EPA 905/R-00/007
June 2000
Prediction of sediment toxicity using
consensus-based freshwater sediment quality guidelines
United States Geological Survey (USGS) final report for
the U.S. Environmental Protection Agency (USEPA)
Great Lakes National Program Office (GLNPO)
Christopher G Ingersoll1, Donald D MacDonald2, Ning Wang3,
Judy L Crane4, L Jay Field5, Pam S Haverland1, Nile E Kemble1,
Rebekka A Lindskoog2, Corinne Severn6, and Dawn E Smorong2
'Columbia Environmental Research Center, USGS
4200 New Haven Road, Columbia, MO 65201
2MacDonald Environmental Sciences Limited,
2376 Yellow Point Road, Nanaimo, British Columbia, Canada V9X 1W5
fisheries and Wildlife Sciences, 302 ABRN Building,
University of Missouri, Columbia, MO 65211
4Minnesota Pollution Control Agency, 520 Lafayette Rd., St. Paul, MN 55155
5National Oceanic and Atmospheric Administration (NOAA),
7600 Sand Point Way ME, Seattle, WA 98115
6EVS Environment Consultants, 200 West Mercer Street, Seattle, WA 98119
Project Officer: Marc Tuchman
USEPA GLNPO
77 West Jackson
Chicago, IL 60604
U.S. Environmental Protection Agency
5 Library (RL-12J)
-------
Abstract
The primary objectives of this study were to: (1) evaluate the ability of consensus-based SQGs
(sediment quality guidelines) to predict toxicity in a freshwater database for field-collected
sediments in the Great Lakes basin; (2) evaluate the ability of SQGs to predict sediment toxicity
on a regional geographic basis elsewhere in North America; and (3) compare approaches for
evaluating the combined effects of chemical mixtures on the toxicity of field-collected sediments.
A database was developed from 92 published reports which included a total of 1657 samples
with high-quality matching sediment toxicity and chemistry data. The database was comprised
primarily of 10- to 14-day or 28- to 42-day toxicity tests with the amphipod Hyalella azteca
(designated as the HA10 or HA28 tests) and 10- to 14-day toxicity tests with the midges
Chironomus tentans or C. riparius (designated as the CS10 test). Endpoints reported in these
tests were primarily survival or growth. Because field-collected sediments typically contain
complex mixtures of contaminants, the predictive ability of a sediment assessment is likely to
increase when SQGs are used in combination to classify toxicity of sediments. For this reason, *v
the evaluation of the predictive ability of probable effect concentrations (PECs) was conducted to ">
determine the incidence of effects above and below various mean PEC quotients (mean quotients "^
of 0.1, 0.5, 1.0, and 5.0). The PECs are SQGs that were established as concentrations of ^
individual chemicals above which adverse effects in sediments are expected to frequently occur. %>
A PEC quotient was calculated for each chemical in each sample in the database by dividing the i
concentration of a chemical by the PEC for that chemical. A mean quotient was calculated for ^
each sample by summing the individual quotient for each chemical and then dividing this sum by H
the number of PECs evaluated.
When mean quotients were calculated using an approach of equally weighting up to 10 reliable
PECs (PECs for metals, total polycyclic aromatic hydrocarbons (PAHs), total polychlorinated
biphenyls (PCBs), and sum DDE), there was an overall increase in the incidence of toxicity with
an increase in the mean quotient in all three tests. For example in the HA 10 test, the toxicity of
samples was 20% at mean quotients of <0.1 and increased to 61% at mean quotients of >5.0.
Similarly, for the CS10 test there was a 20% incidence of toxicity at mean quotients of <0.1
increasing to a 64% incidence of toxicity at mean quotients of >5.0. In contrast, the incidence of
toxicity in the HA28 test was only 8% at mean quotients of <0.1 and increased to 91% at mean
quotients of > 1.0. In all three tests, there was a consistent increase in the toxicity at mean
quotients of >0.5. However, the overall incidence of toxicity was greater in the HA28 test
compared to the short-term tests.
The incidence of toxicity at mean quotients of <0.1 was somewhat higher in the HA 10 and CS10
tests (20%) compared to the HA28 test (8%). This toxicity at low mean quotients does not appear
to be related to total organic carbon in sediment. There was insufficient information in the
database to evaluate effects of grain size on toxicity. Unmeasured contaminants in these field-
collected sediments or contaminants for which we do not have reliable PECs (i.e., pesticides,
herbicides) may have contributed to this toxicity at low mean quotients. Alternatively, the data
for HA 10 and CS10 tests were obtained from numerous laboratories which may have contributed
to variability in the data reported in these studies. In contrast, a limited number of laboratories
conducted most of the HA28 tests.
The reason for the higher incidence of toxicity with increasing mean quotients in the HA28 test
compared to the short-term tests may be due to the duration of the exposure or the sensitivity of
-------
growth in the longer HA28 test. A 50% incidence of toxicity in the HA28 test corresponds to a
mean quotient of 0.63 when survival or growth were used to classify a sample as toxic. By
comparison, a 50% incidence of toxicity is expected at a mean quotient of 3.2 when survival
alone was used to classify a sample as toxic in the HA28 test. In the CS10 test, a 50% incidence
of toxicity is expected at a mean quotient of 9.0 when survival alone was used to classify a
sample as toxic or at a mean quotient of 3.5 when survival or growth were used to classify a
sample as toxic. In contrast, similar mean quotients resulted in a 50% incidence of toxicity in the
HA 10 test when survival alone (mean quotient of 4.5) or when survival or growth (mean quotient
of 3.4) were used to classify a sample as toxic. Results of these analyses indicate that both the
duration of the exposure and the endpoints measured can influence whether a sample is found to
be toxic or not. The longer-term tests in which growth and survival are measured tended to be
more sensitive than shorter-term tests, with acute to chronic ratios on the order of 6 indicated for
H. azteca.
We were also interested in determining the predictive ability of PEC quotients for major classes
of compounds. Therefore, we evaluated the incidence of toxicity based on a mean quotient for
metals, a quotient for total PAHs, or a quotient for total PCBs. Different patterns of toxicity
associated with the various procedures for calculating quotients were observed. For example in
the HA28 test, a relatively abrupt increase in toxicity was associated with elevated PCBs alone or
with elevated PAHs alone, compared to the pattern of a gradual increase in toxicity observed
with quotients calculated using a combination of metals, PAHs, and PCBs. These analyses
indicate that the different patterns in toxicity may be the result of unique chemical signals
associated with individual contaminants. While mean quotients can be used to classify samples
as toxic or non-toxic, individual quotients might be useful in helping to identify substances that
may be causing or substantially contributing to the observed toxicity.
We chose to make comparisons across geographic areas using mean quotients calculated by
equally weighting the contribution of the three major classes of compounds (metals, or PAHs, or
PCBs). This approach assumes that these three diverse groups of chemicals exert some form of
joint toxic action. Use of this approach also maximized the number of samples that were used to
make comparisons across geographic areas. Generally, there was an increase in the incidence of
toxicity with increasing mean PEC quotients within most of the regions, basins, and areas for all
three toxicity tests. For the HA10 and HA28 tests, the incidence of toxicity for samples from
each of the Great Lakes and within the areas of each Great Lake was relatively consistent with
the overall pattern of toxicity in the entire database. However, the relationship between the
incidence in toxicity and mean quotients in the CS10 test was more variable among geographic
areas compared to either the HA10 or HA28 test. The results of these analyses indicate that the
consensus-based PECs can be used to reliably predict toxicity of sediments on both a regional
and national basis.
This paper presents results of the first analyses completed on the entire freshwater sediment
database. Some of the additional analyses planned for the database that are beyond the scope of
this paper include: (1) comparing approaches for designating samples as toxic; (2) evaluating
logistic-regression models; (3) identifying a list of optimal analytes for broad scale application
and testing the relative efficacy of the mean versus the sum PEC quotients; (4) evaluating the
influence of grain size and ammonia on the incidence of toxicity; and (5) developing a guidance
manual for conducting an integrated assessment of sediment contamination.
-------
Contents
Abstract 2
List of Figures 4
List of Tables 4
List of Appendices 4
Acknowledgments 5
Introduction 6
Methods 7
Results and Discussion 11
Summary 21
References 23
Figures 26
Tables 28
Appendices 33
List of Figures
Figure 1. Relationship between the geometric mean of the mean PEC quotients and the
incidence of toxicity in the three tests, based on survival or growth.
Figure 2. Relationship between the geometric mean of the mean PEC quotients and the
incidence of toxicity in the three tests, based on survival or growth, or based on
survival alone.
List of Tables
Table 1. Sediment quality guidelines that reflect probable effect concentrations.
Table 2. Incidence of sediment toxicity within ranges of PEC quotients (calculated using
various approaches) for freshwater tests based on survival or growth.
Table 3. Incidence of sediment toxicity within ranges of PEC quotients for the Hyalella
azteca 10- to 14-day test in various geographic areas within the database based on
survival or growth.
Table 4. Incidence of sediment toxicity within ranges of PEC quotients for the Hyalella
azteca 28- to 42-day test in various geographic areas within the database based on
survival or growth.
Table 5. Incidence of sediment toxicity within ranges of PEC quotients for the Chironomus
spp. 10- to 14-day test in various geographic areas within the database based on
survival or growth.
List of Appendices
Appendix 1. Summary of the data presented in Figures 1 and 2.
4
-------
Acknowledgments
We thank the members of the Sediment Advisory Group on Sediment Quality Assessment for
insight and guidance in developing the procedures for evaluating the predictive ability of
freshwater sediment quality guidelines. We would also like to thank Kathie Adare, Peter
Landrum, and Rick Swartz for providing helpful review comments on the report. Although
information in this report was developed in part by the USGS, USEPA, and NOAA, the report
may not necessarily reflect the reviews of these organizations and no official endorsement should
be inferred. Sediment quality guidelines (SQGs) can be used as one of several tools to assess
contaminated sediments; however, there is no intent expressed or implied that these guidelines
represent USGS, USEPA, or NOAA sediment quality criteria. Information in this report and the
development of the database have been funded in part by the USEPA Great Lakes National
Program Office, USEPA Office of Research and Development, USEPA Office of Water,
Minnesota Pollution Control Agency, and National Research Counc.il of Canada. This report has
been reviewed in accordance with USEPA and USGS policy.
-------
Introduction
Numerical sediment quality guidelines (SQGs) have been developed by a variety of federal, state,
and provincial agencies across North America using matching sediment chemistry and biological
effects data. These SQGs have been routinely used to interpret historical data, identify potential
problem chemicals or areas at a site, design monitoring programs, classify hot spots and rank
sites, and make decisions for more detailed studies (Long and MacDonald 1998). Additional
suggested uses for SQGs include identifying the need for source controls of problem chemicals
before release, linking chemical sources to sediment contamination, triggering regulatory action,
and establishing target remediation objectives (USEPA 1997). Numerical SQGs, when used with
other tools such as sediment toxicity tests, bioaccumulation, and benthic community surveys, can
provide a powerful weight of evidence for assessing the hazards associated with contaminated
sediments (Ingersoll et al. 1997).
