediation Judged to be Operating
Within Bounds of Compliance Criteria
Reject Anomalous Data and
Re-sample/Remeasure for New Dataset
Data Not Erratic
Conduct Chemical Time-Series Testing of
Monitoring Wells Yielding Anomalous Data
Data Erratic
Conduct Special Studies of Flow Rates
and Contaminant Removal Rates
Conduct New Chemical Samplings &
Hydraulic Gradient Measurements
Install Additional Monitoring Wells and
Piezometers Needed to Improve Monitoring
Figure 10. Key efforts needed to identify, verify, and correct inadequate performance of the remediation.
28
-------
been successful in depleting contaminants from the
outermost portions of the plume, it may be sensible
to shrink the remediation wellfield inward. If
additional control is needed, it may need to be
expanded outward.
(4) In these cases particularly, but also in the case
where the overall size remains unaffected and wells
are simply relocated within the existing wellfield, it
is essential to begin to anticipate the new
orientation of the flowline pattern.
(5) Monitoring point locations may need to be
changed, and new trends of contaminant transport
and recovery will have to be identified. Hence, an
increased frequency of sampling/measurement
events will be needed for three to six months
following establishment of routine operation of the
modified system.
(6) Verification of operational viability and
assessment of the impacts of the modification on
depletion rates of contaminant residuals then
follow.
Select Chemical Goals
to Satisfy ARAR's
Select Hydraulic Goals
to Control Source & Plume
MODIFICATION OR TERMINATION
Identify Spatial Gaps & Temporal Trends
in Datasetsfrom Existing Monitoring Wells
Continue Performance Evaluations & Use to
Determine Need for Changes or Termination
Install Wells to
Fill Spatial Gaps
Conduct Bench (Lab)
Treatability Studies
Finalize Operational Goals and Set
Termination & Post Operational Requirements
Design, Install, and Operate a Pilot
Remediation Wellfield & Treatment Plant
Reconfigure the Wellfield and Treatment
Train as Needed to Achieve Best Performance
Monitor Contaminant Concentrations and
Water Table (Potentiometric Surface) Elevations
Use the Chemical and Hydraulic Data to
Evaluate Performance of the Remediation
Figure 12. Key efforts needed to prepare for, implement, and verify modifications of the remediation.
29
-------
References
Abriola, L.M. and G.F. Finder. 1985. AMultiphase
Approach to the Modeling of Porous Media Contamination
by Organic Compounds. Water Resources Research
Andersen, P.P., C.R. Faust, and J.W. Mercer. 1984.
Analysis of Conceptual Designs for Remedial Measures at
Lipari Landfill. Groundwater, 22(2).
Anderson, M.P 1979. Using Models to Simulate the
Movement of Contaminants Through Groundwater Flow
Systems. Critical Reviews in Environmental Controls,
9(2):97-156.
Asbeck, E. andY.Y. Haimes. 1984. The Multiobjective
Risk Method. Large Scale Systems, 13(38).
Bahr, J. 1989. Analysis of Nonequilibrium Desorption of
Volatile Organics during Field Test of Aquifer
Decontamination. Journal of Contaminant Hydrology
4(3):205-222.
Baehr, A.L., G.E. Hoag, and M.C. Marley. 1989. Removing
Volatile Contaminants from the Unsaturated Zone by
Inducing Advective Air-Phase Transport. Journal of
Contaminant Hydrology 4(1): 1-26.
Barker, J.F, J.E. Barbash, and M. LaBonte. 1988.
Groundwater Contamination at a Landfill Sited on Fractured
Carbonate and Shale. Journal of Contaminant Hydrology
Barker, J.F, G.C. Patrick, and D. Major. 1987. Natural
Attenuation of Aromatic Hydrocarbons in a Shallow Sand
Aquifer. Ground Water Monitoring Review 7(1):64-71.
Bear, J. and A. Verruijt. 1987. Modeling Groundwater Flow
and Pollution. D. Reidel Publishing Company; Boston,
Massachusetts.
Bengtsson, G., C.G. Enfield, and R. Lindquist. 1988.
Macromolecules Facilitate Transport of Trace Organics.
Science of the Total Environment 67: 159-169.
Bitton, G. and C.P Gerba (editors). 1984. Groundwater
Pollution Microbiology. Wiley Interscience/John Wiley &
Sons Inc.; New York.
Blatt, H., G.V Middleton, and R.C. Murray. 1980. Origin
of Sedimentary Rocks (2nd Ed.). Prentice-Hall Inc.;
Englewood Cliffs, New Jersey.
Borden, R., M. Lee, J.M. Thomas, P. Bedient, and C.H.
Ward. 1989. In Situ Measurement and Numerical
Simulation of Oxygen Limited Biotransformation. Ground
Water Monitoring Review 9(1):83-91.
