-------
Each Superfund site has unique geologic, hydrologic, biologic
and chemical conditions that may influence the type and mag-
nitude of potential sample errors.  This paper provides an
overview of sample error; types of error potentially important at
each site must be evaluated on an individual basis. Further-
more, while this paper will remain static, the conduct of site
investigations will be in a constant state of flux as new technol-
ogy is developed and as the understanding of contaminant
transport and fate and the sampling process is improved. As a
result, sources of  sampling error  described herein may be
resolved through the application of new technology and meth-
ods while new sources of error are likely to be identified.

MONITORING WELL DESIGN

The design of ground-water monitoring  installations must be
consistent with geologic, hydrologic, and hydrochemical condi-
tions to obtain representative ground-water samples. Important
aspects of  monitoring well design include length of well intake
interval, design of the filter pack and screen, design and instal-
lation of borehole seals, and well location.

Intake Length

The length and location of well intakes have important effects on
the degree with which samples represent ground-water condi-
tions.  Long well  intakes (long screens) are open to a large
vertical interval and therefore are more likely to provide samples
that are a composite of the ground water adjacent to the entire
intake.  Conversely, short intakes (short screens) may be open
to a single strata or zone of contamination and are more likely to
provide samples that represent specific depth intervals. Wells
that are screened over more than one depth interval (multi-
screened wells), regardless of their screen lengths, may impact
ground-water conditions and samples in much the same way as
long-screened wells.

Long-screened wells have been suggested as being more cost
effective in detection monitoring than several short-screened
wells because they sample greater vertical sections of aquifers
(Giddings,  1986).  However, pumping-induced vertical flow in
wells with  long screens can  impact ground-water flow and
contaminant concentrations near the  well (Kaleris, 1989).  In
addition, when ground-water contamination is vertically strati-
fied, composite samples collected from a long-screened well
represent some sort of average of concentrations adjacent to
the screen, and provide little information about the concentra-
tions in individual strata.  In particular, in  cases where contami-
nants may be of low concentration and restricted to thin zones,
long-screened wells may lead to dilution  of the contaminants to
the point where  they may be difficult to detect  (Cohen and
Rabold, 1987).  Likewise, long-screen wells intersecting con-
taminants of differing densities may allow density-driven mixing
within the well bore and subsequent dilution of contaminant
concentrations (Robin and Gillham, 1987). The use of inflatable
packers to isolate specific zones within a long screen may not be
an effective solution because ground water may flow vertically
through  the filter pack  from other zones in response to  the
reduced hydraulic head in the packed-off  zone during sampling.

Vertical  head gradients in  aquifers near long-screened wells
may lead to error in two ways: (1) if contaminants are moving
through  a zone with low hydraulic head,  cleaner water moving
from zones of higher head may dilute the contaminants, leading
to detection of artificially low concentrations, and, (2) if higher
concentrations of contaminants are moving through a zone of
high hydraulic head, cross-contamination between water-bear-
ing  zones may occur via the well bore (Mcllvride and Rector,
1988). These workers describe a case history in which two
aquifer zones were identified at a site, with only the top zone
contaminated with VOCs. Wells screened only in the contami-
nated zone resulted in detection of VOCs in the few hundred |ig/
L range while samples collected from long-screened wells open
to both intervals showed no VOC contamination. A numerical
flow model of a long-screened  well developed by Reilly et al.
(1989) demonstrated that very low head gradients can lead to
substantial cross-flow within long-screened wells.  At sites
where delineation of vertical hydraulic and concentration gradi-
ents is important, errors can be reduced by utilizing a system of
nested short-screened wells that can more accurately charac-
terize the contaminant distribution.

Multilevel sampling devices provide an alternative monitoring
technique in situations where vertical head gradients are impor-
tant or where contamination is  vertically stratified.  These
devices can be installed in such a way that individual zones can
be  sampled  separately without vertical movement of ground
water or contaminants  between zones.  Using a multilevel
device, Smith ef al. (1987) detected a zone containing nitrate
concentrations over 10 mg/Lthat had been previously undetec-
ted by observation wells with two-foot screens. The samples
from the  multilevel sampler also detected  large vertical gradi-
ents in electrical conductivity (EC) and chloride that were not
detected with the monitoring wells.

Residential and  municipal water-supply wells that are  often
used during early phases of Rl programs are generally con-
structed with long screens, therefore concentrations of contami-
nants in samples collected from these wells may not represent
ambient ground-water concentrations. When defining human
receptors this may not be an issue because the overall quality
of ground-water extracted from water-supply wells  may not
reflect the quality of water in individual strata.  In these cases,
dilution may reduce concentrations of contaminants to within
health-based standards.  However, gross errors may be intro-
duced into the analysis if these concentrations are  used for
detailed  delineation of the geometry and  concentrations of
contaminant plumes or detection of contaminants  at very low
concentrations.

To  mitigate hazards, waste management options at Superfund
sites may include remediation of contaminated ground water by
pumping and treatment. Long-screen wells are often the most
effective for extraction of ground water because they are hy-
draulically more  efficient than wells with short screens.  How-
ever, because accurate ground-water contaminant concentra-
tions cannot be determined from these wells it may be neces-
sary to install separate wells for monitoring the  progress of
ground-water extraction and treatment.

Filter Pack and  Well Intake

Suspended solids  that originate from  drilling activities or are
mobilized from the formation during development,  purging, or
sampling may disrupt hydrochemical equilibrium during sample
collection and shipment.  A properly designed combination of

-------
filter pack and welt intake provides an efficient hydraulic connec-
tion to a water-bearing zone and minimizes the suspended
solids content of sampled water. However, to be most effective,
filter pack and well intake design must be based on the sedi-
ments encountered in each borehole. Inadequate well perfor-
mance resulting from application of a generic well design may
lead to incomplete well development and high suspended solids
content in samples.  Descriptions of the methods of filter pack
and intake design can be found in Driscoll (1986) and Aller et at.
(1989).

Artificial filter packs should be composed of a chemically-inert
material so as to reduce the potential for chemical alteration of
ground  water near  the well.  Clean silica (quartz) sand is
generally recommended and  widely used  because it is
nonreactive under most ground-water conditions. Other types
of materials may induce chemical changes. For example, filter
pack  materials containing calcium carbonate, either as a pri-
mary component or as a contam inant, may raise the pH of water
that it contacts and lead to precipitation of dissolved constituents
(Aller etal., 1989).

The use of a tremie pipe to install filter pack materials minimizes
the potential for introducing sample error to this phase of well
construction.  Dropping filter pack materials directly into  an
uncased borehole may lead to cross-contamination by mobiliz-
ing sediments or ground water between depth intervals. Fur-
thermore, installation of filter pack materials by methods which
introduce water to the borehole may modify hydrochemistry to
an unknown extent or add contaminants to the sampling zone.
Water-based methods  may  also lead to cross-contamination
within the borehole.

Borehole Seals

Borehole seals, generally composed of expandable bentonite or
cement grout, are well-known as potential sources of sampling
error. The expandable bentonite clay used in many  seals has
high ion exchange capacity which may alter major ion composi-
tion of water(Gillham et al., 1983) or concentrations of contami-
nants that form complexes with these ions (Herzog et al., 1991).
The effects of these reactions are seldom revealed by measure-
ment of field parameters and  normally-conducted analyses, but
in cases of extreme sodium bentonite contamination may  be
seen  as abnormally high sodium concentrations.

Cement grout can  also significantly influence ground water
chemistry, particularly if thegroutdoesn'tset properly. Contami-
nation by grout seals, which generally results from its calcium
carbonate  content and high alkalinity, may be  identified  by
elevated calcium concentrations, pH (generally over  10  pH
units), EC, and alkalinity (Barcelona and Helfrich, 1986).  These
workers  found that  cement contamination of several wells
persisted for over 18 months after well completion and was not
reduced by ten redevelopment efforts. Barcelona  et al. (1988a)
indicate that solution chemistry and the distribution of chemical
species can be impacted by cement contamination although
these impacts have  not been quantified to date. In low-perme-
ability sediments, the impacts of grout materials may be much
greater due to insuff Sclent flushing of the installation by moving
ground water.

Contamination from borehole seals can be minimized by sepa-
rating the seals from sampling zones by fine-grained transition
sand, estimating the volume of seal material required before
installation  to  more easily  detect bridging  problems during
emplacement, and by allowing sufficienttimeforthe seals to set.
In addition,  cement grout can be isolated from sampling zones
by installation of a bentonite seal. Error can also be reduced by
installing boreholes seals with a tremie pipe.  Dropping seal
materials directly into an uncased borehole may lead to cross-
contamination  by mobilizing sediments or ground water be-
tween depth intervals, or may contaminate sampling zones if the
seal materials are dropped past the sampling zone depth.
Furthermore, installation of  seal  materials by methods which
introduce water to the borehole may modify hydrochemistry to
an unknown extent or introduce contaminants to the sampling
zone. Water-based methods may also lead to cross-contamina-
tion within the borehole.

Well Location

The location of monitoring wells with respect to ground-water
contaminant plumes is important to the accurate depiction of
contaminant movement and concentration distribution, espe-
cially in areas where concentration gradients are large.  A
discussion of optimum well placement is beyond the scope of
this document, but aspects of this topic can be found in the works
of Keith et al. (1983), Meyer and Brill  (1988), Scheibe and
Lettenmaier (1989), Spruill and Candela (1990), and Andricevic
and Foufoula-Georgiou (1991).  These investigators discuss
various aspects  of monitoring well  network  design and how
monitoring well coverage of the area under investigation relates
to accurate quantification of spatial variation  in hydrochemical
parameters. Generally  implied within network design is the
reduction in error associated with delineating spatial variation.
Sampling from wells whose locations were determined without
adequate consideration  of network  design and geologic, hy-
draulic, and hydrochemical conditions may lead to significant
errors  in data  interpretation  and conclusions. For example,
resolution of concentration distribution may be reduced in areas
where wells spacing intervals are too large for the scale of the
investigation.

To summarize the topic of monitoring well design, collection of
accurate ground-water quality data in three dimensions is
strongly dependent on the design of the ground-water monitor-
ing system, including both individual wells and well networks.
Significant errors can be introduced into sampling data, and the
resultant conclusions, if well intakes and filter packs are not
designed for ambient conditions, or are placed at inappropriate
depths or over excessive vertical intervals, or if borehole seals
are improperly installed.  Furthermore, the design of monitoring
well networks may introduce error by inadequately representing
spatial variation through inadequate coverage of the  site. Al-
though the  magnitude of these errors is heavily dependent on
the geologic, hydraulic,  and hydrochemical conditions present
at a particular site, order of magnitude effects are easily within
the realm of possibility.

DRILLING METHODS

Long-term  or  permanent disturbance of hydrogeologic and
hydrochemical conditions may result from the drilling method
used for monitoring well  installation, possibly leading to signifi-
cant error during subsequent ground-water sampling. Drilling
methods may  disturb sediments, allow vertical movement of
ground water and/or contaminants, introduce materials foreign

-------
to the subsurface, and clog void spaces. The extent to which
conditions are altered depends on the drilling method utilized
and the nature of the geologic materials (Gillham. et al., 1983).
In addition, the properties of the contaminants at the  site will
influence their sensitivity to the impacts of drilling.

Monitoring wells are commonly  constructed by auger, rotary,
drill-through casing, and cable-tool  methods.  Auger drilling
methods utilize hollow- or solid-stem auger flights and are
generally restricted to use in unconsolidated materials. Rotary
techniques are classified based on the composition of the drilling
fluid (water, air, and various additives), the mode of circulation
(direct or reverse), and the type of bit (e.g. roller cone,  drag, or
button) and are adaptable to most geologic conditions. The drill-
through  casing method  utilizes rotary or  percussion drilling
techniques but uses a casing  driver to advance temporary
casing in conjunction with the advancing borehole. In cable-tool
drilling, the borehole is advanced by alternately raising and
lowering a heavy string of drilling tools suspended from a cable.
Temporary casing can also be advanced as drilling progresses.

 Some drilling methods may alterthe hydrogeologic environment
by smearing  cuttings  (particularly fine sediments) vertically
along the borehole wall.  This action may form a mudcake that
can reduce the hydraulic efficiency of the borehole wall and
 modify ground-water flow into the completed well (Mcllvride and
 Weiss, 1988). Smearing may alsotransport sediments between
zones and alter the  vertical  distribution of  contaminants
 adsorbed onto these sediments.  In addition, methods  that mix
sediments horizontally near the well bore may affect the trans-
 port of contaminants near the completed  well (Morin, et  al.,
 1988).

 Vertical  movement  of ground water may occur during drilling,
 primarily in situations where the borehole remains uncased
during drilling operations. Ground water can be transported
vertically by circulating drilling fluid or by hydraulic head differ-
ences between zones.   In  situations where contaminated
ground water is vertically stratified, vertical ground-water move-
ment may cause cross-contamination within the well-bore and
adjacent formation (Gillham et al., 1983). Movement of ground
water and  contaminants between  zones may also disrupt
hydrochemical equilibrium near the well.

Drilling activities can alter hydrochemistry as a result of contact
with introduced materials foreign to the subsurface environ-
ment.  For example, lubricants or hydraulic fluids may enter the
borehole directly by falling from the drilling  rig or may enter
indirectly via drilling fluids. In the latter case, contaminants may
originate in mud pumps, air compressors, or down-hole drilling
equipment. Soils or other material from the drilling site may also
enter the open borehole or may adhere to drilling equipment as
it is prepared for use. However, the material most commonly
introduced to boreholes is drilling fluid, which is used to remove
cuttings,  stabilize the borehole wall, and provide cooling, lubri-
cation, and cleaning of the bit and drill pipe  (Driscoll, 1986).
Drilling fluids commonly are composed of water or air alone or
in combination with  clay (usually bentonite) and/or polymeric
additives.

Water from water-based drilling fluids that migrates away from
the borehole and mixes with ambient ground water may alter
hydrochemical  conditions (Aller et al., 1989).  For example,
introduction of a different water type may add contaminants or
disrupt hydrochemical equilibrium and cause precipitation of
dissolved constituents. During sampling, some of these precipi-
tates may be redissolved by ground water flowing toward the
well causing non-representative samples.

The bentonite additives used in many drilling fluids have a high
capacity for ion  exchange  and may alter hydrochemistry of
ground-water samples if not completely removed from the
borehole and surrounding formation (Gillham et al., 1983). Ion
exchange reactions that alter major ion composition may also
affect the concentrations of contaminants that form complexes
with  these ions  (Herzog et al., 1991).  Organic polymeric
additives can introduce organic carbon into ground water and
provide a  substrate for microbial activity  leading to  errors in
water quality observations for long periods.  Barcelona (1984)
reported that total organic carbon (TOC) levels in wells drilled
with fluids containing organic additives remained over three
times higherthan background levels for two years. Inthatstudy,
TOC levels could not be reduced to less than two times back-
ground levels, even after substantial pumping.

The presence of drilling fluids in the formation surrounding well
installations, even after well development, was shown by Brobst
and Buszka (1986). That study, which used chemical oxygen
demand (COD) as an indicator of the presence of drilling fluid,
tested three additives of water-based drilling fluids: guar fluid,
guar fluid with a breakdown additive, and bentonite. Brobst and
Buszka (1986) reported that, using standard well purging and
sampling methods, COD levels were elevated for 50 days in a
well drilled with the guar-and-additive fluid, 140 days in a well
drilled with bentonite, and 320 days in a well drilled with the guar
fluid alone. More intense well purging reduced the COD levels,
but not to background values.

Contaminants present in drilling fluid may also mix with ground
water and bias sampling results.  Mud pumps used with water-
based drilling fluids can add trace quantities of lubricants to the
fluid and deposit them in the wellbore and surrounding forma-
tion. Air compressors used to develop and  maintain pressure of
air-based drilling fluids may have similar impacts. Filtration units
in air-based systems are designed to prevent this occurrence,
however, if feasible, the air stream should be sampled during
drilling to determine the effectiveness of the filter.  Filtration is
generally not possible for water-based systems so if ground
water samples are to be collected for compounds related to
these lubricants it may be necessary to sample the drilling fluid
before it enters the borehole.

An outline of potential impacts of drilling methods on ground-
water sample quality is shown in Table 1, which was compiled
from the work of Scalf et al. (1981), Gillham et al. (1983), Keely
and Boateng (1987), Aller et al. (1989), and Herzog et al. (1991).

WELL DEVELOPMENT

Ground-water monitoring wells are developed to restore the
sampling zone to conditions present prior  to drilling so that
sampled ground water can flow unimpeded and unaltered into
the well. Materials associated with the drilling process, including
borehole wall mudcake, smeared and compacted sediments,
and drilling  and  other fluids, all  must  be removed from the
sampling zone to the extent possible. This can be accomplished

-------
   TABLE 1. POTENTIAL IMPACTS OF DRILLING METHODS ON
             GROUND-WATER SAMPLE QUALITY
Method
Potential Impacts
Auger
Rotary
Drive-Through-Casing
Cable Tool
Drilling fluids generally not used but water or
other materials added if heaving sands are
encountered may alter hydrochemistry

Smearing of fine sediments along borehole wall

Vertical movement of ground water and/or
contaminants within borehole

Lateral mixing of sediments near well bore

Drilling fluids are required and may cause cross-
contamination, vertical smearing of sediments,
alteration of hydrochemistry, and introduction of
contaminants

Smearing of fine sediments along borehole wall

Vertical movement of ground water and/or
contaminants within borehole

Drilling fluids required but advancing casing
reduces potential for drilling fluid loss, cross-
contamination, and vertical smearing of
sediments, ground water, and contaminants.

