SAB-EEC-86-007
REPORT
on the review of the
"RCRA GROUND-WATER MONITORING
TECHNICAL ENFORCEMENT GUIDANCE DOCUMENT"
by the
Environmental Engineering Caranittee
Science Advisory Board
U. S. Environmental Protection Agency
June, 1986
-------
-2-
EPA NOTICE
This report has been prepared as a part of the activities of the Science
Advisory Board, a public advisory qroup providing extramural scientific
information and advice bo the Administrator and other officials of the
Environmental Protection Aqency. The Board is structured to provide a
balanced expert assessment of scientific matters related to problems
facing the Aqency. This report has not been reviewed for approval by
the Aqency, and hence the contents of this report do not necessarily
represent the views and policies of the Environmental Protection Agency,
nor does mention of tr^de names or cormercial products constitute en-
dorsement or recommendation for use.
-------
-3-
TABLE OF CONTENTS
I. Principal Findinas and Recommendations 4
II. Introduction
Background " 8
Committee Review Procedure 8
III. Review of the Technical Enforcement Guidance Document
Chapter 1 - Characterization of Site Hydrogeology 10
Chapter 2 - Placement of Detection Monitoring Wells - 15
Chapter 3 - Monitorina Well Design and Construction 17
Chapter 4 - Sanpling and Analysis 19
Chapter 5 - Statistical Analysis of Detection Monitoring Data .... 22
Chapter 6 - Assessment Monitoring 22
IV. Append ices
A. List of Committee Members
B. OWPE Memo Outlining Issues for Review
C. List of Public Conmenters
-------
-4-
SECTION I
PRINCIPAL FINDINGS AND RECOMMENDATIONS
The Science Advisory Board was asked by Mr. Gene A. Lucero, Director,
Office of Waste Programs Enforcement (CWPE), on Auqust 26, 1985, to review a
draft document entitled "RCRA Ground Water Monitorina Technical Enforcement
Guidance Document" (TEGD), which had been prepared by his staff. The document
concerned the technical aspects of ground water monitorina at Resource Conser-
vation and Recovery Act (PCRA) facilities. A Subcommittee of the Environmental
Engineering Committee of the Science Advisory Board (SAB) was established to
conduct the review, which has new been completed, except for review of Chapter
5 (Statistical Analysis of Detection Monitoring Data), which will be completed
at a later date. The Committee appreciates the opnortunity to review the
document, and has concentrated on a number of technical issues hiqhlighted
for review by CWPE (see Appendix B), as well as some other issues raised by
the Committee themselves.
A summary of the Coirm it tee's principal findings and recoomendations
follows. More detailed comments will be found in Section III.
General
A. The Ccnmittee recognizes that the document is the result of extensive
technical thought ana review of several previous drafts, and is a
significant improvement over previous versions. However, same techni-
cal amendments remain to be made.
B. The Ccnmittee reconvenes that the TEGD be much more explicit in
stating that it is a guidance document only, ana recuires informed
judgement in its application and use.
In the public testimony which was a part of the Ccrmittee's review pro-
cess, many individuals expressed the concern that the document would be used,
particularly by persons with little or no experience in the design and opera-
tion of monitoring systems, to set specific requirements, such as number
and location of monitorinq wells, well materials and screen lengths, where
such reouirements could not be justified by the physical situation. It rrust
be made very clear that the TEGD reguires informed judgement in its applica-
tion and use. This report proposes charges that should substantially reduce
the likelihood of these kinds of problems.
C. Several examples in the TEGD recuire improvement if they are to
serve the purpose intended.
Such graphic exanples as those dealing with well screen length,
placement of upgradient monitoring wells and well desiqn and construction
show details that may be inappropriate or misleading.
-------
-D-
Characterization of Site Hvdrooeology
D. The Carmittee recommends that the procedures specified for the
design of detection monitoring systems be made more efficient,
and that substantially rore flexibility be encouraged in addres-
sing the primary obiective, tnat of determining the direction and
magnitude of flow of potential pollutants.
In non-comolex geology, the cwner/ooerator would obtain most of the infor-
mation needed to design a detection nonitoring system using the procedures
specified in the TEGD. Many ou ice lines -owever, such as well spacings,
continuous samnlinq and the recruiremenc for full Appendix VIII constituent
analyses of potentially contaminated core samples should he re'-evaluated in
the light of increasing both the efficiency and cost-effectiveness of detection
monitoring.
E. In addition to making the characterization and evaluation process
more efficient, EPA should also elaborate and improve the discussion
of a number of important: factors which are inadeguatelv addressed
in the TEGD.
These factors include accuracy recuirements for location surveys, bore-
hole sealing, characterization of the underlying confining layer, the defini-
tion of "oualified geologist" and boring depth. Detailed discussions of
these may be found in Section III.
F. A number of terms used m the TEGD need to be redefined to make
them more specific, consistent with oenerallv-acceoted practice
and consistent with the omective of protecting usable water sup-
plies.
Definitions of terms such as oedrock, aouifer, uppermost aquifer, water
table and hydraulic interconnection are not consistent with standard defini-
t ions.
Placement of Detection Monitorinc Wells
G. The entire discussion in the TEGD related to detection well spacing
should be revised to better reflect the purpose of the monitoring.
There should be a clearer distinction drawn between detection monitoring
systems and assessment monitor inc systems. Arbitrary well spacings should
not be spec if ied, but ra the r should be determined on the basis of site
hydrogeoiogical characteristics (as previously determined) and the requirement
to determine the maqnitude and direction of ground water flow. See Section
III for one approach.
.Monitoring Well Design and Construction
H. Guidance on well design and construction should further emphasize
methods which minimize disturbance of the ground-water svstem and
are aooroorlate for the ^^roceolocic and chemical conditions.
-------
-6-
Dnllina; Emphasis should be placed upon the selection of drillinq
methods which (1) minimize disturbance of the Geologic and hydrcaeoloqic
system and (2) present the least potential for introducing contamination.
Materials: Flourocarbon resins (such as Teflon, PFTE, FEP or PFA) or
stainless steel 304 or 316 should be specified for use in the saturated zone
when potentially sorbing organics are to be determined. In such cases, and
where high potential for corrosion exists or is anticipated, fluorocarbon
resins are preferable to stainless steel. FVC has utility when shown not to
leach or absorb contaminants significantly.
Well Development: Well development and periodic redevelopment effort
should be emphasized in addition to a performance standard.
I. EPA should allow substantially greater flexibility in the recommended
length of well screens.
The maximum limitation of 10 feet on well screen length, particularly
for detection monitoring where the primary purpose is to identify the presence
of pollutants of concern, may not adeauately accomodate all hydroaeoloqic
situations which may be encountered in the field, and may, in fact, lead to
the collection of misleading data. If the screen length is suitable aiven
the hydroaeologic complexity, it should also be sufficient for water guality
sampling.
Sampling and Analysis
J. Sampling protocols in the draft TEGD are substantially acceptable,
but should be further tailored to hydrooeologic conditions and to
the maintenance of sample integrity.
Pureinq: Purging reguirements should be calculated based on an evalua-
tion of tne hydraulic performance of the well, and'confirmed by pH, conductivity
and temperature measurements.
Collection Equipment: Integrated sample withdrawal and purging equip-
ment which minimizes sample disturbance is recommended. See Section III of
this report for further details.
Statistical Analysis of Detection Monitoring Data
K. The Ajencv should use the AR t-test proposed in the draft TEGD in
the short term, put must also acknowledge that there are many situ-
ations encountered in actual practice where the method may not yield
accurate results.
The t-distribution model does not apply to the situation where sampled
concentrations are drawn from a set of uporadient wells with different means,
and then averaged. If the aquifer is not homogeneous with respect to concen-
tration distributions (such as when stratification is present), the average of
sairples from these wells probably will not follow the Student's t-distribution.
The same is true of a set of wells that have seasonal variations.
-------
-7-
L. The Agency should institute a vigorous program of statistical analy-
sis of data collected as the program proceeds to conf inn the adequacy
or inadeguacy of the method proposed in the draft TEGD.
This analysis will allow the Agency to determine how nuch the distributions
of the data deviate from the assumed Students-t, and to evaluate the implica-
tions of such deviations in temis of the robustness of the detection procedure.
M. The Agency should establish a group to devise a statistical test(s)
that will satisfy regulatory needs and will, at the same time, be
technically defensible over the wide range of situations encountered
in actual practice.
The qroup should keep in mind that the goal of the statistical analysis is
to detect leaks at RCRA facilities. Any test should be justified by reference
to site specific factors, and these tests should be based on preselected values
of Type I error (false positives), Type II error (false negatives), and the
magnitude of the difference which defines the event that it is irrportanfc to
detect (the TEGD simply specifies that it must be "larger then background").
Assessment Monitoring
N. Greater emphasis should be placed on using a phased approach on
required assessment wells, based on informed judgement about incom-
ing sample analysis and hydrogeolooic conditions, and local water
use patterns.
Phased placement of additional wells should be utilized to .'.atially
define the composite contamination plume. It is most important to define
the spatial bounds of total contamination, havinq cataloged individual
chemical constituents.
0. The specifications for the initial number of cluster wells are ex-
cessive. This Section should be rewritten to describe a general
approach that contains the flexiblity to fit site-specific conditions.
The existing guidance recommendations call for seven well clusters with
five wells per cluster. Such a configuration may or may not fit any given
site. This degree of specificity in the guidance should not be errployed.
p. The limitations of mathematical modeling to predict contaminant
transport should be specified in terms of site hydrogeolooic
conditions.
The TEGD states, without reservation, that models can be used to predict
hydraulic head and contaminant concentration at any point. A discussion
should be added on model limitations, practical uses, and calibration needs.