Bouchard, D.C., C.G. Enfield, and M.D. Piwoni. 1989.
Transport Processes Involving Organic Chemicals, in
Reactions and Movement of Organics Chemicals in Soils.
Soil Science Society of America, Special Publications Press.
Bouchard, D.C., A.L. Wood, M.L. Campbell, P. Nkedi-
Kizza, andPS.C. Rao. 1988. Sorption Nonequilibrium
during Solute Transport. Journal of Contaminant Hydrology
2(3):209-223.
Boutwell, S.H., S.M. Brown, B.R. Roberts, and D.F.
Atwood. 1985. Modeling Remedial Actions at
Uncontrolled Hazardous Waste Sites. U.S. Environmental
Protection Agency publication EPA/540/2-85/001.
Bouwer, E.J. and J.P Wright. 1988. Transformations of
Trace Halogenated Aliphatics in Anoxic Biofilm Columns.
Journal of Contaminant Hydrology 2(2): 155-169.
Bouwer, H. and R.C. Rice. 1976. A Slug Test for
Determining Hydraulic Conductivity of Unconfined
Aquifers with Completely or Partially Penetrating Wells.
Water Resources Research, v. 12, pp. 423-428.
Brown, J. 1986. Environmental Software Review.
Pollution Engineering, 18(1).
Brusseau, M.L., P.S.C. Rao, R.E. Jessup, and J.M.
Davidson. 1989. Flow Interruption: A Method for
Investigating Sorption Nonequilibrium. Journal of
Contaminant Hydrology 4(3):223-240.
Burden, R.L., J.D. Faires, and A.C. Reynolds. 1981.
Numerical Analysis (2nd Ed.). Prindle, Weber, and
Schmidt; Boston, Massachusetts.
Cannon, J.Z. 1989. Considerations in Ground Water
Remediation at Superfund Sites. Directive No. 9355.4-03.
Office of SolidWaste and Emergency Response, U.S. EPA,
Washington, D.C.
Cedergren, H.R. 1989. Seepage, Drainage, and Flow Nets
(3rd Ed.). John Wiley & Sons Inc.; New York, pp. 125-147.
Chapuis, R.P 1989. Shape Factors for Permeability Tests
in Boreholes and Piezometers. Ground Water 27(5) :647-
654.
30
-------
Chau, T.S. 1988. Analysis of Sustained Ground-Water
Withdrawals by the Combined Simulation-Optimization
Approach. Ground Water 26(4):454-463.
Cheng, S. 1988. Computer Notes - Trilinear Diagram
Revisited: Application, Limitation, and an Electronic
Spreadsheet Program. Ground Water 26(4):505-510.
Compton, R.R. 1962. Manual of Field Geology. John
Wiley & Sons Inc.; New York.
Corey, A.T. 1977. Mechanics of Heterogeneous Fluids in
Porous Media. Water Resources Publications; Fort Collins,
Colorado.
Council on Economic Priorities. 1986. Hazardous Waste
Management: Reducing the Risk. Island Press; Washington,
D.C.
Crow, W.L., E.P Anderson, and E.M. Minugh. 1987.
Subsurface Venting of Vapors Emanating from Hydrocarbon
Product on Ground Water. Ground Water Monitoring
Review 7(l):51-57.
Curtis, G.P, P. V Roberts, and Martin Reinhard. 1986. A
Natural Gradient Experiment on Solute Transport in a Sand
Aquifer: 4. Sorption of Organic Solutes and Its Influence on
Mobility. Water Resources Research 22(13):2059-2067.
Dagan, G. 1982a. Stochastic Modeling of Groundwater
Flow by Unconditional and Conditional Probabilities: Part
1. Conditional Simulation and the Direct Problem. Water
Resources Research 18(4):813-833.
Dagan, G. 1982b. Stochastic Modeling of Groundwater
Flow by Unconditional and Conditional Probabilities: Part
2. The Solute Transport Problem. Water Resources
Research 18(4):835-848.
Daley, PS. 1989. Cleaning Up Sites with On-site Process
Plants. Environmental Science and Technology 23(8):912-
916.
Davis, J.C. 1986. Statistics and Data Analysis in Geology
(2nd Ed.). John Wiley & Sons Inc.; New York, pp.227-250.
Davis, S.N. and R.J.M. DeWeist. 1966. Hydrogeology.
John Wiley & Sons Inc.; New York.
de Marsily, G. 1986. Quantitative Hydrogeology:
Groundwater Hydrology for Engineers. Academic Press
Inc.; Orlando, Florida.
Eaton, J.D. 1982. Real Estate Valuation in Litigation.
American Institute of Real Estate Appraisers; Chicago,
Illinois.
