EPA/600/R-95/110
August 1995
Characterization of Mercury Contamination
at the East Fork Poplar Creek Site, Oak Ridge, Tennessee:
A Case Study
by
C. L. Gerlach, D. Dobb, E. Miller, and D. Cardenas
Lockheed Environmental Systems & Technologies Company
Las Vegas, Nevada
A. K. Singh
University of Nevada - Las Vegas
D. Page
U.S. Department of Energy Oak Ridge, Tennessee
D. Combs
Ogden Environmental and Energy Services Company
Oak Ridge, Tennessee
E. M. Heithmar
U.S. Environmental Protection Agency
Characterization Research Division
Las Vegas, Nevada
Project Officer
K. W. Brown
Technology Support Center for Monitoring and Site Characterization
Characterization Research Division
Las Vegas, Nevada 89119
NATIONAL EXPOSURE RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U. S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
Printed on Recycled Paper
-------
NOTICE
The information in this document has been funded wholly or in part by the United States
Environmental Protection Agency under Contract No. 68-CO-0049 to the Lockheed
Environmental Systems & Technologies Company. It has been subjected to the Agency's peer
and administrative review, and has been approved for publication as an EPA document. Mention
of trade names or commercial products does not constitute endorsement or recommendation for
use.
-------
TABLE OF CONTENTS
NOTICE ii
LIST OF FIGURES iv
LIST OF TABLES v
LIST OF ABBREVIATIONS vi
ACKNOWLEDGMENTS vii
ABSTRACT 1
INTRODUCTION 2
BACKGROUND 4
PHASE Ib: SAMPLING DESIGN OPTIMIZATION BY CONDITIONAL SIMULATION . . 7
PHASE Ib: GEOSTATISTICAL EVALUATION 11
MERCURY SPECIATION TECHNOLOGY 20
CONCLUSIONS AND RECOMMENDATIONS 28
REFERENCES 30
111
-------
LIST OF FIGURES
Figure 1. An overview of the East Fork Poplar Creek site in the context of the Oak Ridge
Reservation, Tennessee 3
Figure 2. Sampling design optimization system overview 9
Figure 3. Linear loss curve for various levels of mercury 10
Figure 4. Subregion 1 sampling surface at East Fork Poplar Creek (as shaded in index map). . 13
Figure 5. Subregion 2 sampling surface at East Fork Poplar Creek (as shaded in index map). . 14
Figure 6. Subregion 3 sampling surface at East Fork Poplar Creek (as shaded in index map). . 15
Figure 7. Subregion 4 sampling surface at East Fork Poplar Creek (as shaded in index map). . 16
Figure 8. Subregion 5 sampling surface at East Fork Poplar Creek (as shaded in index map). . 17
Figure 9. Subregion 6 sampling surface at East Fork Poplar Creek (as shaded in index map). . 18
Figure 10. Subregion 7 sampling surface at East Fork Poplar Creek (as shaded in index map). 19
Figure 11. Distribution of total mercury concentrations (as determined by ICP-MS) with depth. 27
IV
-------
LIST OF TABLES
Table 1. Summary statistics for observed total mercury concentrations in the eleven
subregions at EFPC 12
Table 2. Moisture content, depth, and description of soil samples from East Fork Poplar
Creek 22
Table 3. Comparision of total mercury in dried, pulverized East Fork Poplar Creek
soil samples 23
Table 4. Distribution of mercury species in East Fork Poplar Creek soil samples 24
Table 5. Mercury concentration for each species in as-received East Fork Poplar Creek
soil samples 25
Table 6. Mercury concentration for each species in prepared East Fork Poplar Creek soil
samples 26
-------
LIST OF ABBREVIATIONS
ASV
BLUE
CERCLA
CLP
CRD-LV
CVAAS
DOE
DQO
EFPC
EMSL-LV
EPA
FP-XRF
FS
ICP-MS
LCS
LESAT
NAA
NEPA
NERL
NPL
OK
ORAU
ORD
ORNL
ORR
PARCC
PCB
PRO
QA
QAPJP
QC
RCRA
RI
RSD
SAIC
SARA
SEM
SLB
TSC-LV
XRF
anodic stripping voltametry
best linear unbiased estimator
Comprehensive Environmental Response, Compensation, and Liability Act
EPA Contract Laboratory Program
Characterization Research Division - Las Vegas
cold vapor atomic absorption spectrometry
U.S. Department of Energy
data quality objective
East Fork Poplar Creek
Environmental Monitoring Systems Laboratory-Las Vegas
U.S. Environmental Protection Agency
field portable X-ray fluorescence
feasibility study
inductively coupled plasma-mass spectrometry
laboratory control samples
Lockheed Environmental Systems and Technologies
neutron activation analysis
National Environmental Policy Act
National Exposure Research Laboratory
National Priorities List
ordinary kriging
Oak Ridge Associated Universities
Office of Research and Development (EPA)
Oak Ridge National Laboratory
Oak Ridge Reservation
precision, accuracy, representativeness, completeness, comparability
polychlorinated biphenyl
preliminary remediation goal
quality assurance
quality assurance project plan
quality control
Resource Conservation and Recovery Act
remedial investigation
relative standard deviation
Science Applications International Corporation
Superfund Amendments and Reauthorization Act
scanning electron microscopy
Sewerline Beltway
Technology Support Center - Las Vegas
X-ray fluorescence
VI
-------
ACKNOWLEDGMENTS
The authors are indebted to Mr. Craig Brown and Mr. Tony Abel, Remedial Project
Managers, U.S. EPA Region 4, and to Dr. Richard C. Metcalf, Lockheed Environmental Systems
& Technologies, for his insight into chemical phenomena. The authors gratefully commend Dr.
Wayne Tolbert, Science Applications International Corporation (SAIC), for his coordination of
the DOE's technical response and assistance in writing the second draft. Dr. Ralph Turner, Oak
Ridge National Laboratory (ORNL), provided useful comments on the mercury speciation
section. Thanks to Mr. Mark Barnett, ORNL, for his assistance in the sampling and testing of
mercury. Our gratitude to Mr. Roy Jones, U.S. EPA Region 10, and to Mr. Clark Carlson, the
Bionetics Corporation, who provided field-portable X-ray fluorescence work. Ms. Karen Daniels,
Science Applications International Corporation, provided an insightful technical review of the first
draft. Thanks also to Ms. Nancy Thomas, Lockheed Environmental Systems & Technologies
Company, for her coordination of authors' comments and Ms. Kit Peres, Lockheed Environmental
Systems & Technologies Company, for her superior editorial skills in preparing the final
manuscript.