-------
-8-
SECTION II
IVTRODLJCrriON
Background
On August 26, 1985, Mr. Gene A. Lucero, Director, Office of Waste Programs
Enforcement, requested that the Science Advisory Board review a draft docu-
ment, prepared by his staff, entitled "RCRA Ground Water Monitoring Technical
Enforcement Guidance Document (TEGD)." In a subsecruent memorandum (undated),
Mr. Lucero furnished the document, together with related material, to the
Committee, and provided a detailed, chaoter-by^chapter list of issues on which
he reouested the SAB's review and advice. The SAB was further asked to con-
centrate on specific technical issues in the current version of the document,
and not on whether or not the Aaency had adopted the "best" regulatory an-
proach.
The TEGD is intended to provide guidance on how to evaluate the design
and operation of RCRA interim-status ground water monitorina systems. The
intended audience includes facilitv cwner/operators, permit writers, field
inspectors, attorneys, and enforcement officials (engineers, hydrogeoloq ists,
statisticians), both within EPA and also in State organizations responsible
for administering RCRA-related proarams. The document nrovides guidance on
the evaluation of:
A. Characterization of site hydrogeology (ChaDter 1)
B. Placement of detection monitorina wells (Chapter 2)
C. Monitoring well desian and construction (Chapter 3)
D. Samplina and analysis plans (ChaDter 4)
E. Statistical analysis of detection monitoring data (Chapter 5)
F. Assessment monitoring plans (Chapter 6)
AnDendices to the TEGD also provide details of the statistical methodology,
together with some example apDlications, and descriptions of selected
aeoDhysical methods for sairple analysis.
The Connittse recognizes that the guidance was very difficult to write,
and that it was impossible to describe every hydroaeologic/contamination situ-
ation which miqht face an enforcement official in view of the complexities
typical at specific sites. The document attempts to balance the need for
specific, detailed guidance with the reality that the compliance decision-
making process in the ground water area is enormously complicated.
.Comiittee Review Procedures
A Subcomnittee of the Environmental Engineering Committee was formed to
conduct the review. The Subcommittee consisted of nine individuals (see Roster,
Appendix A). Two, Dr." Haun and Mr. Conway, were members of the Environmental
Engineering CcoTnittee, anr the rest were consultants specifically selected
-------
-9-
for the Subcommittee based on their expertise in the area of qround water moni-
toring, statistics, and environmental law. Mr. Conway and Dr. Cartwright were,
in addition, members of an SAB Committee which recently reviewed the Aaency's
ground water research program. Subcommittee selections were based on recorrmen-
dations from SAB staff, the Office of Waste Programs Enforcement, and outside
experts in the field.
The Subcomnittee held one open public meeting on Octooer 3-4, 1985, in
Washington/ D.C. At that meetina., the Subcommittee was briefed on the contents
of the TEGD by Agency staff, and heard testimony on the proposed guidance from
a number of representatives of the waste management industry, state and local
governments, consulting firms, and environmental groups. A list of corrmenters
is included as Appendix C. The Subcommittee was also furnished extensive writ-
ten testimony by both individuals testifying before the Subconnuttee and others
who were not able to attend the meetina. This written testimony is on file
in the offices of the Science Advisory Board.
Other meetings of the Subcomnittee, and subgroups thereof, were held on
November 14-15, 1985 and December 4, 1985 to draft the final report. The
completed report was submitted for approval to the the full Environmental
Engineering Carmittee at its regular meetina on February 13-14, 1986. At-
that meetina, the Environmental Engineering Comnictee made a number of minor
changes in tne report, and also decided that the portions of the report deal-
ing with statistical methods (Chapter 5 and Appendix B of the TEGD) needed
further review. A Subcomnittee, consisting of Dr. J. William Haun, Chairman,
Dr. Charles 0,Melia, Dr. Mitchell Small, Dr. Carl Silver and Dr. Charles
Norwood, was appointed to conduct this review, and their report is attached
as Appendix D.
While portions of this report were drafted by Mr. Harry C. Torno, Execu-
tive Secretary to tne Environmental Enaineerina Committee, based on Subcom-
mittee input, the report in its final form has been approved by, and represents
the views of, the Environmental Engineering Carmittee as a whole.
-------
-10-
SECTION III
REVIEW OF THE TECHNICAL ENFORCEMENT GUIDANCE DOCUMENT
This review provides detailed ccrnments on a chapter-by-chapter basis, and
supplements the more general ccrnments and recommendations outlined in
Section I.
Chapter 1 - Characterization of Site Hydroaeology
A. Optimum well spacing should be approached in a rational and efficient
manner, rather than by grid sampling. A logical approach could be:
1. Prior to any drilling, regional geology and hvdrogeology must be
reviewed and a conceptual (three dimensional, physical) model of site hydro-
geology devised.
2. Drilling then becomes an iterative process of seeking boring
characteristics to confirm the model. The model is revised as each new bore-
hole/well is drilled and further drilling is undertaken to test the revised
model.
3- As the conceptual model becomes more refined, data frati explora-
tory borings (hyarogeology) should begin to approximate model predictions.
The site can be considered "characterized" at such a time as the geologic
materials, ground water level, and ground water flew direction (in the differ-
ent geologic units), can be accurately predicted before drilling.
It is critical that the quality of all hycirogeologic and chemical results
be knewn and documented. The beginning of the QA/QC (quality assurance/quality
control) program for the monitoring effort should be in the hyarogeologic
characterization. The descriptions of the geologic materials, water levels,
hydraulic conductivities and ground water gradients should conform to high
professional reporting standards. Guidance by way of quantitative statements
on the minimum acceptable levels of accuracy, precision and completeness
should be provided in the document. The extent and confidence in the hydro-
geologic characterization will therefore be driven by the complexity of the
monitoring situation.
This approach will lead to a non-regular spacircj of the exploratory
borings. Relatively canplex areas will have many boreholes, and uniform
(simple) areas will have relatively few.
B. The TEGD should describe a procedure for the delineation of fea-
tures such as bedding planes, foliation planes, joints, shear planes, shear
zones, faults, and fault zones, and their associated secondary permeabilities
as related to the transport of liquids. Again, integrating regional informa-
tion and site-specific observations should be the key. The determination
will often require special exploration activities, such as angle drilling,
to determine the presence of vertical discontinuities.
-------
-11-
C. MaD scale (Pg. 1-3; Table 1-1) - The scale of the required geologic
or soil map should not be held to 1" = 200 ft (1:2400), but should be of an
"appropriate scale to identify properly the areal extent of the various qeo-
loaic or engineering soil units, considering the size of the facility and
complexity of the site neology".
D. Continuous saroling of earth materials at everv boring location, as
recommended in the draft TEGD (Pace 1-8) to define adeouately the subsurface
conditions, may not be necessary. Sampling at 5- to 2^-foot (on-center)
intervals (depending on che vertical heteroneneity of the geologic substrata)
in 50 to 75 percent of trie borings (with continuous sampling in the remainder)
should be adeouate to define subsurface conditions.
E. The TEGD should include, in the list of possible investiaatory tech-
niques, the use of cone penetrometers for differentiating individual, uncon-
solidated substrata. This technicue is gaming in acceptance with the en-
gineering and regulated community, and could be used to obtain continuous
depth information to augment borina data. A precaution, however, should be
added that all boreholes snould be sealed upon abandonment.
F. A competent field description of geolocic units encountered in ex-
ploration, with limited laboratory analysis to confirm field observations, is
aenerally sufficient. The requirement for petroaraphic analysis, bulk geo-
chemical analysis, mineralcgic analysis, and X-ray diffraction is unfounded,
except in certain special cases.
G. Permeability
1. Unsaturated hydraulic conductivity values should be determined
by appropriate laboratory methods for the major soil horizons. Pressure
plate, one-step outflow, or column methods are recommended. Field deter-
minations of saturated hydraulic conductivity are necessary to establish the
validity of laboratory data. Pump tests are preferable, but impractical in
lew permeability layers.
2. The term "intrinsic permeability" and "permeability" should be
replaced by "hydraulic conductivity." In those cases where floating hydro-
carbons are known or could be a contaminant, then sieve (mechanical) analyses
can be performed to augment the hydraulic conductivity measurements in predicting
the ground water flow.
3. An inpression is aiven, in the TEGD, that conducting a pumping
test usina a short screen in a thick aouifer would result in a hydraulic con-
ductivity determination of the formation at the depth .of the screen. This is
not true. Partial penetration factors may give erroneous results.
A more accurate approach for determining the relative hydraulic con-
ductivities of differing layers within a thick aguifer is to measure the
.overall aouifer transmissivity by a pumping test, then to assign relative
hydraulic conductivities on tbe basis of grain size distribution. Other
methods may also be used, but the iirplied method given by the TEGD may lead
to erroneous conclusions.
-------
-12-
We also suggest that the EPA require that all hydraulic conductivities
be reported in cm/sec units, in order to achieve consistency.
H. The time and cost of a full Appendix VIII constituent analysis of
potentially contaminated core samples (as recommended in Table 1-3 of the
Draft TEGD) are not warranted, and seem excessive for a detection monitorinq
Droaram. Anoendix VIII constituent analyses are more properly conducted
aurina a qround water Quality assessment program.
I. For health and safety reasons, it mav fce necessary to collect subsur-
face soil samoles and test for suspected contaminants. Care should fce exer-
cised to verify that OVA measurements are true and not artificially created
by the onerations at the qround surface.
J. An additional factor to consider in defining ground water flow oaths
(Page 1-20) is the effect of surface impoundments and other nearby surface
water todies on local ground water flow. Mounding below impoundments and
manmade canals can be the most sianificant factors influencing the direction
in which ground water contaminants migrate in shallow (nearsurface) aquifers.
Also, the presence of discontinuities should be defined.
K. .Additional factors which are inadecuately addressed in the draft
TEGD include:
1. Accuracy of locational (topoaraphic) surveys. Ground water
levels should be measured to an accuracy of 0.01 foot, in order to arrive at
an accurate determination of flow direction. Accordingly, relative casinq
collar elevations must also be determined to a 0.01-foot accuracy.
