&EPA
United States Office of Research and
Environmental Protection Development
Agency Washington DC 20460
EPA/620/R-93/006
June 1993
Virginian Province
Demonstration Report
EMAP-Estuaries: 1990
Environmental Monitoring and
Assessment Program
-------
-------
EPA/620/R-93/006
June 1993
VIRGINIAN PROVINCE DEMONSTRATION REPORT
EMAP-ESTUARIES - 1990
Prepared by
Stephen B. Weisberg
Versar, Inc.
Columbia, MD 21045
A. Frederick Holland
Marine Resources Research Institute
Charleston, SC 29412
K. John Scott
Science Applications
International Corp.
Narragansett, Rl 02882
Harold T. Wilson
Coastal Environmental Services, Inc.
Linthicum, MD 21090
Douglas G. Heimbuch
Coastal Environmental Services, Inc.
Linthicum, MD 21090
Steven C. Schimmel
U.S. Environmental Protection Agency
Narragansett, Rl 02882
Jeffrey B. Frithsen
Versar, Inc.
Columbia, MD 21045
John F. Paul
U.S. Environmental Protection Agency
Narragansett, Rl 02882
J. Kevin Summers
U.S. Environmental Protection Agency
Gulf Breeze, FL 32561
Raymond M. Valente
Science Applications
International Corp.
Narragansett, Rl 02882
Jeroen Gerritsen
Versar, Inc.
Columbia, MD 21045
Richard W. Latimer
U.S. Environmental Protection Agency
Narragansett, Rl 02882
EPA Contract No. 68-DO-0093
Project Officer
Vernon J. Laurie
Office of Modeling, Monitoring Systems,
and Quality Assurance
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC 20460
-------
NOTICE
The information in this document has been funded wholly or in part by the United States
Environmental Protection Agency (Environmental Monitoring and Assessment Program, Office
of Research and Development) under contract No. 68-DO-0093 to Versar, Inc. It has been
subjected to the Agency's peer and administrative review, and it has been approved for
publication as an EPA document. Mention of trade names or commercial products does not
constitute endorsement or recommendation for use.
The suggested citation for this report is: Weisberg, S.B., J.B. Frithsen, A.F. Holland, J.F.
Paul, K.J. Scott, J.K. Summers, H.T. Wilson, R. Valente, D.G. Heimbuch, J. Gerritsen, S.C.
Schimmel, and R.W. Latimer. 1992. EMAP-Estuaries Virginian Province 1990 Demonstration
Project Report. EPA 600/R-92/100. U.S. Environmental Protection Agency, Environmental
Research Laboratory, Narragansett, Rl.
-------
PREFACE
EMAP is a long-term, integrated, monitoring, research, and assessment program to determine
the condition of our nation's ecological resources. The program is organized into types of
environmental resources, such as estuaries. An important step in the implementation of
EMAP involves conducting regional demonstration projects. This report describes and
evaluates ah EMAP demonstration project conducted in the estuaries of the mid-Atlantic
states in 1990. The project successfully demonstrated that the EMAP sampling design
provides statistically valid measures of environmental condition; 1) a suite of biological
response, pollutant exposure, and habitat indicators can form the core of future EMAP-
Estuaries monitoring activities; 2) the EMAP sampling design provides statistically valid
measures of environmental condition; and 3) the assessment techniques appear to be a good
way to characterize conditions for resource managers.
in
-------
-------
EXECUTIVE SUMMARY
The Environmental Monitoring and Assessment Program (EMAP) is a comprehensive
environmental monitoring network designed to 1) estimate the current status and trends in the
condition of the nation's ecological resources on a regional basis, with known confidence; 2)
seek associations between human-induced stress and ecological condition; and 3) provide
periodic statistical summaries and interpretive reports on ecological status and trends to
resource managers and the public. The program was initiated to provide the information
necessary for formulating future environmental policies by answering the following questions:
What is the current extent of our ecological resources, and how are they distributed
geographically? What proportions of the resources are currently in good or acceptable
condition? What proportions are degrading or improving, in what regions, and at what rates?
Are these changes correlated with patterns in environmental stresses? Are adversely-affected
resources improving overall in response to control and mitigation programs?
Three characteristics of EMAP differentiate it from most previous environmental monitoring
programs. First, sampling in EMAP is probability-based so that estimates of status and trends
can be made with quantifiable confidence. Second, EMAP monitoring and assessments focus
on biological indicators of response to natural and human-induced stress; indicators of
pollutant exposure and habitat condition are sampled simultaneously to provide a context for
interpreting biological indicators. Third, the scale of EMAP monitoring is regional and national,
rather than local.
The program is organized into Resource Groups responsible for conducting assessments of
seven categories of environmental resources: forests, wetlands, surface waters, near coastal
waters, arid lands, agricultural ecosystems, and the Great Lakes. Administered by the EPA's
Office of Research and Development, EMAP is an integrated federal program that is planned
and implemented in cooperation with the National Oceanic and Atmospheric Administration,
the U.S. Fish and Wildlife Service, the U.S. Forest Service, the U.S. Bureau of Land
Management, the U. S. Agricultural Research Service, and the U.S. Geological Survey.
The first stage in implementing EMAP involves conducting demonstration projects for each
Resource Group. Demonstration projects provide the opportunity to illustrate the kinds of
assessments that can be conducted using EMAP data and to work with users of the data to
select the most appropriate indicators for evaluating problems of concern in that resource
category. Demonstration projects also provide information for refining the sampling design
and identifying and resolving logistical difficulties associated with regional monitoring in each
resource category on a limited scale, before EMAP monitoring is implemented nationwide.
This report describes the first EMAP demonstration project, which was conducted in estuaries
of the mid-Atlantic states (the Virginian Province) in 1990 by the Near Coastal Resource
-------
Group. Estuaries were selected for the first demonstration project because of their biological
productivity and the current intense public interest in restoring and maintaining estuarine
resources. The objectives of the project were to demonstrate the utility of regional scale
monitoring using a probability-based sampling design for assessing the condition of estuarine
resources, establish standardized methods for estuarine monitoring, identify and resolve
logistical problems associated with large-scale sampling, test and develop biological indicators
of estuarine environmental quality, and collect the data on regional-scale variability in
ecological parameters needed to evaluate and refine the sampling design. Each of these
objectives is addressed in individual sections of this report. A summary of those sections
follows.
The 1990 Demonstration Project involved 500 sampling visits to 217 sites from Cape Cod to
the mouth of the Chesapeake Bay. A series of indicators that are representative of the overall
health of estuarine resources was measured at each site. These indicators were selected to
address three major attributes of concern: 1) biotic integrity, or the existence of healthy,
diverse, and sustainable biological communities; 2) pollutant exposure, or the condition of the
chemical environment in which biota live; and 3) aesthetics, representing societal values
related to public use of estuarine resources. Habitat indicators, such as depth, salinity,
temperature, and the physical characteristics of the sediment, were also measured.
METHODS (SECTION 2) AND LOGISTICS (SECTION 3)
One characteristic of the EMAP-Estuaries program is that the entire province is sampled
within a limited time period (approximately six weeks) using standardized methods to ensure
comparability of data within and among sampling years. One of the goals of the 1990
Demonstration Project was to develop and document these standardized methods and
associated quality assurance protocols. This objective was accomplished successfully. Field
and laboratory manuals for all EMAP-Estuaries sampling activities are available. EMAP is
promoting these methods and associated QA protocols to EPA Regions, states, and local
agencies responsible for monitoring to facilitate comparability of information from multiple
monitoring programs within the province.
The 1990 Demonstration Project also served as a means for evaluating the logistical feasibility
of conducting a regional monitoring program in estuaries and identifying the most difficult
obstacles involved therein. The 1990 Demonstration Project was a logistical success. More
than 90% of the scheduled samples were collected, and they were all obtained without injury
to crew members or significant loss of equipment. More than 95% of the data collected
passed QA requirements, and specific alterations to the crew training program were identified
to resolve the few data collection problems in future years. Several innovative technologies
incorporated into the 1990 Demonstration Project for evaluation were deemed a success, such
VI
-------
as the use of on-board computers in small boats to record station position at the time of
sampling, and the use of bar-coding to track sample shipments to the numerous laboratories
responsible for processing.
INDICATOR DEVELOPMENT AND TESTING (SECTION 4)
A major accomplishment of the 1990 Demonstration Project was developing and applying a
methodology for calibrating and validating biological indicators of estuarine condition. The
methodology uses discriminant analysis to identify the most effective combination of
measurements for distinguishing assemblages at regional reference sites from those at sites
with known environmental perturbations. Such biological indicators, which are applicable over
a range of latitudes and habitats, had not been identified previously. Developing these
indicators is important for EMAP and should prove helpful in other Agency efforts for
developing biocriteria.
Benthic invertebrate assemblages were the most successful biological indicators. Bottom-
dwelling organisms are particularly useful as indicators because they integrate exposure
conditions over long periods of time (months to years), and because their relative immobility
prevents them from avoiding pollution exposure. Based on studies conducted during the 1990
Demonstration Project, five attributes of the assemblage related to the number of species,
number of amphipods, number of capitelid polychaetes, number of bivalves, and the average
weight per polychaete can correctly differentiate clean reference sites from polluted sites with
about 90% certainty. Initial steps have been taken to validate this index; however, additional
data from future years of the program will be used to further refine the index and continue the
validation process.
Fish assemblage indicators also are promising for distinguishing between polluted and
unpolluted sites, but additional data will be required to reduce uncertainty. The mobility of fish
presents the biggest obstacle to using them for identifying conditions at a particular site. Even
if a fish indicator that successfully discriminates sites of high and low quality cannot be
validated until later years of the program, there are still several fish parameters, such as
prevalence of visible pathological disorders, that convey meaningful information on a regional
or watershed basis.
Measures of pollution exposure also were evaluated during the 1990 Demonstration Project.
For instance, variability in dissolved oxygen concentrations in the province was examined by
deploying continuously-recording meters for periods of up to 60 days at a subset of sites.
Although DO was found to be quite variable at individual sites, it was stable enough on a
regional basis for making provincewide estimates of condition. Improved methods for
measuring dissolved oxygen on a regional scale are being developed and tested in 1991
based on the findings of the 1990 Demonstration Project.
VII
-------
DESIGN EVALUATION (SECTION 5)
One of the primary objectives of the 1990 Demonstration Project was to determine the
precision with which status can be estimated using the present design. At the province level,
confidence intervals surrounding status estimates for each indicator measured during the 1990
Demonstration Project were less than 10% of the areal extent of the province. The
uncertainty associated with these estimates is anticipated to decrease substantially after
completing a four-year sampling cycle. Estimates developed at the subprovince level, such as
for individual estuarine systems like the Chesapeake Bay, were less precise because fewer
samples were collected in the subpopulations, but estimates were still within 15% of the actual
areal extent. Confidence intervals for these subpopulations will also contract when the four-
year sampling cycle is completed.
Another objective of the Demonstration Project was to estimate sources of variability for a
selected indicator (the benthic index) and incorporate this information into a power analysis
model to assess its power for trend detection using the present sampling design and sample
density. Estimates of variability were generated from two sources: the EMAP sampling
program and an ongoing monitoring program in the lower Chesapeake Bay. The power for
trend detection was found to be sensitive to the estimate for interannual variability. Using the
estimate generated from EMAP data, the likelihood of detecting a 2% per year change in the
area of degraded benthic invertebrate assemblages over 12 years was greater than 99%.
Using the lower Chesapeake Bay data set to estimate variance, the power for detecting a 2%
per year change was only about 60%. Neither data set was ideal for estimating interannual
variability; the EMAP data set was limited temporally, and the Chesapeake Bay data set was
limited spatially. As EMAP continues to collect data over the next several years, the power
analysis will be refined to better define the power of the program for detecting trends over
decades.
The stability of several indicators was evaluated across three sampling intervals (mid-June to
late-July; late-July to the end of August; early to late September) to define the boundaries of
the sampling window. The most appropriate sampling period was determined to be late July
through August because 1) dissolved oxygen values are at annual low values; 2) contaminant
exposure is greatest because of low dilution flows and peak metabolic activity associated with
highest water temperatures; and 3) living resources are most abundant, maximizing the
probability of collecting organisms required for assessments.
Sample allocation in the 1990 Demonstration Project was accomplished after stratifying
estuaries into three classes (large estuaries, large tidal rivers, and small estuarine systems).
Without this stratification, the large tidal river and small estuarine system classes, which are
perceived to be at risk from different types of stresses, would not have been sampled
sufficiently to make assessments with acceptable levels of uncertainty. Alternative
VIII
-------
stratification schemes based on more dynamic characteristics, such as salinity and sediment
type, were examined and found to have logistical shortcomings that rendered them less
appropriate than the present design.
Another design alternative examined was the use of index sites in large tidal rivers and in
small estuarine systems. Whereas random sites are designed to estimate areal extent, index
sites are located in the areas of a system most likely to exhibit a problem if one exists (deep,
depositional sites) and are used to estimate the percent of systems experiencing
environmental degradation. Index sites added little value to the program. The biggest
impediment to using index sites was that existing sediment maps of the Virginian Province are
inadequate to define depositional areas accurately.
PRELIMINARY EVALUATION OF ESTUARINE STATUS (SECTION 6)
When fully implemented, EMAP will provide regional and national assessments of ecological
status and trends for the nation's environmental resources based on a four-year sampling
cycle. The multi-year baseline is intended to minimize the effect of natural interannual
variability due to climate and other factors. The preliminary evaluation of the condition of
estuaries in the mid-Atlantic region provided in this report represents a first attempt at
presenting statistically unbiased, regional-scale information to a broad audience and is
intended to elicit discussion about assessment needs. The evaluation is preliminary because
some indicators have not been validated fully, a process that will require several years of
regional-scale data. The estimates presented are based on a single year of data rather than
the four-year running average that is the basic unit of EMAP assessment; nonetheless,
province-level evaluations of ecological condition that are unavailable from other sources are
possible with the data:
• The biotic integrity of estuaries in the Virginian Province was evaluated by
measuring the condition of benthic invertebrate (bottom-dwelling animals)
assemblages. Between 16%-30% of the estuarine area in the Province had
benthic resources that were degraded compared to regional reference sites.
• Biotic integrity was also assessed by examining the prevalence of visible
pathological disorders (lesions, tumors, etc.) of fish. Four of every thousand
fish in the province had a visible pathological disorder. The prevalence in
demersal fish (those living in close contact with the bottom sediments) was
several times that in pelagic fish (those living primarily in the water column).
Less than 0.1% of fish that are commercially or recreationally harvested had
visible pathological disorders.
IX
-------
Ten-day solid-phase toxicity tests using indigenous biota were conducted to
examine the condition of estuarine sediments. Eight percent of the sediments
in the province were estimated to be acutely toxic.
Sediments were screened for contaminants using the same list of analytes used
in the NOAA National Status and Trends Program. Based on this list of
analytes, 39% of the province was estimated to have concentrations of
contaminants in the sediments that potentially cause at least sublethal
biological effects. Metals, lead and zinc in particular, were the most prevalent
contaminants at these concentrations.
High concentrations of Clostridium perfringens, a bacterial tracer of sewage
pollution, were found in an estimated 9% of the province.
Small estuarine systems, including harbors, bays and coastal embayments, had
the highest proportion of toxic sediments, sediments containing contaminant
concentrations of biological concern, and sediments containing high levels of
Clostridium. They also had the highest proportion of fish with pathological
disorders. These small systems typically are overlooked in monitoring
programs that concentrate effort along the main axis of large estuarine systems.
Between 14%-28% of the area in the province had dissolved oxygen
concentrations below 5 ppm, the water quality standard for many states in the
province. Nine percent of the area was estimated to have concentrations below
2 ppm, which is considered stressful to most biota.
Of the largest systems in the Virginian Province, Long Island Sound had the
highest proportion of area with oxygen concentrations less than 5 ppm;
Chesapeake Bay had the highest proportion of area with concentrations below
2 ppm.
Anthropogenic marine debris (trash) was estimated to be present in 9-19% of
the estuarine area of the province. Paper and plastic accounted for most of this
debris.
-------
CONTENTS
Page
NOTICE ii
PREFACE Hi
EXECUTIVE SUMMARY . . . . v
CONTENTS xi
FIGURES xv
TABLES xix
ABBREVIATIONS xxii
1. INTRODUCTION 1-1
1.1 OVERVIEW OF EMAP 1-1
1.2 INTRODUCTION TO 1990 DEMONSTRATION PROJECT 1-2
1.3 OBJECTIVES OF THE 1990 DEMONSTRATION PROJECT 1-4
1.4 PURPOSE AND ORGANIZATION OF THIS REPORT 1-5
2. DEMONSTRATION PROJECT APPROACH AND METHODS 2-1
2.1 SAMPLING DESIGN , 2-3
2.1.1 Base Sampling Sites 2-4
2.1.2 Long-term Dissolved Oxygen (LTDO) Sites 2-5
2.1.3 Supplemental Sampling Sites 2-5
2.1.4 Indicator Testing and Evaluation Sites 2-6
2.1.5 Index Sites 2-6
2.2 SAMPLING METHODS 2-6
2.2.1 Bottom Dwelling (Benthic) Animals 2-8
2.2.2 Sediment Profile Images 2-14
2.2.3 Fish 2-15
2.2.4 Large Bivalve Abundance and Tissue Contamination 2-18
Xli
-------
Table of Contents (Cont.)
2.2.5 Dissolved Oxygen Concentration 2-19
2.2.6 Chlorophyll a 2-20
2.2.7 Sediment Toxicity 2-20
2.2.8 Sediment Contaminant Concentrations . 2-21
2.2.9 Sediment Clostridium Spores 2-21
2.2.10 Sediment Grain Size 2-23
2.2.11 Water Column Toxicity 2-24
2.2.12 Marine Debris . : 2-24
2.3 DATA COLLECTION AND SAMPLE TRACKING 2-25
2.4 ANALYTICAL METHODS FOR STATUS ASSESSMENT 2-25
2.4.1 Assessment Methods for Individual Indicators 2-26
2.4.1.1 Benthic Index 2-26
2.4.1.2 Fish 2-28
2.4.1.3 Sediment Toxicity 2-28
2.4.1.4 Sediment Contaminant Concentrations 2-29
2.4.1.5 Sediment Clostridium Spores 2-29
2.4.1.6 Marine Debris 2-30
2.4.1.7 Water Clarity 2-30
2.4.1.8 Water Column Stratification 2-30
2.4.1.9 Integration of Indicators 2-30
2.4.2 Assessment Methods for Areal Estimation 2-31
2.4.3 Procedures for Estimating Precision 2-33
3. EVALUATION OF LOGISTICAL FEASIBILITY AND QUALITY ASSURANCE 3-1
3.1 WATER QUALITY PARAMETERS 3-3
3.2 SEDIMENT QUALITY AND BENTHIC COMMUNITY PARAMETERS 3-5 ,
3.3 FISH AND BIVALVE SAMPLING 3-12
4. INDICATOR DEVELOPMENT AND TESTING 4-1
4.1 RESPONSE INDICATORS 4-2
4.1.1 General Methodology for Developing Response Indicators 4-4
xii
-------
Table of Contents (Cant.)
4.1.2 Benthic Macroinvertebrates 4-10
4.1.2.1 Development of a Benthic Index of Estuarine Integrity 4-10
4.1.2.2 Discussion of the Benthic Index 4-22
4.1.2.3 Sediment Profile Imaging Techniques 4-23
4.1.3 Fish 4-27
4.1.3.1 Development of a Fish Index of Estuarine Quality 4-26
4.1.3.2 Discussion of the Fish Index 4-31
4.1.3.3 Visible Pathology 4-33
4.1.3.4 Fish Tissue Contaminants . . . 4-34
4.2 EXPOSURE INDICATORS i 4-36
4.2.1 Dissolved Oxygen 4-36
4.2.1.1 Use of dissolved oxygen data for estimating status and trends .... 4-38
4.2.1.2 Use of the dissolved oxygen indicators for associations 4-43
4.2.2 Sediment Toxicity 4-51
4.2.3 Sediment Contaminants 4-54
4.2.3.1 Identifying Anthropogenic Enrichment 4-54
4.2.3.2 Identifying "Above Background" Concentrations 4-58
4.2.3.3 Determining Biologically Significant Concentrations 4-59
4.2.3.4 Comparison of Analytical Approaches 4-60
4.3 HABITAT INDICATORS 4-68
4.3.1 Depth 4-68
4.3.2 Salinity 4-68
5. EVALUATION OF THE SAMPLING DESIGN 5-1
5.1 DETERMINATION OF THE APPROPRIATE TIME FRAME FOR SAMPLING . 5-1
5.2 STRATIFIED SAMPLING 5-3
5.3 PRECISION OF STATUS ESTIMATES 5-12
5.3.1 Improving the Precision of Estimates 5-17
5.3.2 Using Replicates to Improve Precision 5-18
5.3.3 Improving Estimation of Precision 5-21
xm
-------
Table of Contents (Cent.)
5.4 EVALUATION OF POWER FOR TREND DETECTION . 5-22
5.4.1 Formulation of the Model 5-24
5.4.2 The Trend Detection Procedure 5-25
5.4.3 Component Parameter Estimates 5-28
5.4.4 Results from the Trend Detection Model 5-30
5.5 INDEX SITES 5-32
5.6 DETERMINATION OF THE APPROPRIATE SPATIAL SAMPLING SCALE . 5-34
6. STATUS OF VIRGINIAN PROVINCE ESTUARIES 6-1
6.1 INTRODUCTION 6-1
6.2 BIOTIC INTEGRITY 6-3
6.3 POLLUTANT EXPOSURE 6-7
6.4 SPECIFIC ESTUARINE SYSTEMS 6-12
6.5 SPECIALIZED HABITATS 6-19
6.6 ASSOCIATIONS 6-21
6.7 SOCIETAL VALUES 6-22
6.8 INTEGRATION OF ESTUARINE CONDITIONS 6-25
6.9 PROGRAM DIRECTION 6-26
7. SUMMARY AND CONCLUSIONS 7-1
8. REFERENCES 8-1
xiv
-------
FIGURES
Figure 2-1 EMAP biogeographical provinces for estuaries 2-2
Figure 2-2 Schematic summarizing the indicator testing and evaluation strategy for the
1990 Demonstration Project in the Virginian Province 2-7
Figure 2-3 Areas sampled by each team during the 1990 Demonstration Project
in the Virginian Province 2-9
Figure 2-4 Overview of the EMAP indicator strategy giving examples of the types
of indicators used to assess estuarine status 2-10
Figure 4-1 General methodology used to evaluate potential response indicators ..... 4-5
Figure 4-2 Plot showing the regression line used to estimate salinity-
adjusted species richness measures for mean number of species
per site 4-14
Figure 4-3 Relationship of number of species to salinity after adjusting
for habitat variables . 4-15
Figure 4-4 Time series data obtained at two long-term dissolved oxygen
monitoring sites showing fluctuation of dissolved oxygen concen-
trations over the summer index period 4-38
Figure 4-5 Comparison of CDFs for province-wide botom dissolved oxygen
centrations in sampling intervals 2 and 3 4-41
Figure 4-6 Example showing the degree of autocorrelation on samples from
dissolved oxygen concentration . 4-42
Figure 4-7 Differences in dissolved oxygen due to diurnal and tidal periodi-
cities at stations of different depths 4-43
Figure 4-8 CDFs developed by restricting sampling of time series data to
daylight hours and by unrestricted random sampling 4-45
xv
-------
Figure 4-9
Figure 4-10
Figure 4-11
Figure 4-12
Figure 4-13.