Advancing casing reduces potential for cross-
contamination, and vertical smearing of
sediments, ground water, and contaminants.
in monitoring wells by several methods including surging with a
surge block mechanism, surging and pumping with compressed
air, pumping and overpumping with a pump, jetting with air or
water, backwashing with water, and bailing. All of these meth-
ods have the potential (to varying  degrees) to influence the
quality of ground water samples; the extent depends on the
nature of their action and the condition of the sampling zone after
drilling.

Development should be considered complete when representa-
tive samples can be collected and can continue to be collected
indefinitely. Unfortunately, under most ground-water sampling
circumstances determining when samples are representative of
in situ conditions is not possible, so some related criteria are
often chosen.   Ideally, these criteria should include (1) the
production of clear water during development, and (2) the
removal of a volume of water at least equal to the amount lost to
the formation during drilling and well installation (Kraemer et al.,
1991). In addition, certain conditions may require that develop-
ment be continued after the well has been allowed to recover
from the first round of development efforts.  This condition may
exist if the first round of samples exhibit turbidity.

Incomplete or ineffective well development may allow drilling
and other introduced fluids to remain in the sampling zone or
may not remove all mudcake or smeared sediments from the
borehole wall. The presence of these materials may introduce
error by disrupting hydrochemical equilibrium or by introducing
contaminants to the well or sampling zone.  In addition, these
materials can reduce the hydraulic conductivity of the filter pack
and formation and modify ground waterf low nearthe well before
and during sampling.

Development methods that utilize air pressure can entrap air in
the filter pack and formation, disrupt hydrochemical equilibrium
through oxidation, or introduce contaminants from the air stream
to the formation and filter pack. These effects may be reduced
if precautions are taken to eliminate air contact with the well
intake. The addition of water during development may modify
hydrochemistry to an unknown extent or may introduce contami-
nants to the sampling zone, even if all the water  is removed
during development. In light of these potential problems, jetting
methods that inject air or water directly above the well intake are
not recommended (Keely and Boateng, 1987). Likewise, other
methods that  introduce air or water to the well (surging and
pumping with compressed air, and backwashing, for example)
also may not be suitable for monitoring well development (Aller
etal.,1989).

Development of wells at very high rates may displace filter pack
and formation materials and reduce the effectiveness of thef ilter
pack, particularly if the method involves  excessive surging
(Keely and Boateng, 1987). On the other hand, development at
low rates (as is generally attained with sampling pumps) may not
provide enough agitation to  meet development  objectives
(Kraemer etal., 1991). In many monitoring well situations, using
surge-block methods to loosen material and either pumping or
bailing to remove the material has been found to be an effective
development technique (Aller et al.,  1989).

In low-yield wells,  surging methods may result in excessive
mobilization of fine-grained materials.  For example, in  a study
conducted in fine-grained glacial tills, Paul et al. (1988) found
that auger-drilled wells developed  by surge-block methods
produced samples with up to 100 times greater turbidity than
samples from similar wells developed by bailer.  In addition, the
turbidity of samples from the surged wells did not significantly
decrease after a second round of sampling while samples from
the bailed wells showed a four-fold decrease (Paul et al., 1988).
Because these wells were drilled in low permeability sediments
without added fluids, the action of drawing down the water level
within the well by bailing may have  been sufficient to  provide
adequate development. On the other hand, bailing or pumping
techniques alone may not be effective in wells constructed by
drilling methods that introduce fluids  or cause significant distur-
bance of sediments because the development force is dissi-
pated by the filter pack.

The potential  impacts of  monitoring well development on
ground-water sample quality are outlined in Table 2 which is

-------
 TABLE 2.  POTENTIAL IMPACTS OF DEVELOPMENT METHODS ON
             GROUND-WATER SAMPLE QUALITY
Method
Potential Impacts
Surging w'rth surge block
 Surging and pumping
  with compressed air
 Pumping and over-
  pumping with pump
 Jetting with air or water
 Backwashing with water



 Bailing
Displacement of filter pack and formation
materials or damage to the well intake (primarily
a problem in poorly designed and constructed
wells when surging is conducted improperly)

Excessive mobilization of fine-grained materials
from low-permeability formations
 Entrapment of air in filter pack and formation

 Disruption of hydrochemical equilibrium

 Introduction of contaminants


 Low-volume pumps may be incapable of
 sufficient surging action (primarily in high-yield
 wells with little or no drawdown)

 Entrapment of air in filter pack and formation

 Disruption of hydrochemical equilibrium

 Introduction of contaminants

 Excessive mobilization of fine-grained materials
 from low-permeability formations

 Disruption of hydrochemical equilibrium

 Introduction of contaminants

 May be incapable of sufficient development
 action
 based on the work of Keely and Boateng (1987), Paul et al.
 (1988), Aller et al. (1989), and Kraemer et al. (1991).

 MATERIALS

 Transfer of ground water from the subsurface sampling zone to
 a sample container at ground surface often involves contact of
 the sample with a  variety  of materials comprising the  well,
 sampling device, tubing, and container. Some of these materi-
 als have  the potential  to  bias chemical concentrations in
 samples as a result of sorption, leaching, and chemical attack,
 and biological activity (Barcelona et al., 1983). As a result, the
 materials selected for ground-water sampling must be appropri-
 ate for the hydrochemical conditions at the site and the constitu-
 ents being sampled. Other factors that may influence the choice
 of materials, including costs  verses benefits, availability,
 strength, and ease  of handling, can be found in Aller et al.
 (1989).
Materials commonly used in the ground-water sampling train
can be divided  into  five general categories (modified from
Nielsen and Schalla, 1991):

 1. fluoropolymers,  which  include polytetrafluoroethylene
    (PTFE), tetrafluoroethylene (TFE), and f luorinated ethylene
    propylene (FEP);

 2. thermoplastics, which include  polyvinyl  chloride (PVC),
    acrylonitrile butadiene styrene (ABS), polypropylene (PP),
    and polyethylene (PE);

 3. metals, which include stainless steel (SS), carbon steel,
    and galvanized steel;

 4. silicones; and

 5. fiberglass-reinforced, which include fiberglass-reinforced
    epoxy (FRE) and fiberglass-reinforced plastic (FRP);

This document will focus on the most commonly used materials
including the rigid materials PTFE, PVC, and metals (particularly
SS) and the flexible materials PE, PP, PTFE, PVC, and silicone.

Chemical and Biological Impacts

Sorption, which  includes the processes of adsorption and ab-
sorption,  may remove chemical constituents from samples
thereby reducing the concentrations of these constituents from
levels present in the ambient ground water.  If compounds
present in the ground water are removed entirely, false negative
analytical results will be produced.  Additionally, desorption of
compounds previously sorbed can occur if water moving past
the material contains  lower concentrations of the sorbant than
exists in the material.   In this case,  contaminants may be
detected  in samples  that do  not exist in the ground water,
causing false positive analytical results.  Sorption/desorption
processes may  be particularly important in  situations where
contaminant concentrations are at trace levels and change with
time or where samples contact potentially sorbing materials for
long periods (for example, during water level recovery in low-
yield wells or in inadequately purged wells).

Leaching of chemical constituents from some types of materials
may occur under the conditions present at  many hazardous
waste sites. Constituents of the materials' matrix, orcompounds
added during fabrication, storage,  and shipment,  may have
solubilities in water high enough to be leached under  natural
ground-water conditions  (Gillham et al., 1983). Ground water
contaminated by high concentrations of organic solvents may
cause significant degradation  of the matrix of some polymeric
materials,  resulting in  leaching  of  various compounds
(Barcelona et al., 1983).  As a result, false positive analytical
results can be produced if the source of target constituents in
ground-water samples is leaching from casing materials rather
than the ambient ground water. In addition, corrosion of metal
casing may introduce  dissolved   metals  to ground-water
samples and reduce the  integrity of the well.

Under certain ground-water conditions, well-casing materials
may impact biologic activity, and vice versa, in the vicinity of the
well (Barcelona et al., 1988b) and lead to errors that are difficult
to predict.  For example, the presence  of dissolved  iron in
ground-water may favor the growth of iron bacteria near metallic
wells and degrade the casing and  screen (Driscoll,1986).  In
addition, permeation of contaminants or gases through  materi-

-------
als may be a potential source of sample bias with flexible tubing
(Barker et al., 1987; Holm, 1988) but is unlikely with  rigid
materials, as demonstrated by Berens (1985) for organic com-
pounds and rigid PVC pipe over time periods less than 100
years.

Rigid Materials

Rigid materials that contactground-water samples are generally
used in well casings and screens, sampler components, and
filtration equipment.

PTFE

PTFE has been widely considered the best choice for monitoring
well materials because of its apparent resistance to chemical
attack  and  low sorption  and leaching potential.   However,
several recent laboratory studies have shown that rigid PTFE
materials actually demonstrate a significant ability to sorb hydro-
carbons from solution.  Sykes et al. (1986) found that PTFE
materials sorbed several hydrocarbons from a solution contain-
ing concentrations of approximately 100 u.g/L, but did not report
quantities. Parker et al. (1990) found that rigid PTFE materials
sorbed significant quantities of all tested chlorinated organics
and a nitroaromatic; higher, in fact, than PVC materials. These
workers found that losses of some of these com poundsfrom test
solutions (initial concentrations  of each compound were ap-
proximately  2 mg/L) exceeded 10% within eight hours.  Like-
wise, rigid PTFE materials showed significant sorption of aro-
matic hydrocarbons in 24 hours  of exposure for benzene, and
six hours for several  other hydrocarbons (Gillham and
O'Hannesin, 1990).  After eight weeks of PTFE exposure to
benzene, 75% losses from the test solution were observed.

In contrast,  PTFE  materials tend to show lower potential for
interaction with trace metals than PVC or SS (Barcelona and
Helfrich, 1986). For example, lead was the only metal of four
tested (arsenic, chromium, cadmium, and lead) in a laboratory
study to be actively sorbed onto  PTFE materials although only
5% of the lead concentration in the test solution was removed
after 24 hours of exposure (Parker et al., 1990).

PVC

Early studies of PVC materials found  substantial potential for
sample error from sorption and leaching effects.  Many of the
conclusions about sorption were based on flexible PVC, which
has a much  higher sorption potential than rigid PVC. Leaching
of high VOC concentrations was found to be a particular problem
from PVC solvents and cements used for casing joints and bailer
construction.  Boettner  et  al. (1981) found  cyclohexanone,
methylethylketone, and tetrahydrofuran leached into water at
concentrations ranging from 10u.g/L to  10 mg/Lfor more than 14
days after the glue was applied to PVC  pipe. In addition to these
compounds, methylisobutylketone  was detected in  ground-
water samples several months afterthe installation of cemented
PVC casing  (Sosebee et al. (1982). The results of these studies
indicate that alternative methods of joining PVC casing, such as
threaded joints, should be utilized to reduce sample error.

Laboratory investigations show that threaded PVC well materi-
als sorb hydrocarbon compounds, but  often at lower rates than
other polymers,  including PTFE.  Miller (1982)  found little
absorption of six VOCs over a six-week period, with the excep-
tion of tetrachlorethylene which  showed  a 50% decline  in
concentration in solution.  These sorption results were signifi-
cantly lower than those from PE and PP casing  materials.
Subsequent leaching from PVC was found to be at insignificant
levels for all six VOCs. Gillham and O'Hannesin (1990) found
that significant sorption onto rigid PVC from a solution contain-
ing six hydrocarbons did not occur until 12 hours after exposure.
The PVC results were in contrast to three other rigid polymers
(PTFE, FEP, and polyvinylfloride) that showed significant up-
take of at least one of the six compounds within three hours of
exposure. After eight weeks of PVC exposure to benzene, 25%
losses were observed from the original solution concentration of
approximately 1.2 mg/L. Similar results were reported by Parker
etal. (1990) who found that PVC sorption of 10% of initial organic
compound concentrations didn't occur until over 72 hours  of
exposure, while PTFE sorption of 10% of three of the 10 tested
organics occurred within eight hours of exposure. Two dichlo-
robenzene isomers showed the highest sorption rates on PVC:
significant losses were observed within eight hours. Sykes etal.
(1986) found no significant differences between PVC, PTFE,
and SS materials in their tendency  to  sorb six organics  at
concentrations of approximately 100 (ig/L each.

The results of  these research studies indicate that  rigid PVC
materials have relatively low potential for sorption and leaching
of organic compounds relative to other polymers when exposed
to dissolved concentrations generally found at hazardous waste
sites.  However, Berens (1985) demonstrated that  PVC may
soften and allow permeation of organic compounds if exposed
to nearly undiluted solvents or swelling agents for PVC. For this
reason,  PVC  well casing should be avoided under these
conditions.

PVC materials may also react with some trace metals. Miller
(1982) concluded that  in a six-week exposure to test solution,
PVC materials did not affect chromium concentrations but that
lead concentrations declined over 75%. A subsequent experi-
ment  showed that over 75% of the initial lead concentrations
were  desorbed from the PVC material.  Parker et  al. (1990)
found that rigid PVC showed no measurable sorption or leaching
of arsenic or chromium but that cadmium was leached and lead
sorbed.  For example, sorption of lead resulted in a 10% decline
in lead concentration in their test  solution in four hours, while
subsequent desorption resulted  in a 10%  increase in lead
concentration after four hours.

Stainless Steel

SS casing materials are often used when conditions warrant a
strong, durable, corrosion-resistant material. Of the two types
available, Type 316 is somewhat less likely than type 304 to  be
affected  by pitting and corrosion caused by  organic acids,
sulfuric acid, and sulfur-containing species (Barcelona et al.,
1983). However, long exposure  to very corrosive  conditions
may result in chromium and nickel contamination (Barcelona et
al., 1983), or iron, manganese, and chromium contamination
(U.S.  EPA, 1987) of samples. A field study by Barcelona and
Helfrich (1986) found  that stagnant  water samples from SS
installations showed higher levels of ferrous iron  and  lower
levels of dissolved sulfide than nearby PTFE and PVC wells,
suggesting leaching from the SS and precipitation of sulfide by
the excess iron.  However, these workers demonstrated that

-------
proper well-purging techniques eliminated this stagnant water
from ground-water samples, providing representative ground-
water samples.

Laboratory experiments conducted by Parker et at. (1990)
examined the potential for sorption on type 304 and 316 SS
casing materials.  These workers conducted experiments with
aqueous solutions of arsenic, cadmium, chromium, and lead at
concentrations of 50 u,g/L and 100 u,g/L and found that after 10
hours, sorption on both type 304 and type 316 caused a 10%
decline in arsenic concentration in the test solution.  Cadmium
concentrations increased 10% in five hours due to leaching from
type 304, before returning  to initial  concentrations after 72
hours. Cadmium leaching from type  316 caused a maximum
30% increase  after 20  hours,  with concentrations still  20%
above initial values after 72 hours.  No measurable sorption of
chromium occurred for type 304, but 13% losses in 13  hours
were observed for type 316.  Sorption of lead on type 304
materials led to 20% losses after only four hours of exposure,
and approximately 10% for type 316.  Parker et  al. (1990)
concluded from this work that determinations of the concentra-
tions of cadmium, chromium, and lead may be impacted by long-
term contact with stainless steel materials. Unfortunately, these
workers did not address whether well purging would eliminate
these impacts and  provide representative ground-water
 samples.

 In a study with five halogenated hydrocarbons, Reynolds et al.
 (1990) found type 316 SS  caused losses of bromoform and
 hexachloroethane over a five-week period.  Losses of  these
 compounds from the test solution were insignificant until one
 week, after which concentrations dropped up to 70% from initial
 concentrations of 20 to 45 u.g/L. The losses were attributed to
 reactions involving the metal surfaces or metal ions released
 from the surfaces and not to sorption (Reynolds et al., 1990). A
 study by Parker et al. (1990) with ten organic compounds at
 concentrations of approximately 2 mg/L, found that type 304 and
 type 316 SS casing resulted in no detectable  sorption or leach-
 ing effects after six weeks.