2. Proper definition of the uppermost acuifer at the its. Sub-
stantial emohasis is placed on determination of the vertical component of
flow. This factor is inoortant in some cases and not in others. In all
cases, the importance of vertical flow must be determined. This requires,
at a minimum, two sets (clusters) of vertically-spaced piezometers. The
exact number and depths are dependent on site characteristics. Once the_
potential importance of vertical flow has been identified, then the require-
ments for additional well clusters can be determined. For instance, if
strong uoward hydraulic gradients are encountered, no additional information
mav be required. However, if strong downward qradients are encountered,
much additional information is needed on the distribution of heads and the
characteristics of the confining layer.
3. Borehole Sealino. All borings made during site characteri-
zation (or at any other time) and not converted to niezometer installations
or monitoring wells should be adequately sealed so as hot to become an avenue
for contaminant migration. All such wells to be abandoned should be pluaged
with materials that have the same or lower oermeability than the formation
(e.g. very low permeability seal is required in the low permeability confin-
inq beds, while less stringent sealing requirements should apply to a high-
Dermeabilitv aquifer).
4. Recommended addition to site characterization. The characteri-
zation of the underlyino confininq layer has not been adequately defined.
In those situations where t'~e underlvirc confininq layer is relatively
-------
-13-
uncharacterized, sane of the borings should penetrate more than 10 feet into
that layer, and should be planned to:
a. Determine the areal extent and thickness (if less than,
say, 30 feet) of the lcwer confining layer,
b. Measure the hydrostatic pressure in any permeable layer
encountered, and
c. Be grout-sealed tnrcuah the confining layer in order to
preclude migration of contaminants.
5. Definition of qualified geologist. The term Qualified Geolo-
gist should be replaced by qualified geologist, geotechnical engineer, or
other qualified professional person with appropriate geotechnical experience.
High quality borehole logs are essential to identify and fully to define all
of the geologic and hydrcgeologic factors that are critical to site characteri-
zation. The use of trained professionals is indicated on the basis of their
ability to ocserve essential geologic features and because such individuals
generally have a good understanding of the responsibilities inherent in pro-
fessional field work as it relates to the design and long-term performance
of waste management facilities. It is clear that not to specify the qualifi-
cations will De to invite the cwner/operator to utilize the services of
drillers or other unqualified individuals to produce the borehole logs that
are used as the basis of the ground water monitoring program.
6. Depth of Borings. The statement that borings should penetrate
10 feet into bedrock is too specific. Boriros should penetrate sufficiently
deep into the underlying confining layer to define adequately its character-
istics as a confining layer. This assessment should demonstrate an adequate
understanding of such factors as the presence of weathering, core stones,
glacial debris, and dissolution features.
Use of the term bedrock varies in tbe TEED and is often inconsis-
tent with normal geotechnical usage of the term. It is suggested that bed-
rock be rep laced in illustrative examples by petrologic names of rock, but
that such terms as weathered and fractured be retained in connection with
bedrock descriptions. Bedrock is lithified geologic material (rock) or
crystalline rock. The figures in the TEED need to be reviewed for proper
geologic usage. The TEGD sesns to imply that bedrock is impermeable and
then names other rock types which may be permeable.
L. The definition of aquifer and uppermost aquifer need to be more spe-
cific and should be consistent with the objective of protection of usable
water supplies. These terms, ana also mixer (referring to a partially soluble
ana/or miscible waste), should be added to the glossary (TEGD, Appendix G).
1. In sane situations, the uppermost aquifer might extend hun-
dreds or even thousands of feet into the earth, using the loose definition
provided in the Draft TEED. For instance, some slight degree of hydraulic
interconnection or hydraulic camiunication (preferred by the Committee)
between widely separated sand layers might be inferred without a more speci-
-------
-14-
fic definition. In the Gulf Coast area, for example, coastal sedimentary
deDosits can alternate between sand and clay units to a depth of several
thousand feet. Since, in a regional sense, some component of flow might
exist between each adjacent sand unit, the uppermost aauifer might be con-
sidered thousands of feet thick. A more explicit definition of hydraulic
interconnection should be provided.
2. The addition of the nhrase "overlyina or perched water-bearina
zones" to the definition of uppermost aauifer substantially expands the con-
cept of aauifer from that included in the original regulations by including
any water-bearing zones above the acuifer reaardless of their ability to
yield water to a well, regardless of whether or not the zone is saturated,
and regardless of the ability even to sa/role the overlying water. Included
would be overlying clays and other tight formations that are of very low
permeability. This definition of uppermost acruifer is much more expansive
than the definition of aauifer and needs to be reconsidered.
3. The definition of uppermost aquifer provided in the draft TEGD
(water table to lower confining layer) aDplies well to an unconfined, or
"water-table," aauifer, but creates confusion when applied to a confined
aauifer. The piezometric level in a confined aguifer is not necessarily the
level of saturation in the overlying confining layer. Does the "uDpermost
aauifer" include that part of the overlyinG confining layer that is below
the piezometric level in the underlying confined aauifer, or does it include
that part of the overlyina confining aauifer that is saturated? This issue
is particularly relevant when there are otner transmissive zones above the
confined aquifer that do not underlie the waste management facility and
which are not hydraulically connected to the confined aauifer.
4. One addition to the aauifer definition, that in many cases would
restrict the zone of interest, is as follcws:
"If it can be demonstrated that a vertical upward carnonent of flow
exists in permeable layers underlying the uppermost Dermeaole layer,
and that the upward ccrroonent will prooably not be reversed by natural
or man-caused influences, then the monitoring may be restricted to the
uppermost, saturated, permeable layer."
Such a determination c^n be made bv simply noting whether or not the static
level(s) in the lower permeable zone are aoove the static levels in the
uppermost zone. Such determination should be routinely required in order to
assess the danger of downward migration of contaminants, and are a better
evaluation factor than pumpina tests held across an aguitard. Also, Figure
l.T> should be redrawn to show geologically real is tic-perched layers.
For the purposes of this Guidance, we offer the following definition of an
aquifer.
"An aquifer is a permeable and porous geologic unit that can transmit
significant quantities of fluid under ordinary hydraulic gradients, and
is capable of development as a source of water for human, industrial,
agricultural or other beneficial use."
-------
-15-
Chaoter 2 - Placement of Detection Monitoring Wells
A. The horizontal spacing of monitoring wells needs further considera-
tion. There should be a clear distinction made between detection monitoring,
systems and assessment monitorina systems, mainly in the interest of cost
effectiveness. The purposes of detection is, siiroly, to assess the presence
or absence of a contaminant. Assessment monitoring is used to determine the
location and extent of contamination and possible methods of mitigation.
The two monitoring systems need not make use of the same wells.
1. Monitoring well spacing - The purpose of site characterization
work is to identify avenues and direction of ground water (contaminant)
flow. No arbitrary spacing should be specified. Monitoring wells should be
located in those areas where oollution nuaration is most likely to occur,
based on the hydrooeolooical characterization of the site. Alternatives
that should be allowed in lieu of large numbers of wells are:
a. Wells in zones of higher permeabilities.
b. Wells located only on the downgradient side (exceDt, of
course, for the reouired upgradient control wells).
c. Wells located further from the actual site (but still on
the owner's property) that would intercept a dispersing
plume.
The retirement that spacina be close around a double-lined facility is
redundant, as there is a detection system between liners. Nearly all pollution
will oriainate from point sources under waste sites, except for previously-
develoDed unlined licruid-fllled lacoons or from existing impoundment facilities.
2. Soeculative wells - The text on page 2-16 of the Draft TEGD
points out that a sufficiently detailed site characterization may reduce the
need for "soeculative wells" by identifying preferential flowpaths. Figures
2.3 and 2.4, hcwever, shew wells in all dewngradient geologic strata/ not
just those that represent preferential flowpaths.
Additionally, Tables 2-1 and 2-2 do not include the extent of site
hydrogeological characterization in the list of factors that influence the
soacing and number of wells per cluster. More emohasis and more "credit"
should be given for the use of sufficiently detailed site characterization
in establishing the area location and depth of screening of wells. Some of
the wells shewn in the example are "speculative".
3. Minimum longitudinal distance - In certain instances, locating a
well at the uimediate edge of the waste management area (as recommended on
pages 2 and 3 of the draft TEGD) is impractical. For example, pipe racks,
pcwerlines, or underground piping often restrict drill rig access. Some
•guidance regarding the acceptable longitudinal distance from the waste manage-
ment facility ought to be included in the TEGD. For instance, is a distance
of 50 to 100 feet from the toe of a dike acceptable?
B. More flexibility in the length of tr.e well screen should be allowed.
-------
-16-
1. Well screen lenqths should not be limited to a maximum of 10
feet, at least for detection monitoring. The objective of monitoring is to
search for pollutants. If pollutants are discovered, then installation of
depth-specific assessment monitoring wells and screeas is appropriate. Aqui-
fers commonly have zones of higher hydraulic conductivity which prcduce a
large percentaoe of water to the well; these permeable zones will generally
be the zones of dissolved contaminant transport which will be effectively
sampled by long screens with minimal dilution. Sinkers and floaters can be
detected by thief sampling nuch more economically than by well clusters.
2. Illustrations - Additional examples are needed to depict the
thick clay, geologically-simple, and Cretaceous silts tone (Great Plains)
regimes. Every effort should be made to eliminate extraneous monitoring
wells in examples given, and to show why the remaining wells are effective.
For example:
a. Fig. 2-11, TEGD - One fully-penetrating well should be
sufficient, both in the upgradient and dcwngradient locations. Also, the
piezometric surface should be redrawn to reflect the direction of flow..
b. Fiq. 2-12, TEGD - In all probability, one well at each,
location, screened throunh tbe thickness of the sand and gravel layer would
be sufficient. If the silt contains sandy layers, then a second well,
screened through the silt layer, should also be required.
c. Fiq. 2-13, TEGD - In this particular case, one well,
screened throuqh the entire saturated thickness, should be sufficient for the
upqradient side. At the most, two dcwngradient wells should be used - one
screenino the upper half, and one screening the lower half of the porous sand.