El-Kadi, A. 1988. Applying the USGS Mass Transport
Model (MOC) to Remedial Actions by Recovery Wells.
Groundwater, 26 (3).
Enfield, C.G. 1985. Chemical Transport Facilitated by
Multiphase Transport Flow Systems. Water Science and
Technology, 17(9): 1-12.
Enfield, C.G. and G. Bengtsson. 1988. Macromolecular
Transport of Hydrophobic Contaminants in Aqueous
Environments. Ground Water 26(1):64-70.
Faust, C.R. 1985. Transport of Immiscible Fluids Within
and Below the Unsaturated Zone: A Numerical Model.
Water Resources Research 21(4):587-596.
Faust, C.R., L.R. Silka, and J.W Mercer. 1981. Computer
Modeling and Ground- Water Protection. Groundwater, 19
(4).
Feenstra, S., J.A. Cherry, E.A. Sudicky, and Z. Haq. 1984.
Matrix Diffusion Effects on Contaminant Migration from an
Injection Well in Fractured Sandstone. Ground Water
22(3):307-316.
Fetter, C.W 1988. Applied Hydrogeology (2nd Ed.).
Merrill Publishing Company; Columbus, Ohio.
Flathman, P., D. Jerger, and L. Bottomley. 1989.
Remediation of Contaminated Ground Water Using
Biological Techniques. Ground Water Monitoring Review
Freeze, R.A. and J.A. Cherry. 1979. Groundwater.
Prentice Hall, Inc.; Edgewood Cliffs, New Jersey.
Gerald, C.F and P.O. Wheatley. 1984. Applied Numerical
Analysis (3rd Ed.). Addison- Wesley Publishing Company;
Reading, Massachusetts.
Gelhar, L.W 1986. Stochastic Subsurface Hydrology from
Theory to Applications. Water Resources Research
22(9):135S-145S.
Gelhar, L.W. and C.L. Axness. 1983. Three-Dimensional
Stochastic Analysis of Macrodispersion in Aquifers. Water
Resources Research 19(1):161-180.
Gelhar, L.W, A.L. Gutjahr, and R.L. Naff. 1979.
Stochastic Analysis of Macrodispersion in a Stratified
Aquifer. Water Resources Research 1 5 (6) : 1 3 87- 1 3 97 .
CIS World. 1989. CIS Technology '89: Results of the 1989
GIS World Geographical Information Systems Survey. GIS
World, Inc.; Fort Collins, Colorado.
31
-------
Goltz, M.N. and P.V. Roberts. 1986. Interpreting Organic
Solute Transport Data from a Field Experiment using
Physical Nonequilibrium Models. Journal of Contaminant
Hydrology l(l/2):77-94.
Goltz, M.N. and P.V. Roberts. 1988. Simulations of
Physical Nonequilibrium Solute Transport Models:
Application to a Large-Scale Field Experiment. J.
Contaminant Hydrology 3(l):37-64.
Gschwend, P.W. and M.D. Reynolds. 1987. Monodisperse
Ferrous Colloids in an Anoxic Groundwater Plume. Journal
of Contaminant Hydrology, 2(4):309-327.
Gschwend, P.W. and S.C. Wu. 1985. On the Constancy of
Sediment-Water Partition Coefficients of Hydrophobic
Organic Pollutants. Environmental Science and
Technology, 19(1).
Guven, O. and FJ. Molz. 1986. Deterministic and
Stochastic Analyses of Dispersion in an Unbounded
Stratified Porous Medium. Water Resources Research
22(11):1565-1574.
Hallenbeck, W.H., and K.M. Cunningham. 1986. Quan-
titative Risk Assessment for Environmental and Occupa-
tional Health. Lewis Publishers Inc.; Chelsea, Michigan.
Heath, R.C. 1983. Basic Ground-Water Hydrology. U.S.
Geological Survey Water-Supply Paper 2220.
Hess, A.E. and FL. Paillet. 1989. Characterizing Flow
Paths and Permeability Distribution in Fractured-Rock
Aquifers Using a Sensitive, Thermal Borehole Flowmeter.
Proceedings of New Field Techniques for Quantifying the
Physical and Chemical Properties of Heterogeneous
Aquifers. National Water Well Association, Dublin, Ohio.
Hillel, D. 1982. Fundamentals of Soil Physics. Academic
Press; New York, 2nd edition.
Hinchee, R. and H.J. Reisinger. 1987. APractical
Application of Multiphase Transport Theory to Ground
Water Contamination Problems. Ground Water Monitoring
Review 7(l):84-92.
Hossain, M. A. and M.Y. Corapcioglu. 1988. Modifying the
USGS Solute Transport Computer Model to Predict High-
Density Hydrocarbon Migration. Ground Water 26(6) :717-
723.