VII
-------
-------
ABSTRACT
Concern about the adverse health effects of mercury on human and ecological life has led
to a need for more sensitive and reliable analytical and sampling procedures. Mercury
contamination, in various matrices, is present at many Superfund sites throughout the United
States.
The East Fork Poplar Creek (EFPC) Site in Oak Ridge, Tennessee, begins at, but is not
confined to, the Y-12 Plant, a U.S. Department of Energy (DOE) facility. Historic accidental
release of mercury-contaminated material associated with nuclear weapons production at the site
resulted in a complex system of stream and floodplain contamination. The EFPC is designated as
an Oak Ridge Reservation (ORR) operable unit under the Comprehensive Environmental
Response, Compensation, and Liability Act of 1983 (CERCLA) and was included on the National
Priorities List in 1989. As such, its remediation must follow specific procedures mandated by
CERCLA. The EFPC involves off-site release of contaminants from the DOE Y-12 Plant, and
therefore its remediation must also conform to the Resource Conservation and Recovery Act
(RCRA) of 1980, Section 3004(v).
Beginning in 1991, the U.S. Environmental Protection Agency (EPA) Environmental
Monitoring Systems Laboratory in Las Vegas (EMSL-LV) currently the National Exposure
Research Laboratory's Characterization Research Division (CRD-LV) assisted in the Remedial
Investigation/Feasibility Study (RI/FS) for East Fork Poplar Creek. Specifically, CRD-LV helped
to: (1) test the use of field-portable X-ray fluorescence (XRF) as a rapid field analytical method
for mercury; (2) design an optimum soil sampling strategy; (3) select and implement a method of
estimating mercury concentrations using geostatistical methods; and (4) determine the species of
mercury in the floodplain soils using an EPA developmental selective chemical extraction
procedure.
XRF proved not to be a viable field technique for this site, possibly due to the high
moisture content of EFPC soils. Sample design optimization was accomplished using an
innovative application of a Monte Carlo technique known as conditional simulation. Initial
geostatistical evaluation of the soil sample data using kriging had mixed results. Kriging was used
for interpolating contaminant concentration between data points. Improvements were made and
subsequent results using Rank-Order kriging were closer to actual data. The development of a
new selective chemical extraction method based on sequential extraction contributed to the
knowledge about the difficulties associated with mercury speciation.
-------
INTRODUCTION
The U.S. EPA's Characterization Research Division in Las Vegas (CRD-LV) assists the
EPA Regions in the characterization of Superfund and Resource Conservation and Recovery Act
(RCRA) sites through its Technology Support Center (TSC). The CRD-LV TSC is part of the
technical support project that was formed in 1987 by the Office of Research and Development,
Office of Solid Waste and Emergency Response and all of the Regional offices. The function of
the CRD-LV TSC is to provide assistance and ORD technical expertise to Regional personnel in
characterizing site contaminants for remediation.
In 1991, the EPA Region 4, Remedial Project Manager requested the assistance of the
TSC-LV in the Remedial Investigation/Feasibility Study (RI/FS) for East Fork Poplar Creek
(EFPC) (Brown, 1991). Specifically, TSC-LV was asked to help: (1) test the use of field-portable
X-ray fluorescence (XRF) as a rapid field analytical method for mercury; (2) design an optimum
soil sampling strategy; (3) select and implement a method of estimating mercury concentrations
using geostatistical methods; and (4) determine the species of mercury in the floodplain soils using
an EPA developmental selective chemical extraction procedure. From 1991 through early 1994,
TSC-LV assisted the Region and DOE with these technical challenges, beginning with sample
design optimization and ending with mercury speciation. The results of the four related efforts
were intended to be used at various stages in the development of the RI for EFPC.
-------
CO
III
I 1
O
0)
PH
O
-t->
zn
UJ
-------
BACKGROUND
The East Fork Poplar Creek (EFPC) operable unit includes approximately 14.5 miles of
the EFPC, its floodplain, and the Oak Ridge Sewerline Beltway (SLB). The area of concern
(Figure 1) begins at the Y-12 Plant, and continues through the city of Oak Ridge until its
convergence with Poplar Creek back on the Oak Ridge Reservation. Both EFPC and its
floodplain were contaminated by releases of mercury from the Y-12 Plant, primarily during the
1950s and 1960s.
The EFPC is a perennial stream located in Anderson and Roane Counties in Oak Ridge,
Tennessee, approximately 25 miles west of Knoxville. Its headwaters are contained in 54 to 72
inch underground collection pipes that extend from the west end to the central area of the Y-12
Plant, where the above-ground portion of the creek begins. From the Y-12 Plant site, EFPC
flows northward through a gap in Pine Ridge and enters Gamble Valley and the city of Oak
Ridge. From there, the stream flows in a northwesterly manner along Illinois Avenue through
commercial and light industrial areas in Oak Ridge, then trends generally westward, parallel to
Oak Ridge Turnpike in East Fork Valley through primarily residential, agricultural, and
undeveloped forest areas, until it joins Poplar Creek. EFPC waters, after entering Poplar Creek,
flow into a river, which is impounded behind Watts Bar Dam.
The EFPC site is included on the National Priorities List (NPL) as part of the Oak Ridge
Reservation and its environs and, as such, its remediation must follow specific procedures
mandated by the Comprehensive Environmental Response, Compensation, and Liability Act
(CERCLA). EFPC involves, off-site release of contaminants from the Y-12 Plant and, therefore,
its remediation must also conform to the procedures of Section 3004(v) of the RCRA. Actions
taken to remediate EFPC may affect the environment, and the potential environmental impact of
those actions must be publicly addressed in accordance with the values stated in the National
Environmental Policy Act (NEPA). The remedial investigation (RI) of EFPC has integrated the
requirements of these three primary federal regulations as well as those of other federal and
Tennessee state regulations.
The primary steps included in a CERCLA RI are:
to collect data to characterize site conditions,
to determine the nature and extent of contamination, and
to assess current and future risks to human health and the environment, if
no remediation occurs.
The first two steps are generally referred to as the site characterization, and the third step
is termed the baseline human health and ecological risk assessment. These assessments are
described in detail in the Risk Assessment Guidance for Superfund U.S. EPA Vol. 1, Parts A
and B, 1989 and 1991. RCRA and NEPA investigations call for similar activities. After the RI is
completed, work begins on the feasibility study (FS).
