EPA/600/R-94/123
June 1994
METHODS FOR MONITORING
PUMP-AND-TREAT PERFORMANCE
by
Robert M. Cohen, Alex H. Vincent, James W. Mercer,
Charles R. Faust, and Charles P. Spalding
GeoTrans, Inc.
Sterling, Virginia 20166
Prepared under subcontract to Dynamac Corporation
EPA Contract No. 68-C8-0058
Project Officer
John Matthews
Extramural Activities and Assistance Division
Robert S. Kerr Environmental Research Laboratory
Ada, Oklahoma 74820
ROBERT S. KERR ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
ADA, OKLAHOMA 74820
Printed on Recycled Paper
-------
NOTICE
The information in this document has been funded wholly or in part by the United States
Environmental Protection Agency under contract 68-C8-0058 to Dynamac Corporation. This report has
been subjected to the Agency's peer and administrative review and has been approved for publication as
an EPA document. Mention of trade names or commercial products does not constitute endorsement or
recommendation for use.
All research projects making conclusions or recommendations based on environmentally related
measurements and funded by the Environmental Protection Agency are required to participate in the
Agency Quality Assurance Program. This project did not involve environmentally related measurements
and did not involve a Quality Assurance Project Plan.
11
-------
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 soil 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 the 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 the soil arid subsurface environment, for the protection of this resource.
Since the 1980s, numerous pump-and-treat systems have been constructed to: (1) hydraulically
contain contaminated ground water, and/or, (2) restore ground-water quality to meet a desired standard
such as background quality or Maximum Contaminant Level (MCL) concentrations for drinking water.
Although hydraulic containment is usually achievable, experience proves that aquifer restoration will be
hindered at many sites due to Non-Aqueous Phase Liquid (NAPL) dissolution, contaminant desorption,
inefficient hydraulic flushing of heterogeneous media, and other chemical and physical process
limitations. Given the complexity and site-specific nature of ground-water remediation, pump-and-treat
system objectives must be clearly identified and system operation carefully monitored to determine
effectiveness. Typically, monitoring involves measuring hydraulic heads and contaminant concentrations
to evaluate ground-water flow directions, recovery system capture zones, contaminant migration, and
contaminant removal. This document was developed on behalf of the United States Environmental
Protection Agency (EPA) to outline methods for evaluating the effectiveness and efficiency of pump-and-
treat remediation systems.
Clinton W. Hall
Director
Robert S. Kerr Environmental
Research Laboratory
-------
-------
TABLE OF CONTENTS
Page
Notice n
Foreword '"
List of Figures v11
List of Tables x
1. Introduction 1
1.1 Pump-and-Treat Objectives '
1.2 Tailing and Rebound Constraints 4
1.3 How is Success Measured? 9
1.4 Purpose and Format of Report 11
2. Monitoring Hydraulic Containment 13
2.1 Objectives and Process.. 13
2.2 Performance Monitoring Measurements and Interpretation 13
2.2.1 Inward Hydraulic Gradients and Capture Zone Analysis 14
2.2.1.1 Performance Concept 14
2.2.1.2 Methods 15
2.2.1.3 Measurement Locations 15
2.2.1.4 Measurement Frequency 17
2.2.1.5 Some Additional Considerations 19
2.2.2 Vertical Hydraulic Gradients 24
2.2.3 Hydraulic Head Differences 25
2.2.4 Flow Meters 25
2.2.5 Pumping Rates.. 26
2.2.6 Ground-Water Chemistry 26
2.2.6.1 Performance Concept 26
2.2.6.2 Ground-Water Quality Monitoring Locations 27
2.2.6.3 Ground-Water Quality Monitoring Frequency 28
2.2.7 Perimeter Monitoring Using Noninvasive Methods 29
2.2.8 Tracers 30
2.3 Monitoring Location Summary 3 0
2.4 Operations and Maintenance (O&M) Manual 30
2.5 P&T Monitoring Plan 31
2.6 Capture Zone Analysis and Optimization Modeling 31
2.7 Operational Efficiency 35
3. Monitoring Aquifer Restoration 39
3.1 Introduction 39
3.2 Performance Measurements and Interpretation 39
3.2.1 Hydraulic Containment 39
3.2.2 Managing Ground-Water Flow 39
3.2.2.1 Pore Volume Flushing 4 0
3.2.2.2 Minimize Ground-Water Stagnation 41
-------
3.2.2.3 Pulsed Pumping 43
3.2.2.4 Contain the NAPL Zone 44
3.2.3 Contaminant Monitoring 44
3.2.3.1 Ground-Water Sampling and Analysis 45
3.2.3.2 Sampling Aquifer Material 48
3.2.3.3 Treatment System Influent and Effluent 48
3.2.4 Restoration Measurement Frequency Summary 49
3.2.5 Evaluating Contaminant Concentration and Distribution Trends 49
3.2.5.1 Estimating Contaminant Mass-in-Place 49
3.2.5.2 Determining Rate of Contaminant Mass Removal 51
3.2.5.3 Comparing Mass Removal and Mass-in-Place Trends 52
3.3 Projected Restoration Time 52
3.4 Additional Considerations 56
4. Evaluating Restoration Success/Closure 57
4.1 Introduction 57
4.2 System Operation: Short-Term Analyses 59
4.2.1. Parametric Tests 60
4.2.2. Nonparametric Tests 61
4.3 Treatment Termination: Long-Term Analyses 63
4.3.1 Parametric Trend Analyses 63
4.3.2 Nonparametric Trend Analyses 69
4.3.3 Time Series Analysis 69
4.4 Post-Termination Monitoring 72
4.5 Monitoring for Attainment 73
4.6 Conclusions 73
Chem-Dyne Site Case Study 75
5.1 Background 75
5.2 Performance Criteria ..' 77
5.3 Performance Monitoring 77
5.3.1 Hydraulic Head Monitoring 78
5.3.2 Water-Quality Monitoring 78
5.3.3 Monitoring Schedule , 79
5.4 Data Evaluation and Effectiveness 80
5.4.1 Containment 80
5.4.2 Restoration 80
5.4.3 Termination 86
5.4.4 Post Termination Monitoring 92
6. References 95
VI
-------
LIST OF FIGURES
Page
1-1. Examples of hydraulic containment using P&T technology 2
1 -2. Plan view of the mixed containment-restoration strategy. P&T is used to contain
ground-water contamination source areas (e.g., where NAPL or wastes may be present)
and attempt aquifer restoration downgradient 4
1 -3. Concentration versus pumping duration or volume showing tailing and rebound effects
(modified from Keely, 1989) 5
1 -4. Hypothetical examples of contaminant removal from ground water using P&T
(modified from Mackay and Cherry, 1989) 6
1-5. The Langmuir and Freundlich adsorption isotherms (modified from
Palmer and Fish, 1992) 7
1 -6. Relationship between ground-water velocity induced by pumping and the concentration
of dissolved contaminants that (a) desorb from the porous media, (b) dissolve from
precipitates, or (c) dissolve from NAPL (modified from Keely, 1989).
Kinetic limitations to dissolution exacerbate tailing 8
1 -7. Dissolved contaminant concentration in ground water pumped from a recovery well
versus time in a formation that contains a solid phase contaminant precipitate
(from Palmer and Fish, 1992) 9
1-8. Advective velocity, flowpath, and travel time variations (a) to a recovery well
(from Keely, 1989) and (b) induce tailing {from Palmer and Fish, 1992) 10
2-1. Components of a phased design and implementation of a P&T monitoring program 14
2-2. In isotropic media, ground-water flow (b) is orthogonal to hydraulic
head contours '. 15
2-3. Inward gradients are often monitored by comparing hydraulic heads in paired
piezometers near the containment perimeter and primarily in the pre-pumping
downgradient direction 16
2-4. Cross-section showing equipotential contours and the vertical capture zone associated
with ground-water withdrawal from a partially-penetrating well in isotropic media 18
2-5. Near continuous hydraulic head measurements were made in several observation
wells in the vicinity of a recovery well line to examine the transient water table
response to pump cycles and recharge events (modified from ESE, 1992) 19
2-6. Example display of ground-water flow directions and hydraulic gradients determined
between three observation wells 20
2-7. Ground water flows between and beyond the recovery wells even though hydraulic
heads throughout the mapped aquifer are higher than the pumping level 20
2-8. Ground-water and LNAPL flow in anisotropic saprolite soil from a petroleum-product
tank farm in Fairfax, Virginia is offset from the hydraulic gradient toward the strike
of saprolite foliation , 22
2-9. As the initial steep drawdown induced by pumping flattens out with time, hydraulic
containment may be diminished significantly as exemplified by the hydraulic head
contours associated with a line of recovery wells after (a) 5 days of pumping and
(b) after 30 days of pumping 23
2-10. Vertical hydraulic gradients across an aquitard between aquifers are typically
measured using observation well nests 25
vn
-------
2-11. Surface geophysical (EM-Conductivity) surveys were conducted periodically
along transects between monitor wells encircling a sanitary landfill in Maryland
to augment the leak detection monitoring network (from Rumbaugh et al., 1987) 29
2-12. Equations for the dividing streamlines (w = Q/2T) that separate the capture zone
of a single well from the rest of an isotropic, confined aquifer with a uniform
regional hydraulic gradient (modified from Gorelick et al., 1993) 32
2-13. Type curves showing the capture zones of 1 (a), 2 (b), 3 (c), and 4 (d) pump wells
spaced evenly along the y-axis for several values of Q/BU (where Q = pumping rate
(L3/T), B = aquifer thickness (L), and U = Darcy velocity for regional flow (L/T)
(from Javandel and Tsang, 1986) 33
2-14. Example of steady-state, and 10-year and 25-year time-related capture zones
delineated using reverse particle tracking (from Blandford and Huyakorn, 1989) 34
2-15. An example of the Maximal Covering Location Problem applied to monitor well
network design (from Meyer, 1992) 36
2-16. A framework for risk-based decision making regarding P&T system design
and monitoring (modified from Freeze et al., 1990) 37
2-17. The concept of optimal risk (from Freeze et al., 1990) 38
3-1. Examples of stagnation zones associated with single-well and five-spot pumping schemes 42
3-2. Conceptualized ground-water flow patterns and stagnation zones superimposed
on a total VOC isoconcentration contour map at the Lawrence Livermore National
Laboratory site in California (from Hoffman, 1993) 43
3-3. Adaptive modifications to P&T design and operation can reduce clean-up time 44
3-4. The pulsed pumping concept (modified from Keely, 1989) 45
3-5. Example of a ground-water monitoring well network at a P&T remediation site
(modified from USEPA, 1992a) 46
3-6. Simulated trends of VOC concentration in ground water pumped from seven
extraction wells during a P&T operation 47
3-7. Comparison of cumulative mass of TCE removed versus dissolved mass-in-place
at the Air Force Plant 44 in Tucson, Arizona 53
4-1. Determining the success and/or closure of a P&T system 58
4-2. Example contaminant concentrations in a well at P&T site (USEPA, 1992c) 59
4-3. Best-fit regression line and 95% confidence interval for the concentration trend
of data given in Table 4-2 68
4-4. Best fit regression line and 95% confidence interval for the concentration trend
in Table 4-3 71
5-1, Boundary of 0.1 ppm total VOC plume and location of nested piezometers at the
Chem-Dyne site 76
5-2. Nested piezometer hydrograph for 1992 at the Chem-Dyne site (from Papadopulos
&ASSOC., 1993) 81
5-3. Average water table and direction of ground-water flow in the shallow interval
in 1992 at the Chem-Dyne site (from Papadopulos & Assoc., 1993) 82
5-4. Average potentiometric surface and direction of ground-water flow in the intermediate
interval in 1992 at the Chem-Dyne site (from Papadopulos & Assoc., 1993) 83
5-5. Influent and effluent VOC concentrations (mg/L) at the Chem-Dyne treatment
plant from 1987 to 1992 (from Papadopulos & Assoc., 1993) 85
5-6. Cumulative mass of VOCs removed from the aquifer at the Chem-Dyne
site from 1987 to 1992 (from Papadopulos & Assoc., 1993) 87
5-7. Concentrations of VOCs in the shallow interval in December 1992 at the
Chem-Dyne site (from Papadopulos & Assoc., 1993) 88
Vlll
-------
5-8. Concentrations of VOCs in the intermediate interval in December 1992 at the
Chem-Dyne site (from Papadopulos & Assoc., 1993) 89
5-9. Concentrations of VOCs in ug/L in the shallow interval during October/November
1987 (from Papadopulos & Assoc., 1988) 90
5-10. Concentrations of VOCs in ug/L in the intermediate interval during October/November I987
(from Papadopulos & Assoc., 1988) 91
-------
LIST OF TABLES
Page
1-1. Summary of selected ground-water treatment technologies 3
4-1. Distribution of monochlorobenzene 61
4-2. Concentration versus time data showing an asymptotic zero slope (regression only
performed on last six samples.) 67
4-3. Concentration versus time data showing a downward trend 69
5-1. Annual mass of VOCs and volume of ground water extracted from the Chem-Dyne Site
(Papadopulos & Assoc., 1988 and 1993) 84
-------
EVALUATING GROUND-WATER PUMP-AND-TREAT SYSTEMS
Abstract
Since the 1980s, numerous pump-and-treat systems have been constructed to: (1) hydraulically
contain contaminated ground water, and/or, (2) restore ground-water quality to meet a desired standard
such as background quality or Maximum Contaminant Level (MCL) concentrations for drinking water.
Although hydraulic containment is usually achievable, experience suggests that aquifer restoration can
often be hindered at many sites due to the dissolution of Non-Aqueous Phase Liquids (NAPLs),
contaminant desorption, inefficient hydraulic flushing of heterogeneous media, and other chemical and
physical process limitations. Given the complexity and site-specific nature of ground-water remediation,
pump-and-treat system objectives must be clearly identified and system operations carefully monitored to
determine effectiveness. Typically, monitoring involves measuring hydraulic heads and contaminant
concentrations to evaluate ground-water flow directions, recovery system capture zones, contaminant
migration, and contaminant removal. This document was developed on behalf of the United States
Environmental Protection Agency (EPA) to outline methods for evaluating the effectiveness and
efficiency of pump-and-treat remediation systems.
XI
-------
-------
1. INTRODUCTION
1.1 PUMP-AND-TREAT OBJECTIVES
Although this document focuses on the containment or remediation of contaminated ground water
using pump and treat (P&T) systems, other technologies are discussed in a limited way, particularly as
they are used in concert with P&T systems. It is important to note that in the selection and
implementation of any remediation system, or consortia of systems which are designed to contain or
remediate contaminated ground water, that the sources of contaminants must be removed from the site or
sufficiently isolated to assure that they can no longer contribute contaminants to the ground water.
A common remedial strategy to deal with contaminated ground water is to extract the
contaminated water and treat it at the surface prior to discharge or reinjection. This is referred to as
conventional pump-and-treat (P&T) remediation. An overview of pump-and-treat ground-water
remediation technology is provided by Mercer et al. (1990). Between 1982 and 1990, 72 percent (3 14) of
all Superfund site Records of Decisions (RODs) addressing ground-water remediation specified P&T
technology (Steimle, 1992).
P&T systems are designed to: (1) hydraulically contain and control the movement of
contaminated ground water to prevent continued expansion of the contamination zone; (2) reduce
dissolved contaminant concentrations to comply with clean-up standards and thereby "restore'* the
aquifer; or (3) a combination of these objectives.
Hydraulic containment of dissolved contaminants by pumping ground water from wells or drains
has been demonstrated at numerous sites. The concept is illustrated in Figure 1-1. Fluid injection (using
wells, drains, or surface application) and physical containment options (such as subsurface barrier walls
and surface covers) can enhance hydraulic containment systems. Recovered ground water is usually
treated at the surface using methods selected to remove the contaminants of concern (Table 1-1). In many
cases, hydraulic containment systems are designed to provide long-term containment of contaminated
ground water at the lowest cost by optimizing well, drain, surface cover, and/or cut-off wall locations and
by minimizing pumping rates.
P&T designed for aquifer restoration generally combines hydraulic containment with more active
manipulation of ground water (i.e., higher pumping rates) to attain ground-water clean-up goals during a
finite period. As described below, aquifer restoration is much more difficult to achieve than hydraulic
containment.
Selection of P&T objectives depends on site conditions and remedial goals. Hydraulic
containment is preferred where restoration is technically impracticable (e.g., not capable of being done or
carried out) due to the presence of subsurface NAPL, buried waste, formation heterogeneity, or other
factors (USEPA, 1993). Aquifer restoration may be an appropriate goal where these confounding factors
are absent or minimal. At many sites, P&T systems can be used to contain contaminant sources areas and
attempt restoration of downgradient dissolved contaminant plumes (Figure 1-2).
-------
(a)
Static Water Table
Plume
Drain
Capture
Zone Limit
Downgradient
Barrier Wall
Upgradient
Barrier Wall
Figure 1-1. Examples of hydraulic containment in plan view and cross section using P&T technology:
(a) pump well, (b) drain, and (c) well within a barrier wall system.
-------
TABLE 1-1. SUMMARY OF SELECTED GROUND-WATER TREATMENT TECHNOLOGIES
(FROM BOUWER ET AL., 1988).
Ground- Water
Treatment Technology
ORGANIC CONTAMINANTS:
Air stripping
Liquid-phase
Steam stripping
Membranes
Oxidation
Activated sludge
Fixed-film biological
reactors
Biophysical
INORGANIC CONTAMINANTS:
Alkaline precipitation
Coagulation
lonexchange
Adsorption
Filtration
Reduction
Membranes
Oxidation
Representative
Examples
Packed towers, surface or
diffused aeration removal
of volatile compounds;
soil venting
GAC removal of broad
spectrum of VOCs
Packed tower with steam
stripping, removal of low
volatile organics
Ultrafiltration for removal of
selected organics
Ozone/UV. or ozone/H202,
destruction of chlorinated
organics
Oxygen or air biological
oxidation for removal/
destruction of degradable
organics
Fixed-film fluidized bed, for
oxidation of less degradable
organics
Powdered carbon, with
activated sludge, treatment of
high strength wastewaters
Heavy metals removal
Ferric sulfate or alum for
heavy metals removal
Heavy metals; nitrate
Selenium removal on
activated alumina
Removal of clays, other
particulates
SO, reduction of CR(VI)
Reverse osmosis, ultrafiltration
for removal of metals, other ions
Fe(II) and Mn(II)
Residual
Streams
Air stream with
VOCs
GAC for
regeneration or
disposal
Recovered solvent
Concentrated brine
side stream
None
Sludge
Sludge
Powdered carbon
and bacterial
Hazardous sludge
Hazardous sludge
Regeneration stream
Regeneration stream
Backwash wastes
Chromium sludge
Concentrated liquid
waste
Sludge
status of
Technology
Commercial
Commercial
some
commercial
Commercial
some
commercial in
development
stages
Commercial
Commercial
Commercial,
PACT process
Commercial
Commercial
Commercial
Commercial
Commercial
Commercial
Commercial, new
membranes under
development
Commercial
-------
Contaminant
source area
Limit of dissolved plume
\
7
Barrier wall
Initial groundwater flow direction
O Injection well
• Extraction well
Figure 1-2. Plan view of the mixed containment-restoration strategy. P&T is used to contain ground-water
contamination source areas (e.g., where NAPL or wastes may be present) and attempt aquifer
restoration downgradient.
1.2 TAILING AND REBOUND CONSTRAINTS
Although P&T systems continue to be widely used to reduce dissolved contaminants in ground
water, experiences gained in recent years suggest that the efficiency of these systems can be compromised
by a number of factors that are related to the contaminants of interest and characteristics of the site. As a
result, it is often difficult to reduce dissolved contaminants to below drinking-water standards in
reasonable time frames (e.g., less than 10 years) at many sites (Palmer and Fish, 1992; CH2M Hill, 1992;
Haley et al., 1991; Mercer et al., 1990; Mackay and Cherry, 1989; Keely, 1989; Harman et al., 1993; Doty
and Travis, 1991). Monitoring contaminant concentrations in ground water with time at P&T sites
reveals "tailing" and "rebound" phenomena. "Tailing" refers to the progressively slower rate of dissolved
contaminant concentration decline observed with continued operation of a P&T system (Figure 1-3). At
many sites, the asymptotic, apparent residual, contaminant concentration exceeds clean-up standards.
Another problem is that dissolved contaminant concentrations may "rebound" if pumping is discontinued
after temporarily attaining a clean-up standard (Figure 1-3).
Tailing and rebound may result from several physical and chemical processes that affect P&T
remediation (Figure 1-4).
Non-Aqueous Phase Liquid (NAPL) dissolution — Subsurface NAPLs can be long-term
sources of ground-water contamination due to their limited aqueous solubility that may
greatly exceed drinking water standards (Cohen and Mercer, 1993). This long-term
contamination potential is illustrated in Figure l-4(d). If NAPLs are not removed (i.e., by
excavation) or contained, tailing and rebound will occur during and after P&T operation,
respectively, in and downgradient of the NAPL zone. The dissolution of a NAPL source may
require the removal of thousands of equivalent pore volumes.
-------
Max
o
S
I
o
O
0)
*3
19
o>
DC.
Pumping on
Pumping
off
Theoretical removal
without tailing
Removal with tailing
Apparent residual \
contaminant concentration \
Cleanup Standard
Pumping Duration or Volume Pumped
Figure 1-3. Concentration versus pumping duration or volume showing tailing and rebound effects
(modified from Keeiy, 1989).
Contaminant desorption — As dissolved contaminant concentrations are reduced by P&T
system operation, contaminants sorbed to subsurface media desorb from the matrix into
ground water. This equilibrium partitioning process can be described by the Langmuir
isotherm,
Cs = Csmax[(KCw)/(l+KCw)]
or the Freundlich sorption isotherm,
= K Cw"
(1-1)
(1-2)
where Cs and Cw
are the contaminant concentrations associated with the solid and aqueous
phases, respectively, K is the adsorption constant, Csmax is the maximum possible soil
contaminant concentration, and n is a measure of nonlinearity (Figure 1 -5). For the linear
isotherms (n = 1) and for limited ranges of Cw, particularly at low concentration, where n * 1,
the Freundlich constant can be identified as a distribution ratio, Kd, such that
'CS/CW
(1-3)
The Kd values for hydrophobic, nonpolar organic contaminants are frequently represented as
the product of the organic carbon content of the media, f()C (mass of carbon/mass of soil), and
the organic carbon partition coefficient, K(K. (mass of contaminant per unit mass of carbon/
equilibrium concentration in soil) such that
~ K
(1-4)
Values for f(X. and K(K, may be obtained from laboratory analyses of core material and
literature sources (USEPA, 1990), respectively. By assuming a linear isotherm, these
-------
(a) Uniform sand-gravel aquifer
to
t1
Contaminant concentration in extracted
water
tO t1
Time
(b) Stratified sand-gravel aquifer ,
to ~
t1
to ti
(c) Clay lens in uniform sand-gravel aquifer
to
t1
(d) Uniform sand-gravel aquifer
to
t2
tO t1
to
t2
Figure 1-4. Hypothetical examples of contaminant removal from ground water using P&T (modified from
Mackay and Cherry, 1989). Black indicates NAPL presence; stippling indicates contaminant in
dissolved and sorbed phases (with uniform initial distribution); and arrows indicate relative
ground-water velocity. Ground water is pumped from the well at the same rate for each case.
Note that the dotted lines in (a) represent the volume of ground water that would have to be
pumped to flush slightly retarded contaminants from the uniform aquifer.