Huling, S.G. and J.W Weaver. 1991. Ground Water Issue:
Dense Nonaqueous Phase Liquids. EPA/540/4-91/002. Office
of Research and Development, Robert S. Kerr Environmental
Research Laboratory, U.S. EPA; Ada, Oklahoma.
Hunt, J.R., N. Sitar, and K.S. Udell. 1988a. Nonaqueous
Phase Liquid Transport and Cleanup: 1. Analysis of
Mechanisms. Water Resources Research 24(8): 1247-1258.
Hunt, J.R., N. Sitar, and K.S. Udell. 1988b. Nonaqueous
Phase Liquid Transport and Cleanup: 2. Experimental
Studies. Water Resources Research 24(8): 1259-1269.
Huntsberger, D.V and P. Billingsley. 1977. Elements of
Statistical Inference (4th Ed.). Allyn and Bacon, Inc.;
Boston, Massachusetts.
Huyakorn, PS. 1984. "Testing and Validation of Models
for Simulating Solute Transport in Ground Water:
Development and Testing of Benchmark Techniques."
IGWMC Report No. GWMI 84-13, International Ground
Water Modeling Center, Holcolm Research Institute;
Indianapolis, Indiana.
James, M.L., G.M. Smith, and J.C. Wolford. 1977. Applied
Numerical Methods for Digital Computation with
FORTRAN and CSMP (2nd Ed.). Harper & Row,
Publishers; New York.
Javendel, I., C. Doughty, and C.F Tsang, 1984.
Groundwater Transport: Handbook of Mathematical
Models, AGU Water Resources Monograph No. 10.
American Geophysical Union; Washington, D.C.
Jensen, B.K., E. Arvin, and A.T Gundersen. 1988.
Biodegradation of Nitrogen- and Oxygen-Containing
Aromatic Compounds in Groundwater from an Oil-
Contaminated Aquifer. Journal of Contaminant Hydro.
3(l):65-76.
Jorgensen, D.G., T. Gogel, and D.C. Signer. 1982. Deter-
mination of Flow in Aquifers Containing Variable-Density
Water. Ground Water Monitoring Review 2(2):40-45.
Keely, J.F 1982. Chemical Time-Series Sampling. Ground
Water Monitoring Review 2(3):29-38.
Keely, J.F. 1984. Optimizing Pumping Strategies for
Contaminant Studies and Remedial Actions. Ground Water
Monitoring Review 4(3):63-74.
Keely, J.F. 1989. Ground Water Issue: Performance
Evaluations of Pump-and-Treat Remediations. EPA/540/4-
89/005, Office of Research and Development, Robert S.
Kerr Environmental Research Laboratory, U.S. EPA; Ada,
Oklahoma.
Keely, J.F, M.D. Piwoni, and J.T Wilson. 1986. Evolving
Concepts of Subsurface Contaminant Transport. Journal
Water Pollution Control Federation. 58(5):349-357.
32
-------
Keely, J.F. and C.F Tsang. 1983. Velocity Plots and
Capture Zones of Pumping Centers for Ground Water
Investigations. Ground Water 22(6):701-714.
Kipp, K.L., K.G. Stollenwerk, and D.B. Grove. 1986.
Groundwater Transport of Strontium 90 in a Glacial
Outwash Environment. Water Resources Research
22(4):519-530.
Klotz, D., K.P Seiler, H. Moser, and F. Neumaier. 1980.
Dispersivity and Velocity Relationship from Laboratory and
Field Experiments. Journal of Hydrology; 45(1);169-184.
Konikow, L.F. 1986. Predictive Accuracy of a Ground-
Water Model - Lessons from a Postaudit. Ground Water
24(2): 173-184.
Krabbenhoft, D.P and M.P. Anderson. 1986. Use of a
Numerical Ground-Water Flow Model for Hypothesis
Testing. Groundwater, 24(1).
Kueper, B.H. and E.G. Frind. 1988. An Overview of
Immiscible Fingering in Porous Media. Journal of
Contaminant Hydrology 2(2):95-110.
Lee, L.S., P.S.C. Rao, M.L. Brusseau, andR.A. Ogwada.
1988. Nonequilibrium sorption of Organic Contaminants
during Flow through Columns of Aquifer Materials.
Environmental Toxicology and Chemistry 7(10):779-793.
Lesage, S., and P. A. Lapcevic. 1990. Differentiation of the
Origins of BTX in Ground Water Using Multivariate Plots.
Ground Water Monitoring Review, 10(2): 102-105.
Loftis, J.C., J. Harris, and R.H. Montgomery. 1987.
Detecting Changes in Ground Water Quality at Regulated
Facilities. Ground Water Monitoring Review 7(l):72-76.
Macalady, D.L., P.O. Tratnyek, and T.J. Grundl. 1986.