-------
The EFPC RI was conducted in two segments, Phase la and Phase Ib (DOE 1994). Phase
la was designed to determine the nature of contamination and to identify the contaminants of
potential concern - primarily mercury and other metals, radionuclides, and various organic
compounds including polychlorinated biphenyls (PCBs) and pesticides. Completion of this phase
of the RI required the installation of 12 new groundwater monitoring wells and the quarterly
sampling of 22 wells to characterize the groundwater quality and the hydrogeology of the
floodplain. More than 500 surface water, creek sediment, and surface and subsurface soil samples
were taken along the watershed and at an uncontaminated reference site during Phase la. The
samples were analyzed for 182 inorganic, organic, and radionuclide analytes, with mercury being
the primary contaminant of concern. In addition, geotechnical parameters and soil chemistry tests
were performed that relate to treatability studies. Phase Ib was designed to establish the level
and extent of contamination. More than 3,000 field samples were collected and analyzed, and
soil-gas surveys were conducted to monitor mercury volatilization during Phase Ib.
Of all the media at the EFPC, the floodplain soils have the largest volume and highest
concentration of contaminants. The Phase la data demonstrated that mercury accounts for
approximately 85% of the total risk attributable to metals in the soil. A number of analyses were
performed to detect inorganic, organic, and radionuclide compounds. An elevated level of
mercury was observed when high concentrations of other contaminants were present. Mercury
was found to be a surrogate for the presence of other metals and radionuclides, and was used as
such to determine the distribution of other inorganic and even organic contaminants.
Risk potential played a critical role in the decisions made about remediation and,
therefore, it was important to establish pertinent and appropriate data quality objectives (DQOs).
Both qualitative and quantitative DQOs were established for the EFPC RI to meet the objectives
of precision, accuracy, representativeness, completeness, and comparability (PARCC). Analytical
precision and accuracy were controlled by adopting EPA Contract Laboratory Program (CLP)
criteria and quality control (QC) frequency and types. Data verification and validation of the
resulting analytical data packages indicated the quality of the data produced by the laboratories.
Sampling precision was evaluated by the use of colocated samples (field replicates) and split
samples (field duplicates). Representativeness of data from the EFPC RI was accomplished by
selecting sampling methods and performing repetitive sampling events to accurately represent the
characteristic population. Intervals for soil sampling were chosen based on historical site
knowledge and geostatistical premises to obtain the strata with the highest concentration of
contaminants in order to achieve the most conservative representation and to optimize the
number of samples required, avoiding the wasted expense of non-detects. DQOs for
completeness were set at 90% for laboratory completeness for both Phase la and Phase Ib.
Percent completeness for field sampling was established at 90% for Phase la and at 70% for
Phase Ib. To achieve comparability, the EFPC RI used one laboratory to perform its CLP
analyses and applied the same sampling method.
Data uses were identified during the DQO process to include primary input into the site
characterization study, the human health and ecological risk assessments, the initial screening of
alternatives, and the FS. The DOE wanted to obtain the highest quality data possible. The data
collection program was delineated in a sampling and analysis plan for each phase of the RI.
-------
After completion of Phase la, it was obvious that the variability associated with the
distribution of mercury combined with very high analytical costs associated with Level 4 data
would become limiting factors. At this point, TSC-LV assistance was requested in optimizing an
appropriate sampling strategy that would be sufficient to cover nearly 670 acres of floodplain soil.
Concurrent with a sampling design, faster and less expensive techniques for rapid analysis of
mercury were needed.
Phase Ib began with a search for the best analytical method for accurately determining
mercury concentration. In 1991, CRD-LV technical staff, EPA Region 10 personnel, and Science
Applications International Corporation (SAIC; a DOE contractor), tested field-portable XRF
(FPXRF) equipment on-site at EFPC. Similar equipment had proven successful in detecting
metals, including mercury at low concentrations, at some western U.S. sites. Seventeen soil
samples from EFPC were analyzed by CRD-LV using laboratory-based XRF and cold vapor
atomic absorption spectroscopy (CVAAS) to establish calibration standards for the FPXRF
(Hillman, 1991). The field equipment was then tested under actual field conditions at EFPC.
The results of these field tests indicated that there was great disagreement between FPXRF and
CVAAS. It is speculated that the mesic (moist) conditions and other matrix effects associated
with EFPC soil under field conditions caused interferences and resulted in rejection of XRF as a
rapid analytical method for this site. Therefore, in Phase Ib, the DOE relied upon neutron
activation analysis (NAA), a non-CLP method, for the large-scale soil analysis of the floodplain
soil samples to determine the extent and distribution of contamination. DQOs for NAA were
established to meet the precision and accuracy of CLP methods but with an exception for
analytical sensitivity. The desired lower limit of detection for mercury in soils was 11 mg/Kg. This
limit was specified based on previous remediation goals established by the State of Tennessee for
mercury in EFPC soils. DOE demonstrated the equivalency of NAA to CLP methods and this
approach was accepted by the regulators.
-------
PHASE Ib: SAMPLING DESIGN OPTIMIZATION BY CONDITIONAL SIMULATION
The EFPC presented a complex sampling challenge to statisticians. The most desirable
sampling plan would maximize the information obtained and minimize the number of samples
taken. Sampling plan optimization was achieved using a geostatistical approach (kriging) and a
novel application of Monte Carlo simulation call conditional simulation (Englund and Heravi,
1992).
Monte Carlo simulation is a statistical technique that imitates a real-world situation and
uses repetitive iterations to solve complex problems that cannot be solved analytically. In the case
of contaminant concentration data collected from different geographical locations at a site, a
modified version, called conditional simulation, is used. Conditional sumulation is a method that
generates realizations for a study from the available data and a "correlation structure model" that
defines spatial dependency. A computerized site model is created by generating a dense grid of
simulated sample points with contaminant values ("true points") that are consistent with the
existing data. The study area is then divided into discrete blocks whose contaminant
concentrations are estimated by the geostatistical method referred to as kriging. Kriging is a
weighted moving average method used to interpolate values from a data set onto a contouring
grid. The kriging weights are computed from a variogram, which measures the correlation among
sample values as a function of the distance and direction between samples. "True" block means
from the simulation are used for comparison to evaluate the effectiveness of a particular sampling
scheme.
The site model described above is repeatedly sampled in a Monte Carlo fashion for each
potential sampling scheme. In this case, the only difference between schemes was the number of
samples. Each sampling iteration is used to compute block means by kriging. The block means
computed from sampling are then compared against block estimates of the "true" points. The
number of "detects" of blocks correctly identified as being above the action limit, the number of
false negatives and positives, and the number of samples taken are recorded for each scenario.