 Other Metallic Materials

 Steel materials other than stainless steel may be more resistant
 to attack from organic solutions than polymers, but corrosion is
 a significant problem, particularly in high dissolved-solids, acidic
 environments (Barcelona et al.,  1985a). Ferrous materials may
 adsorb dissolved chemical constituents or leach ions or corro-
 sion products such as oxides of iron and manganese (Barcelona
 et al., 1988a).  In addition, galvanized steel may contribute zinc
 and cadmium species to ground-water samples.  The weath-
 ered  steel  surfaces,  as well as the  solid corrosion products
 themselves, increase the surface area for sorption processes
 and may therefore act as a source of bias for both organic and
 inorganic constituents (Barcelona et al., 1985a; Barcelonaet al.,
 1983). Reynolds et al. (1990) determined that galvanized steel
 showed a 99% reduction in concentrations of five  halogenated
 hydrocarbons in a five-week sampling period.  Aluminum casing
 caused concentration reductions of 90% for four of the com-
 pounds. Although many of these aspects of steel materials have
 not been quantified for typical ground-water environments, they
 may be a significant source of sample error.
Alternate Materials

Although not as widely tested or used, FRE may represent a rigid
well material with relatively low potential for sample bias. In a 72-
hour laboratory study, none of the 129 priority pollutants were
detected to be leached from a powdered sample of the material
(Cowgill, 1988).  A three-week dwell-time study of  casing
materials by the same investigator resulted in  detection of  no
base/neutral  or acid compounds.  Gillham and O'Hannesin
(1990) concluded that sorption of benzene and other aromatic
hydrocarbons onto FRE was slightly greaterthan onto rigid PVC
but less than onto PTFE.

Borosilicate glass, another little-used well material, revealed no
sorption effects after a 34-day exposure to five halogenated
hydrocarbons (Reynolds et al., 1990). Of the ten well materials
tested in that study, only the borosilicate glass showed  no
sorption characteristics.  The low potential for sample  error
indicated by that study suggests that further  investigation of
borosilicate glass may be warranted to determine its suitability
for ground-water sampling.

Flexible Materials

Semi-rigid and flexible materials are used fortransfer tubing and
other flexible components of the sampling/analysis train.  In
general, these materials contain plasticizers for flexibility that
give them a higher potential than rigid materials to sorb or leach
compounds.  Latex rubbertubing, flexible PVC, and low density
PE were  all found to sorb greater quantities  than more rigid
materials (Reynolds et al., 1990).

In a study of five tubing materials in solutions of four chlorinated
hydrocarbons, Barcelona et al. (1985b) found that most sorption
occurred in the first 20 minutes of exposure. With the exception
of tetrachloroethylene, the materials ranked in order of increas-
ing sorption PTFE, PP,  PE, PVC, and silicone. PE showed the
highest sorption of tetrachloroethylene.  Desorption from all
materials occurred rapidly with the same ranking: PTFE des-
orbed a maximum of 13% of the sorbed concentrations after one
hour while silicone desorbed 2%. From the results of this work,
Barcelona et al. (1985b) estimated sorptive losses of chlori-
nated hydrocarbons from  sampling tubing under typical flow
rates. As an example, using 15 m of  1/2-inch tubing,  initial
concentrations  of 400  u.g/L for the four halocarbons, and a
sample delivery rate of 100 mL/min, these workers predicted 21,
29,48,67, and 74% sorptive losses for PTFE, PP, PE, PVC, and
silicone tubing, respectively.

Sorption tests conducted  by  Barker et al. (1987) found that
flexible PTFE led to 17% sorptive losses of benzene and 58%
losses of p-xylene  after two weeks.  For PE, 49% losses of
benzene  and 91%  losses of  p-xylene were observed in two
weeks.  As found in other studies, initial rapid losses  were
followed by gradual concentration declines in all compounds.
Desorption of these compounds followed a  similar pattern,
approximately 40% of the initial benzene mass and 20% of the
initial p-xylene masses desorbed.  Laboratory tests conducted
by Gillham and O'Hannesin (1990) showed PVC and PE tubing
caused sorptive losses of over 10% within  five minutes of
exposure to six hydrocarbons in solution. After 24 hours, 90% '
losses for the PVC and 80% losses for the PE had occurred.

-------
These studies suggest that flexible PTFE tubing has  lower
potential for sorption and leaching than other materials, particu-
larly PVC and silicone. However, even PTFE tubing may have
significant impacts on concentrations of organic compounds in
ground-water samples, depending on duration of contact. It is
clear that the sorption and leaching affects of all materials used
as tubing or other flexible portions of the sampling/analysis train
should be considered when designing the sampling program.
Those materials that demonstrate high potential for sorption
and/or leaching should be avoided if those processes  could
impact concentrations of the compounds of interest to the
investigation.

A further  source of sample bias with respect to tubing  is
transmission of compounds or gases through the tubing  mate-
rials.  In a study of PE and PTFE, Barker et al. (1987) detected
2 u.g/L benzene and 15 u,g/L toluene passing through PE tubing
within three days and 15 u.g/L and 100 (ig/L, respectively, after
six days.  Subsequent flushing  of the tubing with three tubing
volumes of clean water reduced the concentrations of both
compounds detectable inside the tubing  but they were still
detectable after twenty volumes were flushed. Under the same
conditions, the compounds did not  pass through the  PTFE
tubing in detectable concentrations.  These workers  suggest
that this mechanism may lead to sample bias in other polymeric
materials,  although perhaps at rates somewhat less than those
exhibited by the flexible PE tubing, and could influence conclu-
sions about when well purging procedures or remediation  activi-
ties are complete.  Holm et al.  (1988) studied the diffusion of
gases through FEP tubing, and found that the amount of gas
transferred is proportional to the tubing length and inversely
proportional to the flow rate through the tube. Calculations by
the authors suggest that, given initially anoxic ground water,
oxygen diffusion through sampling tubing could lead to detec-
tion of DO and changes in iron speciation within tens of feet. The
results of these studies clearly indicate the potential errors that
transmission through flexible tubing might introduce when sam-
pling for both organic and inorganic compounds. This source of
error can be reduced by using appropriate tubing materials for
the sampling conditions and by minimizing tubing lengths.

Selection of Materials

It is clear from  laboratory studies  of casing  materials that
concentrations of trace metals  and hydrocarbons can be im-
pacted by  sorption and leaching from PTFE, PVC, and metallic
casing materials. However, laboratory studies do not attempt to
duplicate the complicated, interrelated physical, chemical, and
biologic conditions present in the field that may cause materials
to behave very differently in the hydrogeologic environment. It
is also important to keep in mind that most of these experiments
were conducted under static conditions and may not adequately
represent  field conditions where stagnant water is generally
replaced with fresh ground water during well purging.  In the
field, sorption of compounds onto casing materials between
sampling  events  may  not affect  subsequent ground-water
samples, as long as adequate purging andsampling procedures
are conducted. Desorption of previously  sorbed compounds
after long-term exposure may be of somewhat greater impor-
tance because continuous desorption may impact trace-level
concentrations, which might have  important implications  to
remedial investigations where concentrations are expected to
eventually reach  non-detectable levels.  But again, proper
selection and implementation of  materials and purging and
sampling methods will reduce the  impact of these processes.

Given the above discussion and current state of research, some
generalizations may be made about the applicability of casing
materials to various ground-water contamination scenarios,
assuming that reducing sample error is the primary criterion for
selection. When monitoring for low hydrocarbon concentrations
in  non-corrosive ground  water, SS and  PVC casing may be
appropriate choices. Because PTFE has been shown to intro-
duce error  into  hydrocarbon determinations, it may be most
applicable under conditions where SS and PVC are not.  As
examples, SS would not be appropriate in corrosive ground
water or where determination of trace metal concentrations is of
primary concern and PVC wells would be  inappropriate in
situations where solvents in moderate to high concentrations
could dissolve the PVC material. A summary of the properties
of rigid PVC, PTFE, and SS materials that may introduce sample
error is shown in Table 3.

Laboratory  studies indicate  that the potential for error from
flexible tubing is much greater than from rigid materials.  For this
reason, efforts should be made to use tubing with low potential
for sorption and leaching and to minimize tubing length and time
of  contact.  It appears that sample  error can be significantly
reduced by substituting  flexible PTFE for PVC and silicone
where possible.

MONITORING WELL  PURGING

Purging stagnant water from monitoring wells priorto sampling
is considered essential to collection of samples representative
of  ambient ground water.  Stagnant water may result from
biological, chemical and physical processes occurring between
sampling events.   These processes may include biological
activity, sorption/desorption reactions with materials of the well,
leaching from the materials of the well, degassing and volatiliza-
tion, atmospheric contamination, and foreign material entering
the well from ground surface.

An effective purging method must allow for flushing of the well
and sampling device of stagnant water without causing undesir-
able physical and chemical changes in the adjacent water-
bearing zone that may bias subsequent samples. Important
aspects of purging include purge volume, pumping rate, depth
of the purging device, and purging methods for low-yield wells.
Field experiments  have shown that  purging has important
impacts on sample chemistry, perhaps greater than other as-
pects of sampling protocol such as sampling device and mate-
rials (Barcelona and Helfrich, 1986).

Purge Volume

To ensure complete purging of a ground-water monitoring well,
there must be established criteria to determine when the water
in  the well is representative of ambient ground water. Three
criteria commonly advocated to determine appropriate purge
volume have been described by Gibs and Imbrigiotta (1990) as:
(1) a specific, predetermined number of well-bore volumes, (2)
stabilization of the values of field chemical indicator parameters
(such as temperature, pH, and EC), and (3) hydraulic equilib-
rium between water stored in the casing and water entering the
casing.

-------
    TABLE 3. PROPERTIES OF COMMONLY-USED WELL CASING
 MATERIALS THAT MAY IMPACT GROUND-WATER SAMPLE QUALITY
                    Properties
 Polytetrafluoroelhylene  Moderate potential for sorption of hydrocarbons.
  (PTFE)
                    Low potential for leaching of organic constituents.

                    Some potential for sorption and leaching of
                    metals, but less than with thermoplastic and
                    metallic materials.

                    Particularly resistant to chemical attack, including
                    aggressive acids and organic solvents.

                    Not subject to corrosion.

                    Resistant to biological attack.

 Stainless Steel (SS)    Very low potential for sorption of hydrocarbons.

                    Not subject to leaching of organic constituents.

                    Significant potential for sorption and leaching of
                    metals.

                    Subject to chemical attack by organic acids and
                    sulfur-containing species.

                    Subject to corrosion.

                    Subject to biological attack.

 Polyvinylchloride (PVC) Potential for sorption of hydrocarbons, but may
                    be less than with fluoropolymers.

                    Leaching of organic constituents may occur
                    through chemical degradation by organic
                    solvents.

                    Sorption and leaching of some metals.

                    Subject to chemical attack by organic solvents.

                    Not subject to corrosion.

                    Resistant to biological attack.
The use of a specific number of well-bore volumes as the sole
criterion for purge volume  has been  applied extensively in
ground-water sampling with recommendations in regulations
and the literature ranging from less than one to over 20 (Herzog
et al., 1991). In addition, definitions of well-bore volume have
included the volume contained within the casing, that volume
plus the pore volume of the filter pack, and the volume of the
entire borehole.  Despite its widespread use, the well-bore
volume  approach does not directly address the issue of obtain-
ing representative ground water because there is no proven
relation between the number of well volumes removed and the
completion of purging.   The combination  of  details of  well
construction,  contaminant  distribution, and geologic  and
hydrochemical conditions result in unique conditions at every
well such that the volume of water required for purging cannot
be determined a priori. It is impossible to predict the magnitude
of error that might be introduced  by arbitrarily  choosing a
number of well volumes that results  in incomplete purging.

Determining purge volume by measuring field parameters is
also widely used. The assumptions implied in this approach are
that; (1) as these parameters stabilize, stagnant water in the well
has been replaced by ambient ground water, and (2) this water
contains representative concentrations of the compounds  of
interest.  However,  field experiments conducted by Gibs and
Imbrigiotta (1990) showed that field parameters often stabilized
before the concentrations of VOCs.  In almost 90% of their
experiments, field parameter measurements stabilized when
three well casing volumes had been purged while VOC concen-
trations stabilized after three well volumes in only about half of
the cases. Likewise, Pearsall and Eckhardt (1987)observed in
a series of field experiments that trichloroethylene concentra-
tions continued to change after three hours of pumping at 1.2 U
min whilef ield parameters stabilized within 30 minutes. Further-
more, measurements of  individual field parameters may not
reach stable values at the same purge volume suggesting that
some parameters are more sensitive to purging than others. For
example, Pionke and Urban (1987) found that temperature, pH,
and EC values of purge water from 14  wells studied generally
stabilized before dissolved oxygen and nitrate concentrations.
Puls et al. (1990) found  that while temperature, pH, and EC
values generally stabilized in less than a single well-bore vol-
ume,  other indicators such as dissolved oxygen and turbidity
required up to three well-bore volumes before stabilization. Puls
et al.  (1990) considered reduction of turbidity to stable values
using low pumping rates as critical to the collection of represen-
tative metals samples.  It should be pointed out that in all of the
cases mentioned above,  reliance on commonly measured pa-
rameters (temperature, pH,  and EC) alone  would apparently
have  underestimated the proper purge volume. These results
suggest that the choice of purge indicator parameters should be
made such  that the indicators are  sensitive  to the purging
process and relate to the hydrochemical constituents of interest.
This can be accomplished by evaluating the patterns of indicator
parameters and ground-water constituents during well purging
(a purge-volume test) to determine the appropriate purge
volume.

Another implied assumption  of the field parameter approach is
that purging  will  result in the stabilization  of all constituent
concentrations at approximately the same  purge volume.  In
many hydrogeologic systems this assumption may not be valid.
For example, in aquifers contaminated by several VOCs, con-
centration trends during pumping  may be very different. In an
evaluation of a purge-volume test, Smith et al. (1988) found that
concentrations of two compounds started relatively high  and
decreased with purging to below detectable levels. Two other
compounds that were undetected at three casing volumes were
detected at four casing volumes and their concentrations in-
creased until stabilizing at ten casing volumes. A fifth compound
remained at a constant concentration  throughout the purge-
volume test.   The authors did not report the  concentrations
observed orthe volumes pumped, but it is clearthat underthese
                                                          10

-------
conditions the choice of purging volume could significantly
impact interpretations of contaminant concentrations.

It is important to keep in mind that the distribution of contami-
nants in limited plumes within a ground-water system is gener-
ally in contrast to the more homogeneous distribution of natural
hydrochemical conditions in space and time.  Consequently,
attaining stable concentrations of field parameters,  or even
gross chemistry, may not indicate a representative sample of the
targeted aquifer volume around a monitoring well (Keely and
Boateng, 1987). As a result,  these workers suggest that the
'inherent variability  of the concentration of contaminants in
many plumes far outstrip the additional variability  potentially
induced by incomplete purging,' and recommend that spatial
and temporal variations in contaminant concentrations be stud-
ied to determine optimum purge volumes.

Methods of determining purge volume by  estimating when
hydraulic equilibrium occurs between water stored in the casing
and water entering the casing may be useful where conserva-
tive, non-varying constituents  are being monitored. However,
determining hydraulic equilibrium by estimating the time at
which water levels in the well are no longer affected by casing
storage (the method of Papadopulos and Cooper, 1967) may
lead to erroneous results (Gibs and lmbrigiotta,1990). These
workers compared the calculated hydraulic equilibrium volume
to measurements of field parameters and VOC concentrations
during several  well purging experiments and found that the
calculated volume consistently underestimated the volumes
required to reach both stable field measurements and stable
VOC concentrations. The casing storage method might provide
an approximation of purge volume under conditions where
conservative, non-varying constituents are being monitored but
the available evidence  suggests that only  sampling for the
constituents of  interest will provide a direct indication of when
their concentrations stabilize.

Recent research reviewed by Puts et al. (1990) demonstrates
that contaminants may be transported in ground water  by
association  with colloidal-sized particles which are generally
described as particles less than  10 u.m in diameter.  Where
contaminant transport by association with colloids is an impor-
tant mechanism, obtaining representative concentrations of
mobile colloids becomes critical to sample representativeness.
However, the acts of purging, sampling, and even placing the
sampling device in the well have been demonstrated to signifi-
cantly impact colloidal suspension in the sampling zones of
monitoring wells (Puls et al.,  1991; Kearle et al., 1992).  If a
significant portion of contaminants are transported in associa-
tion with colloids, the results of these investigations and others
suggest minimizing or eliminating purging, minimizing sampling
flow rates (100  to 500 mL/min), and using dedicated sampling
devices placed within the well intake may all be necessary to
collect representative ground water samples.  This low-volume
approach to purging and sampling was earlier proposed by
Robin and Gillham (1987) when sampling for conservative, non-
varying parameters  in high-yield wells.  Using non-reactive
tracers, these workers demonstrated that natural ground water
movement through the well  intake was sufficient to prevent the
formation of stagnant water with respect to conservative, non-
varying parameters, making purging large volumes unneces-
sary. Robin and Gillham (1987) pointed out that, under these
hydraulic and hydrochemical  conditions, representative
samples can be collected with little or no purging using dedi-
cated devices  positioned within  the well intake.  In order to
resolve the issue of low-volume  purging, however, it appears
that more  research is necessary to better understand colloid
movement in ground-water environments, their importance to
contaminant transport, and their implications to purging and
sampling techniques.