Figure 4-14.
Figure 4-15
Figure 4-16
Figure 5-1
Figure 5-2
Figure 5-3
Figure 5-4
Figure 5-5
Relative performance of the six sampling strategies in estimating
the mean dissolved oxygen concentration within 1 ppm from time
series 4-47
Dissolved oxygen quantities from minimum to 25%, for 13 LTDO
data series
4-48
Misclassification rate associated with DO threshold values for
one sampling strategy 4-49
Overall misclassification rates for the point-in-time sampling
(a) and short-term continuous sampling (b) derived from plots
like the one shown in Figure 5-13 4-50
Relationship between chromium and aluminum from 1990 EMAP data . . . 4-57
Percent of area in each of three resource classes with ele-
vated contaminant concentrations, as defined by four differ-
ent threshold approaches 4-68
Comparison of cumulative distribution function for depth from
sampling intervals 2 and 3 4-70
Comparison of cumulative distribution function for salinity from
sampling intervals 2 and 3 4-71
Comparison of cumulative distribution function for dissolved oxygen
between sampling intervals 2 and 3 5-6
Hypothetical distribution functions used in analysis of stratum
misclassification effects 5-10
Comparison of optimal vs proportional allocation 5-11
Confidence intervals for estimates of benthic index and
sediment toxicity over the entire Virginian Province 5-13
Confidence intervals for estimates of the benthic index and
sediment toxicity by resource classes 5-14
xvi
-------
Figure 5-6 Confidence intervals for estimates of the benthic index and
sediment toxicity by resource system 5-15
Figure 5-7 Comparisons of CDFs for the benthic index between index and
random sites in small estuaries and tidal rivers 5-33
Figure 5-8 Comparison of CDFs for sediment type, benthic index, depth,
and sediment toxicity, between the base and intensified
sampling efforts in the Delaware Estuary 5-35
Figure 6-1. Location of the EMAP 1990 Demonstration Project sample
sites in the estuaries of the Virginian Province 6-2
Figure 6-2. Percent of estuarine area in the three classes of
estuarine resources having degraded benthic communities 6-5
Figure 6-3. Prevalence of fish with pathological disorders in the
Virginian Province 6-7
Figure 6-4. Prevalance of pathological disorders in demersal fish in
the three classes of estuarine resources 6-8
Figure 6-5. Percent of estuarine area in the Virginian Province with
bottom dissolved oxygen concentrations below 2 ppm and 5 ppm 6-9
Figure 6-6. Percent of area in the Virginian Province with sediment
contaminant concentrations greater than Long and Morgan
ER-L values 6-11
Figure 6-7. Pollutant exposure conditions for the three classes of
estuarine resources 6-13
Figure 6-8. Sediment types in the three major estuarine systems
in the Virginian Province 6-14
Figure 6-9. Salinity habitats in the three major estuarine systems in
the Virginian Province: brackish (0 to 5 ppt), transitional
(5 to 18 ppt), and marine (> 18 ppt) 6-15
Figure 6-10. Bottom dissolved oxygen conditions in the three major
estuarine systems in the Virginian Province 6-16
xvn
-------
Figure 6-11. Prevalence of four classes of contaminants in the three
largest estuarine systems in the Virginian Province 6-17
Figure 6-12. Degraded benthic resources in the three largest estuarine
systems in the Virginian Province 6-18
Figure 6-13. Salinity habitats in the Virginian Province as a percent of
total estuarine area 6-20
Figure 6-14. Association between degraded benthic assemblages and the
dissolved oxygen and sediment toxicity pollutant exposure
indicators . 6-22
Figure 6-15. Percent of area having anthropogenic marine debris (trash)
in three classes of estuarine resources 6-24
Figure 6-16. Percent of area in the three classes of estuarine resources
where high concentrations of Clostridum perfringens were
found in the sediment 6-25
Figure 6-17. Conceptual scheme for combining indicators that might be
measured by EMAP into an integrated statement about the
environmental condition of estuaries 6-27
Figure 6-18. Summary of environmental conditions in the estuaries of the
Virginian Province 6-28
xvm
-------
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
3-1
3-2
3-3
3-4
4-1
4-2
TABLES
Summary of the characteristics of estuarine classes in the Virginian Province .... 2-5
Sampling activities accomplished at each type of station in each of
the three sampling intervals during the 1990 Demonstration Project 2-11
Biomass groups for macrofaunal species 2-13
Calculation of the OSI based on sediment profile images 2-15
Targeted species and size ranges of fish retained for tissue analysis
during the 1990 Virginian Province Demonstration Project 2-16
Analytical measurements for fish and bivalve tissue samples collected
during the 1990 Virginian Province Demonstration Project 2-17
Analytical measurements for sediment samples collected during the
1990 Virginian Province Demonstration Project 2-22
Analytical methods used in 1990 for determination of chemical contaminant
concentrations in sediments 2-23
Standardized mean and standard deviation for indicators used in the
discriminant function for the benthic index 2-27
Status of sample collection during the 1990 Virginian Province
Demonstration Project 3-2
Results for SRM 2704 used as a set control for the 1990 EMAP-E
sediment inorganic analyses 3-7
Results for SRM 1941 used as a set control for the 1990 EMAP-E
sediment organic analyses 3-8
Range in detection limits reported for organic compounds in 1990
sediment samples
3-9
Test for significant differences between degraded and nondegraded
reference sites for each of the descriptors of the benthic assemblage 4-13
Summary of correlation between habitat indicators and the candidate
benthic measures
4-15
xix
-------
4-3
4-4
4-5
4-6
4-7
4-8
4-9
4-10
4-11
4-12
4-13
4-14
4-15
5-1
5-2
List of sites used in indicator testing dataset during the 1990
Virginina Province Demonstration Project ' 4-18
Number of indicator testing sites sampled in each of several
habitat types
4-19
Results of discriminant analyses conducted to combine candidate
benthic measures into an index 4-20
Results of validation of four candidate indices 4-22
Correlation coefficients relating the values of each of four indices
and percent of sites that were classified the same among four
indices applies to 152 sampling sites 4-23
Comparison of benthic characterizations produced by sediment profile
image analysis and conventional macroinvertebrate sampling 4-26
Tests for significant differences among degraded and reference sites
for each candidate fish measure 4-28
Results of discriminant analyses conducted for combining candidate
fish measures into an index 4-31
Correlations between the first and second trawl and between intervals 2
and 3 for each candidate fish measure 4-33
Relationship of contaminant concentration to Long and Morgan ER-M
values and toxicity by site, as a function of salinity, grain size,
and holding time
4-53
Relationship between sediment metal concentration and aluminum 4-58
Comparison of threshold contaminant levels based upon the "above
background", ER-L, and ER-M values 4-62
Percent of the area in the Virginian Province exceeding each of
four threshold sediment contaminant concentrations 4-65
T-test comparison of dissolved oxygen concentrations and the benthic
index among periods 5-3
Average dissolved oxygen concentration by week for stations with
relatively complete Hydrolab records 5-4
xx
-------
5-3
5-4
5-5
5-6
5-7
5-8
Weekly minimum dissolved oxygen concentrations for stations with
relatively complete Hydrolab records 5-5
Comparison of sediment type predicted from NOAA database with that
measured for EMAP sampling sites 5-8
Confidence intervals associated with estimates around critical values
of response and exposure indicators 5-16
Optimal percent of sample size devoted to annual sampling 5-23
Estimates of variance components used in the power
anlaysis based on EMAP data 5-29
Power for trend detection in the benthic index over three EMAP
sampling cycles
5-31
XX T
-------
ACE
ANOVA
APHA
ARS
Bl
BLM
CDF
CEQ
CPR
CTD
DO
EMAP
EMAP-E
EPA or USEPA
ER-L
ER-M
FDA
FS
FWS
GFAA
IBI
ABBREVIATIONS
U.S. Army Corps of Engineers
analysis of variance
American Public Health Association
Agricultural Research Service
benthic index
Bureau of Land Management
cumulative distribution function
Council on Environmental Quality
cardio-pulmonary resuscitation
conductivity, temperature, depth
dissolved oxygen
Environmental Monitoring and Assessment Program
EMAP Estuaries Resource Group
U.S. Environmental Protection Agency
effects range-low value
effects range-median value
Food and Drug Administration
U.S. Forest Service
U.S. Fish and Wildlife Service
graphite furnace atomic adsorption
Index of Biotic Integrity
xxn
-------
ICP-AES inductively coupled plasma-atomic emission spectrometry
ITE Indicator Testing and Evaluation
LTDO Long-term Dissolved Oxygen
loran long range navigation
MSE mean squared error
NOAA National Oceanic and Atmospheric Administration
NRC National Research Council
ORD Office of Research and Development
OSI Organism-Sediment Index
OTA Office of Technology Assessment
PAH polycyclic aromatic hydrocarbons
PAR photosynthetically active radiation
PCS polychlorinated biphenyls
ppm parts per million
ppt parts per thousand
QA/QC quality assurance/quality control
REMOTS® Remote Ecological Monitoring of the Seafloor
RPD redox potential discontinuity
SDD Secchi disc depth
SRM standard reference material
USGS U.S. Geological Survey
xxm
-------
-------
SECTION 1
INTRODUCTION
Pollution control programs in the United States have been estimated to cost more than $80
billion annually (CEQ 1990). Most of these programs address specific local pollution problems;
however, the means to assess the effectiveness of these programs for protecting the
environment at national and regional scales and over extended periods of time do not exist.
The U.S. Environmental Protection Agency (EPA) considers it critical to establish monitoring,
research, and assessment programs to determine the effectiveness of pollution control
strategies and to substantiate the science upon which these strategies are based (USEPA
1992).
1.1 OVERVIEW OF EMAP
The Environmental Monitoring and Assessment Program (EMAP) is a nationwide program
initiated by EPA's Office of Research and Development (ORD). EMAP was developed in
response to the demand for information about the degree to which existing pollution control
programs and policies protect the nation's ecological resources. EMAP is an integrated
federal program; ORD is coordinating, planning, and implementing EMAP in conjunction with
other federal agencies, including the Agricultural Research Service (ARS), the Bureau of Land
Management (BLM), the U.S. Fish and Wildlife Service (FWS), the U.S. Forest Service (FS),
the U.S. Geological Survey (USGS), and the National Oceanic and Atmospheric
Administration (NOAA), and with other offices within EPA (e.g., Office of Water). These other
agencies and offices participate in the collection and analysis of EMAP data and will use it to
guide their policy decisions, as appropriate.
The objectives of EMAP are to estimate the current status and trends in the condition of the
nation's ecological resources on a regional basis, with known confidence; to seek
associations between human-induced stresses and ecological condition; and to provide
periodic statistical summaries and interpretive reports on ecological status and trends to
resource managers and the public. It is designed to provide the information required to
formulate the environmental protection policies of the 1990s and beyond by providing answers
to the following questions:
• What is the status, extent, and geographical distribution of the nation's
ecological resources?
• What proportion of these resources is declining or improving? Where? At what
rate?
1-1
-------
• What factors are likely to be contributing to declining condition?
• Are pollution control, reduction, mitigation, and prevention programs achieving
overall improvement in ecological condition?
Assessment of status and trends in the condition of the nation's ecological resources requires
collecting data in a standardized manner, over large geographic scales, and for long periods
of time. Such assessments cannot be accomplished by aggregating data from the many
individual, short-term monitoring programs that have been conducted in the past or are being
conducted currently (NRG 1990). Differences in the parameters measured, collection
methods, timing of sample collection, and program objectives severely limit the value of
historical monitoring data and existing monitoring programs for making integrated regional and
national assessments.
EMAP was proposed and developed by EPA/ORD because an integrated monitoring and
assessment program that samples ecological resources in a probability-based manner offers
considerable advantages over historical monitoring approaches. One advantage is improved
definition of the extent and magnitude of pollution problems at regional and national scales.
Simultaneous, probability-based sampling of pollution exposure, environmental condition, and
biological resources, however, is most important. This approach enables estimates to be
made of the uncertainty associated with assessments and will improve our ability to identify
ecological responses to pollution. EMAP, therefore, will provide objective assessments of the
severity and extent of environmental problems and the degree to which degraded resources
are responding to efforts to protect or restore them.
1.2 INTRODUCTION TO 1990 DEMONSTRATION PROJECT
EMAP is being initiated in phases, beginning in 1990 with demonstration and pilot projects in
the estuaries of the Virginian Province (i.e., the mid-Atlantic region) and the forests of New
England. These projects afford EMAP the opportunity to refine the selection of indicators and
evaluate and modify the sampling design. They also provide the opportunity to resolve
difficulties associated with regional-scale sampling. Demonstration projects combine the
elements of several pilot projects (such as testing selected indicators and specific design
aspects) with a demonstration of the EMAP approach to monitoring and assessment of
ecological condition.
Estuaries were selected as one of the first resources to be sampled by EMAP. Estuaries are
among the most productive of ecological systems. Historically, more than 70% of commercial
and recreational landings of fish and shellfish were taken from estuaries (NOAA 1987).
Estuaries also provide feeding, spawning, and nursery habitats, and migratory routes for many
commercially and recreationally important fish and wildlife, including threatened and
1-2
-------
endangered species (Lippson et al. 1979; Olsen et al. 1980). The public values estuarine
ecosystems for recreation (e.g., boating, swimming, hunting, and fishing) and aesthetic
appeal. Approximately $7 billion in public funds are spent annually on outdoor marine and
estuarine recreation in the 33 coastal states (NOAA 1988). Millions of tourists visit coastal
beaches annually, and coastal property is among the nation's most valuable real estate.
About half of the nation's population now lives in coastal areas, and by the year 2010, the
population in these areas will have grown by almost 60% to 127 million people (Culliton et al.
1990). The estuarine and coastal environment also provides cooling waters for energy
production, transportation routes for ships, and space for economic development. Most of the
nation's major ports are located in estuaries.
Estuaries are complex transition zones between streams, rivers, and coastal oceans. They
have physical features that concentrate and retain pollutants, and they tend to serve as
repositories for the many pollutants released into the nation's waters and into the atmosphere.
The ecological condition of estuaries is influenced strongly by human activities in the
watershed, particularly land use patterns and the release of pollutants to the environment. In
many coastal regions, water and sediment quality and the abundance of living resources are
perceived to have declined over the past 10 to 15 years, despite the implementation of more
strict pollution control programs. Increasingly, reports appear in the popular press and
scientific journals (Morganthau 1988; Toufexis 1988; Smart et al. 1987) describing the decline
of estuarine and coastal environmental quality, as exemplified by:
• Increases in the frequency, duration, and extent of water containing insufficient
oxygen to sustain living resources (USEPA 1984; Officer et al. 1984; Parker et
al. 1986; Rabalais et al. 1985; Whitledge 1985);
• Accumulation of contaminants in bottom sediments and in the tissues of fish
and shellfish to levels that threaten human health and the sustainability of fish
and shellfish populations (OTA 1987; NRC 1989);
• Increased evidence that many restoration and mitigation efforts have not
replaced losses of critical habitats (Sanders 1989; The Conservation
Foundation 1988); .
• Increased incidence of pathological problems in fish and shellfish (Sinderman
1979; O'Connor et al. 1987; Buhler and Williams 1988; Capuzzo et al. 1988);
• Increased frequency and persistence of algal blooms and associated decreases
in water clarity (USEPA 1984; Pearl 1988; Smayda and Villareal 1989);
• Increased incidence of closures of beaches, shellfishing grounds, and fisheries
because of pathogenic and chemical contamination (Smart et al. 1987; FDA
1-3
-------
1971, 1985; Hargis and Haven 1988; Broutman and Leonard 1988; Leonard et
al. 1989); and
• Increased incidence of human health problems from consuming contaminated
fish and shellfish or swimming in contaminated waters (Fein et al. 1984; Malins
1989).
The Virginian Province was selected as the testing ground for the EMAP monitoring effort for
estuaries because there is a general public perception that estuaries in this region of the
country are deteriorating more rapidly than in other regions. Many of the estuaries in this
province have been investigated intensively by scientists, and a considerable amount of
information was available for use in designing the 1990 Demonstration Project. Six National
Estuary Programs are in place in the Virginian Province. In addition, many management
decisions are forthcoming, including development of a restoration plan for the New York
Harbor complex, and development of management plans and evaluation of previous
management actions for many large estuaries, including Delaware Bay, Chesapeake Bay, and
Long Island Sound. Development of such plans presents an opportunity to demonstrate how
EMAP monitoring data can assist in the formulation of environmental programs and policies.
1.3 OBJECTIVES OF THE 1990 DEMONSTRATION PROJECT
The specifics of the 1990 Virginian Province Demonstration Project are documented in
Holland (1990). A critical issue addressed during the 1990 Demonstration Project was how
best to represent the ecological condition of estuarine resources on a regional scale with the
limited financial resources available. All the environmental parameters of concern to the
public, scientists, and regulators cannot be measured. A limited set of parameters that serve
as indicators of overall estuarine condition needed to be identified, calibrated and verified. It
is obvious that one or two samples from a few locations, collected at one time of day, in a
single season of a particular year cannot characterize the ecological condition of an estuary.
The 1990 Virginian Province Demonstration Project was used to identify which indicators and
design attributes are most effective for meeting program objectives and forming the basis for
developing programs in other provinces.
The objectives of the 1990 Virginian Province Demonstration Project were to:
• Demonstrate the value of regional monitoring using an unbiased sampling
approach as a basis for assessing the condition of estuarine resources;
• Evaluate the ability of a suite of indicators of environmental quality to
discriminate between polluted and unpolluted sites over the regional scale;
1-4
-------
• Establish standardized methods for monitoring indicators of ecological status
and trends in estuaries;
• Obtain data on regional-scale variability in ecological parameters to evaluate
and refine the sampling design;
• Develop analytical procedures for using regional-scale monitoring data to
assess the ecological status of estuaries, and apply the procedures to establish
baseline conditions in the Virginian Province; and
• Identify and resolve logistical problems associated with conducting a regional-
scale monitoring program in estuaries.
Clearly, all of these objectives cannot be accomplished in a single year, and most
require a process of continuing adjustment and improvement using data collected in
subsequent years. This report will describe the current status of analyses for the 1990
Virginian Province Demonstration Project and provide evaluations of the sampling design and
the parameters measured during the Demonstration Project.
1.4 PURPOSE AND ORGANIZATION OF THIS REPORT
The purpose of this report is to present results for each of the six objectives of the 1990
Demonstration Project and to evaluate the project itself. Because of its dual purpose as an
evaluation of a process and a preliminary product of that process, this report is expected to
have many audiences with varying interests and levels of expertise in estuarine science and
monitoring; consequently, the degree of detail and description of analyses presented varies
among sections.
This report is organized in sections addressing the objectives of the 1990 Demonstration
Project, Sections 2 through 5 are evaluations of specific elements of the 1990 Demonstration
Project including methods, logistics, indicator development, and sampling design. A
description of the 1990 Demonstration Project is presented in Section 2 along with a summary
of the standardized sample collection and processing methods and data management
procedures developed for monitoring indicators of ecological status and trends in estuaries
(Holland 1990; Rosen et al. 1990; Strobel 1990; USEPA 1991). This section also contains a
description of the analytical methods used to complete the evaluation of estuarine status.
Logistical problems associated with conducting a regional-scale estuarine monitoring program
are identified in Section 3. Specific problems relating to quality control are also included in
this section (Valente et al. 1990).
1-5
-------
Section 4 presents an evaluation of the indicators developed for the Demonstration Project
(Holland 1990; Hunsaker and Carpenter 1990; Knapp et al. 1990) and describes a general
approach for developing biological indices that can be used to discriminate between polluted
and unpolluted sites over regional scales. Data from the Demonstration Project are used in
Section 5 to define the regional-scale variability in ecological parameters and evaluate the
EMAP sampling design used for conducting a regional-scale monitoring program in estuaries.
Additional data from other environmental monitoring programs are used with EMAP data to
make preliminary estimates of the power to detect trends using this sampling design.
Sections 2 through 5 use results from the Demonstration Project to build upon material
previously presented by the EMAP-Estuaries Resource Group and are intended primarily for
technical audiences within EMAP and for reviewers of the program. Section 6 is a preliminary
evaluation of the condition of estuarine waters in the Virginian Province in 1990. It is intended
primarily for non-technical users of EMAP data, including Congress, the EPA Administrator,
the EPA Regions and program offices, and state and local resource managers. The section
presents information, distilled from the 1990 Demonstration Project data, that may be of use to
regulators and resource managers for determining the scope and utility of EMAP monitoring.
It also provides a summary of the information that specialists in estuarine science and
monitoring believe to be most pertinent for assessing estuarine condition. The style of Section
6 is purposely different from the other sections to highlight that it is written for a different
audience; however, cross-reference is made to later sections to provide technical justifications
and explanations as appropriate.
The juxtaposition of the preliminary assessment with the evaluation of the 1990 Demonstration
Project affords scientific reviewers and managers alike the opportunity to determine the utility
and value of regional monitoring data collected with a probability-based sampling approach,
thus addressing one of the primary goals of the 1990 Virginian Province Demonstration
Project.
Section 7 summarizes the conclusions that can be drawn from the 1990 Demonstration
Project as they relate to its stated objectives. This section is not meant to present major
results as does the Executive Summary. Rather, it describes how the results may be of use
to potential clients of EMAP.
1-6
-------
SECTION 2
DEMONSTRATION PROJECT APPROACH AND METHODS
Three elements of the EMAP approach to monitoring are distinctive and were the guiding
forces in designing the sampling plan for the 1990 Demonstration Project. First, probability-
based sampling sites are selected in an unbiased manner so that resources are sampled in
proportion to their abundance in a size class. Probability-based sampling permits estimation
of the condition of the portion of the resource that was not sampled based on knowledge of
the sampled portion. Estimates of the proportion of the total area sampled that is in degraded
condition can be made with quantifiable confidence, and the level of confidence in the
estimate can be increased by increasing the number of sites sampled, if interest and
resources allow.
Second, EMAP monitoring focuses on indicators of biological response to stress and uses
measures of exposure to stress or contamination as a means for interpreting that response.