From these data, the cost of additional samples can be weighed against the resulting improvement
in performance.
After considering the heterogeneity of contaminant concentrations within the soil medium
as indicated by the results of the Phase la soil investigation, a sampling plan was developed that
defined the extent and distribution of the primary contaminants throughout 670 acres of the
EFPC floodplain. Historical data DOE (1994) was used as input to the conditional simulation
software. These data include mercury analysis of 321 samples.
-------
An overview of the sampling design employed by the CRD-LV is displayed in Figure 2. A
semivariogram (model for spatial dependence) was computed from the historical data, and a
conditional simulation was performed in which a representative map was constructed with tens of
thousands of data points representing mercury values. This simulated data set represented the
reality against which multiple sampling scenarios were tested. Sampling scenarios ranged from
100 to 4,096 samples randomly distributed throughout the reality study area. The sampling results
were block-kriged and compared to block-kriging results for the simulated map.
A linear loss function provides a means of estimating the "point of diminishing returns",
where the value of the improved decision performance obtained from additional sampling is no
longer worth the additional sampling cost. The linear loss function had three cost components:
(1) sampling cost, which is simply the cost per sample and is linear with the number of samples
taken; (2) remediation cost, which is an estimated cost to clean up each contaminated block; and
(3) linear cost, which is an estimate of the societal/environmental cost associated with neglecting
to identity and remediate a contaminated block. From this curve, the number of samples yielding
the minimum total cost is obtained. Figure 3 shows linear loss curves for various levels of
mercury. Within the range of the action levels considered in the simulation program, the optimal
number of samples appears to be independent of the action level. The optimal number of
samples is 484 for the study area, which translates to about a 20-m grid size. Thus, the
geostatistical methods employed assisted in determining sampling and data analysis optimization
for the EFPC study.
-------
pilot study *
. . -s.
date
mmuJnaum m***»mm*»*
»**»». »g.l»j5*42I!SI
« 1 | 1 y H MM WIN** MMP "
? 7 rr i
-do." ' " "ab."
2*>. ato
six as
j
vanogram
model
Conditional Simulation
i
Simulation Map Mercury (ppm)
-> Simulated Sampling
148.:
W.J
4S.J
-S.J
6S.J
-108.
' ''m * »."»».
Bf» ir ^. » M.
** **' Mk2 ''*' ***«.
«*** 1M. "*»JW "flj."0- ',* **' W.
" TJ. a*5**' i*. 4. ' «
1 »M- * . V».
>. M. > °' 0.°- 40. |
-*. ab. 180. 860. 380.
Block Estimation
Krig0d Estimate* Mercury (ppm) Block AveraoM-Mercury (pom)
<-Loop
Simulated Remediation
and Cost Evaluation
Linear Cost Curve Misdassification Histogram
I-
False Negative Map
225 Samples
Figure 2, Sampling design optimization system overview.
-------
1800
1700-^
1600-^
1500-
g 1400-
o
1300-
1200-
1100-
1000
0 1000 2000 3000 4000
Number of Samples
5000
Figure 3. Linear loss curve for various levels of mercury.
One disadvantage of conditional simulation modeling is that it optimizes for a single round
of sampling based on a given set of data. An underlying assumption is that the distribution and
contaminant levels are consistent throughout the site and can be defined with a single sampling
scenario derived from the model output. As it was not feasible to sample the entire area within
the 100-year floodplain on a 20-m grid (65-ft), a compromise was agreed upon to devise a
sampling plan based upon the results of the conditional simulation. The next step was to use a
systematic sampling design where soil sampling transects were situated across the floodplain. The
sample distance between points across the floodplain was 20 m (65 ft), and the transects were
placed every 100 m (328 ft) along the length of the creek. A total of 159 transects were sampled
over the length of the creek. Soil samples were obtained at the edge of the creek and every 20 m
(65 ft) away from the creek until the elevation of the 100-year flood event was reached. Surface
soil cores from 0 to 40.6 cm (0 to 16 in.) were collected at each station on the transects and each
core was composited in a stainless steel bowl prior to the sample being placed in a 250-mL glass
jar. At the even-numbered transects (i.e., every other transect), second- and third-layer samples
were attempted (see figures 4-10 for the locations of transects and sampling locations). Due to
the physical difficulties encountered in collecting samples from 40.6 to 81 cm (16 to 32 in.) and 81
to 122 cm (32 to 48 in.), the number of samples taken declined with depth. However, sufficient
data were obtained to determine the vertical distribution of contaminants.
10
-------
PHAS5E Ib: GEOSTATISTICAL EVALUATION
The EFPC RI employed the geostatistical method of kriging for spatial evaluation of
contaminant concentrations. Geostatistics is a methodology used for the analysis of spatially
correlated data. Its characteristic feature is the use of a semivariogram or related technique to
quantify and model the spatial correlation structure. Kriging is then implemented using the
selected mathematical function for the semivariogram as a weighted-moving average interpolator.
Kriging produces estimates of contaminant concentration at a point or average concentration over
a block. These estimates are linear functions of observed concentrations and have minimum
estimation variance. This is often referred to as best linear unbiased estimator (BLUE).
The semivariogram modeling and kriging of the data was performed by the TSC-LV. A
validated data set was supplied to TSC-LV from the NAA results of the Phase Ib soil transect
samples. EFPC was subdivided into 11 sections to compensate for the sinuous nature of the
creek and to permit processing; subsets of about 3,000 samples. The summary statistics for total
mercury concentration in the 11 sections (subregions) of the site are shown in Table 1.
The software package Geo-EAS, developed by CRD-LV, was used for semivariogram
modeling and kriging of the data. This software package performs Ordinary kriging (OK). In
OK, it is assumed that the concentration data have no spatial trends (e.g., an increasing or
decreasing contaminant concentration trend over the site in question). The symmetry of data
distribution is an implicit assumption. The data distribution was not normally distributed but was
heavily skewed and, therefore, log transformation was used to obtain a symmetric distribution.
The estimates of block averages, obtained by using log-normal OK, were then back-transformed
to mercury concentrations. The back-transformation formula involves not only the block-averages
of log-data, but also their kriging standard deviation. For this reason, the initial attempt at log-
normal OK resulted in back-transformed block averages that were much higher than the observed
maximum concentrations of mercury in areas farthest from the actual sample locations. The
method of Rank-Order kriging, developed by Singh, et al. (1993) was then used. This method is a
nonparametric statistical method and is accomplished by sorting the concentration data by their
ordered rank from 1 to n. A variogram model is then fitted to the rank data, and OK is
performed on the concentration data using the rank variogram model. The kriging variance is
multiplied by a correction factor (equal to the variance of concentration data/variance of rank
data). The variance of the concentration distribution is estimated from the raw concentration
data.