A critical component in the application of SQGs for assessing sediment quality is a
demonstration of the ability of the guidelines to predict the absence or presence of toxicity in
field-collected sediments (Ingersoll et al. 1996, Smith et al. 1996, Long et al. 1998a, Swartz
1999, Fairey et al. 2000, MacDonald et al. 2000a,b). This paper is the fourth in a series that is
intended to address the ability of various SQGs to predict toxicity in contaminated sediments.
The first paper in the series focused on resolving the "mixture paradox" that is associated with
the application of empirically-derived SQGs for individual polycyclic aromatic hydrocarbons
(PAHs). In this case, the paradox was addressed by developing consensus-based SQGs for total
PAHs (Swartz 1999). A second paper developed and evaluated consensus-based SQGs for total
polychlorinated biphenyls (PCBs) to address a similar mixture paradox for that group of
contaminants (MacDonald et al. 2000b).
A third paper developed consensus-based SQGs for freshwater sediments (MacDonald et al.
2000a). The published SQGs for 28 chemical substances were assembled and classified into two
categories in accordance with their original narrative intent. These published SQGs were then
used to develop two consensus-based SQGs for each contaminant, including a threshold effect
concentration (TEC; below which adverse effects are not expected to occur) and a probable effect
concentration (PEC; above which adverse effects are expected to frequently occur; Table 1;
MacDonald et al. 2000a). A preliminary evaluation of the predictive ability of these consensus-
based SQGs for freshwater sediment was conducted using a database of 347 samples obtained
from 15 separate studies. The results of these three previous investigations demonstrated that the
consensus-based SQGs provide a unifying synthesis of the existing guidelines, reflect causal
rather than correlative effects, and account for the effects of contaminant mixtures in sediment
(Swartz 1999, MacDonald et al. 2000a,b).
The primary objective of this fourth paper is to further evaluate the predictive ability of the
consensus-based PECs developed by MacDonald et al. (2000a). The database used by
MacDonald et al. (2000a) was expanded to include a total of 1657 samples from 92 published
reports with high-quality matching sediment toxicity and chemistry data. The majority of the data
from these reports were for 10- to 28-day toxicity tests with the amphipod Hyalella azteca or 10-
to 14-day toxicity tests with the midges Chironomus tentans or C. riparius. The endpoints
-------
measured in these toxicity tests primarily included survival or growth of test organisms at the end
of the sediment exposures. A second objective of this paper is to evaluate the predictive ability of
these PECs on a regional basis within the larger database. We were interested in determining if
there are differences in the predictive ability of the PECs across the entire database compared to
various geographic areas within the database, such as all of the samples from the Great Lakes or
all of the samples from an area within a Great Lake, such as Indiana Harbor or Waukegan Harbor
located within Lake Michigan.
Methods
Development of consensus-based sediment quality guidelines
Individual SQGs for freshwater ecosystems have previously been developed using a variety of
approaches (Table 1). Each of these approaches has certain advantages and limitations which
influence their application in the sediment quality assessment process (Ingersoll et al. 1997). In
an effort to focus on the agreement among these various published SQGs, consensus-based SQGs
were developed by MacDonald et al. (2000a) for 28 chemicals of concern in freshwater
sediments (i.e., metals, PAHs, PCBs, and pesticides). For each contaminant of concern, two
consensus-based SQGs were developed from published SQGs, including a threshold effect
concentration (TEC) and a probable effect concentration (PEC). The TECs were calculated by
determining the geometric mean of the SQGs that were included in this category (MacDonald et
al. 2000a). Likewise, consensus-based PECs were calculated by determining the geometric mean
of the PEC-type values (Table 1). The geometric mean, rather than the arithmetic mean or
median, was calculated because it provides an estimate of central tendency that is not unduly
affected by extreme values and because the distributions of the SQGs were not known
(MacDonald et al. 2000a). Consensus-based TECs or PECs were calculated only if three of more
published SQGs were available for a chemical substance or group of substances. The evaluations
of toxicity in the present study were based on the use of PECs because TECs were developed to
provide an estimate of conditions where toxicity would not be expected and PECs were
developed to provide an estimate of conditions where toxicity would be expected. Evaluations of
SQGs in the present study were based on dry-weight concentrations because previous studies
have demonstrated that normalization of SQGs for PAHs or PCBs to total organic carbon
(Barrick et al. 1988, Long et al. 1995, Ingersoll et al. 1996) or normalization of metals to acid-
volatile sulfides (Long et al. 1998b) did not improve the predictions of toxicity in field-collected
sediments.
The consensus-based PECs listed in Table 1 were critically evaluated by MacDonald et al.
(2000a) to determine if they would provide effective tools for assessing sediment quality
conditions in freshwater ecosystems. The criteria for evaluating the reliability of the consensus-
based PECs were adapted from Long et al. (1998a). Specifically, the individual TECs were
considered to provide a reliable basis for assessing the quality of freshwater sediments if >75%
of the sediment samples were correctly predicted to be not toxic. Similarly, the individual PEC
for each substance was considered to be reliable if >75% of the sediment samples were correctly
predicted to be toxic using the PEC. Therefore, the target levels of both false positives (i.e.,
samples incorrectly classified as toxic) and false negatives (i.e., samples incorrectly classified as
-------
not toxic) was 25% using the TEC and PEC. To assure that the results of this evaluation were
not unduly influenced by the number of sediment samples, the various SQGs were considered to
be reliable only if a minimum of 20 samples were included in the evaluation (i.e., 20 samples
above at a PEC or 20 samples below a TEC; CCME 1995). The results of this evaluation
described in MacDonald et al. (2000a) indicated that most of the TECs (i.e., 21 of 28) provide an
accurate basis for predicting the absence of sediment toxicity. Similarly, most of the PECs (i.e.,
16 of 28) provide an accurate basis for predicting sediment toxicity (Table 1).
A reliable TEC or PEC was not available for mercury, an important contaminant of concern in
sediments (Table 1). This lack of reliability is most likely due to the speciation of mercury in the
sediments, as well as the ability of methyl mercury to bioaccumulate in organisms. Sediment
quality guidelines developed using a tissue residue approach are needed to establish safe
sediment concentrations for human health and piscivorus-wildlife receptors (Ingersoll et al.
1997). For this report, only direct toxic effects on benthic invertebrates are considered in the
evaluation of the predictability of the consensus-based SQGs.
Development of the sediment toxicity database
To support the development of the sediment toxicity database, matching sediment chemistry and
biological effects data were compiled for various freshwater locations across North America (in
addition to the data that were used in the analyses performed by MacDonald et al. 2000a).
Candidate data sets were identified by reviewing the published literature and by contacting
individuals active in the field of sediment quality assessment. More than 1500 documents were
reviewed and evaluated to obtain the data required to evaluate SQGs in the present study.
Because these data sets were generated for a wide variety of purposes, each study was critically-
evaluated to assure the quality of the data used for evaluating the predictive ability of the SQGs
(Long et al. 1998a, Ingersoll and MacDonald 1999). Data from individual studies were
considered acceptable for use in the present study if:
• The study was conducted in a freshwater location in North America;
• Appropriate procedures were used to collect, handle, and store sediments (e.g., ASTM
2000, USEPA 2000);
• Matching sediment chemistry and biological effects data were reported and
concentrations of contaminants were measured in each sample or treatment group.
• Minimum data quality requirements were reported. For example, analytical detection
limits were lower than freshwater probable effect levels (PEL; Smith et al. 1996),
accuracy and precision were within acceptable limits, and analytes were not present at
detectable levels in method blanks;
• Appropriate analytical methods were used to generate chemistry data. For metals,
concentrations of total metals needed to be reported. However, other measures of metal
concentrations were used (i.e., simultaneously extracted metals) if sufficient information
was available to demonstrate that these measures are comparable to total metal
concentrations (Ingersoll et al. 1996, 1998). For organic compounds, the concentrations
needed to be measured using gas chromatography-mass spectroscopy, high pressure
liquid chromatography, or comparable methods; and,
8
-------
• Toxicity tests needed to meet test acceptability criteria outlined in ASTM (2000) and
USEPA (2000) and endpoints measured needed to be ecologically relevant (likely to
influence the viability of the organism in the field) or needed to be indicative of
ecologically-relevant endpoints.
Using these selection criteria, a total of 92 freshwater data sets were incorporated into a database
that included 1657 individual sediment samples. The toxicity tests in this database primarily
include tests with the amphipod, Hyalella azteca; the midges, Chironomus tentans or C. riparius;
the mayfly, Hexagenia limbata; the oligochaete, Lumbriculus variegatus; the daphnids,
Ceriodaphnia dubia or Daphnia magna; and, the bacterium, Vibrio fisheri (Microtox). We
selected a subset of the samples from this database that reported sediment chemistry for at least
one of the substances for which reliable SQGs were listed in Table 1. We then selected a subset
of these resultant samples with matching toxicity data for the amphipod Hyalella azteca or the
midges Chironomus tentans or C. riparius. Tests with H. azteca and Chironomus spp. were
selected because the samples that were tested represented a broader geographic area compared to
the other tests in the database. The selected studies provided 670 samples for H. azteca 10- to 14-
day tests (designated as HA 10), 160 samples for H. azteca 28- to 42-day tests (designated as
HA28) and 632 samples for Chironomus spp. 10- to 14-day tests (CS10; 556 of the samples were
for tests with C. tentans and 76 of the samples were for tests with C. riparius). We combined the
data for the two midge species due to the limited amount of data for C. riparius. Preliminary
analyses of the database indicated similar sensitivity for these two species of midge. The selected
studies represented a broad range in both sediment toxicity and contamination. A total of 28% of
the samples were toxic in the HA10 test, 35% of the samples were toxic in the HA28 test, and
27% of the samples were toxic in the CS10 test (28% of the samples were toxic in the C. tentans
tests and 21% of the samples were toxic in the C. riparius tests). Toxicity of samples was
determined as a significant reduction in survival or growth relative to a control or reference
sediment (as designated in the original study or determined using appropriate statistical
procedures). Sexual maturation and reproduction were periodically reported in the HA28 test;
however, these two additional endpoints did not identify any additional samples as toxic relative
to effects reported on survival or growth of amphipods.
The total PCB concentration in each sediment sample in the database was calculated by summing
dry-weight concentrations of individual congeners. If only aroclors concentrations were reported,
total PCBs were calculated as the sum concentration of all individual aroclors. If both congeners
and aroclors were reported, the congeners were used to calculate the concentration total PCBs in
a sample. If only total PCBs was reported for a sample, then this value was used. The total PAH
concentration in each sediment sample was generally calculated by summing the dry-weight
concentrations from as many of the following 13 compounds that were reported: acenaphthene,
acenaphthylene, anthracene, fluorene, 2-methylnaphthalene, naphthalene, phenanthrene,
benz(a)anthracene, dibenz(a,h)anthracene, benzo(a)pyrene, chrysene, fluoranthene, and pyrene.