Traditionally, estuarine monitoring has focused on measures of exposure (e.g., concentration
of contaminants in sediment) and attempted to infer biological impacts based upon laboratory
bioassays. The advantage of the ecologically-based approach emphasized in EMAP is that it
can be applied to situations where multiple stressors exist, acting separately or in
combination, and where natural processes cannot be modeled easily. This is certainly the
case in estuarine systems, which are subject to an array of anthropogenic inputs and exhibit
great biotic diversity and complex physical, chemical, and biological interactions.
Third, EMAP monitoring is conducted on regional and national scales. Standardized methods
are employed, and an entire region is sampled simultaneously within a defined time period to
ensure comparability of data within and among sampling years. EMAP identified boundaries
for eight estuarine regions (Fig. 2-1) based upon biogeographic provinces defined previously
by NOAA and the U.S. Fish and Wildlife Service using major climatic zones and prevailing
ocean currents (Bailey 1983; Terrell 1979). The 1990 Demonstration Project in the Virginian
Province, which includes the wide expanse of irregular coastline from Cape Cod,
Massachusetts, to the mouth of the Chesapeake Bay (Cape Henry, Virginia), was designed to
evaluate the feasibility of regional sampling and to evaluate and improve the sampling design
and indicators for future monitoring. This section summarizes the sampling plan and specific
methods used to collect data for a preliminary assessment of the ecological condition of the
estuaries of the Virginian Province in 1990. More detailed methods are described by Holland
(1990).
2-1
-------
0)
.o
o
I
Q.
75
.g
I
O)
o
s,
o
S
Q_
LU
CM
g>
LL
2-2
-------
2.1 SAMPLING DESIGN
An objective of the 1 990 Demonstration Project in the Virginian Province was to collect data at
the spatial scale and using the collection methods expected to be used when EMAP
monitoring in the Virginian Province is fully implemented. This sampling approach permits
evaluation of the logistical feasibility of the sampling plan and generation of preliminary status
estimates of the type that will be described in Section 6. The 1990 Demonstration Project
also included several pilot elements designed to evaluate the indicators and sampling design.
Accomplishment of the pilot study objectives required augmenting the probability-based
sampling sites used for developing regional status estimates with judgement samples. This
resulted in sampling five types of sites during the Demonstration Project:
• Base sites were the probability-based sampling sites forming the core of the
EMAP monitoring design for estuaries. Data collected from these sites are the
basis of the preliminary status evaluation for the Virginian Province presented in
Section 6.
• Long-term Dissolved Oxygen (LTDO) sites were a subset of the base sample
sites. Oxygen meters were deployed to obtain a continuous record of dissolved
oxygen concentrations at these sites. These continuous records were used to
evaluate the variability of oxygen concentrations and the feasibility of using single,
point-in-time measurements of dissolved oxygen to determine status (see Section
Supplemental sampling sites were part of a pilot study to define an adequate
spatial scale for full implementation of EMAP monitoring in estuaries and to define
spatial variability in small estuaries. These sites were selected in a probability-
based manner and were used in the preliminary status evaluation.
Indicator testing and evaluation (ITE) sites were part of a pilot study to determine
the reliability, sensitivity, specificity, and repeatability of indicator responses for
discriminating between polluted and unpolluted conditions. These sites were
selected based on historical data and expert judgement to represent the extremes
of environmental exposure (degraded to pristine).
Index sites were part of a pilot study to evaluate whether individual
"representative" sites can be used to portray the status of individual small
estuaries or tidal river segments in the way that the oxygen content of
hypolimnetic water can be used to describe the trophic state of small lakes.
Whereas base sites were designed to quantify the areal extent of pollution effects
(i.e., percent of degraded estuarine area), index sites were designed to estimate
the percent of affected small estuarine systems or tidal river segments.
2-3
-------
Potentially, information from base and index sites can be used together to
distinguish whether the estimated areal extent of identified problems results from
extensive problems in a small number of systems, or from problems affecting a
small amount of area in a large number of systems.
Figure 2-2 presents a map of all the probability-based (base and supplemental) sites sampled
in the 1990 Demonstration Project. All sites were restricted to a depth greater than 1.5 m to
allow for boat access. Therefore, the sampling results may not be applicable to habitats
characterized by a depth of less than 1.5m. Methods used to select sampling locations for
each type of sampling site are provided below.
2.1.1 Base Sampling Sites
The sampling design for base sites was stratified by size into large estuaries, large tidal rivers,
and small estuarine systems. Stratification permitted customizing the sampling frame to the
specific geographic features of these different classes of estuaries (Table 2-1). It also allowed
allocation of a strata-specific number of samples so that class estimates could be derived with
a desired level of precision. The boundaries of these classes were defined using NOAA
maps, resulting in 12 large estuaries, 5 large tidal rivers (i.e., Hudson, Potomac, James,
Delaware, Rappahannock), and 137 small estuarine systems (Holland 1990). Methods for
selecting sampling sites within each stratum are described below.
Large Estuaries — Large estuary sampling sites were selected using an enhancement of the
systematic sampling grid proposed for use throughout EMAP (Overton 1989). This grid was
placed randomly over a map of the United States and intensified to make 280 km2 hexagonal
grids. Fifty-four base sampling sites were selected using this grid. The sampling sites were
the center points of the hexagons, which were 18 km apart.
Large Tidal Rivers - Base sampling sites in large tidal rivers were selected using a "spine"
and "rib" approach that is a linear analog of the sampling grid for large estuaries. The
starting point of the spine was at the mouth of the river, and the first transect ("rib") was
located at a randomly selected river-kilometer between 0 and 25. Additional upstream
transects were placed every 25 km from the first. Sampling sites were selected at random
along the rib of each transect. A total of 25 base sampling sites were identified in the large
tidal rivers of the Virginian Province.
Small Estuarine Systems - A list frame was used to select 32 (23%) of the 137 small
estuarine systems in the Virginian Province for sampling during the 1990 Demonstration
Project. To ensure that the selected systems were dispersed geographically, all small
estuarine systems in the province were listed in order of latitude from north to south and
combined into groups of four. One system was selected at random from each group.
2-4
-------
Table 2-1 . Summary of the characteristics of estuarine classes in the Virginian Province
Characteristics
Surface Area
Shape
Salinity
Watersheds
Management Regions
Contaminant Sources
Total area in Virginian
Province
Percent of total area in
Virginian Province
Large Estuaries
> 260 km2
Aspect ratio < 20
Strong salinity
gradients
Large, complex
Multi-state
Multiple
16,889km2
72
Large Tidal Rivers
> 260 km2
Aspect ratio > 20
Partial salinity
gradients
Large, complex
Multi-state
Multiple
1,810km2
8
Small
Estuaries
2.6 - 260 km2
Any
Generally does
not have salinity
gradients
Small
Usually a single
state
Relatively few
4,879 km2
21
2.1.2 Long-term Dissolved Oxygen (LTDO) Sites
Thirty of the 116 base sampling sites located throughout the province were selected as LTDO
sites. Sites were selected for large estuarine systems (13 sites), large tidal rivers (5 sites),
and small estuarine systems (5 sites) and represented a range of estuarine habitats as
defined by salinity, sediment type, and depth.
2.1.3 Supplemental Sampling Sites
Available data were insufficient to ascertain the spatial sampling scale necessary to represent
the ecological status of estuarine systems in the Virginian Province with adequate precision.
To address this problem in large estuaries, the Delaware Bay was sampled at a density four
times greater (i.e., sample points were located approximately 9 km apart, yielding 20
additional sampling sites) than other large estuaries. To address the problem in tidal rivers,
sampling intensity was doubled in the Delaware River. Supplemental samples were placed
2-5
-------
exactly half-way (12.5 km) between base sampling sites there, resulting in four additional
samples.
Two types of supplemental sampling sites were selected in small estuaries to determine the
appropriate scale for representing resource conditions in this class: 1) an additional
randomly-located site was selected in each of five separate systems, and 2) four randomly-
located supplemental sites were selected in a single small estuarine system (i.e., Indian
River).
2.1.4 Indicator Testing and Evaluation Sites
Based on a review of existing data and expert opinion, 23 indicator testing and evaluation
sites were selected for specific combinations of geographic location, salinity, concentrations of
sediment contaminants, and dissolved oxygen concentrations. These sites were sampled to
investigate the reliability of indicator responses for discriminating between degraded and
nondegraded sites across the range of habitats that occurs throughout the Virginian Province
(Fig. 2-3).
2.1.5 Index Sites
Index sites were located in deep, muddy, depositional areas, the places believed to be most
likely to accumulate contaminants or to experience the most severe stress from low dissolved
oxygen concentrations within the sampling period. Index sites were located in every river
segment and every small estuary in which there were base sites, resulting in 25 index sites in
tidal rivers and 32 in small estuaries.
2.2 SAMPLING METHODS
All sampling was conducted from 8-m (24-ft), twin-engine, Romarine workboats. Specific
sampling sites were located using the onboard navigation equipment (Raytheon RAYNAV-780
loran-C) and, when applicable, dead reckoning. Loran-C was used to locate most sites. Dead
reckoning was used to locate sites where signal interference prevented reliable performance
of loran. Most sites in large tidal rivers and small estuarine systems had obvious landmarks,
channel buoys, and other fixed structures that could be used to obtain ranges and bearings.
2-6
-------
c:
o
1
co
o
CD
Q
O
O5
CD
O)
1
o
1
76
&
to
D>
'•£3
I
I
73
0)
O)
'N
Is
3 I
S c
cd co
E 'E
CM •
o "G
3 -2.
o) o-
LL a.
2-7
-------
Sampling was performed by three sampling teams working simultaneously. Team 1 sampled
stations from Cape Cod west to the Hudson River. Team 2 sampled from New York Harbor,
south to the Delaware Bay, and the northern-most part of the Chesapeake Bay. Team 3
sampled the remaining stations in coastal Delaware, Maryland, and Virginia (Fig. 2-4). Teams
circuited their assigned regions every ten days. Stations were organized into clusters, and
specific sites were selected at random from each cluster to be sampled on each circuit.
All sampling for the 1990 Demonstration Project was conducted between 19 June and
23 September. The sampling period was divided into three sampling intervals (19 June to 18
July, 19 July to 31 August, 1 September to 23 September). Sampling was repeated in three
Intervals as a pilot element to determine the stability of various indicators over the summer.
Only a subset of parameters was remeasured at all stations; therefore, the number and type
of sites sampled, and the types of sampling activities that took place at those sites varied
among intervals (Table 2-2). A description of the parameters measured at each site and the
specific methods used to conduct the sampling follows.
The EMAP indicator strategy involves four types of ecological indicators (Hunsaker and
Carpenter 1990): response, exposure, habitat, and stressor (Fig. 2-5). Response indicators
are ecological characteristics that integrate the responses of living resources to specific or
multiple pollutants and other stresses and are used by EMAP to assess overall estuarine
condition. Exposure indicators quantify pollutant exposure and habitat degradation and are
used mainly to identify associations between stresses on the environment and degradation in
response indicators. Habitat indicators provide basic information about the natural
environmental setting and are used to normalize exposure and response indicators to natural
environmental gradients. Stressor indicators are used to quantify pollution inputs or stresses
and identify the probable sources of pollution exposure. Examples of the relationships
between response, exposure, and habitat indicators sampled during the 1990 Demonstration
Project are given in Fig. 2-5. Descriptions were taken from the Near Coastal Program Plan
(Holland 1990), the Near Coastal Field Methods Manual (Strobel 1990), and the Near Coastal
Laboratory Procedures Manual (USEPA 1991). Readers also are referred to the overall
quality assurance plan for the 1990 Demonstration Project (Valente et al. 1990).
2.2.1 Bottom Dwelling (Benthic) Animals
Benthic invertebrate assemblages are composed of diverse taxa with a variety of reproductive
modes, feeding guilds, life history characteristics, and physiological tolerances to environ-
mental conditions (Warwick 1980; Frithsen 1989; Bilyard 1987). As a result,
benthic populations respond to changes in conditions, both natural and anthropogenic, in a
variety of ways (Pearson and Rosenberg 1978; Rhoads et al. 1978; Boesch and Rosenberg
2-8
-------
SAMPLING TEAM 1
SAMPLING TEAM 2
ANNAPOLIS, MD
SAMPLING TEAM 3
Figure 2-3. Areas sampled by each team during the 1990 Demonstration Project in the
Virginian Province. Base Stations are indicated by stars.
2-9
-------
SOR
ORS
NDICAT
S
EXPOSU
NDICATO
CO
S
H
-
n $
c 5^ f
•i S | .3
RESPONSE
NDICATORS
1
.
Q E
en i
Q. 'oEF
E £
S CD
cB £
2 =6
t/5 .E
$£
cd CD
ll
.E -Q
Q. W
< -9-
CD
"»
> "»
n-3
C«
«
«
2-10
-------
Table 2-2. Sampling activities accomplished at each type of station in each of the three
sampling intervals during the 1990 Demonstration Project
Station Type
Base Sampling Site
Supplement Sites
Continuous Dissolved
Oxygen Monitoring
Sites
Index Sites
Indicator Testing/
Evaluation Sites
Interval 1
CTD Profile*^
Fish Trawling^
Not sampled
CTD Profile
Fish Trawling
Benthic Assemblage
Not sampled
Not sampled
Interval 2
CTD Profile^3)
Fish Trawling* '
Benthic Suite*0)
CTD Profile*3)
Fish Trawling*b)
Benthic Suite*0)
CTD Profile*3)
Fish Trawling*b)
Benthic Suite*0)
CTD Profile
Benthic Assemblage
CTD Profile^
Fish Trawling*b)
Benthic Suited
Bivalve Dredging
Water Column Toxicity
Interval 3
CTD Profile*3)
Fish Trawling* )
Shellfish Dredging
Not sampled
CTD Profile
Fish Trawling
Benthic Assemblage
Shellfish Dredging
Not sampled
Not sampled
(a) CTD profile includes dissolved oxygen, temperature, salinity, pH, PAR, transmissometry, and
f luorometry.
' ' Fish trawling includes measurement of fish assemblage, tissue contaminants, and gross
pathology.
(°) Benthic suite sampling includes samples for determining benthic invertebrate assemblage,
sediment contaminants, and sediment toxicity.
1981). Responses of some species (e.g., filter feeders, species with pelagic life stages)
indicate changes in water quality and others (e.g., organisms that burrow in or feed on
sediments) changes in sediment quality; furthermore, because most benthic species have
limited mobility, they cannot avoid exposure to pollution stress. Benthic community studies
have been used in many regional estuarine monitoring programs (Bilyard 1987; Holland et al.
1987) and have proven to be an effective indicator for describing the extent and magnitude of
pollution impacts in estuarine ecosystems, as well as for assessing the effectiveness of
management actions.
2-11
-------
Benthic samples for measures of species composition, abundance, and biomass were col-
lected at all sampling sites during the second interval and at continuous monitoring sites
during the first and third intervals. Samples were collected with a Young-modified Van Veen
grab that samples a surface area of 440 cm2; three grabs were collected at each site. The
Young grab was selected because it is deployed easily from small boats, and it samples both
mud and sand habitats adequately. The maximum depth of penetration of the grab was 10
cm, and only grabs that penetrated deeper than 7 cm were accepted. A small core was taken
from each grab to determine silt-clay content. The remaining sample was sieved through a
0.5 mm screen using a backwash technique that minimized damage to soft-bodied biota.
Samples were preserved in 10% buffered formaldehyde-rose bengal solution and stored for at
least 60 days prior to processing.
In the laboratory, macrobenthos were sorted, identified to the lowest practical taxonomic level,
and counted. All macrobenthos were identified to species, except for the following groups:
Taxonomic Group Level of Identification
Class Anthozoa
Subclass Copepoda
Phylum Nemertinea
Subclass Ostracoda
Class Turbellaria
Class
Order
Phylum
Subclass
Class
For samples collected in low salinity (less than 5 ppt) water, oligochaetes were identified to
species, and chironomids to genus. Above 5 ppt salinity, oligochaetes were identified to class
(most were assumed to belong to the genus Tubificoides), and chironomids were identified
only to family. Biomass was measured for approximately 30 species that were expected, prior
to the beginning of the 1990 Demonstration Project, to be dominant (Table 3-3). All other
species were combined into biomass groups defined by feeding type and major taxonomic
group (i.e., sub-surface, deposit-feeding polychaetes). Bivalves and polychaetes were
classified into feeding categories defined by Fauchald and Jumars (1979), Jorgenson (1966),
and Bousfield (1973). Shell-free dry weight after drying at 60°C was determined using an
analytical balance with an accuracy of 0.1 mg. Large bivalves (greater than 2-cm long) were
shucked prior to determining biomass. Smaller shells were removed by acidification using a
10%HC1 solution.
2-12
-------
Table 2-3. Biomass groups for macrofaunal species. Biomass determined for each
species or group of species listed.
Amphipods
Ampelisca sp.
Corophium sp.
Gammaridae
Isopods
Cyathura sp.
Haustoriidae
Leptocheirus sp.
Monoculodes sp.
Unciola sp.
Other Amphipods
Other Isopods
Decapods
All
Chironomids
Intact individuals and
Posterior fragments
anterior fragments
Polychaetes
Glycera sp. Marenzelleria viridis
Heteromastus filiformis Mediomastus ambiseta
Leitoscoloplos sp. Neanthes succinea
Maldanidae Nephtys sp.
Other sub-surface deposit feeding polychaetes
Other surface feeding polychaetes
Oligochaetes
Intact individuals and
Posterior fragments
anterior fragments
Bivalves
Corbicula fluminea Mulinia lateralis
Ensis directus Nucula sp.
Gemma gemma Rangia cuneata
Mercenaria mercenaria Tellinidae
Paraprionospio pinnata
Polydora sp.
Streblospio benedicti
Other carnivore/omnivore polychaetes
Unidentified polychaetes and fragments
Yoldia limatula
Other deposit feeding bivalves
Other suspension feeding bivalves
Unidentified bivalves
Gastropods
Acteocina canaliculata
Hydrobia sp.
Other Gastropods
Miscellaneous
Echinodermata
Hemichordata
Phoronis sp.
Nemertinea
2-13
-------
2.2.2 Sediment Profile Images
Benthic animals are important regulators of the deposition and resuspension of bottom
sediments, the exchange of constituents between bottom sediments, and the exchange of
constituents between bottom sediments and the overlying water (Rhoads 1974; Rhoads et al.
1978; Rhoads and Boyer 1982; Aller 1982). By ventilating and displacing sediments during
burrowing and feeding, they affect geochemical profiles in sediments and pore waters. This is
particularly true in higher salinity habitats with fine-grained sediments, where there is no wave
disturbance or tidal scour. In these habitats, the apparent depth to the redox potential
discontinuity (RPD), the oxygenated layer of sediments, is positively associated with the
presence of larger, deep-burrowing, longer-lived organisms, and inversely related to the
presence of smaller, short-lived, surface dwelling, opportunistic species. The former condition
is representative of acceptable ecological conditions, and the latter may be associated with
physical or anthropogenically induced disturbance. Sediments exhibiting a shallow RPD that
are dominated by shallow-burrowing, short-lived species have been shown to be either
chemically or organically enriched (Rhoads and Germano 1986; Scott et al. 1987; O'Connor et
al. 1989; Valente et al. 1992).
Sediment profile images were collected at 20 stations in the northern half of the Virginian
Province. Replicate photographic images were taken using a Benthos Model 3771 sediment
profile camera and the Remote Ecological Monitoring of the Seafloor (REMOTS)® imaging
analysis system (Rhoads and Germano 1982, 1986). The camera photographs the sediment-
water interface in the vertical plane. The photograph is processed to quantify apparent RPD
depth, grain size, relative abundance of surface tube structures, boundary roughness,
penetration depth, and the presence of feeding voids and methane gas bubbles. The depth of
the apparent RPD is determined in the sediment profile images as the boundary between
light-colored, oxidized surface sediments and underlying grey to black reduced sediments.
The successional stage is characterized by the types of fauna present, or whose presence is
inferred, in the REMOTS® image. Stage I communities typically consist of small tubicolous
spionid or capitellid polychaetes that exploit recently disturbed or open space. Stage II
represents a transitional community that commonly consist of dense aggregations of
tubicolous amphipods (e.g., Ampelisca) and tellinid bivalves. Stage III fauna are represented
by relatively large, head-down deposit feeders. These are rarely observed directly on the
images, but their presence is inferred by feeding pockets or voids at depth. Chemical
parameters are indicated by the presence of highly reflective methane gas bubbles in the
sediment, and low reflectance (sulfitic) sediment near the sediment-water interface (high
sediment oxygen demand). A multi-parameter index, the Organism-Sediment Index (OSI),
which ranges from -10 to +11, was calculated to characterize sediment quality (Table 2-4).
The OSI is based on measurements of the RPD depth, infaunal successional stage, and the
apparent presence of selected chemical attributes (e.g., methane gas).
2-14
-------
Table 2-4. Calculation of the OSI based on sediment profile images
Mean Apparent RPD Depth Measurement
0.00 cm
>0 to 0.75 cm
0.76 to 1 .50 cm
1.51 to 2.25 cm
2.26 to 3.00 cm
3.01 to 3.75 cm
>3.75 cm
Index Score
0
1
2
3
4
5
6
Infaunal Successional Stage
No visible macrofauna
Small, short-lived, surface-dwelling opportunists: Stage I
Transitional, surface-dwelling, tubicolous amphipods, tellinid
bivalves: Stage II
Large, head-down, deposit feeding polychaetes and bivalves:
Stage III
Stage I on III
Stage II on III
-4
1
3
5
5
5
Chemical Parameters
Methane gas present
High apparent sediment oxygen demand
-2
-4
OSI = Total of chosen index values (range: -10 to +1 1)
2.2.3 Fish
There are several advantages to using fish as potential indicators of estuarine condition.
Because of their longevity and dominant position at the upper end of the food web, fish
responses integrate many short-term and small-scale environmental perturbations. They are
2-15
-------
known to respond to most of the major environmental problems of concern in estuaries,
including eutrophication, habitat modification, and pathogenic or toxic contamination.
Eutrophication can affect fish adversely by reducing dissolved oxygen below levels that are
critical for growth or survival. Habitat modification, such as the loss of submerged aquatic
vegetation, has been linked to decreased fish productivity through loss of important nursery
areas. Toxic and pathogenic contaminants can decrease fish growth, reproduction, or survival
and can make fish unsafe for human consumption. Fish also are valuable as indicators
because of their importance for determining the public perception of estuarine quality.
Reports of fishery closures due to chemical and viral contamination alarm the public.
Fish were collected by trawling with a 16-m (footrope), high-rise net with 5-cm mesh wings
and a 2.5-cm mesh cod end. The net was towed for 10 minutes against the tide at speeds
between 0.6 and 1.0 m/s (i.e., speed over the bottom between 0.3 and 1.0 m/s). All fish
caught in a trawl were identified to species and counted; up to 30 fish of each species from
each collection were measured to the nearest millimeter. Up to five individuals in
predetermined size ranges from each of 10 target species (Table 2-5) were retained from
each trawl for tissue analysis. The specimens were gutted, labeled, frozen on dry ice, and
shipped to the laboratory, where they were stored frozen to await analysis for the same
chemicals measured in the NOAA National Status and Trends Program (Table 2-6). No PAHs
were planned for tissue residue analyses because of the low probability of finding PAHs in
muscle tissue. Delays in the laboratory resulted in exceeding the holding time for frozen
samples; consequently, fish tissue samples were not processed as part of the 1990
Demonstration Project.