-------
II
•Csmax
Langmuir isotherm:
Cs = Csmax
1+KCw
Aqueous Concentration
Aqueous Concentration
Figure 1-5. The Langmuir and Freundlich adsorption isotherms (modified from Palmer and Fish, 1992).
relationships can be used to estimate: (1) the retardation factor, Rf, or velocity of dissolved
contaminant movement, vc, relative to ground-water flow,
(2) the retardation coefficient, R, which is the reciprocal of Rf,
R - 1 + (Kdp,/n)
(1-5)
(1-6)
and (3) the equilibrium distribution of contaminant mass between the solid and aqueous
phases
CWVW
[(CWVW)-KCSMS)] = Vw / (Vw + KdMs) (1-7)
where pb is the dry bulk density, n is the porosity, Vw is the volume of water in the total
subject volume, Ms is the mass of solids in the total subject volume, and fw is the fraction of
mass residing in the aqueous phase.
Sorption and retardation are site-specific. Field retardation values vary between different
contaminants at a given site and between different sites for a given contaminant (Mackay and
Cherry, 1989). As illustrated in Figure 1-4, desorption and retardation increase the volume of
ground water which must be pumped to attain dissolved contaminant concentration
reductions. Tailing and rebound effects will be exacerbated where desorption is slow relative
to ground-water flow and kinetic limitations prevent sustenance of equilibrium contaminant
concentrations in ground water (Palmer and Fish, 1992; Haley et al., 1991; Brogan, 1991;
Bahr, 1989). This concept is illustrated in Figure 1-6. Kinetic limitations to mass transfer are
likely to be relatively significant in the high ground-water velocity 'zone in the vicinity of
7
-------
c
CO
c
Eg
CO :j3
1
> o
oO
(0
(0
-^-- Equilibrium Concentration
.. 4^H» •»»»»••*•• •«^^^^HHBHH^^^^^^^^^^^M^M^^^^ •••»
\
Long contact time produces
equilibrium partitioning
concentrations
Kinetic limitations limit
dissolved concentrations
-Contact Time-
— Groundwater Velocity—
Figure 1-6. Relationship between ground-water velocity induced by pumping and the concentration of
dissolved contaminants that (a) desorb from the porous media, (b) dissolve from precipitates, or
(c) dissolve from NAPL (modified from Keely, 1989). Kinetic limitations to dissolution
exacerbate tailing.
injection and extraction wells. Under such conditions, insufficient control time is available
between the adsorbed contaminants and ground water to allow the development of maximum
concentrations.
Precipitate dissolution — Large quantities of inorganic contaminants, such as chromate in
BaCrG4, may be bound with crystalline or amorphous precipitates on porous media (Palmer
and Fish, 1992). Dissolution of contaminant precipitates may cause tailing (Figure 1-7) and
rebound. These effects may increase due to mass transfer limitations where the dissolution
rate is slow relative to ground-water flow.
Ground-water velocity variation - Tailing and rebound also result from the variable travel
times associated with different flow paths taken by contaminants to an extraction well
(Figures 1-4 and 1-8). Ground water at the edge of a capture zone travels a greater distance
under a lower hydraulic gradient than ground water closer to the center of the capture zone.
Additionally, contaminant-to-well travel times vary as a function of the initial contaminant
distribution and differences in hydraulic conductivity. If pumping is stopped, rebound will
occur wherever the resulting flow path modification causes the magnitude of contaminant
dilution to be reduced.
Matrix diffusion - As contaminants advance through relatively permeable pathways in
heterogeneous media, concentration gradients cause diffusion of contaminant mass into the
less permeable media (Gillham et al., 1984). Where contamination persists for long periods,
this diffusion may cease when contaminant concentrations equilibrate between the different
strata. During a P&T operation, dissolved contaminant concentrations in the relatively
permeable zones may be quickly reduced by advective flushing relative to the less permeable
zones as illustrated in Figure l-l(c). This causes a reversal in the initial concentration
8
-------
gradient and the slow diffusion of contaminants from the low to high permeability media.
This slow process can cause long-term tailing, and rebound after the termination of pumping.
Tailing and rebound patterns associated with these different physical and chemical processes are
similar. Multiple processes (i.e., dissolution, diffusion and desorption) will typically be active at a P&T
site. Diagnosis of the cause of tailing and rebound, therefore, requires careful consideration of site
conditions and usually cannot be made by examination of concentration versus time data alone.
1.2
§0.8
o
O
1 0.6
a 0.4
I 0.2
Contaminant concentration
controlled by solubility
Solid phase
reserve deplete'
-Pumping Duration or Volume Pumped
Figure 1-7. Dissolved contaminant concentration in ground water pumped from a recovery well versus time
in a formation that contains a solid phase contaminant precipitate (from Palmer and Fish, 1992).
1.3 HOW IS SUCCESS MEASURED?
A successful P&T system is a design and implementation that has been determined capable of
accomplishing the remedial action objectives of containment and/or restoration in a desired time period.
For containment, success is usually defined as the achievement of hydrodynamic control at the outer
limits (horizontal and vertical) of the contaminant plume such that hydraulic gradients are inward to the
pumping system. Measuring the effectiveness of a restoration program is generally more difficult due to:
(1) limitations of methods used to estimate contaminant mass distribution prior to and during remediation,
and (2) the inherent difficulty of aquifer restoration as discussed in the previous section.
Tracking the performance of a containment or restoration P&T system is achieved by setting
performance criteria, monitoring to assess these criteria, and assessing operational efficiency.
Performance measures such as induced hydraulic gradients and contaminant concentration reductions are
monitored to verify that the system is operating as designed and achieving remediation goals. If the
performance criteria have not been adequately formulated, perhaps due to a flawed site conceptual model,
-------
(a)
MODERATE
FAST
MODERATE
(b)
1.2
0.8
8
o
O
1 0.6
o
*„„
Q 0.4
.1
0.2
Tailing due to different travel
times along flow paths to
recovery well
-Pumping Duration or Volume Pumped
Figure 1-8. Advective velocity, flowpath, and travel time variations (a) to a recovery well (from Keely, 1989)
and (b) induce tailing (from Palmer and Fish, 1992).
10
-------
then meeting specified criteria may provide a misleading sense of system effectiveness. Operational
efficiency is also a measure of success for a P&T system, It refers to the cost-effectiveness of a system,
and can be measured by monitoring costs and assessing related environmental benefits. For example, a
highly efficient and cost-effective hydraulic containment system may extract ground water at the
minimum rate required to demonstrate attainment of hydraulic gradient objectives. Ideally, a phased
remedial approach, whereby system improvements evolve from performance monitoring, will maximize
both the performance effectiveness and efficiency of a P&T system.
1.4 PURPOSE AND FORMAT OF REPORT
The purpose of this report is to provide guidance for monitoring the effectiveness and efficiency
of P&T systems. Related complementary guidance is given by USEPA (1992a). Emphasis herein is
placed on the "pump" portion of P&T technology. Chemical enhancements to P&T remediation, such as
injection of cosolvents or surfactants, are discussed by Palmer and Fish (1992). For details on ground-
water treatment techniques and strategies, see AWWA (1990), Nyer (1992), and USEPA (1987), among
others. It is assumed that the reader is familiar with basic concepts of hydrogeology and P&T technology.
The report is divided into six main sections: (1) Introduction, (2) Monitoring Hydraulic
Containment, (3) Monitoring Ground-Water Restoration, (4) Evaluating Restoration Success/Closure, (5)
A Case Study, and (6) References. Examples and illustrations are provided to convey concepts. This
section provides an overview of P&T use, objectives, and limitations. Sections 2 and 3 describe
performance criteria, monitoring objectives, data analysis, system enhancements, and protocols for
evaluating the effectiveness of the P&T systems designed for containment and restoration, respectively.
Methods for determining the timing of system closure are addressed in Section 4. In Section 5,
monitoring data from the Chem-Dyne site in Hamilton, Ohio are presented as an example of a P&T
system effectiveness evaluation.
11
-------
-------
2. MONITORING HYDRAULIC CONTAINMENT
2 . 1 OBJECTIVES AND PROCESS
Monitoring programs are designed to measure the effectiveness and efficiency of P&T system
performance in achieving hydraulic containment objectives. For successful hydraulic containment,
contaminants moving with ground water in the containment zone must follow pathlines that are captured
by the P&T system (Figure 1-1). In addition to P&T systems designed to remove dissolved contaminants
and contaminants that may be adsorbed to mobile colloids, remedial designs should be developed to
preclude the migration of NAPLs, if present, beyond the containment perimeter.
In general, containment monitoring involves: (1) measuring hydraulic heads to determine if the
P&T system affects hydraulic gradients in such a way as to prevent ground-water flow and dissolved
contaminant migration across the containment zone boundary; and (2) ground-water quality monitoring to
determine if temporal and spatial variations in contaminant distribution are consistent with hydraulic
containment (i.e., no contaminant movement or increase of contaminant mass across the containment
zone boundary). Containment monitoring activities, therefore, typically include some combination of
hydraulic head measurement, ground-water sampling and analysis, tracer monitoring, and pumping rate
measurement.
Containment monitoring plans are developed and revised during a phased remedial program. As
outlined in Figure 2-1, the first step in establishing performance criteria, after characterizing pre-remedy
ground-water flow patterns and contaminant distributions, is to determine the desired containment area
(two-dimensional) and volume (three-dimensional). These should be clearly specified in site remedial
action and monitoring plans.
At any particular site, there may be multiple separate containment areas, or a contaminant source
'containment area within a larger dissolved plume containment area (e.g., Figure 1-2), or a containment
area that does not circumscribe the entire ground-water contamination zone. As shown in Figure 1-1,
barrier walls are often used along the containment perimeter, while drains and recovery wells are located
within the containment area. After defining the containment area, a capture zone analysis (Section 2.6) is
conducted to design a P&T system and a performance monitoring plan is developed based on the
predicted flow system (Figure 2-1). The monitoring plan may be revised as improvements to the site
conceptual model and the P&T system evolve, and, if the containment area/volume is modified based on
changes in contaminant distribution with P&T operation.
2.2 PERFORMANCE MONITORING MEASUREMENTS AND INTERPRETATION
Various hydraulic containment performance criteria are described in this section. Monitoring of
these criteria is done to determine if the containment system is functioning as designed and to provide
guidance for P&T system optimization. Performance is monitored by measuring hydraulic heads and
determining gradients, ground-water flow directions, pumping rates, ground-water chemistry, and,
possibly, tracer movement.
13
-------
Characterize Initial Groundwater Flow
Field and Contaminant Distribution
[identify Desired Containment Area/Volume |
Specify Monitoring Locations and Performance Criteria
Conduct Capture Zone and Remedy Analyses |
I
[Determine Initial P&T Design |
i
[initiate Phased P&T Implementation^
* ^
[Operate and Monitor P&T System!]
Modify P&T System and Monitoring Based on
Conceptual Model Improvements
Figure 2-1. Components of a phased design and implementation of a P&T monitoring program.
2.2.1 Inward Hydraulic Gradients and Capture Zone Analysis
2.2.1.1 Performance Concept
Inward hydraulic gradients across the boundary of, and/or within, the desired containment area
may be specified as part of the performance standard. An inward gradient indicates that the ground-water
flow is inward, thus allowing the capture of dissolved contaminants by the P&T system.
Hydraulic head and gradient data are interpreted within the context of capture zone analysis
(Section 2.6). The capture zone concept is illustrated in Figure 2-2. Note that the capture zone of a well
is not coincident with its zone of influence (ZOI) except in those incidences where the hydraulic gradient
is negligible prior to pumping. Therefore, there can be locations in the vicinity of a pumping well where
a drawdown within that well does-not indicate that the ground water will be contained by the capture
zone. It should also be noted that successful containment does not require the establishment of inward
hydraulic gradients all around the containment zone when it is larger than the contaminated zone. In
either case, the subsurface volume showing inward hydraulic gradients will not correspond to the actual
capture volume (Larson et al., 1987).
14
-------
(a)
(c) (d)
Figure 2-2. In isotropic media, ground-water flow lines (b) are orthogonal to hydraulic head contours
(a) (modified from Gorelick et al., 1993). Pumping causes drawdowns and a new steady-state
potentiometric surface (c). Following the modified hydraulic gradients, ground water within the
shaded capture zone flows to the pump well (d). The stagnation point is designated sp.
2.2.1.2 Methods
Depth-to-water measurements can generally be made to +/- 0.01 or 0.02 ft. The accuracy of
depth-to-water measurement methods is discussed by Thornhill (1989), USGS (1977), and Dalton et al.
(1991). Well reference point elevations should be surveyed to +/- 0.01 ft and checked periodically due to
the potential for settlement of surface materials, compaction of pumped strata, or physical damage to the
well. This is particularly important when measuring small head differences because the flow direction
may be misinterpreted due to slight elevation errors.
2.2.1.3 Measurement Locations
In relatively simple hydrogeologic settings inward hydraulic gradients can be estimated by
comparing hydraulic heads in paired piezometers near the containment perimeter, primarily in the pre-
pumping downgradient direction (Figure 2-3). For more complex flow systems, this may not always be
true and gradients can only be determined by using three or more wells. Capture zone analysis
incorporating aquifer tests and potentiometric surface data should be used to help select inward gradient
control monitoring locations.
15
-------
Plan View
t
Containment Perimeter
•r Gradient w
Control
Monitor Well
Pairs
CD
*<
03
3
o
Cross Section
Figure 2-3. Inward gradients are often monitored by comparing hydraulic heads in paired piezometers near
the containment perimeter and primarily in the pre-pumping downgradient direction.
Inward gradients can also be evaluated by interpreting potentiometric surface maps developed
using all available and comparable hydraulic head data (measured in wells within and outside of the
containment area). Since ground-water flow is perpendicular to the equipotential lines in the direction of
decreasing potential, containment is inferred if flow lines at the containment boundary converge at
extraction wells. However, it is critical that potentiometric surface maps be developed using hydraulic
heads measured in comparable stratigraphic intervals to avoid misinterpreting horizontal flow directions,
especially where significant vertical gradients are present. For this reason, care should be exercised with
regard to incorporating measurements from wells with unknown or inconsistent completions.
Potentiometric surface maps developed from wells completed in different geologic units may result in
misleading interpretations and containment.
In addition to focusing on the downgradient side of the plume, containment boundary monitoring
should also target the more permeable portions of the subsurface. Ground-water flow and contaminant
migration occur preferentially in these zones. Ideally, the spatial distribution of preferential pathways
will be identified during the remedial investigation. However, additional site characterization may be
warranted to allow adequate performance monitoring.
16
-------
Hydraulic gradients across the containment volume should be measured in three dimensions.
This may be difficult to accomplish in areas lacking a sufficient number of observation wells to define the
convoluted potentiometric surface that may develop due to complex site conditions (i.e., multiple
pumping or injection wells, heterogeneity, anisotropy, transient effects, etc.). In addition to horizontal
flow divides near pumping wells, flow divides also exist in the vertical dimension (Figure 2-4) because
the hydraulic influence of each well extends only a limited depth (Larson et al., 1987; Keely, 1989). As
shown in Figure 2-4, capture zone volume may be misinterpreted by neglecting vertical hydraulic
gradients. Monitoring vertical hydraulic gradients is discussed further in Section 2.2.2.
In general, the number of observation wells needed to evaluate hydraulic containment increases
with site complexity and with decreasing gradients along the containment perimeter. This latter factor is
of particular concern with P&T systems that seek to minimize ground-water treatment and/or disposal
costs by decreasing pumping to impose the smallest gradients needed for capture. In some cases, it may
be practical (and necessary) to use a modeling analysis to interpret hydraulic head measurements and
evaluate containment performance (Larson et al., 1987). In other cases, it will be cost-effective to
overpump to achieve more demonstrable containment.
It is often easier to demonstrate that inward hydraulic gradients exist toward such systems as
recovery drains than toward recovery wells (Figure 1-1). In some cases, this is a significant advantage of
P&T systems that incorporate drains and walls.
2.2.1.4 Measurement Frequency
Inward gradients and hydraulic containment may be affected by hydraulic head fluctuations
caused by the startup and cycling of P&T operations, offsite well pumping, tidal and stream stage
variations, and seasonal factors. If the P&T site is located in an active hydrogeologic setting, hydraulic
heads may rise and fall on the order of feet several times a day. To adequately monitor inward gradients
and hydraulic containment, consider the following strategies.
(1) Monitor intensively during system startup and equilibration to help determine an appropriate
measurement frequency. This may involve using pressure transducers and dataloggers to
make near-continuous head measurements for a few days or weeks, then switching
sequentially to daily, weekly, monthly, and possibly quarterly monitoring. Data collected
during each phase is used to examine the significance of hydraulic head fluctuation and
justify any subsequent decrease in monitoring frequency. An example of the use of frequent
measurements to assess transient effects of daily pumping cycles on hydraulic gradients is
shown in Figure 2-5.
(2) Make relatively frequent hydraulic head measurements when the P&T system pumping rates
or locations are modified, or when the system is significantly perturbed in a manner that has
not been evaluated previously. Significant new perturbations may arise from extraordinary
recharge, flooding, drought, new offsite well pumping, improved land drainage, etc.
(3) Acquire temporally consistent hydraulic head data when measuring inward hydraulic
gradients or a potentiometric surface so that differences in ground-water elevations within the
well network represent spatial rather than temporal variations.
17
-------
Partially-Penetrating Pump Well
00
ertical Capture Zone
Figure 2-4. Cross-section showing equipotential contours and the vertical capture zone associated with
ground-water withdrawal from a partially-penetrating well in isotropic media. Ground-water
flow at depth beneath the well is not captured by pumping despite the presence of apparent
upward gradients. In stratified, anisotropic media (e.g., Kx>Kz), the vertical hydraulic control
exerted by a partially-penetrating well will be further diminished.
-------
CO
I
1
•a
>»
369.5
368.5
367.5
3:00 FM
-MW-9-
MW-
MW-2
MW-7
MW-4
MW-1 PW-20 quick cycles
PW-19 pumping-
-PW-21 pumping-
9:00 T
9/23/92
3:00 A!
9:00 AM
9/24/93
3:00 PM
Figure 2-5. Near continuous hydraulic head measurements were made in several observation wells in the
vicinity of a recovery well line to examine the transient water table response to pump cycles and
recharge events (modified from ESE, 1992). The data reveal that ground-water flow directions
are fairly constant during pump cycles. In conjunction with weekly data, it was determined that
the frequency of hydraulic head surveys should be reduced to monthly.
If inward gradients are not maintained during P&T operation, an analysis should be made to
determine if containment is threatened or lost. Rose diagrams can be prepared to display the variation
over time of hydraulic gradient direction and magnitude based on data from at least three wells (Figure
2-6). Transient capture zone analysis, perhaps using a numerical model and particle tracking, may be
required to assess containment effectiveness. Even where the time-averaged flow direction is toward the
P&T system, containment can be compromised if contaminants escape from the larger capture zone
during transient events or if there is a net component of migration away from the pumping wells over
time.
2.2.1.5 Some Additional Considerations
Use of Pump Well Data - Hydraulic heads and extraction rates associated with recovery wells
should be factored into capture zone analysis. It is generally inappropriate, however, to interpret inward
gradients by comparing the hydraulic head measured in a piezometer to that in a pump well (Figure 2-7).
Rather, hydraulic gradients and flow patterns should be interpreted primarily based on head
19
-------
North
West
East
South
Figure 2-6. Example display of ground-water flow directions and hydraulic gradients determined between
three observation wells.
""
...-•' ...-•' Q>' ..-•" .,.-''* ..--Constant-heat
V-" -••"' ..-"" -•'"' "^ .-''' X' -'•"" -•• Pumping
""" \.-'""" ..---"" ...--"' S' ..^' /'"" x--'"" ...---•'' ^ elevation in
.-••"'\..-••'" ..--•"" ^^ ..-••'' ^''" .-•••''' .--"'' _b^' each well is
.- 360.0 ft
::||£«^^
fe
Figure 2-7. Ground water flows between and beyond the recovery wells even though hydraulic heads
throughout the mapped aquifer are higher than the pumping level. Rely primarily on
observation well data to determine flow directions.
20
-------
measurements in observation wells or piezometers. Useful estimates of hydraulic heads in the vicinity of
a pump well with a known pumping rate and level can be derived using well hydraulics equations (i.e.,
the Theis or Theim equations, Bear, 1979) or a ground-water flow model; but uncertainties associated
with formation properties and well loss may confound the analysis.
Horizontal Anisotropy - Where strata are inclined or dipping, particularly foliated media such as
schist with high-angle dip, significant horizontal anisotropy may be present. The directions of maximum
and minimum permeability are usually aligned parallel and perpendicular, respectively, to foliation or
bedding plane fractures. In anisotropic media, the flow of ground water (and contaminants moving with
ground water) is usually not perpendicular to the hydraulic gradient. This is demonstrated at a petroleum
tank farm site in Virginia where the flow of leaked LNAPL and ground water is offset significantly from
the hydraulic gradient toward the direction of maximum permeability (Figure 2-8). Interpretation of
hydraulic head data and capture zone analysis must account for anisotropy to evaluate containment
effectiveness. Various well hydraulics equations incorporate anisotropy (Papadopulos, 1965; Kruseman
and deRidder, 1990) and many numerical models can treat anisotropic conditions.
Transient Loss of Capture during Early Pumping — Given the steep initial hydraulic gradient
induced by pumping, hydraulic containment provided by P&T operation may decrease with time due to
the flattening of the drawdown cone(s) as illustrated by the computer simulations shown in Figure 2-9.
Early demonstration of inward hydraulic gradients, therefore, does not ensure continued containment.
Long-term monitoring must be relied upon to assess long-term P&T system performance.
Drawdown Limitations - Under some conditions, inward hydraulic gradients cannot be
maintained unless barrier walls are installed and/or water is injected (or infiltrated) downgradient of or
within the contaminated zone. Limited aquifer saturated thickness, a relatively high initial hydraulic
gradient, a sloping aquifer base, and low permeability are factors that can prevent hydraulic containment
using wells or drains (Saroff et al., 1992). Where these conditions exist and hydraulic containment is
planned, particular care should be taken during pilot tests to assess this limitation.
Injection/Extraction Cells - Two prime objectives of aquifer restoration are to contain and/or
remove contaminant plumes. Hydraulic controls provide an opportunity to concurrently accomplish both
of these objectives. Recharging upgradient of the contaminant plume and flushing it toward
downgradient collection points creates a ground-water recirculation cell that isolates the plume from the
surrounding ground water. By properly adjusting recharge and extraction rates, these cells can minimize
the volume of water requiring treatment, thereby reducing the flushing time. If permitted, water injection
can greatly enhance hydraulic control of contaminated ground water. Options associated with selecting
injection locations and rates provide great containment flexibility (e.g., Wilson, 1985).
Highly Permeable and Heterogeneous Media — In highly permeable media, high pumping rates
are usually required to attain hydraulic containment and performance monitoring can be complicated by
flat hydraulic gradients. Barrier walls and containment area surface covers installed to reduce the rate of
pumping needed for containment also facilitate demonstration of inward gradients (Figure 1-1). Complex
heterogeneous media are difficult to characterize. Ideally, monitor wells are installed in the more
21
-------
/ HYDRAULIC GRADIENT DIRECTION
Woler table contours h H obow USt with a 2-It
362 contour hilervol bosjd on the J/3'/»2 M»v»
Estimoled extent of separate phase pevoleum
• LNAPL present
T INAPl absent
Figure 2-8. Ground-water and LNAPL flow in anisotropic saprolite soil from a petroleum-product tank
farm in Fairfax, Virginia is offset from the hydraulic gradient toward the strike of saprolite
foliation.
-------
a a"
-S .9
"
o sa .2
ill
8" 3 1
.9
ja &>
~E^
'a 1 S g>
•B 8 a !