Figures 4 through 10 display the average surface mercury concentrations for 7 of the 11
subregions in 20-m blocks overlain on the hand-drawn estimates of the distribution of mercury for
50 and 200 ppm levels (only surface concentrations were kriged). The class boundaries (i.e.,
cutoffs) for mercury are taken from the preliminary remediation goals (PRGs) which are based on
risk to human health.
Based on a comparison of the kriged results to actual field measurements, the DOE
decided not to rely solely on the kriging results to define the extent of contamination. Some
results showed poor predictions, possibly due to the large area of "edge" associated with the sides
of the creek. Background levels of inorganics in soils and sediments by Breckenridge and
Crockett, 1995 provides a useful insight into the complexity of sites with widely varying
contaminant concentrations. Kriging, however, was a valuable tool for interpolating contaminant
concentration between data points, and the final results better reflected the actual data.
11
-------
TABLE 1. SUMMARY STATISTICS FOR OBSERVED TOTAL MERCURY
CONCENTRATIONS IN THE ELEVEN SUBREGIONS AT EFPC
Subregion
Subregion 1
Subregion 2
Subregion 3
Subregion 4
Subregion 5
Subregion 6
Subregion 7
Subregion 8
Subregion 9
Subregion 10
Subregion 11
Number
of
samples
(n)
135
148
148
140
148
148
143
239
148
150
47
Minimum
(ppm)
1.4
0.4
0.9
0.5
0.6
0.8
0.4
0.4
0.5
0.5
0.7
Maximum
(ppm)
216.0
186.0
99.0
103.0
10902.2
180.0
198.0
1070.0
387.0
1320.0
1170.0
Mean
(ppm)
18.6
9.9
4.5
7.6
93.7
15.3
11.7
49.5
37.0
51.7
43.1
Standard
deviation
(ppm)
37.9
24.7
11.3
16.4
898.9
34.1
26.5
139.5
68.3
153.0
180.0
12
-------
EAST FORK POPLAR CREEK (OAK RIDGE)
Surface Mercury Contamination within the 100 Year Floodplain
M Marcury 50 200 PPM
Marcury > 200 PPM
Z] Sampto 810s
Z3 Road*
SI CrMk AIYfeutarlM
IndaxMap
Figure 4. Subregion 1 sampling surface at East Fork Poplar Creek (as shaded in index map).
13
-------
EAST FORK POPLAR CREEK (OAK RIDGE)
Surface Mercury Contamination within the 100 Year Floodplain
1- /
M MffCIvy 50 - 200 PPM
Mtrcury > 200 PPM
ED SvnptoSttt
E3 RoKto
G3 CrMk & THbutartos
ImtocMap
Figure 5. Subregion 2 sampling surface at East Fork Poplar Creek (as shaded in index map).
14
-------
EAST FORK POPLAR CREEK (OAK RIDGE)
Surface Mercury Contamination within the 100 Year Floodplain
M Mvcury 50 - 200 PPM
Mercury > 200 PPM
Z) Stmpto Stet
SI Roach
22 Creak ATttoutartes
Z3 Roodptaln
bidwMap
Figure 6. Subregion 3 sampling surface at East Fork Poplar Creek (as shaded in index map).
15
-------
EAST FORK POPLAR CREEK (OAK RIDGE)
Surface Mercury Contamination within the 100 Year Floodplain
H Mtrcury 50 - 200 PPM
Mercury > 200 PPM
CD Sampto SIM
El Roads
El Creak & TMbutartes
CD Floodplain
Figure 7. Subregion 4 sampling surface at East Fork Poplar Creek (as shaded in index map)
16
-------
EAST FORK POPLAR CREEK (OAK RIDGE)
Surface Mercury Contamination within the 100 Year Floodplain
mi Mvcury 50 - 200 PPM
Mtrcury > 200 PPM
ED Sarnpto Stes
O ROMfc
E3 Creak & TTtouurtes
CD Roodpum
Figure 8. Subregion 5 sampling surface at East Fork Poplar Creek (as shaded in index map).
17
-------
EAST FORK POPLAR CREEK (OAK RIDGE)
Surface Mercury Contamination within the 100 Year Floodplain
H Mtrcury 60 - 200 PPM
Mercury > 200 PPM
CD Sample Stt*
O Ro«to
E2 OrMk ITtbuUries
C3 Hoo^ptaln
Mot Map
Figure 9. Subregion 6 sampling surface at East Fork Poplar Creek (as shaded in index map).
18
-------
EAST FORK POPLAR CREEK (OAK RIDGE)
Surface Mercury Contamination within the 100 Year Floodplain
f " /
&=3 /
M Mtrcury 50 200 PPM
Mercury > 200 PPM
CD Sampto Stf*
\ffl Roatto
1221 CrMk
IndoxMap
Figure 10. Subregion 7 sampling surface at East Fork Poplar Creek (as shaded in index map).
19
-------
MERCURY SPECIATION TECHNOLOGY
The relative toxicity of the different species of mercury makes speciation methods
necessary and, in the long run, cost effective. Thoughtful and safe remediation efforts can be
based on the results of speciation methods. When mercury is present, it is important to know if it
is present in the highly toxic, and bioavailable, organic form or in the most stable, and least
bioavailable form, mercury sulfide. Mercury speciation technology, based on sequential selective
extraction, is the process whereby the various forms, or species, of mercury are separated so that
individual quantitative results may be obtained. There are currently at least three sequential
extraction methods used for mercury speciation (Revis et al., 1989; Sakamoto et al., 1992; Miller
et al., 1994). One extraction method, which is under development by CRD-LV scientists, was
tested in response to the analytical requirements of the EFPC Site. The innovation of the CRD-
LV method is twofold: it is able to separate more species individually, and it does not rely upon
heating, which can drive off more than one species.
In 1994, 20 additional soil samples were taken from the EFPC site and sent to the CRD-
LV for total mercury and mercury species analysis. Mercury species were sequentially removed, in
order of increasing stability and decreasing toxicity, by exposing a soil sample to solvents of
increasing acidity and oxidizing strength. The residue from each extraction step was separated by
centrifugation and subjected to the next extraction step. Extracts were analyzed by inductively
coupled plasma-mass spectrometry (ICP-MS) With confirmatory analyses by anodic stripping
voltametry (ASV). Step 1 is extraction with toluene, which removes organo-mercury compounds
such as methyl mercury. Step 2 is extraction of water-soluble compounds including mercuric
chloride. Step 3 is extraction with dilute nitric acid, which removes compounds like mercuric
oxide and mercuric sulfate. Step 4 is extraction with strong nitric acid, which removes mercury
metal and amalgams. The final step is extraction in modified aqua regia, which removes the least
soluble compound: mercuric sulfide. For a complete discussion of the method, see Miller, 1993.