Table 2-5. Targeted species and size ranges of fish retained for tissue analysis during
the 1990 Virginian Province Demonstration Project. If fish from the primary
size range were not captured, fish were selected from the secondary range.
Target Species
Channel Catfish (Ictalurus punctatus)
Atlantic Croaker (Micropogon undulatus)
Hogchoker (Trinectes maculatus)
Summer Flounder (Paralichthys dentatus)
Spot (Leiostomus xanthurus)
White Catfish (Ictalurus catus)
Weakfish (Cynoscion regalis)
Winter Flounder (Pseudopleuronectes americanus)
Windowpane Flounder (Scophthalmus aguosus)
White Perch (Morone americana)
Primary Size
Range (mm)
200-300
200-300
100-150
350-450
150-250
200-300
300-400
300-400
300-400
150-250
Secondary
Range (mm)
300-400
300-400
150-250
200-350
100-150
300-400
200-300
200-300
200-300
100-150
2-16
-------
Table 2-6. Analytical measurements for fish and bivalve tissue samples collected during
the 1990 Virginian Province Demonstration Project
DDT and its Metabolites
o,p'-DDD
p,p'-DDD
o,p'-DDE
p,p'-DDE
o,p'-DDT
P)p'-DDT
Aldrin
Alpha-Chlordane
Trans-Nonachlor
Dieldrin
Heptachlor
Chlorinated
18
PCB No.
8
18
28
44
52
66
101
105
118
128
138
153
170
180
187
195
206
209
Trace Elements
Aluminum Mercury
Arsenic Nickel
Cadmium Selenium
Chromium Silver
Copper Tin
Iron Zinc
Lead _
Pesticides other than DDT
PCB
Heptachlor epoxide
Hexachlorobenzene
Lindane (gamma-BHC)
Mirex
Congeners:
Compound Name
2,4'-dichlorobiphenyl
2,2',5-trichlorobiphenyl
2,4,4' -trichlorobiphenyl
2 2'
*-jc- >
3,5'-tetrachlorobiphenyl
2,2',5,5'-tetrachlorobiphenyl
2,3',
2,2',
4,4'-tetrachlorobiphenyl
4,5,5'-pentachlorobiphenyl
2,3,3',4,4'-pentachlorobiphenyl
2,3',
4,4',5-pentachlorobiphenyl
2,2',3,3',4,4'-hexachlorobiphenyl
2,3',
2,2',
2,2',
2 2'
C-tC- 1
2,2',
2,2',
2 2'
£•*£- j
3,4,4' ,5-hexachlorobiphenyl
3,4,4',5'-hexachlorobiphenyl
4,4',5,5'-hexachlorobiphenyl
3,3',4,4',5-heptachlorobiphenyl
3,4,4',5,5'-heptachlorobiphenyl
3,3',4,4',5,6-octachlorobiphenyl
3,3',4,4',5,5',6-nonachlorobiphenyl
decach lorobiphenyl
At all stations where fish were collected, up to 30 individuals of each target species were
inspected for visible external pathological disorders. This inspection included checking the
2-17
-------
body surface and fins for skin discoloration, raised scales, white or black spots, ulcers, fin
erosion, lumps or growths, parasites, and opercular deformity; the branchial chamber for gill
discoloration, erosion, deformity, parasites, and lumps or growths; the buccal cavity for
hemorrhages, parasites, and lumps or growths; the overall morphology of the fish for skeletal
malformations; and the condition of the eyes (e.g., cloudiness, hemorrhages, enlarged or
sunken). Specimens with pathologies were preserved in Dietrich's solution for laboratory
verification. At ITE sites, up to 30 specimens of each target species and 20 specimens of
non-target species that were free of visible pathological abnormalities were preserved for
quality control checks of field observations. At indicator testing and evaluation sites, tissue
samples for histological examination were collected from all target fish that exhibited visible
pathology. For comparison, tissue samples were taken from up to 25 target fish and 10 non-
target fish that did not exhibit visible pathology. Tissue samples were dehydrated in an
ethanol gradient, cleared in a xylene substitute, infiltrated, and embedded in paraffin.
Sections were cut at 6 |im on a rotary microtome, stained with Harris' hematoxylin and eosin,
and examined microscopically. These examinations were intended to determine the
relationship between the incidence of visible external abnormalities and the presence and
kinds of internal histopathological changes, and to evaluate the ability of this indicator to
discriminate between polluted and unpolluted sites.
2.2.4 Large Bivalve Abundance and Tissue Contamination
The occurrence of large, older bivalves at a site generally indicates that environmental
conditions have been relatively stable over time. The relative immobility of bivalves makes
them good integrators of long-term environmental conditions at the site from which they were
collected. The burrowing habit of many bivalves places them where exposure to stress, such
as low dissolved oxygen concentrations and contaminants, is likely to be high. Filter feeding
bivalves concentrate contaminants ingested with food to levels many times higher than those
in the water. Tissue contamination can reduce growth and survival, which can adversely
affect production. It also reduces the value and quality of bivalve meats for human
consumption.
Attempts were made to collect large infaunal bivalves from ITE sites and from base sampling
sites during interval 3 to determine if a sufficient number of specimens could be collected to
evaluate the magnitude and primary sources of variation in the concentrations of contaminants
in bivalve tissue. If most of the natural variation in bivalve tissue contamination can be
accounted for, then contamination in large bivalves may be a valuable indicator of suitability
for human use. A 30-cm rocking chair dredge was used to collect specimens of large infaunal
species. The dredge, equipped with a 2.5-cm mesh cod-end bag, was towed over the bottom
for five minutes at approximately one knot. Mollusks were identified to species and counted.
Shell length was measured to provide an indication of the age structure of the population. Up
to 10 individuals of each species were labeled, frozen on dry ice, and shipped to the analytical
2-18
-------
laboratory, where they were stored frozen to await analysis. Bivalve tissue samples were
analyzed using standard methods to determine the concentrations of the contaminants listed
in Table 2-6.
2.2.5 Dissolved Oxygen Concentration
Dissolved oxygen (DO) concentration is an important indicator of environmental condition
because it is a fundamental requirement for maintenance of populations of benthos, fish,
shellfish, and other aquatic biota. DO concentrations are affected by environmental stresses,
such as point and nonpoint sources of nutrients, particulates, and dissolved organic matter.
Stresses that occur in conjunction with low DO concentrations may be even more detrimental
to biota (e.g., exposure to hydrogen sulfide, decreased resistance to disease and
contaminants). DO levels, however, are highly variable over time, fluctuating widely due to
the tide, winds, and biological activity (Kemp and Boynton 1980; Sanford et al. 1990; Welsh
and Eller 1991). The objective of the 1990 Demonstration Project was to collect regional data
to determine the best measure for representing dissolved oxygen exposure (e.g., percent of
time below a critical value, mean over a defined time period), to evaluate the stability of DO
levels over the sampling period, and to determine if the occurrence, magnitude, and duration
of exposure to extreme low dissolved oxygen stress can be predicted using short-term
continuous records of dissolved oxygen concentration.
Dissolved oxygen was sampled in two ways during the 1990 Demonstration Project: 1) point-
in-time water column profiles, and 2) continuous bottom measurements. Point-in-time
measurements were made at all sampling sites during every visit. Data from the point
measurements were used to generate regional status estimates. Continuous monitoring was
conducted at 23 base sampling sites'selected to represent a range of geographic and habitat
types in the Virginian Province. These sites were known as long-term dissolved oxygen
(LTDO) sites. Continuous monitoring data were used in Monte Carlo simulations to evaluate
alternative sampling strategies for measuring dissolved oxygen cost-effectively (Section 4).
A Seabird model SEE 25 CTD equipped with a Beckman polarographic DO electrode was
used to make the point-in-time measurements. A CTD cast was performed during all station
visits in each of the three sampling intervals to obtain vertical profiles of the water column. In
addition to producing profiles of dissolved oxygen, the CTD measured salinity, temperature,
pH, transmissometry, fluorometry, and photosynthetically active radiation (PAR). At each site,
the CTD was held at the water surface for approximately one minute to allow the DO probe to
reach thermal equilibrium. Upon reaching equilibrium, the CTD was lowered through the
water column at a rate of approximately 1 m per second, until it reached 1 m from the bottom.
The CTD was held at this position for two minutes, then retrieved and downloaded to an
onboard laptop computer.
2-19
-------
Hydrolab DataSonde 3 data loggers were used at the 23 continuous-monitoring sites. These
instruments were programmed to record DO measurements every 30 minutes and were
suspended approximately 1 m from the bottom at each continuous-monitoring site for 70 days
between 19 June and 30 August (sampling intervals 1 and 2). The Hydrolab was retrieved
approximately every 10 days. The stored data was downloaded to a computer, and a
calibrated Hydrolab was redeployed.
2.2.6 Chlorophyll a
The concentration of chlorophyll a in the water column is a measure of the biomass of
phytoplankton communities (Parsons et al. 1977). This measurement was included in the
1990 Demonstration Project as a potential indicator of eutrophication, since phytoplankton
biomass is expected to increase in eutrophic areas receiving large inputs of nutrients.
A Go-Flo sample bottle was used to collect water samples from 1 m below the surface. A
portion of the water sample was filtered through a 25-mm glass fiber filter (type GF-F, 0.7-|im
nominal pore size) using a 60-ml syringe filter assembly. The filter was removed from the
syringe assembly using forceps, wrapped in aluminum foil, placed in a small zip-lock bag, and
frozen on dry ice. Samples were kept frozen until analysis.
2.2.7 Sediment Toxiclty
Sediment toxicity testing is the most direct measure available for determining the toxicity of
contaminants in sediments to indigenous biota. It improves upon direct measurement of
sediment contamination because many contaminants are tightly bound to sediment particles or
are chemically complexed and, therefore, are not biologically available (USEPA 1989).
Sediment toxicity testing cannot be used to replace direct measurement of the concentrations
of contaminants in sediment because such measurements are an important part of interpreting
the results of toxicity tests.
Sediment for toxicity tests was collected with the Young-modified Van Veen grab used for
benthic invertebrate sampling. Grabs were conducted only during the second sampling
interval at all but index sites. The top 2 cm of sediment from five or more grabs were
removed and placed into a teflon bowl. Care was taken to avoid collecting material from the
edge of grabs and to use only samples that had undisturbed sediment surfaces. The teflon
bowl was kept on ice in a cooler between grabs to reduce the temperature of the sample and
to prevent accidental contamination. After approximately 3,000 ml of sample was collected,
the composite was homogenized and distributed to appropriate sediment chemistry, toxicity,
and grain size sample containers.
Toxicity tests were performed on the composites of surface sediments collected from each
station. Tests were conducted using the standard 10-day acute test method (Swartz et al.
2-20
-------
1985; ASTM 1990) and the tube-dwelling amphipod Ampelisca abdita. Five replicate tests
were completed. For each toxicity test, 200 ml of sediment from the composited sample was
placed in 1-1 glass jars and covered with 600 ml of 30 ppt seawater. The bioassays were
conducted under static conditions for 10 days at 20°C. An additional series of toxicity tests
was conducted to determine whether contaminated sediments from low salinity waters
became less toxic in the higher salinity waters (30 ppt) used for these tests; thus, for all
samples from stations with bottom salinities of 5 ppt or less, a second 10-day acute test was
conducted concurrently using the freshwater amphipod, Hyalella azteca. The procedure for
both species was identical, except that the low salinity test with Hyalella used well water.
2.2.8 Sediment Contaminant Concentrations
Metals, organic chemicals, and fine-grained sediments entering estuaries from freshwater
inflows, point sources of pollution, and various nonpoint sources including atmospheric
deposition, generally are retained within estuaries and accumulate in the sediments (Turekian
1977; Forstner and Wittman 1981; Schubel and Carter 1984; Nixon et al. 1986; Hinga 1988).
This is because most contaminants have an affinity for adsorption onto particles (Hinga 1988;
Honeyman and Santsche 1988). Chemical and microbial contaminants generally adsorb to
fine-grained materials in the water and are deposited on the bottom, accumulating at
deposition sites, including regions of low current velocity, deep basins, and the zone of
maximum turbidity. The concentration of contaminants in sediments is dependent upon
interactions between natural (e.g., physical sediment characteristics) and anthropogenic
factors (e.g., type and volume of contaminant loadings; Sharp et al. 1984).
Sediment samples for contaminant analysis were collected during the second sampling
interval at all sites except index sites. Samples were collected as a subsample of the
sediment slurry collected for sediment toxicity testing. The sediment was placed in clean
glass jars with teflon lid liners, shipped on ice, and stored frozen in the laboratory prior to
analysis. Sediments were analyzed for the NOAA National Status and Trends suite of
contaminants for sediments (Table 2-7) using standard analytical methods (Table 2-8).
2.2.9 Sediment Clostridium Spores
Clostridium perfringens is an obligate-anaerobic, spore-forming bacterium found in the feces of
warm-blooded animals. Spores accumulate in sediments and have been interpreted as
conservative tracers of fecal contamination because the spores survive longer than other
common indicators of fecal pollution (Bisson and Cabelli 1980; Duncanson et al. 1986).
2-21
-------
Table 2-7. Analytical measurements for sediment samples collected during the 1990
Virginian Province Demonstration Project
Polyaromatlc Hydrocarbons (PAHs)
Aconaphthone
Anlbraceno
8onz(o)anthraceno
Bonzo(a)pyronQ
Bonzo(o)pynjna
Blphonyl
Cnrysono
Dfoonz(a,b)anlhracon9
2,6-dImethylnaphthalens
Fluoranthene
Fluorene
ldeno(1,2,3-c.d)pyrene
2-methylnaphthalene
1-melhylnaphthalene
1-methylphenanthrene
Naphthalene
Perylens
Phenanthrene
Pyrene
Benzo(b)fluoranthene
Acenaphthylene
Benzo(k)fluoranthene
Benzo(g,h,i)perylene
2,3,5-Trimethylnaphthalene
DDT and Its metabolites Chlorinated pesticides other than DDT
o,p'-DDD
p,p'-DDD
0,p'-DDE
Aluminum
Iron
Manganese
p,p'-DDE Aldrin
o.p'-DDT Alpha-Chlordane
p,p'-DDT Trans-Nonachlor
Dleldrin
Heplachlor
Major Elements
Antimony Copper
Arsenic Lead
Cadmium Mercury
Chromium Nickel
Heptachlor epoxlde
Hexachlorobenzene
Llndane (gamma-BHC)
Mlrex
Trace Elements
Selenium
Silver
Tin
Zinc
18 PCB Congeners:
No.
8
18
28
44
52
66
101
105
118
128
138
153
170
180
187
195
206
209
Compound Name
2,4'-dichloroblphenyl
2,2',5-trlchloroblphenyl
2,4,4Mr!chloroblphenyl
2,2',3,5'-tetrachloroblphenyl
2,2',5,5'-tetrachloroblphenyl
2.3',4,4'-tetrachloroblphenyl
2,2',4,5,5'-pentachloroblphenyl
2,3,3',4,4'-pentachloroblphenyl
2,3',4,4',5-pentachloroblphenyl
2,2',3,3',4,4'-hexachloroblphenyl
2,3',3,4,4',5-hexachloroblphenyl
2,2',3,4.4',5'-hexachloroblphenyl
2,21.4,4'.5,5'-hexach!oroblphenyl
2,2',3,3',4,4',5-heptachloroblphenyl
2,2',3,4,4',5,5'-heptachloroblphenyl
2,2',3,3',4,4',5,6-octachloroblphenyl
2,2',3,3',4,4',5,5',6-nonachloroblphenyl
decachloroblphenyl
Other measurements
Tributyltfn
Acid volatile sulfides
Total organic carbon
Grain size distribution
2-22
-------
Table 2-8. Analytical methods used in 1990 for determination of chemical contaminant
concentrations in sediments
Compound(s)
Inorganics:
Ag, Al, Cr, Cu, Fe,
Mn, Ni, Pb, Zn
As, Cd, Sb, Se, Sn
Hg
Organics:
Extraction/Cleanup
PAH measurement
PCB/pesticide
Method
Total digestion using HF/HNO3 (open vessel hot plate) followed
by inductively coupled plasma-atomic emission spectrometry
(ICP-AES) analysis.
Microwave digestion using HNOg/HCI followed by graphite furnace
atomic absorption (GFAA) analysis.
Cold vapor atomic absorption spectrometry
Soxhlet extraction, extract drying using sodium sulfate, extract
concentration using Kuderna-Danish apparatus, removal of
elemental sulfur with activated copper, removal of organic ,
interferents with GPC and/or alumina.
Gas chromatography/mass spectrometry (GC/MS)
Gas chromatography/electron capture detection (GC/ECD) with
second column confirmation
The concentration of Clostridium spores was measured from subsamples of the homogenized
sediment samples collected for toxicity tests and contaminant measurements. Spores were
extracted from sediment samples by vortexing sediment slurries following the addition of
deionized water and sodium metaphosphate. Extracted water was filtered through sterile
membrane filters, which were cultured on agar (Emerson and Cabelli 1982). Colonies were
counted after staining with ammonium hydroxide.
2.2.10 Sediment Grain Size
The physical characteristics of estuarine sediments (e.g., grain size, silt-clay content) influence
the distribution of benthic fauna and the accumulation of contaminants in sediments (Rhoads
1974; Plumb 1981). Sediment grain size and silt-clay content data were collected to help
interpret the responses of these parameters. A subsample from the benthic invertebrate and
sediment contaminant grabs was retained to determine the silt-clay content and the
2-23
-------
distribution of grain sizes in the sediment. Samples for the determination of silt-clay content
and grain size distribution were sieved through 63-|im mesh. Both the filtrate and the fraction
retained on the sieve were dried in an oven at 60°C and weighed to calculate the proportion
of silts and clays in the sample. Procedures for determining grain size distribution generally
followed the framework described for the silt-clay analyses; however, the fraction retained on
the sieve was processed through an additional sieve analysis to determine specific grain size
percentages. The filtrate was processed using pipette analysis.
2.2.11 Water Column Toxicity
Water column toxicity tests were used as indicators of the presence of contaminant
concentrations potentially toxic to planktonic estuarine organisms. Samples for water column
toxicity tests were collected from the ITE sites during the second sampling interval. A 2-I
water sample was taken from 1 m below the surface using a clean Go-Flo sampling bottle.
The water sample was placed on ice and shipped immediately to the laboratory so that the
tests could be performed within 48 hours of sample collection.
Three water column bioassays were conducted: 1) sea urchin fertilization test (Nacci et al.
1987), 2) red algae sexual reproduction test (Thursby and Steel 1987), and 3) bivalve
fertilization and larval growth test (APHA 1985). The results of the three tests were compared
to determine their relative sensitivity and were correlated with the results of other indicators to
determine whether this indicator should be included in future EMAP monitoring in estuaries.
2.2.12 Marine Debris
Aesthetic appeal is an important factor determining the suitability of an estuary for human use.
The presence of trash in the water and the clarity of the water are the primary means by
which the public assesses the aesthetic quality of an estuary. Observations of these factors
were made at all sites during the 1990 Demonstration Project because of their importance to
the public and the relatively small incremental cost of the effort. The kinds and relative
amounts of floating and submerged (i.e., collected in otter trawls) trash were noted for all
stations. Trash was categorized as paper and plastics, cans and bottles, medical, and other
wastes.
2-24
-------
2.3 DATA COLLECTION AND SAMPLE TRACKING
Field crews were supplied with two personal computers and appropriate software to facilitate
collection and electronic recording of data, and sample tracking. The software included a
navigation system linked to loran-C and Global Positioning System (GPS) navigation aids.
The navigation system recorded the location and time of day for each observation and was
used to identify boat speed and distance for all fish trawls.
All samples, data diskettes, shipments, and equipment were bar coded to facilitate sample
tracking and to reduce transcription errors. Field computers were equipped with bar code
readers to record sample identification information. The field computer system included
software to capture sample identifications, visual observations made by field crews, and all
electronic data from the CTD profiles and the continuous dissolved oxygen meter.
Copies of all data were made on diskettes at the end of each day to minimize data loss from
potential equipment failure. With the exception of the CTD and continuous dissolved oxygen
meter data, records also were kept on paper data sheets. Communications software and
commercial carrier phone lines were used to transfer data daily to the Field Operations Center
at the USEPA Environmental Research Laboratory in Narragansett, Rl. All electronic records
were verified with paper forms.
The bar-coded information was used to follow all sample shipments from point-of-origin to
point-of-analysis. Samples were followed through processing and analysis steps using an
individual sample identification number. Further details on data management for the
Demonstration Project are presented in Rosen et al. (1990).
2.4 ANALYTICAL METHODS FOR STATUS ASSESSMENT
Three types of analyses were conducted for this report, including those to 1) test and develop
indicators, 2) evaluate the overall sampling design, and 3) conduct an initial evaluation of the
status of estuaries within the Virginian Province. The analytical steps used to develop
indicators and evaluate the sampling design and the results of those analyses are described
in Sections 5 and 6 of this report. The analytical methods used to complete the evaluation of
estuarine status that appeared in Section 2 are given below in three parts. The first describes
the analytical procedures used to translate measurements made in the field and data
generated in the laboratory into information for estimating estuarine status. The second
describes the general analytical methods used to generate areal estimates of estuarine
condition, and the third contains methods for calculating confidence intervals associated with
the areal estimates.
2-25
-------
2.4.1 Assessment Methods for Individual Indicators
2.4.1.1 Benthic Index
The benthic assemblage at each sampling site was characterized as degraded or
nondegraded based on a benthic index described in Section 4. Degraded and nondegraded
conditions were defined relative to regional reference conditions identified from indicator
testing stations (Section 4). The index was developed from a discriminant function that
incorporates five characteristics of the benthic assemblage: proportion (expressed as a
percent) of the expected number of species adjusted for salinity, number of amphipods,
percent of total abundance in molluscan taxa, number of capitellids, and average weight per
individual polychaete. Only benthic infaunal species were included in this analysis and all
data were transformed [Log10 (value +1)] prior to conducting the analysis.
The discriminant function combining the five indicators was:
Discriminant Score =
(0.011 * Proportion of expected number of species) +
(0.817 * Number of amphipods) +
(0.671 * Percent of total abundance as bivalves) +
(0.465 * Number of capitellids) +
(0.577 * Average weight per individual polychaete)
Prior to calculating discriminant scores, the log transformed values for each indicator were
normalized using a standard normal (Z) transformation:
Z =
where Z is the transformed indicator value and x is the log of the indicator. The standardized
means and standard deviations for each indicator were calculated using the indicator testing
stations and are given in Table 2-9.