— <•> "" S
II gg.
s »r
I'l 11
C*l
S
S,
23
-------
permeable strata to provide optimal chemical detection and gradient control monitoring capability.
Hydraulic containment, inward gradient monitoring, and site characterization are also facilitated in
heterogeneous media by installing barrier drains and walls, particularly if done in a manner that allows
subsurface examination during construction.
NAPL Containment — Inward hydraulic gradients will contain LNAPL migration. DNAPL,
however, may migrate under the influence of gravity in directions that are counter to the hydraulic
gradient. Unless of sufficient magnitude to overcome the gravitational force, therefore, inward hydraulic
gradients cannot be relied upon to contain DNAPL movement. Cohen and Mercer (1993) describe
several approaches for estimating hydraulic gradients required to arrest DNAPL migration.
Ambiguous Gradient Data — At many P&T sites, interpretation of hydraulic gradients will
provide an ambiguous measure of containment effectiveness. To raise confidence in the monitoring
program, consider: (1) increasing the frequency and locations of hydraulic head measurements; (2)
conducting more robust data analysis, perhaps using models; (3) relying more on chemistry monitoring;
or (4) modifying the P&T system (e.g., by increasing the pumping rate) to provide more demonstrable
containment.
2.2.2 Vertical Hydraulic Gradients
Inward gradients may also be specified as upward gradients at the base of the contaminant plume
or containment volume. This is important because a P&T system may fail to prevent downward
contaminant migration (e.g., where remediation wells are too shallow or have insufficient flow rates). For
dissolved contaminants, in many cases, the magnitude of the upward gradient need only be measurable.
For DNAPLs, the inward gradient must be large enough to overcome the potential for DNAPL to move
via gravity and capillary pressure, forces (Cohen and Mercer, 1993). At sites where upward hydraulic
gradients sufficient to arrest DNAPL migration cannot be developed, consideration must be given to other
containment strategies. For example, if DNAPL can be reduced to residual saturation by pumping,
capillary forces may be sufficient to overcome gravitational forces and prevent downward migration.
Upward gradients across the bottom of the containment volume can be monitored by comparing
(1) hydraulic head differences measured in adjacent nested wells that are screened at different depths and/
or (2) potentiometric surfaces developed for different elevations, stratigraphic units, or flow zones.
Generally, a nested cluster of wells consists of three monitoring wells/piezometers completed at different
depths. However, the required number of wells depends on site-specific monitoring objectives,
contaminant distribution, P&T system design, and the degree of site complexity.
In a layered multiaquifer system, where the entire thickness of a contaminated upper aquifer is
within the containment volume, upward gradient control wells can be completed above and below the
underlying aquitard to determine the direction of flow across the aquitard (Figure 2-10). If, however, the
containment volume bottom is within a flow zone of significant thickness, nested wells will generally
be required at different elevations (above and below the containment volume bottom) within the flow
zone. For this case, upward gradients may not ensure containment (Figure 2-4), and it may be necessary
24
-------
Lower Aquifer jjjj
Potentiometric |
Surface II
Figure 2-10. Vertical hydraulic gradients across an aquitard between aquifers are typically measured using
observation well nests.
to rely on a careful three-dimensional analysis of flow and chemical monitoring to evaluate containment
-effectiveness.
A more thorough analysis of upward hydraulic gradients can be made by comparing
potentiometric surface maps for different elevations (or stratigraphic units) to develop a contour map of
vertical hydraulic gradients. A vertical gradient contour map can be used to delineate areas of upward
and downward flow components.
Special precautions should be taken when drilling monitor wells into and/or below a
contaminated zone to minimize the potential for cross-contamination. Where DNAPL is present, it may
be advisable to monitor potentially uncontaminated, deep units by installing wells beyond the DNAPL
zone limit even though this will diminish the upward gradient monitoring capability.
2.2.3 Hydraulic Head Differences
True hydraulic gradients may be difficult to determine; therefore, the objective may revert to
determining a measurable quantity, such as hydraulic head. Hydraulic head differences may be specified
as performance criteria at pumping or observation wells as either differences in head between different
locations at the same time or as time-dependent drawdown in particular wells. In any event, hydraulic
head performance criteria must be developed within the context of capture zone analysis based on an
understanding of the relationship between hydraulic heads at specific locations and local hydraulic
gradients. Otherwise, they may be poor indicators of system performance.
2.2.4 Flow Meters
A few techniques and tools have been developed recently to measure horizontal ground-water
flow directions directly in a single well. Such techniques include using a special flowmeter in a well to
25
-------
measure horizontal flow direction (Kerfoot, 1984; Melville et al., 1985; Guthrie, 1986) and a colloidal
borescope that measures the movement of naturally occurring colloids in ground water (Kearl and Case,
1992). If these tools are found to be reliable at a site, then flow directions (and hence inward hydraulic
gradients) can be determined directly in wells placed along the containment boundary and elsewhere.
Technologies for measuring vertical flows within wells under ambient and pumping conditions
hence also have been developed (Molz and Young, 1993). These tools allow better characterization of the
relative permeability distributions and hence preferential flow paths.
2.2.5 Pumping Rates
For hydraulic containment, the placement and extraction or injection rates of wells are determined
so that ground water in the containment area/volume follow pathlines to the P&T system. The initial
design may be based on the results of ground-water modeling (Section 2.6) and may designate pumping
rates, pump well drawdowns, or high-low pumping level ranges for the P&T system. However, it is not
appropriate to specify model-determined pumping rates or levels as long-term performance criteria,
because these may be too high or too low if the model is inaccurate. The feasibility of pumping rates and
levels determined using a model must be verified during onsite aquifer testing, upon initiation of the P&T
system, and by long-term monitoring.
Pumping rates and levels are monitored to: (1) demonstrate that the system is operational (or
alert managers to make necessary repairs if pumps are found to be inoperable); (2) determine if pumping
rates and levels are within specified tolerances; and, (3) provide data necessary for system optimization.
Pumping rates must be maintained to control hydraulic gradients. As discussed in Section 2.2, if the rates
are "optimized" to reduce P&T costs, it may become very difficult to demonstrate containment by
measuring hydraulic gradients. When analyzing P&T system behavior, particularly where there are
multiple pumping wells, it is important to monitor (and document) pumping rates, times, and levels on a
well-specific basis (rather than simply monitoring totalized flows from multiple wells).
Well discharge rates can be determined by several methods, including the use of a pipe orifice
weir, weirs and flumes, and flowmeters (Driscoll, 1986). During P&T system operation, however,
pumping rates are usually monitored in a closed system using flowmeters which provide pumping rate
and totalized discharge data. Several different types of flowmeters (e.g., rotameters, ultrasonic Doppler
flowmeters, turbine/paddlewheel flowmeters, magnetic flowmeters, etc.) and automated data logging and
alarm systems are available.
2.2.6 Ground-Water Chemistry
2.2.6.1 Performance Concept
Ground-water quality monitoring is performed at nearly all P&T operations to determine if
temporal or spatial variations in contaminant distribution are consistent with effective hydraulic
containment. If not, the monitoring identifies areas and temporal conditions of inadequate containment
which should then be improved by a P&T system upgrade.
26
-------
At sites where contamination is enclosed by the containment volume perimeter, a detection
monitoring program can be implemented at or beyond this perimeter to evaluate P&T performance.
Chemical analysis should target the most mobile site contaminants. Detection of contaminants above
background concentrations (if any) indicates a lack of containment, unless the contaminant presence can
be attributed to an alternate source.
Ground-water quality monitoring to assess containment may provide ambiguous results if some
site contaminants are located beyond the containment volume perimeter prior to P&T system startup.
Given this scenario, containment failure is suggested if: (1) the estimated total contaminant mass in
ground water beyond the containment perimeter increases with time (see Section 3 and Appendix A); (2)
contaminant concentrations change with time (e.g., increase) in perimeter or downgradient monitor wells
in a manner that is inconsistent with effective containment; and/or (3) relatively retarded contaminants,
that were previously restricted to the containment area, are detected in perimeter monitor wells. If the
spatial distribution of contaminants or the ground-water flow field is ill-conceived, then each of these
criteria is subject to misinterpretation. Where ground-water chemistry data limitations are significant,
greater reliance is placed on hydraulic gradient monitoring.
Tracers can be injected within the plume and monitored outside the containment volume to
discriminate between lack of containment, pre-existing contamination beyond the containment limit, and
potential offsite contaminant sources. Detecting a unique tracer beyond the containment area indicates a
lack of containment. The use of tracers is discussed in Section 2.2.8.
2.2.6.2 Ground-Water Quality Monitoring Locations
Monitor well locations and completion depths are selected to provide a high probability of
detecting containment system leaks in a timely manner. Site characterization data and capture zone
analysis are used to identify potential areas and pathways of contaminant migration across the
containment volume perimeter during P&T operation and inoperation (due to mechanical failure or
routine system maintenance). These potential migration routes may include the more permeable media,
areas and depths subject to relatively weak ground-water flow control, and manmade or natural drainage
features (e.g., sewers, streams, etc.). Using this hydrogeologic approach, site-specific conditions are
evaluated to choose optimum ground-water sampling locations. Various geostatistical methods (e.g.,
Haug et al., 1990) and plume generation models (e.g., Wilson et al., 1992; Meyer and Brill, 1988) can
also be used to help assess well spacing and depths. Loaiciga et al. (1992) present a review of the
application of hydrogeologic and geostatistical approaches to ground-water quality network design. In
general, as with mapping hydraulic gradients, the number of ground-water quality monitor wells needed
to assess containment effectiveness increases with plume size and site complexity.
Ideally, P&T system failure will be detected before contaminants migrate far beyond the
containment perimeter toward potential receptors. Consequently, monitor wells with a relatively close
spacing are usually located along or near the potential downgradient containment boundary. Inward
gradient control wells (discussed in Section 2.2.1) are frequently used for ground-water sampling. Public
or private water supply wells located downgradient of the contamination may also be used to monitor
27
-------
containment effectiveness and to determine the quality of ground water being consumed by local
residents.
Modifications to monitoring locations and criteria may be needed to complement changes in P&T
operation, ground-water flow directions, contaminant distributions, and/or the specified containment
volume.
2.2.6.3 Ground-Water Quality Monitoring Frequency
Ground-water quality surveys are usually conducted less frequently than hydraulic head surveys
because: (1) contaminant movement is a slower process than that controlling transient hydraulic head
propagation; and (2) ground-water quality surveys are much more expensive to conduct than ground-
water elevation surveys. Determining ground-water sampling frequency requires consideration of site-
specific conditions. It should not be assumed that all wells must be sampled at the same time, for the
same parameters, or during every sampling episode.
In general, it is good practice to sample at a higher frequency and perform more detailed chemical
analyses in the early phase of the monitoring program, and then to use the information gained to optimize
sampling efficiency and reduce the spatial density and temporal frequency of sampling in the later phases.
For example, consider the following strategies.
(1) Monitor ground-water quality in perimeter and near-perimeter leak detection wells more
frequently than in wells that are more distant from the contaminant plume limit.
(2) Specify sampling frequency based on potential containment failure migration rates that
consider the hydraulic conductivity (k) and effective porosity (n) of the different media, and
maximum plausible outward hydraulic gradients (i). If appropriate, account for the
retardation factor, Rf (Section 1.2, Equation 1-4). Use modeling results or simple
calculations of contaminant average linear velocity (vc, where vc = Rjki/n) to estimate
potential contaminant transport velocities. Consider sampling more permeable strata in
which migration may occur relatively quickly more frequently than less permeable media.
(3) After performing detailed chemical analyses during the remedial investigation or the early
phase of a monitoring program, increase monitoring cost-effectiveness by focusing chemical
analyses on site contaminants of concern and indicator constituents. Conduct more detailed
chemical analyses on a less frequent basis or when justified based on the results of the more
limited analyses.
At sites with inorganic contamination or where organic site contaminants are present initially
beyond the containment perimeter, it may be necessary to use statistical methods to: (1) distinguish
contaminant detections from background concentrations; and (2) assess the influence of various temporal
and spatial factors (e.g., recharge rate and heterogeneity, respectively) on contaminant concentration
variability. Sampling locations and frequency, therefore, may be dictated by the requirements of
28
-------
statistical analyses. Guidance on applying statistics to differentiate contamination from background and
to identify concentration trends with statistical significance is provided by USEPA (1986, 1988a, 1989,
1992b, 1992c) and Gilbert (1987). At some sites, identifying background contaminant concentrations and
trends may not be cost-effective given monitoring program objectives.
2.2.7 Perimeter Monitoring Using Noninvasive Methods
At sites where contaminants have not migrated beyond the containment perimeter, it may be cost-
effective to enhance P&T monitoring by conducting surface geophysical or soil gas surveys along
transects between monitor wells (Figure 2-11). Using this approach, an initial baseline survey is made
along well-defined transects. Repeat surveys are then conducted periodically to detect changes from the
baseline condition that evidence contaminant migration.
Electrical geophysical methods (EM-conductivity and resistivity) can be used to detect the
migration of conductive contaminants in ground water. An application of this strategy using quarterly
EM-conductivity surveys along transects between wells to augment a landfill leachate detection
monitoring network is described by Rumbaugh et al. (1987). Similarly, under appropriate conditions,
volatile organic contaminant movement in the upper saturated zone can be inferred by analysis of soil gas
samples (Devitt et al., 1987; Cohen and Mercer, 1993).
LEGEND
EM PERIMETER TRAVERSE
I FENCE EXPERIMENTS
; POWER LINE EXPERIMENT
' MICROWAVE EXPERIMENT
seal*
f««t
Figure 2-11. Surface geophysical (EM-Conductivity) surveys were conducted periodically along transects
between monitor wells encircling a sanitary landfill in Maryland to augment the leak detection
monitoring network (from Rumbaugh et al., 1987).
29
-------
Although often less costly, data acquired using noninvasive methods is also less definitive than
direct ground-water sampling data. As a result, inferences derived from these techniques must be
confirmed by ground-water sampling and analysis.
2.2.8 Tracers
Tracers are used in ground-water studies to determine flow path, velocity, solute residence time,
and formation properties such as hydraulic conductivity, dispersivity, and effective porosity (Davis et al.,
1985). At sites where contaminants are present beyond the containment zone, ground-water tracers can
be used to enhance performance monitoring. A tracer can be released periodically into ground water
inside the containment zone where hydraulic control is considered least effective. Subsequent tracer
detection in ground water beyond the containment perimeter (e.g., during regular monitoring surveys)
would indicate containment failure and possibly the general location of the failure. Tracers can also be.
used to help delineate the P&T capture zone by releasing tracer in areas of uncertain capture and
monitoring for tracer presence in pumped ground water.
A detailed discussion of tracer selection and use for ground-water investigations is provided by
Davis et al. (1985). Important ground-water tracers include particulates (spores, bacteria, and viruses),
ions (chloride and bromide), dyes (Rhodamine WT and Fluorescein), radioactive tracers, fluorocarbons,
and organic anions. Tracers are selected based on their properties (e.g., toxicity and mobility) and the
availability of reliable analytical techniques. Determination of the amount of tracer to inject is based on
its background concentration, the analytical detection limit, and the expected degree of tracer dilution at
sampling locations. Tracer concentration should not be increased so much that density effects become a
problem for the particular application.
2.3 MONITORING LOCATION SUMMARY
Hydraulic head and ground-water chemistry monitoring locations are discussed in Section 2.2 for
each performance measure. In summary, monitoring is conducted within, at the perimeter, and
downgradient of the containment zone to interpret ground-water flow, contaminant transport, and P&T
system performance. Containment area monitoring is used particularly to assess extraction/injection
impacts and hydraulic control at the containment volume bottom. Perimeter monitoring facilitates
contaminant leak detection and evaluation of inward gradients. Downgradient monitoring provides
additional containment failure detection capability and helps assess potential contaminant migration to
water-supply wells and/or surface water.
2.4 OPERATIONS AND MAINTENANCE (O&M) MANUAL
Many P&T systems may be dysfunctional due to a lack of adequate monitoring and maintenance.
O&M manuals should be prepared for each P&T system. Elements of an O&M plan should: (1) provide
an introductory description of the P&T system; (2) identify and describe system components (e.g., pumps,
controllers, piping, wiring, treatment system parts, alarms, etc.); (3) include detailed drawings of system
layout, equipment schematic diagrams, and parts listings; (4) enumerate system installation, startup, and
30
-------
operation procedures; (5) provide a troubleshooting guide and problem call-down or contact list; and (6)
detail system monitoring, maintenance, and record-keeping requirements and schedules. Much of this
information is available from equipment vendors.
2.5 P&T MONITORING PLAN
As noted in Section 2.1, a written monitoring plan should also be developed for P&T system
operation. The plan should describe: (1) monitoring objectives; (2) the types of measurements to be
made (e.g., pumping rates, hydraulic heads, ground-water chemistry, precipitation); (3) measurement
locations; (4) measurement methods, equipment, and procedures; (5) measurement schedules; and (6)
record-keeping and reporting requirements. It is important that the monitoring plan be revised as data is
collected and improvements are realizedwith respect to the site conceptual model and knowledge of the
distribution of contaminants is enhanced.
2.6 CAPTURE ZONE ANALYSIS AND OPTIMIZATION MODELING
In recent years, many mathematical models have been developed or applied to compute capture
zones, ground-water pathlines, and associated travel times to extraction wells or drains (Javandel et al.,
1984; Javandel and Tsang, 1986; Shafer, 1987a,b; Newsom and Wilson, 1988; Blandford and Huyakom,
1989; Pollock, 1989; Strack, 1989; Bonn and Rounds, 1990; Bair et al., 1991; Rumbaugh, 1991; Bair and
Roadcap, 1992; Fitts, 1993; Gorelick et al., 1993). These models provide insight to flow patterns
generated by alternative P&T schemes and the selection of monitoring locations and frequency.
Additionally, linear programming methods are being used to optimize P&T design (Ahlfeld and Sawyer,
1990; Hagemeyer et al., 1993; Gorelick et al., 1993) by specifying an objective function subject to various
constraints (e.g., minimize pumping rates but maintain inward hydraulic gradients). Given their
application to the design, evaluation, and monitoring of P&T systems, a brief overview of a few capture
zone analysis and optimization techniques follows. It must be kept in mind, however, that the accuracy of
modeling predictions is dependent on the availability and validity of the required input data.
Several semianalytical models employ complex potential theory to calculate stream functions,
potential functions, specific discharge distribution, and/or velocity distribution by superposing the effects
of multiple extraction/injection wells using the Thiem equation on an ambient uniform ground-water flow
field in a two-dimensional, homogeneous, isotropic, confined, steady-state system (e.g., RESSQ, Javandel
et al., 1984; DREAM, Bonn and Rounds, 1990, and, RESSQC, Blandford and Huyakom, 1989). Based
on this approach, the simple graphical method shown in Figure 2-12 can be used to locate the stagnation
point and dividing streamlines, and then sketch the capture zone of a single well in a uniform flow field.
The extent to which these results represent actual conditions depends on the extent to which the
assumptions vary from actual site conditions.
This analysis is extended by Javandel and Tsang (1986) to determine the minimum uniform
pumping rates and well spacings needed to maintain capture between two or three pumping wells along a
line perpendicular to the regional direction of ground-water flow. Their capture zone design criteria and
type curves given in Figure 2-13 can be used for capture zone analysis, but more efficient P&T systems
31
-------
can be designed with nonuniform pump well orientations, spacings, and extraction rates. Streamlines and
capture zones associated with irregular well spacings and variable pumping rates can be simulated by the
complex potential flow models, RESSQ, RESSQC, and DREAM. Reverse particle tracking is
implemented in RESSQC to derive steady-state capture zones by releasing particles from the stagnation
point(s) of the system and tracking their advective pathlines in the reversed velocity field. Similarly,
time-related captures zones (Figure 2-14) are obtained by tracing the reverse pathlines formed by particles
released all around each pumping well (Blandford and Huyakom, 1989; Shafer, 1987a).
Application of semianalytical models to field problems requires careful evaluation of their
limiting assumptions (e.g., isotropic and homogeneous hydraulic conductivity, fully-penetrating wells, no
recharge, no vertical flow component, and constant transmissivity). Several analytic models relax these
restrictive assumptions by superposition of various functions to treat recharge, layering, inhomogeneity,
three-dimensional flow, etc. (Fitts, 1989; Strack, 1989; Rumbaugh, 1991). Where field conditions do not
conform sufficiently to model assumptions, the simulation results will be invalid (e.g., Springer and Bair,
1992).
Numerical models are generally used to simulate ground-water flow in complex hydrogeologic
systems (e.g., MODFLOW, McDonald and Harbaugh, 1988; and SWIFT/486, Ward et al., 1993). For
example, the benefits of using partially-penetrating recovery wells to minimize pumping rates and
unnecessary vertical spreading of contaminants can be examined using a three-dimensional flow model.
w = Q/2T
y = 0
w = -Q/2T
x = 0
Figure 2-12. Equations for the dividing streamlines (w=Q/2T) that separate the capture zone of a single well
from the rest of an isotropic, confined aquifer with a uniform regional hydraulic gradient
(modified from Gorelick et al., 1993). Note that T=transmissivity (L2/T), Q=pump rate (L3/T),
and i=initial uniform hydraulic gradient).
32
-------
DOUBLE-WELL CAPTURE-ZONE TYPE CURVES
SINGLE-WELL CAPT.URE-ZONE TYPE CURVES
1000.
-1000. r,.i.Iiii.I,,iiIi..iI
-500. 0. 500.1000.1500.2000.2500.
1500.
1000.
500.
0.
-500.
-1000.
-1500. ". .
THREE-WELL CAPTURE-ZONE TYPE CURVES
-500. 0. 500. 1000. 1500. 2000. 2500. 3000.
FOUR-WELL CAPTURE-ZONE TYPE CURVES
1500.
1000. -
500. -
in
c.
0)
0. -
-500. -
-1000. -
-1500. r, , , , .1 . . . , i . . i , i i . i . i . , i . i . i i . i . . . , i
-500. 0. 500. 1000. 1500. 2000. 2500. 3000.
-500. -
-1000. -
-1500.
-500.
0. 500. 1000. 1500. 2000. 2500. 3000.
Number
of Wells
1
2
3
Recommended
Distance between
Each Pumping Well
Q/CrfTi)
1.26Q/(nTi)
Distance between
Dividing Streamlines
at Well Line
Q/(2Ti)
Q/Ti
3Q/(2Ti)
Distance between
Dividing Streamlines
Far Upstream from
Well Line
Q/Ti
2Q/Ti
3QTi
Distance to
Stagnation Point at the
Center of the
Capture Zone
Q/(2JtTi)
Q/(2jrt1)
3Q/(47tTi)
Figure 2-13. Type curves showing the capture zones of 1 (a), 2 (b), 3 (c), and 4 (d) pump wells spaced evenly
along the y-axis for several values of Q/BU (where Q = pumping rate L3/T), B = aquifer
thickness (L), and U = Darcy velocity for regional flow (LT) (from Javandel and Tsang, 1986).
To assess the number of wells, pumping rates, and well spacings needed to capture a plume using
evenly spaced recovery wells along a line: (1) Construct a plume map at the same scale as the
type curves; (2) Superimpose the 1-well type curve over the plume with the x-axis parallel to the
regional flow din and overlying the center of the plume such that the plume is enclosed by
one Q/BU curve; (3) Calculate the required single well pumping rate as Q=B*U*TCV where
TCV is the bounding Type Curve Value of Q/BU; and, (4) If a single well cannot produce the
calculated pump rate, repeat the steps using the 2,3, and 4 well type curves until a feasible single
well pump rate is calculated. Use the above equations to determine optimum well spacings. gee
Javendel and Tsang (1986) for details.