The method development and application followed the guidelines set forth in the quality
assurance project plan (Ecker et al, 1994). Total mercury analyses were determined by
summation of individual compounds and confirmed using CVAAS and ICP-MS. Complete
extraction and separation of mercury compounds were confirmed through spike recoveries and the
analysis of spent residues for residual mercury by XRF.
Samples were taken at two depth ranges: shallow (0-3 inches), and varying deeper depths
(down to 32 inches) (Table 2). These samples (Table 3) were found to contain total mercury
concentrations between 20 mg/Kg and 3000 mg/Kg. The deeper soil samples generally contained
higher concentrations of mercury. In most of the shallow samples, the selective extraction
indicated that 80 to 90 percent of the total mercury was present primarily as the metal or as an
amalgam. In deeper samples, the CRD-LV method indicated that mercury was predominantly
mercury metal and mercury sulfide in approximately equal proportions. Minor, but significant,
amounts of acid-soluble mercury compounds (probably mercury oxide) were seen in most of the
samples. In two of the samples, mercury oxide was the predominant form of mercury.
The analytical results indicate that three sample parameters are strongly correlated: sample
depth, total mercury concentration, and the ratio of mercury sulfide to metallic mercury (tables 2
through 6). Specifically, these parameters tend to be correlated: deeper samples have greater
mercury concentrations and higher ratios of mercury sulfide to mercury metal. The correlation
20
-------
between greater depth and higher total mercury concentrations is particularly strong, with nine of
the ten sample pairs containing ten times as much mercury at greater depth. At these depths
there was no decrease in contamination. The extent of the contamination with increasing depth
could not be definitely determined because mercury was detected in all samples. In the single
apparently aberrant sample pair, ZE3770810 and ZE3770822, the shallower sample has a total
mercury concentration approximately ten times that of the deeper sample. However, subsequent
discussion with ORNL personnel indicated that this sample pair was taken at a point where the
creek bank had eroded and the overhanging embankment had collapsed so that the original strata
were reversed: the "deep" soil was at the surface and the original surface soil was buried. This
could explain the reversal in the mercury measurement data. If this is the case, then all samples
exhibited higher mercury levels with greater depth. Figure 11 shows the correlation between
depth and total mercury concentrations as determined by ICP-MS. The depth ranges for Figure
11 are shown in Table 2, and the summed total ICP-MS data used in the figure are shown in
Table 3.
Total mercury concentrations were determined by four methods and, for three samples, by
five methods (Table 3). Methods included: strong-acid digestion followed by ICP-MS; acidic
permanganate/persulfate digestion followed by CVAAS, EPA Method 7471 SW-846, summation
of results from the ICP-MS analysis of the speciation extracts; summation of results of the ASV
analysis of the speciation extracts; and summation for results for three samples by XRF analysis.
The ASV technology only responds to dissolved species. The ICP-MS detects total mercury
available in a sample, dissolved as well as suspended colloidal compounds. The agreement among
method results is evidence that the extracts analyzed by ICP-MS only contained dissolved species.
This is also evidence that centrifugation efficiently separated dissolved species from undissolved
species. Concentrations of individual mercury species as determined by ICP-MS were in good
agreement with the values obtained by ASV. Although a few discrepancies were noted,
agreement was good among all the methods, indicating that the results for total mercury are
accurate. It is believed that sampling error is a major contributor to most disparate results
between duplicate samples. Details on the analytical results from the study can be found in Dobb
et al, 1994.
With a few exceptions, the results of quality control (QC) sample analyses conformed to
the requirements specified in the quality assurance project plan (QAPP) associated with the task.
The only exceptions occurred in the replicate precision where two questionable values were
generated by ICP-MS. The two extracts in question were reanalyzed by ASV and the results,
which indicated that there was no precision problem, were confirmed. Overall, method precision
was near 10 to 15 percent relative standard deviation (RSD). Speciated matrix spike and
laboratory control sample (LCS) results indicated a satisfactory recovery of all mercury species
from both spiked samples and the spiked blank. Recoveries of 102 to 112 percent were typical.
Preparation blanks demonstrated a satisfactory freedom from positive interferences for all mercury
species. All QC samples for total mercury determinations were satisfactory.
21
-------
TABLE 2. MOISTURE CONTENT, DEPTH, AND DESCRIPTION OF SOIL SAMPLES FROM EAST
FORK POPLAR CREEK.
Soil Sample
Number
ZE3181013
ZE3 181025
ZE3740918
ZE3740920
ZE3770810
ZE3770822
ZE5150717
ZE5150729
ZE5340619
ZE5340621
ZE5380512
ZE5380524
ZE5470412
ZE5470424
ZN3210115
ZN3210127
ZN3340211
ZN3340223
ZN3340312
ZN3340324
Moisture
Content,
Percent (%)
25.9
22.0
23.4
26.8
20.9
21.4
23.4
30.8
26.0
28.9
22.3
42.4
27.1
35.1
26.7
33.0
24.4
35.6
30.0
45.8
Depth
(inches)
0-3
16-32
0-3
8-12
0-3
10-16
0-3
5-8
0-3
10-16
0-3
10-14
0-3
10-20
0-3
13-19
0-3
5-16
0-3
3-9
Description
Dark Brown, Medium Texture, Artifacts (roots)
Reddish Brown, Fine Texture (clay)
Dark Brown, Medium Texture, Artifacts (roots)
Dark Brown, Medium Texture
Dark Brown, Medium Texture, Artifacts (roots)
Light Brown, Medium Texture
Dark Brown, Medium Texture, Artifacts (roots)
Dark Clay, Medium Texture
Dark Brown, Medium Texture
Dark Brownish-Gray, Medium Texture
Brown, Fine Texture (clay)
Dark Gray, Medium Texture (clay)
Dark Brown, Medium Texture
Dark Gray, Medium Texture, Artifacts (roots)
Grayish-Brown, Medium Texture, Artifacts (roots)
Brownish-gray, Medium Texture
Grayish-Brown, Medium Texture
Brownish-Gray, Medium Texture
Grayish-Brown, Medium Texture
Dark Gray, Medium Texture, Artifacts (roots)
22
-------
TABLE 3. COMPARISION OF TOTAL MERCURY IN DRIED, PULVERIZED EAST
FORK POPLAR CREEK SOIL SAMPLES.