Purge Rate and Depth

It was suggested previously  that the  pumping rate at which
purging is  conducted may impact sampling results.  Although
few detailed studies have been conducted to directly address
this issue,  the results of a few specific field studies suggest the
types of impacts that purging rates might have on sampling
results.  For example, Imbrigiotta  et al. (1988) reported that
purging rates of 40 L/min were found to produce VOC concen-
trations  up to  40% higher than concentrations  obtained at
purging rates of 1  L/min. Likewise, purging with a high-speed
submersible pump at a rate of 30 L/min was found to generally
produce higher colloid concentrations and larger particle sizes
than a low-speed pump at rates lower than 4 L/min (Puls et al.,
1990). Despite these colloid differences, however, metals and
cation concentrations did not  necessarily correlate to pumping
rate.  Both investigators attributed the  variability to the effects
that different  pumping  rates  had  on  the  distribution  of
hydrochemical conditions nearthe well. Imbrigiotta etal. (1988)
further concluded that the variability  in VOC concentrations
caused by purging rate  was  of the same magnitude as that
observed in a comparison of seven types of sampling devices,
suggesting that purging rate may  be at least as important to the
collection  of representative samples  as the type  of  device
utilized. Puls et al. (1990) suggested that the colloid differences
might also have resulted from entrainment of normally non-
mobile suspended particulates in the wells.

Although the issue remains unresolved, it appears that employ-
ing pumping rates that allow sample  collection with  minimal
disturbance of  the sample and the hydrochemical environment
in and nearthe well may aid in minimizing sampling error. To this
end, it has been suggested that the purging rate be chosen such
that the rate of ground water entering the well intake  is not
significantly higher than the  ambient  ground-water flow rate
(Puls and Barcelona, 1989). Under typical hydraulic conditions,
this may be possible with pumping  rates between 100 and 500
mL/min.

The depth at which purging is conducted may also affect sample
representativeness.  At high pumping  rates or  in low- and
medium-yield wells, purging at depths far below the air-water
interface may introduce error  because  stagnant water from the
well above the pump may be drawn into the pump inlet. Under
these conditions, pumping near the air-water interface signifi-
cantly reduces the time required to remove stagnant water by
reducing  mixing from above the pump intake  (Unwin  and
Huis.1983; Robin and Gillham,  1987).  Keely and Boateng
(1987) suggest lowering the pump during purging so as to
further reduce the possibility of migration of stagnant water into
the intake during sample collection. On the other hand, under
high-yield  conditions, placing the pump at the well intake and
utilizing low pumping rates may serve to isolate the stagnant
water in the well bore  above  the pump thereby providing
representative samples with minimal purging (Barcelona et al.,
                                                        11

-------
 1985; Robin and Gillham, 1987).  Unwin and Maltby (1988)
 reported that pumping  at virtually any depth within a well,
 including the intake, may lead to contamination of samples by
 stagnant water from above the pump inlet although their labora-
 tory investigation demonstrated that at a pumping rate of 1  U
 min,  samples collected within the well intake contained less
 stagnant water than samples collected above the well intake.
 Regardless of the depth of the pumping device, if a stagnant
 water zone develops near the water surface subsequent move-
 ment of the pumpor placement of a sampling device through this
 zone may cause contamination of the device by stagnant water.

 As suggested above in the discussion of purge volume, certain
 hydrogeologic conditions and chemical constituents may re-
 quire that samples be collected with little or no purging  using
 dedicated devices positioned within the well intake. Underthese
 circumstances, it would also be necessary to utilize low purging
 and sampling rates so as to minimize disturbance of the sample
 and sampling environment and to prevent migration of stagnant
 water from the well bore down into the sampler intake.

 Purging in Low-Yield Wells

 Purging low-yield wells introduces conditions that by definition
 don't occur in medium- to high-yield wells. These conditions,
 which tend to have their greatest impact on constituents that are
 sensitive to pressure changes and/or exposure to construction
 materials or the atmosphere, often result from dewatering the
 filter pack and well intake.  Dewatering may produce a large
 hydraulic gradient between the adjacent water-bearing zone
 and the filter pack as a result of the large drawdown  in the well
 and  the low hydraulic conductivity of the formation.   One
 consequence of this condition may be the formation  of a seep-
 age face at the borehole wall causing ground water entering the
 borehole to flow down the borehole wall and fill the dewatered
 filter pack from the bottom up.  Formation of a seepage face
 increases the surface area of the interface between the liquid
 phase (ground water) and vapor phase (headspace in the well)
 available for transfer of solutes. Another consequence of the
 large hydraulicgradient is the sudden pressure decline from the
 pressure head in the water-bearing zone to atmospheric pres-
 sure in the pumped well. The sudden release of this pressure
 may  cause losses from solution (by degassing or volatilization)
 of solutes that have combined partial pressures, with that of
 water, greater than atmospheric. Finally .because water levels
 recover slowly in low-yield wells, significant changes  in the
 chemical composition of the ground water may occur through
 sorption, leaching, or volatilization before sufficient  volume is
 available for sample collection.

 In a  field study of purging and sampling  in low-yield wells,
 Herzog et  al. (1988) found that some VOC concentrations
 increased significantly from pre-purging conditions during the
 first two hours of water level recovery. For example, chloroben-
 zene  concentrations increased from 25 u.g/L before purging to
 over  125 u^g/L at two hours after purging.  Concentrations
 generally did not change significantly after two hours, although
 some concentrations declined.  Although Herzog (1988) pro-
 vided no explanation for the observed concentration trends,
they were likely caused by more representative ground water
 entering the  well and replacing the purged  stagnant water.
Smith et al. (1988) reported very different results in  their field
study of a trichloroethylene plume. Concentrations of trichloro-
ethylene declined from 100 ^g/L directly after purging to 10 ng/
L 24 hours after purging. In a laboratory study, McAlary and
Barker (1987) found that if the water level in a simulated well was
drawn  down  below the intake, VOC concentrations during
recovery declined 10% in five minutes and 70% in one hour.
These changes were attributed to volatilization from the water as
it entered and filled the  well.

In summary, aspects of well purging important to collection of
representative samples include purging volume, pumping rate,
depth of the purging device, and time of sampling in low-yield
wells. Although error is strictly dependent on individual well and
site conditions, the available evidence suggests that order-of-
magnitude  errors may easily  result from improper purging
techniques.  In low-yield wells, time of sampling is clearly  an
important source of  error although there are too  few data
available to completely understand concentration  trends in
these situations.

Contamination concentrations during purging vary in ways that
are often difficult to predict, and various compounds may even
exhibit  opposite  trends.  To estimate the appropriate purge
volume, it may be necessary to conduct preliminary purge-
volume tests with sampling at regular intervals during purging.
These tests may be useful fordetermining how indicator param-
eters and constituent concentrations respond to purging rates,
purging volumes, and the distribution of contaminants around
the well. In addition, for certain sensitive constituents such as
trace metals under certain hydrogeologic and hydrochemical
conditions, low-volume purging and sampling should be consid-
ered with dedicated sampling devices installed atthe well intake.

SAMPLE COLLECTION

Sample collection involves physical removal and transport of
ground water from depth (generally from a monitoring well) to
ground surface and into a sample container. As such, collection
methods may have great potential for alteration of the sample's
chemical state.  Sampling devices must be chosen and used
carefully to ensure that error is minimized.  Important aspects of
sample collection include sampling device, collection time after
purging, and sampling depth.

Chemical Impacts

Sampling devices can cause chemical changes in the sample by
contact  with materials of the device (sorption, desorption,  or
leaching) or by the physical action of the device. Although the
materials of the  device are a potentially significant source of
sample  error,  that topic was discussed  previously and the
following discussion will address chemical changes produced
only by the operation of the sampling device.

Because fluid  pressure in the saturated zone is greater than
atmospheric, ground-water samples brought to the surface will
tend to be under higher pressure conditions than the ambient
atmosphere.  Exposure of these  samples to the lower atmo-
spheric  pressure will cause degassing and/or loss of volatile
constituents until the partial pressures of the contained volatile
components reaches equilibrium  with atmospheric pressure.
Degassing may cause losses of oxygen (O2), methane (CHI
nitrogen (N2), or carbon dioxide (CO.,), while volatilization migm
affect any solute that exists as a liquid, solid, or gas under in situ
                                                        12

-------
ground-water temperature and pressure conditions (Gillham et
al., 1983). Furthermore, toss of CO2 may raise the pH which can
lead to precipitation of dissolved constituents, particularly iron
(Gibb et al., 1981). Constrictions in the flow path within a device
may also raise the sample pH by changing the partial pressure
of CO2 (Herzog  et al., 1991).

Exposure of samples to the atmosphere, or the driving gas used
in some devices, may introduce oxygen causing oxidation of
iron, manganese, cadmium,  or other species.  Oxidation of
ferrous iron to  ferric iron has important implications to the
speciation and concentrations of many constituents in ground
water samples (Herzog etal., 1991). Contaminants may also be
added to the sample by exposing it to the atmosphere or driving
gas.

Sampling Devices

Sampling devices designed for use in conventional monitoring
wells can be divided into four general types:  grab, positive
displacement (no gas contact), suction lift, and gas contact
(Pohlmann and Hess,  1988).  Grab samplers  include open
bailers, point-source bailers,  and syringe samplers. Positive
displacement samplers are usually submersible pumps such as
bladder pumps, gear-drive pumps, helical-rotor pumps, and
piston pumps. Suction lift devices include peristaltic pumps and
surface centrifugal pumps while gas contact pumps include
those devices that lift waterto the surface by direct gas pressure.
Submersible centrifugal pumps, which operate on the principle
of positive displacement at low flow rates, develop a partial
vacuum at the  pump impellers at higher flow rates.  For this
reason, high-speed submersible centrifugal pumps without vari-
able motor speed capability should be considered  as distinct
from positive displacement pumps.  On the other hand, sub-
mersible  centrifugal pumps are now available that can be used
in 5.1-cm (2-inch) diameter wells and that allow  adjustment of
the motor speed to produce very low flow rates.  If used  at low
flow rates, these low-speed pumps could conceivably eliminate
the application of a partial vacuum to the sample and thereby
can be considered as positive displacement pumps. Discussion
of the operating principles of many of ground-water sampling
devices,  and their potential for sample bias, can be found in
Gillham etal. (1983).

Sampling devices for conventional monitoring wells can be used
either portably or in a  dedicated mode. Portable devices are
used to collect samples in more than one well and so may cause
cross-contamination between installations or sampling events if
not properly decontaminated.  Dedicated devices are perma-
nently installed in a single well and are generally not removed for
cleaning between sampling events. Dedicated samplers, when
also used for well purging, may not have adequate flow control
for effective purging in large wells (high  discharge rate) and
sampling (low discharge rate). Furthermore, parts of dedicated
samplers may sorb contaminants during periods of contact with
ground water between sampling events and then release them
during sample collection. Alternatively, if inappropriate materi-
als are used in the construction of dedicated samplers, contami-
nants may leach  from these materials between  sampling
events.

To study the effects of sampling devices on sample quality,
investigations have been conducted both in the laboratory and
in the field. Laboratory studies can provide values of absolute
sample error by testing under controlled conditions, particularly
constituent concentration. However, by their very nature, labo-
ratory experiments represent ideal conditions that can never be
duplicated in the field and therefore may not include important
field-related errors.  On the other hand, field studies include all
the physical,  chemical, biological, and operating conditions
present in field sampling efforts, but the true concentration of the
constituents of interest are unknown. As a result, field compari-
son studies cannot provide values of absolute sample error, only
the relative ability of individual devicesto recover the constituent
of  interest.

Values of field chemical indicator parameters can often be the
first indication of sample errors due to sampling device. Labo-
ratory investigations of a wide range of sampling devices by
Barcelona et al. (1984) revealed that pH and redox potential (Eh)
were the most sensitive to sampling device. The largest errors
were produced by a peristaltic pump (an increase of 0.05 pH
units and a 20 mV decline in Eh). All tested devices had O2 and
CH  losses of 1% to 24%, although positive displacement
devices and an open-top bailer resulted in the lowest losses and
the highest precision in that study. A field study by Schuller et
al. (1981) found that, as a result of CO stripping, an air-lift pump
and a nitrogen-lift pump produced pH values up to 1.0 pH unit
higher than a  peristaltic pump and opentop bailer. Other field
studies concluded that open-top and dual-valve bailers pro-
duced no  more error in field parameter values than  bladder
pumps (Houghton and Berger, 1984). In that study, which used
bladder pump values as a standard for comparison, a peristaltic
pump and a  high-speed submersible centrifugal pump had
increases  in pH of about 0.06 pH units and approximately 20%
declines in dissolved oxygen (DO) concentrations. A gas-driven
piston pump had an increase in DO of 8% to 36%. Temperatures
increased  up to 5% in samples collected with the peristaltic and
piston pumps and 14% in samples collected with the high-speed
submersible centrifugal pump.

Most major dissolved ions are relatively stable and not greatly
affected by collection method. Schuller etal. (1981) determined
that concentrations of calcium, chloride, fluoride, potassium,
magnesium, and sodium collected at two field sites were not
significantly affected by the choice of suction, gas-contact, or
bailer device.  Dissolved metals, on the other hand, are very
sensitive to sample aeration and degassing during  sampling.
Schuller etal.  (1981) found that iron and zinc concentrations in
samples collected with two gas contact devices were, at most,
30% of those collected with either a peristaltic pump or a bailer.
Field studies  of 18  wells with seven sampling devices by
Houghton  and Berger (1984) showed significant declines in
metals concentrations for a gas contact device when compared
to positive displacement pumps, grab samplers, and a peristaltic
pump.  Houghton  and Berger (1984)  also  found that
coprecipitation of arsenic and zinc with iron led to significant
losses of these constituents in samples collected with a high-
speed submersible centrifugal pump.

Sampling device impact on VOC concentrations is of particular
importance because of the high sensitivity of these compounds
to  sample aeration  and  degassing and the critical need for
accurate VOC data in many site investigations. Several labora-
tory experiments have shown that positive displacement de-
vices (bladder, piston, and helical-rotor pumps)  and  conven-
                                                        13

-------
 tional grab samplers (open-top and dual-valve bailers) provide
 the most accurate VOC concentrations (Barcelona et al. 1984;
 Unwin, 1984; Schalla et al., 1988; Unwin and Maltby, 1988).
 Although the bladder pump and bailers that  Barcelona et al.
 (1984) tested produced less than 3% losses in VOC concentra-
 tions, these same devices produced  up to 10% losses in other
 studies, even undercaref ully-controlled conditions. Suction and
 gas-contact devices tested  in these studies, and a study of
 peristaltic pumps by Ho (1983), resulted in 4% to 30% losses in
 VOC concentrations.  Of those devices that performed well, no
 relation was found between sampler accuracy and VOC con-
 centration  over a range of 80 to 8000 u.g/L (Barcelona et al.,
 1984; Unwin, 1984). The devices that performed poorly, how-
 ever, often revealed significant increases in error as concentra-
 tion increased (Barcelona et al., 1984). From these laboratory
 studies it appears that certain classes of samplers, specifically
 suction and gas-contact, can lead to significant error in VOC
 concentrations as a  result  of volatilization from the sample
 during collection.

 A positive relation  between increased losses of VOCs from
 solution with increase in Henry's law constant was predicted by
 Pankow (1986)  based  on theoretical considerations of the
 factors leading to bubble formation  in water  during  sampling.
 Physical experiments have shown a  strong positive correlation
 between compound volatility and Henry's law constant for a
 peristaltic pump, some correlation for a helical-rotor pump, but
 no correlation for a bailer and bladder pump (Unwin and Maltby,
 1988). On the other hand, Barker et al. (1987) found no clear
 correlation for  a peristaltic pump and gas-drive sampler and
 Barker and Dickhout (1988) found  no clear correlation for a
 peristaltic, bladder, or inertial-lift pump, although the range of
 Henry's law constants was small.  These findings suggest that
 compound volatility may not be an important source of bias for
 some positive displacement and grab samplers but there  may
 be potential for losses for samplers that impose a suction on the
 sample.

 Many field comparisons of sampler effectiveness  verify the
 findings of laboratory experiments, despite the increased num-
 ber of variables  involved in the field studies.  Investigations
 involving a variety of field conditions by Muska et al. (1986),
 Pearsall and Eckhardt (1987), Imbrigiotta et al. (1988), Liikala et
 al.  (1988), Yeskis et  al. (1988), and Pohlmann et al. (1990)
 concluded that positive displacement devices produced the
 highest VOC concentrations, and therefore introduced the least
 error into VOC determinations. The accuracy of grab samplers
 was more variable:  some studies showed little difference
 between the VOC recoveries of bailers and positive displace-
 ment pumps (Muska  et al.  (1986);  Imbrigiotta et al. (1988);
 Liikala et al. (1988)), but Imbrigiotta et al. (1987), Yeskis et al.
 (1988), and Pohlmann et al. (1990)  reported that bailer VOC
 concentrations were significantly lower than positive displace-
 ment pumps; 46% to  84%  lower in  the work of Yeskis et al.
 (1988). Pearsall and Eckhardt (1987) found that a bailer was as
 accurate as a positive displacement pump at concentrations in
 the range of 76 to 79 u.g/L but recovered 12% to 15% lower
 concentrations in the range 23 to 29  u.g/L.
et al. (1988) concluded that syringe sampler accuracy was lower
than the pumps but that precision was comparable.  Other
samplers field-tested produced significant error:  a peristaltic
pump and surface centrifugal pump were found by Pearsall and
Eckhardt (1987) to be less accurate, but not necessarily less
precise than the other samplers tested. Imbrigiotta et al. (1988)
found the same for a peristaltic pump.