33
-------
3000
2400
1800
1200
600
600
1200
1SOO
2400
300C
Figure 2-14. Example of steady-state, and 10-year and 25-year time-related capture zones delineated using
reverse particle tracking (from Blandford and Huyakom, 1989).
Numerical flow model output is processed using reverse or forward particle-tracking software such as
MODPATH (Pollock, 1989), GWPATH (Shafer, 1987b), STLINE (Ward et al., 1993), FLOWPATH
(Franz and Guiguer, 1990), PATH3D (Zheng, 1989), and the GPTRAC module of WHPA (Blandford and
Huyakom, 1989) to assess pathlines and capture zones associated with P&T systems at sites that cannot
be accurately modeled using simpler techniques. Solute transport models are primarily run to address
aquifer restoration issues such as changes in contaminant mass distribution with time due to P&T
operation (e.g., Ward et al., 1987).
Ground-water flow models can be coupled with linear programming optimization schemes to
determine the most effective well placements and pumping rates for hydraulic containment much more
quickly than a trial-and-error approach. The optimal solution maximizes or minimizes a user-defined
objective function subject to all user-defined constraints. In a P&T system, a typical objective function
may be to minimize the pumping rate to reduce cost, while constraints may include specified inward
gradients at key locations, and limits on drawdowns, pumping rates, and the number of pump wells.
Gorelick et al. (1993) present a review of the use of optimization techniques in combination with ground-
water models for P&T system design. Available codes include AQMAN (Lefkoff and Gorelick, 1987) an
optimization code that employs the Trescott et al. (1976) two-dimensional ground-water flow model, and
MODMAN (GeoTrans, 1992), which adds optimization capability to the threedimensional USGS
MODFLOW model (McDonald and Harbaugh, 1988) and others (USEPA, 1993 a). A case study of the
application of an optimization code to assist P&T design is given by Hagemeyer et al. (1993).
34
-------
Coupled ground-water flow-optimization models can also be used to evaluate monitoring well
network design (Meyer and Brill, 1988; Meyer, 1992). Objectives might be to (1) minimize the total
number of monitor wells, (2) maximize the probability of detecting contaminant migration, and (3)
minimize the area of expected contamination at the time of leak detection. The first two objectives are
addressed using the Maximal Covering Location Problem method illustrated in Figure 2-15 to find well
locations and depths that maximize the probability of future plume detection (Meyer, 1992). Another
approach, the Extended P-Median Problem, addresses all three objectives by tracking plume size as it
grows with time (Meyer, 1992).
Although P&T and monitoring design can be aided by the use of ground-water models, actual
field monitoring must be carried out in order to provide information necessary to evaluate model
predictions. As described in this Chapter, hydraulic containment effectiveness is determined by
monitoring hydraulic heads and ground-water chemistry.
2.7 OPERATIONAL EFFICIENCY
Operational efficiency refers to the cost-effectiveness of actions taken to attain remedial
objectives. These actions include P&T system design, operation, monitoring, and modification. Efficient
P&T performance requires that there be a clear statement of remedial objectives.
For, perpetual hydraulic containment, an appropriate objective might be to minimize the total cost
required to maintain hydraulic containment and satisfy associated regulatory requirements. Given this
objective, installing low permeability barriers to reduce pumping rates might be cost-effective. At sites
with an economic incentive to remove contaminant mass (i.e., where the containment area size may be
diminished or P&T discontinued if clean-up goals are met), a more complex cost-effectiveness trade-off
exists between minimizing hydraulic containment costs and maximizing contaminant mass removal rates.
Comparative cost-benefit analysis requires evaluation of the benefits, costs, and risks of each
design alternative based on P&T component and site specific factors. A framework for risk-based
decision analysis applicable to P&T system design (Figure 2-16) is provided by Massmann and Freeze
(1987), Freeze et al. (1990), and Massmann et al. (1991). Using this method, an objective function, j, is
defined for each remedial alternative, j = 1 . ..N. as the net present value of the anticipated stream of
benefits, costs, and risks taken over a remedial time period and discounted at the market interest rate. The
goal is to maximize the objective function (Freeze et al., 1990):
a
t-0
where 0= - the objective function for alternative j [$]; Bj(t) - benefits of alternative j in year t [$]; Cj(t) =
costs of alternative j in year t [$]; Rj(t) - risks of alternative j in year t [$]; T - time horizon [years]; and i
- discount rate [decimal fraction]. The probabilistic risk cost, R(t), is defined as (Freeze et al., 1990):
35
-------
Compliance
Boundary
O O
O
(a)
(b)
(c)
O Potential Well Location
• Well Location Detects 1 Plume
• Well Location Detects 2 Plumes
(d)
Figure 2-15. An example of the Maximal Covering Location Problem applied to monitor well network design
(from Meyer, 1992). The capability of different monitor well locations to detect random plumes
generated using a Monte Carlo simulator in (a), (b), and (c) are combined to indicate optimum
well locations in (d).
36
-------
Field Investigation Program:
Data Acquisition System
Geological
Uncertainty Model
Parameter
Uncertainty Model
Hydrogeological
Simulation Model
Engineering
Reliability Model
I Decision Model I
il======_!^==.===Bfl
Figure 2-16. A framework for risk-based decision making regarding P&T system design and monitoring
(modified from Freeze et al., 1990.)
R(t)-Pr(t) Cr(t)tfCr)
(2-3)
where Pr(t) = the probability of failure in year t [decimal fraction] Q(t) = costs associated with failure in
year t [$]; and 7 (Cr) = the normalized utility function [decimal fraction, Y> 1] which can be used to
account for possible risk-averse tendencies of decision makers. The benefits of an alternative, B(t), can
similarly be formulated as probabilistic benefits. Trade-offs between cost and risk and the concept of
optimal risk are illustrated in Figure 2-17. Note that acceptable risk, from a societal or regulatory
perspective, may be less than an owner-operator's optimal risk.
Example applications of this risk-based decision analysis approach to P&T system design are
given by Massmann et al. (1991) and Evans et al. (1993). Variables pertaining to P&T monitoring
design, such as well spacing and sampling frequency, can also be evaluated using this methodology, as
can proposed modifications to system design that might be derived from monitoring data. Monitoring
contributes to the objective function by reducing the probability of failure, or equivalently, increasing the
probability of detection (Meyer and Brill, 1988).
Remedial efficiency can be also be enhanced by applying total quality management practices to
P&T operation. Hoffman ( 1993) recommends nine steps to increase the efficiency of a P&T system
designed for hydraulic containment and contaminant mass removal: (1) perform a thorough site
characterization; (2) establish a decision support system that allows rapid interpretation and integration of
new data; (3) locate and remove or contain shallow sources of ground-water contamination; (4) design the
37
-------
$
Objective Function,
Probabilistic Risk
Cost,
Cost of P&T system
and monitoring,
Optimal Risk
-L&creasing, Risk and Increasing Reliability-+
Figure 2-17. The concept of optimal risk (from Freeze et al., 1990).
P&T system to contain and remove contaminant mass; (5) phase in the remedial program to take
advantage of ongoing conceptual model improvements; (6) maintain extensive monitoring of the P&T
system; (7) design the well field such that extraction and injection rates and locations can be varied to
minimize ground-water stagnation; (8) use reinjection of treated ground water and other techniques to
enhance contaminant mass removal; and (9) set contaminant concentration goals (e.g., at the containment
area perimeter) that will allow appropriate water standards to be met at the downgradient point of use.
Although the applicability of various monitoring and remedial measures depends on site-specific
conditions, active P&T system management will usually be cost-effective and lead to enhanced
operational efficiency.
38
-------
3. MONITORING AQUIFER RESTORATION
3.1 INTRODUCTION
The challenge of aquifer restoration is presented in Chapter 1. Restoration P&T design will
typically reflect a compromise between objectives that seek to: (1) reduce contaminant concentrations to
clean-up standards, (2) maximize contaminant mass removal, (3) minimize clean-up time, and (4)
minimize cost. At many sites, P&T systems cannot be relied upon to reduce ground-water contaminant
concentrations to comply with clean-up standards within a short time frame. Aquifer restoration efforts
are made more difficult by concentration tailing and rebound caused by NAPL dissolution, contaminant
desorption, precipitate dissolution, ground-water velocity variations, and/or matrix diffusion (Section
1.2). Consequently, P&T for aquifer restoration requires a high degree of performance monitoring and
management to identify problem areas and improve system operation.
Hydraulic containment generally is a prerequisite for aquifer restoration. Reference, therefore,
should be made to discussions of hydraulic containment design, monitoring, and management in Chapter
2. This chapter focuses on managing and monitoring P&T technology to clean up ground water in the
containment area/volume. Statistical analysis of monitoring data is discussed in Chapter 4.
3.2 PERFORMANCE MEASUREMENTS AND INTERPRETATION
Various restoration performance criteria are described in this section. These criteria are
monitored to determine if the P&T system is functioning as designed and to provide guidance for system
optimization. Performance is monitored by measuring hydraulic heads and gradients, ground-water flow
directions and rates, pumping rates, pumped water quality, contaminant distributions in ground water and
porous media, and, possibly, tracer movement.
33.1 Hydraulic Containment
Hydraulic containment is a design objective of nearly all restoration P&T systems. That is, the
plume is contained to prevent further spread during restoration efforts. In addition, as shown in
Figure 1.2, for some ground-water contamination problems, restoration and containment are used for
different sections of the aquifer. Refer to Chapter 2 for guidance on hydraulic containment performance
monitoring.
3.2.2 Managing Ground-Water Flow
Restoration P&T ground-water flow management typically involves optimizing well locations,
depths, and injection/extraction rates to maintain an effective hydraulic sweep through the contamination
zone, minimize stagnation zones, and flush pore volumes through the system. Wells are installed in lines
or five-spot patterns to enhance hydraulic flushing efficiency; drains are installed to effect line sweeps. In
the following sections, various aspects of ground-water flow management are discussed including
39
-------
(1) pore volume flushing, (2) stagnation zone control, (3) pulsed pumping, and (4) pumping in the
presence of NAPL.
3.2.2.1 Pore Volume Hushing
Restoration requires that sufficient ground water be flushed through the contaminated zone to
remove both existing dissolved contaminants and those that will continue to desorb from porous media,
dissolve from precipitates or NAPL, and/or diffuse from low permeability zones until the sum of these
processes and dilution in the flow field yields persistent acceptable ground-water quality at compliance
point locations.
The volume of ground water within a contamination plume is known as the pore volume (PV),
which is defined as
I
PV - bn dA (3-1)
where b is the plume thickness, n is the formation porosity, and A is the area of the plume. If the
thickness is relatively uniform, then
PV = BnA (3-2)
where B is the average thickness of the plume.
The number of pore volumes (NPV) which must be extracted for restoration is a function of the
clean-up standard, the initial contaminant distribution, and the chemical/media complexities discussed in
Section 1.2. Estimates of the NPV required for clean up can be made by modeling analysis and by
assessing the trend of contaminant concentration versus the NPV removed. At many sites, many PVs
(e.g., 10 to 100) will have to be flushed through the contamination zone to attain clean-up standards.
The NPV withdrawn per year is a useful measure of the aggressiveness of a P&T operation. It is
calculated as
-Qy/PV (3-3)
where Qyr is the total annual pumping rate. Systems are typically designed to remove between 0.3 and
2.0 PVs annually. Low permeability conditions or competing uses for ground water may restrict the
ability to pump at higher rates. Additionally, kinetic limitations to mass transfer (Figure 1-7) may
diminish the benefit of higher pumping rates. If limiting factors are not present, pumping rates may be
increased to improve P&T performance.
40
-------
Water flushing will be limited to infiltration rates where P&T operation has dewatered
contaminated media. As a result, dissolved contaminant concentrations may rebound as the water table
rises when pumping is reduced or terminated. Water can be injected or infiltrated to both minimize this
potential problem and increase the rate of flushing. Where injection is not feasible, soil vapor extraction
or other vadose zone remedial measures might be needed to remove contaminant mass above the water
table.
Where the P&T design is appropriate, but concentration reduction is very slow, monitoring data
should be evaluated to determine if it is technically impracticable to meet remedial action objectives
(Section 3.4). In order to demonstrate technical impracticability, it must be shown that poor or
inappropriate remedial design is not responsible for tailing. Additional information on technical
impracticability is provided in U.S. EPA, 1993.
Poor design factors include low pumping rates and improper location of pump wells and
completion depths. A simple check on the total pumping rate is to calculate the NPV^. Inadequate
location or completion of pump wells (or drains) may lead to poor P&T performance even if the total
pumping rate is -appropriate. For example, wells placed at the containment area perimeter may withdraw
a large volume of clean ground water from beyond the plume via flowlines that do not flush the
contaminated zone. Similarly, pumping from the entire thickness of a formation in which the
contamination is limited vertically will reduce the fraction of Q^. that flushes the contaminated zone. In
general, restoration pump wells or drains should be placed in areas of relatively high contaminant
concentration.
Well placement can be evaluated by: (1) applying expert knowledge linked to a proper
conceptual model of the hydrogeologic system and contaminant distribution; (2) comparing contaminant
mass removed to contaminant mass dissolved in ground water; and (3) using ground-water flow and
transport models. P&T system modifications should be considered if any of these methods indicate that
different pumping locations or rates will improve system effectiveness.
3.2.2.2 Minimize Ground-Water Stagnation
Ground-water flow patterns need to be managed to minimize stagnation zones during P&T
operation. Stagnation zones develop in areas where the P&T operation affects low hydraulic gradients
(e.g., downgradient of a pump well and upgradient of an injection well) and in low permeability zones
regardless of hydraulic gradient. Stagnation zones caused by low hydraulic gradients can be identified by
measuring hydraulic gradients, tracer movement, ground-water flow rates (e.g., with a downhole
flowmeter), and by modeling analysis. Low permeability heterogeneities should be delineated as
practicable during the site characterization study and during ongoing P&T operation. Flow modeling
results can be used to generate either Darcy or interstitial velocities. These can then be contoured or used
with particle tracking to help identify and locate potential stagnation zones. Examples of stagnation zones
associated with different pumping schemes simulated by a ground-water model are given in Figure 3-1.
and the distribution of potential stagnation zones at a complex field site is shown in Figure 3-2.
41
-------
— ©
M M n
Lf> ni U> t/> tO tO to LO i) I/) l/> 10 tO tf) LT>
HYDRAULIC HEAD
CONTOUR MAP
GROUNDWATER VELOCITY
CONTOUR MAP
^Extraction
-/ Well
535 , /
—536 537
I HYDRAULIC HEAD Extraction
CONTOUR MAP \ We||
\
.
-
GROUNDWATER VELOCITY
CONTOUR MAP
Figure 3-1. Examples of stagnation zones (shaded where the ground-water velocity is less than 4 L/T
associated with single-well and five-spot pumping schemes.
42
-------
Lagand
->• Ground watar flow llna
© Extraction location
Araaa of potential ground
watar atagnatlon
^_ Total VOC (ppb)laocon-
,A cantratlon contour, dished
whara kilarrad
Scala : Faat
All walls within VOC plum** ara
contoured without regard to depth
Rhonewood Subdivision
X>
Figure 3-2. Conceptualized ground-water flow patterns and stagnation zones superimposed on a total VOC
isoconcentration contour map at the Lawrence Liver-more National Laboratory site in California
(from Hoffman, 1993).
Once identified, the size, magnitude, and duration of stagnation zones can be diminished by
changing pumping (extraction and/or injection) schedules, locations, and rates. Again, flow modeling
based on field data may be used to estimate optimum pumping locations and rates to limit ground-water
stagnation. An adaptive pumping scheme, whereby extraction/injection pumping is modified based on
analysis of field data, should result in more expedient cleanup (Figure 3-3).
3.2.2.3 Pulsed Pumoing
Pulsed pumping can be used to increase the ratio of contaminant mass removed to pumped
ground-water volume where mass transfer limitations restrict dissolved contaminant concentrations
(Figure 1-7). The concept of pulsed pumping is illustrated in Figure 3-4. Dissolved contaminant
concentrations increase due to diffusion, desorption, and dissolution in slower-moving ground water
during the resting phase of pulsed pumping. Once pumping is resumed, ground water with higher
concentrations of contaminants is removed, thus increasing mass removal during pumping. Pulsed
pumping may also help remediate stagnation zones by cycling certain well schemes and altering flow
paths. Detailed information can be obtained from Keely, 1989.
Pulsed pumping schedules can be developed based on highly monitored pilot tests, modeling
analysis, or ongoing performance monitoring of hydraulic heads and contaminant concentrations. Special
care must be taken to ensure that the hydraulic containment objective is met during pump rest periods.
43
-------
1000
Fixed well
configuration
I 1
5 10 15 20 25 30 35 40 45 50
Time (y)
Figure 3-3. Adaptive modifications to P&T design and operation can reduce clean-up time (from Hoffman,
1993).
3.2.2.4 Contain the NAPL Zone
Subsurface NAPL can be, a long-term source of ground-water contamination (Figure 1-5) due to:
(1) its low aqueous solubility (that may greatly exceed clean-up standards); and, (2) the inability to
remove all NAPL that is trapped at residual saturation by capillary forces and in dead-end pores. The
mixed containment-restoration strategy shown in Figure 1-2 should be used to contain the NAPL zone
and prevent NAPL migration (that may, perhaps, be induced by pumping) into the P&T restoration area.
Within the NAPL zone, pumping may be used to reduce NAPL mobility by lowering NAPL saturation to
residual. An overview of NAPL pumping techniques is provided by Mercer and Cohen (1990).
3.2.3 Contaminant Monitoring
Samples of ground water taken from wells, soil (or rock) from borings in the contaminated zone,
and treatment plant influent and effluent should be analyzed periodically for contaminant presence to
monitor restoration P&T performance. Sampling locations and frequencies depend on the distribution of
ground-water and contaminant flow velocities within the study area. Mathematical models can be used to
help determine appropriate locations and schedules for sampling ground water and formation solids.
Treatment plant influent and effluent are generally analyzed on a relatively frequent basis to ensure
proper treatment system performance. The degree of monitoring should increase with site complexity.
Various contaminant monitoring considerations are discussed below. Additional relevant information is
provided in Chapter 2.
44
-------
Q.
2
ON
OFF
MAX
z
o
I-
cc
h-
z
UJ
o
z
o
o
TIME
Figure 3-4, The pulsed pumping concept (modified from Keely, 1989).
3.2.3.1 Ground-Water Sampling and Analysis
Ground-water sampling is performed to monitor changes in the contaminant concentration and
distribution during remediation. As described in Section 2.2.6, ground-water samples taken from bej^ond
the restoration area are analyzed to assess hydraulic containment. For restoration P&T, samples should
also be taken from all pump wells and selected observation wells within the contaminant plume to
interpret clean-up progress. An example of a restoration monitoring well network is shown in Figure 3-5.
The number of observation Wells at which samples are taken (in addition to all pump wells) and sampling
frequency depends on site-specific conditions and cost-benefit trade-offs (Section 2.7). In general, greater
sampling density and frequency allows for more adaptive and effective P&T remediation (Figure 3-3).
Turning off pumping wells that produce clean water or do not significantly contribute to hydraulic
containment allows greater resources to be allocated to more highly contaminated zones.
Parameters analyzed should include: (1) the chemicals of concern (or indicator chemicals), (2)
chemicals that could affect the treatment system (such as iron which may precipitate and clog treatment
units if ground water is aerated), and (3) chemicals that may indicate the occurrence of other processes of
interest (e.g., dissolved oxygen, carbon dioxide, nutrients, and degradation products where biodegradation
is considered). As described in Section 2.2.6, relatively detailed analyses should be performed during the
early phase of P&T and sampling frequency should account for probable contaminant velocities.
Background wells located upgradient or cross-gradient of contaminated ground water should be
monitored to indicate if contaminants have migrated beyond the containment zone (e.g., as might occur
where injected water drives contaminated ground water outward). These wells should also be monitored
45
-------
<§>
Potential
Receptors
X
®
®
® Extraction Wells
*» Gradient Control Points
© Injection Wells
X Monitoring Wells
• Off-site Background
Quality Wells
.r.rr.rr.rr
v
X'
\
1 \
V
A «
x x • •
Y Bldg ' Bldg
Y 21
i
j ' ' Bldg 3
IL. \
. . . .
/
/
'I
-^P/operty Boundary—'
^ 0
e
Groundwater
Flow Direction
Figure 3-5. Example of a ground-water monitoring well network at a P&T remediation site (modified from
USEPA, 1992a).
-------
to detect offsite contamination that may be confused with onsite data. Additional guidance on monitoring
ground-water quality beyond the restoration area is given in Section 2.2.6.
Increasing or decreasing contaminant concentration trends in individual wells may not directly
reflect overall clean-up performance. A heterogeneous initial contaminant distribution and flow pattern
changes caused by pumping will result in different portions of the restoration area becoming more or less
contaminated with system operation (e.g., Figure 3-6). For example, contaminant concentrations in
restoration pump wells near the plume perimeter will generally decrease quickly as clean water from
beyond the perimeter flows inward to these wells. Conversely, concentrations may increase at locations
along the flowpath of highly contaminated ground water to pumped wells.
Projections of concentration trends from individual wells can be used to assess clean-up times.
The first indication of contaminant tailing is usually revealed by concentration histories of individual
wells. The statistical methods discussed in Chapter 4 can be applied to evaluate trends and test for an
asymptote (near zero-slope) on an individual well basis. In many cases, individual well results will show
contradictory trends due to plume movement and/or statistical errors associated with sampling and
analysis. The difficulty with variable projections from individual wells can partially be overcome by
evaluating the total restoration P&T performance as described in Sections 3.2.3.3 and 3.2.4.
1000003
.0 10000
Q.
Q.
C
o
o
o
o
o
1000i
1 10 100 1000
Elapsed Time (days)
Figure 3-6. Simulated trends of VOC concentration in ground water pumped from ten extraction wells
during a P&T operation.
47
-------
3.2.3.2 Sampling Aquifer Material
Periodic sampling and chemical analysis of aquifer materials from representative locations in the
contamination zone provides a measure of contaminant removal during P&T operation. The
heterogeneous distribution of subsurface materials, including contaminants, must be considered when
determining sampling requirements, selecting sample locations, and interpreting contaminant mass data.
Unfortunately, high costs will usually preclude acquiring sufficient data to reliably estimate the
magnitude of trends when dealing with sorbed and residual phase contaminants. At most sites, it will be
preferable to analyze soil samples from many locations infrequently (e.g., at intervals needed to sweep at
least two PVs through a formation) than to analyze fewer samples more frequently. Even where mass-in-
place cannot be reliably estimated, a consistency check can be performed by comparing contaminant mass
data with other P&T monitoring data.
Measuring natural organic carbon content in the formation can also provide useful information
for estimating sorption of hydrophobic contaminants (see Section 1.2). Determining natural total organic
carbon, however, is confounded where the porous media are contaminated with anthropogenic organic
contaminants. Methods and considerations for collecting total organic carbon data are provided by
Powell (1990). Retardation of hydrophobic contaminants migrating toward recovery wells and desorption
of hydrophobic contaminants from organic carbon can greatly extend the time required for aquifer
restoration using P&T (see Section 3.3).