Total PAHs were calculated using eight or less of these individual PAHs for <5% of the samples
in the database. In calculating total PCBs or total PAHs, half the detection limit was used for
compounds reported below the detection limit. Similarly, half of the detection limit was used for
concentrations of metals below the detection limit. For DDTs, the concentrations of p,p'-DDD
and o,p'-DDD, p,p'-DDE and o,p'-DDE, and p,p'-DDT and o,p'-DDT were summed to calculate
-------
the concentrations of sum DDD, sum DDE, and, sum DDT, respectively. Total DDT was
calculated by summing the concentrations of sum DDD, sum DDE, and, sum DDT. If all
individual PCBs, PAHs, DDD, DDE, or DDT were less that the detection limit, the detection
limits were summed and reported as a less than value for the sum.
Analysis of data
The initial evaluation of predictive ability by MacDonald et al. (2000a) focused primarily on
determining the ability of each SQG, when applied alone, to correctly classify samples as toxic or
not toxic. Because field-collected sediments typically contain complex mixtures of
contaminants, the predictive ability of these sediment quality assessment tools is likely to
increase when the SQGs are used in combination to classify toxicity of sediments. For this
reason, the evaluation of the predictive ability of the SQGs in the present study was conducted to
determine the incidence of effects above and below various mean PEC quotients (mean quotients
of 0.1, 0.5, 1.0, and 5.0; Ingersoll et al. 1998, Long et al. 1998a, Fairey et al. 2000).
A PEC quotient was calculated for each chemical in each sample in the database by dividing the
concentration of a chemical by the PEC for that chemical. A mean quotient was then calculated
for each sample by summing the individual quotient for each chemical and dividing this sum by
the number of reliable PECs evaluated. MacDonald et al. (2000a) found that some PEC values
were more reliable predictors of toxicity and that use of only these PECs reduced the variability
in the prediction of sediment toxicity compared to using all available PECs. The PEC for total
PAHs, instead of the PECs for the individual PAHs, was used in the calculation to avoid double
accounting of the PAH data (MacDonald et al. 2000a). This resulted in the use of up to 10
reliable PECs in calculating the mean quotient (arsenic, cadmium, chromium, copper, lead,
nickel, zinc, total PAHs, total PCBs, and sum DDE; Table 1 and designated "Mean - all" in Table
2). This approach to the calculation of mean quotients weighs each of the chemicals and
chemical classes equally (Ingersoll et al. 1998, Long et al. 1998a, MacDonald et al. 2000a).
In the present study, a second approach was also used to calculate mean PEC quotients. We were
interested in equally weighting the contribution of metals, PAHs, and PCBs in the evaluation of
sediment chemistry and toxicity (assuming these three diverse groups of chemicals exert some
form of joint toxic action). For this reason, we first calculated an average PEC quotient for up to
seven metals in a sample. A mean quotient was then calculated for each sample by summing the
average quotient for metals, the quotient for total PAHs, and the quotient for total PCBs, and
then dividing this sum by three (n = 3 quotients/sample; designated "Mean - MPP (and)" in Table
2). Another approach for evaluating mean quotients was to calculate the mean of the average
quotient for metals, the quotient for total PAHs, or the quotient for total PCBs (n = 1 to 3
quotients/sample; designated "Mean - MPP (or)" in Table 2). Hence, the "Mean - MPP (or)"
approach uses any or all of the three classes of contaminants as available. For example, if a
sample only had a measure of total PAHs and total PCBs, then the mean quotient would be
calculated using just the quotients for these two classes of compounds. In contrast, the "Mean -
MPP (and)" approach only uses samples with measures of metals, total PAHs, and total PCBs.
Sum DDE was not included in these calculations of the mean quotient because there were a
limited number of samples in the database with elevated concentrations of DDE and we were
10
-------
interested in equally weighting the contributions of the major groups of contaminants in the
database (metals, total PAHs, and total PCBs). Therefore, the differences in this "MPP
approach" from the approach used by MacDonald et al. (2000a) are: (1) an average quotient for
metals was used instead of the individual quotients for metals and (2) sum DDE was not used in
the calculation.
Results and Discussion
Evaluation of approaches for calculation of mean PEC quotients
The results of the evaluations of the incidence of sediment toxicity within ranges of mean PEC
quotients in the entire database for each of the three tests (HA10, HA28, CS10) are summarized
in Table 2. When mean quotients were calculated using the approach of weighting equally up to
10 reliable PECs (designated "Mean - all" in Table 2), there was an increase in the incidence of
toxicity with an increase in the mean quotient in all three tests. For the HA 10 test, the incidence
of toxicity was 20% at mean quotients of <0.1 and increased to 67% at mean quotients of >5.0.
Similarly, for the CS10 test there was a 20% incidence of toxicity at mean quotients of <0.1
increasing to a 64% incidence of toxicity at mean quotients of >5.0. hi contrast, the incidence of
toxicity in the HA28 test was only 8% at mean quotients of <0.1 and increased to 91% at mean
quotients of > 1.0 (the incidence of toxicity at a quotient >5.0 was not calculated for the HA28
test due to a limited number of samples above a quotient of 5.0; Table 2).
Long et al. (1998a) conducted a similar analysis of the incidence of toxicity in sediment tests
using a database developed for 10-day marine amphipod tests (n=1068). The incidence of
toxicity was only 12% at mean quotients of <0.1 (quotients calculated using either marine effect
range median (ERM) or probable effect level (PEL) guidelines; Long et al. 1998a). In the present
study, the incidence of toxicity at mean quotients of <0.1 was somewhat higher in the HA10 and
CS10 tests (20%) compared to the tests with marine amphipods (12%; Long et al. 1998a) or the
HA28 test (8%; Table 2).
The reason for this higher incidence of toxicity at mean quotients of <0.1 in the HA10 and CS10
tests is not clear. This toxicity at low mean quotients does not appear to be related to total
organic carbon in sediment. There was insufficient information in the database to evaluate
effects of grain size on toxicity. USEPA (2000) and ASTM (2000) reported that amphipods and
midges are relatively intolerant to effects of sediment grain size. Unmeasured contaminants in
these field-collected sediments or contaminants for which we do not have reliable PECs (i.e.,
pesticides, herbicides) may have contributed to this toxicity at low mean quotients (see
discussion of Figures 1 and 2 below). Alternatively, the data for HA10 and CS10 tests were
obtained from numerous laboratories which may have contributed to variability in the data
reported in these studies, hi contrast, a limited number of laboratories conducted most of the
toxicity tests for the marine amphipod or HA28 tests.
hi all three tests, there was a consistent increase in the toxicity at mean quotients of >0.5
(designated "Mean - all" in Table 2). However, the overall incidence of toxicity was greater in
the HA28 test (91% toxicity at mean quotients of >1.0) compared to the short-term tests (57%
11
-------
toxicity at mean quotients of > 1.0 and 67% toxicity at mean quotients of >5.0 in the HA 10 test
and 51% toxicity at mean quotients of > 1.0 and 64% toxicity at mean quotients of >5.0 in the
CS10 test). Similarly, Long et al. (1998a) reported a 56 to 71% incidence of toxicity at mean
quotients of > 1.0 in the 10-day sediment tests with marine amphipods. The reason for the higher
incidence of toxicity in the HA28 test compared to the short-term tests may be due to the
duration of the exposure or the sensitivity of growth in the longer HA28 test (see discussion of
Figures 1 and 2 below). However, comparisons of the sensitivity between these tests needs to be
made with some caution. There were very few samples in the database where tests were
conducted using splits of the same samples. Therefore, the differences observed in the responses
of organisms may also be due to differences in the types of sediments evaluated in the individual
databases for each test.
We were also interested in determining the predictive ability of PEC quotients for major classes
of compounds. Therefore, we evaluated the incidence of toxicity based on an average quotient
for metals, a quotient for total PAHs, or a quotient for total PCBs (second, third, and fourth rows
for each toxicity test listed in Table 2). For the HA 10 test, the incidence of toxicity across
quotients of <0.1 to >1.0, based on metals alone, total PAHs alone, or total PCBs alone were
similar to the incidence of toxicity that was calculated for the mean quotient using up to 10 PECs
(designated "Mean - all" in Table 2). The incidence of toxicity in the HA 10 test was somewhat
higher at quotients >5.0 for total PAHs (80%) or total PCBs (73%) compared to metals alone
(62%).
For the CS10 test, the incidence of toxicity was also similar across quotients of <0.1 to >5.0,
calculated based on metals alone compared to a mean quotient calculated using up to 10 PECs
(designated "Mean - all" in Table 2). However, the incidence of toxicity at a total PCB quotient
<0.1 was 46%, suggesting that other compounds may be contributing to toxicity at low
concentrations of PCBs. The incidence of toxicity in the CS10 test was somewhat higher for
quotients based solely on total PAHs compared to quotients based on metals alone, total PCBs
alone, or mean quotients calculated using up to 10 PECs. These analyses suggest that the CS10
test may be more sensitive to PAHs compared to the other chemical classes.
For the HA28 test, the incidence of toxicity was similar across the quotients of <0.1 to >1.0,
calculated based on metals alone compared to a mean quotient calculated using up to 10 PECs
(designated "Mean - all" in Table 2). However, the incidence of toxicity at quotient of 0.1 to
<0.5 was higher for PAHs (61%) compared to the other three quotients (6 to 20% toxicity for
metals alone, total PCBs alone, or mean quotients based on up to 10 PECs). The incidence of
toxicity at PCB quotients of < 1.0 was lower (4 to 17%), while the incidence of toxicity at a PCB
quotient of > 1.0 was higher (97%) compared to the other three quotients. Results of these
analyses indicate a relatively abrupt increase in toxicity associated with elevated PCBs alone or
elevated PAHs alone compared to the pattern of a gradual increase toxicity observed with
quotients calculated using the up to 10 PECs (designated "Mean - all" in Table 2). These results
suggest that H. azteca may be more sensitive to PAHs and PCBs in longer-term tests than it is to
metals.
hi the next analysis, we were interested evaluating the incidence in toxicity by equally weighting
12
-------
the combined influence of metals, PAHs, and PCBs in a sample. A mean PEC quotient was
calculated for each sample by summing the average quotient for metals, the quotient for total
PAHs, and the quotient for total PCBs, and then dividing this sum by three (designated as "Mean
- MPP (and)" in Table 2). This calculation was done for only those samples with reported
concentrations of metals, total PAHs, and total PCBs (266 of 670 samples in the HA 10 test, 109
of 160 samples in the HA28 test, and 177 of 632 samples in the CS10 test). Results of this
analysis indicate a higher incidence of toxicity in the HA10 (66%), HA28 (100%), and CS10
(60%) tests at mean quotients >1.0 based on an equal weighting of metals, total PAHs, and total
PCBs compared to a mean quotient calculated using up to 10 PECs (57, 91, and 51%,
respectively; designated "Mean- all" in Table 2).