Abiotic Reduction Reactions of Anthropogenic Organic
Chemicals in Anaerobic Systems: A Critical Review.
Journal of Contaminant Hydrology 1(1/2): 1-28.
Mackay, D.M., W.P Ball, and M.G. Durant. 1986.
Variability of Aquifer Sorption Properties in a Field
Experiment on Groundwater Transport of Organic Solutes:
Methods and Preliminary Results. Journal of Contaminant
Hydrology 1(1/2):119-132.
Mackay, D.M., D.L. Freyberg, P.V. Roberts, and J.A.
Cherry. 1986. A Natural Gradient Experiment on Solute
Transport in a Sand Aquifer: 1. Approach and Overview of
Plume Movement. Water Resources Research 22(13):2017-
2029.
Major, D.W., C.I. Mayfield, and J.F Barker. 1988.
Biotransformation of Benzene by Denitrification in Aquifer
Sand. Ground Water 26(1):8-14.
Matheron, G. and G. de Marsily. 1980. Is Transport in
Porous Media always Diffusive? A Counterexample. Water
Resources Research, 16(5):901-917.
McCarthy, J.F. and J.M. Zachara. 1989. Subsurface
Transport of Contaminants. Environmental Science and
Technology, 23(5):496-502.
Mendenhall, W. 1968. Introduction to Linear Models and
The Design and Analysis of Experiments. Duxbury Press;
Belmont, California.
Mercado, Abraham. 1985. The Use of Hydrogeochemical
Patterns in Carbonate Sand and Sandstone Aquifers to
Identify Intrusion and Flushing of Saline Water. Ground
Water 23(5):635-645.
Mercer, J.W. and C.R. Faust. 1981. Ground-Water
Modeling. National Water Well Association; Dublin, Ohio.
Mercer, J.W., D.C. Skipp, and D. Giffin. 1990. Basics of
Pump-and-Treat Ground-Water Remediation Technology.
EPA/600/8-90/003, Office of Research and Development,
Robert S. Kerr Environmental Research Laboratory, U.S.
EPA; Ada, Oklahoma.
Mercer, J.W., P.S.C. Rao, and I.W Marine. 1983. Role of
the Unsaturated Zone in Radioactive and Hazardous Waste
Disposal. Ann Arbor Science Publishers; Ann Arbor,
Michigan.
Mercer, J.W., S.D. Thomas, B. Ross. 1982. Parameters and
Variables Appearing in Repository Siting Models. U.S.
Nuclear Regulatory Commission NUREG/CR-3066.
Miller, M.M., S.P Wasik, G. Huang, W. Shiu and D.
Mackay. 1985. Relationships between Octanol-Water
Partition Coefficients and Aqueous Solubility.
Environmental Science and Technology 19(7):522-529.
Miller, S. 1980. Cost-Benefit Analyses. Environmental
Science and Technology, 14(12).
Moltyaner, G.L. 1989. Hydrodynamic Dispersion at the
Local Scale of Continuum Representation. Water Resources
Research 25(5):1041-1048.
Molz, F.J., O. Guven, J.G. Melville, and J.F. Keely. 1986a.
Performance and Analysis of Aquifer Tracer Tests with
Implications for Contaminant Transport Modeling. U.S.
EPA Office of Research and Development: EPA/600/2-86-
062.
33
-------
Molz, F.J., M.A. Widdowson, and L.D. Benefield. 1986b.
Simulation of Microbial Growth Dynamics Coupled to
Nutrient and Oxygen Transport in Porous Media. Water
Resources Research 22(8): 1207-1216.
Molz, F.J., O. Guven, J.G. Melville, J.S. Nohrstedt, and J.K.
Overholtzer. 1988. Forced-Gradient Tracer Tests and
Inferred Hydraulic Conductivity Distributions at the Mobile
Field Site. Ground Water 26(5):570-579.
Nkedi-Kizza, P., M.L. Brusseau, P.S.C. Rao, and A.G.
Hornsby. 1989. Nonequilibrium Sorption during
Displacement of Hydrophobic Organic Chemicals and 45Ca
through Soil Columns with Aqueous and Mixed Solvents.
Environmental Science and Technology 23(7):814-820.
Nkedi-Kizza, P., P.S.C. Rao, and A.G. Hornsby. 1985.
Influence of Organic Cosolvents on Sorption of
Hydrophobic Organic Chemicals by Soils. Environmental
Science and Technology, 19(10):975-979.
Nkedi-Kizza, P., P.S.C. Rao, and A.G. Hornsby. 1987.
Influence of Organic Cosolvents on Leaching of
Hydrophobic Organic Chemicals through Soils.
Environmental Science and Technology, 21(11):1107-1111.