In ground-water environments charged with dissolved gases,
collection of accurate VOC samples can be even more problem-
atic. VOC losses of 9% to 33% were produced by a peristaltic
pump in laboratory and field studies of water containing high CO2
(laboratory study) and CH4 (field study) concentrations (Barker
and Dickhout, 1988).  Losses of 13% to 20% were produced by
a bladder pump in the laboratory study, while an inertial-lift pump
produced no losses. No differences between results from these
two pumps were observed in the field. The CO, concentrations
used in the laboratory investigation were higher than under
environmental conditions, but this study nonetheless suggests
that degassing during sample collection, even with a positive
displacement pump, can lead to significant error in VOC concen-
trations (Barker and Dickhout, 1988).

Several "in situ" devices have been developed to alleviate some
of the problems inherent to conventional monitoring wells and
sampling devices.   These  devices generally utilize sample
containers under reduced pressure to collect samples directly
from the water-bearing zone, without exposure to the atmo-
sphere or excessive agitation.  In afield study, Pohlmann et al.
(1990) found that two types of in situ devices delivered samples
with VOC concentrations that were not significantly different
from those collected by a  bladder pump  in a conventional
monitoring well.

Although the field studies outlined above cannot provide values
of absolute sample error, they do  provide information on the
effectiveness of various devices under actual operating condi-
tions. The results of the laboratory studies, in conjunction with
field studies,  indicate that suction pumps are very likely to
introduce  significant  error  into VOC  determinations.
Grabsamplers,  especially bailers,  are also likely to produce
errors if not operated with great care because their successful
operation is closely related to operator skill.  Under  certain
conditions,  for certain parameters,  and if operated by skilled
personnel, bailers can produce representative samples. How-
ever, much of the research outlined here indicates that positive
displacement pumps consistently provide the lowest potential
for  sample error.  Appropriate application of most types of
positive displacement pumps can reduce sampling device con-
tribution to error well below the levels of some other aspects of
ground-water sampling protocol.

A summary of the impacts that some commonly-used sampling
devices have on ground-water sample  quality is shown in
Table 4 which was compiled from the sources referenced in this
section and Nielsen and Yeales (1985).

Collection Depth and Time after Purging
Another grab sampler, the syringe sampler,  also produced   The length of time between well purging and sample collection
mixed  results.  Muska et al. (1986) concluded that syringe   may influence the representativeness of samples by exposing
sampler accuracy and precision were not significantly different   ground water to the effects of atmospheric diffusion, interaction
from those of the positive displacement pumps while Imbrigiotta   with well materials, and contaminant volatilization.  Smith et al.
                                                        14

-------
 TABLE 4. SOME IMPACTS THAT THE OPERATING PRINCIPLES OF
  GROUND-WATER SAMPLING DEVICES MAY HAVE ON GROUND-
    WATER SAMPLE QUALITY (WITH THE EXCEPTION OF GRAB
 SAMPLERS, FT IS ASSUMED THAT THESE DEVICES REMAIN IN THE
           WELL DURING THE SAMPLING PROCESS).


Operating Principle    Impacts
Gas Contact
Grab
Contact with drive gas may cause loss of
dissolved gases and increase pH.

Contact with drive gas may volatilize sensitive
solutes.

Exposure to driving gas may introduce
contaminants or oxidize sensitive constituents.

Contact with atmosphere during sample recovery
and transfer may cause loss of dissolved gases
and increase pH.

Contact with atmosphere during sample recovery
and transfer may volatilize sensitive solutes.

Exposure to atmosphere during sample recovery
and transfer may introduce contaminants or
oxidize sensitive constituents.

May be contaminated when passing through
zone of stagnant water.
Positive Displacement   Minimal if discharge rate is low.

Suction Lift
Application of suction to sample may cause loss
of dissolved gases and increase pH.

Application of suction to sample may volatilize
sensitive solutes.
High-Speed           Suction applied at pump intake may cause loss
Submersible Centrifugal of dissolved gases and increase pH.

                    Suction applied at pump intake may cause
                    volatilization of sensitive solutes.

                    Application of excessive head to the sample may
                    cause degassing or volatilization.

                    Heat produced by pump motor may increase
                    sample temperature.
(1988) found that trtehloroethane concentrations in a well de-
clined from 170 u.g/L immediately after purging to 10 u.g/L 24
hours later.  To  ensure consistency and to reduce potential
errors when sampling in high-yield wells, it is generally recom-
mended that samples be collected immediately following
completion of well bore purging. In low-yield wells, however, low
water level recovery rates may require that sampling be delayed
until sufficient volume is available.  Determination of sample
collection time in low-yield wells is more problematic and may
require site-specific sampling experiments.

To reduce potential errors caused by mixing with stagnant well
water during sampling, research has suggested that the sam-
pler intake be located either  within the screened  interval
(Giddings, 1983; Bryden et al., 1986; Robin and Gillham, 1987)
or at the top of the screened interval (Unwin, 1982; Barcelona
and Helf rich, 1986) so samples can be obtained soon after fresh
ground water enters the well bore.  However, in cases where
wells are screened over a  long interval,  it is important to
determine if contaminants are vertically stratified in the well.
Pearsall and Eckhardt (1987) found that TCE concentrations of
samples collected at the top of a 10-foot screened interval were
30% lower than those collected at the bottom and attributed the
difference to vertical stratification of VOCs within the screened
interval.  Errors associated with sampler intake placement have
not been quantified to date but are likely strongly controlled by
conditions at each well.

The use of samplers that must pass through the zoneof stagnant
water that invariably remains near the water level, even in a
properly-purged well, may also introduce error.  For example,
grab samplers, which often require repeated entry and retrieval
from the well during sampling, may be contaminated by this
zone of stagnant water or may mix stagnant water into the water
column.  Likewise, if the purging device is not used for sampling,
removal of the purging device and installation of the sampling
device may have a similar effect. The use of a dedicated device
for  both purging and sampling would  significantly reduce this
source of error but may introduce others.

SAMPLE FILTRATION

Ground-water samples collected for analysis of certain constitu-
ents are often filtered in the field prior to transfer to the appropri-
ate container.  Reasons for filtration include prevention of
geochemical reactions that might occur with particulates during
sample shipment  and storage,  removal of  suspended sedi-
ments so as to analyze only dissolved constituents, and removal
of fine-grained sediments which might interfere with laboratory
analyses. Because filtration may contribute to sample error by
the method employed or by the choice to filter, it is of the utmost
importance to confirm the objectives of the sampling program
and the implications of filtering when choosing whether to filter
and, if so, the filtration technique.

Puls and Barcelona (1989) point out that if. mobile trace metal
species are of interest to the investigation filtration may remove
metals adsorbed onto some colloidal particles, leading to under-
estimates of  dissolved metals concentrations  and, therefore,
concentrations of mobile species. Conversely, if the objective of
metals analysis is to quantify total dissolved metals concentra-
tions, colloids with sorbed  metals that pass through the filter
material may result in overestimates of dissolved metals con-
centrations (Puls and Barcelona, 1989). These workers indicate
that filtration should not be used as  a means of removing from
the sample particulates that result from poor well construction,
purging, or sampling procedures because the misapplication of
filtration may introduce substantial bias to trace metal determi-
nations. If filtration is deemed necessary, it should be conducted
soon after sample collection as temperature changes,  CO
                                                          15

-------
invasion, or the presence of particulates may have adverse
effects on trace metal concentrations or dissolved solids content
(Unwin, 1982). Factors important to properfield filtration include
filter pore size, material, and method, and holding time prior to
filtration.

Filter pore size has very important implications for determina-
tions of metal species and major ions in ground-water samples
as a result of the inclusion of undissolved material. Constituents
showing the greatest  sensitivity to filter pore size include iron
and zinc (Gibb et al., 1981), iron and aluminum (Wagemann and
Brunskill, 1975), and iron, aluminum, manganese, and titanium
(Kennedy et al., 1974).  In all cases, larger filter pore  sizes
produced higher concentrations of these constituents because
the larger pore-size filters allowed more particulates to pass. In
fact, Kennedy et al. (1974) found that concentrations of some
 metal species in samples filtered through 0.45 u.m filters were up
to five times higher than in samples filtered through 0.10 u,m
filters.  These results suggest that if field-filtering is deemed
 necessary, smaller pore size filters may reduce sample error.

 Sorptive losses of trace metals during filtration can also  intro-
duce error into metals determinations. Truitt and Weber (1979)
found that both cellulose acetate and polycarbonate 0.4 u.m filter
 membranes sorbed copper and lead from solution. For ex-
 ample,  losses of copper averaged 8.6% with cellulose acetate
 membranes  and 1.1% with polycarbonate  membranes.
 Gardner and  Hunt (1981) found that sorption  of  lead onto
 cellulose acetate membranes resulted in losses of 20 to 44%
from a  synthetic solution.  These losses  were reduced to 5 to
 24% by pre-rinsing the filter apparatus with the test solution
 (Gardner and Hunt, 1981). Studies by Jay (1985) found that
 virtually all filters require pre-rinsing to avoid sample contamina-
tion by  leaching of anions from the filter material.

Although filter material and pore size have been  the subject of
considerable research, less effort has been directed toward
understanding the effects  of filtration method  on  dissolved
constituents.  Of the  few studies available, Stolzenburg and
Nichols (1985) investigated the effects of sampling and filtration
method on concentrations of iron and arsenic. Their laboratory
study showed that samples that were vacuum-filtered after a 10-
minute  holding time delay experienced iron losses of 20% to
90% and arsenic losses of 45% to 100% compared to in-line
filtered samples. The ranges of percentages were due to the use
of several types of sampling devices. Later experiments by
Stolzenburg and Nichols (1986)  added immediate vacuum
filtering  of samples.   Both immediate and delayed vacuum-
filtration produced similar iron concentrations but these concen-
trations were 17% to 67% lower than concentrations produced
by in-line filtration. In both the 1985 and 1986 reports, in-line
filtering  produced concentrations that were comparable to the
source concentrations of approximately 8 mg/L  iron and 0.05
mg/L arsenic suggesting that in-line filtration methods were the
most effective of those tested. These experiments also sug-
gested that filtration method may cause greater losses of certain
constituents than the type of sampling device used.  Unfortu-
nately, commonly-used pressure filtration methods were  not
compared to in-line and vacuum filtration  methods in these
experiments.

Clearly,  sample filtration can lead to substantial  error in  trace
metal determinations even if procedures are carefully followed.
Because of this great potential for error, filtration should not be
used to correct for sedimentation problems that result from
poorly designed or constructed wells or incomplete develop-
ment. If filtration is deemed necessary, pre-cleaning the filters
can reduce error.  In addition, the limited research into filtration
methods in ground-water investigations suggests that in-line
methods may result in the least sample error. However, even
under ideal conditions, sample filtration may lead to significant
error in  determinations of  metals concentrations, suggesting
that analysis of both filtered and non-filtered samples should be
considered.

EQUIPMENT DECONTAMINATION

Contaminants on  equipment that contacts ground water and
samples, including drilling equipment, well materials, sampling
devices, and sample bottles may be another source of sample
error. Error may be introduced by the addition of contaminants
to ground water or samples (contamination) or by the convey-
ance of ground water and/or contaminants from one sampling
installation  or zone to another (cross-contamination).  Cross-
contamination is most often a problem when equipment, particu-
larly sampling devices, is used portably but not properly cleaned
between installations.  The process of  cleaning equipment
before installation or after sampling is generally referred  to as
decontamination.

Drilling equipment can be a source of gasoline, diesel fuel,
hydraulic fluid, lubricating  oils and greases, and paint,  all of
which can  be  introduced  into the subsurface during drilling
operations. In addition, contaminated soil, scale, orwaterfrom
the site may enter the borehole directly or by adhering to drilling
pipe or  other down-hole equipment.  If  these contaminants
originate from  other  sites  or  boreholes,  cross-contamination
may result (Fetter, 1983).  Steam cleaning  is often  recom-
mended as a method of decontaminating the drilling  rig and
equipment  before use  and between boreholes.  In addition,
placing  down-hole drilling  equipment on plastic sheeting or
other appropriate material while not in use may reduce contami-
nation from soils or other sources of contaminants at ground
surface.

Well casing and screen materials may contain residues of the
manufacturing process including cutting oils, cleaning solvents,
lubricants, and waxes (Aller et al., 1989). These residues must
be removed prior to well installation to prevent contamination or
other chemical impacts on samples.  A  procedure generally
recommended is  to  wash the casing in  a strong  detergent
solution followed by a tap water rinse (Barcelona et al., 1983;
Curran and Tomson,  1983) although steam cleaning or a high-
pressure hot water wash may be required for removal of some
oils, lubricants, and solvents (Aller et al., 1989).

Equipment used portably can lead to cross-contamination by
transferring water and contaminants from one installation to
another. In a survey of state and federal environmental regula-
tory agencies, Mickham et  al. (1989) found that procedures for
decontamination of sampling equipment generally include a tap
water rinse, acid or solvent rinse (depending on type of contami-
nation), organic-free water rinse, and airdrying. The survey also
showed that equipment that does not directly contact samples
is generally cleaned by detergent washes and steam cleaning.
These workers found little research  into  the effectiveness of
decontamination procedures.
                                                        16

-------
Korte and Kearl (1985) suggest that high-volume pumping may
sufficiently clean sampling pumps.  In contrast, field experi-
ments conducted by Matteoli and Noonan (1987) determined
that 90 minutes of pumping clean water through 200 feet of
PTFE tubing was  required to reduce the concentrations of
several organic and inorganic constituents to below detection
levels. These workers found that the time required for effective
decontamination was generally related to the type of constitu-
ent.  Freon was still detectable after 120 minutes of pumping.

The effects of cross-contamination can be reduced or elimi-
nated by utilizing equipment dedicated to individual monitoring
wells. As discussed previously, a potential disadvantage of this
approach may be interactions  between the device and ground
water in the well between sampling events.

The  use of plastic sample bottles may be another potential
source of contamination through leaching of organic and inor-
ganic constituents from the bottle materials (Gillham  et al.,
1983).   An  experiment comparing acid-washed and water-
washed plastic sample containers determined that the  risk of
contamination from trace elements in the bottles was greatest
for cadmium, copper, and zinc (Ross, 1986).  In some  cases
copper concentrations  were 50 times higher in  samples col-
lected  in bottles that  were  not  acid-washed.   Moody  and
Lindstrom (1977) suggested that plastic sample containers are
most effectively cleaned with  rinses  in both hydrochloric acid
and nitric acid to leach impurities from the plastics. Their study
further determined that, after acid-washing,  PTFE  and PE
containers were the least contaminating plastic or polymeric
materials.

Interference of ground-water sample chemistry may result from
direct introduction  of foreign  materials  to ground water and
samples or from crosscontamination. Although it appears that
currently used decontamination procedures are adequate in a
general way, little research has been conducted to determine
the effectiveness of specific procedures for individual contami-
nants.  Because they are not standardized, the contribution to
sample  error of a particular  procedure must be evaluated,
perhaps on a case-by-case basis.

To prevent crosscontamination when using sampling devices
portably, rinsate blanks (also referred to as equipment blanks)
should be collected to ensure the effectiveness of decontamina-
tion procedures. This may be accomplished by flushing or filling
the device with Type  II reagentgrade water and collecting a
sample of the rinsate water. Analysis of rinsate blanks for the
contaminants being sampled  will provide an indication  of the
effectiveness of the cleaning  method (U.S. EPA, 1986) and
indicate if modifications of the procedures are required.

SAMPLE TRANSPORT AND STORAGE

Ground-water samples require proper containers, treatment,
transport, and storage to ensure the chemical and physical state
of the sample is preserved until analysis.  Factors that could
potentially lead to error include volatilization, adsorption, diffu-
sion, precipitation, photodegradation,  biodegradation,  and
cross-contamination (Parr et al., 1988).  Methods developed,
and widely accepted, to minimize these effects are summarized
in U.S. EPA (1986) and Herzog et al. (1991).
To reduce the potential for bias during sample handling, appro-
priate chemical preservation of samples  should  take place
immediately upon collection. Increases in pH of 0.3 to 0.4 units
and declines in iron and zinc concentrations of several orders of
magnitude have been observed within seven hours of sample
collection (Schulleretal., 1981). These investigators also noted
slight declines in the concentrations  of calcium,  potassium,
magnesium, manganese, and sodium in unpreserved samples
within 48 hours of collection. To ensure immediate preservation,
it may be advisable in some cases to add chemical preserva-
tives to  bottles immediately before sample collection.  If this
method  is utilized it is important to prevent  the  bottle from
overflowing which might cause the loss(of some of the preser-
vative.