2-26
-------
Table 2-9. Standardized mean (p.) and standard deviation (o) for indicators used in the
discriminant function for the benthic index. All values were calculated with data
from the indicator testing stations only.
Indicator
Proportion of affected
number of species
Number of Amphipids
Percent of total abundance
as bivalves
Number of Capitellids
Average weight per
individual polychaete
Transformed Mean (u)
1 .4503
0.7986
0.4647
0.7187
0.0003967
Transformed Standard
Deviation (a)
0.5498
0.8460
0.5422
1 .0031
0.0005539
In this formulation, the number of species in each sample was expressed in terms of the
expected number of species to normalize for the effects of salinity on the number of species
present (Remane and Schlieper 1971). The number of species expected at a site was
calculated from a third order polynomial fit to the 90th percentile of the distribution of the
number of species vs. salinity for sites sampled during the 1990 Demonstration Project (n =
34, r2 = 0.96, p < 0.0001) (see Section 4) . The polynomial was developed for a 3 ppt salinity
running average of the number of species:
Expected Number of Species = 13.733 - (1.101 * Salinity) +
(0.132 * Salinity2) - (0.002 * Salinity3)
The dimensionless scale of the discriminant score was converted to a range of 0 to 10 to
clarify the presentation of results using the following algorithm:
_, ..... Discriminant Score - Minimum Discriminant Score 1n
Benthic Index = :—— — =-— x 10
Discriminant Score Range
The threshold boundary between degraded and nondegraded benthos was set at 3.4 This
was the midpoint between the mean of nondegraded and degraded discriminant values (see
Section 5).
2-27
-------
2.4.1.2 Fish .
Visible pathological disorders were separated into two categories: those considered to be
inducible by exposure to polluted conditions, and those that may not be the result of
environmental conditions at the site, such as parasites. Only the first category, which included
spots, ulcers, fin erosion, lumps or growths, skeletal malformations, opercular deformities, or
abnormal condition of the eyes (e.g., cloudiness, hemorrhages, enlarged or sunken), was
used in making status estimates.
The preliminary status evaluation contained two estimates of condition based on visible
pathology: 1) the area with an abnormally high rate of pathological disorders, and 2) the
prevalence of disorders per 1000 fish. The areal estimates were calculated using the same
methodology as for the other parameters. The threshold criterion for an individual site was
established at greater than 1% of the fish with visible pathological disorders. The threshold
criterion also required observing disorders in at least two fish at the site to minimize false
positives at sites where few fish were collected.
The estimate of prevalence of pathological disorders in fish differed from the generic
methodology in that it involved a weighting factor in addition to the areal inclusion probability
for each sampling site. This factor was relative abundance of fish caught at each site in the
standardized trawl. This estimate involved two steps. First, a between-species prevalence
rate for the station was calculated as the abundance-weighted average of the prevalence of
each individual species. Next, the provincewide prevalence rate was estimated as a weighted
average for the individual station values, where the weighting factor was the product of
average catch-per-haul at the station and the station's inclusion probability. Prevalence rates
for resource classes within the province were calculated in the same manner, except that
inclusion probabilities for the class were used rather than for the province as a whole.
2.4.1.3 Sediment Toxicity
Estimates of area in the Virginian Province containing toxic sediments were based oh the
results of bioassays using the amphipod Ampelisca abdita; tests conducted with the
freshwater amphipod Hyalella azteca were used only to evaluate the response of Ampelisca in
low salinity habitats (see Section 5). Sediment toxicity was treated as a categorical variable in
the status evaluation. A relative measure of toxicity was employed to facilitate comparisons
between sites over a series of bioassays. Sediments were considered toxic if the survival of
Ampelisca in test sediments was less than or equal to 80% of the survival observed in clean,
control sediments, and if the percentage of survival in test and control sediments was
significantly different (p <, 0.05). These criteria were consistent with those established in
USEPA/ACE (1991).
2-28
-------
2.4.1.4 Sediment Contaminant Concentrations
Sediment contaminants were analyzed categorically, and concentrations higher than the ER-L
values of Long and Morgan (1990) were classified as being of biological concern. Analyses
were performed separately by contaminant and by classes of contaminants. The classes
included contaminants having both natural and anthropogenic sources, such as metals;
contaminants having mainly anthropogenic sources, such as polycyclic aromatic hydrocarbons
(PAHs); and synthetic contaminants from industrial applications, such as polychlorinated
biphenyls (RGBs). If any of the contaminants within a class were high, then the site was
considered high for the whole class. Concentrations of all contaminants within the class had
to be low to consider the site uncontaminated with that class of chemicals. The information
from each site was used to estimate the status of contaminants by area throughout the
province.
ER-L values represent concentrations at which some type of biological effects were noted in
at least 10% of exposure studies reviewed by Long and Morgan (1990). These values were
interpreted as concentrations at which biological effects begin to occur and were used to
evaluate the extent of area where sublethal effects may occur. ER-L values are not available
for all of the contaminants measured by EMAP, and data for chemicals without ER-L values
were not included in the analysis. The estimates in Section 2, therefore, represent minimum
estimates. Section 5 examines the effect of choosing alternative critical contaminant
concentrations.
2.4.1.5 Sediment Clostridium Spores
The concentration of Clostridium spores in sediments was used to identify estuarine areas
influenced by anthropogenic sewage inputs. Clostridium spores are ubiquitous in coastal
marine environments and are found along the western Atlantic continental shelf at
concentrations of about 10 colony forming units (CPU) per gram dry weight of sediment
(Cabelli and Pedersen 1982; Duncanson et al. 1986); therefore, a threshold concentration had
to be defined to separate background spore concentrations from elevated concentrations due
to sewage inputs. For the assessment, Clostridium spore concentration was treated as a
categorical variable, and sediments were considered to be influenced directly by sewage
inputs if Clostridium concentrations were greater than 250 CPU per gram wet weight of
sediment. This value is approximately equivalent to the background level of 100 CPU per
gram dry weight of sediment identified in other studies (Cabelli, personal communication).
Sediments with Clostridium spore concentrations less than or equal to 250 CPU per gram wet
weight of sediment were not considered to be directly influenced by sewage inputs. These
areas may, however, be indirectly influenced by farfield sewage inputs upstream or in other
areas of the estuary from which spores were transported by currents.
2-29
-------
2.4.1.6 Marine Debris
In the status evaluation presented in Section 2, marine debris was treated as a categorical
variable, indicating the presence or absence of trash at a site. Only debris of anthropogenic
origin (e.g. bottles, cans) was included in the estimates. Estimates were based on debris
found either on the surface or on the bottom (i.e., collected in fish trawls).
2.4.1.7 Water Clarity
Water clarity was calculated using the vertical profiles of photosynthetically active radiation
(PAR). Light extinction coefficients (k) were calculated from the profiles of PAR taken at each
station. The light extinction coefficient was converted to Secchi disc depth (SDD) using the
relationship k=1.7/SDD (Poole and Atkins 1929). A Secchi disc depth of 0.3 m (1 ft) was the
threshold for water clarity.
2.4.1.8 Water Column Stratification
Density (sigma-t) was calculated using the standard equation of state for sea water (Millero
and Poisson 1981), and the difference between surface and bottom water density was used
as an indicator of water column stratification. A change in density greater than 2 was
considered evidence for moderate to strong stratification, whereas a change less than 1 was
considered evidence for weak stratification or no stratification.
2.4.1.9 Integration of Indicators
One objective of EMAP-Estuaries is to develop indices of environmental condition that
integrate information from multiple indicators. Although individual response indicators provide
information concerning specific aspects of environmental condition, overall statements
regarding the condition of resources can be more useful to managers and nontechnical
audiences. Single integrated statements can be communicated and understood more easily
and are more appropriate for establishing and measuring progress towards environmental
goals.
Analyses were completed to integrate information from indicators to focus on the two
environmental attributes of interest (biological integrity and societal value). The integrated
picture of biological integrity was developed by combining information from the benthic index
and measures of fish pathology. Biological integrity was considered degraded at a site if the
value of the benthic index was below the threshold value defining degraded benthic
assemblages, or if the data for pathological disorders in fish exceeded the criteria of at least
2-30
-------
two fish with pathological disorders and an incidence of 1 % or greater. Biological integrity
was considered nondegraded at a site if the benthic index indicated the presence of
nondegraded benthic assemblages, and if the occurrence of pathological disorders did not
exceed the criteria stated above.
Information about societal value was included in a similar manner. If trash was found at a
site, Clostridium perfringens was present at greater than 250 CPU, or water clarity was less
than 0.3 m Secchi disc depth, then societal value at the site was classified as degraded. If no
trash was found, Clostridium was present only at background levels, and water clarity was
greater than 0.3 m, then societal value was considered acceptable.
The indices of biological integrity and aesthetics were combined further to describe overall
conditions within the estuaries of the Virginian Province. At each site, if either biological
integrity or societal value was categorized as degraded, then the environmental conditions at
the site were considered degraded. If both biological integrity and societal value were
nondegraded, then environmental conditions at the site were considered nondegraded. This
approach, in which each component indicator is weighted equally, is preliminary. EMAP
presently is exploring ways to develop an overall index that weights indicators according to
their relative importance to environmental managers and society.
2.4.2 Assessment Methods for Areal Estimation
Estimates of status for estuaries of the Virginian Province (Section 6) were made using data
collected at the probability-based (base and supplemental) sampling sites. Each site was
weighted according to the area associated with its sampling unit. For large estuaries, the area
associated with each sample was equal to the size of the hexagon (280 km2 for most sites;
70 km2 in the Delaware Estuary, where supplemental sites were sampled). For large tidal
rivers, the area was equal to the area of each river segment corresponding to the sampling
site as measured using a geographic information system. For small estuarine systems, the
area was equal to the area for that particular small system. For small estuarine systems
having both base and supplemental sites (Back River, Elizabeth River, Indian River Bay,
Mattopani River, Mullier River, and Mystic River), the area given to each was equal to the total
area of the system divided by the number of base and supplemental sites in that system. To
generate estimates within a resource class (large estuaries, large tidal rivers, small estuarine
systems), the proportional weighting factors for each sample were calculated by dividing the
individual area associated with each sampling site by the total area sampled in that resource
class.
Large estuaries, large tidal rivers, and small estuarine systems were defined as sample strata
in the sampling design; however, the stratification scheme was modified to define reporting
classes because of perceived misclassification during stratification. Large tidal rivers originally
2-31
-------
were designated on the basis of overall size and aspect ratio, but aspect ratio was considered
only for rivers as a whole. The result was that river segments were ecologically dissimilar in
some cases. In most cases, the biological resources in river segments were dependent upon
conditions in upstream reaches of the river; however, in some cases, biological resources,
especially benthic resources, were more dependent upon conditions downstream in the larger
estuarine systems into which the river flowed.
These differences were particularly acute in the Potomac River, where the lower portion is
more like an extension of the central axis of Chesapeake Bay than like a tidal river. The
average width of the most upstream Potomac segment is only 1 km, whereas in the most
downstream segment it averages more than 20 km. The channels in the most upstream
segments have to be dredged to 8 m for navigation; natural channels exist to more than 20 m
in the two lower segments. Water in the upper segments is not stratified and consists largely
of runoff from upstream areas. The lower portion is highly stratified, with saltier, denser water
from Chesapeake Bay flowing under the river runoff in classic, two-layer estuarine circulation
(Lippson et al. 1979). Salinity differences from surface to bottom in the two lower Potomac
segments were more than 5 ppt and 3 ppt during the 1990 sampling. No other tidal river
segment had a difference of even 0.5 ppt.
Because of their size and ecological similarity to the adjacent Chesapeake Bay, the two lower
stations in the Potomac River were included in the large estuary reporting class. This
reclassification demonstrates the flexibility of the EMAP design. Because each sample has an
associated area weight, relative inclusion probabilities can be calculated for almost any
combination of sampling sites; thus, the large tidal river subpopulation need not be defined by
the sampling stratum but can be defined on the basis of ecological properties. With this
reclassification, the total area for the three classes was: large estuaries - 16,889 km2, tidal
rivers - 1809 km2, and small estuarine systems - 4875 km2.
To generate estimates across classes (strata), weights for sites within each class were
adjusted so that the total of the weights for that class was equal to the total area represented
by the sites within that class. On occasion, data were missing due to missed or lost samples,
or failed quality control standards. In those circumstances, the total area represented by
sampling did not equal the total area in the three resource classes, and adjustments were
made for the evaluation of status to avoid undefinable sample area. A correction factor was
applied to the previous weightings to compensate for missing data. The correction factor was
the ratio of the area represented by all possible samples to the area represented by all
samples for which data were available. Correction factors were calculated for each reporting
class. For the small estuarine systems class, the area representing all possible samples was
the area of all estuaries in the list frame of small systems. The 1990 Demonstration Project
sampled approximately one-fourth (32 out of 137) of these estuaries.
2-32
-------
Estimates in the preliminary evaluation were based only on data collected during sampling
intervals 2 (19 July to 31 August) and 3 (1 September to 23 September). The first occurrence
of data from each site was selected for inclusion in the estimates. Data from sampling interval
1 were excluded from the calculations because examination of the data revealed that
exposure conditions were not as severe then as they were later in the summer (see Section 6-
1).
2.4.3 Procedures for Estimating Precision
The approximate 90% confidence intervals for the province were calculated based on the
assumption that the CDF estimates were distributed normally. The confidence intervals were
obtained by adding and subtracting 1.645 times the estimated standard error (square root of
the variance) to the estimated CDF value.
For small estuarine systems, estimates of CDFs and associated variances were computed
based on a random selection of small systems within the province, with replicate samples
taken from a subset of the selected systems (Cochran 1977). The resulting CDF estimate is:
* O v —
M
EA
where
A
S,x
Yi
m,
y*
A,
n
CDF estimate for value x
m'
= number of samples at 'small system /
1 if response is less than x
0 otherwise
area of small system /
number of small systems sampled
Since replicate samples were obtained only at a subset of the sampled small estuarine
systems, the formula for the estimated variance taken from Cochran (1977 eq. 11.30) was
modified to produce the following estimate of the approximate mean squared error (MSE) of
the CDF estimate:
2-33
-------
Az
K.I "* A Q
-—V ' '
n* fei m,
where
N =
ft -
n* =
number of small systems in the province (137)
n/N
number of small systems with replicate samples
m,
/77/-1
A = the total area of small systems in the province (4875 km2)
Estimates of CDFs for large tidal rivers were obtained by applying Horvitz-Thompson
estimation (Cochran 1977) with selection probabilities being inversely related to station area.
Data from all base stations and from supplemental stations sampled in the Delaware River
were used in this analysis. Areas of base stations in the Delaware River were adjusted to
reflect the inclusion of the supplemental stations. Estimates of CDFs were:
where
r,*
P - 1 T '
T'x '
= Estimated CDF at value x
y\ =
1 if response is less than x
0 otherwise
itf = inclusion probability for station/(1/area)
A = total area of the sampled tidal rivers
n = number of stations sampled
Variance estimators that are commonly used with general probability samples, such as the
Horvitz-Thompson or Yates-Grundy (Cochran 1977), require that all joint event probabilities
must be non-zero, which is not the case for systematic random designs, such as EMAP's.
2-34
-------
Estimates derived from such designs do not have unbiased estimators of variance. Instead,
the variance is approximated based on a model or set of assumptions. The variance
approximation to estimate confidence intervals presented in Section 6 for the status estimates
was derived from the Yates-Grundy estimator:
where
and
= probability that sites / and j are selected for sampling
1t» =
The resulting approximation no longer depends on the joint event probabilities, but only on the
first order inclusion probabilities.
The approximation was developed by Stehman and Overton (1989) and was derived under
the assumption that the population is randomized between successive draws. The use of this
approximation based on the Yates-Grundy variance estimator amounts to assuming that the
grid-based sample is nearly a simple random sample. Simulation studies (Overton and
Stehman 1987; Stehman and Overton 1987a,b) have demonstrated that the approximation
performs well in EMAP-like sampling circumstances. The systematic sample used in large
systems and tidal rivers should provide more precise estimates than simple random sampling,
so that the approximation will provide conservative estimates of precision.
Estimates of CDFs for large systems also were obtained by applying Horvitz-Thompson
estimation with selection probabilities being inversely related to station area. Data from all
base stations in the Virginian Province and from supplemental stations in the Delaware Bay
were used in this analysis. Areas of all stations in the Delaware Bay were assumed to be 70
km2 due to the inclusion of the supplemental stations. Areas for other large estuary base
stations were assumed to be 280 km2. The areas for the two lower Potomac River stations
included in the large estuary class were the areas of the corresponding tidal river segments.
Formulae for the CDF estimates and corresponding variances are analogous to those
presented for large tidal rivers.
2-35
-------
Estimates of CDFs for a particular geographic system within the province (e.g. Chesapeake
Bay system) were obtained by applying the above procedures to the small estuarine systems,
tidal rivers, and large estuaries sampled within that geographic system. Estimates of the
CDFs for the entire province or for a geographic system within the province were computed as
weighted averages of the relevant station class CDFs:
where
WS,WT,WL = relative areas for small systems, tidal rivers, and large estuaries,
respectively.
The variance of the estimate is:
In applying these procedures, variance estimation was based on the assumption of a fixed
sample size within each resource class. For large tidal rivers and large estuaries, the sample
size is a random element depending on the position of the sampling grid. This variance
component has not been incorporated into the estimation of variances of CDFs.
Estimating the percentage of fish with pathological disorders represents a different analytical
problem because it does not involve an estimate of area. The calculation was treated as two-
phase sampling. For each sample station, the calculated estimate of the proportion of fish
with pathologies was:
where
m
PI = the estimated proportion of fish pathology at station /'
m = number of species caught at station /
IVfc = n,3ln,
ni* = number of fish of species s caught at station /
n, = total number of fish caught at station /
2-36
-------
G-a = number of fish of species s that were examined for pathologies at station /
Xf,, = number of examined fish of species s that had pathologies at station /
Applying the estimation methods of two-phase sampling, the estimated variance of this
estimate was calculated by (Cochran 1977, eq. 12.24):
m
n,
For large estuaries, the estimate of the fish pathology rate was calculated by:
"L
A
B
where
nL = number of large estuaries sampled
n, = average catch per unit effort at station /
The estimated variance of the estimate was calculated by:
B
Az
Bz
AB
2-37
-------
where
n n
=EE
M />/
?l \mjfL\ +f HL
1 ( */ */ J M
M
The estimate of the proportion of fish pathology for large tidal rivers (tfT) and associated
variance ( Vi4/?(/^)) was calculated analogously to the estimation for large estuaries.
For small systems, the estimated proportion of fish pathology was estimated by:
where
= the estimate of fish pathology for small estuarine systems
ns = the number of small estuarine systems sampled
R, = A,n,
Aj = area of system /
R =
M
The estimated variance of the estimate (Cochran 1977, eq. 11.30) was calculated by:
n.
n.
R*
n-1
2-38
-------
Provincewide estimates for proportion of fish pathology were calculated by weighted averages
of small estuarine systems, large tidal rivers, and large estuaries estimates.
2-39
-------
-------
SECTION 3
EVALUATION OF LOGISTICAL FEASIBILITY AND
QUALITY ASSURANCE
One of the primary objectives of the 1990 Demonstration Project was to evaluate the
feasibility of collecting, within a limited sampling period, the kinds and volume of data required
to produce a regional evaluation of the status of estuarine ecosystems in the Virginian
Province. This section uses the results of the 1990 Demonstration Project to address two
fundamental questions pertaining to logistics and quality assurance for future EMAP
monitoring in estuaries:
• Could the data required for developing regional status estimates be collected with
the level of effort and sampling methods employed in the Demonstration Project?
• Does the sampling plan ensure collection of data that satisfy criteria defining
suitability for the intended use (i.e., quality assurance)?
The section also identifies major logistical impediments encountered during the 1990
Demonstration Project and provides recommendations for improvement in future years.
Logistical successes and failures are identified from both the sample collection and processing
perspectives for each indicator.
Only results from the second and third sampling intervals (19 July to 23 September) were
considered in evaluating the logistical feasibility of the EMAP sampling plan. Together, these
two intervals represent a "typical" sampling effort for future years of EMAP monitoring in the
Virginian Province in terms of duration, the skill and experience of the sampling crews, and
the types and number of expected station visits (including an allowance for station revisits, if
necessary). Although it produced some usable data, the first sampling interval during the
1990 Demonstration Project was used primarily for additional crew training and refinement of
procedures. As a result, only about half of the site visits originally scheduled for interval 1
were completed successfully. Including this first sampling interval in the evaluation of
logistical feasibility would needlessly underestimate the effectiveness of the logistical plan for
use in future implementation of the program, when sampling procedures will be fully tested
and refined, and crews will be fully equipped and experienced.
In general, the field crews succeeded in obtaining the required data within the allotted time.
The overall effectiveness of the 1990 sampling plan is reflected in the high percentage of
stations for which usable data were obtained for all parameters (Table 3-1). The number of
stations with usable data is sufficient for addressing the major objectives of the 1990
Demonstration Project, including the evaluations of sampling design, indicator
3-1
-------
Table 3-1. Status of sample collection during the 1990 Virginian Province Demonstration
Project
Parameter
Expected
Number of
Stations
Sampled
Percent of Stations*8'
Collected
Received0"
Analyzed
Passed
QA
Water Quality:
CTD Profiles
Suspended Solids
Chlorophyll a
Water Column Toxicity
Clostridlum
Datasonde Deployments
217
23
217
23
160
30
99
100
95
100
95
93
96
100
95
100
95
90
96
100
0
100
93
90
86
100
0
100
93
73
Sediment:
Inorganic Chemistry
Organic Chemistry
Grain Size
Sediment Toxicity
160
160
160
160
95
95
96
96
95
95
94
95
93
93
93
91
93(c)
73(0)
93
88
Benthic:
Infaunal Assemblages
Silt-clay Content
Sediment Profile Images
217
217
20
96
96
85
96
95
85
96
95
85
96
95
85
Fish:
Assemblages/Pathology
Chemistry
160
160
89
70
89
69(d)
89
89
Bivalve:
Chemistry
126
29
29
10 !
10
w Values represent the percent of stations sampled in intervals 2 or 3 for which the various steps
were completed successfully.
(b> This accounts for samples lost at collection or during shipment.
<°» Based on expected number of analyte results rather than expected number of stations.
-------
suitability, and estuarine status. Although data collection rates were uniformly high, there
were some notable QA/QC deficiencies in the field sampling component of the Demonstration
Project. The following sections offer detailed descriptions of these deficiencies and provide
recommendations for overcoming them in future years.
It is noteworthy that in over 5,000 man-hours of field activity in three months, there were no
serious injuries or significant losses of equipment. Overall, these results confirm that small
boats (24 ft) can be used to collect samples in estuaries safely and effectively. It must be
stressed that there was considerable emphasis on safety procedures during crew training
(including first aid and CPR) and throughout field operations; furthermore, there was an
appropriate allowance in the schedule for inclement weather and equipment malfunction, and
back-up equipment was available at all times throughout the 1990 Demonstration Project.