3.2.3.3 Treatment System Influent and Effluent
Sampling and analysis of treatment system influent and effluent must be performed regularly to
assess: (1) treatment system performance, (2) changes in influent chemistry that may affect treatment
effectiveness, and (3) dissolved contaminant concentration trends. The performance of individual
treatment units within a treatment train (e.g., where water is pumped through a clarifier to remove metal
hydroxides, and then into an air stripper followed by an activated carbon filter to remove VOCs) are
similarly monitored by periodic analysis of samples taken between units. Such monitoring will provide
data necessary to: (1) estimate total mass removed from system, individual treatment unit loadings, and
estimated breakthrough times; (2) document compliance with discharge requirements; and (3) identify the
need to modify, replace, or regenerate system components. Treatment system monitoring criteria should
be specified in the O&M manual (Section 2.4).
The concentration of influent to the treatment plant can be plotted versus time to evaluate the
trend of ground-water cleanup. Careful consideration, however, must be given to the contaminant
distribution and ground-water flow patterns to the pump wells when interpreting this data. A variation of
this analysis involves computing the trend of contaminant concentration versus the NPV extracted, rather
than concentration versus time. This approach accounts for variations in pumping rates. A limitation of
focusing on treatment plant influent data is that it may not be representative of clean-up progress
throughout the plume.
48
-------
3.2.4 Restoration Measurement Frequency Summary
The hydraulic head and chemical sampling frequency recommendations for containment
discussed in Sections 2.2.1.4 and 2.2.6.3 apply to restoration P&T. Some additional aspects of
measurement frequency, however, need to be considered for restoration monitoring. As described above,
determining the frequency and density of sampling for chemical analysis depends on site-specific
conditions (including the distribution of contaminant velocities and pore volume sweep rates induced by
P&T operation) and cost-benefit trade-offs. Adaptive modification of pumping locations and rates means
that it may also be beneficial to revise sampling locations and frequency.
Minimum restoration measurement frequencies cannot be reasonably specified due to the site-
specific nature of P&T remediation. Typical measurement frequencies, however, include: (1) daily to
monthly analyses of contaminant concentrations (or indicator parameters) in treatment system influent
and effluent; (2) monthly to yearly analysis of contaminant concentrations in ground water sampled from
all pump wells and specified observation wells; (3) infrequent analyses of aquifer solids (e.g., at intervals
needed to sweep at least two PVs through a formation volume); (4) weekly to monthly hydraulic head
surveys to monitor flow directions and rates; (5) continuous (using flowmeters) to weekly monitoring of
individual well pumping rates; and, (6) continuous flowmeter measurement of the combined inflow to
treatment units.
5.2.5 Evaluating Contaminant Concentration and Distribution Trends
Contaminant distribution trends in ground water and aquifer materials should be examined to
assess restoration progress. Performance measures based on concentration decreases are discussed in
Chapter 4. Other performance measures are based on mass removal rates and contaminant mass-in-place
trends. Specifically, these include: (1) the rate of contaminant mass removed by pumping (mass/year);
(2) the rate of reduction of contaminant mass-in-place (mass/year); and (3) the rate of reduction of the
volume of aquifer contaminated above MCLs or other standards (volume/year). A determination of
contaminant mass-in-place, both dissolved and total, is necessary to apply these performance criteria.
3.2.5.1 Estimating Contaminant Mass-in-Place
A meaningful analysis of P&T performance can be obtained by comparing the contaminant mass
removed versus dissolved contaminant mass-in-place. The dissolved mass-in-place (Mw) of a
contaminant at a specific time is given by:
€wbdxdy (3-4)
where n is the formation porosity, Cw is the dissolved contaminant concentration, b is the plume
thickness, and A is area of the plume.
49
-------
The total contaminant mass-in-place (MT) in the saturated zone, discounting NAPL presence, is
more difficult to estimate than dissolved contaminant mass-in-place because of the additional data
requirements. MT can be estimated based on chemical analyses of ground water and solid samples as:
I
Mr = I (nCw + pbCs)b dxdy (3.5)
where Cs is contaminant concentration in the solid media, and pb is the formation bulk density.
Alternatively, MT can be approximated using the partition coefficient, Kd (see Section 1.2), as:
-I
Mr = I (nCw + pbKdCw)b dxdy (3-6)
Determining mass-in-place prior to and during remediation is frequently complicated by a paucity
of available data, particularly with regard to estimating Kd, and Cs distributions, and therefore, MT. The
presence of NAPL can also confound application of mass-in-place performance measures. Where
present, NAPLs will usually account for a dominant portion of the M,, but estimation of NAPL mass is
subject to a very high level of uncertainty. If undetected, NAPL presence may cause misinterpretation of
mass removed versus mass-in-place trends.
Determining mass-in-place necessitates defining the "plume". This is generally not
straightforward because it involves interpolating sparse data to develop a continuous plume distribution.
There are several means to interpolate sparse data (Jones et al., 1986), including hand contouring which
takes into account the experience, knowledge, and bias of the individual performing the contouring.
Computer software packages are used to contour large amounts of data (Hamilton and Jones,
1992). To determine mass-in-place, interpolation is usually performed on contaminant concentration
values or the logarithm of these values. It is especially important that a log transformation be made for
"spiked" plumes to improve data fitting without significant loss of peak values. Ground-water quality
analyses at contamination sites determine "detect" and "non-detect" values. Whereas significant detects
are the basis for interpolation, the non-detects pose problems. Although non-detect sample locations may
clearly indicate the outermost possible extent of the plume, it is often difficult to delineate the true extent
of contamination. Additionally there are often areas lacking any data. In these areas, the contouring
packages are unbounded and may extrapolate data poorly. It may be necessary to provide boundary
clarification with "dummy" zero concentration points.
Even computer-based contouring is subjective in that different contouring methods produce
different results. Most applications are based on contouring two-dimensional isopleths, although
contouring of three-dimensional isopleths is possible. Numerous contouring software products, many of
which were developed in the mining, petroleum and civil engineering fields, are available commercially.
Contouring routines are also incorporated as modules of Geographical Information Systems.
50
-------
Several interpolation techniques are available for estimating mass-in-place including:
. Graphical methods baaed on contoured concentration data (e.g., inverse distance raised to a
power of 2,4,6, or higher);
. Kriging (universal and unique variograms); and
. Triangulated Irregular Network (TIN).
A brief introduction to these methods follows.
The graphical method involve! calculating the mass within each interval cf a concentration
contour map by measuring the interval area and multiplying it by the plume thickness, porosity, and
contour concentration (or mean of the contour values bounding the interval area). This method cannot
easily account for nonuniform porosity or plume thickness. Kriging is an advanced geostatistical
technique that potentially can provide the best estimate of mass-in-place. Kriging, however, requires
considerable experience for proper application. The TIN method is a simple numerical integration
approach commonly used to estimate volumes in civil engineering applications. The procedure involves
determining the optimum network of triangles to connect monitor and extraction wells and then
evaluating a mass-in-place equation (such as 3-4) for each triangle. Different numerical approximations
are obtained using different interpolation functions. Appendix A is the documentation of a TIN computer
program that assumes linear interpolation over the triangle area. The program is included on a computer
disk with this document. The TIN method can account for nonuniform plume thickness and porosity.
3.2.5.2 Determining Rate of Contaminant Mass Removal
The rate of contaminant mass removal (M^) can be determined by sampling treatment plant
influent for the constituents of concern and then multiplying the dissolved concentration (C^) of
contaminant (i) by the total flow rate (QT):
(3-7)
This estimate can be compared to a calculation of M^ using data collected at each extraction well (j):
n
W = X qjCwij (3-8)
where n is the number of extraction wells, qj is the pumping rate of well (j), and Cwij is the dissolved
concentration of contaminant (i) pumped from well (j). These two estimates of mass removal rate should
be comparable, but not necessarily identical, due to (1) variability of analytical results and (2) difference
in the sum of individual well flow measurements and the measurement of treatment plant inflow.
51
-------
3.2.5.3 Comparing Mass Removal and Mass-in-Place Trends
Restoration progress can be assessed by comparing the rate of contaminant mass removal (e.g.,
plotted as cumulative mass removed) to the dissolved and/or total contaminant mass-in-place. If the rate
of contaminant mass extracted approximates the rate of dissolved mass-in-place reduction, then the
contaminants removed by pumping are primarily derived from the dissolved phase. This is illustrated for
trichloroethene in Figure 3-7, which shows mass removed as a mirror image of mass-in-place.
Conversely, a contaminant source is indicated where the mass removal rate greatly exceeds the rate of
dissolved mass-in-place reduction. The source may be NAPLs, contaminants sorbed to formation solids,
an uncontained disposal area, or dissolved contaminants diffusing from low-permeability strata. Site
hydrogeology and contaminant properties should be evaluated to determine if source removal and/or
containment, and/or system modifications could improve P&T performance.
The time needed to remove dissolved contaminants can be projected by extrapolating the trend of
the mass removal rate curve or the cumulative mass removed curve. If the mass removal trend indicates a
significantly greater clean-up duration than estimated originally, the conceptual model of contaminant
distribution may need to be reevaluated, and system modifications may be necessary. The effect (or lack
of effect) of P&T system modifications will be evidenced by the continuing mass removal rate and
cumulative mass removed trends.
Progress inferred from mass removal rates can be misleading, however, where NAPL and sorbed
contaminants are present (e.g., the mass removed will exceed the initial estimate of dissolved mass-in-
place). Interpretation suffers from the high degree of uncertainty associated with estimating NAPL or
sorbed contaminant mass-in-place. Stabilization of dissolved contaminant concentrations while mass
removal continues is an indication of NAPL or solid phase contaminant presence. Methods for evaluating
the potential presence of NAPL are provided by Cohen and Mercer (1993), Feenstra et al. (199 I), and
Newell and Ross ( 1992).
Mass removal rates are also subject to misinterpretation where dissolved contaminant
concentrations decline rapidly due to: (1) mass transfer rate limitations to desorption, NAPL or
precipitate dissolution, or matrix diffusion; (2) dewatering a portion or all of the contaminated zone;
(3) dilution of contaminated ground water with clean ground water flowing to extraction wells from
beyond the plume perimeter; or (4) the removal of a slug of highly contaminated ground water.
Contaminant concentration rebound will occur if pumping is terminated prematurely in response to these
conditions.
3.3 PROJECTED RESTORATION TIME
The projected restoration or clean-up time is site specific and varies widely depending on
contaminant and hydrogeologic conditions and the clean-up concentration goal. For example, clean-up
time in homogeneous transmissive aquifers contaminated with mobile dissolved chemicals may be on the
order of several years. NAPL sites or sites with sorbed contaminants in heterogeneous aquifers, however,
may require decades or centuries of P&T operation to reach clean-up levels with currently available
52
-------
CQ
Q.
CL
(0
a
4.0E + 12
3.5E+12
3.0E+12
2.SE+12
Dissolved TCE mass-in-place
Cumulative mass of TCE
removed from extraction wells
83 Jun-84 Jun-85 Jun-86 Jun-87 JuiJBS
B 2.0E+12
•«"••
5
O
39 Jun-90 Jun-91 Jun-92
Figure 3-7. Comparison of cumulative mass of TCE removed versus dissolved TCE mass-in-place during
P&T operation at the Air Force Plant 44 in Tucson, Arizona.
-------
technology. Further, the length of time for restoration is usually difficult to estimate due to complications
associated with characterization of the processes that limit cleanup (see Section 1.2).
The determination of restoration time is necessary to evaluate whether clean-up goals are
practical and for choosing the most efficient remediation system. To demonstrate the magnitude of clean-
up time variation at sites with different conditions, several examples are included herein. The first
example illustrates a simple method to estimate the time required to extract mobile dissolved
contaminants in a homogeneous aquifer (Hall, 1988). Assume that ground water in a 55-ft thick aquifer
with a 0.3 storage coefficient is contaminated by conservative solutes throughout a ten acre area. The
pore volume of the contamination zone is approximately 54,000,000 gallons, which, under ideal
conditions, could be removed after one year of pumping at approximately 100 gpm. In reality, however,
actual sites are not this simple and P&T hydraulics cannot be managed to prevent inflow of ground water
from beyond the plume perimeter. To remove one pore volume from the plume requires pumping a
greater volume of ground water. Geologic and chemical complexities can add years, decades, or longer to
clean-up time due to processes described in Chapter 1 that cause tailing.
Diffusion also complicates clean-up time calculation. Conservative contaminants that have
migrated (by any process) into less permeable strata in heterogeneous media will slowly diffuse into the
more permeable zones during P&T operation. This diffusion may dictate the time necessary for complete
remediation. For example, consider an aquifer with clay lenses that was contaminated for a long time
before P&T operation reduced dissolved concentrations in the permeable strata, but not in the clay, to
below clean-up standards. The area! extent of the clay is such that an approximation of one-dimensional
diffusion out of each lens can be used to help estimate the time needed to deplete contaminants in the
clay. The concentration gradient from the center to the edge of each clay lens can be approximated as
unity if we assume relative dissolved contaminant concentrations of one (maximum concentration) in the
center of each clay lens and zero (clean) in the permeable strata. The time for conservative contaminants
to diffuse out from the clay center under these circumstances is:
t = m2/Da
where m is half of the clay lens thickness, Da is the contaminant's apparent diffusion coefficient, and
where
Da = D/aR
where R is the retardation coefficient, a is tortuosity (usually = 1.3 to 1.5), and D is the aqueous diffusion
coefficient. Da is the water diffusion coefficient modified to reflect tortuosity of the porous medium and
sorption of the contaminant. The water diffusion coefficient for tetrachloroethene (a nonconservative
contaminant), for example, is 7.5 x 10" cm /sec (Lucius et al., 1990), yielding a corresponding Da value
of 1 x 10" cm/sec. Using this value, for clay lenses that are 0.2, 1,2, and 4 ft thick, the times for
contaminants to diffuse from the center of the clay lenses are 0.29, 7.36, 29, and 118 years. In reality, the
time required to reduce contaminant concentrations to very low levels may be much longer because the
concentration within the clay will decline slowly and the concentration gradient will be less than unity.
54
-------
Sorption and desorption also cause tailing, concentration rebound, and slow ground water
restoration. The number of pore volumes which must be passed through a contamination zone to attain
clean-up standards increases with the sorptive tendencies of a contaminant and kinetic limitations to the
rate of desorption (Keely, 1989). An example of this process is demonstrated by a numerical model to
evaluate a proposed P&T configuration for the Chem-Dyne site in Hamilton, Ohio (Ward et al, 1987; see
Chapter 5). Due to the simulation of linear contaminant partitioning between soil and water, a nearly
linear relationship was found to exist between retardation and the duration of P&T operation (or the NPV
pumped) needed to attain a specific ground-water clean-up standard. Other investigators have concluded
that nonlinear sorption may further increase the time required for ground-water cleanup using P&T
technology. For example, Stephanatos et al. (1991) recommend using site-specific leaching tests to assess
sorption, and that, in lieu of such tests, they suggest using USEPA's Organic Leachate Model (OLM) (51
Fed. Reg. 21,653, June 13, 1986; 51 Fed. Reg. 27,062, July 29, 1986; 51 Fed. Reg. 41,088, November 13,
1986) as a more realistic approach to setting ground-water based soil clean-up goals. To illustrate their
point, Stephanatos et al. (1991) present data from the Whitmoyer Laboratories CERCLA site. Nonlinear
sorption for an iron-arsenic compound was determined from soil leaching tests. Based on these results, an
estimated clean-up time of 50,000 years would be required to reduce arsenic concentrations in ground
water to below 0.05 mg/1 using conventional P&T technology. Assuming linear sorption, the restoration
time was underestimated to be about 160 years.
Another complexity in estimating clean-up times for P&T systems involves the presence of
NAPL. Where NAPL is present, it will slowly dissolve, creating a continuing source to ground-water
contamination until the NAPL mass is depleted. Flow rates during P&T may be too rapid to allow
residual NAPL to dissolve to its effective aqueous solubility limits. As such, the contaminated water is
advected away from the NAPL residuals prior to reaching chemical equilibrium and is replaced by fresh
water from upgradient. This has the same ramifications as other processes that cause tailing in that large
volumes of water with low concentrations may be pumped during P&T operation. Several relationships
have been derived to predict dissolved concentrations and time required to deplete residual and pooled
NAPL sources (Cohen and Mercer, 1993). These indicate that NAPL can persist as a source of ground-
water contamination for decades or longer.
Guidance for estimating ground-water restoration times using batch and continuous flushing
models is provided by USEPA (1988b). The batch flushing model is based on a series of consecutive
discrete flushing periods during which contaminated water in equilibrium with adsorbed contaminants is
displaced from the aquifer pore space by clean water. Values of contaminant concentration in soil and
water are calculated after each flush. An example of an analogous method (and corrections) to this batch
flushing model are provided by Zheng et al. (1991, 1992). The batch and continuous models assume that:
( 1) zero-concentration influent water displaces contaminated ground water from the contamination zone
by simple advection with no dispersion; (2) the clean ground water equilibrates instantaneously with the
remaining adsorbed contaminant mass; (3) the sorption isotherm is linear; and (4) chemical reactions do
not affect the sorption process. Care must be taken to avoid relying on misleading estimates of restoration
time that may be obtained by using these simplified models.
55
-------
The relatively simple calculations provided in this section demonstrate some of the difficulty in
estimating clean-up time. It is obvious that long periods of P&T operation will be required to attain
drinking water clean-up standards at many sites. Although more sophisticated modeling techniques are
available (NRC, 1990), their application usually suffers from data limitations, resulting in uncertain
predictions. Nevertheless, clean-up time analyses are needed to assess alternative remedial options and
to determine whether or not clean-up goals are feasible.
3.4 ADDITIONAL CONSIDERATIONS
Pilot tests and phased implementation of restoration P&T are recommended to improve
understanding of site conditions and thereby address complex and costly remediation in an effective and
efficient manner. Discussions of modeling and operational efficiency provided in Sections 2.6 and 2.7 are
very relevant to restoration P&T. Similarly, O&M and monitoring plans noted in Sections 2.4 and 2.5,
respectively, should be developed for restoration P&T systems.
56
-------
4. EVALUATING RESTORATION SUCCESS/CLOSURE
4.1 INTRODUCTION
Ground-water restoration (as operationally defined) is achieved when a predefined clean-up
standard is attained and sustained. To ensure that these conditions are met, the procedure as outlined in
Figure 4-1 should be followed. To protect human health and the environment, clean-up standards and/or
containment objectives first must be set to define the goals of the remediation. Clean-up standards are
site-specific and depend on the contaminants present, the risk imposed by those contaminants, and the
fate of those contaminants in the subsurface. Clean-up standards include Maximum Contaminant Levels
(MCLs), Alternate Concentration Limits (ACLs), detection limits, and natural water quality. Guidelines
for selection of clean-up standards are provided in Guidance on Remedial Actions for Contaminated
Ground Water at Superfund Sites (USEPA, 1988).
Much of the information in this chapter follows closely material in USEPA (1992c), with an
attempt to minimize duplication. Figure 4-2 is provided to show the stages (indicated by the circled
numbers) of remediation using water quality data from a single well. During the first stage, the site is
evaluated to determine the need for and conditions of a remedial action. Once the remediation system is
started, concentrations at most wells will decrease as shown for stage 2 in Figure 4-2. Concentrations will
fluctuate around the trend due to seasonal changes, fluctuations due to the heterogeneous distribution of
chemicals in the subsurface, changes in pumping schedules, variations in sample collection, and lab
measurement error.
Based on both expert knowledge of the ground-water system and data collected during P&T, the
time to terminate treatment will be determined (stage 3). For system termination, all wells on the site
should be monitored and analyzed individually for compliance unless site-specific conditions dictate
otherwise. Data analysis may indicate that clean-up standards will not be achieved, and other
technologies and/or goals may be assessed. For remediation systems that have terminated, the transient
effects resulting from remediation will take time to dissipate (stage 4). Monitoring during this time
period is referred to as post-termination monitoring. After the ground-water flow system has reached a
post-remediation equilibrium, sampling to assess attainment of the clean-up standards begins (stage 5).
At stage 6, data collected during stage 5 is used to determine if the clean-up standard has been attained.
Due to fluctuating concentrations over time, the average concentration over a short period of
time may be different from the average over a long period of time. Statistical decisions and estimates that
only apply to the sampling period of approximately one year or less, are referred to as short-term
estimates (USEPA, 1992c). Decisions and estimates that apply to the foreseeable future are called
long-term estimates, and assume that ground-water processes can be described in a predictable manner.
Long-term estimates are used to assess attainment, whereas short-term estimates are used to make interim
management decisions.
Short-term analyses as applied to the P&T system operation (stage 2) are presented in Section
4.2. The methods described include both parametric and nonparametric analyses. Long-term analyses
57
-------
Define Attainment Objectives and
Clean-up Standard
Develop Sampling and Analysis
Plan for Performance Evaluation
Continue/
Modify
Treatment
Assess and Revise
Treatment Design as
Necessary
Monitor System Performance
Terminate Treatment
Allow System to Reach
Steady-State
Verify the Attainment of
Clean-up Standard
Is the Clean-up
Standard Maintained
Over Time?
Is the Clean-up
tandard Reached?
Demonstration of
Technical Impracticability
Modify Remedial
Action Objectives
,Yes
Monitor as Necessary
Figure 4-1. Determining the success amlAw-dosure of a P&T system.
58
-------
Start
Treatment
Bid Start
Treatment Sampling
End Sampling
Declare Clean or
Contaminated
Measured
Ground
Water
Concentration
0.2
Date
Figure 4-2. Example contaminant concentrations in a well at P&T site (USEPA, 1992c).
used to determine the time of treatment termination (stage 3) are presented in Section 4.3. Long-term
analyses include parametric trend analyses, nonparametric trend analyses, and time-series analyses. Post-
remediation monitoring of ground-water concentrations and water levels (stage 4) is discussed in Section
4.4. Section 4.5 discusses the general statistical methods used to determine if ground-water conditions
after P&T system termination will remain below the site clean-up standard (stage 5 and 6).
4.2 SYSTEM OPERATION: SHORT-TERM ANALYSES
Statistical methods for analyzing short-term trends (stage 2) answer questions of the following
nature:
Are concentrations in individual wells at the site currently below the clean-up standard? To
what degree of confidence is this true?
Is the average sitewide concentration currently below the clean-up standard? To what degree
of confidence is this true?
Is the current sampling program sufficient to make inferences about concentration trends?
Are there sections of the plume where clean-up standards have been met with confidence?
Short-term analyses consist of parametric and non-parametric techniques; i.e., those statistical
analyses that can be performed on data that has a known distribution, and those data whose distribution is
unknown or non-normal, respectively.
59
-------
A set of concentration measurements taken over a year (short-term) can be described through
simple sample statistics such as sample mean, standard deviation, standard error and percentile. The
sample mean for this population characterizes the average concentration for all wells over the year.
Sample-based comparisons can be made using hypothesis testing of differences between the sample mean
and the site clean-up goal or other standards. Statisticians use the standard error, or sample variability to
characterize the precision of samples-based comparisons through confidence intervals. Confidence
intervals delineate a range of values within which the true value is expected to exist within a specified
level of confidence. The standard error of the mean concentration provides a measure of the precision of
the mean concentration obtained from ground-water samples taken over the year. The appropriate
method used to calculate the standard error of the mean for a short-term analysis depends on the behavior
of contaminant measurements over time, and the sampling design used for sample collection. Corrections
to the standard error of the mean must be made if the data are collected systematically (at specified
intervals), if there are seasonal patterns, if the data are serially correlated, and if there are trends in the
data (see USEPA, 1992c).
4.2.1. Parametric Tests
Once sample statistics have been developed, simple hypothesis testing can be used to determine if
the mean of the sample population (ground-water concentrations) is less than the clean-up standard. The
following procedure describes simple hypothesis testing (USEPA, 1992c).
(1) Assume that the mean concentration of the collected data is greater than the clean-up standard
as the null hypothesis. The clean-up standard therefore represents the null hypothesis of the
analysis.