The different patterns of toxicity associated with these various procedures for calculating
quotients may be the result of unique chemical signals associated with individual contaminants in
each sample. For example, there was a higher incidence in toxicity with quotients calculated
using total PCBs alone compared to quotients calculated using metals alone in the HA 10 test.
Alternatively, these different patterns may also be influenced by the total number of samples used
to make these comparisons. For example, there were 670 total samples for the HA 10 test. Of
these 670 samples, 623 had metal chemistry data, 488 had measured concentrations of total
PAHs, 326 had measured concentrations of total PCBs, and all three of these classes of
compounds were measured in only 266 samples. In order to determine the influence of sample
number on the observed incidence of toxicity, we first selected the same samples used in the
analysis described in the previous paragraph (samples in which metals, total PAHs, and total
PCBs were all measured). Quotients were then calculated for these subsets of samples: (1) using
up to 10 PECs (designated "Mean - all (select 1)"), (2) using up to 9 PECs (not including DDE;
designated "Mean - all (select2)", or (3) using metals alone ("mean - metals (select2)"), total
PAHs alone ("total PAHs (select2)"), or total PCBs alone ("total PCBs (select2)"; Table 2).
Results of these analyses indicate that the incidence of toxicity for all three tests was similar in
these subsets of samples using the three procedures to calculate mean quotients ("Mean - MPP
(and)" versus "Mean - all (select 1)" versus "Mean - all (select2)"). In the HA 10 test, the
incidence of toxicity in the subset of samples (n=266) was similar based on an average quotient
for metals alone, a quotient for total PAHs alone, or a quotient for total PCBs alone compared to
the "Mean - all (select2)". In contrast, there were different patterns of toxicity associated with
individual classes of compounds in the subsets of samples in the HA28 (n=109) or CS10 (n=177)
tests where metals, PAHs, and PCBs were all reported. In the HA28 test, a relatively abrupt
increase in toxicity was associated with elevated PCBs alone or with elevated PAHs alone,
compared to the pattern of a gradual increase toxicity observed with quotients calculated using
the "Mean - all (select2)". Similarly in the CS10 test, a increase in toxicity was observed at
lower quotients of PAHs alone or metals alone compared to the pattern of a gradual increase
toxicity observed with quotients calculated using the "Mean - all (select2)".
These analyses indicate that the different patterns in toxicity may be the result of unique chemical
signals associated with individual contaminants in samples rather than the result of a limited
number of samples used to make these comparisons. However, we could only make these
comparisons with a limited number of samples where metals, PAHs, and PCBs were all reported.
13
-------
Fairey et al. (2000) conducted a similar analysis of a larger marine database for amphipods and
found that the number and type of SQGs used in the calculation of a mean quotient influenced
the predictions of sediment toxicity to amphipods. The incidence of toxicity to amphipods
increased with increasing numbers of contaminants included in the derivation of the mean
quotient (Fairey et al. 2000). While mean quotients can be used to classify samples as toxic or
non-toxic, individual quotients might be useful in helping to identify substances that may be
causing or substantially contributing to the observed toxicity (MacDonald et al. 2000b).
To use all samples in the database and equally weight the contribution of metals, PAHs, and
PCBs, a final analysis was conducted where the mean quotient was calculated as the average of
the three major classes available in a sample. For example, if a sample only had a measure of
total PAHs and total PCBs, then the mean quotient would be calculated using just the quotients
for these two classes of compounds (designated as "Mean - MPP (or)" in Table 2). Results of
these analyses indicate that the incidence of toxicity in all three tests was similar when either
"Mean - all" (based on up to 10 PECs) or "Mean - MPP (or)" were used to calculate the mean
quotients (Table 2).
Evaluation of exposure duration and endpoints measured in toxicity tests
In Figures 1 and 2, we evaluated the relationship between mean PEC quotients and the incidence
of toxicity as a function of the duration of the exposure or of the endpoints measured in the
toxicity tests. In this analysis, a mean quotient for each sample was calculated using the "Mean -
MPP (or)" approach. The samples within each test were ranked in ascending order by mean
quotient. The incidence of toxicity and geometric mean of the mean quotients within groups of
20 samples for the HA 10 and CS10 tests or within groups of 10 samples for the HA28 test was
then plotted (Figures 1 and 2, Appendix 1). The geometric mean, rather than the arithmetic mean
or median of the quotients, was calculated because it provides an estimate of central tendency
that is not unduly affected by extreme values and because the distributions of the mean quotients
were not known.
In Figure 1, samples were classified as toxic based on an adverse effect on survival or growth in
the three tests. Results of these analyses plotted in Figure 1 are consistent with the analyses
presented in Table 2. Importantly, the incidence of toxicity increases with increasing level of
contamination in all three tests. This increase was particularly pronounced at mean quotients of
>0.5 in all three tests. There was a slightly elevated incidence of toxicity at the very lowest mean
quotient in all three tests. Long et al. (1998a) also observed an elevated incidence of toxicity
with marine amphipods at low mean quotients. Long et al. (1998a) suggested that these samples
were sometimes fine-grained sediments with low concentrations of organic carbon and detectable
concentrations of butyltins, chlorinated pesticides, alkyl-substituted PAHs, ammonia, or other
substances not accounted for with the SQGs. In the present study, the incidence of toxicity at low
mean quotients did not appear to be related to total organic carbon in sediment. There was
insufficient information in the database to evaluate effects of grain size on toxicity. USEPA
(2000) and ASTM (2000) reported that amphipods and midges were relatively intolerant to
effects of sediment grain size.
14
-------
We also evaluated relationships between the toxicity and mean quotients calculated using up to
10 reliable PECs or mean quotients calculated using all of the available PECs in Table 1
regardless of their reliability (plots not included in this paper). An increase in toxicity was
observed with increasing contamination using either 10 reliable PECs or all of the PECs to
calculate mean quotients. However, the variability was higher when all of the PECs listed in
Table 1 were used in this analysis. Therefore, the use of reliable PECs improved the relationship
between mean PEC quotients and the incidence of toxicity.
hi Figure 2, samples in the three tests were classified as toxic based on an adverse effect on
survival alone or based on an adverse effect on survival or growth. The relationship between the
incidence of toxicity and the geometric mean of the mean quotients was best described by a three
parameter logistic model (SigmaPlot 1997; Figure 2; see Appendix 1 for the equations and
coefficients). The best fit of the data was observed in the HA28 test (r2 = 0.79 based on survival
or 0.93 based on survival or growth) relative to the HA10 test (r2 = 0.73 based on survival or 0.78
based on survival or growth) or CS10 test (r2 = 0.56 based on survival or 0.76 based on survival
or growth; Figure 2). In the HA 10 test, the relationship between toxicity and mean quotient was
similar when either survival alone or survival or growth together were used to classify a sample
as toxic. However, in the HA28 and CS10 tests, the relationship between the incidence of
toxicity and mean quotient was different when survival or growth were used to classify a sample
as toxic compared to survival alone (Figure 2).
The incidence of toxicity in the HA28 and CS10 tests based on survival or growth was often
double the incidence of toxicity based on survival alone at mean quotients of >0.3. A 50%
incidence of toxicity in the HA28 test corresponds to a mean quotient of 0.63 when survival or
growth were used to classify a sample as toxic (Figure 2, Appendix 1). By comparison, a 50%
incidence of toxicity was estimated at a mean quotient of 3.2 when survival alone was used to
classify a sample as toxic in the HA28 test. In the CS10 test, a 50% incidence of toxicity was
estimated at a mean quotient of 9.0 when survival alone was used to classify a sample as toxic or
at a mean quotient of 3.5 when survival or growth were used to classify a sample as toxic, hi
contrast, similar mean quotients resulted in a 50% incidence of toxicity in the HA 10 test when
survival alone (mean quotient of 4.5) or when survival or growth (mean quotient of 3.4) were
used to classify a sample as toxic.
Results of these analyses indicate that both the duration of the exposure and the endpoint
measured can influence whether a sample is found to be toxic or not. Again, comparisons of the
sensitivity between these tests needs to be made with some caution. There were very few
samples in the freshwater database where tests were conducted using splits of the same samples.
Therefore, the differences observed in the responses of organisms may also be due to differences
in the types of sediments evaluated in the individual databases for each test. Nevertheless, it
appears that longer-term tests in which survival and growth are measured tend to be more
sensitive than short-term tests, with acute to chronic ratios on the order of 6 indicated for H.
azteca. Similar differences in sensitivity of//, azteca have been observed in 10- and 42-day
water-only exposures to cadmium, fluoranthene or DDD (unpublished data).
15
-------
Evaluation of the predictive ability of mean PEC quotients across various geographical
areas with the database
A primary objective of this paper was to determine if there were differences in the predictive
ability of PEC quotients within geographic areas compared to the entire database. We chose to
make these comparisons of geographic areas using the "MPP (or)" mean quotients calculated as
the mean of the average quotient for metals, the quotient for total PAHs, or the quotient for total
PCBs (Table 2). Use of this approach maximized the number of samples used in the evaluation
and equally weighted the contribution of the three major classes of compounds (metals, or PAHs,
or PCBs) to the observed incidence of toxicity. The relationship between the incidence of toxicity
and mean quotients is presented in Tables 3 to 5 for the entire database and for various
geographical areas represented within the database. Designation of region, basin, and area for
each geographic area in Tables 3 to 5 is based on information obtained from the original report or
by contacting the authors of the report. Comparisons of toxicity among the entire geographic
areas listed in Tables 3 to 5 should be done with caution given the limited number of samples
from each area (i.e., only 5 samples for the HA10 test from the entire Buffalo River). Samples
were grouped by geographic areas in Table 3 to 5 to determine how well toxicity and mean
quotients correspond in the entire database compared to subsets of samples within the database.
Control samples were not included within each geographical area. However, the incidence of
toxicity with and without control samples was similar for the entire database (first and second
rows for each of the three tests listed in Tables 3 to 5).