Novak, S.A. and Y. Eckstein. 1988. Hydrogeochemical
Characterization of Brines and Identification of Brine
Contamination in Aquifers. Ground Water 26(3):317-324.
Osiensky, J.L., K.A. Peterson, and R.E. Williams. 1988.
Solute Transport Simulation of Aquifer Restoration after In
Situ Uranium Mining. Ground Water Monitoring Review
8(2):137-144.
Osiensky, J.L., G.V. Winter, and R.E. Williams. 1984.
Monitoring and Mathematical Modeling of Contaminated
Ground-Water Plumes in Fluvial Environments. Ground
Water 22(3):298-306.
Ophori, D.U., and J. Toth. 1989. Patterns of Ground-Water
Chemistry, Ross Creek Basin, Alberta, Canada. Ground
Water 27(l):20-26.
Papadopulos, S.S., J.D. Bredehoeft, andH.H. Cooper, Jr.
1973. On the analysis of "slug test" data. Water Resources
Research, 9:1087-1089.
Finder, G.F. and L.M. Abriola. 1986. On the Simulation of
Nonaqueous Phase Organic Compounds in the Subsurface.
Water Resources Research 22(9):109S-119S.
Piwoni, M.D. and P. Banerjee. 1989. Sorption of Volatile
Organic Solvents from Aqueous Solution onto Subsurface
Solids. Journal of Contaminant Hydrology 4(2): 163-179.
Pollock, D.W. 1988. Semianalytical Computation of Path
Lines for Finite Difference Models. Ground Water
26(6):743-750.
Poole, V.L., K. Cartwright, and D. Leap. 1989. Use of
Geophysical Logs to Estimate Water Quality of Basal
Pennsylvanian Sandstones, Southwestern Illinois. Ground
Water 27(5):682-688.
Press, W.H., B.P Flannery, S.A. Teukolsky, and WT.
Vetterling. 1986. Numerical Recipes: The Art of Scientific
Computing. Cambridge University Press; New York.
Remson, I., G.M. Hornberger, and FJ. Molz. 1971.
Numerical Methods in Subsurface Hydrology with an
Introduction to the Finite Element Method. Wiley -
Interscience/John Wiley & Sons Inc.; New York.
Roberts, P.V., M.N. Goltz, and D.M. Mackay. 1986. A
Natural Gradient Experiment on Solute Transport in a Sand
Aquifer: 3. Retardation Estimates and Mass Balances for
Organic Solutes. Water Resources Research 22(13):2047-
2058.
Roberts, P.V., L. Semprini, G.D. Hopkins, D. Grbic-Galic,
PL. McCarty, and M. Reinhard. 1989. In-Situ Aquifer
Restoration of Chlorinated Aliphatics by Methanotropic
Bacteria. U.S. Environmental Protection Agency Project
Summary EPA/600/S2-89/033.
Rogers, R.J. 1989. Geochemical Comparison of Ground
Water in Areas of New England, New York, and
Pennsylvania. Ground Water 27(5):690-712.
SAS Institute Inc. 1988a. SAS Introductory Guide for
Small Computers, Release 6.03 Edition. SAS Institute Inc.;
Gary, North Carolina.
SAS Institute Inc. 1988b. SAS/GRAPHR User's Guide,
Release 6.03 Edition. SAS Institute Inc.; Cary, North
Carolina.
SAS Institute Inc. 1988c. SAS/STATR User's Guide,
Release 6.03 Edition. SAS Institute Inc.; Cary, North
Carolina.
SAS Institute Inc. 1988d. SAS/ETSR User's Guide,
Version 6, First Edition. SAS Institute Inc.; Cary, North
Carolina.
SAS Institute Inc. 1989a. SAS/ORR User's Guide, Version
6, First Edition. SAS Institute Inc.; Cary, North Carolina.
SAS Institute Inc. 1989b. SAS/QCR User's Guide, Version
6, First Edition. SAS Institute Inc.; Cary, North Carolina.
34
-------
Satlin, R.L. andP.B. Bedient. 1988. Effectiveness of
Various Aquifer Restoration Schemes Under Variable
Hydrogeologic Conditions. Ground Water 26(4):488-499.
Scalf, M.R., J.F. McNabb, W.D. Dunlap, R.L. Cosby, and J.
Fryberger. 1981. Manual of GroundWater Quality
Sampling Procedures, Robert S. Kerr Environmental
Research Laboratory, U.S. EPA; Ada, Oklahoma.
Scanlon, B.R. 1989. Physical Controls on Hydrochemical
Variability in the Inner Bluegrass Karst Region of Central
Kentucky. Ground Water 27(5):639-646.
Siegrist, H. and PL. McCarty. 1987. Column
Methodologies for Determining Sorption and
Biotransformation Potential for Chlorinated Aliphatic
Compounds in Aquifers. Journal of Contaminant
Hydrology 2(1):31-50.