3.1 WATER QUALITY PARAMETERS
Both the vertical profile and continuous near-bottom water quality measurements were
recorded internally, using the Seabird CTD and Hydrolab DataSonde 3, respectively; the
resultant electronic files were downloaded, stored, and transferred using the field computers.
A major logistical objective was to determine the feasibility of deploying and retrieving moored
Hydrolabs on a 10-day cycle, with the ultimate goal of obtaining continuous 70-day records at
30 sites throughout the province.
The field crews successfully retrieved approximately 90% (110 of 122) of the Hydrolabs
deployed throughout the 1990 Demonstration Project. Despite deployment of units at a wide
variety of locations, only eight were lost permanently. Several others were displaced from
their mooring sites but were found later or retrieved by divers. Units that were lost
permanently were generally from a few high-traffic sites.
The high success rate in retrieving deployed Hydrolabs and the general reliability and ease of
operation of the instruments indicate the overall feasibility of using Hydrolabs to obtain
continuous near-bottom data in future years. Some Hydrolab records were lost, however, as a
result of internal .electronic problems or failure to communicate with the field computers. Field
QC .checks and subsequent review of the continuous records suggested that biological fouling
of the dissolved oxygen sensors, which typically occurred several days after deployment,
frequently diminished the accuracy of readings. This suggests the need to shorten the
deployment interval in future years.
CTD .files were received successfully from .more than 90% of stations (Table 3-1). At less
than 10% of stations, CTD files were not available for finarQA/QC review due either to failure
to deploy the instrument, or inability to download data because of intermittent electronic or
field computer software problems. Diskettes containing CTD files from a small percentage of
3-3
-------
stations were lost during shipment. These data could not be recovered because the files were
deleted from the field computers after the diskettes were shipped. A greater level of effort
devoted to troubleshooting and debugging the computer software and better crew training prior
to field operations should result in lower rates of data loss in future years.
The original plan for QA/QC review of the CTD profile data was to compare daily Winkler DO
measurements against the simultaneous CTD reading as a check on the accuracy of the
latter. This check was unsuitable for several reasons. First, there is considerable doubt about
the accuracy of the Winkler titrations performed by the field crews due to insufficient training in
this technique. Second, there is evidence that exposure to excessive heat caused the Winkler
chemicals to become unstable as the summer progressed. Third, and most important, the
Winkler accuracy check was performed only once a day while the CTD was at the surface;
therefore, its ability to detect inaccurate readings during deployment of the instrument on
station was limited.
The overall accuracy of each of the Seabird CTDs was confirmed in a laboratory calibration
check immediately following the completion of the 1990 sampling effort. In the absence of a
reliable, daily, field QC check, the vertical plot of each parameter in the CTD data files (e.g.,
temperature, salinity, pH, dissolved oxygen, PAR, percent transmission, and chlorophyll a
fluorescence) was reviewed visually for consistency between descending and ascending
profiles. Rejection of the water column profile data from 20 stations (Table 3-1) is attributed to
problems with physical deployment of the CTDs at specific stations. At most of these stations,
sharp, anomalous drops in both dissolved oxygen and salinity, and lack of agreement between
the downcast and upcast profiles suggested that sediments drawn into the CTD pump system
upon contact with the bottom may have clogged it completely. Anomalous dissolved oxygen
and salinity readings helped to identify this problem because only these two sensors require
flowing water to obtain proper readings. The manual review of CTD data resulted in rejection
of profiles that 1) indicated that field crews did not allow enough time for the dissolved oxygen
sensor to equilibrate at depth before taking measurements, or 2) exhibited improbable (e.g.,
supersaturated) or unstable readings as a result of intermittent electronic problems.
Several recommendations could improve the success rate for obtaining usable vertical profiles
with the Seabird CTD in the future. First, to prevent clogging problems, field crews should
avoid letting the instrument contact the bottom during deployment. Second, a field QC check
should be done at every station and should provide an independent measure of bottom
dissolved oxygen concentration for comparison with the CTD reading. A Nisken bottle could
be used to obtain a bottom water sample from which to measure dissolved oxygen
concentration using either Winkler titration or a second instrument (e.g., a hand-held YSI
meter). This check would reveal inaccurate readings associated with improper deployment of
the CTD and provide instantaneous confirmation of acceptable vertical profiles. Third, field
crews should undergo more intensive training and be given greater responsibility for
determining the acceptability of CTD profiles on station. This would provide greater
3-4
-------
opportunity for the crews to redeploy the instrument if an initial cast is suspected to be
unacceptable. Finally, back-up instruments should always be available to replace field
instruments in need of service or recalibration.
Usable data on the concentration of suspended solids and water toxicity were obtained at
most stations despite the stringent preservation requirements (shipment below 4°C) and
narrow holding-time limits (less than 48 hours after sample collection for water toxicity). In
contrast, none of the chlorophyll a samples were suitable for processing. The filters that
arrived at the laboratory were either thawed or too wet due to inadequate training of field
crews in use of the simple syringe filtering apparatus. Measurement of chlorophyll a concen-
tration will be reassessed for its suitability as an indicator for estuaries.
3.2 SEDIMENT QUALITY AND BENTHIC COMMUNITY PARAMETERS
Sediment samples for sediment quality and benthic community analyses were obtained
successfully at over 95% of the scheduled stations (Table 3-1). The roughly 5% of sites
where samples could not be obtained were characterized by rocks or cobbles; the sampling
gear used in the Demonstration Project is not designed to work in these bottom types.
Chemistry, grain size, and toxicity samples from an additional 2% to 5% of the stations either
were lost in shipping or could not be processed upon receipt at the laboratory due to container,
breakage. Sturdier containers will be used in the future, particularly for the sediment toxicity
and chemistry samples; furthermore, greater emphasis will be placed on instructing field crews
in packing techniques to avoid breakage.
Sediment profile images were collected from 20 stations in the northern half of the Virginian
Province. Organism-Sediment Index (OSI) values could not be calculated from data collected
at three of the stations (69, 77, 93) because of poor camera penetration. At Station 69, the
sand content was greater than 75%, and the surface dwelling gastropod Crepidula was one of
the dominant forms. No grain size data are available for Station 77; sand content at Station
93 was greater than 95%. OSI data were collected from the remaining four high sand content
(greater than 75%) stations in this survey. It appears from these data that sediment profile
imaging may be restricted at sites that have very high sand content or heterogeneous,
unconsolidated deposits.
Average laboratory error rates of less than 4% were achieved for sediment grain size and
benthic analyses (including sorting, species identification and enumeration, and biomass).
These error rates represent the cumulative results of a series of internal QA/QC checks
designed to ensure consistent production of usable data. Error rates above acceptable limits
triggered a series of corrective actions, including re-analysis of samples. Because the overall
laboratory processing error rate was low for the sediment grain size and benthic analyses, the
3-5
-------
final data were deemed usable for a high percentage of the stations from which samples were
collected.
Acceptability of the sediment chemistry data depended upon the ability of the laboratory to
perform the analyses within pre-established control limits. For the inorganic analyses, the
laboratory generally was able to meet control limits for the required QA/QC samples (e.g.,
calibration check samples, blanks, matrix spikes). Laboratory results for Standard Reference
Materials (SRMs), which were analyzed along with every sample batch as a QC check on
both accuracy and precision, are presented in Table 3-2. These results indicated acceptable
recoveries for the metals analyzed by inductively coupled plasma-atomic emission
spectrometry (ICP-AES; Al, Cr, Cu, Fe, Mn, Ni, Pb, Zn) and acceptable, although somewhat
lower and more variable, recoveries for the metals analyzed by graphite furnace atomic
adsorption (GFAA; As, Cd, Sb, Se, Sn). The increased variability in GFAA analyses was due
both to the lower concentrations of these metals in the SRM and the less rigorous digestion
procedure used; therefore, some of the metal concentrations determined by GFAA may
slightly under- or overestimate the true amount present in the sample. Silver was not
detected in most of the samples; however, the laboratory's detection limit of 1 ppm was above
the EMAP-E target detection limit of 0.1 ppm for this metal.
The expected number of analyte results for inorganics was 2235, based on analysis of 149
samples for 15 different metals. Because results for QA/QC samples generally fell within the
control limits established for the program, a high percentage (93%) of the associated metals
data was deemed to have "passed QA" (Table 3-1) and was used to evaluate sediment
contamination in the Virginian Province. Only results for silver failed to meet QA criteria, due
to the laboratory's inability to achieve the required detection limit for this metal.
The laboratory that analyzed the 1990 sediment samples experienced some early difficulties
with the organic analyses, however, the final QA/QC results for the SRM (Table 3-3)
suggested that the analyses were performed within acceptable control limits for accuracy and
precision. Very high and variable rates of recovery were experienced for the pesticides
heptachlor epoxide, cis-chlordane, trans-nonachlor, and 4,4'-DDT, which were present in the
SRM at concentrations very close to the laboratory's detection limit. Such a high degree of
uncertainty and variability is to be expected from any laboratory trying to quantify organic
analytes at concentrations close to the detection limit in a complex matrix like SRM 1941 (i.e.,
Baltimore Harbor sediment). As analyte concentrations increase above the detection limit,
there is a concomitant increase in a laboratory's ability to detect and quantify the compounds
accurately (Keith 1991). This is reflected in the SRM 1941 results for 1990: good recoveries
were achieved for PAHs and PCB congeners with high SRM concentrations relative to the
laboratory's detection limit. Based on these results, it is reasonable to conclude that the 1990
organics data is reliable for samples having similarly elevated organic contaminant
concentrations.
3-6
-------
Table 3-2. Results for SRM 2704 (Buffalo River Sediment) used as a set control for the
1 990 EMAP-E sediment inorganic analyses.
Element
ICP-AES Metals (n
= 18 analysis sets
or "batches")
Al
Cr
Cu
Fe
Mn
Ni
Pb
Zn
GFAA Metals (n =
1 8 analysis sets)
As
Cd
Sb
Se
Sn
Average00
96
87
96
88
94
90
92
96
78
100
79
97
80
Stdv(b)
1.8
2.7
2.1
1.6
2.1
5.5
4.3
1.5
4.1
6.8
11.4
12.1
29.4
CV(C)
1.9
3.1
2.2
1.8
2.2
6.2
4.6
1.6
5.3
6.8
14.4
12.5
36.7
(a) Average percent recovery relative to the SRM certified value
(b) Standard deviation of the percent recovery values
(c) Coefficient of variation of the percent recovery values
The laboratory's failure to meet the target detection limits consistently is a major deficiency in
the organics data sets (Table 3-4). The analytical method resulted in high detection limits,
and detection limits varied because the laboratory analyzed a different amount (i.e., dry
weight) of sediment from each sample. As a result, the target analytes were not detected in a
large number of samples, and the "calculated" detection limits (i.e., the theoretical
concentration of each analyte necessary for detection) differed significantly from sample to
sample (Table 3-4).
3-7
-------
Table 3-3. Results for SRM 1941 (Baltimore Harbor sediments) used as a set control
for the 1990 EMAP-E sediment organic analyses.
Compound(a)
PAHs (n = 19 analysis sets or "batches")
Phenanthrene
Anthracene
Fluoranthene
Pyrene
Benz{a]anthracene
Benzo[b+k]fluoranthene
Benzo[a]pyrene
Perylene
Benzo[ghi]perylene
lndeno{1 ,2,3-cd]pyrene
PCBs/pestlcldes (n = 15 analysis sets)"
PCS 18
PCS 28
PCB52
PCS 66
PCB 101
PCB118
PCB 153
PCB 105
PCB 138
PCB 187
PCB 180
PCB 170
PCB 195*
PCB 206*
PCB 209
Heptachlor epoxide*
Alpha-Chlordane*
Trans-Nonachlor*
p.p'-DDE
p.p'-DDD
p,p'-DDT*
Average(b)
95.8
69.1
97.7
85.0
93.8
101.9
63.1
62.3
84.9
120.8
80.6
53.5
96.2
66.2
69.2
99.3
95.2
91.0
72.1
78.5
89.5
81.6
135.2
97.2
88.1
238.0
304.0
547.0
97.3
86.2
202.0
Stdv(c)
20.6
17.1
22.8
18.0
21.6
17.6
16.3
16.8
23.3
29.5
14.1
11.3
23.4
11.6
15.4
15.4
15.3
18.9
15.4
17.7
17.6
22.8
39.1
29.7
19.9
78.3
55.3
835.0
31.3
23.2
141.0
cv(d)
21.5
24.7
23.3
21.2
23.0
17.3
25.8
27.0
27.4
24.4
17.5
21.1
24.3
17.5
22.2
15.5
16.1
20.8
21.3
22.5
19.7
27.9
28.9
30.5
22.5
32.9
18.1
152.0
32.2
26.9
69.8
w SRM 1941 has certified concentrations for only a subset of the PAH compounds of interest to EMAP-E.
w Average percent recovery relative to the SRM certified value
w Standard deviation of the percent recovery values
<
-------
Table 3-4. Range in detection limits (in ng/g dry weight) reported for organic
compounds in 1 990 sediment samples. The target detection limits were 1 0
ng/g for each PAH compound and 0.5 ng/g for each PCB congener and
pesticide.
Polycyclic Aromatic Hydrocarbons
(PAHs)
Acenaphthene
Anthracene
Benz(a)anthracene
Benzo(a)pyrene
Benzo(e)pyrene
Biphenyl
Chrysene
Dibenz(a,h)anthracene
2,6-dimethylnaphthalene
Fluoranthene
Fluorene
2-methylnaphthalene
1 -methylnaphthalene
1 -methylphenanthrene
Naphthalene
Perylene
Phenanthrene
Pyrene
Benzo(b+k)fluoranthene
Acenaphthlylene
Benzo(g,h,i)perylene
ldeno(1 ,2,3-c,d)pyrene
2,3,5-trimethylnaphthalene
DDT and its metabolites
o,p'-DDD
p,p'-DDD
o,p;-DDE
P)P'-DDE
o,p'-DDT
p,p'-DDT
Minimum
21
17
17
23
23
23
22
24
24
16
25
25
23
13
30
27
16
15
22
22
31
26
23
0.13
0.12
0.10
0.04
0.12
0.18
Maximum
207
121
72
151
153
150
72
252
156
114
176
162
150
86
54
189
44
39
145
212
325
249
219
1.93
6.10
1.11
0.45
' 1.26
3.22
Median
34
28
28
38
37
36
35
43
38
24
43
39
34
21
39
46
26
22
33
38
55
43
38
0.24
0.20
0.18
0.07
0.22
0.58
3-9
-------
Table 3-4. Continued
Chlorinated pesticides other than DDT
Aldrin
Alpha-Chlordane
Trans-Nonachlor
Dieldrin
Heptachlor
Heptachlor epoxide
Hexachlorobenzene
Undane (gamma-BHC)
Mirex
18 PCB Congeners
PCB08
PCB 18
PCB 28
PCB 44
PCB 52
PCB 66
PCB 101
PCB 105
PCB 118
PCB 128
PCB 138
PCB 153
PCB 170
PCB 180
PCB 187
PCB 195
PCB 206
PCB 209
Minimum
0.10
0.09
0.04
0.04
0.10
0.08
0.03
0.16
0.03
0.08
0.37
0.08
0.06
0.11
0.09
0.12
0.07
0.06
0.12
0.11
0.11
0.09
0.11
0.08
0.10
0.10
0.12
Maximum
1.78
1.16
0.87
0.52
1.47
1.85
7.23
27.5
1.9.3
4.46
5.89
1.03
1.50
2.70
1.01
1.39
0.60
0.65
1.62
1.31
1.03
2.15
1.30
0.72
1.23
1.38
1.09
Median
0.27
0.19
0.07
0.08
0.19
0.19
0.09
0.64
0.08
0.63
0.94
0.17
0.17
0.38
0.18
0.20
0.14
0.12
0.23
0.18
0.19
0.32
0.19
0.13
0.19
0.20
0.20
The expected number of results for organics was 8344, based on analysis of 149 samples for
23 PAH compounds, 15 pesticides, and 18 PCB congeners. Any results reported as "not
detected" in samples for which the laboratory failed to meet the EMAP-E target detection limits
(i.e., 10 ppb for individual PAHs, 0.5 ppb for individual pesticides/PCBs) were considered to
have "failed" QA; therefore, the overall success rate for the organics analyses was 73% (Table
3-1). If the target detection limits had been achieved and consistent sample sizes had been
used, quantifiable amounts of the organic analytes of interest probably would have been
3-10
-------
detected in most of the 1990 samples, and a higher percentage of results would have passed
QA.
Concentrations near the program detection limits represent the "background" levels of many of
the EMAP-E organic contaminants in the estuarine sediments of the Virginian Province. The
difficulty encountered in consistently meeting the target detection limits impairs our ability to
make quantitative evaluations of the percent of bottom area in the Virginian Province that is
contaminated with organic compounds. Chemicals may have been present that were not
detected and quantified. This limits the comparability of the 1990 chemistry data with other
data sets for which lower detection limits were achieved.
A number of corrective actions and changes in methodology have been implemented by the
laboratory responsible for the 1990 analyses. For example, future analyses will be performed
using a constant dry weight of sample, which should result in a consistent detection limit. The
1990 experience helps to illustrate some of the difficulties inherent in performing low-level (i.e.,
low parts-per-billion) analyses of organic contaminants in complex samples from estuarine
environments.
Of the 160 sediment toxicity samples scheduled, 146 were processed. Results from 35 of
these samples are qualified (i.e., "flagged") in the database because average survival of the
test organism (the amphipod Ampelisca abdita) in the controls for these stations was less than
the pre-established limit of 90%. In 30 of these qualified results, however, control survival
was greater than 86%, and they were used in the preliminary evaluation. When the control
survival requirement was reduced to 85%, only five stations (i.e., less than 4% of the stations
tested) were unusable due to unacceptable control survival. The original 90% survival
criterion, which was adopted from the ASTM (1991) procedure for the amphipod Rhepoxynius
abronius, may not be achievable for all species, especially those for which a large database is
being formed only now. For example, the ASTM control survival requirement for Hyalella is
80%. Data loss can be avoided to an even greater extent in the future by collecting enough
sediment to allow for retesting in the event of unacceptable control survival; however, even
without this additional sample effort, the 1990 Demonstration Project results indicate that the
Ampelisca test can provide usable sediment toxicity data for a high percentage of stations.
Clostridum spore concentrations in sediments were determined from subsamples taken from
the same homogenized sediment samples used for the determination of sediment contaminant
concentrations and sediment toxicity. Subsamples were collected when sediment samples
were thawed for extraction of contaminants. Established methods for Clostridium spore
counts require conducting assays on the day samples are thawed; however, because of the
number of sediment samples scheduled for contaminant extraction each day, not all
subsamples for Clostridum counts could be processed on the day they were thawed.
Recognizing this problem, laboratory personnel obtained some Clostridium samples by
removing chips of sediment from samples prior to thawing and keeping them frozen until they
could be processed. The time delay prior to processing some thawed Clostridium samples
and analysis of subsamples that had not been homogenized with the whole sample after
3-11
-------
thawing (i.e., frozen chips) both represent departures from established methods. Samples so
processed were flagged with appropriate quality assurance codes in the database.
Additional analyses were conducted to determine how these departures from established
procedures influenced spore counts. Tests demonstrated that storing thawed samples for up
to four days did not significantly reduce Clostridium spore concentrations. Since no thawed
samples were stored for more than four days, the time delay probably did not influence spore
counts. The additional analyses demonstrated that the frozen sediment chips contained about
43% fewer spores than subsamples drawn from sediments that were homogenized after
thawing. Spore concentrations were determined from frozen sediment chips for 25 of 177
samples. Doubling spore concentrations to correct for using frozen sediment chips would
have changed the interpretation for only one of those 25 samples; consequently, no
corrections were made in the data set, and all samples were used in the evaluation presented
in Section 2.
3.3 FISH AND BIVALVE SAMPLING
A successful trawl was defined as one in which the crew was able to deploy and "fish" the 16-
m (footrope length) high-rise trawl net for 10 ± 2 minutes, regardless of whether fish were
caught. This condition was met at 143 (89%) of the 160 stations where fish trawls were
scheduled (Table 3-1). Fish were caught at 140 (98%) of the 143 stations where trawls were
successful. These high percentages indicate the logistical feasibility of using the chosen
equipment and standard trawl duration to catch fish at a variety of estuarine locations.
Despite the high rate of fish sampling success, the Demonstration Project identified some
limitations of the chosen methodology. Trawls were not even attempted at 15 stations. This
includes 11 sites that were too deep (greater than 80 ft) for proper deployment of the gear, 2
shallow sites that lacked sufficient room to maneuver the boat, and 2 sites where equipment
failure prevented deployment of the net. At the remaining two stations where trawls were
unsuccessful, trawling was attempted but had to be aborted because the net either filled with
weeds or got entangled on an underwater obstruction.
Target fish species were collected and saved for chemical analyses at 112 (78%) of the 143
stations where trawls were successful; this constitutes 70% of the stations where trawls were
scheduled (Table 3-1). At 13 of the sites where no target fish were caught, only one trawl
was performed. The success rate for capturing target fish could improve in the future if
additional sampling effort is expended at such sites. This might involve doing an additional
trawl right away or returning to the site on another day. At an additional 13 sites, the field
crews caught only scup or butterfish. These species initially were included among the target
fish but were eliminated later to reduce handling time on the boat. They have since been
reinstated as target species, which should result in a higher overall capture rate for target
species in the future.
3-12
-------
Some of the target fish samples, representing collections from ten stations, were lost because
they were not saved or were not shipped to the laboratories. The entire collection of target
species was lost at only two of these stations. Improvements in the sample tracking system
and in crew training have been developed to avoid this loss in the future. The 110 stations
(69% of the anticipated stations; Table 3-1) for which chemistry samples were received by the
analytical laboratory would have been sufficient for assessing the feasibility of the fish
contaminant indicator; however, the fish tissue samples were scheduled for analysis following
the completion of the sediment samples, and delays in finalizing the sediment organic
analyses resulted in exceeding the recommended holding time for the fish tissue. For this
reason, fish contaminant analyses were not performed on the 1990 samples.
Besides saving fish for chemical analyses, the field crews recorded species composition and
abundance for each trawl, measured the length of each individual, and examined each fish for
visible external pathologies. Approximately 150 specimens, representing 48 species, were
sent to the laboratory for independent verification of the field identification. The laboratory
taxonomic expert had to identify 11 of the 150 specimens because they were not identified by
the field crews. Nine specimens were identified incorrectly by the field crews; thus, 130 of the
150 (87%) individuals received for taxonomic verification were identified correctly by the field
crews.