Soil
Sample
ZE3181013
ZE3181025
ZE3740918
ZE3740920
ZE3770810
ZE3770822
ZE5150717
ZE5150729
ZE5340619
ZE5340621
ZE5380512
ZE5380524
ZE5470412
ZE5470424
ZN3210115
ZN3210127
ZN3340211
ZN3340223
ZN3340312
ZN3340324
Total Mercury (mg/Kg)
ASV
Summed
Total
35.0
483.0
102.0
782.0
528.0
114.0
237.0
895.0
192.0
1,310.0
70.0
1,730.0
99.0
1,100.0
247.0
2,250.0
224.0
1,860.0
201.0
1,810.0
ICP-MS
Summed
Total
34.0
476.0
82.9
812.0
581.0
26.9
243.0
1,000.0
189.0
1,670.0
69.2
1,960.0
75.5
1,230.0
326.0
3,140.0
350.0
2,050.0
305.0
2,420.0
ICP-MS
Measured
27.8
442.0
64.7
339.0
281.0
23.8
217.0
(a)
157.0
1,670.0
73.5
2,090.0
96.4
1,060.0
294.0
3,300.0
301.0
2,130.0
247.0
2,630.0
CVAAS
Measured
29.2
376.0
56.2
589.0
454.0
21.4
183.0
804.0
129.0
1,360.0
64.8
1,830.0
85.6
1,190.0
258.0
2,850.0
281.0
2,020.0
226.0
2,060.0
XRF
Measured
864(b)
2,830
2,360
(a) uuring microwave digestion, this sample generated gases causing the container pressure-relief
valve to vent, resulting in loss of sample. Repeated digestion attempts were also unsuccessful.
(b) Values are extrapolated beyond the XRF calibration curve high standard.
23
-------
TABLE 4. DISTRIBUTION OF MERCURY SPECIES IN EAST FORK POPLAR CREEK SOIL
SAMPLES.
Species Distribution, as Percentages of the Total Summed Mercury
Soil Sample
ZE3181013
ZE3181025
ZE3740918
ZE3740920
ZE3770810
ZE3770822
ZE5150717
ZE5 150729
ZE5340619
ZE5340621
ZE5380512
ZE5380524
ZE5470412
ZE5470424
ZN3210115
ZN3210127
ZN3340211
ZN3340223
ZN3340312
ZN3340324
Organic
<0.1
<0.1
<().!
<().!
<().!
<().!
<().!
<().!
<().!
<().!
<().!
<0.1
<().!
<0.1
<0.1
<0.1
<0.1
<0.1
<0.1
<0.1
Water-Soluble
0.1
1.0
0.1
0.3
<0.1
0.4
0.1
0.1
0.2
<0.1
0.1
<0.1
<0.1
0.1
0.1
0.3
<0.1
<0.1
<0.1
<0.1
Acid-Soluble
71
25
49
13
11
19
11
11
3.9
0.7
1.1
3.2
2.8
14
19
7.1
<0.1
5.8
<0.1
6.2
Metallic
21
36
26
66
81
76
83
56
90
57
91
32
88
49
74
35
92
25
94
36
Sulfide
7.7
38
26
20
7.6
5.4
5.8
33
6.2
42
7.6
65
8.9
37
7.2
57
8.4
69
6.5
57
24
-------
TABLE 5. MERCURY CONCENTRATION FOR EACH SPECIES IN AS-RECEIVED EAST FORK
POPLAR CREEK SOIL SAMPLES.
Soil Sample
ZE3181013
ZE3181025
ZE3740918
ZE3740920
ZE3770810
ZE3770822
ZE5150717
ZE5 150729
ZE5340619
ZE5340621
ZE5380512
ZE5380524
ZE5470412
ZE5470424
ZN3210115
ZN3210127
ZN3340211
ZN3340223
ZN3340312
ZN3340324
Species Concentration (mg/Kg Mercury)
Organic
<0.003
<0.003
0.005
0.004
O.003
<0.003
0.008
0.004
0.006
<0.003
0.002
<0.001
<0.002
<0.001
<0.002
<0.001
<0.002
<0.001
O.001
<0.001
Water-Soluble
0.019
3.88
0.059
1.85
0.028
0.094
0.221
0.347
0.264
0.146
0.041
0.123
0.016
0.520
0.166
6.90
0.058
0.077
0.038
0.163
Acid-Soluble
18.0
93.6
30.9
79.8
51.9
3.92
20.1
77.5
5.48
8.60
0.611
36.5
1.52
111
45.7
145
0.039
76.0
0.045
81.3
Metallic
5.26
135
16.3
395
373
16.0
155
386
126
675
49.0
357
48.6
390
176
717
242
327
200
477
Sulfide
1.95
140
16.2
118
35.0
1.13
10.9
23.0
8.66
501
4.07
737
4.91
295
17.2
1,170
22.2
914
13.8
753
25
-------
TABLE 6. MERCURY CONCENTRATION FOR EACH SPECIES IN PREPARED EAST FORK
POPLAR CREEK SOIL SAMPLES.
(a
Soil
Sample
ZE31810
ZE31810
ZE37409
ZE37409
ZE37708
ZE37708
ZE51507
ZE51507
ZE53406
ZE53406
ZE53805
ZE53805
ZE54704
ZE54704
ZN32101
ZN32101
ZN33402
ZN33402
ZN33403
ZN33403
Species Concentration (mg/Kg Mercury)
Organic
<0.004
<0.004
OJ007
0.005
< 0.004
< 0.004
0.011
0.006
0.008
<0.004
0.003
<0.002
<0.002
<0.002
<0.002
<0.002
-------
o
^
Jm
O
_c
c
15
0
(/>
-O-i
-1
-2-
-3-
-4-
-5-
-6-
-7-
-8-
-9-
-10-
-11-
-12-
-13-
-14-
-15-
-16-
-17-
-18-
-19-
-20-
-21-
-22-
-23-
-24-
-25-
-26-
-27-
-28-
-29-
-30-
31-
32-
Total M«rcury Concentrations in mg/kg
1001-2000
2001-3000
^Sample number and concentration correspond to data in Table 3.
Note that shallow samples are 0-3 inches and deeper samples vary in depth.
Figure 11. Distribution of total mercury concentrations (as determined by ICP-MS) with depth.