For the HA 10 test, there was typically an increase in the incidence of toxicity with an increase in
the mean quotient within most of the regions, basins, and areas (Table 3). The incidence of
toxicity for samples from each of the Great Lakes and within the areas of each Great Lake was
relatively consistent with the overall pattern of toxicity in the entire database. No one area
influenced the overall incidence of toxicity for the HA 10 test. However, the absolute incidence of
toxicity differed somewhat between areas. For example, the incidence of toxicity for the Great
Lakes samples in the HA10 test (n=313) was 14% at mean quotients of <0.1, 68% at mean
quotients of >1.0, and 78% at mean quotients of >5.0. This compares to an incidence of toxicity
for the entire database (n=654) of 19% at mean quotients of <0.1, 54% at mean quotients of > 1.0,
and 71% at mean quotients of >5.0. Hence, at mean quotients of >1.0 there was about a 14%
higher incidence of toxicity for samples from the Great Lakes compared to the entire database for
the HA 10 test. The lower incidence of toxicity for samples from non-Great Lakes areas was
primarily due to the lower incidence of toxicity at mean quotients >1.0 for samples from the
states of New York (7%, n=15), Oregon (40%, n=20), and Washington (50%, n=34).
In the HA 10 test, there was also a higher incidence of toxicity at mean quotients of <0.1 for a
limited number of samples from the lower Mississippi River (89%, n=9) and from the Temblader
River in California (75%, n=4) compared to the entire database. These samples contributed to
the 22% incidence of toxicity that was observed for samples from non-Great Lakes areas at
quotients of <0.1 (n=89). By comparison, only a 14% incidence of toxicity was evident for
samples with quotients of <0.1 for Great Lakes areas (n=51). One of the reasons for the
difference between non-Great Lakes samples and Great Lakes samples may be due to
16
-------
contaminants associated with agricultural practices in the non-Great Lakes areas (i.e., herbicides
or pesticides) that contributed to the toxicity of these samples independently of elevated metals,
PAHs, or PCBs.
hi the HA28 test, there was also an increase in the incidence of toxicity with an increase in the
mean quotient within most of the regions, basins, and areas (Table 4). The incidence of toxicity
for samples from each of the Great Lakes and within areas of each Great Lake was relatively
consistent with the overall pattern of toxicity in the entire database. Similarly, the incidence in
toxicity for samples from the Great Lakes was consistent with the incidence in toxicity for
samples from non-Great Lakes areas. Therefore, no one area influenced the overall incidence of
toxicity within the HA28 test. There were a relatively high number of non-toxic samples (n= 51
of 53) with low mean quotients from the upper Mississippi River. The incidence of toxicity for
the Great Lakes samples (n=42) was 13% at mean quotients of <0.5, 71% at mean quotients of
0.5 to <1.0, and 100% at mean quotients of >1.0. This pattern is consistent with incidence of
toxicity for the entire database (n=151) of 17% at mean quotients of <0.5, 56% at mean quotients
of 0.5 to <1.0, and 97% at mean quotients of >1.0 (Table 4).
hi the CS10 test, there was typically an increase in the incidence of toxicity with an increase in
the mean quotient within most of the regions, basins, and areas (Table 5). The incidence of
toxicity for the Great Lakes samples (n=463) was 14% at mean quotients of <0.1, 56% at mean
quotients of > 1.0, and 71% at mean quotients of >5.0. This pattern is consistent with incidence
of toxicity for the entire database (n=611) of 20% at mean quotients of <0.1, 52% at mean
quotients of > 1.0, and 68% at mean quotients of >5.0. The incidence of toxicity for samples from
each of the Great Lakes and within the areas of each Great Lake was relatively consistent with
the overall pattern of toxicity hi the entire database; however, there were some exceptions to this
pattern. A lower incidence of toxicity was observed at mean quotients of > 1.0 for samples from
the St. Mary's River (41%, n=17), Fox River and Green Bay (0%, n=8), Menominee River (38%,
n= 8), and White Lake Montague (25%, n=8). However, at mean quotients of >5.0, there was a
more consistent pattern in the incidence of toxicity in the CS 10 test. An exception to this pattern
at mean quotients >5.0 was observed for samples from the St. Mary's River (38%, n=8) and
White Lake Montague (25%, n=4). Similarly, a lower incidence in toxicity was observed at
mean quotients of >5.0 for samples from the state of Washington (50%, n=4).
hi the CS10 test, there was also a higher incidence of toxicity at mean quotients of <0.1 for a
limited number of samples from Sheboygan Harbor (75%, n=4), from the lower Mississippi
River (89%, n=9), and from the Trinity River (67%, n=3). The samples from the Lower
Mississippi River and Trinity River contributed to the 53% incidence of toxicity that was
observed for samples from non-Great Lakes areas with quotients of <0.1 (n=19). By comparison,
only a 14% incidence of toxicity was evident for samples with quotients of <0.1 for Great Lakes
areas (n=99). Again, one of the reasons for the difference between non-Great Lakes samples and
Great Lakes samples may be due to contaminants associated with agricultural practices in the
non-Great Lakes areas (i.e., herbicides or pesticides) that contributed to the toxicity of these
samples independently of elevated metals, PAHs, or PCBs.
17
-------
In summary, there was generally an increase in the incidence of toxicity with increasing mean
PEC quotients within most of the regions, basins, and areas for all three tests (Tables 3 to 5). For
the HA 10 and HA28 tests, the incidence of toxicity for samples from each of the Great Lakes and
within the areas of each Great Lake was relatively consistent with the overall pattern of toxicity
in the entire database. However, the relationship between the incidence in toxicity and mean
quotients in the CS10 test was more variable among geographic areas compared to either the
HA 10 or HA28 test. The results of these analyses indicate that the consensus-based PECs can be
used to reliably predict toxicity of sediments on both a regional and national basis.
Future analyses planned for the database
This paper presents results of the first analyses completed on the entire freshwater sediment
database. Some of the additional analyses planned for the database (beyond the scope of this
paper) are listed below:
Approaches for designating samples as toxic. In the present study, samples were
designated as toxic in the three tests based on a significant reduction in survival or a
sublethal endpoint (typically growth) relative to a control or reference sediment. This
designation of toxicity utilized the results of statistical analyses presented in each of the
original studies. Alternatively, Long et al. (1998a) classified sediments in a marine
amphipod database as either marginally toxic (significantly reduced relative to the
control) or as highly toxic (significantly reduced relative to the control with a reduction
greater than a minimum significant difference; MSD). The MSD was established by Long
et al. (1998a) using a power analysis of data from 10-day marine amphipod tests (Thursby
et al. 1997). Long et al. (1998a) and Field et al. (1999) reported reduced variability in the
relationship between toxicity and sediment contamination when toxicity was evaluated
using a standardized approach. Future analyses of the freshwater database will compare
relationships between toxicity and contamination using marginally toxic versus highly
toxic sediments. This classification may be based on a power analysis to establish an
MDS for each endpoint in each of the three tests. Alternatively, the MDS for each
endpoint may be established using results of round-robin testing (USEPA 2000). This
latter procedure is currently being investigated for use in classifying the toxicity of
freshwater samples in a revision to the report on the incidence and severity of sediment
contamination in surface waters of the United States (USEPA 1997; Scott Ireland,
USEPA, Washington, DC, personal communication). Classification of sediments as
either marginally or highly toxic may help to reduce the variability in the observed
relationship between toxicity and mean quotients. Additionally, this analysis may help to
address the higher incidence of toxicity at mean quotients of <0.1 in the HA 10 and CS10
tests compared to the HA28 tests.
• Logistic-regression modeling of the freshwater database. Field et al. (1999) described
a procedure for evaluating matching marine sediment chemistry and toxicity data using
logistic regression models. These models can be used to estimate the probability of
observing an effect at any contaminant concentration. The models were developed for
marine amphipods using a large database (n=2524) from a variety of geographic areas.
18
-------
The results of preliminary analyses using these techniques indicate that the freshwater
database may have too few samples to adequately develop these regression models.
However, evaluations are ongoing to determine how well the regression models
developed from the marine database can be used to predict responses of organisms in the
freshwater database.
Optimal list of analytes for broad scale application and test of the relative efficacy of
the mean versus the sum PEC quotient. Further comparisons are needed of approaches
for calculating the PEC quotient. Calculation of a mean quotient may include analytes
which have limited toxicological importance. The significant toxicological contribution
of a few chemicals may be averaged out by the use of the mean quotient. Analyses need
to be conducted to identify an optimal list of analytes for broad scale application and to
test the relative efficacy of the mean versus the sum PEC quotient. Even if the mean
quotient continues to be the best at predicting toxicity, identification of an optimal list of
analytes would be useful. If a consistent set of analytes is applied, the sum and the mean
quotient will mathematically equivalent.
Influence of sediment grain size and ammonia on the incidence of toxicity. Data on
grain size and pore-water ammonia, pH, and water hardness were obtained for some
studies evaluated in this paper. Future analyses of the database will evaluate the influence
of either grain size or ammonia on the response of the test organisms in the HA10, HA28,
or CS10 tests. ASTM (2000) and USEPA (2000) provide the following guidance for
dealing with the influence of grain size or ammonia in toxicity tests with freshwater
sediments.
Natural physico-chemical characteristics such as grain size or organic carbon can
potentially influence the response of test organisms. ASTM (2000) and USEPA (2000)
summarize results from a variety of studies and conclude H. azteca can tolerate a wide
range in grain size and organic matter in 10- to 42-day tests with sediments. Larvae of C.
teutons in 10-day tests were tolerant of a wide range of grain size if ash-free dry weight
was used to account for the influence of inorganic material in the gut. The content of
organic matter in sediments does not appear to affect survival of C. tentans larvae in
sediments; but, may be important with respect to larval growth. Future analyses of the
database will evaluate potential relationships between grain size or organic carbon of
sediments on the toxicity in the HA10, HA28, or CS10 tests.
Ammonia in pore water may contribute to the toxicity of some sediments in fresh water.
The toxicity of ammonia to C. tentans is dependent on pH whereas the toxicity of
ammonia to H. azteca is also dependent on water hardness (ASTM 2000, USEPA 2000).
Water-only LC50 values may provide suitable screening values for potential ammonia
toxicity; however, higher concentrations may be necessary to actually induce ammonia
toxicity in sediment exposures, particularly for H. azteca due to avoidance. ASTM (2000)
and USEPA (2000) cite studies which describe procedures for conducting toxicity
identification evaluations (TIEs) for pore-water or whole-sediment samples to determine
if ammonia is contributing to the toxicity of sediment samples. Future analyses of the
19
-------
database will evaluate potential relationships between pore-water ammonia, pH, and
water hardness on toxicity in the HA 10, HA28, or CS10 tests. These analyses will focus
on the HA28 test due to limited data available for the HA10 or CS10 tests.
Guidance manual for conducting an integrated assessment of sediment
contamination. Work is in progress to develop a guidance manual for USEPA Great
Lakes National Program Office that can be used to assess sediment quality and determine
the need for remediation at a site. Specifically, this guidance manual will describe
procedures for combining results of sediment toxicity, bioaccumulation, benthic
communities surveys, and sediment chemistry in an integrated evaluation of ecological
risk. Development of this guidance manual is being coordinated with ongoing efforts to
develop similar guidance by the Department of the ulterior and by Environment Canada.
Additional analyses of the freshwater database will be used to help develop this guidance.