Plastic bottles are usually used  for metals and  major ions
samples to avoid the sorption effects that may occur with glass.
Most types  of plastic bottles can be cleaned with hydrochloric
acid and nitric acid rinses which effectively leach impurities from
the material. PTFE and PE bottles tend to not leach impurities
to samples (Moody and  Lindstrom, 1977) and therefore are the
easiest to clean and have the lowest potential to contaminate
samples. The quantities of impurities leached in these studies
are in the very low ng/cm2 range, generally below the levels in
most site investigations. Sorption of metals onto plastic bottles,
although normally not a problem,  is reduced by  acidifying the
sample and thereby keeping the metals ions in solution (Parr et
al., 1988). Clearly, if adequate cleaning is carried out and pre-
analysis holding times  are not exceeded, contamination  of
major ion and trace metal samples by sample bottles is unlikely.

Organic samples are usually placed in glass containers to avoid
the chemical interferences that may occur with plastic bottles.
The borosilicate  glass used in bottles for  water samples for
organic analyses is easily cleaned and has very little potential for
contamination of samples or sorption from samples.

Cross-contamination of VOC  samples during transport and
storage  can be minimized if accepted procedures are carefully
followed. The evidence presently available indicates that cross-
contamination of VOC  samples at concentrations typical  of
hazardous waste sites is negligible under conditions normally
present  during sample storage (Levine  et al., 1983; Maskarinec
and Moody, 1988). Levine et al. (1983) did note, however, the
thickness of the PTFE lining under the VOC vial septum was
critical to the prevention of cross-contamination  and that con-
tamination was evident when samples were stored near vials
containing saturated aqueous solutions of VOCs.  Trip blanks
can be  utilized to evaluate the potential for contamination of
samples during shipment to the laboratory. These blanks, which
consist of reagent-grade water in bottles of the same type used
for sampling,  can be shipped to the site and laboratory in the
same shipping containers used for samples.

The length of time that a sample can be stored without degrada-
tion is related  to the potential sources of error covered here.  If
adequate measures are taken to reduce these errors, chemical
alteration of the sample during storage can be minimized. Using
commonly-accepted storage methods, concentrations of VOCs
have been shown to be stable after 34 days (Friedman et al.,
1986) and 56  days (Maskarinec and Moody, 1988).
                                                        17

-------
ANALYTICAL TECHNIQUES

To gain perspective into the relative magnitude and importance
of errors introduced during ground-water sampling, it is useful to
quantify the errors involved in laboratory analysis.  Potential
sources of error in the laboratory include glassware, reagents,
laboratory preparation techniques,  and analytical equipment
and apparatus  (Lewis, 1988).  It is beyond the scope of this
document to discuss how each of these aspects of laboratory
operation can impact sample quality except to say that errors
can be detected and controlled by the use of various quality-
control checks.  Vitale et al., (1991)  describe the blanks, dupli-
cate samples,  and  spikes that  ensure the  identification  of
laboratory error. Through the use of these checks, analytical
errors often can be quantified, unlike many aspects of sampling
protocol  where comparison to 'true' concentrations is  usually
impossible.

In a review of  the EPA Contract Laboratory Program (CLP)
database for gas chromatograph/mass spectrometer (GC/MS)
analysis of VOCs, Flotard et al. (1986) analyzed the deviations
in reported concentrations from actual concentrations in  blind
performance evaluation samples.  These deviations can be
considered  measures of analytical  errors, with underreported
concentrations  considered negative error and overreported
concentrations  considered positive error.  The Flotard et al.
(1986) study found errors in reported concentrations of 22 VOCs
from -46.4% for 1,1-dichloroethane to +6.5% for  bromoform.
The results for  methylene chloride exhibited an apparent  error
of +36.6% but this value was attributed to laboratory contamina-
tion of samples and not analysis error. Their review indicated
that 55% of the 22 evaluated VOCs resulted in reported concen-
trations that were more than 20% lower than actual concentra-
tions.  Interlaboratory errors from 35 laboratories were found to
bef ram-3.9% to zero, although datafrom only three compounds
were analyzed.

A similar review of the CLP database for semi-volatile analyses
conducted by Wolff et al. (1986) concluded that the greatest
analytical errors were associated with phenolic compounds,
whose concentrations were consistently underreported. Other
classes of semi-volatiles showed no general trends.  In that
study, analytical errors ranged from -48% for 1,3-dichloroben-
zene  and 2,6-dinitrotoluene to +12% for 4-chlorophenyl-
phenylether. The review indicated that 60% of the 33 com-
pounds evaluated showed analytical errors in excess of -20%,
slightly more than for VOC analyses. Interlaboratory errors for
six compounds  ranged from -51% for phenol-ds to -16% for p-
terphenyl, considerably greater than for the volatile analyses.

The CLP database  has also been  evaluated for errors intro-
duced by inorganic analytical methods (Aleckson et al., 1986).
These workers found that analytical  errors ranged from  -26.5%
to +10.0%, with  most errors falling in the range -10.0% to zero.
The greatest negative errors were found for selenium, silver,
and  thallium.

Barcelona et al.  (1989) tabulated laboratory errors for inorganic
constituents  during  an intensive time-series  investigation of
ground-water chemistry variation.  They found that errors in
determinations of major ions in external performance samples
ranged from -8.1%  (potassium)  to +12.1% (total iron).  An
evaluation of eight analytical laboratories was conducted by
Rice et al. (1988) as part of a uranium mill tailings ground-water
quality investigation.   Constituents of  interest included total
dissolved solids, major ions, trace metals, and radionuclides.
Analysis  of external performance samples during the study
showed that 67% of all analyses were within the acceptable
range but that 60% of the reported values were higher than the
known concentrations. Iron and aluminum were among the
constituents showing the highest analytical errors.

SUMMARY AND CONCLUSIONS

As shown here, many aspects of ground-water investigations
may introduce error into determinations of concentrations of
hydrochemical constituents.  The  potential errors associated
with many of these aspects are summarized in Table 5.

Errors produced during certain aspects of sampling programs
can be identified, quantified, and controlled through the use of
accepted procedures in conjunction with performance evalua-
tion samples.  For example, equipment decontamination and
sample transport and storage have considerable potential for
introducing sample error if not  conducted in  a careful and
consistent manner. In the case of equipment decontamination,
collection and  analysis of rinsate blanks from cleaned equip-
ment can be useful for evaluating the effectiveness of  decon-
tamination procedures. Likewise, errors that may occur during
sample transport can be identified by the use of trip blanks that
are transported to the site and laboratory in the same shipping
containers  as  field samples.   An aspect that  may  require
particular attention and further research is the effectiveness of
decontamination of flexible tubing used for conveying samples
from the sampler to sample bottle.

The potential errors associated with other aspects of sampling
programs are relatively well understood and can be minimized
through appropriate choice of equipment and materials. For
instance, advances in sampling device design and construction
have resulted in the development and widespread use of posi-
tive displacement sampling devices whose operation generally
introduces little sample error. For most compounds, including
VOCs, positive displacement devices allow collection of accu-
rate and precise samples, with concentrations of VOCs typically
within 10% of true concentrations. Some grab samplers, par-
ticularly bailers, may also produce representative samples but
their effectiveness is  highly dependent on mode of operation
and the constituents of interest. Under unfavorable field condi-
tions or when operated improperly, bailers may produce errors
in VOC concentrations from -10% to -80% or more. Most other
types of samplers produce errors of unpredictable magnitude
but show VOC errors of at least -20% in controlled laboratory
experiments. The unpredictable magnitude of errors associated
with many of these devices also  means that they often cannot
provide the precise, or repeatable, measurements usually asso-
ciated with positive displacement devices. As a result, the use
of positive displacement sampling devices may minimize the
introduction of  error into determinations of the concentrations of
sensitive hydrochemical constituents.  Use of other types of
devices may introduce error of unpredictable magnitude.

Potential impacts of materials used in well and sampler con-
struction have been demonstrated, but the implications of these
effects in afield setting remain unclear. Laboratory comparison
studies conducted under static conditions have demonstrated
                                                        18

-------
the potential for rigid PTFE,  PVC, and metallic materials to
introduce error into concentrations of some trace metals and
hydrocarbon compounds.  However, little work has been con-
ducted under conditions simulating dynamic ground- or sample-
water flow or, more importantly, well-purging effects. Despite
these unresolved issues, materials' impacts can be minimized
by choosing well materials compatible with the objectives of the
sampling program  and the hydrogeologic and hydrochemical
conditions of the site. The proper choice of materialscan reduce
chemical effects on water stored in the well between sampling
events and make removal of stagnant water during well purging
less difficult. When monitoring for low hydrocarbon concentra-
tions in non-corrosive ground water, SS and PVC casing may be
the most appropriate choices.  Because  PTFE has been shown
to introduce error into hydrocarbon determinations, it may be
most  applicable under conditions  where SS and PVC are not
considered appropriate. For example, SS would probably not be
considered an appropriate material in corrosive ground water or
where determinations of  trace metal  concentrations  are of
primary concern. Likewise, PVC probably would not be consid-
ered  an appropriate  material in situations where solvents in
moderate to high concentrations might dissolve the PVC  mate-
rial.

Flexible tubing can introduce significant error through sorption
of contaminants onto tubing material, leaching of constituents of
the tubing material into sampled water,  and possibly transmis-
sion of organic compounds and gases through tubing  walls.
These errors are generally greater than for rigid materials and
may be particularly important during site remediation efforts
when declines in ground-water concentrations may be masked
by desorption of previously sorbed compounds.  Laboratory
research has demonstrated the potential for errors under static
conditions, but further research is required to understand how
sorption/desorption mechanisms  can impact samples during
the dynamic sampling process. These studies suggest, how-
ever, that sample error can be minimized by substituting  PTFE
for other types of flexible materials.

Filtration of samples for trace metals determinations may intro-
duce sample error either by the equipment and methods utilized
or by the actual decision to filter. Due to the presence of colloidal
sized particles in ground  water, filtration can have  dramatic
impacts on determinations of the concentrations of both mobile
and total dissolved metals. Indiscriminate filtration of metals
samples may lead to gross errors in these concentrations and
result in erroneous conclusions about ground-watertransport of
metals. In view of this, the objectives of the sampling program
must be carefully considered before samples are filtered. If it is
decided to filter samples, in-line filtration with pro-cleaned, lower
pore-size filters can reduce errors associated with filtration.

In contrast to most aspects of the sampling process,  errors
introduced during  laboratory  analysis  may be relatively well
quantified. Analysis of the CLP database has shown errors in
reported concentrations of performance samples  of -20% to
-30% for volatile and semivolatile compounds and -10% to zero
for inorganic constituents.  Errors in analytical methods, as with
sample transport, sample storage, and  equipment decontami-
nation, can be quantified for individual investigations by analyz-
ing standards and  blind quality evaluation samples. Although
the magnitude of analytical error may be greater than the error
introduced during some aspects of sample collection, analysis
of quality evaluation samples leads to easier identification and
quantification of analytical error.

Errors  associated with other aspects  of site investigations,
including well drilling and  construction,  are more difficult to
identify because true concentrations of hydrochemical constitu-
ents are unknown  in field  investigations. During the drilling
phase of site investigations, hydrogeologic disturbances can be
minimized by utilizing appropriate drilling methods.  Likewise,
drilling-related hydrochemical disturbances can be reduced by
avoiding the use of fluids that might alter ground-water chemis-
try through  ion exchange  reactions or exposure to organic
polymers. Well construction and development methods appro-
priate to the site hydrogeologic conditions are capable of remov-
ing artifactsf rom the drilling process and  improving the hydraulic
efficiency of the well with minimal impact on subsequent
samples. Proper design, installation, and isolation of cement or
bentonite seals reduces the potential for chemical alterations
from these materials. Any of these aspects of drilling and well
construction can lead to large errors if not carefully controlled,
however, the magnitude  of error is directly related to site
conditions and the extent to which methods have been misap-
plied.  Careful consideration and application of methods and
materials during well drilling and construction can significantly
reduce sample error.

Well purging method, purging rate, and the volume purged prior
to sample collection all possess great potential for introducing
significant error when sampling for sensitive constituents. For
example, setting the purging device far below the air-water
interface and using a high purge rate may contaminate samples
by allowing stagnant water to mix with sampled water.  However,
it is  possible to identify these potential sources of  error and
modify purging procedures to minimize  the errors.  Conducting
a preliminary purge test may aid in identification of the depth and
rate that results in the most representative samples, however,
determination of when purging is complete (purge volume) may
be more difficult. Although purge volume can be calculated by
several indirect methods, this volume may not directly correlate
with  the volume of  water required to  provide representative
samples. In particular, stabilization of the values of field chemi-
cal indicator parameters such as temperature, pH, and EC may
not coincide with stabilization of other  hydrochemical param-
eters and constituents. Due to the often complex three-dimen-
sional  distribution of many contaminants, concentrations of
individual constituents may not stabilize  at the same time, or
may never stabilize.  Despite these possibilities, the potential for
sample error can be reduced by choosing indicator parameters
that  are sensitive to the  purging  process and relate to the
constituents of interest.

To reduce error  when sampling for constituents that may be
associated with colloids, or other very sensitive constituents, it
is particularly important to minimize disturbance of the samples
and the sampling environment during the purging and sampling
process. To this end, reducing or eliminating purging, minimiz-
ing purging and sampling  flow rates, and using  dedicated
sampling devices placed within the well intake interval should all
be considered. Becausethis issue remains unresolved, general
recommendations are not possible and it may be necessary to
conduct preliminary purge tests to  determine  how indicator
                                                         19

-------
parameters and concentrations of important constituents vary
with purging rate, volume, method, and distribution of contami-
nants  around the well.  Inadequate determination of these
factors may lead to order-of-magnitude, or more, errors  in
concentration determinations, especially in low-yield wells.

The errors most critical to sampling programs are those that are
difficult or impossible to identify because important conclusions
may be unknowingly based on erroneous or inadequate data.
Well location and design are aspects of sampling that are very
likely to produce undetected errors.  Errors produced  by  well
location are virtually impossible to identify because their magni-
tude is entirely specific to that particular location. The appropri-
ate  placement of a well can mean the  difference between
detection of a contaminant plume or missing it entirely, so the
potential for error is virtually infinite.  Even if a well is located in
the  targeted zone  of contamination or  plume, little can be
deduced  about small-scale hydrogeologic properties or con-
taminant  distribution without a  well-designed monitoring  net-
work that accounts for individual site characteristics and  pro-
gram objectives.

Well design, particularly the depth and interval of the well intake,
can also be a large potential source of undetectable errors. To
delineate the vertical  distribution of contaminants at a single
location, samples must be collected at specific depths, hence,
wells  must be screened over short  intervals and adequately
                   sealed between sampling zones. Dilution and cross-contamina-
                   tion resulting from long-screened wells or poor well seals may
                   produce order-of-magnitude errors in concentrations that far
                   outweigh errors produced in all other aspects of the sampling
                   process.  For example, dilution of samples collected from long-
                   screened remediation wells may mask true contaminant con-
                   centrations, leading to erroneous conclusions about the effec-
                   tiveness of remedial efforts.

                   In conclusion, it can be stated that virtually all aspects of ground-
                   water investigations, from well location to laboratory analysis,
                   have the potential to introduce error into the determinations of
                   concentrations of hydrochemical constituents. General defini-
                   tion of the magnitude of potential  errors is difficult because
                   errors will be influenced by the complex interaction of geologic,
                   hydraulic, and hydrochemical conditions unique to each site, as
                   well as the design and performance of the sampling program.
                   Potential  sources of error  related to site conditions must be
                   identified during early phases of the remedial investigation (Rl)
                   and then minimized  by careful design of the sampling program.
                   Modifications to the  program design may then be necessary to
                   address issues that might arise as the Rl proceeds. Methods of
                   detecting errors that may be introduced during the performance
                   of the sampling prog ram must be utilized sothat these errors can
                   be identified and minimized. However, errors that are difficult or
                   impossible to detect may provide the greatest obstacles to the
                   collection of representative data.
         TABLE 5.  POTENTIAL SOURCES OF ERROR ASSOCIATED WITH ELEMENTS OF GROUND-WATER SAMPLING PROGRAMS
                                               AT HAZARDOUS WASTE SITES.
Program Element   Type of Error
Ability
to Avoid
Error       Methods for Error Avoidance
Ability
to Detect
Error      Methods for Error Detection
Well Intake Length   Long-screened and multi-     Easy to
                  screened wells may lead to    Moderate
                  cross-contamination or
                  contamination dilution.

Well Intake Depth    Well intake may miss zone    Easy to
                  of interest.                 Moderate
Well Intake Design   Presence of particulates      Easy to
                  in samples.                Moderate

Filter Pack          Presence of particulates in    Easy to
                  samples. Reaction with filter  Moderate
                  pack materials or introduced
                  contaminants may alter
                  hydrochemistry. Vertical
                  connection of naturally
                  isolated zones if filter pack
                  too long. Invasion of borehole
                  seal materials if filter pack
                  too short.
           Identify specific zones of interest.
           Use intake length appropriate to
           program objectives and hydrogeologic
           and hydrochemical conditions.

           Identify specific zones of interest.
           Use intake length appropriate to
           program objectives and hydrogeo-
           logic and hydrochemical conditions.