Smith, L. and R.A. Freeze. 1979a. Stochastic Analysis of
Steady State Groundwater Flow in a Bounded Domain: 1.
One-Dimensional Simulations. Water Resources Research
15(3):521-528.
Smith, L. and R.A. Freeze. 1979b. Stochastic Analysis of
Steady State Groundwater Flow in a Bounded Domain: 2.
Two-Dimensional Simulations. Water Resources Research
15(6):1543-1559.
Spain, J.C., J.D. Milligan, D.C. Downey, and J.K. Slaughter.
1989. Excessive Bacterial Decomposition of H2O2 During
Enhanced Biodegradation. Ground Water 27(2): 163-167.
Spayed, S.E. 1985. Movement of Volatile Organics
Through a Fractured Rock Aquifer. Ground Water
23(4):496-502.
Srinivasan, P. and J. W. Mercer. 1988. Simulation of
Biodegradation and Sorption Processes in Ground Water.
Ground Water 26(4):475-487.
Staples, C.A. and S.J. Geiselmann. 1988. Cosolvent
Influences on Organic Solute Retardation Factors. Ground
Water 26(2): 192-198.
Starr, R.C., R.W. Gillham, and E.A. Sudicky. 1985.
Experimental Investigation of Solute Transport in Stratified
Porous Media: 2. The Reactive Case. Water Resources
Research 21(7): 1043-1050.
Steinhorst, R.K. and R.E. Williams. 1985. Discrimination
of Groundwater Sources Using Cluster Analysis,
MANOVA, Canonical Analysis, and Discriminant Analysis.
Water Resources Research 21(8): 1149-1156.
Stover, E. 1989. Co-produced Ground Water Treatment
and Disposal Options During Hydrocarbon Recovery
Operations. Ground Water Monitoring Review 9(1):75-
82.
Stumm, W. and J.J. Morgan. 1981. Aquatic Chemistry: An
Introduction Emphasizing Chemical Equilibria in Natural
Waters (2nd Ed.). Wiley Interscience/John Wiley & Sons,
Inc.; New York.
Sudicky, E.A., R.W. Gillham, and E.G. Frind. 1985.
Experimental Investigation of Solute Transport in Stratified
Porous Media: 1. The Nonreactive Case. Water Resources
Research 21(7): 1035-1041.
Taylor, J.K. 1987. Quality Assurance of Chemical
Measurements. Lewis Publishers Inc.; Chelsea, Michigan.
Testa, S. and M. Paczkowski. 1989. Volume Determination
and Recoverability of Free Hydrocarbon. Ground Water
Monitoring Review 9(1): 120-128.
Thomas, J.M. and C.H. Ward. 1989. In-situ Biorestoration
of Organic Contaminants in the Subsurface. Environmental
Science and Technology 23(7):760-766.
Thomsen, K., M. Chaudhry, K. Dovantzis, and R. Riesing.
1989. Ground Water Remediation Using an Extraction,
Treatment, and Recharge System. Ground Water
Monitoring Review, 9(l):92-99.
Thorstenson, D.C. and D.W Pollock. 1989. Gas Transport
in Unsaturated Porous Media: The Adequacy of Pick's Law.
Reviews of Geophysics 27(l):61-78.
Tinsley, I.J. 1979. Chemical Concepts in Pollutant
Behavior. John Wiley & Sons Inc.; New York.
Todd, O.K. 1980. Groundwater Hydrology (2nd Ed.). John
Wiley & Sons Inc.; New York.
Tufte, E.R. 1983. The Visual Display of Quantitative
Information. Graphics Press; Cheshire, Connecticut.
U.S. Environmental Protection Agency. 1988a. Guidance
for Conducting Remedial Investigations and Feasibility
Studies Under CERCLA (Interim Final). EPA/540/G-89/
004, Office of Emergency and Remedial Response, U.S.
EPA; Washington, D.C.
U.S. Environmental Protection Agency. 1988b. Guidance
on Remedial Actions for Contaminated Ground Water at
Superfund Sites. EPA/540/G-88/003, Office of Emergency
and Remedial Response, U.S. EPA; Washington, D.C.
35
-------
U.S. Environmental Protection Agency. 1989a. Evaluation
of Ground-Water Extraction Remedies, Volume 1, Summary
Report. EPA/540/2-89/054, Office of Emergency and
Remedial Response, U.S. EPA; Washington, D.C.
U.S. Environmental Protection Agency. 1989b. Criteria for
Identifying Areas of Vulnerable Hydrogeology Under the
Resource Conservation and Recovery Act, Appendix B,
Ground-Water FlowNet/Flow Line Construction and
Analysis (Interim Final). Office of Solid Waste and
Emergency Response, U.S. EPA; Washington, D.C.
