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

-------