(2) Collect a set of data representing a random sample from the population of interest (e.g.,
concentrations over the year).
(3) Develop a statistical test from the sample data. Assuming that the null hypothesis is true,
calculate the expected distribution of the statistic.
(4) If the value of the statistic is consistent with the null hypothesis, conclude that the null
hypothesis provides an acceptable description of the analyses made.
(5) If the value of the statistic is highly unlikely given the assumed null hypothesis, conclude that
the null hypothesis is incorrect.
If the chance of obtaining a value of a test statistic beyond a specified limit is, for example,
5 percent, and the null hypothesis is true, then if the sample value is beyond this limit, substantial
evidence exists that the null hypothesis is not true and the mean concentration is less than or equal to the
clean-up standard. An example of simple hypothesis testing for a short-term concentration mean of the
data provided in Table 4-1 is presented in Box 4-1. If comparisons of means to clean-up standards are
repeated periodically, a general evaluation of the remediation can be made. With time, the variance of
60
-------
concentration, as measured by standard deviation, should decrease as the system reaches "steady-state"
and concentrations are reduced by dispersion effects and remediation. For this reason, confidence levels
about the mean concentration may be increased. Hypothesis testing of this nature assumes the data
roughly represent a normal distribution. If the sample data set is limited, and the sample distribution is
unknown, then nonparametric analyses must be performed.
4.2.2. Nonparametric Tests
A nonparametric analysis is used when the raw concentration data have been found to violate the
normality assumption (based on a chi-squared or other normality test), a log-transformation fails to
normalize the data, and no other specific distribution is assumed (USEPA, 1989). Similar to the
parametric analysis, a nonparametric analysis produces a simple confidence interval that is designed to
contain the true or population median concentration with specified confidence. If this confidence interval
contains the clean-up standard, it is concluded that the median concentration does not differ significantly
from the clean-up standard. If the interval's lower limit exceeds the clean-up standard, this is statistically
significant evidence that the concentration exceeds the clean-up standard.
To compare the median site concentration to the site clean-up standard using a nonparametric
analysis, an approach outlined in USEPA (1989) for compliance at RCRA facilities can be applied.
TABLE 4-1. DISTRIBUTION OF MONOCHLOROBENZENE.
Concentration of Concentration of
Monochlorobenzene Monochlorobenzene
Well ID in ppb WellW in ppb
MW-1
MW-2
MW-3
MW-4
MW-5
MW-6
MW-7
MW-8
MW-9
MW-IO
MW-11
MW-12
MW-13
86
109
85
84
91
65
99
107
115
167
58
66
89
MW-14
MW-15
MW-16
MW-17
MW-18
MW-19
MW-20
MW-21
MW-22
MW-23
MW-24
MW-25
MW-26
76
55
87
105
75
53
135
113
84
83
19
118
21
61
-------
Box 4-1 . Short-term Hypothesis Testing
Table 4-1 represents the distribution of monochloroben/ene at 26 wells at a site.
Wells at the site were sampled at approximately the same time. The data are assumed to
be normally distributed, not affected by seasonal effects, and not serially correlated.
The mean concentration value of these data is 86.3 ppb. The standard deviation of these
data is 32 ppb. The standard error of the mean concentration is 6.3 ppb and is given by
the following equation:
where s is standard deviation, &j is the standard error of the mean, and N represents the
number of samples taken.
A one-sided confidence interval can be calculated by
x + l l-o; N-l ^jc
where x is the mean value of the sample population, and t^ N_, is the t statistic for N-l
degrees of freedom at an a level of significance.
The clean-up standard for monochlorobenzene at the site is 100 ppb. For 25 degrees
of freedom (N-l) and a 95 percent confidence level (a- 0.05), t^ N-1 is 1.671 (USEPA,
1992c; Appendix A.I). The one-sided confidence interval yields 96.8 ppb. If, in the
null hypothesis, the mean concentration was assumed to be greater than 100 ppb, then
the null hypothesis is incorrect to a 95 percent confidence level; i.e., 96.8 ppb is less
than 100 ppb. The site mean concentration is highly likely (to a 95 percent confidence
level) to be less man the clean-up standard.
This method requires a minimum of seven observations. This procedure as outlined is as follows
(USEPA, 1989):
(1) Order the n data from least to greatest, denoting the ordered data by X( l),...X(n), where X(i)
is the critical value in the ordered data.
(2) Determine the critical values of the order statistics. If the minimum seven observations are
used, the critical values are 1 and 7. Otherwise, find the smallest integer, M, such that the
62
-------
cumulative binomial distribution (see Conover, 1980) with parameters n (the sample size) and
p = 0.5 (0.5 quantile) is at least 0.99. For sample sizes from 4 to 11 values of M and the n+1-
M together with the exact confidence coefficient can be found in Table 6.3 of USEPA (1989).
For larger sample sizes, take as an approximation the nearest integer value to
M = n/2+l+Zo.99 V(n/4) (4-1)
where Z099 is the 99th percentile from the normal distribution and equals 2.33.
(3) Once M has been determined in Step 2, find n+l-M and take as the confidence limits the
order statistics, X(M) and X(n+l-M).
(4) Compare the confidence limits found in Step 3 to the clean-up standard. If the lower limit,
X(M) exceeds the compliance limit, there is statistically significant evidence of
contamination. Otherwise, ground water is within the clean-up standard.
Both the nonparametric and parametric tests for short-term analyses described provide
comparison between the mean of the site data and the site clean-up standard. Other comparisons can be
made against the median, percentiles or proportions of concentration data for both parametric and
nonparametric analyses (see USEPA, 1992c; Helsel and Hirsch, 1992; and Gilbert, 1987).
4.3 TREATMENT TERMINATION: LONG-TERM ANALYSES
Analyses of long-term concentration trends provide models that can be used in P&T system
termination decisions (Stage 3) and to determine if the goals of the remediation are feasible. Several
statistical methods of evaluating long-term concentration trends exist. These methods include regression
analyses (trend analyses) and time-series analyses.
It is important to note that changes in system stress (e.g., pumping rate changes or an external
influence, such as seasonal fluctuations in recharge), can result in changes in concentration variation, and
correlation. These "fluctuations" can make regression analyses difficult. However, certain trends can be
removed from the data prior to regression analyses. To determine if the data are serially correlated, the
Durbin-Watson test can be applied. Methods for correcting for serial correlation are described in USEPA
(1992c, Section 6.2.4).
4.3.1 Parametric Trend Analyses
A regression or trend analysis of ground-water contaminant levels provides information on
concentration level trends over time and predicted concentration levels in the future. Regression analysis
techniques fit a theoretical curve or model to a set of sample data. Actual time-concentration data is
replaced by a model that can be used to predict concentrations within a specified confidence or prediction
interval. By applying confidence intervals to a regression line fit, the following assumptions are made:
63
-------
The assumed model or fitted-curve form is correct.
The data used to fit the model are representative of the data of interest.
The variance of the residuals is constant with time. A residual is the difference between the
observed concentration measurement and the corresponding concentration value predicted by
the regression model.
Residuals are independent, and, therefore, free from serial correlation. Serial correlation is
the interdependence of residuals in a time sequence.
Residuals are normally distributed.
Sources of variability that can cause the data collected not to be normally distributed include
(API, 1991):
Seasonal or short-term natural fluctuations.
Spatial heterogeneity in the contaminant distribution in the aquifer so that water volumes
containing variable amounts of contaminants flow past a fixed sampling point.
Sampling errors such as the collection of non-representative samples, or not using the
sampling technique consistently over time.
Sample handling or preservation problems so samples contain different amounts of
contaminant at the time of analysis than were present at the time of collection.
Analytical variability caused by (a) differences in analytical technique and instrumentation
among different laboratories or within a given laboratory over the long term, and (b) intrinsic
imprecision in analytical measurements.
Formal tests for normality include the Shapiro-Wilk test, the Shapiro-Francis test and the
Kolomogorov-Smimov test (USEPA, 1989). A relatively simple way for checking the normality of
residuals is to plot the residuals ordered by size against their expected values under a normal distribution
(USEPA, 1992c). Under normality, the residuals against their expected values should plot as a straight
line.
Both straight-line and curvilinear regression models can be used. The initial choice of regression
model can be made by observing a plot of the sample data over time (USEPA, 1992c). Straight-line
regressions are appropriate if a plot of concentration versus time forms a straight line. For most P&T
systems, long-term concentration declines will be curvilinear; i.e., concentration versus time does not
form a straight line when plotted. Under certain circumstances, however, the concentrations versus time
relationship can be modeled as a straight line by transforming either the dependent or independent
variable (USEPA, 1992c) (i.e., log linear).
64
-------
Regardless of the model of regression analysis, an assessment of the fit of the theoretical curve or
regression model to existing data is required. The diagnostic statistical parameters that test the fit of the
theoretical curve include:
SSE = Sum of Squares Due to Error is a measure of how well the model fits the data. If the
SSE is small, the fit is good, if it is large, the fit is poor.
MSB = Mean Square Error provides an estimate of the variance about the regression. The
lower the MSB the better the fit.
•^
R = Coefficient of Determination represents the proportion of the total variance in the
observed value that is accounted for by the regression model. A value of R close to 1
9
represents a good fit of the data to the regression line. Low values of R can indicate either a
relatively poor fit of the model or no relationship between the concentration levels and time
(USEPA, 1992c). The fit of the model should not be judged based solely on the
corresponding R value.
A full description of the development and application of these parameters for both straight and curvilinear
regression is provided in the USEPA (1992c) guidance document entitled "Methods For Evaluating the
Attainment of Cleanup Standards, Volume 2: Ground Water."
Once the fit of the regression line has been assessed, predictions and conclusions about trends and
future concentration values can be made. These determinations can be compared to clean-up standards to
decide whether or not remediation can be terminated.
One termination analysis method that can be applied is the zero-slope method (USEPA,. 1989).
This method requires the demonstration that contaminants have stabilized at a level below the clean-up
standard and will remain at that level with time (zero slope). Typically, ground-water concentrations in a
P&T system "level off with time and trend toward an asymptotic limit (with a slope of zero). An
example of the application of the zero-slope method to concentrations trends in Box 4-2 uses data in
Table 4-2. In this example, the slope calculated for the best fit regression line is compared to zero by
determining the standard error of the estimated slope of the regression line. By knowing the standard
error, the degree of confidence in the estimated regression line slope can be determined to quantify the
degree of potential error of the slope estimate. To statistically test if the "steady-state" concentration
level reached at the zero slope point is below the clean-up standard, confidence intervals about the
regression line can be determined. By applying confidence intervals to any conclusion derived from the
regression line, the fitted curve residuals are assumed to be normally distributed as described above. The
concentration trends presented in Table 4-2, the corresponding best fit regression line, and the
corresponding upper 95 percent confidence interval line are depicted in Figure 4-3.
A computer program, REGRESS, has been developed through funding by the American
Petroleum Institute (API, 1992) to assess asymptotic conditions with first order and polynomial
(exponential) regression techniques. This program performs sequential linear regression analyses (e.g.,
65
-------
Box 4-2. Analysis of Zero-Slope Trending Data
The estimated regression line for the last six data points in Table 4-2 and Figure 4-3 has the following
hear equation based on least squares estimates:
C = -0.00022x + 1.93053
where C is the concentration in ppb at a given time and x is the time in months.
A confidence interval about the slope -0.00022 can be used to determine if a downward trend exists.
The confidence interval about the slope is given by:
b, ± ii.s;N-a sl*H)
where b, is the estimated slope of the regression line,
tl-ffl;N-2
is the student t statistic for N-2 degrees of freedom with an a significance level, and s(bO is the Standard
Error of the estimated slope.
The Standard Error of the estimated slope can be determined by
where
MSE
MSE'
C; - the actual concentration at time i; C; - the estimated concentration based on the estimated regression
line at an equivalent time as C; ; ^ - the total number of samples taken; N-2 - the degrees of freedom; i -
the sample time;
i-l
N
i-l
For the data presented in Table 4-2 s(b,) - 0.008815. For a 95 percent confidence level, a - 0.05. The
student t statistic t,. ^ N. 2 for an a of 0.05 is 2.776 (see Appendix A of USEPA, 1992c).
The confidence interval about the slope is -0.00022 ± 2.776 (0.008815), and the slope will range from
0.02425 to -0.02469 within a 95 percent confidence level. The upper confidence value of the slope is
greater than zero and the lower confidence value is less than zero. This signifies that the slope is not
significantly different from zero (i.e., no positive or negative trend exists).
The data presented in Table 4-2 suggest that the clean-up standard (10 ppb) was reached at time 104
days. Based on this information and the fact that mere is a zero slope in concentration, the treatment system
can be terminated, and post-termination monitoring can proceed.
66
-------
TABLE 4-2. CONCENTRATION VERSUS TIME DATA SHOWING AN ASYMPTOTIC ZERO
SLOPE (REGRESSION ONLY PERFORMED ON LAST SIX SAMPLES).
Time, i
(days)
1
8
16
21
37
48
62
104
134
161
189
217
272
302
Actual
Concentration,
a (ppb)
150.0
41.0
41.0
24.0
15.0
51.0
7.0
10.0
3.0
0.3
3.0
0.5
2.0
2.3
Estimated Regressed
Cocentration,
cinppb
1.90
1.89
1.89
1.88
1.87
1.86
upper 95%
Confidence
Concentration in ppb
6.18
5.92
5.76
5.71
5.96
6.27
subsets consisting of the last five data points, the last six data points, etc.) until the final data set regressed
includes all the data. The subset of the regression curve assigned as having approximately a zero slope is
defined to be the asymptote of the concentration values.
Hirsch et al. (1982) showed that if the seasonal cycles are present, and/or the data are not
normally distributed, and/or the data are serially correlated, the true slope as calculated by confidence
intervals may not be correct, in fact, a zero slope may actually occur and not be detected by the regression
analysis.
If asymptotic concentration levels exceed the clean-up standard, then a reassessment should be
made of the P&T methods and goals. P&T operation may need to be modified by increasing pumping
rates, adding new recovery wells, etc. Note, however, that, in some cases, the clean-up standards may be
unobtainable using the best available technology. For this case, if the standards cannot be relaxed based
on institutional controls or a reevaluation of risk, then the remedial goal should be reevaluated and may be
modified to long-term hydraulic containment. Additional guidance is provided in USEPA, 1993.
In a similar application to Box 4-2, it can be statistically determined if concentration levels will
follow a downward trend after reaching the clean-up standard. If this condition occurs, the P&T system
may be terminated. An example of this method, as applied to the concentrations in Table 4-3, is
presented in Box 4-3, and Figure 4-4.
67
-------
12.00 -i
11.00 H
10.00 --
JQ
CL
Q.
C
o
9.00 -
o
"£ 8.00
o
C
o
0
7.00 -
6.00 -
5.00
Upper 95Vi Confidence
Lower 95% Confidence
I i I I i i i i i i i i i I i I
0 1 2 3 4 5 67 8 910111213141516
Time (Months)
Figure 4-3. Best-fit regression line and 95% confidence interval for the concentration trend of data given in
Table 4-2.
68
-------
TABLE 4-3. CONCENTRATION VERSUS TIME DATA SHOWING A DOWNWARD TREND.
Time,i
(months)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Actual
Concentration, Ci
(ppb)
10.6
10.4
9.5
9.6
10.0
9.5
8.9
9.5
9.6
9.4
8.75
7.8
7.6
8.25
8.0
Estimated
Regressed
Concentration, c
inppb
10.44
10.26
10.07
9.89
9.71
9.53
9.34
9.16
8.98
8.80
8.61
8.43
8.25
8.07
7.89
Upper 95%
Confidence
Concentration
inppb
11.49
11.29
11.09
10.89
10.70
10.51
10.32
10.14
9.96
9.78
9.61
9.44
9.27
9.10
8.94
Lower 95%
Confidence
Concentration
inppb
9.38
9.22
9.06
8.89
8.72
8.54
8.37
8.18
8.00
7.81
7.62
7.43
7.23
7.03
6.83
4.3.2 Nonpammetric Trend Analyses
When the residuals from a regression analysis are not normally distributed, or of an unknown
distribution, then nonparametric trend analyses are recommended. Examples of nonparametric trend
analyses include the Mann-Kendall trend test, Sen's nonparametric procedure, and a curve smoothing
procedure, LOWESS (Locally Weighted Scatterplot Smoothing). Each of these methods can be used to
calculate a model for concentration trends over time. Sen's nonparametric procedure can be used to
estimate the magnitude of the trend. When seasonal variation is present in the data, then the seasonal
Kendall test and seasonal Kendall slope estimator may be used to adjust for seasonal effects (Carosone-
Link et al., 1993). Several references which describe nonparametric trend analyses, as applied to water
studies, include Gilbert (1987), Helsel and Hirsch (1992), and USEPA (1992c).
43.3 Time Series Analysis
Time series analysis is very similar in use to regression, except that time series makes predictions
based on serial correlation with trends removed, whereas regression tries to eliminate these correlations
and analyze trends. Three time-series techniques, the general linear model (GLM), auto-regressive
moving average (ARMA), and auto-regressive integrated moving average (ARIMA), may provide some
additional information about the direction in which the mean is trending, and its stability (USEPA, 1989).
These methods are usually computer intensive; their use requires a familiarity with time-series analysis.
69
-------
Bex 4-3. Analysis of Downward Trending Data
The estimated regression line for the data in Table 4-3 and Figure 4-4 has the following linear equation
based on least squares estimates:
C--G.1832X+ 10.62
where C is the concentration in ppb at a given time, x is the time in months.
A confidence interval about the slope -0. 1 832 can be used to determine if a downward trend exists. The
confidence interval about the slope is given by:
where b, is the estimated slope of the regression line;
tj-S; N-2
is the student t statistic for N-2 degrees of freedom with an a significance; and s(bj) is the Standard Error of
die estimated slope.
The Standard Error of the estimated slope can be determined by
where
**• \2
i-i
C; - the actual concentration at time i; C; - the estimated concentration based on the estimated regression
line at an equivalent time as C{; N - the total number of samples taken; N-2 - the degrees of freedom; i - the
N
N
For the data presented in Table 4-3, s(b,) - 0.026173
For a 95 percent confidence level, a - 0.05. The student t statistic t, a f ^ N 2 for an a of 0.05 is 2. 1 60
(see Appendix A of USEPA, 1992c). The confidence interval about die slope is -0.1823 ±2.160 (0.026173)
and the slope will range from -0.1258 to -0.2388 within a 95 percent confidence level. The negative slope
within the confidence interval strongly suggests mat concentrations are on a downward trend.
The upper confidence interval line (Figure 4-4) suggests that the clean-up standard (10 ppb) was reached
at time 8.9 months. Based strictly on this information and the fact that there is a downward slope in
concentration, the treatment system can be terminated, and post-termination monitoring can proceed.
70
-------
Clean-up Standard
iu —
1 8-
>^x 7 —
C 6-
O
^ 5-
O
-t 4-
C
- ^__ _ —-
Upper 95% Confidence
A A
Best-Fit Line A A
A A
1 1 ' 1 ' 1 ' 1 /
i5 175 225 275 325
/ Time (days) //
/ /
, /
. /
A - y/
AA /
/ /
AA Clean-up Standard /
A^^A*A If /
I 1 1 1 1 1 ' 1 * \ ' I
6 50 100 150 200 250 300 350
Time (days)
Figure 4-4. Best-fit regression line and 95% confidence internal for the concentration trend in Table 4-3.
71
-------
4.4 POST-TERMINATION MONITORING
After terminating P&T operation, a period of time must pass to ensure that any transient effects of
treatment on the ground-water system no longer exist (stage 4 in Figure 4-2). This period allows ground
water to reequilibrate hydraulically and chemically with the new flow field. Ground water can only be
judged to attain the clean-up standard if both present and future contaminant concentrations are
acceptable.
Changes in ground-water flow velocities and flow paths are induced when a P&T system is
initiated. These changes redistribute contaminant pathways and affect the rate at which ground water will
travel. Any change to the P&T system (e.g., increased pumping rates) will change ground-water flow
velocities and contaminant pathways. Following system termination, ground-water sampling may
continue, but only data collected after steady-state conditions have been reached (attainment sampling)
may be representative of long-term conditions.
Steady-state conditions occur when ground-water concentrations and elevations no longer are
influenced by the effects of the P&T system. When sampling to determine whether the ground-water
system is at steady-state, three decisions are possible (USEPA, 1992c):
Steady-state conditions exist and sampling for assessment attainment can begin;
The current contaminant concentrations indicate that the clean-up standard is unlikely to be
reached, and further treatment must be considered; or
More time and sampling must occur before it can be confidently assumed that the ground
water has reached steady state.
To determine if post-remediation steady-state conditions have occurred, it is useful to have a
knowledge of steady-state conditions prior to initiation of the remediation. Ground-water elevations may
not, however, return to preremediation conditions if the remediation includes permanent features such as
slurry walls.
The frequency of data collection will depend on the correlation among consecutively obtained
values (USEPA, 1992c). If serial correlation seems to be high, the time interval between data collection
efforts should be lengthened. With little or no information about seasonal patterns or serial correlations in
the data, at least six observations per year are recommended (USEPA, 1992c).
Underlying trends in ground-water chemistry and elevation data will suggest whether steady-state
conditions exist. All data should be plotted over time for visualization of potential trends. Statistical
methods for determining trends include parametric trend analyses and nonparametric trend analyses, and
were discussed previously. Other formal procedures for testing for trends also exist, including the
Seasonal Kendall Test, Sen's Test for Trend, and a Test for Global Trends. All three of these tests require
the assumption of independent observations. If this assumption is violated, these tests tend to indicate
that there is a trend when one does not actually exist (USEPA, 1992c).
72
-------
4.5 MONITORING FOR ATTAINMENT
After ground-water conditions have reached a new equilibrium or steady-state, long-term clean-
up attainment can be assessed (stages 5 and 6 in Figure 4-2). Long-term post-remediation monitoring is
critical in ensuring no future impact from contaminants gradually leaching out of the remediated matrix.
Post-operational monitoring may be required for a period of two to five years or more after termination,
depending on site conditions. As discussed in Chapter 1, contaminant concentrations can rebound
significantly after terminating a P&T operation (e.g., Robertson, 1992).
Two potential measures of long-term site cleanup consist of comparisons between clean-up
standards and mean concentration, or comparisons between clean-up standards and a selected percentile
of all samples. The procedures used to make these comparisons depend on whether a fixed number of
samples is to be analyzed (e.g., 20 samples over a two-year period), or samples are to be taken
sequentially at set intervals without specifying a total number of samples. Methods for determining if
clean-up standards have been maintained are similar to those methods for short-term comparisons.
Additional considerations include corrections for seasonal effects, determination of appropriate sample
sizes, determination of appropriate sampling frequency and, for sequential analysis, determining an
appropriate rate of data analysis. Guidance for statistical analysis of fixed and sequential sampling is
provided by USEPA (1992c).
4.6 CONCLUSIONS
General descriptions of the statistical techniques used to determine the time of remediation
termination have been presented in this chapter. For further guidance on the application of statistical
methods to assessing environmental data, refer to USEPA (1989 and 1992c), Helsel and Hirsh (1992), and
Gilbert (1987).
73
-------
-------
5. CHEM-DYNE SITE CASE STUDY
It is important to note that selected data from the Che&Dyne site in Hamilton, Ohio, are used
only to illustrate some of the monitoring methods described within this document. Much of the data was
obtained a number of years ago. Issues, such as the potential presence and sources of NAPLs, the effects
of near-site pumping, and changes in the remediation system since its initiation, are not considered for
this purpose. Extraction wells placed along and within the contaminant plume at the site are designed to
hydraulically contain and remove contaminated ground water for treatment. An overview of the P&T
design, monitoring, and results are provided in this chapter. Other P&T case studies are provided by
CH2M Hill (1992).