Some of the fish that either were not identified or were misidentified in the field were
uncommon or exotic species (e.g., blue runner, striped cusk eel, spotted eagle ray) of which
only a few representatives were collected throughout the summer. The crews consistently
identified the most abundant species correctly, including the target species. The only
exceptions to this were identifications of catfish (Family Ictaluridae) and clupeids (Family
Clupeidae, including menhaden, herring and anchovies). These misidentifications were
infrequent and limited to within-family errors (e.g., blue catfish identified as white catfish,
alewife identified as Atlantic herring). These results will be used to refine and focus crew
training in future years.
At 30 of the 140 stations where fish were collected, at least one fish identified by the field
crews as having one or more visible external pathologies was sent to the EPA's Gulf Breeze
laboratory for confirmation of the pathologies. A total of 572 fish identified as having no
visible external pathologies (i.e., reference fish) were collected at an additional 22 stations and
sent to Gulf Breeze. Upon examination by the Gulf Breeze specialists, only 8 (4.1%) of the
197 fish identified by the field crews as having at least one pathology of any type were found
to lack the condition. This low overall error rate suggests that field crews were successful in
identifying "diseased" fish (i.e., fish having one or more pathologies).
The field crews also were reasonably successful in identifying healthy fish; only 54 (9.4%) of
the 572 reference fish were found by the Gulf Breeze specialists to have one or more
pathologies. The presence of parasites in the branchial chambers of reference fish was the
pathology most frequently missed in the field. A microscope normally is required to detect
branchial parasites; therefore, field crews cannot be held accountable for these errors. The
3-13
-------
field crews were negligent, however, in identifying fin erosion on 7 reference fish and gill
erosion on 10. These results will be used to focus training efforts in future years.
Although field crews generally were successful in identifying both diseased (n = 197) and
healthy fish (n = 572), an analysis of the QC data for all fish (n = 769) on the basis of
individual pathologies revealed relatively high error rates for certain disorders. Two types of
identification error were possible in the field: 1) identification of pathologies that were not
confirmed by the Gulf Breeze experts (false positives), and 2) failure to identify pathologies
that were identified at Gulf Breeze (false negatives). As indicated previously, the highest rate
of false negatives occurred for branchial chamber parasites, for which the field technicians
were not responsible. Pathologies that should have been identified but were missed
frequently by the field crews included body lumps/bumps, skin discoloration, and erosion of
fins and gills. Interviews with crew members suggested that crews became less diligent about
looking for additional abnormalities after detecting a single pathology. False negatives,
therefore, reflected a lack of effort on the part of field technicians rather than an inability to
recognize common pathologies. In the future, crews will be trained to identify all pathologies
on each fish. The field technicians also exhibited relatively high rates of false positive
identifications for several pathologies, including body ulcerations, fin and gill erosion, skeletal
malformation, and microphthalmia. Most of these misidentifications occurred on fish that,
according to the Gulf Breeze experts, exhibited a different pathology than the one identified in
the field.
Even though the identification of specific pathologies was often incorrect, the field crews
achieved a low overall error rate for identification of "diseased" fish (i.e., fish exhibiting one or
more pathologies of any kind). The high rate of false positives probably reflects insufficient
training of field technicians in distinguishing pathologies. The 1990 QC results will be used to
improve training in future years. Although the 1990 data were useful for making reliable
estimates of the number of diseased fish in the Virginian Province, it was not possible to
quantify the prevalence of specific pathologies with certainty.
The rocking chair dredge was deployed successfully at 104 of the 126 scheduled stations
(Table 3-1) in an attempt to collect large (greater than 2.5 cm) bivalves. Although this
indicates the logistical feasibility of deploying the equipment, the overall success rate for
collecting bivalves was relatively low. Bivalves were collected and saved for chemical
analysis at only 36 (29%) of the expected stations. Often, no bivalves were collected in areas
where they were expected to occur. Although the dredge appeared to be operating properly,
it often became clogged with mud and was ineffective. Extensive modifications of this gear
are needed if collecting bivalves continues to be of interest to the program in future years.
Sufficient biomass for chemical analysis of bivalve tissue was obtained from only 13 (10%) of
the expected number of stations. Although these samples were analyzed, the data were
insufficient to assess the utility of bivalve tissue contamination as an indicator, and it could not
be used in the assessment of estuarine condition.
3-14
-------
SECTION 4
INDICATOR DEVELOPMENT AND TESTING
EMAP monitoring focuses on four types of ecological indicators (Hunsaker and Carpenter
1990): response, exposure, habitat, and stressor (Fig. 2-4). Response indicators are
ecological characteristics that integrate the responses of living resources to individual
pollutants or combinations of pollutants and other stresses and provide quantitative evidence
of the condition of ecological resources (Messer 1990). Exposure indicators quantify pollutant
exposure and habitat degradation and are used mainly to identify probable causes of poor
environmental quality that may explain observed responses of ecological indicators. Exposure
indicators and response indicators are used to describe environmental status. Habitat
indicators provide basic information about the natural environmental setting and are used to
normalize exposure and response indicators to natural environmental gradients. Stressor
indicators are used to quantify pollution inputs and stresses and identify the likely sources of
pollution exposure.
This approach differs from many historical monitoring and assessment programs in that it
emphasizes biological indicators for conducting ecological assessments (Hunsaker and
Carpenter 1990). The biological indicators are selected based on their known responsiveness
to anthropogenic pollutants, habitat modifications caused by human activities, and other
influences believed to be causing degradation. This ecologically-based approach is
emphasized in EMAP because it can be applied to situations where multiple stressors are
acting separately or in combination and where natural processes affecting pollution exposure
can not be modeled easily (e.g., the bioavailability of contaminants in sediments). This is
certainly the case in estuarine systems, which are subject to an array of anthropogenic inputs
and exhibit great biotic diversity and complex physical, chemical, and biological interactions.
Using an ecologically-based approach may also reveal the effects of emerging environmental
problems, even when causal relationships are poorly understood.
An important component of the 1990 Virginian Province Demonstration Project was an attempt
to develop response indicators that can be used to discriminate between polluted and
unpolluted environments on a regional scale. Although establishing biocriteria has been
identified as an Agency priority, biotic indicators of ecological condition have not yet been
developed for estuaries. The major reason is the lack of sufficient historical information with
which to calibrate and verify the reliability of candidate response indicators over the large
geographic and habitat gradients (e.g., salinity) that characterize estuaries. Rather than
developing generic response indicators to estimate the degree of degradation, estuarine
ecologists generally have relied on site-specific knowledge about differences in the kinds and
abundances of organisms between polluted and unpolluted habitats and expert opinion to
define the extent and severity of degradation. Because different kinds of organisms occur in
different parts of a province, this approach is not appropriate on regional scales.
4-1
-------
In contrast, methods for measuring the exposure variables selected by EMAP are well-
developed and have been used in a number of monitoring programs throughout the province;
however, some of the exposure variables, such as sediment toxicity, have not been applied on
a regional scale. Others, such as dissolved oxygen, have been measured on regional scales
but never using a probability-based approach to develop regional status estimates of the kind
being made in EMAP.
Recognizing that measuring indicators on regional and national levels differs from the way
such information has been used in many previous programs, EMAP developed a strategy to
select and incorporate indicators into the program (Knapp et al. 1990; Holland 1990). In this
strategy, potential indicators are incorporated into the measurement program if they meet a
series of evaluation criteria that apply to all resource groups in EMAP (Hunsaker and
Carpenter 1990). This section describes analysis of data collected during the 1990
Demonstration Project to determine how well some of the potential indicators for estuaries met
the EMAP evaluation criteria.
4.1 RESPONSE INDICATORS
As previously described, response indicators are characteristics of the environment that
provide quantitative evidence of the status of ecological resources and the biological integrity
of the site at which they are measured (Messer 1990). Ecosystems with a high degree of
biotic integrity (i.e., healthy ecosystems) are composed of balanced populations of indigenous
organisms with species composition, diversity, and functional organization comparable to
natural habitats (Karr and Dudley 1981; Karr et al. 1986). Response indicators include
measurements of the kinds and abundances of biota present, the health of individual
organisms, and the sustainability of critical ecological processes. They are the empirical data
collected in EMAP that are integrated to measure the status and trends in the biological
integrity of ecological resources.
Two categories of response indicators were measured during the 1990 Demonstration Project:
(1) benthic response indicators, and (2) fish response indicators. Benthic organisms are
invertebrates that live in the sediments of aquatic habitats. In estuaries, they are a major link
between primary producers and higher trophic levels, including fish, shellfish, birds, and other
wildlife (Carriker 1967; Rhoads 1974). They are a particularly important source of food for
juvenile fish and crabs (Chao and Musick 1977; Bell and Coull 1978; Holland et al., 1989).
Estuarine benthos also play important roles in ecological processes that affect water quality
and productivity. For example, the feeding and burrowing activities of macrobenthos affect
sediment depositional processes and chemical transformations (Carriker 1967; Rhoads 1974;
Kemp and Boynton 1981). Benthic feeding activities also remove large amounts of paniculate
material from some shallow estuaries, which may improve water clarity (Cloern 1982; Officer
et al. 1982; Holland et al. 1989).
4-2
-------
Benthic assemblages have many attributes that make them reliable and sensitive indicators of
the ecological condition of estuarine environments (Carriker 1967). For example, most
macrobenthic species have limited mobility and cannot avoid exposure to anthropogenic or
natural stress. Benthos live in bottom sediments, where exposure to contaminants is highest.
Benthic assemblages are composed of a diverse array of species that respond to pollution
stress in a variety of ways. Some species are especially sensitive to exposure to pollution.
These species experience adverse effects due to this exposure, such as mortality, reduced
growth, and decreased reproduction. Others are tolerant of pollution, and respond by
increasing in abundance following exposure to pollutants. Because they generally depend
upon and interact with biological processes in the water column, the responses of benthic
organisms are representative of overall ecosystem responses (Rhoads 1974; Boesch and
Rosenberg 1981; Holland etal. 1988).
Fish have several advantages as potential indicators of estuarine condition. Because fish
have long life-spans and dominate the upper end of the food web, their responses integrate
many short-term and small-scale environmental perturbations. They are known to respond to
most of the major environmental problems of concern in estuaries (NOAA 1988), including
eutrophication, habitat modification, and the presence of pathogenic or toxic contaminants.
For example, eutrophication can affect fish adversely by diminishing dissolved oxygen
concentrations below critical levels for growth or survival. Habitat modification, such as the
loss of submerged aquatic vegetation, has been linked to decreased fish productivity through
loss of important nursery areas and shelters from predation. Toxic and pathogenic
contaminants can decrease fish growth, reproduction, or survival and can make fish unsafe for
human consumption.
Fish also are valuable as indicators because of their importance in determining public
perception of estuarine quality. Reports of fishery closures due to chemical and viral
contamination alarm the public. In addition, the public has an economic interest in estuarine
and coastal fisheries. More than seven billion dollars are spent annually in this country on
saltwater recreational fishing, the vast majority of which occurs in estuaries or within three
miles of the coast. Combining recreation and commercial fishing, estuaries and coastal
waters account for 70% of U.S. fisheries landings (NOAA 1987).
4-3
-------
4.1.1 General Methodology for Developing Response Indicators
The approach for developing response indicators consisted of five steps (Fig. 4-1):
1) developing a list of candidate measures for discriminating among sites of differing
environmental quality,
2) identifying a set of sample sites that could be categorized with confidence as
"nondegraded regional reference sites" or "degraded sites with known pollution
exposure,"
3) using discriminant analysis to identify combinations of candidate measures that
distinguished reliably between the degraded and nondegraded sites identified in
step 2,
4) validating the index, and
5) scaling the index from 0 to 10.
The activities that occurred in each step during planning for the 1990 Demonstration Project
are described below.
Step 1: Identify candidate measures
The first step consisted of two activities: (1) identifying candidate measures for differentiating
between nondegraded reference sites and degraded sites, and (2) characterizing and
removing variation in candidate measures associated with variation in natural habitat factors.
A list of candidate measures was developed by reviewing the relevant scientific literature and
conducting a series of workshops attended by ecologists knowledgeable about northeastern
U.S. estuaries and pollution assessment techniques. Measures that were applicable across
multiple habitats and broad regions (e.g., measures of species richness) were sought.
Measures were selected to represent the range of ecological attributes that characterize
estuarine assemblages, including biodiversity, abundance, species composition, trophic
interactions, and health of individual organisms. Measures that were applicable to a narrow
range of habitats (e.g., salinities) or particular subregions were not included on the list.
A dominant feature of estuarine environments is the large spatial variability in physicochemical
conditions that occurs over relatively small distances. For example,
4-4
-------
Step 1
identify Candidate Measures
• Define parameters to be
measured
• Adjust data to account for natural
habitat gradients
Step 2
Develop Test Data Set
• Regional reference sites
• Degraded sites
Step 3
Identify combinations of candidate
measures that discriminate between
reference and degraded test sites
• Univariate statistical analyses
• Stepwise discriminant analysis
Feedback loop
Optimize for Long-term
Performance Critieria
Step 4
Validate observed responses
using independent data
Step 5
Scale index from 0 to 10
Figure 4-1. General methodology used to evaluate potential response indicators.
4-5
-------
salinity variation that occurs from the headwaters of estuaries to their seaward boundary is a
major factor controlling the biodiversity and abundance of estuarine biota (Carriker 1967;
Boesch 1973; Lippson et al. 1979). Such large natural variation can obscure the responses of
candidate measure to pollution exposure and must be identified and controlled for before the
responses of candidate measures to pollution exposure can be characterized accurately.
Procedures for determining the influence of natural environmental factors on candidate
measures involved: 1) examining scatter plots of the distribution of candidate measures (Y
axis) against habitat factors, including salinity, water depth, sediment silt-clay content, and
latitude (X-axis); and 2) conducting linear or log-linear regressions to measure the magnitude,
direction, and significance of relationships identified from scatter plots. Candidate measures
were considered to be significantly influenced by habitat factors if the correlation coefficient
(r2) in regressions was greater than or equal to 0.25, and the slope was statistically different
from 0. Regressions with correlation coefficients less than 0.25 were considered biologically
insignificant. Data collected at all of the EMAP sites were used for these analyses.
Candidate measures were corrected for the influence of habitat factors by reformulating them
in a manner that was insensitive to habitat conditions (see Section 4.1.2)
Prior to performing regression analyses or other statistical examinations of the data, the
distribution of each candidate measure was tested for normality and homogeneity of variance.
Measures that were not normally distributed or had unequal variances were re-examined for
assumptions of normality after making logarithmic, arcsin, and square root transformations. In
all cases where adjustment was necessary, the Log10 (value +1) transformation was found to
be an appropriate transformation.
Step 2: Develop a test data set
Step 2 identified EMAP sites that could be used to determine which candidate measures and
what combinations of measures were most effective for discriminating among degraded and
nondegraded conditions. Ideally, nondegraded reference sites should have been established
by identifying pristine estuaries that are free of anthropogenic influence and selecting sampling
sites within them. Such estuaries, however, could not be identified in the highly populated
and heavily industrialized northeast United States. Instead, the reference sites were areas
believed to be "least impacted" on the basis of known exposures, and degraded sites were
those known to have substantial exposure to anthropogenic stresses. A preliminary list of
such sites was established during the planning process (Fig. 2-3) based on historical water
and sediment quality data, and the knowledge of local scientific experts.
Despite considerable effort to select appropriate test sites, the expected environmental
conditions were not verified by exposure measures at almost half of the preselected sites
(e.g., some supposedly nondegraded reference sites had high concentrations of contaminants
in sediment, and the sediment was toxic to biota in laboratory tests). The data collected at
4-6
-------
randomly selected base sites, however, provided a large database that was used for
identifying additional sites that could be included in the indicator testing program. These
additional sites ensured that the indicator testing sites represented the full range of
environmental conditions occurring in the Virginian Province. Inclusion of these additional
sites in the indicator testing data set also ensured that the number of samples was sufficient
to calibrate indicators with an acceptable level of precision.
Data for sediment contaminant concentrations, sediment toxicity, and dissolved oxygen
measurements (particularly the continuous measurements) were used to document
environmental conditions at indicator test sites. These data were used to select indicator test
sites representing the
• range of habitats that exists in the Virginian Province, including all major salinity
zones, sediment types, and biogeographical divisions;
• major categories of pollution stress, including contaminated sediments, stressful
low dissolved oxygen concentrations, and the cumulative impact of stressful low
dissolved oxygen concentrations and contaminated sediments; and
• relatively unpolluted reference sites that were exposed to little anthropogenic
perturbation.
To establish a matrix of test sites, sampling locations were classified as nondegraded regional
reference sites if:
1) summertime bottom DO measurements were never less than 1 ppm, 90% of the
dissolved oxygen observations were greater than 3 ppm, and 75% of the dissolved
oxygen observations were greater than 4 ppm;
2) no contaminant was apparent in the sediment at a concentration higher than that
identified by Long and Morgan (1990) as the median effects concentration (ER-M
value) for biological responses; and
3) survival of the amphipod Ampelisca abdita in the 10-day acute sediment bioassay
was greater than 75% and not significantly different from controls.
Sites were classified as degraded if:
1) dissolved oxygen concentrations below 0.3 ppm were recorded at the site during
the sampling period, or if 10% of the continuous bottom dissolved oxygen
observations was less than 1 ppm, or 20% of bottom dissolved oxygen
4-7
-------
observations was less than 2 ppm, or bottom dissolved oxygen observations of less
than 2 ppm occurred at the site for 24 consecutive hours; or
2) the concentration for at least one sediment contaminant exceeded the ER-M value
defined by Long and Morgan (1990), and the survival for a 10-day acute sediment
bioassay using the amphipod Ampelisca abdita was less than 75% of, and
significantly different from, control survival.
These criteria do not represent any EPA standards and are intended only for use in this
report.
Application of the above criteria identified a subset of sample sites that probably were
classified accurately as relatively undisturbed regional reference areas and degraded sites
with known pollution exposure. The number of sites identified as indicator test sites was,
however, considerably fewer than the total number of sites sampled. This is because of the
conservative nature of the criteria used to assign sites to categories. This conservative
procedure was necessary because of the poor scientific understanding of biotic responses to
intermediate levels of pollution stress.
Degradation resulting from habitat modification and other causes was not included in defining
indicator test sites. The data required to document the degree of habitat modification at a site
quantitatively are not available currently, and the resources were not available to collect these
data. The responses of candidate response indicators to habitat modification and other
causes of degradation were assumed to be similar to the responses of the pollution stresses
tested.
Regional reference sites were selected to calibrate the responses of indicators identified in
step 1 into an index that defines biological condition. Defining biological condition relative to
regional reference sites believed to contain nondegraded resources has a two important
implications with respect to interpretating the resulting index. First, degraded and nondegra-
ded were defined relative to present-day conditions. The conditions at present day reference
sites may not represent nondegraded conditions that existed decades or centuries ago.
Second, degradation determined by comparison to biological condition at reference sites does
not lend itself to distinguishing degradation attributable to human activity from degradation due
to natural processes. For example, the presence of low dissolved oxygen may cause a
biological response regardless of whether the low DO is due solely to the presence of deep,
highly stratified waters or results, in part, from anthropogenic enrichment of nutrients and
organic carbon. Determining the relative contribution of natural and anthropogenic factors
would require intensive, site-specific or system-specific studies of metabolic processes that
are beyond the present scope of EMAP.
4-8
-------
Step 3: Identify combinations of candidate measures that discriminate between
degraded and nondegraded areas
In Step 3, discriminant analysis was used to identify combinations of candidate measures from
the list generated in step 1 that reliably distinguished between the degraded and nondegraded
sites identified in step 2. Discriminant analysis provides an unbiased tool for selecting index
components from among the list of candidate measures. The approach also provides a
means for weighting component measures that is free from investigation bias. Finally, an
unbiased threshold value for distinguishing between sites having degraded and nondegraded
condition could be defined using discriminant analysis.
Prior to conducting discriminant analysis, however, a t-test was used to define the direction
and magnitude of differences for each candidate measure. Only candidate measures for
which there was a significant (p<0.05) difference in parameter values between reference areas
and degraded sites were included in the discriminant analysis. This procedure prevented
including measures with high signal to noise ratios. It also reduced the list of candidate
measures to a manageable number from which it was highly probable that a subset(s) could
be identified to discriminate reliably between reference and degraded areas.
In the stepwise discriminant analysis, a p-value of 0.15 was the criterion for including a
measure into the model. In addition, the direction of the coefficient for each measure included
was required to be consistent with the direction of the difference observed in the t-test. If the
direction of a measure was inconsistent between the t-test and discriminant analysis, the
stepwise procedure was repeated after removing the measure in question. This procedure
reduced the possibility of identifying a combination of measures that optimized only for the
calibration data set. Canonical discriminant analysis was used to determine the direction and
magnitude of coefficients.
Many of the candidate measures contain redundant information (i.e., are highly correlated);
thus, although the combination of measures selected by discriminant analysis maximizes
separation between degraded and reference sites, it may fail to incorporate easily interpreted
and highly valued measures (e.g., species richness) that also would be useful for discriminat-
ing among sites of differing quality. For this reason, the discriminant procedures were applied
twice. The first analysis was conducted allowing the analysis procedure to include measures
only on the basis of the degree to which they differentiated between degraded environments
and nondegraded regional reference areas. In the second discriminant analysis, selected
measures were forced into the model. These measures were not included in the first model,
but were reported in the scientific literature to discriminate between degraded and nondegra-
ded environments (e.g., species richness). Further, the F-values and partial r2 from the first
analysis indicated these forced measures were highly correlated with measures that were
included. This iterative process identified several combinations of candidate measures that
distinguished between degraded and nondegraded sites. For each of these models, canonical
4-9
-------
discriminant analysis was used to calculate the frequency with which reference sites were
incorrectly classified as degraded (i.e., false positive), and the frequency with which degraded
sites were classified as reference areas (i.e., false negatives).
As a means for maximizing discrimination among two-predefined groups, stepwise discrimi-
nant analysis is sensitive to the incorrect assignment of sites. To evaluate the degree to
which individual stations influenced the discriminant analysis results, canonical discriminant
analysis was conducted multiple times. Each time, a different station was eliminated from the
calibration data set. Stations that had a large influence on discriminant coefficients were
identified and examined to determine if they were incorrectly assigned to test groups. Stations
that were judged to be misclassified were removed from the test data set and step 3 was
repeated.
Step 4: Validating the index
Step 4 involves validating the indices developed in the previous steps. Validation requires
testing the index with an independent data set to ensure that the multivariate solution is not
specific to the indicator testing sites sampled in 1990. Validation will be accomplished with
data collected in future years of the program; however, it was possible to validate the indices
preliminarily in two ways with the 1990 data: 1) the cross-validation procedure of Lachenbr-
uch (1975), in which each station is removed from the calibration data set and used as a test-
case for validation, and 2) evaluating the performance of the indices using data collected at 16
of the test sites that were revisited in interval 3 (September, 1990). Period 3 data are not
entirely independent because they were collected at the same sites as the period 2 data;
nonetheless, these data do permit examination of consistency of the response, which is a
necessary attribute of an index.