A well is not felt to be justified for the claystone. Again, the piezometric
surface sbould be redrawn to reflect the flow direction indicated.
3. Placement of upgradient monitoring wells.
a. The number of backaround wells will be a function of hydro-
geologic conditions at the site and the statistical requirement to measure
variance, and usually more than two wells (and almost always more than one)
will be necessary. In addition, upqradient wells should match dcwngradient
wells with respect to formation screening for data to be amenable to comparison.
This section of the TEGD is written in such a way as to be biased tcward
the exceptional requirements, rather than toward the more common circumstances.
The section should be rewritten to accomodate simple situations, and then
progress to reasonable requirements for the more conplicated situations.
b. .As an illustration, see Figure 2-14, TEGD - Only if the
potentiometrie surface in the lower porous sand is lower than the piezometric
surface in the upper gravelly sand should any monitoring wells be required in
the lower sand. If the head is lower in the lower sand, then the upper half
'of that sand should be monitored. Also, if the thickness of the upper sand
is greater than about 50 feet, or if the formation/unit is not homogeneous,
then the upper and lower halves should be screened by separate wells.
-------
-17-
In this illustration, the number of wells actually needed might be only
one, under the most favorable circumstances, or three wells, under the worst
circumstances. This is a qood exanple of why illustrations must not only be
carefully conceived, but nust be thorouchly explained, in order for the con-
cepts not to be misused.
Chapter 3 - Monitoring Well Design and Construction
A. Drilling Methcds
1. Table 3-1, TEGD - Drillmq methcds should not be ranked nu-
merically without seme rationale. Air rotary drilling rated as 1 (one) is
in conflict with both the need to keep foreian matter cut of the monitoring
zone and the statement on pace 3-13 acainst air development technicues. A
suaqested ranking scheme could consist of appropriate versus nonaopropriace
technicues. An excellent reference for additional information on drilling
methcds is the NWKA/EPA "Manual of Ground Water Sampling Procedures."
MOTS: The use of decision trees can be very valuable in providing advice to
evaluators of monitoring plans. For example the chart on pane 3-14, Figure
3-3 of the TEED is effective in communicating the need for careful handling
of turbid samples though it contains several errors (a feedback loop should
be included in the left branch, and "repurqe" should actually be redevelop).
Similar charts miqht be useful for other decisions.
2. Reasonable field alternatives to the use of a hollow stem
auqer for drillinq in heaving sands should be included in the guidance for
drilling methods. An example is the development of a positive head within
the auger by filling the auqer with water, thereby displacing the heaving
sands when the knockout plug is removed.
3. A prcmisinci methed for both characterization and monitoring
well construction is the dual-wall drill stem air rotary. Formation sam-
pling is excellent, the outer drill pine provides a temporary casing, and
monitoring well casing can be installed before the drill stem is withdrawn.
These rigs are scarce at the present time, but this could chanqe with demand.
4. Conditional statements concerning the use of air rotary drill-
ina for monitoring wells (e.g., sloughing of sidewalls when the air pressure
is removed and its inappropriate use when contaminated soil in the upper
horizons is suspect) should be included. The exposure hazards to personnel
of drilling through contaminated zones by air rotary methcds should also be
ment ioned.
5. A statement regarding the potential for distributing a contami-
nant throughout the entire borehole if it is encountered durina drilling
operations (except with cable tool drilling) should be included and an ap-
propriate response to this case included.
6. The use of bentonite may contribute to the long term TOC con-
tent of the around water even with proper well developipent and purging. The
use of bentonite should also be discouraged in situations where heavy metals
mav be fcunc. These potential sources of chemical interference should be
rrent .
-------
-18-
B. Monitoring Well Construction Materials
1. Statements concerning the appropriateness of well materials
should include the statement that in all cases they are to be examined in
terms of the anticipated lifetime cf the monitoring program. The Ccmmittee
recommends that fluorocarbon resins or stainless steel 304 or 316 be specified
for use in the saturated zone when potentially sorbing organics are to be de-
termined, or may be tested, during a 30-year period. In such cases, and where
high corrcsion potential exists or is anticipated, fluorocarbon resins are
preferable to stainless steel (304 or 316). NSF-(National Sanitation Founda-
tion) and ASTM-approvea PVC for well casing may be appropriate if only trace
metals or nonsorbircg organics are the contaminants anticipated. As research
demonstrates the apprcprlateness of other materials for screens or casing in
the saturated or unsaturated zones, they may be acceptable. PVC, stainless
steel or teflon are appropriate casing materials in the unsaturated zone.
2. Figure 3.1, TEGD - Suggest removing the 8"-10" dense phase
sampling cup frcm the diagram. The cap at the bottom of the monitoring well
will serve primarily to accumulate sediment and act as a source of persistent
turbidity in the water samples.
3. Steam cleaning of a fluorocarbon resin casino/screen may not
be necessary if it has already been washed with detergent, rinsed with metha-
nol, and rinsed thoroughly with deionized water before packaging. Augers
should be steam cleaned off-site if at all possible. The casing/screen shculd
be enclosed in seme type of protective wrap until the sections are actually
lowered into the borehole, and shculd not be brought onto the site until this
time.
4. Fluorocarbon resins, PVC and other plastic materials do not cor-
rode in the strict sense of the definition. We suggest the use of the term
corrcsion for stainless steel and either weathering or deterioration when
discussing plastics.
5. Distinguishing the difference in actual monitoring performance
of stainless steel 304 and 316 would be difficult? either material shculd be
appropriate if conditions call for stainless steel.
6. The terms vadose and unsaturated zone are used throughout the
document. Vadcse zone is recommended as the preferred terminology.
7. The reference to sodium bentonite as a recarmended material
sesns overly specific and should be renoved. The criteria for a proper seal
shculd be one which is chemically carpatible with the anticipated wastes and
one to two orders of magnitude less permeable than the surrounding formation.
Of the bentonite clays, calcium bentonite is less susceptible to metal and
organic 'attack than sodium bentonite. Neat cement may also be an appropriate
sealing material, although only calcium bentonite shaild be mixed with the
cenent/ not sodium bentonite.
8. We recommend that uncontaminated water, rather than formation
water, be used for mixing cement, mixing with bentonite, developing or any
other use in construction of moriicring wells.
-------
-19-
C. Well Development
1. Air develocment is quite useful and does not exert long tern
effects on subsurface conditions as long as proper purainn and samolina oro-
cedures are followed. Despite frequent development, seme wells may never
provide turbiditv-free (less than 5 NTU) samples. In these cases, some ex-
ception to the turbidity limitation should be allowed if repeated development
does not improve the cuality of the samples.
2. Some mention should be made of the need to evaluate the hydrau-
lic properties of the well and the use of initial performance as a background
level for annual redevelopment and continued maintenance.
D. Well Desiqn
1. Placement of the sampling pump intake could properly be at the
top of the screen, as well as midway in the screened interval, and actual
conditions should override specific guidance. Pumps placed in the screen
often cause turbulence wnich results in suspended solids in the sample ana
eventual plugging of screens.
2. The subcasinq is an unusual suggestion, as there is little
or no literature to support its use. This type of complicated construction
is certain to encounter serious problems in installation, detection of mal-
functions, and verification of performance or repair. If water level devices
are to be lowered into tne well, this entire strina would probably have to
be removed or some unconventional well desiqn adopted.
E. Evaluation of existing wells - The field demonstration for existing
and new wells will be extremely difficult to evaluate in practice. Differ-
ences in construction may or not manifest themselves during the field test.
The results may lead to false conclusions in view of the normal variabilities
inherent in water quality parameters or sampling which may be attributed to
differences between old and new wells. Similarly, differences in well con-
struction, development, etc., which can never be duplicated may also result
in negative or positive biases due to causes other than well construction.
We recormiend that when sach situations arise and when the wells are suspect
that the wells be inactivated, sealed, and replaced.
Chapter 4 - Sampling anc Analysis
A. Water Level Measurements
1. Water level measurements should be made to plus or minus 0.01
feet.
2. Reccnmenc eliminating the use of an acoustical sounder as a
water level measurement device as the resolution of such devices at less than
25 feet below the ground surface is qe.nerally very poor.
3. Where a dedicated pump is being used, then a permanently^
installed pipe should be used as a quice for the water-level sounder.
-------
-20-
4. A description of the type of manometer which would provide
the resolution between aqueous ana nonaqueous layers should be included.
Electronic devices are also available for detection of conductance inter-
faces.
B. Well Evacuation
1. The removal or isolation of the stagnant water in the well
should be encouraged. If the pump intake is placed at the top of the screen,
well purqina will be easier and -nore efficient. The cwner/operator should
establish the hydraulic and yield characteristics of the monitoring wells,
calculate a purgino reauirement and then verify the Duraina process by measur-
ing pH, conductance, and temperature in the field. This is the best way to
get conductance and pH values for RCRA compliance. Both TOC and TOX have
volatile fractions which should be acknowledged in the discussion of purging
and sample collection.
2. Purging pumping rates should not be exceeded during sanding,
or additional development will occur. The three—veil-volume-ratio rule-of
thumb is not consistent with aood practice. The purge volume should be cal-
culated based on the hydraulic conductivity of the screened formation. Bail-
ers are very poor devices for well purging, since continual mixing of the
water column occurs during water collection. If any of these devices are
deemed acceptable for purging, it will be difficult to aet the regulated
community to be more careful in the selection of sampling mechanisms. It is
unclear as to how the water will "stabilize" within an hour or so after using
a qas lift pump to purge a monitoring well.
C. Sample Withdrawal
1. Sanclinq mcnitormq wells and piezometers may be the source of a
large nart of the total error in the total sampling and analysis procedures.