27
-------
CONCLUSIONS AND RECOMMENDATIONS
The major aspects of these technology applications are noted below:
XRF can be used as a confirmatory device. However, in 1991, at the time of the ORNL field-
samples analysis, FPXRF was not a viable method for analyzing EFPC soils in the field for mercury.
Caution should be used when applying FPXRF to the analysis of metals in moist soils.
Conditional simulation was demonstrated to be a valuable method that optimized sampling
efforts and saved time and costs while: maintaining the integrity of the data. The successful use of
conditional simulation at the EFPC site illustrates the pertinence of the Technology Support Center as
liaison between research and practical field applications.
The use of kriging at the EFPC site is an example of a geostatistical process applied to
characterizing contaminants at a hazardous waste site. The use of semivariograms, Ordinary and
Rank-Order kriging, and other geostatistical methods is being applied by environmental scientists. The
results of the work at the EFPC site are further proof of the validity and strength of these techniques.
The use of geostatistics at the EFPC was integrated with the conceptual model of contaminant
distribution, and all factors were considered in the final interpretation. A combination of geostatistics
and professional judgment was employed to provide the final analysis.
The CRD-LV mercury speciation method gave good results and was successfully applied to the
EFPC matrix. The deeper samples from the EFPC site had greater mercury concentrations. Almost
universally, ratios of mercury sulfide to mercury metal were in approximately equal proportions. In
most of the shallow samples, the selective extraction indicated that 80 to 90 percent of the total
mercury present was predominately mercury metal or an amalgam.
The CRD-LV method gives reliable information about mercury concentrations, and the
strength of the speciation technology was validated by the results from the EFPC study. This method
provides a new way to determine and quantify mercury species which can be verified through total
mercury concentrations tests.
Knowledge of mercury metal speciation not only provides a more realistic assessment of risk
but also assists environmental decision-makers in determining the most appropriate remediation
technologies. For example, when mercury oxide or mercury sulfate are present, remediation by
volatilization can be a problem. After volatilization at temperatures sufficient to remove most other
mercury species, significant amounts of the bioavailable and highly toxic mercury oxide or mercury
sulfate can remain. The higher temperatures needed to remove mercury oxide and mercury sulfate
will present practical problems.
This mercury speciation method is particularly valuable because the species are sequentially
released in order of decreasing toxicity by solvents of increasing harshness. Therefore, levels of
toxicity can be identified allowing scientists to better ascertain the degree of contamination. The
method is relatively fast and simple allowing delivery of a preliminary site assessment in a shorter
period of time.
28
-------
The EPA and DOE worked in tandem on this project and, together, were successful in
bringing the best in research, data analysis, sampling design, and risk assessment to meet the needs of
the EFPC site. This would be an important site for future study where mercury speciation could be
implemented to help decision-makers determine the type or extent of remediation needed at the site.
This interdepartmental cooperation is the key to future work. The DOE facilities pose special
challenges to technical personnel - challenges that often can be met by a combined task force.
The Technology Support Center is an integral part of the research laboratory network,
providing a format for testing, demonstrating, evaluating, and establishing innovative technologies.
The TSC can serve the Regions and the researchers by bringing together the problems of one with the
solutions of the other. The application of research projects to site characterization serves two
purposes: the in situ validation of an innovative technology and the opportunity to refine that
technology to meet the rigorous conditions of field analytical procedures.
29
-------
REFERENCES
Breckenridge, R.P., and A.B. Crockett, "Detection of Background Levels of Inorganics in Soils and
Sediments at Hazardous Waste Sites", EPA report, in press, 1995.
Brown, C.S. to K.W. Brown, Letter of Request for Technical Assistance at the East Fork Poplar
Creek Site, U. S. EPA Internal Memorandum, March 1991.
Ecker, V., D. Dobb, E. Miller, and D. Cardenas, Quality Assurance Project Plan for Determination of
Speciated and Total Mercury on Poplar Creek Samples, Report to EPA, EMSL-LV, Lockheed
Environmental Systems & Technologies Document #QAO 30-94-01, 1994.
Dobb, D.E, E.L. Miller, and D. Cardenas, Determination of Mercury with Speciation, in Poplar Creek
Soil Samples, U. S. EPA Internal Report, March 1994.
Englund, E. J. and N. Heravi, Conditional Simulation: Practical Application for Sampling Design
Optimization, Proceedings of 1992 Troia Geostatistics Conference, Troia. Portugal, 1992.
Hillman, D.C., Determination of Mercury in Oak Ridge Reservation East Fork Poplar Creek
Sediment Samples by X-ray Fluorescence Spectrometry, U.S. EPA Environmental Monitoring Systems
Laboratory - Las Vegas, Superfund Technical Support Product TSC-10, May 1991.
Miller, E. L., Speciation of Mercury in Soil, U. S. EPA Environmental Monitoring Systems
Laboratory - Las Vegas Internal Report, Las Vegas, NV, 1993.
Miller, E., D. Dobb, D. Cardenas, arid E. Heithmar, Method Characterization During Determination
of Mercury, with Speciation, in Poplar Creek Soil Samples, U.S. EPA EMSL-LV Internal Report,
April 1994.
Revis, N. W., T. R. Osborne, G. Holdsworth, and C. Hadden, "Distribution of Mercury Species in Soil
from a Mercury-Contaminated Site", Water, Air and Soil Pollution , 45:105-113, 1989.
Sakamoto, H., T. Tomiyasu,, and N. Yonehara, "Differential Determination of Organic Mercury,
Mercury (II) Oxide, and Mercury (D.) Sulfide in Sediments by Cold Vapor Atomic Absorption
Spectrometry", Analytical Sciences. Vol. 8, February 1992.
Singh, A K., M. M. Ananda, and A R. Sparks, "Superfund site characterization using non-parametric
variogram modeling", Anal. Chim. Acta. 277, 255-266, Elsevier Science Publishers, Amsterdam, 1993.
U.S. DOE East Fork Poplar Creek - Sewer Line Beltway Remedial Investigation Report, U.S.
DOE/OR/02-1119 & D2. Volumes 1-2, January 1994.
U.S. EPA Risk Assessment Guidance for Superfund: Volume I - Human Health Evaluation Manual
(Part A), EPA/540/1-89/002, EPA, Office of Solid Waste and Emergency Response, 1989.
U.S. EPA Risk Assessment Guidance for Superfund: Human Health Evaluation Manual (Part B),
EPA Office of Solid Waste and Emergency Response, 1991.
3Q i^TUS. GOVERNMENT PRINTING OFFICE: IMS C50-OM/22064
------- |