Specifically, these analyses will focus on determining which endpoints provide the most
sensitive and cost effective measures of sediment contamination.
20
-------
Summary
The primary objectives of this study were to: (1) evaluate the ability of consensus-based PECs to
predict toxicity in a freshwater database for field-collected sediments in the Great Lakes basin;
(2) evaluate the ability of these PECs to predict sediment toxicity on a regional geographic basis
elsewhere in North America; and (3) compare approaches for evaluating the combined effects of
chemical mixtures on the toxicity of field-collected sediments. When mean quotients were
calculated using an approach of equally weighting up to 10 reliable PECs (PECs for metals, total
PAHs, total PCBs, and sum DDE), there was an increase in the incidence of toxicity with an
increase in the mean quotient in all three tests. A consistent increase in the toxicity in all three
tests occurred at mean quotients of >0.5. However, the overall incidence of toxicity was greater
in the HA28 test compared to the short-term tests. The reason for the higher incidence of toxicity
in the HA28 test compared to the short-term tests may be due to the duration of the exposure or
the sensitivity of growth in the longer HA28 test. However, comparisons of the sensitivity
between these tests needs to be made with some caution. There were very few samples in the
database where tests were conducted using splits of the same samples. Therefore, the differences
observed in the responses of organisms may also be due to differences in the types of sediments
evaluated in the individual databases for each test. Nevertheless, it appears that longer-term tests
in which survival and growth are measured tend to be more sensitive than shorter-term tests, with
acute to chronic ratios on the order of 6 indicated for H. azteca.
We were also interested in determining the predictive ability of PEC quotients for major classes
of compounds. Therefore, we evaluated the incidence of toxicity based on a mean quotient for
metals, a quotient for total PAHs, or a quotient for total PCBs. Different patterns of toxicity
associated with the other procedures for calculating quotients were observed. For example in the
HA28 test, a relatively abrupt increase in toxicity was associated with elevated PCBs alone or
with elevated PAHs alone, compared to the pattern of a gradual increase toxicity observed with
quotients calculated using a combination of metals, PAHs, and PCBs. These analyses indicate
that the different patterns in toxicity may be the result of unique chemical signals associated with
individual contaminants in samples. While mean quotients can be used to classify samples as
toxic or non-toxic, individual quotients might be useful in helping to identify substances that may
be causing or substantially contributing to the observed toxicity.
We chose to make comparisons across geographic areas using mean quotients calculated by
equally weighting the contribution of the three major classes of compounds (metals, or PAHs, or
PCBs). This approach assumes that these three diverse groups of chemicals exert some form of
joint toxic action. Use of this approach also maximized the number of samples that were used to
make comparisons across geographic areas. Generally, there was an increase in the incidence of
toxicity increasing with mean PEC quotients within most of the regions, basins, and areas for all
three toxicity tests. The incidence of toxicity for samples from each of the Great Lakes and
within the areas of each Great Lake was relatively consistent with the overall pattern of toxicity
in the entire database for the HA 10 and HA28 tests. However, the relationship between the
incidence in toxicity and mean quotients in the CS10 test was more variable among geographic
areas compared to either the HA10 or HA28 test. Results of these analyses indicate that the
21
-------
PECs developed using a database from across North America can be used to reliably predict
toxicity of sediments on a regional basis.
One of the primary goals of sediment quality assessments is to evaluate the effects of
contaminated sediments on benthic communities in the field (Ingersoll et al. 1997). Swartz et al.
(1994) evaluated sediment quality conditions along a sediment contamination gradient of total
DDT using information from 10-day toxicity tests with amphipods, sediment chemistry, and the
abundance of benthic amphipods in the field. Survival of amphipods (Eohaustorius estuarius,
Rhepoxynius abronius, and H. azteca) in laboratory toxicity tests was positively correlated to the
abundance of amphipods in the field and negatively correlated to total DDT concentrations. The
toxicity threshold for amphipods in 10-day sediment toxicity tests was about 300 ug total DDT/g
organic carbon. The threshold for reduction in abundance of amphipods in the field was about
100 ug total DDT/g organic carbon. Therefore, correlations between toxicity, contamination, and
the status of benthic macroinvertebrates in the field indicate that 10-day sediment toxicity tests
can provide a reliable indicator of the presence of adverse levels of sediment contamination in
the field. However, these short-term toxicity tests may be under protective of sublethal effects of
contaminants on benthic communities in the field.
Similarly, Canfield et al. (1994, 1996, 1998) evaluated the composition of benthic invertebrate
communities in sediments in a variety of locations including the Great Lakes, the upper
Mississippi River, and the Clark Fork River in Montana. Results of these benthic invertebrate
community assessments were compared to SQGs (ERMs) and 28-day sediment toxicity tests with
H. azteca. Good concordance was evident between measures of laboratory toxicity, SQGs, and
benthic invertebrate community composition in extremely contaminated samples. However, in
moderately contaminated samples, less concordance was observed between the composition of
the benthic community and either laboratory toxicity tests or SQGs. The laboratory toxicity tests
better identified chemical contamination in sediments compared to many of the commonly used
measures of benthic invertebrate community structure. As the status of benthic invertebrate
communities may reflect other factors such as habitat alteration in addition to effects of
contaminants, the use of longer-term toxicity tests in combination with SQGs may provide a
more sensitive and protective measure of potential toxic effects of sediment contamination on
benthic communities compared to use of 10-day toxicity tests.
This paper presents results of the first analyses completed on the entire freshwater sediment
database. Some of the additional analyses planned for the database that are beyond the scope of
this paper include: (1) comparing approaches for designating samples as toxic; (2) evaluating
logistic-regression models; (3) identifying a list of optimal analytes for broad scale application
and testing the relative efficacy of the mean verses the sum PEC quotient; (4) evaluating the
influence of grain size and ammonia on the incidence of toxicity; and (5) developing a guidance
manual for conducting an integrated assessment of sediment contamination.
22
-------
References
American Society for Testing and Materials (ASTM). 2000. Standard test methods for measuring
the toxicity of sediment-associated contaminants with freshwater invertebrates. E1706-
00. In ASTM Annual Book of Standards, Vol. 11.05, Philadelphia, PA.
Barrick R, Becker S, Brown L, Seller H, Pastorok, R. 1988. Sediment quality values refinement:
1988 Update and Evaluation of Puget Sound AET, Vol. I. PTI Contract C717-01, PTI
Environmental Services, Bellevue, WA.
Canadian Council of Ministers of the Environment (CCME). 1995. Protocol for the derivation of
Canadian sediment quality guidelines for the protection of aquatic life. Prepared by the
Technical Secretariat of the CCME Task Group on Water Quality Guidelines, Ottawa,
Canada.
Canfield TJ, Kemble NE, Brumbaugh WG, Dwyer FJ, Ingersoll CG, Fairchild JF. 1994. Use of
benthic invertebrate community structure and the sediment quality triad to evaluate metal-
contaminated sediment in the Upper Clark Fork River, Montana. Environ Toxicol Chem
13:1999-2012.
Canfield TJ, Dwyer FJ, Fairchild JF, Haverland PS, Ingersoll CG, Kemble NE, Mount DR, La
Point TW, Burton GA, Swift MC. 1996. Assessing contamination in Great Lakes
sediments using benthic invertebrate communities and the sediment quality triad
approach. J Great Lakes Res 22:565-583.
Canfield TJ, Brunson EL, Dwyer FJ, Ingersoll CG, Kemble NE. 1998. Assessing sediments from
the upper Mississippi river navigational pools using a benthic community invertebrate
evaluations and the sediment quality triad approach. Arch Environ Contain Toxicol
35:202-212.
Environment Canada and Ministere de rEnvionnement du Quebec (EC and MENVIQ). 1992.
Interim criteria for quality assessment of St. Lawrence River sediment. ISBN 0-662-
19849-2. Environment Canada, Ottawa, Ontario.
Fairey R, Long ER, Roberts CA, Anderson BS, Phillips BM, Hunt JW, Puckett HR, Wilson CJ,
Kapahi G, Stephenson M. 2000. A recommended method for calculation of sediment
quality guideline quotients. Manuscript in preparation.
Field LJ, MacDonald DD, Norton SB, Severn CG, Ingersoll CG. 1999. Evaluating sediment
chemistry and toxicity data using logistic regression modeling. Environ Toxicol Chem
18:1311-1322.
Ingersoll CG, Haverland PS, Brunson EL, Canfield TJ, Dwyer FJ, Henke CE, Kemble NE. 1996.
Calculation and evaluation of sediment effect concentrations for the amphipod Hyalella
azteca and the midge Chironomus riparius. J Great Lakes Res 22:602-623.
23
-------
Ingersoll CG, Ankley GT, Baudo R, Burton GA, Lick W, Luoma S, MacDonald DD, Reynoldson
TB, Solomon KR, Swartz RC, Warren-Hicks WJ. 1997. Work group summary report on
uncertainty evaluation of measurement endpoints used in sediment ecological risk
assessment. In: Ingersoll CG, Dillon T, Biddinger GR, editors. Ecological risk assessment
of contaminated sediment. Pensacola FL: SET AC Press, p 297-352.
Ingersoll CG, Brunson EL, Dwyer FJ, Hardesty DK, Kemble NE. 1998. Use of sublethal
endpoints in sediment toxicity tests with the amphipod Hyalella azteca. Environ Toxicol
Chem 17:1508-1523.
Ingersoll CG, MacDonald DD. 1999. An assessment of sediment injury in the West Branch of the
Grand Calumet River. Report prepared for the Environmental Enforcement Section,
Environment and Natural Resources Division, U.S. Department of Justice, Washington,
DC, January 1999.
Long ER and Morgan LG. 1991. The potential for biological effects of sediment-sorbed
contaminants tested in the National Status and Trends Program. NOAA Technical
Memorandum NOS OMA 52. National Oceanic and Atmospheric Administration,
Seattle, WA, 175 pp + appendices.
Long ER and MacDonald DD. 1998. Recommended uses of empirically-derived sediment quality
guidelines for marine and estuarine ecosystems. Human Ecolog Risk Assess 4:1019-
1039.
Long ER, MacDonald DD, Smith SL, Calder FD. 1995. Incidence of adverse biological effects
within ranges of chemical concentrations in marine and estuarine sediments.
Environmental Management 19:81-97.
Long ER, Field LJ, and MacDonald DD. 1998a. Predicting toxicity in marine sediments with
numerical sediment quality guidelines. Environ Toxicol Chem 17:714-727.
Long ER, MacDonald DD, Cubbage JC, Ingersoll CG. 1998b. Predicting the toxicity of
sediment-associated trace metals with simultaneously extracted trace metal: acid volatile
sulfide concentrations and dry weight-normalized concentrations: A critical comparison.
Environ Toxicol Chem 17:972-974.