Good guality samples recuire both correct eouipment and skill in using the
equipment, and a basic understanding of the well sampling process. Bailers
are pcor sancling devices for volatiles and the bottom-draw models are tricky
to handle adequately. Sample reproducibility is very poor in most field
situations. The use of bailers for sampling should be carefully evaluated
before allowing their use for sample collection. They should be discouraged
where volatile oraanics are being determined.
2. The use of a steel chain should be mentioned in connection with
the use of bailers. It is easy to clean, and more manageable than cable or
wire under field conditions.
3. Regardless of whether a bladder pump or bailer is used, the use
of procedures or eouipment that minimize sample agitation and that reduce/
eliminate contact with the atmcsohere curing sairole transfer should be en-
couraged.
4. Cleaning of non-dedicated samplers and tubing for either inorganic
or orqanic contaminants should begin with detergent/soap mixture and end with
distilled water rinses before storina for the next use.
-------
-21-
5. 8Iadder pumps do not provide a "continuous" sample, but rather
provide a pulsating stream. They should therefore he .operated carefully, so
that the sample bottle is filled with either one conclete "pulse" or as few
pulses as possible by turnina dcwn the nressure on the"compress ion cvcle.
This is possible with a careful operator and a lew discharge rate.
6. A flow rate of 100 ml/fainute for sample discharge is about the
most rapid rate at which a 40 ml vial can be filled without undue agitation.
This rate should apply to all samples where oases or volatile cons titutents
are of interest (e.g., pH, alkalinity, TOC, TOX, 601 and 602 organics, etc.).
The sampling flow rate should not exceed the puree rate and the purge pumping
rate should be below the flew rate used during well development. Otherwise,
continued well development or well damage may occur.
7. Field electrode measurements of pH and conductance should be
made before and after sairole collection as a measure of purcina efficiency
and as a check on the stability of the water sampled over time.
D. Sairple Preservation and Handling
1. MultiDle transfers recjuired by the use of bailers for sampling
volatiles are not oooa practice. The transfer of samples from one container
to another in the field should not be encouraged.
2. A trip blank for organics should be glass distilled, not deion-
ized, water.
3. The sample blank should also be reported and more guidance
should be aiven on the procedures for identifyina sources of contamination.
For examnle, methylene chloride will be virtually impossible to avoid in
many cases.
4. Field loabooks should include documentation of the puroing
process (e.g., time started, initial water level, purge volume pumping rate,
time finished, etc.).
5. Climatic conditions on the day of all sampling activites,
including air temperature, should be reported.
E. Evaluation of t^e Quality of Ground ^ater Data
1. The LT (less than) detection limit values will vary sanewhat
with time as will the LCO (limit of cruantif ication) and LOD (limit of detec-
tion). That is, they should be established each day of analysis by the use
of both procedural and field standards and reported with the data.
2. Extra digits carried through any computation provide an inaccu-
rate impression of the quality of t*e data. The significant fiaures should
reflect actual analytical precision.
3. Accuracy and precision determine significant figures, not vice
versa. The discussion on Page 4-30 of the TEGD is, in addition, incorrect
numerically. The precision of the range 65-73 +_ 1 ug/L is about 1.5%, not
11%.
-------
-22-
Chapter 5 - Statistical Analysis of Detection Monitoring Data
This Chapter was reviewed separately, and a report on this review is
attached as Appendix D.
Chapter 6 - Assessment Monitoring
A. The specification of the initial number of wells required for assess-
ment monitorinq seems excessive. The investiqation would be more efficient
if it were accorolished in stages, with progressive well placements deter-
mined on kncwledae gained. This comment, of course, does not recognize the
implications of enforcement action should the cwner/operator prove to be
uncooperative with EPA Regional or State regulatory officials.
The guidance for establishina both the minimum number of initial well
clusters (Section 6.5.2) and the minimum number of wells per cluster (Section
6.5.3) needs to be improved. The text recomnends that seven well clusters be
installed, with four inside the plume and three outside the plume. The text
also reconroends five wells per cluster, three within the plume and one each
above and below the plume. These are initial wells, installed before the
plume is defined. Hew can initial wells be specifically installed inside or
outside of a plume before the plume is defined? Also, five wells per cluster
is excessive in many situations (e.g., where the transmissive zone is less
than twenty feet thick). The information qained from a five-well cluster
may not be critical or even relevant to the design of a corrective action
system (see also the discussion of imiscible compounds in the detection
monitoring section).
B. The analytical effort should minimize the number of full Appendix
VIII analyses. Given its hiqh cost, "speculative" Appendix VIII jonstituent
analysis should be avoided on all samples. Appendix VIII analysis seems
appropriate on a few samples, with the plume then defined using tracer
corrpounds.
C. The discussion on modeling in Section 6.4.3 of the Draft TEGD is too
general and does not clarify either the practical uses or limitations of ground
water models. The suggestion that models can be used to predict future events
(page 6-11) is misleadina and provides no additional precautions regarding the
need for prior calibration and verification of any model (against geologic,
hydrcgeologic, and waste characterization data) before its use as a predictor.
-------
APPENDIX A
U.S. ENVIRONMENTAL PROTECTION AGENCY
SCIENCE ADVISORY BOARD
ENVIRONMENTAL ENGINEERING COMMITTEE
GROUND WATER MONITORING GUIDANCE REVIEW SUBCOMMITTEE
CHAIRMAN
Dr. William Haun
13911 Ridqedale Drive
Suite 343
Minnetonka, MN 55343
MEMBERS
Dr. Michael Barcelona
Illinois State Water Survey
2204 Griffith Drive
Champaign, IL 61820
Dr. Keros Cartwriqht
Illinois State Geological Survey
615 East Peabody Street
Champalqn, II 61820
Mr. Richard A. Conway
Coporate Development Fellow
Union Carbide Corporation
P.O. Box 8361 (770/342)
South Charleston, WV 25303
Mr. John Fryberger
Fngineerina Enterprises Inc.
1225 West Main
Norman, OK 73069
Dr. James E. Krier
University of Michiqan Law School
Hutchins Hall
800 Monroe Street
Ann Arbor, MI 48109-1215
Dr. Allen Hatheway
Dept. of Geological Engineering
University of Missouri - Rolla
Rolla, MO 65401
Mr. Robert Morrison
2001 Reetz Road
Madison, WI 53711
Dr. Carl Silver
Department of Quantitative Methods
506C Matheson Hall
Drexel University
Philadelphia, PA 19104
EXECUTIVE SECRETARY
Mr. Harry C. Torno
Environmental Protection Acericy
499 South Capitol Street SW - Suite 508
Was hire cor, TC 20460
-------
APPENDIX B
^°3r%
\ UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
XNl/^L. 5 WASHINGTON. D.C. 204S0
PRO14,
OFFICE OF
SOLID WASTE ANO EMERGENCY HE5PONSE
MEMORANDUM
SUBJECT: Science^MV^>^Y_Board Review of the "Draft RCRA Ground-Water
w-m-i-1 "" "" 1 Enforcement Guidance Document"
ran
f J.ny
.Xdi
FRCM:
TO:
5 A. Lucero, Director
Office of Waste Programs Enforcement
Addressees
_ , ^ttfdlec? for ^ur review is a cnpy of the Draft RCRA Ground-Water Hinitnrinn
Technical Enforcement Guidance Ebc-snent (TEGD). Also attached is a copy of the
RCRA Ground-Water Monitoring Compliance Order Guidance (COG). "together, these docu-
ments provide comprehensive guidance on tow to identify and rectify grourai-water
monitoring violations at RCRA facilities. "
The purpose of the Science Advisory Board meeting on October 3-4 is to review
technical issues in the draft TEGD. We have, however, included the COG in this
package for background reading. The COG provides an overview of the RCRA cround-
^er_^nitorIng reflations and explores the interrelationship of the Part 265
Part 270, and Part 264 regulations. It also describes the Agency's strategy for
correcting ground-water nonitoring violations at interim status lard disposal facil-
1 J, 1that you read the COG since this document introduces*the Aqencv's
ground-water monitoring enforcement strategy and provides you with the go lie/ and
regulatory context that you need to conduct your review of the draft TEGD.
DrD. Thf d^aft TEGD provides guidance on how to evaluate the design art! operation of
RCRA interim status groundwater monitoring systems. The audience for the draft TEGD
includes permit writers, field inspectors, attorneys and enforcement officials (enqP
neers, h}drogeologistsr statisticians). The TEGD provides guidance on how to evaluate:
° characterization of site hydrogeology (Chapter One);
placement of detection monitoring wells (Chapter Two);
° monitoring well design and obstruction (Chapter Three];
0 sampling and analysis plans (Chapter Fbur);
statistical analysis of detection monitoring data (Chapter Five); and
° assessnent monitoring plans (Chapter Six).
-------
-2-
The draft TEGD represents the consensus views of an EPA workgroup made up of
representatives from the Office of Waste Programs Enforcement, the Office of Solid
Waste, the Office of Enforcement and Compliance Monitoring, the National Enforcement
Investigations Center and EPA regions 3,5,6,8 and 9. Wbrk on the draft TEGD began
last fall after EPA completed a study on the compliance status of the regulated
comnunity in regards to the RCRA ground-water monitoring requirements. The study
revealed serious problems. The level of noncompliance in the regulated community
was alarmingly high. Confusion existed bo tin witnin the regulated comrrunity arri within
EPA as to what constituted compliance. This confusion was contributing to a breakdown
in our ability to efficiently issue permits to RCRA land disposal facilities.