           Design in conjunction with filter
           pack for hydrogeologic conditions.

           Design in conjunction with well
           intake for hydrogeologic conditions.
           Use clean, non-reactive materials.
           Install with tremie pipe and measure
           depths and volumes during installation
           to ensure correct placement.
Difficult    Compare with data from short-
          screen wells or field-screening
          methods.
Difficult    Compare with data from other
          wells or field-screening
          methods.
Easy to    Turbid samples.
Moderate

Easy to    Turbid samples.
Moderate  Sorption/leaching studies of
          materials before installation.
                                                                                                                (Continued)
                                                           20

-------
                                                          TABLES. CONTINUED.
Program Element    Type of Error
                             Ability
                             to Avoid
                             Error
Methods for Error Avoidance
Ability
to Detect
Error       Methods for Error Detection
Borehole Seals
Well Location
Drilling
Well Development
Materials
 If improperly placed,          Moderate
'bentonite materials may alter
 hydrochemistry through ion
 exchange. If improperly
 placed, cement may elevate
 values of ground-water pH,
 EC, alkalinity, calcium
 concentration.

 Inadequate coverage of       Moderate
 area of investigation.
 Depends on method.         Moderate
 Contamination by drilling or
 other fluids may alter
 hydrochemistry. Smearing
 and mixing of fluids and
 sediments at borehole
 wall. Cross-contamination
 within borehole.
 Depends on method.         Easy to
 Incomplete development may  Moderate
 lead to turbid samples or poor
 hydraulic efficiency.  Alteration
 of hydrochemistry by develop-
 ment action. Introduction of
 contaminants (including air
 and water).

 Depends on material,         Easy to
 contaminants, hydrochemical  Moderate
 conditions, and time of contact.
 Sorption/desorptbn of
 chemical constituents.
 Leaching of constituents from
 materials'matrix. Biologic
 activity.  Possible transmission
 through flexible materials.
Design for hydro-geologic conditions.      Moderate
Isolate seals from sampling zone.        to Difficult
Install with tremie pipe and measure
depths and volumes during installation
to ensure correct placement.
Careful design of monitoring well         Difficult
network.
Careful consideration and application     Moderate
of methods that are appropriate for       to Difficult
program objectives and hydrogeologic
and hydrochemical conditions.
Minimize use of water-based drilling
fluids and additives.  If constituents
sensitive to atmospheric exposure will
be sampled, minimize use of air-based
drilling fluids. Determine the chemical
quality of drilling fluids used. Use
appropriate development methods to
minimize impacts of drilling.

Careful consideration and application     Moderate
of methods that are appropriate for
program objectives and hydrogeologic
and hydrochemical conditions. Avoid
adding fluids to well.  If adding fluids is
necessary, determine the chemical
quality of the fluids used.
Select materials that are appropriate      Difficult
for program objectives and hydro-
geologic and hydrochemical conditions.
Use appropriate well purging techniques.
            Bentonite:  High sodium con-
            centrations if sodium bentonite
            used and samples are highly
            contaminated. Cement:
            Sample pH over 10, and high
            EC, alkalinity, and calcium
            concentrations.
           Compare with data from
           nearby wells or field-
           screening methods.

           Drilling fluid contamination:
           Depends on composition of
           fluid.  Compare with data from
           nearby wells and field-
           screening methods. Evaluate
           chemical quality of fluids used.
           Turbid samples and production
           of sedments during pumping
           may indicate incomplete
           development or inadequate
           design of filter pack and well
           intake.  If fluids were added,
           evaluate chemical quality of
           fluids used.

           Sorption/leaching studies of
           materials before installation.
           Detection after installation
           depends on material,
           contaminants, hydrochemical
           conditions, and time of contact.
                                                                                                                                (Continued)
                                                                    21

-------
                                                         TABLES.  CONTINUED.
Program Element   Type of Error
                            Ability
                            to Avoid
                            Error       Methods for Error Avoidance
                                      Ability
                                      to Detect
                                      Error      Methods for Error Detection
Well Purging
Sampling Device
 Sample Collection
 Time and Depth
 Sample Filtration
Incomplete removal of        Easy to
stagnant water (water         Moderate
affected by contact with       (Moderate
atmosphere and well and      to Difficult
sampling device materials),    under
Disturbance of ambient       low-yield
hydrochemical conditions.     conditions)
Depends on operating        Easy
principle of sampling device.
Sorption, desorption, and
leaching from materials.
Degassing or volatilization
from sample. Atmospheric
contamination.

Mixing with stagnant water    Easy
in well.  As time after purging
increases, water in well
becomes more stagnant.
Type of filter system used     Easy to
and length of pre-filtratbn     Moderate
holding time determines
extent of temperature
changes, atmospheric
contamination, degassing,
and sorption onto particulates.
Filter pore size may affect
passage of certain constituents
and suspended material.
Filter material and filter pre-
cleaning may affect results.
Erroneous conclusions about
metals concentrations may
result from association of
metals with colloids.
Choose indicator parameters that are     Easy to
sensitive to purging process and relate   Moderate
to the chemical constituents of interest.   (Moderate
Conduct purge-volume test to determine  to Difficult
when parameters or constituents of      under
interest reach stable values. Determine  low-yield
if low flow-rate and/or low volume        conditions)
purging is appropriate. If not, minimize
volume of stagnant water above device
intake by purging near water surface or
lower device during purging or before
sampling. Avoid drawing water level
below top of well intake.

Select device that is appropriate for      Moderate
sample type, hydrochemical conditions,   to Difficult
and program objectives.
            Conduct purge-volume test to
            determine when parameters or
            constituents of interest reach
            stable values
            Depends on sampler type.
            Compare with data collected
            with other devices.
Collect samples from within or im-
mediately above well intake.  Use
appropriate sampling rate.  Avoid
moving sampler within water column
during sampling. High-yield wells:
Sample immediately after purging.
Low-yield wells: Determine
appropriate time based on response
of well and purge-volume test.

Determine of filtration is necessary
for the objectives of the program.
Minimize pre-filtration holding time.
Use pre-deaned in-line filters. Some
situations may warrant use of pore
sizes other than 0.45|om.
Moderate   Test different scenarios and
to Difficult   compare results, although may
            be very difficult to determine
            which results are the most
            representative.
Moderate   Compare analytical results of
            filtered and unfiltered samples.
            Compare analytical results of
            different filtration methods.
                                                                                                                                (Continued)
                                                                    22

-------
                                                        TABLES. CONTINUED.
Program Element    Type of Error
                           Ability
                           to Avoid
                           Error       Methods for Error Avoidance
                                                Ability
                                                to Detect
                                                Error       Methods for Error Detection
Equipment           Cross-contamination
Decontamination     between wells if sampling
                    equipment is used portably.
                    Incomplete removal of
                    residues from manufacture
                    or contaminants from
                    storage, transport, or use.

Sample Preservation  Changes in hydrochemistry
                    during sample shipment
                    and storage.
                           Easy       Use appropriate cleaning and
                                       decontamination procedures.
Sample Transport
and Storage
Cross-contamination
between sample bottles.
Materials' effects from
sample bottles. Loss of
volatile constituents.
                                                 Easy       Collect rinsate blanks after
                                                            cleaning.
Laboratory Analysis  Deviation from true
                    concentrations.
                           Easy       Use appropriate physical and
                                       chemical preservation procedures.
Easy        Use appropriate sample bottle type
            and cleaning procedure.
            Do not exceed sample holding times.
                           Moderate   Use appropriate analytical methods
                                       and laboratory procedures.
Moderate   Indirectly identified by
to difficult,  evaluating how well
           procedures are being
           followed.

Easy       Transport trip blanks with
           samples.
                                                 Easy to     Analyze blind performance
                                                 Moderate   evaluation samples, blanks,
                                                            and standards.
                                                                  23

-------
REFERENCES

Aleckson.K. A., J.W. Fowler, and Y.Joyce Lee, 1986. Inorganic
analytical methods performance and quality control consider-
ations.  In:  Quality Control in Remedial Site Investigation:
Hazardous and Industrial Solid Waste Testing, Fifth  Volume,
ASTM STP 925, C. L Perket, ed. American Society for Testing
and Materials, pp. 112-123.

Aller,  L, T. W. Bennett, G. Hackett, R. J. Petty, J. H.  Lehr, H.
Sedoris, D. M. Nielsen, and J. E. Denne, 1989. Handbook of
Suggested Practices for the Design and Installation of Ground-
Water Monitoring Wells.   U.S.  Environmental Protection
Agency. EPA/600/4-89/034. 221 p.

Andricevic, R.  and E.  Foufoula-Georgiou, 1991.   A  transfer
function approach to sampling network design for groundwater
contamination.  Water Resources Research. Vol. 27, No. 10.
pp. 2759-2769.

Barcelona, M. J., J. P. Gibb, and R. A. Miller, 1983. A Guide to
the Selection of Materials for Monitoring Well Construction and
Ground-Water Sampling. Illinois State Water Survey Contract
Report No. 327. U.S. Environmental Protection Agency. EPA/
600/52-84/024. 78 p.

Barcelona, M. J., 1984. TOC determinations in ground water.
Ground Water. Vol. 22, No. 1. pp. 18-24.

Barcelona, M. J., J. A. Helf rich, E. E. Garske, and J. P. Gibb,
1984. A laboratory evaluation of ground water sampling mecha-
nisms. Ground Water Monitoring Review. Vol. 4, No. 2. pp. 32-
41.

Barcelona, M. J., J. P. Gibb, J. A. Helfrich, and E. E.  Garske,
1985a.  Practical Guide for Ground-Water Sampling.  Illinois
State Water Survey Contract Report No. 374. 94 p.

Barcelona, M. J., J. A. Helfrich,  and E. E. Garske,  1985b.
Sampling  tubing effects on groundwater samples. Analytical
Chemistry. Vol. 57, No. 2. pp. 460-464.

Barcelona, M. J. and J. A. Helfrich, 1986. Well construction and
purging effects on groundwater samples. Environ. Sci. Technol.
Vol. 20, No. 11. pp. 1179-1184.

Barcelona, M. J., J. A. Helfrich,  and E. E. Garske,  1988a.
Verification of sampling methods and selection of materials for
groundwater contamination studies.  In:  Ground-Water Con-
tamination: Field Methods, Special Technical Publication 963,
A. I. Johnson, ed. American Society for Testing and Materials.
pp. 221-231.

Barcelona, M.  J., G. K. George,  and M. R. Schock, 1988b.
Comparison of Water  Samples from PTFE, PVC,  and SS
Monitoring Wells. U.S. Environmental Protection Agency. EPA/
600/X-88/091.37p.

Barcelona, M. J., H. A. Wehrmann, M. R. Schock, M. E. Sievers,
and J. R. Karny, 1989.  Sampling Frequency for Ground-Water
Quality Monitoring.  U.S. Environmental Protection  Agency.
EPA/600/4-89/032. 191 p.
Barker, J. F., G. C. Patrick, L. Lemon, and G. M. Travis, 1987.
Some biases in sampling multilevel  piezometers for volatile
organics. Ground Water Monitoring Review. Vol. 7, No. 2. pp.
48-54.

Barker, J. F. and R. Dickhout, 1988. An evaluation of some
systems for sampling gas-charged ground water for volatile
organic analysis. Ground Water Monitoring Review. Vol. 8, No.
4. pp. 112-120.

Berens, A. R., 1985. Prediction of organic chemical permeation
through  PVC pipe.  Journal of the American Water Works
Association. Vol. 77, No. 1. pp. 57-64.

Boethner, E. A., G. L.  Ball, Z. Hollingsworth, and R. Aquino,
1981. Organic and Organotin'Compounds Leached from PVC
and CPVC Pipe. U.S. Environmental Protection Agency. EPA/
600/1-81/062. 102 p.

Brobst, R. B. and P. M. Buszka, 1986. The effect of three drilling
fluids on groundwater sample chemistry.  Ground Water Moni-
toring Review. Vol. 6. No. 1. pp. 62-70.

Bryden, G. W., W. R. Mabey, and K. M. Robine, 1986. Sampling
for toxic contaminants in ground water. Ground Water Monitor-
ing Review. Vol. 6, No. 2. pp. 67-72.

Cohen, R. M. and R. R. Rabold, 1987. Numerical evaluation of
monitoring well design.   In:   Proceedings of First  National
Outdoor Action Conference, National Water Well Association,
Dublin, Ohio. pp. 267-283.

Cowgill, U. M., 1988. Sampling waters: The impact of sampling
variability on planning and confidence levels. In:  Principles of
Environmental Sampling, L. H. Keith, ed. American Chemical
Society Professional Reference Book. pp. 171-189.

Curran, C.  M. and M.  B.  Tomson, 1983.  Leaching of trace
organics into water from five common plastics. Ground Water
Monitoring Review.  Vol. 3, No. 3. pp. 68-71

Driscoll, F. G., 1986. Ground Water and Wells, 2nd Ed. Johnson
Division.  St. Paul, Minnesota. 1108 p.

Fetter, C. W., 1983.   Potential sources of contamination in
ground-water monitoring.  Ground Water Monitoring Review.
Vol. 3, No. 2. pp. 60-64.

Flotard, R. D., M. T. Homsher, J. S. Wolff, and J. M. Moore, 1986.
Volatile organic analytical methods performance and quality
control considerations.  In:  Quality Control in Remedial Site
Investigation: Hazardous and Industrial Solid Waste Testing,
Fifth Volume, ASTM  STP 925, C. L. Perket, ed. American
Society for Testing and Materials, pp. 185-197.

Friedman, L.  C., L J.  Schroder,  and  M. G. Brooks, 1986.
Recovery of several volatile organic compoundsf rom simulated
water samples:  Effect of transport and storage. Environ. Sci.
Technol. Vol. 20, No. 8. pp. 826-829.
                                                       24

-------
Gardner, M. J. and D. T. Hunt, 1981. Adsorption of trace metals
during filtration of potable water samples with particular refer-
ence  to the determination of filterable lead concentrations.
Analyst. Vol. 106. pp. 471-474.

Gibb, J. P., R. M. Schuller, and R. A. Griffin, 1981. Procedures
for the Collection of Representative Water Quality Data from
Monitoring Wells. Cooperative Groundwater Report #7, State of
Illinois, Department  of  Energy  and  Natural Resources,
Champaign, Illinois. 61 p.

Gibs, J. and T. E.  Imbrigiotta, 1990.  Wellpurging criteria for
sampling purgeable organic compounds. Ground Water. Vol.
28, No. 1. pp. 68-78.

Giddings, T., 1983.  Bore-volume purging to improve monitoring
well performance: an often mandated myth. In: Proceedings of
the Third National Symposium on Aquifer Restoration and
Ground Water Monitoring, National Water Well Association,
Columbus, Ohio. pp. 253-256.

Giddings, T., 1986.  Screen length selection for use in detection
monitoring  well networks.  In:  Proceedings of Sixth National
Symposium and Exposition on Aquifer Restoration and Ground
Water Monitoring,  National Water Well Association, Dublin,
Ohio. pp. 316-319.

Gillham, R. W., M.  J. L Robin, J. F. Barker, and J. A. Cherry,
1983. -Groundwater Monitoring and Sample Bias. API Publica-
tion No. 4367, American Petroleum Institute. 206 p.

Gillham, R. W. and S. F. O'Hannesin, 1990. Sorption of aromatic
hydrocarbons by materials used in construction of ground-water
sampling wells. In:  Ground Waterand VadoseZone Monitoring.
D. M.  Nielsen and A.I.  Johnson, eds.  ASTM STP  1053,
American Society for Testing and Materials, Philadelphia, Penn-
sylvania, pp. 108-122.

Herzog, B. L, S. J. Chou, J. R. Valkenburg,  and R. A. Griffin,
1988. Changes in volatile organic chemical concentrations after
purging slowly  recovering wells.   Ground Water Monitoring
Review.  Vol. 8, No. 4. pp. 93-99.

Herzog, B., J. Pennine, and G. Nielsen, 1991. Ground-water
sampling. In: Practical Handbook of Ground-Water Monitoring.
D. M. Nielsen, ed.  Lewis Publishers.  Chelsea, Michigan,  pp.
449-499.

Ho, J. S., 1983. Effect of sampling variables on recovery of
volatile organics in water. Journal of the American Waterworks
Association. Vol. 12. pp. 583-586.

Holm, T. R., G. K. George, and M. J. Barcelona, 1988. Oxygen
transfer through flexible tubing and its effects on ground water
sampling results. Ground Water Monitoring Review. Vol. 8, No.
3. pp. 83-89.

Houghton, R. L. and M. E. Berger, 1984. Effects of well-casing
composition and sampling method  on  apparent quality of
ground water. In: Proceedings of The Fourth National Sympo-
sium and Exposition on Aquifer Restoration and Ground Water
Monitoring, National Water Well Association, Dublin, Ohio.  pp.
203-213.
Imbrigiotta, T. E., J. Gibs, J. F. Pankow, and M. E. Rosen, 1987.
Field comparison of downhole and surface sampling devices for
purgeable organic compounds in ground water.  Abstract in:
U.S. Geological Survey Program on Toxic Waste-Ground-Wa-
ter Contamination: Proceedings of the Third Technical Meeting,
Pensacola, Florida.  B. J. Franks, ed. U.S. Geological Survey.
Open-File Report 87-109.  pp. E5-E6.