U.S. Environmental Protection Agency. 1989d. Statistical
Analysis of Ground-Water Monitoring Data at RCRA
Facilities (Interim Final Guidance). Office of Solid Waste,
Waste Management Division, U.S. EPA; Washington, D.C.
Usunoff, E.J., and A. Guzman-Guzman. 1989. Multivariate
Analysis in Hydrochemistry: An Example of the Use of
Factor and Correspondence Analyses. Ground Water
27(l):27-34.
Valocchi, AJ. 1988. Theoretical Analysis of Deviations
from Local Equilibrium during Sorbing Solute Transport
through Idealized Stratified Aquifers. Journal of
Contamination Hydrology 2(3):191-208.
van der Heijde, P.K.M., Y Bachmat, J.D. Bredehoeft, B.
Andrews, D. Holtz, S. Sebastian. 1985. Groundwater
Management: The Use of Numerical Models, 2nd Edition.
AGU Water Resources Monograph No. 5, American
Geophysical Union; Washington, D.C.
van der Heijde, P.K.M. and P. Srinivasan. 1983. Aspects of
the Use of Graphic Techniques in Ground-Water Modeling.
IGWMC Report No. GWMI 83-11. International Ground
Water Modeling Center, Holcolm Research Institute;
Indianapolis, Indiana.
vanZyl, D.J.A., S.R. Abt, J.D. Nelson, and T.A. Shepherd
(editors). 1987. Geotechnical & Geohydrological Aspects
of Waste Management. Lewis Publishers, Inc.; Chelsea,
Michigan.
Villeneuve, J.P, P. Lafrance, O. Banton, P. Frechette, and C.
Robert. 1988. A Sensitivity Analysis of Adsorption and
Degradation Parameters in the Modeling of Pesticide
Transport in Soils. Journal of Contaminant Hydrology
3(l):77-96.
Vomvoris, E.G. andL.W. Gelhar. 1986. Stochastic
Prediction of Dispersive Contaminant Transport. U.S.
Environmental Protection Agency publication EPA/600/2 -
86/114.
Vouk, V.B., G.C. Butler, D.G. Hoel, andD.B. Peakall.
1985. Methods for Estimating Risk of Chemical Injury:
Human and Non-Human Biota and Ecosystems. John Wiley
& Sons, Inc.; New York.
Walton, WC. 1985. Practical Aspects of Ground Water
Modeling (2nd Ed.). National Water Well Association;
Dublin, Ohio.
Walton, WC. 1987. Groundwater Pumping Tests: Design
& Analysis. Lewis Publishers, Inc.; Chelsea, Michigan.
Wang, H. and M.P Anderson. 1982. Introduction to
Groundwater Modeling: Finite Difference and Finite
Element Methods. WH. Freeman and Company; San
Francisco.
Ward, D.S., D.R. Buss, J.W. Mercer, and S.S. Hughes.
1987. Evaluation of a Groundwater Corrective Action at the
Chem-Dyne Hazardous Waste Site using a Telescopic Mesh
Refinement Modeling Approach. Water Resources
Research 23(4):603-617.
Watson, Ian. 1984. Contamination Analysis - Flow Nets
and the Mass Transport Equation. Ground Water 22(1):31-
37.
Weast, R.C. (editor). 1985. Handbook of Chemistry and
Physics (66th Ed.). CRC Press Inc.; Boca Raton, Florida.
White, J.A., M.H. Agee, and K.E. Case. 1984. Principles of
Engineering Economic Analysis, 2nd Edition. John Wiley
and Sons Inc.; New York.
White, W.B. 1988. Geomorphology and Hydrology of
Karst Terrains. Oxford University Press; New York.
Woodburn, K.B., P.S.C. Rao, M. Fukui, and P. Nkedi-Kizza.
1986. Solvophobic Approach for Predicting Sorption of
Hydrophobic Organic Chemicals on Synthetic Sorbents and
Soils. Journal of Contaminant Hydrology, l(l/2):227-241.
Yeh, H.D., and H.Y Han. 1989. Numerical Identification
of Parameters in Leaky Aquifers. Ground Water 27(5):655-
663.
Young, S.C. and W.R. Waldrop. 1989. An Electromagnetic
Borehole Flowmeater for Measuring Hydraulic
Conductivity Variability. Proceedings of New Field
Techniques for Quantifying the Physical and Chemical
Properties of Heterogeneous Aquifers. National Water Well
Association, Dublin, Ohio.
36
-------
Zheng, C., K.R. Bradbury, and M.P. Anderson. 1988. Role
of Interceptor Ditches in Limiting the Spread of
Contaminants in Ground Water. Ground Water 26(6):734-
742.
37
------- |