The Office of Waste Programs Enforcement was given the job of writing enforcement
guidance that would definitively describe how enforcement officials should apply the
interim status ground-water monitoring regulations in compliance decisionmaking. As
you can imagine, writing guidance of this type was very difficult. A "ocoktook"
which lays out every possible situation the* enforcement official may encounter and
describes how to make decisions in each situation was impossible to write. There are
simply boo many site specific situations enforcement official may encounter ¦and too
many variables which complicate decisionmaking. We wrote the draft TEGD very carefully
to try to balance the need of the our enforcement program for specific, detailed
Guidance with the reality that conpliance decisionmaking in the ground-water area is
very complicated. In those areas where tne enforcement official must consider
numerous variables in decisionmaking, we have tried to identify all the important
factors which may affect the decision and have tried, through example, to show how
the enforcement official may consider individual variables. An example of this
approacn can be found in the discussion on the horizontal spacing between downgradient
monitoring wells (2-5 to 2-15).
There are a number of technical issues on which we would like your review and
advice. They are listed individually in an attachment with a brief synopsis which
should aid you in your review. As ysu conduct your rev lev I encourage yDu to keep
in mind that this guidance was developed within the context of the RCRA interim
status ground-water monitoring regulatory structure. This Science Advisory Board
review should not be a forum on whether or not the Agency has adopted the "best"
regulatory approach for interim status ground-water monitoring systems. Instead
your review should focus on how we have dealt with specific technical issues in the
document and how you feel we can improve the technical aspects of this document.
If you have questions regarding the draft TEGD please feel free to contact either
Michael Barclay at (202) 475-9315 or Dr. Ken Jennings at (202) 475-9374.
addressees: Dr. Michael Barcelona
Dr. Keros Cartwright
Richard Cbnway
John Fryberger
Dr. Allen Hathewav
Dr. William Haun
James Krier
Robert Morrison
Dr. Carl S.iver
-------
1. Level of Site Hidrogeo logic Investigation
Our objective in Chapter One was to describe the quality arid quantity of hyjro-
geologic information needed by owner/operators - i.e. How much information is''
enough? What are appropriate techniques bo collect information? Table 1-1 illus-
trates h^drogeologic investigatory techniques owner/operators may use to collect
data as well as preferred presentation ferrets. We expect this chapter will be
Host useful to those enforcement officials who rtust o trier owner/operators to
perform a hydrogeologic investigation (refer to phased order approach in CCG).
If an owner/operator followed the approach described in this chapter for
conducting a hydrogeologic investigation, will the investigation yield
enough information for the owner/operator to design a detection monitoring
system? Will it provide erouch information to design an assessment ironitoring
program?
Can the approach this chapter describes for collecting information be made
nore efficient? What additional investigatory techniques might be useful
in RCRA. site investigations?
2. Definition of Uppermost Aquifer
The discission beginning on pace 1-33 describes how the enforcement official can
decide if the owner/operator has correctly identified the uppermost aquifer beneath
his site.
Do you feel the criteria we have established for identifying the uppermost
aquifer (water table to confining layer and perched zones of saturation)
adequately balances the need for specific guidance with the need ,for-'f lexibility*
"enforcement d"e^sy"nma]
-------
4. Well Cbnstruction Materials
Section 3.2.1 describes our preference for construction materials for RCRA
rronitoring wells. In those cases where the enforcement official will require
the owner/operator to install wells as a condition of an order, we reconniend
that the enforcement official require the owner/operator to use teflon, stainless
steel 316 or other proven chenically and physically stable materials for those
portions of the well casing in the saturated zone. Other materials such as PVC
may be used as construction materials in the unsaturated zone. Current research
indicates that teflon and stainless steel 316 are nore highly resistant to corrosion
from chsnical species likely to be encountered in RCRA monitoring and are less
likely to.interfere^chemically with_the quality of ground-water samples. j^heif^
^research results "are "ipjt.def initiyeg3Dw|yfef3 Also, there are problems associated
with constructing wells with teflon and with the long term structural viability
of teflon wells (shearing of threads, deformation of screen slots).
Is it practical_to recommend that_ RCRA we1Is be constructed of teflon given
the engineering ar5 strueturaJ"ptoblaxis as30clated j^iffiTeflon}'£
The Agency currently does not have a program for testing materials to
determine their suitability for use as well casing materials. Assume the
Agency was interested in establishing criteria that manufacturers oould use
to test their products. How difficult would it be to establish such criteria?
Oould such criteria be developed quickly? Oould the criteria be developed
such that manufacturers oould get quick turnaround on product testing?
The cost of stainless steel 316 and teflon is often cited as ¦-factor which
mitigates against their use as construction materials. Is the added cost of
these materials to the overall monitoring program worth the added benefit
they supply in terms of increased confidence in monitoring data?
Monitoring wells should be constructed of materials that will last 30 or
nore year's. Also, monitoring wells should be constructed of materials that
are chemically inert. Data available to us indicates that 9ome materials
comnonly in use as construction materials will degrade upon exposure to
chemical constituents such as those found in Appendix VIII. We do not,
however, have extensive material testing data or definitive research findings
regarding the longterm degradation one might expect various materials to
experience as a result of exposure to different chemical compounds. Also,
there is aonflicting research regarding the adsorption and de3orption properties
of construction materials such as PVC. Should the Agency limit the type of
materials that can be used in well construction (in the context of a compliance
order) to those which we know to be chemically and structurally stable?
Should we allow owner/operators to use construction materials which later
research may prove to be structurally and/or chemically unstable?
5. Statistical Analysis - False Positives
The statistical analysis (i.e. student's t test) required by Part 265 yields a
high rate of false positives - that is, an indication of contamination when none
exists. We acknowledge this in the draft TEGD and have taken steps to mitigate
pnoble-ns associated with false positives. We have suggested an alternative
-------
statistical test which will reduce the incidence of false postives (the average
replicate t test proposed by the Chenical Manufacturers Association). We have
also recommended that owner/operators install more upgradient wells which will
cut down on the likelihood of false positives. Finally, we have developed guidance
on how owner/operators may prove cr disprove false positives in the context of an
assessnent monitoring program (Section 6.3).
Co you have any recommendations as to how we can further reduce the incidence
of false positives and at the same time avoid creating unaccaotably high
levels of false negatives?
6. Sampling and Analysis Plans
Chapter Pour describes the type and amount of information owner/operators should
include in their written sampling and analysis plans. This chapter more than any
other chapter in the draft TEGD contains specific requirenents we feel owner/operators
should follow.
Have we included all the elements of sampling and analysis plans that you
feel are necessary for the owner/operator to carry out an adequate sampling
and analysis program (see section 4.1)? Are there other aspects of sampling
and analysis programs that you feel we should include?
Would you suggest modification to any of the specific requirenents related
to sampling or analysis in this chapter?
7. Assessment tonitoring Plans
Chapter Six describes the type and amount of information owner/operators should
include in their written assessment monitoring plans.
Have we included all the elements of assessment monitoring plans that you
feel are necessary for the owner/operator to carry out a successful assessment
monitoring program? What additional elements would you include?
We are interested in promoting the idea that owner/operators should use a
variety of procedures and investigative tools in their assessment monitoring
programs (see Section 6.4). Are there additional assessment methods other
than those in Section 6.4 that we should have worked into the document?
Section 6.7 describes the type of analyses owner/operators should conduct
in an assessment monitoring prcgram. This section encourages owner/operators
to consider conducting a sampling prcgram concentrating on a limited set of
parameters in the early phases of the assessment monitoring prcgram and
later expanding the sampling effort after the geometric dimensions of the
plume(s) have been established. Should we provide more explicit guidance
perhaps in the form of examples to illustrate this concept?
-------
APPENDIX C
U. S. ENVIRONMENTAL PROTECTION AGENCY
SCIENCE ADVISORY BOARD
ENVIRONMENTAL ENGINEERING COMMITTEE
The following is a list of individuals who offered Dublic conment on the draft
of an Aqency document entitled "RCRA Ground Water Monitoring Technical Enforce-
ment Guidance Document (TEGD) at a meeting on October 3-4, 1985.
Mr. Peter Vardy
Waste Management, Inc.
Mr. Swep Davis
Environmental Testing and Control
Dr. Jay Lehr
National Water Well Association
Mr. David Alexander
Envirosafe Services, Inc.
Ms. Sue Moreland
Association of State and Territorial
Solid Waste Management Officials
Mr. Tony Breard
CECOS International
Mr. David Mioduszewski
QED Environmental Systems, Inc.
Mr. Geoffrey Hunk in
Ground Water Sampling, Inc.
Ms. Linda Greer
Environmental Defense Fund
Dr. Charles Johnson
National Solid Wastes Management Association
Dr. Brian Meade
Consultant, duPont Corporation
Mr. James Gustin
Law Engineering Services
-------
APPENDIX D
SAB-E£C-86-007
REPORT
on the review of the
"RCRA GROUND-WATER MONITORING
TECHNICAL ENFORCEMENT GUIDANCE DOCUMENT"
(SUPPLEMENT - Statistical Discussion)
by the
Environmental Engineering Committee
Science Advisory Board
U. S. Environmental Protection Agency
April, 1986
-------
INTRODUCTION
In February, 1986, the Environmental Engineering Committee completed its re-
view of the Office of Waste Programs Enforcement's draft "RCRA Ground Water
Monitoring Technical Enforcement Guidance Document (TEGD).
At its regular meeting on February 13-14, 1986, the Environmental Engineering
Committee decided that the portions of the report dealing with statistical
methods (Chapter 5 and Appendix B of the TEGD) needed further review. A Sub-
committee, consisting of Dr. J. William Haun, Chairman, Dr. Charles O'Melia,
Dr. Mitchell Small, Dr. Carl Silver and Dr. Charles Norwood, was appointed,
and this document is a report on their review.
In the introduction to Chapter Five of the TEGD, the Agency notes that RCRA
facility owner/operators must implement a ground water monitoring program
capable of determining if a facility has had a significant effect on the
quality of the ground water, and states that this determination is based on
the results of a statistical test.
The Committee has followed standard SAB review practice and concentrated on
the scientific/technical aspects of the draft Guidance. The Committee recog-
nizes that some of the following comments may encourage solutions not permit-
ted by existing regulations, but we assume that the Agency's primary concern
is to arrive at a technically acceptable method.