Imbrigiotta, T. E., J. Gibs, T. V.  Fusillo,  G. R. Kish, and J. J.
Hochreiter, 1988.  Field evaluation of seven sampling devices
for purgeable organic compounds in ground-water. In: Ground-
Water Contamination: Field Methods, A. I. Johnson, ed. Special
Technical Publication 963, American Society for Testing and
Materials, pp. 258-273.

Jay, P. C.( 1985. Anion contamination of environmental water
samples introduced by filter media. Analytical Chemistry.  Vol.
57, No. 3. pp. 780-782.

Kaleris, V., 1989. Inflow into monitoring wells with long screens.
In:  Contaminant Transport in Groundwater. H.  E. Kobust and
W. Kinzelbach, eds.  A. A. Balkema, Rotterdam, pp. 41-50.

Kearl, P.  M., N. E. Korte, and T. A. Cronk, 1992.  Suggested
modifications to ground water sampling procedures based on
observations  from the colloidal  horoscope.   Ground  Water
Monitoring Review.  Vol. 12, No. 2. pp. 155-161.

Keely, J. F. and K. Boateng, 1987. Monitoring well installation,
purging, and sampling techniques - part 1: conceptualizations.
Ground Water. Vol.  25, No. 4. pp. 427-439.

Keith, S. J., L. G. Wilson, H. R. Fitch, and D. M. Esposito, 1983.
Sources of spatial-temporal variability  in  ground-water quality
data and methods of control. Ground Water Monitoring Review.
Vol. 3, No. 2. pp. 21-32.

Kennedy, V. C., G. W. Zellweger, and B. F. Jones, 1974. Filter
pore-size effects on the analysis of Al,  Fe, Mn and Ti in water.
Water Resources Research. Vol. 10, No. 4. pp. 785-790.

Korte, N. and P. Kearl, 1985.  Procedures  for the Collection and
Preservation of Groundwater and Surface Water Samples and
for the Installation of  Monitoring Wells.  GJ/TMC-08, U.S.
Department of Energy. 68  p.

Kraemer, C. A., J. A. Shultz, and J. W. Ashley, 1991.  Monitoring
well post-installation considerations. In: Practical Handbook of
Ground-Water Monitoring.  D. M. Nielsen,  ed. Lewis Publishers.
Chelsea,  Michigan, pp. 333-365.

Levine, S. P., M. A. Puskar, P. O. Dymerski, B. J. Warner, and
C. S. Friedman, 1983.  Cross-contamination of water samples
taken for  analysis of purgeable organic compounds. Environ.
Sci. Technol.  Vol. 17, No. 2. pp.  125-127.

Lewis, D. L., 1988. Assessing and controlling sample contami-
nation. In:  Principles of Environmental Sampling, L.H.Keith,
ed. American Chemical Society Professional Reference Book.
                                                       25

-------
Liikala, T. L, D. S. Daly, and A. P- Toste, 1988. An Evaluation
of the Effects of Well Construction Materials and Groundwater
Sampling Equipment on Concentrations of Volatile Organic
Compounds. PNL-6585 UC-11. Pacific Northwest Laboratory.
25 p.

Maskarinec, M. P. and  R. L  Moody,  1988.  Storage and
preservation of environmental  samples.  In:  Principles of
Environmental Sampling, L. H. Keith, ed. American  Chemical
Society Professional Reference Book. pp. 145-155.

Matteoli, R. J.  and J. M. Noonan, 1987. Decontamination of
rubberhoseandteflontubingforgroundsampling. In: Proceed-
ings from First National Outdoor Action Conference, National
Water Well Association, Dublin, Ohio. pp. 159-183.

McAlary, T. A.  and J. F. Barker, 1987. Volatilization losses of
organics during ground water sampling from low permeability
materials.  Ground Water Monitoring Review. Vol. 7,  No. 4. pp.
63-68.

Mcllvride, W. A. and R. B. Rector, 1988.  Comparison of short-
and long-screen  monitoring wells in alluvial sediments.   In:
Proceedings of Second National Outdoor Action Conference,
National Water Well Association, Dublin, Ohio. pp. 277-287.

Mcllvride, W. A. and R. B. Weiss, 1988. Drilling method effects
on well performance. In:  Proceedings from Second National
Outdoor Action Conference, National Water Well Association,
Dublin, Ohio. pp. 277-287.

Meyer, P. D. and E. D. Brill, Jr., 1988. A method for locating wells
in a groundwater monitoring network under conditions of uncer-
tainty. Water Resources Research.  Vol. 24, No. 8.  pp. 1277-
1282.

Mickham, J. T., R. Bellandi, and E. C. Tiff, 1989.  Equipment
contamination  procedures for ground water and vadose zone
monitoring  programs: Status and prospects.  Ground Water
Monitoring Review.  Vol. 19, No. 2. pp. 100-121.

Miller, G. D., 1982, Uptake and release of lead, chromium, and
trace  level volatile organics exposed to synthetic well casings.
In:  Proceedings of the Second National Symposium on Aquifer
Restoration and Ground Water Monitoring, National Water Well
Association, Dublin, Ohio. pp. 236-245.

Moody, J. R. and R. M. Lindstrom, 1977. Selection and Cleaning
of Plastic Containers for Storage. Analytical Chemistry. Vol.49,
No. 14. pp. 2264-2267.

Morin, R. H., D. R. LeBlanc, W. E. Teasdale, 1988. A statistical
evaluation of formation disturbance produced by well-casing
installation methods.  Ground Water. Vol. 26, No. 2. pp. 207-
217.

Muska, C. F., W.  P. Colven, V. D. Jones, J. T. Scoglin, B. B.
Looney, and V. Price, Jr., 1986. Field evaluation of ground water
sampling devices for volatile organic compounds. In: Proceed-
ings of the Sixth National Symposium and Exposition on Aquifer
Restoration and Ground Water Monitoring, National Water Well
Association, Dublin, Ohio. p. 235-246.
Nielsen, D. M. and G. L. Yeates, 1985.  A comparison of
sampling mechanisms available f or smalldiameter ground water
monitoring wells.  Ground Water Monitoring Review. Vol.5. No.
2. pp. 83-99.

Nielsen, D. M. and R. Schalla, 1991. Design and installation of
ground-water monitoring wells.  In:  Practical Handbook of
Ground-Water Monitoring. D. M. Nielsen, ed. Lewis Publishers.
Chelsea, Michigan, pp. 239-331.

Pankow, J. F., 1986. Magnitude of artifacts caused by bubbles
and headspace in the determination of volatile organic com-
pounds in chemistry. Analytical Chemistry. Vol. 58. pp. 1822-
1826.

Papadopulos, I. S. and H. H. Cooper, Jr., 1967. Drawdown in a
well of large diameter. Water Resources Research. Vol. 3. pp.
241-244.

Parker, L. V., A. D. Hewitt, and T. F. Jenkins, 1990. Influence of
casing materials on trace-level chemicals in well water. Ground
Water Monitoring Review. Vol. 10, No. 2. pp. 146-156.

Parr,  J., M.  Bellinger, O. Callaway,  and  K.  Carlberg,  1988.
Preservation techniques for organic and inorganic compounds
in water samples. In: Principles of Environmental Sampling, L.
H. Keith, ed. American Chemical Society Professional Refer-
ence Book, pp?221 -230.

Paul, D. G., C. D. Palmer, and D. S. Cherkauer, 1988. The effect
of construction, installation, and development on the turbidity of
water in monitoring wells in fine-grained glacial till.  Ground
Water Monitoring Review. Vol. 8, No. 1. pp. 73-82.

Pearsall, K. A. and A. V. Eckhardt, 1987.  Effects of selected
sampling equipment and procedures on the concentrations of
trichlorethylene and related  compounds in ground  water
samples. Ground Water Monitoring Review. Vol. 7, No. 2. pp.
64-73.

Pionke, H. B. and J. B. Urban, 1987. Sampling the chemistry of
shallow aquifer systems - a case study. Ground Water Monitor-
ing Review.  Vol.  7, No. 2. pp. 79-88.

Pohlmann, K. F. and J. W. Hess, 1988. Generalized Ground
Water Sampling  Device Matrix.   Ground Water  Monitoring
Review. Vol. 8, No. 4.

Pohlmann, K. F., R. P. Blegen, and J. W. Hess, 1990.  Field
Comparison of Ground-Water Sampling Devices for Hazardous
Waste Sites:  An  Evaluation Using Volatile Organic Com-
pounds. U.S. Environmental Protection Agency. EPA/600/4-
90/028. 102 p.

Puls, R. W. and M. J. Barcelona, 1989. Ground Water Sampling
for Metals Analyses.  U.S. Environmental Protection Agency.
Superfund Issue Paper. EPA/540/4-89/001. 6 p.

Puls, R. W., J. H. Eychaner, and R. M. Powell, 1990. Colloidal-
Facilitated Transport of  Inorganic Contaminants  in Ground
Water: Part I. Sampling Considerations.  U.S. Environmental
Protection Agency. Environmental Research Brief. EPA/600/
M-90/023. 12 p.
                                                       26

-------
Puls, R. W., R. M. Powell, D. A. Clark, and C. J. Paul, 1991.
Facilitated Transport of Inorganic Contaminants  in Ground
Water:  Part II.   Colloidal  Transport.  U.S. Environmental
Protection Agency. Environmental Research Brief. EPA/600/
M-91/040. 12 p.

Reilly, T. E., O. L Franke, and G. D. Bennett, 1989.  Bias in
groundwater samples caused by wellbore flow.  Journal of
Hydraulic Engineering.  Vol.  115, No. 2. pp. 270-277.

Reynolds, G. W., J. T. Hpff, and R. W. Gillham, 1990. Sampling
bias caused by  materials used  to monitor haiocarbons in
groundwater. Environ. Sci. Technol. Vol. 24, No. 1. pp 135-
142.

Rice,  G.,  J. Brinkman, and  D.  Muller, 1988.   Reliability of
chemical analyses of water samples - the experience of the
UMTRA project. Ground Water Monitoring Review. Vol. 8, No.
3. pp. 71-75.

Robin, M. J. L. and R. W. Gillham, 1987.  Field evaluation of well
purging procedures.  Ground Water Monitoring Review.  Vol.7,
No. 4. pp. 85-93.

Ross, H. B., 1986. The importance of reducing sample contami-
nation in routine monitoring of trace  metals in atmospheric
precipitation.  Atmospheric Environment. Vol. 20, No. 2. pp.
401-405.

Scalf, M. R., J. F. McNabb, W. J. Dunlap, and R. L. Crosby, 1981.
Manual of Ground-Water Quality Sampling Procedures.  Na-
tional Water Well Association, Dublin, Ohio. 93 p.

Schalla, R., D. A. Myers, M. A. Simmons, J. M. Thomas, and A.
P. Toste, 1988. The  sensitivity of four monitoring well sampling
systems  to  low  concentrations  of three volatile organics.
Ground Water Monitoring Review.  Vol. 8, No. 3. pp. 90-96.

Scheibe, T. D. and D.  P  Lettenmaier, 1989.   Risk-based
selection of monitoring wells for assessing agricultural chemical
contamination  of ground water.   Ground Water Monitoring
Review. Vol. 9, No.  4. pp. 98-108.

Schuller, R.  M.. J. P. Gibb,  and R. A. Griffin, 1981.  Recom-
mended sampling procedures for monitoring wells.  Ground
Water Monitoring Review. Vol. 1, No. 1. pp. 42-46.

Smith, R. L, R. W. Harvey, J. H. Duff, and D. R. LeBlanc, 1987.
Importance of  close-interval vertical sampling  in delineating
chemical  and microbiological gradients  in ground  water
Ssudies.  In:  Toxic  Waste-Ground-Water Contamination Pro-
gram  Third Technical Meeting, OFR87-109, U.S. Geological
Survey, pp. 33-35.

Smith, F., S. Kulkarni, L. E.  Myers, and M. J. Messner, 1988.
Evaluating and presenting quality assurance sampling data. In:
Principles of Environmental Sampling, L. H. Keith, ed. American
Chemical Society Professional Reference Book.

Sosebee, J. B., P. C. Geiszler,  D. L Winegardner, and C. R.
Fisher, 1982. Contamination of groundwater samples with poly
(vinyl chloride) adhesives and poly (vinyl chloride) primer from
monitoring wells.  In:  Hazardous and  Industrial Solid Waste
Testing:   Second Symposium, January 28-29, 1982, R. A.
Conway and W. P. Gulledge, eds. ASTM Technical Publication
805. pp. 38-50.

Spruill, T. B. and L. Candela, 1990. Two approaches to design
of monitoring networks. Ground Water. Vol. 28, No. 3. pp. 430-
442.

Stolzenburg, T. R. and D. G. Nichols, 1985. Preliminary Results
on Chemical Changes in Groundwater Samples Due to Sam-
pling Devices.  EPRI EA-4118, Electric Power Research Insti-
tute. 84 p.

Stolzenburg, T. R. and D. G. Nichols, 1986. Effects of filtration
method  and sampling  devices  on  inorganic chemistry  of
sampled  well water.  In:  Proceedings of the Sixth National
Symposium and Exposition on Aquifer Restoration and Ground-
Water Monitoring, National Water  Well  Association, Dublin,
Ohio.  pp. 216-234.

Sykes, A. L., R. A. McAllister, and J. B. Homolya, 1986. Sorption
of organics by monitoring well construction materials.  Ground
Water Monitoring Review.  Vol. 6, No. 4. pp. 44-47.

Truitt, R. E. and J. H. Weber, 1979.  Trace metal ion filtration
losses at pH 5 and 7. Analytical Chemistry. Vol. 51, No. 12. pp.
2057-2059.

Unwin.J.P., 1982. AGuideto Ground-Water Sampling. Stream
Improvement Technical Bulletin No. 362.  National Council of
the Paper Industry for Air and Stream Improvement, Inc. 124 p.

Unwin, J. P. andD. Huis, 1983. A laboratory investigation of the
purging behavior of small-diameter monitoring wells. In:  Pro-
ceedings of the Third National Symposium on Aquifer Restora-
tion and Ground Water Monitoring, National Water Well Asso-
ciation, Columbus, Ohio. pp. 257-262.

Unwin, J. P, 1984. Sampling ground water for volatile organic
compounds: The effects of sampling method, compound vola-
tility and concentration. In: Proceedings of the Fourth National
Symposium and Exposition on Aquifer Restoration and Ground
Water Monitoring, National Water  Well  Association, Dublin,
Ohio.  pp. 214-220.

Unwin, J. P. and V. Maltby, 1988. Investigations of techniques
for purging ground-water monitoring wells and sampling ground
water for volatile  organic  compounds.   In:   Ground-Water
Contamination:  Field Methods, Special Technical Publication
963, A.  I. Johnson, ed. American  Society for Testing and
Materials, pp. 240-252.

U.S. EPA. 1986.  RCRA Ground-Water Monitoring Technical
Enforcement Guidance Document.  OSWER-9950.1, U.S. En-
vironmental Protection Agency,  Office of Solid  Waste and
Emergency Response. 208 p.

U.S. EPA, 1987. Handbook: Groundwater. EPA/625/6-87/016
212 p.

Vitale, R. J., O. Braids, and R. Schuller,  1991. Ground-water
sample analysis.  In:  Practical  Handbook of Ground-Water
Monitoring. D. M.  Nielsen, ed. Lewis Publishers.  Chelsea
Michigan, pp. 501-539.
                                                       27
                                                                    •&U.S. GOVERNMENT PRINTING OFFICE: 1992 - 7SO-OOZ/60IOI

-------
Wagemann, R. andG. J. Brunskill, 1975. The effect of filter pore   Waste Testing, Fifth Volume, ASTM STP 925, C.L. Perket, ed.
size on analytical concentrations of some trace elements in   American Society for Testing and Materials, pp. 157-171.
filtrates of natural water. IntemationalJoumal of Environmental
Analytical Chemistry. Vol. 4, No. 1. pp. 75-84.                Yeskis, D., K. Chiu, S. Meyers, J. Weiss, and T. Bloom, 1988. A
                                                        field study of various sampling devices and their effects on
Wolff, J. S., M. T. Homsher, R. D. Flotard, and J. G. Pearson,   volatile organic contaminants.  In:  Proceedings of Second
1986. Semi-volatile organic analytical methods performance   National Outdoor Action Conference, National Water Well Asso-
and quality control considerations.   In:   Quality  Control in   ciation, Dublin, Ohio. pp. 471-479.
Remedial Site Investigation: Hazardous and Industrial Solid
 United States
 Environmental Protection Agency                                                               EPA
 Center for Environmental Research Information                                           PERMIT No G-35
 Cincinnati, OH 45268
      BULK RATE
POSTAGE & FEES PAID
 Official Business
 Penalty for Private Use
 $300

 EPA/540/S-92/019

-------