5.1 BACKGROUND
The Chem-Dyne site occupies approximately 20 acres along the Great Miami River within
Hamilton. Hazardous waste, accepted for solvent reprocessing from 1974 to 1980, resulted in
contamination of soil and ground water. The hydrogeologic environment at the site consists
predominantly of glaciofluvial sand and gravel, lacks extensive clay layers, and receives induced
infiltration from the Great Miami River. Generally, two hydrostratigraphic units exist: (1) a lower unit
consisting of medium gravel or sand and gravel, and (2) a shallow unit comprised of silts, clayey silts, and
silty and fine sands.
A Remedial Action Plan implemented in 1985 included: (1) excavation and disposal of
contaminated surficial soils; (2) installation of a low-permeability cap; and (3) development of a P&T
system to hydraulically contain and remove contaminated ground water within the 0.1 ppm total Priority
Pollutant volatile organic compounds (VOCs) isopleth (Figure 5-1). Priority Pollutant VOCs account for
approximately 96% of the contaminant mass detected mostly in the shallow depth wells open to the upper
five to ten feet of the aquifer (Papadopulos & Assoc., 1985). Sampling and analysis of intermediate depth
wells screened between a depth of approximately 55 to 65 ft also detected concentrations greater than 0.1
ppm total VOCs. The P&T design utilizes extraction wells within the zone of contamination and along
the plume boundary to hydraulically contain and remove the contaminated ground water. A portion of the
treated ground water is reinjected upgradient of the extraction wells to increase the pore volume flushing
rate. The system was originally designed to pump an average 2.6 pore volumes per year through the
contaminated zone. A ten year clean-up time was projected to reduce total dissolved Priority Pollutant
VOCs to below 0.1 ppm throughout the aquifer.
Limited operations of the ground-water P&T system began in February 1987. Data were
collected to assess the initial mechanical and operational performance of the system. Beginning January
1988, the Chem-Dyne site P&T system was fully operational. Five full years (1988 to 1992) of
operational and monitoring data have been collected and can be utilized to evaluate the effectiveness of
P&T remediation at Chem-Dyne.
75
-------
500
FEET
Shallow piezometer
Nested shallow, intermediate and deep piezometer
Plume boundary of total VOC concentrations
in excess of 0.1 ppm as defined in 1986
Figure 5-1. Boundary of 0.1 ppm total VOC plume and location of nested piezometers at the Chem-Dyne site
(from Papadopulos & Assoc., 1993).
76
-------
5.2 PERFORMANCE CRITERIA
The P&T system was designed based on the following performance goals and criteria which were
specified in a Consent Decree.
(1) After defining the 0.1 ppm total Priority Pollutant VOC plume limits, the outermost
downgradient extraction wells shall be placed at or beyond the contaminant plume boundary.
(2) The extraction/injection system shall establish and maintain an inward hydraulic gradient,
both vertically and horizontally, to ensure that the contaminants within the 0.1 ppm total
VOC plume boundary are contained for removal and treatment.
(3) The P&T system shall be operated for a minimum of ten years and shall be capable of
reducing the total Priority Pollutant VOC concentration within the plume boundary to 0.1
ppm.
(4) Ground-water quality shall not exceed water quality criteria for the protection of human
health (based on 10" risk or background, whichever is higher using best analytical
techniques) at compliance points outside of the zone of hydraulic control.
The responsible parties can terminate the P&T system after ten years of operation if the total
Priority Pollutant VOCs in all monitor and extraction wells within the 0.1 ppm plume have been reduced
below 0.1 ppm. If the total Priority Pollutant VOC concentrations are not maintained effectively constant
below 0.1 ppm after the cessation of pumping, additional corrective actions may be required. If the
concentration reduction goals are not met after 20 years of operation, then the regulatory and responsible
parties will determine whether further P&T operation or modification would produce significant
improvement.
5.3 PERFORMANCE MONITORING
Detailed P&T monitoring requirements were specified in the Consent Decree. The performance
monitoring at the Chem-Dyne site provides an example of the locations, frequency, and type of data to be
collected for measuring containment and restoration performance. The Chem-Dyne monitoring program
is designed to provide data to (Papadopulos & Assoc., 1985):
(1) Evaluate the performance of the extraction/injection system with respect to its design criteria
and to facilitate timely adjustments;
(2) Determine whether the system will be terminated after the initial ten year period, or at what
time thereafter;
(3) Assess whether performance goals have been met at compliance points and within the defined
plume boundary after termination; and
77
-------
(4) Develop a reliable predictive model that can be used to assess the effects of system
adjustments and the impacts of residual contamination, and of noncompliance, if any, on
potential receptors.
Water-level and waterquality data are collected to achieve these objectives.
5.3.1 Hydraulic Head Monitoring
Water-level data are measured regularly in approximately 130 wells. Locations for monitoring
include 25 extraction wells both within the plume and along the plume boundary, 31 inside-plume
monitor wells, 12 outside plume monitor wells, 18 compliance monitor wells, 6 water-supply wells, and 9
injection wells. Piezometer networks were installed at six locations along the containment area perimeter
to determine whether or not inward and upward hydraulic gradients are being maintained. Each
piezometer has a maximum screen length of five feet. Shallow, intermediate, and deep piezometers are
completed 10 to 15, 35 to 40, and 70 to 75 ft below the mean annual water table, respectively. As shown
in Figure 5-1, there are three shallow piezometers arranged in a triangle and three vertically nested
piezometers within each of the plume boundary piezometer networks.
5.3.2 Water-Quality Monitoring
Water-quality data are obtained from monitor wells, extraction wells, compliance wells, and five
nearby production wells. Concentrations detected during the remediation are compared to the "baseline"
conditions represented by the contaminant plume boundary and concentrations in compliance and
production wells in 1986. Baseline ground-water quality conditions were determined for offsite
production wells and three compliance points through three consecutive monthly sampling events and in
accordance with 40 CFR § 264.97. In addition, three consecutive, monthly ground-water quality
sampling events at six new (1985) and existing monitor wells were completed to redefine the contaminant
plume boundary. These sampling events resulted in a revised conceptualization of both the shallow and
intermediate depth VOC plumes. The greater lateral extent of the redefined VOC plumes required
modification of the remedial design. This example demonstrates the importance of ongoing
characterization during the remedial design stages and changes that can occur in the contaminant
distribution between initial characterization and remediation implementation.
Because treated ground water is injected into the aquifer and discharged to surface water (Ford
Hydraulic Canal), effluent water quality sampling is performed. VOC loading to the Ford Hydraulic
Canal is determined to fulfill NPDES permit requirements. Similarly, influent water quality analysis is
performed to determine the chemical loading to the treatment plant and the mass of contaminants
removed from the aquifer. Flow rates and water quality are used to determine these loadings. For the
extraction and injection wells, flow rates are measured at individual wells. At the ground-water treatment
facility, flow rates and ground-water quality are determined for the influent and effluent to the system.
78
-------
5.3.3 Monitoring Schedule
The monitoring schedule for hydrodynamic and chemical data collection provides an example of
performance monitoring for the containment objective and the restoration objective. Contingencies for
modification of the sampling events are also included to facilitate changing concentrations during the
operational and post-termination periods. The following monitoring schedule has been implemented at
the Chem-Dyne site.
For the first year of operation, water-levels were measured with water-level probes semi-monthly
at wells and piezometers and recorded hourly by pressure transducers/data loggers at the six piezometer
clusters at the boundary of the plume. Since 1989, water-levels have been measured by hand monthly and
water-levels from the central shallow piezometer from each of the six piezometer clusters have been
recorded hourly by dataloggers. Semi-monthly water levels are measured for three months (within a 250-
ft radius of the affected point) if any significant modification to the extraction/injection system or if
unstable water levels in the monitoring network have occurred. Daily extraction and injection rates are
measured at individual wells with flow meters. A remote recording system also provides a continuous
registry at the treatment facility of the volumes of ground water extracted/injected from individual wells.
Water flow rates are continuously measured for the treatment facility influent and treated effluent
discharged to the Ford Hydraulic Canal.
During the P&T remediation, water quality sampling is performed semi-annually for Priority
Pollutant VOCs and annually for all other Priority Pollutants at compliance point monitor wells as well as
annually for VOCs at monitor wells within the initial plume boundary. This sampling interval will
continue for five years after system termination. To facilitate determination of system termination after
ten years of operation, all monitor wells and extraction wells at and within the plume boundary will be
sampled quarterly for Priority Pollutant VOCs for the last three years of the ten year period. Upon
termination of the system, ground-water quality sampling will continue at these wells for five years:
quarterly for the first two years and semi-annually for the next three years. For wells beyond the defined
0.1 ppm total Priority Pollutant VOC isopleth, VOC sampling will be performed annually during system
operation and for five years after system termination.
The Chem-Dyne ground-water quality sampling plan also incorporates contingencies for the
monitoring program. For example, if the concentrations of VOCs at compliance point monitor wells
exceed compliance standards during operation or post-termination monitoring, sampling frequency will
be increased to quarterly for a minimum of six months. Also, if concentrations of total VOCs exceed 0.1
ppm at monitor wells outside the plume boundary during operation or post-termination monitoring, the
sampling frequency will be increased to quarterly for a minimum of six months. Further, if this
exceedance occurs during two consecutive sampling events, the significance of the occurrence will be
determined.
Water quality analysis of the effluent of the ground-water treatment facility is currently
performed monthly for Priority Pollutant volatile organics (EPA Methods 601 and 602), quarterly for
Priority Pollutant organics (EPA Methods 624,625, and 608). and semi-annually for Priority Pollutant
heavy metals.
79
-------
5.4 DATA EVALUATION AND EFFECTIVENESS
Several methods are utilized to demonstrate the effectiveness of the P&T system in attaining
containment and progressing toward restoration. To verify containment, water-level data are evaluated to
assess whether hydraulic control of the plume is maintained laterally and vertically. To evaluate
restoration, water quality of influent, effluent, monitor wells, and extraction wells is monitored.
5.4.1 Containment
Hydrographs of the monitored piezometer clusters (located at the original plume boundary) are
prepared using the hourly data from the shallow piezometer and the time-weighted averages of monitor
well water levels. The monthly and semi-monthly water-level measurements are also plotted on each
hydrograph. Average ground-water levels of the monthly data are utilized at the Chem-Dyne site to
reduce the effects of short-term disturbances and represent the conditions that are commensurate with the
rates of ground-water migration and indicative of significant patterns of flow (Papadopulos & Assoc.,
1993). The averages are time-weighted to reflect the relative duration of water-level conditions
associated with measurements made at different days of the months.
The vertical capture of the plume is assessed by evaluating the relative hydraulic head values at
the different screened depths within the piezometer clusters. Figure 5-2 illustrates an example of a
piezometer nest hydrograph for 1992. Vertical containment is inferred by the upward net vertical
hydraulic gradient between the deep and intermediate and between the intermediate and shallow horizons
of the aquifer.
Lateral containment is verified by preparing potentiometric surface maps of the shallow,
intermediate, and deep horizons'of the aquifer using average ground-water levels from the monthly water-
level data. Figures 5-3 and 5-4 present the average water-level conditions and the direction of ground-
water flow for the shallow and intermediate zones during 1992, respectively. Figure 5-3 demonstrates
that the extraction system has created a cone of depression in the shallow horizon at the leading edge of
the plume boundary. The heavy dashed line illustrates the approximate limit of the shallow horizon
capture zone indicating that water from within the plume boundary is captured by the extraction system.
Figure 5-4 demonstrates that the intermediate interval plume is also contained by the extraction wells.
The potentiometric surface contours during low and high ground-water conditions in the shallow and
intermediate intervals demonstrate that containment is maintained under low and high ground-water
conditions.
5.4.2 Restoration
The rate of VOC removal, the total mass of contaminants removed from the aquifer, and
contaminant concentrations in extraction and monitor wells are measured to evaluate restoration progress.
Table 5-1 illustrates the rate of VOC removal for the previous five years. Yearly mass removal rates have
decreased from 7500 to 1435 Ibs/year as the total volume of water treated has increased since the
initiation of the P&T system. However, the mass removal rates have not stabilized. Stabilization of
80
-------
568
562
Jan Feb Mar Apr May
June July
1992
Aug
Sep
Nov
Dec
Water-level measurement in deep piezometer
Water-level measurement in intermediate piezometer
Water-level measurement in shallow piezometer
Daily average water level from hourly data recorded
in shallow piezometer
Time-averaged water level in deep piezometer
Time-averaged water level in intermediate piezometer
Time-averaged water level in shallow piezometer
Figure 5-2. Nested piezometer hydrograph for 1992 at the Chem-Dyne site (from Papadopulos & Assoc.,
1993).
-------
N
feet
EXPLANATION
O Shallow Monitoring Wall
x Shadow Injection Wall
G Shadow Extraction Wan
A Shallow Ptazomatar
-^^— Direction of Ground-water Flow
-564- Lin* of Equal Watar-Tabla Bavation in Fact Abova MSL
A Ptazomatar dustar
— _ Unit of Capture Zona
•—— Pluma Boundary as dafinadtnl 986
9 Surfaca-Watar Gaga
O Watar-TaMa Oapracsion
. Watar-Tabla Mound
Figure 5-3. Average water table and direction of ground-water flow in the shallow interval in 1992 at the
Chem-Dyne site (from Papadopulos & Assoc., 1993).
82
-------
N
EXPLANATION
ATM wMh NgrMr than 100 \ujfL of VOC», Owwnbw, 1992
0 250 500
Scale feet
->"~567 Contour On Watar Surfaea
^ .. Intvrmtdtatt Phim* Boundary (1966)
^«^ O«prMfion
Figure 5-4. Average potentiometric surface and direction of ground-water flow in the intermediate interval
in 1992 at the Chem-Dyne site (from Papadolpulos & Assoc., 1993).
83
-------
TABLE 5-1. ANNUAL MASS OF VOCs AND VOLUME OF GROUND WATER
EXTRACTED FROM THE CHEM-DYNE SITE (PAPADOPULOS & ASSOC.,
1988 AND 1993).
Year
1987
1988
1989
1990
1991
1992
TOTAL
Average
Flow
Rate1 (GPM)
6252
--
791
726
7%
833
~~
Total
Volume
Treated
(million
gallons)
240
270
362
355
381
423
2,031
Average
Concentration
of PPVOCS
in Treatment
Effluent (ug/L)
2936 to 8205
254
6.5
92.7
131
39
--
Average
Concentration
Of PPVOCs
in Treatment
Influent (ug/L)
11580 to 06081s
3565 to 14426
2000
1630
1660
1294
414
—
Moss of
VOCs
Remove&
(pounds)
7500
4630
4970
4685
3794
1435
27,014
1 Average extraction rate based on operating hours and volume pumped to treatment plant
2 Average between June and December 1987
3 Net mass removed after injection
4 After modifying the air stripping system, the average effluent concentration decreased to 10 ug/L
5 Range during March and April
6 Range during September through December
contaminant mass removal rates, contaminant mass-in-place, and ground-water concentrations might
indicate that a P&T system is approaching a point of diminishing returns. This stabilization is not
suggested by the performance monitoring data at the Chem-Dyne site.
Performance of the P&T remediation is also demonstrated by the total mass of contaminant
removed from the saturated subsurface. A determination of the total mass removed is obtained from
extracted ground-water quality and extraction rates. At the Chem-Dyne site, influent water quality
combined with flow rates are used to provide time-weighted calculations of VOCs delivered to the
treatment plant (Figure 5-5). Because treated ground water (containing VOCs) is returned to the aquifer
by injection, the mass of VOCs delivered to the treatment plant does not represent the mass of
contaminants removed from the aquifer. Water quality and flow rates of the effluent and of discharges to
the Ford Hydraulic Canal are used to determine the net mass removal of VOCs from the aquifer. For
example, during 1992, the mass of VOCs in plant influent, effluent, discharges to the canal, and injectate
was calculated to be 1,470 ± 55, 140 ± 4, 105 ± 4, and 35 Ibs, respectively (Papadopulos & Assoc., 1993).
Therefore, the net mass removal for 1992 is approximately 1,435 ± 55 Ibs. These calculations are
84
-------
UJ
1987
1988
1989
1990
1991
1992
YEAR
INFLUENT
EFFLUENT
NOTES: 1. Concentrations for October, November and December 1992 are estimates
2. Concentrations include 1,2 - cis - DCE, not a priority pollutant.
Figure 5-5. Influent and effluent VOC concentrations (mg/L) at the Chem-Dyne treatment plant from 1987
to 1992 (from Papadopulos & Assoc., 1993).
-------
performed monthly to monitor the cumulative mass of Priority Pollutant VOCs removed since pumping
commenced in 1987 (Figure 5-6). Approximately 27,000 Ibs of VOCs have been removed from the
aquifer since the system became operational.
To evaluate this removal performance, the mass removed is compared to the original mass-in-
place. The estimated mass of VOCs dissolved in the ground water prior to the system operation within
the 0.1 ppm plume boundary was 4,500 Ibs (Papadopulos, 1993). The mass of VOCs sorbed on the
aquifer materials was estimated to be 36,000 Ibs, therefore, the total mass of VOC contaminants in the
aquifer (assuming no NAPL) prior to P&T remediation was 40,500 Ibs (Papadopulos, 1993). The P&T
system has removed 67% of the estimated original mass-in-place in less than half of its planned
operational period. However, the dissolved contaminant mass remaining in the aquifer, discussed below,
must be evaluated to confirm this apparent progress.
Performance of the restoration is also demonstrated at the Chem-Dyne site by comparing the
mass of dissolved contaminants remaining in the aquifer through time. The dissolved mass-in-place
provides an average of the distribution of contaminants, therefore, tracking the decrease in the dissolved
mass provides a basis for evaluating the effectiveness of the restoration. The estimates of dissolved mass
are based on contoured ground-water quality data, thickness of the contaminated zone, and porosity.
Figures 5-7 and 5-8 present the December 1992 concentrations of Priority Pollutant VOCs in the shallow
and intermediate wells in comparison to the 1986 0.1 ppm VOC plume boundary. Figures 5-9 and 5-10
present the October/November 1987 concentrations of Priority Pollutant VOCs in the shallow and
intermediate wells. The reduction of contaminants in the aquifer and significant reduction of
contaminants in individual wells evidences the effectiveness of the P&T system. Based on the
concentrations of total VOCs detected in extraction and monitor wells within the 0.1 ppm plume
boundary, the mass of contaminants dissolved in ground water prior to the commencement of the P&T
system was 4,500 pounds (Papadopulos, 1993). Sampling results indicate that the mass of dissolved
contaminants was reduced to approximately 235 pounds in 1992 (Papadopulos & Assoc., 1993). This
reduction of mass of dissolved contaminants appears promising, however, a comparison of the reduction
in dissolved mass of contaminants in the aquifer (4,265 pounds) with the amount of contaminants
removed (27,000 pounds) reflects the presence of NAPL and/or sorbed contamination. Physical and
chemical conditions discussed in Chapter 1 may result in future extraction with limited dissolved mass-in-
place reduction.
5.4.3 Termination
Data evaluation will also be performed to determine whether the Chem-Dyne P&T system can be
terminated after the 1 O-year operation period or at any time thereafter. This determination will be based
on the attainment of the performance goals stated above. The determination of whether VOC
concentrations within the plume have become effectively constant will be made for each extraction and
monitor well within the plume boundary according to the following procedures (Papadopulos & Assoc.,
1985).
86
-------
30.000
oo
-j
1987
1988
1989
1990
1991
1992
YEAR
NOTES 1. September, October, November and December 1992 cumulative weights are
based on estimated concentrations; potential range too small to show at this scale.
2. Mass of removed VOCs includes, 1,2-cis-DCE, not a priority pollutant.
Figure 5-6. Cumulative mass of VOCs removed from the aquifer at the Chem-Dyne site from 1987 to 1992
(from Papadopulos & Assoc., 1993).
-------
Scale
feet
EXPLANATION
Q SMawExtaMtonW*
O Shatow Mentoring W««
x ShaJtow Injactton Wai
-—— Plum* Boundary a* Dafinad in 1986
496 Total VOC* in iigfl-
.100. Lin* at Equal CuncaiUialiuii.. ttg/L
NO NotOalactad
NOTES:
I.Tha first of 1ha two numbarsrapoftad for
comptanctwcNs indieatM rmutts based on
analysts by method 8260 and tha sacond
by mathods 601/602
2.M«thyt«i»chtorid«in«noun1sl«s8t»wi5
MO/t not raportad if only contaminant praaant
Figure 5-7. Concentrations of VOCs in the shallow interval in December 1992 at the Chem-Dyne site (from
Papadopulos & Assoc., 1993).
88
-------
N
EXPLANATION
• Intermediate Monitoring W«N
B. ln»rmedtata Extraction Wai
54 c Total VOCs in itg/L
—— Plume Boundary a« defined in 1986
'100. un« of Eoual Concentration., iia/L
NO NotOMKtcd
NotSwnptad
NOTES:
1. Th« first of tn« two numtMrs rvpoitad for
compHanc* «wH* indkcatM rawlts bas*d on
anarysn by m*thod 8260 and tn* second
by method* 601/602
2. Metnylww chkxio* in •mounts IMS than $
HO/I, not reported H arty contaminant present
Figure 5-8. Concentration of VOCs in the intermediate interval in December 1992 at the Chem-Dyne site
(from Papadopulos & Assoc., 1993).
89
-------
N
xx /$,&<& ''•
Figure 5-9. Concentrations of VOCs in ug/L in the shallow interval during October/November 1987 (from
Papadopulos & Assoc., 1988).
90
-------
N
Figure 5-10. Concentrations of VOCs in jig/L in the intermediate interval during October/November 1987
(from Papadopulos & Asrac, 1988).
91
-------
(1) Totals of Priority Pollutant VOCs for the 12 most recent sampling events will be plotted
versus time.
(2) If the curve indicated by the concentrations is linear, a straight line will be fitted to the data
using a least squares regression model. The slope of the fitted curve will be computed as the
estimated slope.
(3) If the curve suggested by the data is nonlinear, then an exponential curve using a least squares
regression model will be fitted to the data. The estimated slope will be the first derivative of
the curve at the midpoint between the last two sample points.
(4) The estimated slope will be defined as zero if (a) the slope is < 0 and > -0.02 ppm/year, and
(b) the rate of change of that slope = 0 or indicates a continuously decreasing concentration.
(5) If the mean concentration in a well is < 0.02 ppm and the above procedure results in a
positive slope, then the 95 percent confidence interval will be calculated for the slope of the
regression line; if a zero slope is within this confidence interval, then the estimated slope will
be deemed to be zero.
(6) The concentrations in a well will be declared to be effectively constant if the estimated slope
is defined as zero.
If the concentration of total VOCs has become effectively constant (as defined above) in each
monitor and extraction well within the defined plume, but at a higher concentration than 0.1 ppm
(performance goal No. 1) after ten years of operation or any time thereafter, the system will be terminated
if the following two conditions are met (Papadopulos & Assoc., 1985):
(1) Substantial compliance with the performance goal of 0.1 ppm VOCs has been achieved
(considering factors which may include but are not limited to variations in permeability
which result in the persistence of high concentrations in certain wells, and the averaging of
concentrations in wells); and
(2) Periodic evaluation of data during system operation indicates that no reasonable modification
or adjustment to the system will produce significant improvement within a total operational
period of 20 years.
If both performance goals are not met after the 20 years of operation, the evaluation as to whether further
operation and modification would be cost-effective will be made by the parties involved.