MAJOR ISSUES
The Committee recognizes the extreme difficulty in prescribing a single suit-
able test, given the wide variation in the hydrogeology of the sites, and in
the data which have been collected. The Committee also believes that the
Agency has done its best to prescribe a test which will meet existing interim
status regulations (40 CFR Parts 264 and 265 - which require that a Student's
t-test be used for this determination, but does not prescribe the specific
test), and which recognizes many of the practical limitations on the data
available.
It is unlikely, however, that any statistical test will meet, for most of the
cases encountered in the field, the stated intention of the TEGD, which is to
determine directly whether there has been a "significant increase in any ground
water contamination indicator parameter in any well" (it is important to note
that "significant" is defined as statistically larger than background).
One reason for this is that the Agency has not yet defined:
1. The magnitude of the difference which defines the event that it is
important to detect (other than to state that it must be statistically larger
than background); and
2. The probability with which that difference should be detected.
-------
-2-
Statistical tests are characterized by the magnitude of the difference tested
for, the false positive (Type I) error rate, and the false negative (Type II)
error rate. The Agency discussion to date has focused primarily on the false
positive rate and the concern that this not become too high for a facility.
The magnitude of the difference tested for is by implication any positive
difference whatsoever, given the manner in which the t-tests are currently
formulated in the TEGD. It is not clear whether this is the result of thought-
ful intent or simply the default condition. Finally, the allowable false
negative rate is not addressed by the Agency, this being the important consi-
deration from the perspective of protecting the public health (i.e., how
likely is.it that a leak will go undetected). All three parameters should be
explicitly considered in the EPA formulation.
Another point that needs to be clarified in determining whether there has
been a "significant increase" is the precise identification of the population
with which the downgradient wells are being compared. Is it the distribution
of concentrations at a single well, or should the baseline population variabi-
lity incorporate the spatial and temporal variability which is inherent to the
"upgradient" or more generally, the "no-leak-influence" condition. The general
consensus is that the latter is what is intended. The implication of this is
that the CABF t-test based on aliquot/replicate variability is inappropriate
for use in the TEGD, because it fails to account for the inherent well-to-well
and temporal variability of ground water quality. In addition, the high degree
of correlation between replicate samples violates the assumption of independent
sampling required in the t-test. The Agency has recognized this and intends to
recommend that the CABF t-test be replaced by the AR t-test, which can incorpor-
ate upgradient spatial variability, and averages samples which are at least not
obviously correlated to a high degree. The Agency has also been re jonsive to
other apparent drawbacks in the draft TEGD, such as the need to use data trans-
formations when concentration distributions deviate significantly from normality.
While the AR t-test does eliminate some of the difficulties associated with the
CABF replicate t-test, there are limitations which still must be recognized.
In particular, the underlying t-distribution model does not theoretically apply
to the situation where sampled concentrations are being drawn from populations
(i.e., different upgradient wells) with different means, and then averaged.
Also, the AR t-test as currently formulated requires a single downgradient
observation for comparison with the upgradient distribution (which, given the
first argument may actually be a mixture of distributions). We consider each
of these drawbacks in turn.
The AR-test is a physically appropriate model for the case where the ground-
water aquifer (unaffected by a leak) is homogeneous in terms of concentration
distributions. When inherent stratification is present, however, i.e., there
are systematic differences in concentration distributions among wells, the
average of samples from these wells may not, indeed most likely will not,
follow the Student's t-distribution. One may conceptually skirt this issue
by indicating that he is not interested in comparing downgradient wells to a
single, uniform upgradient distribution, but rather to the average of the up-
gradient field. Nevertheless, the t-distribution no longer directly applies
as before, and if the procedure is used, there is a clear need for secondary
-------
-3
analysis of the ongoing RCRA data being collected to see how far the distri-
butions do deviate from the assumed Students-t, and to evaluate the implica-
tions of this in terms of the robustness of the detection procedure.
The second problem involves the use of single (albeit, replicate averaged)
downgradient observations to judge leak occurrence. This problem is difficult
to address because we are interested m leaks even if they affect only one
well (indeed, wel1-conflned plumes are likely to do just that), and because
we desire rapid detection and do not necessarily want to wait for evidence
of contamination over a long-term period before triggering assessment moni-
toring. Two suggestions have been made for addressing this issue-. The
first is to average the downgradient observations. It is argued that the
added sample size and associated increase in power would allow a greater
ability to detect changes, even if they occur at only one of the downgradient
wells. This procedure should be coupled with a check on the downgradient var-
iance to ensure that a significant increase in one downgradient well is not
being masked by random decreases in the others. The second suggestion (put
forth by R.D. Gibbons, attached as Appendix A) is to abandon the t-test for-
mat and replace it with a tolerance interval test commonly used in quality
control studies. This procedure would have the added advantage that it would
eliminate the dependence on the t-distribution assumption for the upgradient
mean, which as noted above, is questionable. The Agency should consider
these recommendations as possibilities for alternative detection monitoring
procedures.
ADDITIONAL CONSIDERATIONS
The application of the AR t-test with multiple upgradient wells averaged over
the four quarterly first-year measurements allows proper consideration of
spatial variability (between wells) and temporal variability (between sea-
sons), so long as the temporal variations occur randomly. However, if sea-
sonal variations occur deterministically, i.e., one or more of the four
indicator parameters tend to be higher in one season and lower in another,
then a bias is introduced in the test. False positives are more likely to
be triggered during the season when background concentrations are high, and
false negatives are more likely to occur during the season when background
concentrations are low. Possible mechanisms for dealing with this problem
include the use of more advanced Analysis of Covariance (ANCQVA) procedures,
or the incorporation of seasonal adjustment factors in certain cases. The
ANCOVA procedure would allow seasonality to be one of the dimensions of
variations considered. The skill level required for field implementation of
an ANCOVA would be high, though software is available. Seasonal adjustments
could be considered by the EPA on a site-specific basis. Procedures should
be developed by the Agency that would allow this in future regulations.
As part of the discussions of the Committee, some more general questions were
raised as to whether the quarterly sampling protocol, using four indicator
parameters, was really best for meeting the Agency's goal: identifying leaks
at facilities. It was suggested that a broad review of the procedure be made.
This could consider:
-------
-4-
1. The appropriateness of the four chemical indicator parameters for
detecting leaks. In particular, a more specific set of parameters (e.g.,
organic chemicals or heavy metals) may be appropriate on a site-specific'
basis.
2. The use of other hydrogeollcal approaches for detecting leaks, such
as tracer studies in conjunction with expert surveys.
RECOMMENDATIONS
1. The Agency should use the AR t-test method proposed in the TEGO in
the short term. It should be explicitly acknowledged in the TEGD, however,
that there are situations where the method may not yield accurate results,
such as where seasonal variations occur, and examples should be provided. In
these cases other statistical procedures, such as analysis of covariance
(ANCOVA), may be appropriate.
2. The Agency should institute a vigorous program of secondary analysis
(of data collected as the program proceeds) performed by skilled statistical
analysts to confirm the adequacy or inadequacy of the proposed method.
3. The Agency should establish a group to attempt to solve the much
greater overall problem of devising a statistical test that will satisfy
regulatory needs as defined above, and will at the same time be technically
defensible over the wide range of situations encountered in actual practice.
This group should keep the following points in mind:
a. The goal of the statistical analysis should be to detect leaks
from RCRA facilities.
b. Any test (statistical or not) should be justified by reference
to site-specific factors. Those factors should define a physical system
consistent with the assumptions required by the test selected.
c. Statistical tests should be based on preselected values of Type
I error (false positives), Type II error (false negatives) and an environ-
mentally significant difference. When these are specified, knowledge of
variability on the site allows direct calculation of the number of wells
required.
d. Lab bench (analytic) error should not be used to test whether
two wells (or two regions) differ. Error components should include both
temporal and spatial variation, where such variation is random. Where this
variation is not random, the significant difference referred to in in c.
above should be defined relative to the spatial and temporal variation ap-
propriate to each site.
-------
APPENDIX A
(to Statistical Supplement
Petition:
A Methodology for the Statistical Evaluation of Groundwater Data
Robert D. Gibbons
University of Illinois
1
-------
The critical flaw associated with the CABF t-test is that it
uses the aliquot or replicate sample as it's unit of observation.
Since replicate samples of the same groundwater are almost
perfectly correlated, the assumption of independent observations
is violated, the degrees of freedom and sampling variability are
incorrectly computed and the resulting test statistic has a
false positve rate (i.e. type I error rate) that approaches
unity. Statistics, such as, the AR t-test average the replicate
samples and use a single number to represent the level of a
compound at a particular well on a particular occassion. if the
occassions are reasonably well separated (eg. quarterly
measurements) the assumption of independence is tenable.
However, unless we pool downgradient wells, (a practice that is
doomed to failure), the use of any t-statistic is inappropriate,
because we have no method of estimating downgradient variability
and we must therefore assume that a single new downgradient
observation represents a sample mean value. Clearly this in not
the case. Our statistical problem is not one of comparing two
mean values (i.e. obtained from a sample of upgradient wells and
a sample of downgradient wells respectively) but rather a problem
of determining the probability that a single new downgradient
observation was drawn from an upgradient population distributed,
with mean and variance that can be estimated from a sample of
upgradient observations (which reflect both spatial and temporal
variability). The only rigorous statistical solution to this
problem is to construct tolerance intervals (see pp. 224ff. in
Bowker A.H. and Lieberman G.J., Engineering Statistics. Englewood
2
-------
Cliffs, N.J.: Prentice-Hall, 1959) around the upgradient mean
value and compare each new downgradient observation to the upper
bound of this interval (upper and lower bound in the case of pH)..