MacDonald DD, Ingersoll CG, Berger T. 2000a. Development and evaluation of consensus-
based sediment quality guidelines for freshwater ecosystems. Arch Environ Contam
Toxicol 39:20-31.
MacDonald DD, DiPinto LM, Field J, Ingersoll CG, Long ER, Swartz RC. 2000b. Development
and evaluation of consensus-based sediment effect concentrations for polychlorinated
biphenyls (PCBs). Environ Toxicol Chem 19:1403-1413.
24
-------
Persaud D, Jaagumagi R, Hayton A. 1993. Guidelines for the protection and management of
aquatic sediment quality in Ontario. Water Resources Branch, Ontario Ministry of the
Environment, Toronto, ONT, 27 p.
SigmaPlot 1997. Transforms and Regressions. SigmaPlot 4.0 for Windows 95, NT, and 3.1.
SPSS Inc., Chicago, IL.
Smith SL, MacDonald DD, Keenleyside KA, Ingersoll CG, and Field J. 1996. A preliminary
evaluation of sediment quality assessment values for freshwater ecosystems. J Great
Lakes Res 22:624-638.
Swartz RC, Cole FA, Lamberson JO, Ferraro SP, Schults DW, DeBen WA, Lee H, Ozretich RJ.
1994. Sediment toxicity, contamination and amphipod abundance at a DDT and dieldrin-
contaminated site in San Francisco Bay. Environ Toxicol Chem 13: 949-962.
Swartz RC. 1999. Consensus sediment quality guidelines for PAH mixtures. Environ Toxicol
Chem 18:780-787.
Thursby GB, Heltshe J, Scott KJ. 1997. Revised approach to toxicity test acceptability criteria
using a statistical performance assessment. Environ Toxicol Chem 14:1977-1987.
U.S. Environmental Protection Agency (USEPA). 1996. Calculation and evaluation of sediment
effect concentrations for the amphipod Hyalella azteca and the midge Chironomus
riparius. EPA 905/R-96/008, Chicago, IL.
U.S. Environmental Protection Agency (USEPA). 1997. The incidence and severity of sediment
contamination in surface waters of the United States, Volume 1: National Sediment
Quality Survey. EPA 823/R-97/006, Washington, DC.
U.S. Environmental Protection Agency (USEPA). 2000. Methods for measuring the toxicity and
bioaccumulation of sediment-associated contaminants with freshwater invertebrates,
second edition, EPA 600/R-99/064,Washington, DC.
25
-------
(A) 10- to 14-d Hyalella azteca
100 T
8° -
60 H
o
8
40 -
•g 20 H
_c
0 -
0.01
0.1
10
100
(B) 28- to 42-d Hyalella azteca
100 n
1 60 H
8 4(H
0
I 20 H
0 -
0.01
0.1
10
100
(C) 10- to 14-d Chironomus spp.
100 -i
80-
.o
I
M—
O
8
(!)
'O
60 -
40-
20 -
0 -
0.01
0.1 1 10
Geometric mean of mean PEC-Q
100
Figure 1. Relationship between the geometric mean of the mean PEC quotients and the incidence
of toxicity in the three tests, based on survival or growth. The average is based on 20 samples/data
point in Figures 1A and 1C (except for the data point for the highest average of the mean quotients
where n=19 to 30), and 10 samples/data point for Figure 1B (see Appendix 1 for detail).
-------
(A) 10- to 14-d Hyalella azteca
100 -I
80 •
•8
O
40H
o
2 20 H
c
0 -
1^=0.73
0.01
0.1
1^=0.78
—•—Survival or growth
—^—Survival only
10
100
(B) 28- to 42-d Hyalella azteca
100 n
0.01
(C) 10- to 14-d Chironomus spp.
100 -,
80 •
60 H
40H
O
£
8
V
I 20 H
0-
= v
0.01
0.1 1 10
Geometric mean of mean PEC-Q
100
Figure 2. Relationship between the geometric mean of the mean PEC quotients and the incidence
of toxicity in the three tests, based on survival or growth, or based on survival alone. The dotted line
represents a 50% incidence of toxicity. See legend for Figure 1 for additional detail.
-------
Table 1. Sediment quality guidelines that reflect probable effect concentrations (PECs; i.e., above which harmful effects are likely to be
observed; MacDonald et al. 2000a). An "*" designates a reliable PEC (>20 samples and >75% correct classification as toxic).
Substance
Metals (in rag/kg D W)
Arsenic
Cadmium
Chromium
Copper
Lead
Mercury
Nickel
Zinc
PEL
17
3.53
90
197
91.3
0.486
36
315
SEL
33
10
110
110
250
2
75
820
TET
17
3
100
86
170
1
61
540
ERM
85
9
145
390
110
1.3
50
270
PEL-HA28
48
3.2
120
100
82
NG
33
540
Consensus-Based PEC
33.0*
4.98*
111*
149*
128*
1.06
48.6*
459*
Polycyclic Aromatic Hydrocarbons (in ug/kg D\V)
Anthracene
Fluorene
Naphthalene
Phenanthrene
Benz[a]anthracene
Benzo(a)pyrene
Chrysene
Fluoranthene
Pyrene
Total PAHs
NG
NG
NG
515
385
782
862
2355
875
NG
3700
1600
NG
9500
14800
14400
4600
10200
8500
100000
NG
NG
600
800
500
700
800
2000
1000
NG
960
640
2100
1380
1600
2500
2800
3600
2200
35000
170
150
140
410
280
320
410
320
490
3400
845
536
561*
1170*
1050*
1450*
1290*
2230
1520*
22800*
Polychlorinated Biphenyls (in fig/kg D W)
Total PCBs
277
5300
1000
400
240
676*
Organochlorine Pesticides (in fig/kg DW)
Chlordane
Dieldnn
Sum ODD
Sum DDE
Sum DDT
Total DDTs
Endrin
Heptachlor Epoxide
Lindane (gamma-BHC)
8.9
6.67
8.51
6.75
NG
4450
62.4
2.74
1.38
60
910
60
190
710
120
1300
50
10
30
300
60
50
50
NG
500
30
9
6
8
20
15
7
350
45
NG
NG
NG
NG
NG
NG
NG
NG
NG
NG
NG
17.6
61.8
28.0
31.3*
62.9
572
207
16.0
4.99
PEL = Probable effect level, dry weight (Smith et al. 1996).
SEL = Severe effect level, dry weight (Persaud el al. 1993).
TET = Toxic effect threshold; dry weight (EC & MENVIQ 1992).
ERM = Effects range median; dry weight (Long and Moigan 1991).
PEL-HA28 = Probable effect level for Hyalella azteca; 28-day test; dry weight (USEPA 1996).
NG = No guideline.
-------
Table 2. Incidence of sediment toxicity within ranges of PEC quotients (calculated using various approaches) for
freshwater tests based on survival or growth.
, Incidence of toxicity (%) based on mean PEC quotients (number of samples in parentheses)
PEC quotient
<0.1
Hvalella azteca 10- to 14-day tests
Mean -all 20 (102)
Mean - metals
Total PAHs
Total PCBs
Mean - MPP (and)
Mean - all (selectl)
Mean - all (selectl)
Mean - metals (select2)
Total PAHs (selectl)
Total PCBs (select2)
Mean - MPP (or)
20 (104)
20 (178)
26 (109)
19 (79)
19 (36)
22 (46)
23 (40)
25 (123)
20 (98)
18 (147)
Hvalella azteca 28- to 42-day tests
Mean -all 8 (51)
Mean - metals
Total PAHs
Total PCBs
Mean - MPP (and)
Mean - all (selectl)
Mean - all (select2)
Mean - metals (select2)
Total PAHs (select2)
Total PCBs (selectl)
Mean - MPP (or)
8 (50)
17 (98)
4 (26)
4 (45)
5 (44)
5 (40)
5 (40)
8 (57)
4 (26)
10 (63)
Chironomus snp. 10- to 14-day tests
Mean - all 20 (90)
Mean - metals
Total PAHs
Total PCBs
Mean - MPP (and)
Mean - all (selectl)
Mean - all (select2)
Mean - metals (select2)
Total PAHs (select2)
Total PCBs (select2)
Mean - MPP (or)
22 (88)
14 (178)
46 (91)
29 (21)
29 (7)
33 (6)
8 (12)
27 (64)
48 (58)
20 (121)
0.1 to <0.5
16 (336)
19 (354)
26 (160)
21 (91)
26 (89)
23 (128)
22 (121)
24 (139)
33 (76)
25 (61)
16 (288)
11 (47)
20 (51)
61 (46)
6 (35)
6 (18)
5 (17)
5 (21)
25 (24)
64 (37)
6 (35)
13 (39)
19 (333)
23 (338)
33 (133)
22 (49)
35 (78)
37 (98)
35 (98)
43 (107)
33 (73)
23 (31)
17 (313)
0.5 to <1.0
36 (80)
39 (72)
23 (47)
46 (49)
38 (34)
32 (38)
33 (40)
33 (45)
35 (20)
47 (43)
37 (73)
57 (28)
62 (37)
56 (9)
17 (12)
50 (18)
50 (22)
50 (22)
60 (33)
55 (9)
17 (12)
56 (27)
33 (95)
25 (89)
57 (28)
36 (33)
35 (26)
29 (34)
33 (36)
22 (36)
77 (13)
34 (32)
43 (63)
1.0 to <5.0
52 (103)
63 (72)
37 (54)
51 (47)
49 (35)
67 (42)
64 (39)
81 (31)
49 (33)
47 (34)
41 (92)
NC2
NC
NC
NC
NC
NC
NC
NC
NC
NC
NC
44 (75)
39 (61)
67 (33)
53 (51)
50 (34)
47 (17)
61 (23)
75 (12)
85 (20)
35 (34)
43 (88)
>1.0
57 (152)
63 (51)
57 (103)
60 (77)
66 (64)
70 (64)
73 (59)
86 (42)
64 (47)
59 (64)
54 (162)
91 (34)
86 (22)
86 (7)
97 (36)
100 (28)
100 (26)
100 (26)
100 (12)
100 (6)
97 (36)
97 (31)
51 (114)
44 (84)
74 (53)
58 (74)
60 (52)
66 (38)
68 (37)
82 (22)
81 (27)
48 (56)
52 (132)
>5.0 "
67 (49)
62 (21)
80 (49)
73 (30)
86 (29)
77 (22)
90 (20)
100 (11)
100 (14)
73 (30)
71 (70)
NC
NC
NC
NC
NC
NC
NC
NC
NC
NC
NC
64 (39)
57 (23)
85 (20)
70 (23)
78 (18)
81 (21)
79 (14)
90 (10)
71 (7)
68 (22)
68 (44)
Total
lumber of
samples
670
623
488
326
266
266
266
266
266
266
670
160
160
160
109
109
109
109
109
109
109
160
632
599
392
247
177
177
177
177
177
177
629
'Description of quotients:
Mean - all: based on the PECs for each of the reliable metals, total PAHs, total PCBs, or sum DDE listed in Table 1 (n = 1 to 10
quotients/sample).