5.4.4 Post Termination Monitoring
The Chem-Dyne site post termination monitoring plan provides an example of the verification of
continued "success" of the P&T system after the operational period. Water quality analyses at onsite
92
-------
monitor wells and offsite compliance points will be conducted to confirm the completion of the
remediation. Monitoring will be performed for five years after termination as specified below.
The concentrations of total Priority Pollutant VOCs within and on the defined plume boundary
will be monitored for five years after P&T termination to verify that concentrations do not rebound. To
determine compliance with this criterion, water-quality data collected from monitor wells within the
defined plume at the termination of the system will be statistically analyzed as follows (Papadopulos &
Assoc., 1985):
(1) The mean value and standard deviation of total concentration from all wells within and
on the plume boundary at the time of P&T termination will be used as baseline conditions.
(2) The mean value and standard deviation of the total VOC concentration will be determined for
each sampling event after termination.
(3) Statistical tests will be performed to determine if the variance of each sampling event is
statistically equal to the variance of the baseline value and if the baseline and sampling event
data are normally distributed.
(4) If the variances are equal and the data are normally distributed, a t-Test will be performed to
determine whether the mean value of the sampling event is significantly different from the
baseline mean at a five percent level of significance.
(5) If the variances are not statistically equal and/or the data are not normally distributed, then an
appropriate statistical test will be used to determine whether the mean value of the sampling
event is significantly different from the baseline mean value at a five percent level of
significance.
(6) If the mean value from the sampling event is not significantly different from the baseline
mean value, the concentration of total VOCs has been maintained effectively at or below the
levels reached at the time of P&T termination.
(7) If a significant increase in the mean value is determined, a second round of sampling will be
conducted within 30 days of receipt of the laboratory results. If this second round of
sampling confirms the significant increase in the mean value, corrective action will be taken.
The concentrations of total Priority Pollutant VOCs at offsite compliance points will also be
monitored for five years after P&T termination to verify that concentrations at receptors are not above the
water quality criteria.
93
-------
-------
6. REFERENCES
Ahlfeld, D.P., and C.S. Sawyer, 1990. Well location in capture zone design using simulation and
optimization techniques, Ground Water, 28(4):507-512.
American Water Works Association, 1990. Water Quality and Treatment, McGraw-Hill, NY, 1194 pp.
API, 1991. Technological limits of groundwater remediation: A statistical evaluation method, American
Petroleum Institute Publication Number 4510, Washington, D.C.
API, 1992. User's manual for REGRESS: Statistical evaluation of asymptotic limits of groundwater
remediation, American Petroleum Institute Publication Number 4543, Washington, D.C.
Bahr, J., 1989. Analysis of nonequilibrium desorption of volatile organics during field test of aquifer
decontamination, Journal of Contaminant Hydrology, 4(3):205-222.
Bair, E.S., A.E. Springer, and G.S. Roadcap, 1991. Delineation of travel time-related capture areas of
wells using analytical flow models and particle-tracking analysis, Ground Water, 29(3):387-397.
Bair, E.S., and G.S. Roadcap, 1992. Comparison of flow models used to delineate capture zones of wells:
1. Leaky-confined fractured-carbonate aquifer, Ground Water, 30(2): 199-211.
Bear, J., 1979. Hydraulics of Groundwater, McGraw-Hill, Inc., New York, NY, 569 pp.
Blandford, T.N., and P.S. Huyakorn, 1989. WHPA: A modular semi-analytical model for delineation of
Wellhead Protection Areas, USEPA, Office of Ground-Water Protection, Washington, DC,
Bonn, B.A., and S.A. Rounds, 1990. DREAM - Analytical Ground Water Flow Programs, Lewis
Publishers, Boca Raton, FL, 109 pp.
Brogan, S.D., 1991. Aquifer remediation in the presence of rate-limited sorption, Masters Thesis,
Department of Applied Earth Sciences, Stanford University, Palo Alto, CA, 238 pp.
Bouwer, E., J. Mercer, M. Kavanaugh, and F. DiGiano, 1988. Coping with groundwater contamination,
Journal Water Pollution Control Federation, 60(8): 1415-1427.
Carosone-Link, P.H.R. Horsey, J.C. Loftis, L.D. Rainey, 1993. Ground water quality statistical analysis:
Implementing the new RCRA regulations, Presented at the NGWSE National Outdoor Conference,
Las Vegas, NV.
CH2M Hill, 1992. Evaluation of ground-water extraction remedies, NTIS PB92-963346 and PB92-
963347, USEPA Office of Emergency and Remedial Response, Washington, D.C.
Clay, D.R., 1992. Considerations in ground-water remediation at Superfund sites and RCRA facilities ~
Update, USEPA memorandum, 13 pp.
Cohen, R.M., and J.W. Mercer, 1993. DNAPL Site Evaluation, C.K. Smoley, Boca Raton, FL.
95
-------
Conover, W.J., 1980. Practical Nonparametric Statistics, 2nd ed., Wiley, New York, NY.
Dalton, M.G., B.E. Huntsman, and K. Bradbury, 1991. Acquisition and interpretation of water-level data,
in Practical Handbook of Ground-Water Monitoring, D.M.,Nielsen, ed., Lewis Publishers, Boca
Raton, FL, pp. 367-394.
Davis, S.N., D.J. Campbell, H.W. Bentley, and T.J. Flynn, 1985. Ground Water Tracers, National Water
Well Association, Dublin, OH, 200 pp.
Devitt, D.A., R.B. Evans, W.A. Jury, T.H. Starks, B. Eklund, and A. Gholson, 1987. Soil gas sensing for
detection and mapping of volatile organics, EPA/600/8-87/036, USEPA Environmental Monitoring
Laboratory, Las Vegas, NV, 281 pp.
Doty, C.B., and C.C. Travis, 1991. The effectiveness of groundwater pumping as a restoration
technology, Waste Management Research and Education Institute, Document ORNL/TM-11866.
Driscoll, F.G., 1986. Ground Water and Wells, Johnson Division, UOP, St. Paul, MN, 1089 pp.
ESE, 1992. Baseline monitoring report and continuous monitoring report summary for the Stockbridge
remediation system and Mantua segmented trench system, Environmental Science & Engineering,
Hemdon, VA.
Evans, E.K., G.M. Duffield, J.W. Massmann, R.A. Freeze, and D.E. Stephenson, 1993. Demonstration of
risk-based decision analysis in remedial alternative selection and design, Proceedings 1993 Ground
Water Modeling Conference, IGMWC, Colorado School of Mines, Golden, CO.
Feenstra, S., D.M. Mackay, and J.A. Cherry, 1991. A method for assessing residual NAPL based on
organic chemical concentrations in soil samples, Ground Water Monitoring Review, 11(2): 128-136.
Fitts, C.R., 1989. Simple analytic functions for modeling three-dimensional flow in layered aquifers,
Water Resources Research, 25(5):943-948.
Fitts, C.R., 1993. Well discharge optimization using analytical elements, Proceedings 1993 Ground
Water Modeling Conference, IGMWC, Colorado School of Mines, Golden, CO.
Franz, T., and N. Guiguer, 1990. FLOWPATH, Two-dimensional horizontal aquifer simulation model,
Waterloo Hydrogeologic Software, Waterloo, Ontario.
Freeze, R.A., and J.A. Cherry, 1979. Groundwater, Prentice Hall, Inc., Edgewood Cliffs, NJ, 604 pp.
Freeze, R.A., J.W. Massmann, L. Smith, T. Sperling, and B. James, 1990. Hydrogeological decision
analysis: 1. A framework, Ground Water, 28(5):738-766.
GeoTrans, Inc., 1992. MODMAN: An optimization module for MODFLOW, Version 2.1,
Documentation and user's guide, Sterling, VA.
Gilbert, R.O., 1987. Statistical Methods for Environmental Pollution Monitoring, Van Nostrand
Reinhold Company, NY, 320 pp.
96
-------
Gillham, R.W., E.A. Sudicky, J.A. Cherry, and E.O. Frind, 1984. An advective-diffusion concept for
solute transport in heterogeneous unconsolidated geologic deposits, Water Resources Research,
20(3):369-378.
Gorelick, S.M., R.A. Freeze, D. Donohue, and J.F. Keely, 1993. Groundwater Contamination: Optimal
Capture and Containment, Lewis Publishers, Boca Raton, FL, 385 pp.
Guthrie, M., 1986. Use of a geoflowmeter for the determination of ground water flow direction, Ground
Water Monitoring Review, 6(2):81-86..
Hagemeyer, R.T., P.P. Andersen, R.M. Greenwald, and J.L. Clausen, 1983. Evaluation of alternative
plume containment designs at the Paducah Gaseous Diffusion Plant using MODMAN, a well
pumpage optimization module for MODFLOW, Proceedings 1993 Ground Water Modeling
Conference, IGMWC, Colorado School of Mines, Golden, Colorado.
Haley, J.L., B. Hanson, C. Enfield, and J. Glass, 1991. Evaluating the effectiveness of ground water
extraction systems, Ground Water Monitoring Review, 11(1): 19- 124.
Hall, C.W., 1988. Practical limits to pump and treat technology for aquifer remediation, Proceedings of
the Groundwater Quality Protection Pre-Conference Workshop, Water Pollution Control Federation,
61st Annual Conference, Dallas, TX, pp. 7-12.
Hamilton, D.E., and T.A. Jones, eds., 1992. Computer Modeling of Geologic Surfaces, The American
Association of Petroleum Geologists, Tulsa, OK, 296 pp.
Harman, J., D.M. Mackay, and J.A. Cherry, 1993. Goals and effectiveness of pump and treat
remediation, Volume 1, Final Draft, Robert S. Kerr Environmental Research Laboratory, Ada, OK,
27pp.
Haug, A., R.H. Petrini, G.E. Grisak, and K. Klahsen, 1990. Synthetic semivariograms ~ A novel
approach to assessing positions and spacing of ground-water monitoring wells, Proceedings of
Conference on Minimizing Risk to the Hydrologic Environment, American Institute of Hydrology,
Minneapolis, MN, pp. 224-233.
Helsel, D.R., and R.M. Hirsch, 1992. Statistical Methods in Water Resources, Elsevier, New York.
Hirsch, R.M., J.R. Slack, and R.A. Smith, 1982. Techniques for trend analysis for monthly quality data,
Water Resources Research, 18(1): 107-12 1.
Hoffman, F., 1993. Ground-water remediation using 'smart pump and treat', Ground Water, 31(1):98-
106.
Istok, J., 1984. Groundwater Modeling by the Finite Element Method, Water Resources Monograph,
American Geophysical Union, Washington, D.C., 495 pp.
Javendel, I., C. Doughty, and C.F. Tsang, 1984. Groundwater Transport: Handbook of Mathematical
Models, American Geophysical Union Water Resources Monograph No. 10, Washington, D.C.,
228 pp.
97
-------
Javendel, I, and C.F. Tsang, 1986. Capture-zone type curves: A tool for aquifer cleanup, Ground Water,
24:616625.
Johnson, R.L., and J.F. Pankow, 1992. Dissolution of dense immiscible solvents in groundwater:
2. Dissolution from pools of solvent and implications for the remediation of solvent-contaminated
sites, Environmental Science & Technology, 26(5): 896-901.
Jones, T.A., D.E. Hamilton, and C.R. Johnson, 1986. Contouring Geologic Surfaces with the Computer,
Van Nostrand Reinhold, NY, 314 pp.
Kearl, P.M., and C.M. Case, 1992. Direct field measurement of groundwater velocities, Interdisciplinary
Approaches in Hydrology and Hydrogeology, M.E. Jones and A. Laenen, eds., American Institute of
Hydrology, Minneapolis, MN, pp. 91-02.
Keely, J.F., 1989. Performance evaluation of pump-and-treat remediations, USEPA/540/4-89-005,
Robert S. Kerr Environmental Research Laboratory, Ada, OK.
Kerfoot, W.B., 1984. Darcian flow characteristics upgradient of a kettle pond determined by direct
ground water flow measurement, Ground Water Monitoring Review, 4(4):188-192.
Kruseman, G.P., and N.A. deRidder, 1990. Analysis and Evaluation of Pumping Test Data, International
Institute of Land Reclamation and Improvement, Bulletin 11, 2nd Ed., Wageningen, The Netherlands.
Larson, S.P., C.B. Andrews, M.D. Rowland, and D.T. Feinstein, 1987. A three-dimensional modeling
analysis of ground water pumping schemes for containment of shallow ground water contamination,
Proceedings of Solving Ground Water Problems -with Models, National Water Well Association,
Dublin, OH, pp. 5 17-531.
Letkoff, L.J., and S.M. Gorelick, 1987. AQMAN: Linear and quadratic programming matrix genera&or
using two-dimensinal groundwater flow simulation for aquifer management modelling, USGS Water-
Resources Investigations Report 87-4061.
Loaiciga, H.A., R.J. Charbeneau, L.G. Everett, G.E. Fogg, B.F. Hobbs, and S. Rouhani, 1992. Review of
ground-water quality monitoring network design, Journal of Hydraulic Engineering, ASCE
Hydraulics Division, 118(l):ll-37.
Lucius, J.E., GR. Olhoeft, P.L. Hill, and S.K. Duke, 1990. Properties and hazards of 108 selected
substances, USGS Open-File Report 90-408, 559 pp.
Mackay, D.M., and J.A. Cherry, 1989. Groundwater contamination: Pump-and-treat remediation,
Environmental Science & Technology, 23(6):620-636.
Mackay, D.M., W.Y. Shiu, A. Maijanen, and S. Feenstra, 1991. Dissolution of non-aqueous phase liquids
in groundwater, Journal of Contaminant Hydrology, 8(l):23-42.
Massmann, J., and R.A. Freeze, 1987. Groundwater contamination from waste management sites: The
interaction between risk-based engineering design and regulatory policy, 1. Methodology,
2. Results, Water Resources Research, 23(2):351-380.
98
-------
Massmann, I, R.A. Freeze, L. Smith, T. Sperling, and B. James, 1991. Hydrogeological decision
analysis: 2. Applications to ground-water contamination, Ground Water, 29(4):536-548.
McDonald, M.G., and A.W. Harbaugh, 1988. A modular threedimensional finite-difference groundwater
flow model, USGS Techniques of Water-Resources Investigations, Book 6, Chapter Al, USGS,
Reston, VA, 586 pp.
Melville, J.G., F.J. Molz, and 0. Guven, 1985. Laboratory investigation and analysis of a ground-water
flowmeter, Ground Water, 23(4):486-495.
Mercer, J.W., D.C. Skipp, and D. Giffin, 1990. Basics of pump-and-treat groundwater remediation,
USEPA-600/8-90/003, Robert S. Kerr Environmental Research Laboratory, Ada, OK, 31 pp.
Meyer, P.D., 1992. The optimal design of groundwater quality monitoring networks under conditions of
uncertainty, Ph.D. Thesis, Civil Engineering Department, University of Illinois, Urbana-Champaign,
122 pp.
Meyer, P.O., and E.D. Brill, Jr., 1988. A method for locating wells in a groundwater monitoring network
under conditions of uncertainty, Water Resources Research, 24(8): 1277-1282.
Molz, F.J., and S.C. Young, 1993. Development and application of borehole flowmeters for
environmental assessment, The Log Analyst, January-February 1993: 13-23
National Research Council, 1990. Ground Water Models: Scientific and Regulatory Applications,
National Academy Press, Washington D.C., 303 pp.
Newell, C.J., and R.R. Ross, 1992. Estimating potential for occurrence of DNAPL at Superfund sites,
USEPA Quick Reference Fact Sheet, #9355.4-07 FS, Robert S. Kerr Environmental Research
Laboratory, Ada, OK.
Newsom, J.M., and J.L. Wilson, 1988. Flow of ground water to a well near a stream ~ Effect of ambient
ground-water flow direction, Ground Water, 26(6): 703-711.
Nyer, E.K., 1992. Groundwater Treatment Technology, 2nd Ed., Van Nostrand Reinhold Co., NY,
188 pp.
Palmer, C.D., and W. Fish, 1992. Chemical enhancements to pump-and-treat remediation, EPA/540/S-
92/001, USEPA Ground Water Issue Paper, Robert S. Kerr Environmental Research Laboratory, Ada,
OK, 20 pp.
Papadopulos & Associates, Inc., 1985. Remedial action plan, Chem-Dyne Site, Hamilton, OH, May.
Papadopulos & Associates, Inc. and Conestoga-Rovers & Associates, Ltd., 1988. Chem-Dyne Site Trust
Fund, 1987 Annual Report, Chem-Dyne Site, Hamilton, OH, January.
Papadopulos & Associates, Inc. and Conestoga-Rovers & Associates, Ltd., 1993. Chem-Dyne Site Trust
Fund, 1992 Annual Report, Chem-Dyne Site, Hamilton, OH, April.
Papadopulos, S.S., 1965. Nonsteady flow to a well in an infinite anisotropic aquifer, Proceedings of
Symposium International Association of Scientific Hydrology, Dubrovnik, pp. 21-31.
99
-------
Papadopulos, S.S., 1993. Personal communication, August 9.
Pollock, D.W., 1989. Documentation of computer programs to compute and display pathlines using
results from the USGS Modular Three-Dimensional Finite-Difference Groundwater Flow Model,
USGS Open File Report 89-381, 188 pp.
Powell, R.M., 1990. Total organic carbon determination in natural and contaminated aquifer materials,
relevance and measurement, Proceedings of the Fourth National Outdoor Action Conference on
Aquifer Restoration, Ground Water Monitoring and Geophysical Methods, NWWA, Dublin, OH,
pp. 1245-1258.
Robertson, C.G., 1992. Groundwater extraction system case history, IBM Corporation, Dayton, New
Jersey, Presentation to Committee on Groundwater Extraction Systems, National Research Council,
March 24, Washington, D.C.
Rumbaugh, J.O., 1991. Quick Flow: Analytical ground-water flow model, Version 1.0, Geraghty &
Miller, Plainview, NY.
Rumbaugh, J.O., J.A. Caldwell, and S.T. Shaw, 1987. A geophysical ground water monitoring program
for a sanitary landfill: Implementation and preliminary results, Proceedings of the First National
Outdoor Action Conference on Aquifer Restoration, Ground Water Monitoring, and Geophysical
Methods, National Water Well Association, Dublin, OH, pp. 623-641.
Saroff, S.T., H. He, and G. Powell, 1992. Hydrogeological implications of saprolite aquifer remediation
at Fairfax, Virginia, oil spill, Proceedings ofHMCRI's 13th Annual National Conference and
Exhibition, Hazardous Materials Control Research Institute, Silver Spring, MD, pp. 407-412.
Shafer, J.M., 1987a. Reverse pathline calculation of time-related capture zones in nonuniform flow,
Ground Water, 25(3):283-289.
Shafer, 1987b. GWPATH: Interactive groundwater flow path analysis, Illinois State Water Survey
Bulletin 69.
Springer, A.E., and E.S. Bair, 1992. Comparison of methods used to delineate capture zones of wells:
2. Stratified-drift buried-valley aquifer, Ground Water, 30(6):908-917.
Steimle, R., 1992. An inventory of research, laboratory studies and field demonstrations of in situ
treatment of contaminated ground water, Preprint submitted to In Situ Treatment of Contaminated
soil and Water, Cincinnati, OH.
Stephanatos, B.N., K. Walter, A. Funk, and A. MacGregor, 1991. Pitfalls associated with the assumption
of a constant partition coefficient in modeling sorbing solute transport through the subsurface,
Proceedings of the International Symposium on Ground Water, American Society of Civil Engineers,
ASCE, Nashville, TN, pp. 13-20.
Strack, O.D.L., 1989. Groundwater Mechanics, Prentice Hall, Englewood Cliffs, NJ, 732 pp.
Thornhill, J.T., 1989. Accuracy of depth to water measurements, EPA/540/4-89/002, Ground-Water
Issue Paper, Robert S. Kerr Environmental Research Laboratory, Ada, OK.
100
-------
Trescott, PC., G.F. Finder, and S.P. Larson, 1976. Finite-Difference Model for Aquifer Simulation in
Two Dimensions with Results of Numerical Experiments, USGS Techniques of Water Resource
Investigations, Book 7, Chapter Cl.
USEPA, 1986. RCRA ground-water monitoring technical enforcement guidance document, OSWER
Directive 9950.1, EPA Office of Emergency and Environmental Response, Washington, D.C.
USEPA, 1987. A compendium of technologies used in the treatment of hazardous wastes, EPA/625/1-86/
060, Washington, D.C.
USEPA, 1988a. Statistical methods for evaluating ground-water monitoring from hazardous waste
facilities: Final rule, Federal Register, 53(196):39728-39731, October 11.
USEPA, 1988b. Guidance on remedial actions for contaminated ground water at Superfund sites, EPA/
540/G-88/003, OSWER Directive 9283.1-2, Office of Solid Waste and Emergency Response,
Washington, D.C.
USEPA, 1989. Statistical analysis of ground-water monitoring data at RCRA facilities, Interim Final
Guidance, Office of Solid Waste, Washington, D.C.
USEPA, 1990. Subsurface Remediation Guidance, EPA/540/2-90/01 b, Office of Emergency and
Remedial Response, Washington, D.C.
USEPA, 1992a. General methods for remedial operations performance evaluations, EPA/600/R-92/002,
Robert S. Kerr Environmental Research Laboratory, Ada, OK, 37 pp.
USEPA, 1992b. Statistical analysis of ground-water monitoring data at RCRA facilities, Addendum to
Interim Final Guidance, Office of Solid Waste, Washington, D.C.
USEPA, 1992c. Methods for evaluating the attainment of cleanup standards, Volume 2: Ground water,
EPA/230-R-92-014, Environmental Statistics and Information Division, Office of Policy, Planning,
and Evaluation, Washington, DC.
USEPA, 1993a. Compilation of Ground-Water Models, EPA/600/R-93/118, May 1993.
USEPA, 1993b. Guidance for Evaluating the Technical Impracticability of Ground-Water Restoration.
Directive 9234.2-25, Office of Solid Waste and Emergy Response, Washington DC.
USGS, 1977. National handbook of recommended methods for water-data acquisition, U.S. Geological
Survey, Reston, VA.
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.
Ward, D.S., A.L. Harrover, A.H. Vincent, and B.H. Lester, 1993. Data input guide for SWIFT/486,
GeoTrans, Inc., Sterling, VA.
101
-------
Watson, D.F., 1982. ACORN: Automatic contouring of raw data, Computers and GeoSciences,
Wilson, C.R., C.M. Einberger, R.L. Jackson, and R.B. Mercer, 1992. Design of ground-water monitoring
networks using the monitoring efficiency model (MEMO), Ground Water, 30(6): 965-970.
Wilson, J.R., 1985. Double-cell hydraulic containment of pollutant plumes, Proceedings of the Fourth
National Symposium and Exposition on Aquifer Restoration and Ground Water Monitoring, NWWA,
Dublin, OH, pp. 65-70.
Zheng, C., 1989. PATH3D ~ A ground-water path and travel-time simulator, Version 2.0, User's
manual, S.S. Papadopulos and Associates, Bethesda, Maryland.
Zheng, C., G.D. Bennett, and C.B. Andrews, 1991. Analysis of ground-water remedial alternatives at a
Superfund site, Ground Water, 29(6):838-848.
Zheng, C., G.D. Bennett, and C.B. Andrews, 1992. Reply to discussion of analysis of ground-water
remedial alternatives at a Superfund site, Ground Water, 30(3):440-442.
102 * U'S- GOV£RMMENT PRIHTlMn OFFICE: 1995 - 650-006/00226
------- |