This statistical strategy, which is common in quality control
studies where the integrity of a new manufactured product is
determined by comparing one or more of its measureable
characteristics to a historical sample of such products, has also
been suggested by the TEGD in their Feb 6, 1986 response to the
Science Advisory Board:
Quality control techniques described by (Burr, 1976) may
also be appropriate. Quality control techniques generally
operate by using baseline measurements to establish
tolerance limits which represent the bounds of acceptable
performance based on future measurements.
We propose to use tolerance intervals for the analysis of
all of our groundwater data. In the case of the four indicator
parameters (TOC, TOX, pH and specific conductance) , we will use
tolerance intervals based on the normal distribution (specific
details of this method are given in Appendix A). Briefly, the
method simply requires that we multiply the upgradient standard
deviation by a tabled value that depends only on the number of
independent upgradient observations that we have (eg. 2 wells
measured quarterly for 1 year produces 8 independent upgradient
observations). The value produced by multiplying the upgradient
standard deviation by the tabled value is then added to the
upgradient mean. Any downgradient observation that exceeds this
-------
value is significant, otherwise it is considered to be within
upgradient limits. Suitable transformation of the data (eg
natural logarithm) is suggested to better approximate the assumed
normality of this procedure (although it should be reasonably
robust to minor departures from normality).
The previous methodology is not appropriate for data that
exhibit truncated distributions due to detection limit problems.
The 32 commonly measured volatile organic compounds all have
detection limit problems and are therefore inappropriate for
analysis using the tolerance interval described in Appendix A. A
simple modification of this procedure in which the Poisson
distribution is substituted for the normal distribution
completely remedies this problem. The Poisson distribution which
is the limiting form of the binomial distribution and has been
used to describe rare event data such as mutation rates, is
appropriate for the analysis of chemical data in which compounds
are commonly found at levels at or below the detection limit or
are measured at relatively few parts per billion. Unlike the
normal distribution, the Poisson distribution yields unbiased
estimates of the mean and variance despite severe truncation of
the observed frequency distribution at the detection limit. We
propose that all volatile organic compounds be statistically
evaluated using the Poisson tolerance intervals. Complete
details for the computation of both the fit of the Poisson
distribution to an observed sample of upgradient data and
estimation of the upper bound of the tolerance interval are given
in Appendix B. As in the previous example, the significance of a
-------
new downgradient observation is determined by simply comparing
it's value to the upper bound of the tolerance limit.
D
-------
Appendix A
Gaussian Tolerance Limits for Individual Downgradient Observations
In groundwater management problems we are not interested in
knowing the true value of a downgradient population mean, but
rather, in estimating the highest likely downgradient value based
on our knowledge of the upgradient population. Of course, if we
knew the population values ^ and cr, the areas of the normal curve
would provide the required estimates. However, when only x and s
are available from a sample of N independent upgradient
observations, there are two kinds of uncertainty: the exact value
of (j is unknown and
-------
Factors (K) for One-Sided Tolerance Limits*
95%
99%
N
K
K
3
7.66
10. 6
4
5.14
7 . 04
5
4.20
5.74
6
3 .71
5 . 06
7
3.40
4 . 64
8
3. 19
4.35
9
3.03
4 . 14
10
2 .91
3 .98
11
2 .82
3 .85
12
2.74
3 .75
13
2.67
3 . 66
14
2.61
3 . 58
15
2.57
3 . 52
16
2.52
3 .46
Extracted from A.H. Bowker and G.J. Lieberman, Engineering
Statistics (Englewood Cliffs, N.J.: Prentice-Hall,1959) Table 8.3
For an upgradient sample of size N, from which x and s are
obtained (ie. the sample mean and standard deviation), the table
gives the factor (K) which is used to compute the upper tolerance
limit x + K(s). The resulting tolerance .limit will include 95%
or 99% of the observations and have a type 1 error rate of less
than 5%. In this case N reflects the number of upgradient wells
multiplied by the number of quarterly measurements; Hence, N=16
could be obtained by a single measurement of 16 upgradient wells,
4 wells measured quarterly for 1 year or 2 wells measured
quarterly for two years. Each new downgradient observation is
simply compared to this limit.
2
-------
Appendix B
Poisson Tolerance Intervals for Rare Volatile Organic Compounds
A family of statistical models that is particularly well
suited to "isolated" or rare event data are models that are based
on the Poisson distribution. The Poisson distribution which is
the limiting form of the binomial distribution (i.e. the
distribution that describes the probability of the presence of "an
event when that probability is extremely small) has been
effectively used to characterize rare event data such as mutation
frequencies and radioactive count data. Using the Poisson
distribution, we can therefore 1) test for the randor-.ess of a
process such as the occurrence of detectable levels of volatile
organic compounds in upgradient wells, 2) develop interval
estimates that will tell us how many parts per billion of a
specific compound in a new downgradient observation are
consistent with chance expectations and, 3) allow us to compare
upgradient observations from different sites so that similar
sites can be combined to enlarge our sample of upgradient or
background observations and in turn increase the precision of our
statistical estimates. In the following, the relevant hypotheses
will be outlined and the appropriate statistical methods will be
described.
3
-------
Hypotheses and Questions of Interest
Question 1
Is the sampling distribution of the upgradient wells at a
particular site for one or more volatile organic compounds
consistent with expectations of a Poisson distribution?
If the answer to this question is yes, we can have
confidence in assuming that the observations are independent and
that the observed levels of these compounds are consistent with
random sampling fluctuations. If the answer to this question is
no, there is evidence for possible laboratory error or
contamination of the upgradient well. Upgradient populations
that do not fit a Poisson distribution are generally
inappropriate for comparisons that will be described in this
report and sources of laboratory error or contamination should be
investigated.
To test hypothesis 1, we rely on the fact that based on the
Poisson distribution, the probability of a sample with r parts
per billion is
p(r) = ur e"u r=0,1,2,... (1)
r! e=2.718
The term u in (1) is the mean of the Poisson distribution
which is equal to
N
u = E fr / N (2)
i=l
where N is the total number of samples
f is- the observed frequency of each response
4
-------
and
r is the number of parts per billion.
Based on these results, the expected number of samples with r
parts per billion is
F = NP.
For the case of 0 parts per billion we therefore have
1 part per billion
Fx = N ue~U
2 parts per billion
F = Nu2 e~u
* 2
3 parts per billion
F = N u3 e~u
(2)(3)
and 100 parts per billion
F = m u100 e~u
4 Too!
To test if the observed counts are consistent with the expected
counts, we compute the chi-square statistic
2 k 2
X = E (fi-Fi) /Fi
i=l ± x x
where k is the number of uniquely observed counts.
This chi-square statistic is distributed on k-2 degrees of
freedom.
5
-------
Question 2
Once we have obtained historical data on a population of
upgradient wells, and have demonstrated that they fit a Poisson
distribution, what is the probability that a new downgradient
observation was drawn from the upgradient population?
This second question directly adresses the central issue of
contamination. If a new downgradient observation falls within
the 95% or 99% limits of the upgradient population the well
should not be considered contaminated; however, if the new
observation lies outside of this interval (i.e. higher than the
upper bound) the well should be further examined for sources- of
contamination. Tolerance intervals for a Poisson distribution
have a particularly simple form since the mean and variance of a
Poisson distribution are identical. The 95% tolerance limit is
therefore:
= u ± 1.96Vu
9o%
and the 99% tolerance interval is
V99% = u ± 2,58 Vu
In the present situation, we are only interested in increases in
downgradient wells, therefore the one sided tolerance limit is
given by
V95% = U +
or
V99% = u + 2-33Vu
6
-------
Example
Table 1, displays the observed and expected frequency
distributions for each of the three datasets: 1) upgradient data
from 6 facilities and 29 wells, 2) 82 field blanks and 3) 65 trip
blanks. The tabled frequencies represent the combination of
levels obtained from 32 volatile organic compounds.
Table 1
Observed and Expected Frequencies and Relevant Test Statistics
ppb upgradient wells field blanks trip blanks
observed
expected
observed
expected
observed
expected
0
- 10
5016
5013.89
1983
1982.45
1770
1766.15
10
- 20
164
168.26
41
42.10
18
22.71
20
- 30
5
2.82
1
.45
0
. 15
30
- 40
0
.03
0
.00
0
. 00
40
- 50
0
.00
0
.00
0
. 00
50
- 100
0
.00
0
.00
1
.00
100
- 500
0
.00
0
.00
0
. 00
500
-
0
.00
0
. 00
0
.00
x2
1.72
0.01
0. 66
df
1
1
1
P
ns
ns
ns
upper bound
tolerance 362 ppb 358 ppb 356 ppb
limit
7
-------
*
Methylene Chloride deleted
**
Detection limit = 10 ppb (ie if all 31 compounds were not
detected we would expect 310 ppb)
The results displayed in Table 1 indicate the following.
First, all three datasets fit a Poisson distribution once
methylene chloride is deleted (ie. the chi-square statistic for
the null hypothesis that the Poisson distribution fits these
observed data was not significant indicating that the null
hypothesis could not be rejected).
Second, given a detection limit of 10 ppb, we would expect
310 ppb if all 31 compounds (ie. 32 - methylene chloride) were 'at
or below the detection limit. The upper bounds of the tolerance
limits were 362 ppb for upgradient wells, 358 ppb for field
blanks and 356 ppb for trip blanks. These findings indicate that
for all 31 compounds considered simultaneously, new dc---igradient
samples may exceed the detection limit (ie. 310 ppb) by 52 ppb
for each well at each sampling period in which all 31 compounds
are measured. For example, 3(5 compounds may be at or below the
detection limit and a single compound may exist at 62 ppb.
Alternatively, 29 compounds may be at or below the detection
limit and 2 compounds may exist at 31 ppb each. Conversely, if
all 31 compounds exhibit 12 ppb each, this would produce a total
of 372 ppb which would exceed the upper bound of the tolerance
limit and be rejected. Inspection of tolerance limits for
upgradient wells, field blanks and trip blanks revealed virtually
identical results.
8
------- |