Mean - metals: based on the PECs for reliable metals listed in Table 1 (n = 1 to 7 quotients/sample).
Total PAHs: based on the PEC for total PAHs listed in Table 1 (n = 1 quotient/sample).
Total PCBs: based on the PEC for total PCBs listed in Table 1 (n = 1 quotient/sample).
Mean - MPP (and): based on reliable PECs for the average metals quotient, the total PAH quotient, and the total PCB quotient (n = 3
quotients/sample).
Mean - all (selectl): based on the PECs for each of the reliable metals, total PAHs, total PCBs, or sum DDE listed in Table 1 using only samples
with measured metals, total PAHs and total PCBs (n = 4 to 10 quotients/sample).
Mean - all (select2): based on the PECs for each of the reliable metals, total PAHs, or total PCBs listed in Table 1 using only samples with
measured metals, total PAHs and total PCBs (n = 3 to 9 quotients/sample).
Mean - metals (selectl), total PAHs (selectl), or total PCBs (selectl): based on the average quotient for metals, the quotient for total PAHs, or
the quotient for total PCBs using only samples with measured metals, total PAHs and total PCBs.
Mean - MPP (or): based on the reliable PECs for the average metals quotient, total PAH quotient, or total PCB quotient (n = 1 to 3
quotients/sample).
2NC: Not calculated.
-------
y S
a "
0*
S I
l«
£• 5
*3 S
ill
III
lit
lil
u S .tS
fl * *
^. J2 T3
fO fl U
**?
•ssi
H &.S
w
1
|
w
y
a.
a
i
B
O
•2
y,
^
'3
1
o
a
v
2
*£
=.
o
A1
0
g
0
o
7
0
tfl
Oi
Z
p
M
*
o
^1
o
m
O
— .
O
«
o
Fox River
^
8
f!
S
m
p;
t^
CN
^>
CN
—
0
—
^
^
Indiana Harbor
o
0
0
o
^>
o
0
rt
o
Lake Michigan
0
V)
i
v»
O
OO
o
0
0
V)
i
Menominee River
—
8
CN
O
—
O
o
CN
£
TT
Vl
(^
00
•^
CN
Sheboygan Harbor
—
8
-H
8
0
— -
8
^-i
0
0
CO
s
Waukegan Harbor
0
CN
O
(N
O
O
0
0
(N
O
Waushara County
TT
8
bo'
w
f-
^
S
O
0
0
oo
Vl
White Lake Montague
m
0
^
0
—
o
~H
8
—
Os
o
\o
"-•
—
0
§
OJ
0
o
0
—
8
CN
m
0
m
— -
oo
Genesee River
o
o
o
o
K
o
0
r^
o
Oswego River
m
0
•^
0
—
o
0
CN
0
o
^o
o
Sturgeon Lake
P
00
o
VI
m
00
—
oo
m
~
CN
VI
0\
K'
;•",
OO
St. Louis River
i*
?!
9
•n
s
Vl
3
CN
Os
Vl
CN
*
^>
O
VI
fN
"-'
O
S
I
fl
1
o
CN
O
CN
O
_4
0
— •
0
^>
VI
t-
00
oo
A
i
1
0
i
K
o
CN
O
CN
O
~M
O
—
o
0
<3-
o
1
B
M
1
O
O
0
0
o
•<±
V(
TJ-
^
Temblader River
cs
CN
O
Vl
CN
O
Vl
K'
Z
o
CN
Os
CN
00
"- '
CN
CN
Miami-Dade County
«
E
c
1
-+'
o
Vl
^
^
^
CN
0
^O
0
VI
O
00
^"
I*
New York Lakes
l_
V
2
§
Tl
B
^
1
C
JU
1
f\
K
o
°
Vl
O
0
O
OS
n
*-
o
o
-"
VI
Lower Columbia River
2
V}
Trinity River
«;
1
O
U
£
^)
£
*fr
o
VI
OO
00
CN
m
CN
VI
2
O
O
CN
"-'
CN
B
R
b.
z
M
C
V.
fN
O
VD
O
^>
O
CN
8
t^-
oo
o
o
Vl
•^
•"
Columbia River*
•
-------
u S;
«!
* s
£ S
an ,_
L. e
•a S
11
C **
s ^
ts
ll
•S .§
II
1 S;
'* s
2 i
U 0
A 3
« g
Is
II
ll
Ij
Q "J3
a 2
|1
"3 4)
5j o
V o«
5 *
•21
a •§
1 I
u -c
0. 0
"S &
Ir
2 i!
.s *
2 J-
l-s
2- B
:M
s|»
| S f
111
8 « 1
B « _
U U .S
1 » s
^ B T)
•» B 5
* '* %
£ &.s
1
•c
er
U
w
a.
JJ
i
B
e
•g
*
a
^
oftoxicitj
1
I
0
V)
A
o
A1
O
2
s
o
7
2
V)
e
1 to <0.5
0
i"7
1 ^
o
s
s
^
1
cs
en
«3
8
—
S;
in
i^.
ON
m
m
0
°
r*i
All (with controls)
North America
M)
8
F-H
&
*n
C*w
o
£
m
2
«
m
All (w/o controls)
v,
8
K
g
fS
g
t-
£
oo
2
o
(N
r*.
mit
Great Lakes
o
fS
g
(S
g
<*1
5
O
o
»n
o
00
Buffalo River
Lake Erie
o
(N
g
Lake Micl
m
8
•^
g
.—
g
0
0
0
Tf
8
Indiana Harbor
IN
8
ON
g
r-
g
(N
g
ts
o
VI
0
m
Waukegan Harbo
_
8
rf-
r-
r^
r-*
NO
o
o
n
TT
m
fc.
Upper Mississippi Rive
o
,— i
o
^H
O
o
^>
o
0
*n
O
Mobile Bay
|
Gulf of M
State of Alabama
o
o
0
o
(N
S
•&
«
Pend Orel
State of Montana
0
o
o
0
CN
O
-------
i*
i?
| 3
S t
II
if
i?
li
n
<£ "•"
tl
o er
II
E «
II
HI
111
lei
« f-g
« .s ^
1
g
•S
•3
V
00
ft
CN
OO
ct
ct
ft
oo
CN/
r-
00
o
CN
1
s
"aS
i
.0
oo
p
2
o
5
VI
K
— 1
CN,
2
1
reat Lakes
O
— "
o
s
o
o
t^
r-
NO
*-"
oo
ON
CS
ft
•?
H
S
J
o
o
CN
o
ft
£
o
o
Q
Buffalo River 4(
o
8
Tf
8
OO
1
o
o
CN
Clinton River IOC
o
o
o
-*
§
o
CN
S
Elmira, Ontario, Canada 33
c~
o
s
TT
£
NO
ft
oT
"-'
"*
CN
s.
S
1
O
O
O
NO
o
o
s
£
oo
oo
s
—
s
NO
s
ON
— '
r-
CN
S
B
e
8
S
J
o
o
fN
O
CN
O
NO
o
o
o
>
s
i
1
oo
00
ct
*
ON
3
^-
S
m
—
S
*
oo
c.
St. Mary's River i.
oo
P!
r*
R
VI
CN
O
^
'"
ft
ft
ON
P
U
.s
S
jj
J
CN
O
O
O
o
M
o
o
C"
j
1
i
o
o
o
o
0
cr
o
o
°
s
1
o
o
oo
o
ct
0
G^
o
o
oo
1
U
1
1
~"
s
o
~-
o
o
~-
o
0
ft,
s
5
I
oo
8
8
NO
1
o
o
o
2
§
1
o
o
—
o
o
o
°
Kenosha Harbor (
o
o
o
o
CN
I
cc
"5
q
«
o
a
CM
S
—
o
o
rv
r—
o
^
VI
I
•3
S
O
O
—i
O
S
o
°
Lake Superior <
o
o
0
—
o
o
"
Wisconsin State (
NO"
o
a
rt
oo
oo
-------
e
1
E
c
•o
=
0*
a.
2
•o
1
(M
9
t
g
_
^
9
1
a
^
o.
W
tn
g
O
£
s
&
•3
•
2
0
,
2
2
i
jS
(U
1
1
rj-
3
OO
(N
1
1
^
"w
|
W
i
—
0
o
£»
o
1
g
CO
-g
8
Q
g
*c5
1
CO
_£,
*c!
o
1
E
3
^
P
™
O
1
>
5
v»
1
1
'g
CO
^
&
1
bo
O
"c3
>
'S
3
2
u ^
gb
c o
S a
o ^
O t*-
0
Z
u ^
H
c ry
g n
1 1
O <5
Z
n
H
cx
g
1 i
o
.a ;?-
x =••
o ^ —
H
§ 1
E S
8 g
o
0 ^
§ 1
i i
O noo>no*A»noooil/^'nino*noovC>
rn^-fSr-j— -fS »—cN^---fS — •— < ts P*I m CN •- ' fi'^r^*'O
vir^mooOs — t^-r^o^ft^-r^ioooov'^Tj-(Nt^ov£>c\ooo — \oo^t^^t~-o oo^rs in
OOOOOOOOOOOOOOOOOOOOOOOOO— '— 'f-lfiv^r^ ON O fi
(N ON '
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOON
oooooooooooooooo
vovo — -^-.Tj-^rJ-Ht^OvO-HfSONO -ONON fS
oooooo22S®£Sooo^o ooSfs ™
ooooooooooooo ON o O
OOOOOO-H— '•—r->>nioooo—'O oooir- ^
oo'o'o'o'o'o'o'o'o'o'o'.o'fNfir^ ^H^-O
oooooooooooooooo
O^^OO^O^O^OOO^OOO^^jO^OOO^^OOOOp^O
fr
[G
ssIsisisSBIBSsSSsill^is^issisSRSI ilS 1 5
OOOOOOOOOOO O'o'o'o'o'o'o'o'o'o'o'o'o'o*'— -H.-.rvifi^O'-'fS (S O O\ *o
~* "^ (S ' ^
§
•^~~1 *2
^ 'y
1 -1
O O *f) O O O O Vi O fi O vi O Vi vi O O V> vi O Vi O O vi in vi vj vi O vi vi v> O + S£
"^ •— ' •— ' (N t^- CN ON ^ o ON rt cs ^~* r^~ ft ^J~ rs '"O *** ^o ON o t^ ON h^ ••* *&
OOOO'oo'o'oOOOOOO'doo'OOOOO'o'oO — ^H^fNft^O^HfN TTOvi 0*
n %
g s
•a tS
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO J3 «
olcsrsr-lf^rscNr^r^rscNr^fStSfSfStscsr^r^olcNfSrscsr^cscsf^r^cNtNf^ g^ 5
w « js >^ £
------- |