Step 5: Scaling the index
Scores from canonical discriminant analysis are unsealed and are difficult for general
audiences to understand; therefore, discriminant scores were normalized to a scale of 0 to 10
using the following algorithm:
4-10
-------
4.1.2 Benthic Macroinvertebrates
4.1.2.1 Development of a Benthic Index of Estuarine Integrity
Characteristics of benthic assemblages have been used to measure and describe ecological
status and trends of marine and estuarine environments for the last several decades (Sanders
1956; Sanders 1960; Rosenberg 1976; Boesch 1973; Pearson and Rosenberg 1978; Rhoads
et al. 1978; Boesch and Rosenberg 1981; Holland et al. 1988). This literature has identified a
diverse array of measures of benthic assemblages that represent ecological status and trends,
including 1) measures of biodiversity/species richness; 2) changes in species composition; 3)
changes in the relative abundance or productivity of functional groups; 4) changes in the
relative abundance and/or productivity of "key" species; 5) changes in biomass; and 6) relative
size of biota. Estuarine benthic ecologists have used site-specific knowledge about differenc-
es in the kinds, abundances, and physiological condition of individual benthic organisms
between degraded sites and undegraded reference sites to estimate the extent and degree of
degradation. Although many different attributes of benthic assemblages have been identified
and used previously to measure ecological status and trends, a generally accepted (i.e.,
calibrated and validated) benthic indicator that integrates the information available from all the
potential measures into a single measure (i.e., a benthic index value) that is applicable over
broad geographical areas and sensitive to a broad range of stresses has not been developed
previously.
Differences in the specific attributes of the benthic assemblage measured, collection methods
used, timing of sample collections, and specific objectives of programs limit the usefulness of
the existing data for developing a benthic index. For example, the level of taxonomic
identification and size of organisms sampled frequently varies from study to study, making the
data inappropriate for developing a benthic index. Data used to develop a benthic index must
be collected in a standardized way over broad geographical scales. Most of the existing
benthic data do not meet these criteria and cannot be used to develop a broadly accepted
benthic index.
Although the historical data for benthos cannot be used to develop a benthic index, these
studies provide valuable information that was used to design the EMAP-E benthic indicator
testing program. Historical data was used to identify candidate measures for inclusion in the
measurement program and develop cost-effective sampling and processing methods. Several
indicator workshops were held to review and refine the proposed benthic indicator testing
program. A goal of the Demonstration Project was to collect the data and conduct the
analyses necessary to develop a benthic index that would produce information that is
understood easily by technical and nontechnical audiences.
4-11
-------
Step 1: Identify candidate benthic measures
Benthic abundance, biomass, and species composition data were used to define 22 descrip-
tors of the major ecological attributes of the benthic assemblages occurring at each sample
site (Table 4-1). Many of these descriptors were formulated in several ways. The measures
of taxonomic composition and functional groups generally were calculated on the basis of both
biomass and counts. In addition to absolute values, measures also were expressed as
proportions of total abundance or biomass, where appropriate. Some of the measures, such
as biodiversity, were calculated by averaging and by compositing the three replicate samples.
When the 22 descriptors were reformulated in these ways, 58 candidate measures were
defined.
The 1990 Virginian Province Demonstration Project sampled only benthic organisms that do
not pass through a 0.5-mm mesh sieve. This component of the benthos constitutes greater
than 90% of the biomass and is relatively stable over long periods of time (Mare 1942;
Rhoads 1974). Only data for infauna (organisms that live in or on the sediment surface) were
used when computing values for candidate measures. Data for epifauna (organisms that live
on hard surfaces such as shells) were excluded from computations because 1) the sampling
methods did not sample epifauna efficiently; 2) the exposure of epifauna to pollution insults,
particularly chemical contaminants in sediments, is different than the infauna; and 3) shell and
other hard bottom habitats were found at relatively few sample sites (less than 1 %). Exclud-
ing epifauna reduced the possibility that the presence of small pieces of shell or other
structure that occurred at some sites would introduce unnecessary variability or bias analysis
results. Over 90% of the individuals collected were infauna.
Almost all of the candidate benthic measures were significantly (p<0.05) correlated with at
least one of the habitat factors measured. Over 75% of the candidate measures were related
significantly to salinity distributions (Table 4-2). Relationships between candidate measures
and other habitat factors (i.e., latitude, silt-clay content of sediments, water depth) occurred
less frequently and did not account for as much of the total variation as relationships with
salinity (Table 4-2).
Correlations between candidate measures and habitat factors, however, explained little of the
total variation in candidate measures. Only two of the correlations accounted for more than
25% of the total variation. Both of these were measures of species richness: 1) mean
number of species per site, and 2) total number of species per event. Salinity accounted for
up to 35% of the total variation for these measures. Figure 4-2 shows the distribution of the
number of species per site in response to salinity.
4-12
-------
Table 4-1 . Test for significant differences between degraded and nondegraded reference
sites for each of the descriptors of the benthic assemblage. P-values are
shown for the formulation of each descriptor that best distinguished sites of
different quality.
Measures of Biodiversity
Shannon-Weiner index
Pielou's evenness index
Proportion of expected number of species
Measures of Community Condition
Total benthic biomass
Total benthic abundance
Measures of Individual Health
Biomass/abundance ratio
Weight per individual polychaete
Weight per individual mollusc
Weight per individual bivalve
Measures of Functional Groups
Suspension feeding species abundance
Deposit feeding species abundance
Omnivore/predator species abundance
Opportunistic species abundance
Equilibrium specie^ abundance
Measures of Taxonomic Composition
Amphipod abundance
Bivalve abundance
Gastropod abundance
Molluscap abundance
Polychaete abundance
Capitellid polychaete abundance
Spionid polychaete abundance
Tubificid oligochaete abundance
t-test
(p-value)
0.06
0.18
< 0.001
0.03
0.005
0.52
0.02
0.97
0.98
< 0.001
0.02
0.001
0.16
0.001
< 0.001
< 0.001
0.02
< 0.001
0.02
0.006
0.9
0.49
Direction
(+ = greater value
at
reference sites)
+
+
+
+
+
+
+
0
0
+
+
+
+
+
+
+
+
+
+
+
+
4-13
-------
t. •
0)
»««
;Wto
CD O
W §
S^
(Q CD
OQCC
«
L O
co
1 IO
CM
CO
o
CM
- in
_ o
TT-
o
CO
I ' I
J- O
s
o
Tf
O
CO
£-0
II
•SB
CD 0)
E CD
.
o
'c
CO
CD
>
w
--
co "co
-S £
'co o
CO T3 O
.Z^ CO Q-
12* O> CD
Q 'c "2
a 1 S
CO
O
©US J©d sepeds jo
- CD $
ill
•la 8
CO CD O
CD *- CD
H— Q_
Q ° CO
"1°
3 =3 CD
S- is E
O CD "C3
CO h- tj
CD CD
0)0) x"
i- '« CD
CD •- CD
S g.«
o> to .52
.£ .9 CD
§ CD =
•S w J^
tj O^
Q- CD CZ
cvi E o
^ i «
£ c .co
D) CD CD
U- E ~
4-14
-------
Table 4-2. Summary of correlation between habitat indicators and the candidate benthic
measures
Habitat Indicator
Salinity (ppt)
Latitude
Silt-clay Content
Water Depth
All
Number of
Significant
Correlations
43
37
23
19
55
Number of
Correlations with
r2>0.10
12
2
5
3
18
Number of
Correlations with
r2 > 0.25
2
0
0
0
3
The two measures of benthic species richness were corrected for the effects of salinity by
calculating adjusted measures of species richness that were insensitive to salinity variation.
The adjusted measures consisted of calculating the percent deviation in species richness from
a baseline condition that represented the response of benthos to the estuarine salinity
gradient. The adjusted measurement was renamed percent expected number of species.
The baseline for estimating the percent expected number of species was obtained by fitting a
polynomial through the 90th percentile of a 3 ppt running average of species richness
measures using all the data collected in 1990. The r2 for this relationship was 0.98. The 90th
percentile was chosen because it did not differ substantially from the actual species richness
values observed for nondegraded reference sites and represented the number of species that
would be expected to occur if the only factor influencing species richness measures was the
estuarine salinity gradient. Figure 4-3 is the scatter plot showing the insensitivity of percent
expected mean number of species per site to salinity. This measure was also insensitive to
sediment type, depth, and latitude. Baselines based on other percentiles (75th, 50th) had
similar distributions and would have produced similar results.
The scientific literature suggests that the relative abundance of benthic organisms frequently is
related to sediment characteristics (e.g., percent fine sediment particles); however, only five of
the candidate measures had relationships with the silt-clay content of sediments that
accounted for more than 10% of the total variance (Table 4-2). All of these were negative
relationships with species richness measures, suggesting that mud habitats are inhabited by
fewer species than sand habitats. These variations were not
4-15
-------
• V
«. •«
. *_*.
I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I
ooooooooooooooooo
o
CO
Q.
Q.
V—'
o ^.
CM J
B;
(0
CO
_ o
- o
CO
CO
"I
.1
!Q
cfl
s
CO
'o
Q.
CO
E
ZJ
c
M—
O
Q.
CO
g
i
CO
4
I
O)
(pejoedxg 10 •/) sepods |o
4-16
-------
corrected because these relationships did not account for a significant proportion of the
variability in the data. In addition, mud sites had the greatest exposure to low dissolved
oxygen and contaminant stress. The mud habitats represented by lowest species richness
and diversity values were the same stations exposed to most severe low dissolved oxygen
and contaminant stress. Adjusting for this variation would have decreased the probability of
accurately representing the responses of benthic assemblages to pollution.
Step 2: Develop a test data set
Thirty-three sites were selected from base and ITE sampling sites for use in the indicator
testing data set; 14 of these were nondegraded reference sites (Table 4-3). The 19 remaining
sites were classified as degraded based on the criteria established for low dissolved oxygen
stress, exposure to contaminants, or both. Twenty-seven percent of the test sites were north
of the Hudson River. Both the degraded and nondegraded test sites encompassed a wide
range of habitats (Table 4-4), including all major salinity zones and sediment types that occur
in the Virginian Province; however, no test sites were located in oligohaline sand. Mud
substrate was more common at degraded test sites than at reference sites, which probably
reflects the affinity of contaminants for fine-grained particles.
Step 3: Identify combinations of candidate benthic measures that discriminate between
degraded and nondegraded areas
Twenty-eight of the 58 candidate benthic measures differed significantly between degraded
and reference sites in the t-tests (Table 4-1). These differences were in concordance with
expectations established when the measures were proposed (e.g., total abundance,
abundance of sensitive groups, and species richness measured were higher at regional
reference sites than degraded sites). These 28 measures were the only candidates used in
the discriminant analyses.
Discriminant analysis results are summarized in Table 4-5. Four variables were included in
the model for the first stepwise discriminant analysis (Index 1): 1) number of amphipods per
event, 2) the percent of total benthic abundance in molluscan taxa, 3) the mean weight per
individual polychaete, and 4) the number of capitellids per event. Using this combination of
measures, 7% of the nondegraded reference sites were classified as degraded (i.e., false
positives), and 11% of degraded sites were classified as reference sites (i.e., false negatives)
(Table 4-5). The canonical r2, which approximates the total variance explained by this
analysis, was 60%.
4-17
-------
CD
o
c
(X
s
O
o>
OJ
0)
O)
,C
•O
I
Q.
I
co
CO
i
O)
1
s
.1
w 1
° p
In 1
3
CD
CO
!
s1
o
•§
CO
|
5
ce Sites
u
S
T- D
io ra
m CM § CD
•O |JO .S T—
S!?0^
oS-gco
Lg
0
o
-c K p tn
° ••" S "*
a ?- g en
« •* 5 co
w
"Sr's
,?~ O CM
s
o CM o
^,0 (0 CO CO CM
cnoco'-coo
S. to C- 1> C. t>
OJ CO)
.CO fl)CO
§• ^ -g
a: uj m
PCO P
^TO n
% •§
a. o.
ill
I co to
to .
p
cr
CD
0
co
^»CO
to
. en
p
o
cu SJ b
T- .> O
to cr to
r o?
151»
rt («
Q.
I CM
IS
.a'
• S '^ 1
CB N.
cnco c co co co
a
co
CO
rp
cu
'
°
- .g
m l~~ T^ l~— "T" 1^
.S U) Jj8. tb ~P CO
OC IO 3- CO o CO
,- en .* b> :g co
uj m
CM O5
T- IO
CM CD
co "
O
CM
_N
CO TO 0
- o u>
O T—
•z? ^— - o o s- i—
CO O O T— O O^,O
coin >,ioi--cM s^rr ft *-
^ & ^ s-r- «s r- 2 r-
COCO i_ CO ^ "o TO ffiCM
CDCM.&U) 2cO Of-- g
^
r f£ ?
en n
-------
Table 4-4. Number of indicator testing sites sampled in each of several habitat types
Habitat Type
Sand
Oligohaline
Mesohaline
Polyhaline
Mud
Oligohaline
Mesohaline
Polyhaline
Number of Sites
Degraded
0
2
1
4
5
7
Nondegraded
0
1
3
6
0
4
Changes in the F-statistic and partial r2 at steps 1 and 2 of the first discriminant analysis
indicated that measures of amphipod abundance were correlated highly with salinity- adjusted
measures of species richness (percent expected number of species); therefore, in the second
discriminant analysis, the percent expected number of species was forced into the model.
Five measures were included in the second discriminant model (Index 2): 1) percent
expected number of species, 2) the percent of total benthic abundance as amphipods, 3) the
percent of total benthic abundance in molluscan taxa, 4) the mean weight per individual
polychaete, and 5) the number of capitellids per event. Classification efficiency with the
salinity-adjusted species richness measure included was the same as in the previous analysis,
and the canonical correlation coefficient was slightly higher (Table 4-5).
Sensitivity analysis indicated that two sites were having a disproportionate effect in
determining calibration of both discriminant models. These two sites (94 and 106), both of
which were classified as degraded sites, were examined more closely to determine whether
their disproportionate effect potentially resulted from misclassification of the sites. Site 94 is
located in the Arthur Kill estuary in northeastern New Jersey, and the data used to classify the
station as degraded were compelling: 1) 99% of the amphipods in the bioassay died; 2)
several metals and organic chemical concentrations exceeded Long and Morgan's (1990) ER-
M values; and 3) historic water quality data indicated that the site frequently experienced
hypoxic conditions in summer months. Based on this information, it was concluded station 94
was classified correctly and should remain in the calibration data set.
4-19
-------
g
T3
C
C
CO
o
4-1
asures i
CD
E
.£3
JC
B
CO
1
i conducted to combine <
w
CD
"co
s
ro
"c
CO
E
•c
CO
T3
**—
O
"5
CO
a
DC
uj
4
CD
i
H
^^
rtj
*S c « 'S «:
g 2 & £ «
irsi
*^™ >W *7V
o ^ ^* §
Hill
£a oo
Selected Measures
(0
T??
c
§
-5
s
v^
O3
00
CD
F amphipods
: individuals as molluscs
jht per individual polychael
f capitellids
00-5 o
CD "£ * CD
o nj *- n
E S § E
3 CD .g 3
Z Q-S Z
-a
CD .2
ll
(0 •—
^J eo
I™ .2
> ^
T- O =
-§ • <•
_c
CM
CD
^o
O5
S^
g
(pected number of species
: individuals as amphipods
: total abundance as bivalvi
jht per individual polychael
f capitellids
® "5 -5 -5 o
'£ '£; '£ S» CD
CD O CD c -D
E £J £> co E
CD CD CD ® 3
Q. 0. Q. S Z
"8
E^
« "O
CO ^
CD O
f* "~-
0 03
•E C
«l
•gfi
CM 0.=
x co <
•8- -
,c
r^»
CO
^
CD
'CO
S^
§
f amphipods
: total abundance as bivalvi
jht per individual polychael
f capitellids
00-50
U. ^_j > &»
CD C ^ CD
_Q 0 r- -Q
E ££ « E
3 CD ® 3
Z QL S Z
• • • •
TJ CD
(D ^
O O
£ §
W *"
_g> co
•= ^
£0 g
^ "^
CO O S
X Z CO
1- .
CO
-
CO
00
0^
§
0? CD
(pected number of species
f amphipods
: total abundance as bivalvi
3ht per individual polychael
f capitellids
Q 0^-5 o
C CD C & CD
0) -Q
-------
Site 106 is located in the Mystic estuary, Connecticut, and reexamination of the exposure data
for this site suggested it was probably misclassified. The sediment chemistry and sediment
toxicity data did not indicate a contaminant problem at this site. The original basis for
classifying this site as degraded was the occurrence of dissolved oxygen concentrations of 0
at the site. These extreme low dissolved oxygen readings were less than one hour in duration
and occurred on only one day. Dissolved oxygen concentrations of less than 2 ppm were
observed infrequently at site 106 during the rest of the summer (i.e., <6% of the time).
Because low dissolved oxygen concentrations were rare at this site, discriminant analyses
were repeated after removing station 106 from the calibration data set.
When station 106 was removed, the measures selected by the stepwise discriminant analysis
remained basically the same as those selected previously (Table 4-5). For the analysis in
which no variables were forced (Index 3), percent of total abundance in molluscan taxa was
replaced with a similar measure, percent of total abundance as bivalves, and all other
measures remained the same. For the analysis in which species richness was forced (Index
4), percent of total abundance as amphipods was replaced with total number of amphipods
per event. Removing station 106 from the calibration data set also improved the canonical r2
in both the forced and unforced discriminant analyses. The percent of sites classified
correctly, however, declined slightly (Table 4-5).
Step 4: Validating the model
Four potential indices were developed in step 3. The cross-validation efficiency for all of them
was about 80% (Table 4-6). The second means for validating indices was to determine the
classification efficiency for 16 sites that were revisited in September. For the first two indices,
15 of the 16 sites were classified correctly, and for the third and fourth indices, 14 of the 16
sites were classified correctly. One site (#223), a degraded site located in the Delaware
River, was misclassified by all four indices in the September data. This site was classified
correctly by all indices using the August data. These findings suggest that small-scale spatial
patchiness, or measurement error, rather than an inherent shortcoming in the index
parameters, may have been responsible for the error associated with this site. All of the
indices had correlation coefficients that exceeded 0.8 between August and September (Table
4-6).
4-21
-------
Table 4-6. Results of validation of four candidate indices. Description of measures
comprising each index is shown in Table 4-5.
index
1
2
3
4
Cross-validation Efficiency
Percent of
Degraded
Sites
Correctly
Classified
84%
79%
83%
83%
Percent of
Reference
Sites
Correctly
Classified
71%
86%
86%
86%
September Samples
Overall
Percent
Correctly
Classified
79%
82%
84%
84%
Index
Correlation
with August
Samples
0.80
0.81
0.84
0.84
Percent of
Sites
Correctly
Classified
94%
94%
88%
88%
Step 5: Scaling the index
The final step in developing the benthic index was to select one of the indices based on the
calibration and validation information, calculate discriminant scores for all sample sites, and
normalize the calculated scores to a scale of 0 to 10. Although all four candidate indices were
calibrated and validated preliminarily at an acceptable level, the fourth alternative was used to
assess the status of benthic resources of the Virginian Province. This index consisted of
forcing species richness and dropping the questionable station from the calibration step. The
discriminant function for this index was:
Discriminant Score =
(0.011 * Proportion of expected number of species +
(0.817 * Number of amphipods) +
(0.671 * Percent of total abundance as bivalves) +
(0.465 * Number of capitellids) +
(0.577 * Average weight per individual polychaete).
When applied to all 152 sites sampled, the range of this index was from -2.12 to +4.75 with a
critical value for discriminating between degraded and reference sites (calculated as the
midpoint between mean discriminant scores for degraded and reference sites) of +0.22.
When these discriminant scores were normalized to a range of 0 to 10, the critical value
between degraded and nondegraded was 3.40.
4-22
-------
4.1.2.2 Discussion of the Benthic Index
Although the benthic index developed above appears to work well for distinguishing sites of
differing environmental quality, it may not be the only effective index, or the most effective
one. First, covariance among many of the candidate measures was high, indicating that many
alternative combinations could produce comparable results. Second, index development was
based on only 33 indicator testing sites that, although representative of a wide range of
physicochemical and stress conditions, did not represent all possible conditions. In future
development of the benthic index, additional sites will be added to the calibration data set so
that it includes the full range of environmental habitats and stressors present. Third, there
may be other important measures of the benthic assemblage that were not included as
candidates in the stepwise discriminant analysis. As more is learned in local studies about
other measures that are effective for discriminating sites of differing environmental quality,
they can be incorporated into the calibrations. The index development process is flexible; as
other studies suggest increased confidence in selected measures, they can be incorporated
into the index through the forced stepwise discriminant analysis.
Although further index development is warranted, particularly validation with an independent
data set, it is not clear that the predictive capability of the index will change dramatically. The
process described above identified four possible indices. When they were applied to all sites
for which there was benthic data, the correlation between the four index values exceeded 90%
for all pain/vise combinations (Table 4-7). Similarly, over 90% of stations in this data set were
classified the same (degraded vs nondegraded), regardless of the index used.
Table 4-7. Correlation coefficients relating the values of each of four indices (top of table)
and percent of sites that were classified the same (degraded vs. nondegraded)
among four indices (bottom of table) applied to 1 52 sampling sites.
Components of each index are provided in Table 4-5.
Index 1
Index 2
Index 3
Index 4
Index 1
94%
95%
95%
Index 2
.97
91%
91%
Index 3
.96
.96
100%
Index 4
.96
.96
.99
As development of a benthic index of estuarine condition continues, several questions need to
be addressed. One of these is, "how good is good enough?" In the analyses described
4-23
-------
above, 90% accuracy and about 84% cross-validation were the best classification efficiency
rates that could be achieved. These may be acceptable because, even with a 10% to 15%
error rate, biological data may be interpreted more easily than exposure information such as
sediment chemistry alone.
The degree of control that investigators have for achieving a higher degree of accuracy may
be limited. Some error may be due to sampling or measurement variability, as suggested by
the fact that 9% of the sites switched classification from August to September. This error
probably can be reduced by taking a iarger sample or by enforcing greater quality control in
the laboratory; however, doing so adds considerable cost. The sample size with three
replicates used in the Demonstration Project already exceeds 1200 cm2, and the quality
control protocol required over 90% accuracy in sample sorting and identification. Some of the
error may be due to the way sites in the calibration dataset were classified using dissolved
oxygen and contaminant exposure. This error could be reduced by collecting additional
measures, such as those that describe habitat alteration, but again at considerable cost.
Finally, achieveable levels of sensitivity may be limited by the nature of the attributes being
examined; assemblage measures may not be sensitive to some types of exposure. This
limitation could be addresed by adding new types of indicators, such as measures of the
sublethal responses of individuals, but these types of biomarkers are just beginning to be
developed for invertebrates.
Several other factors that have not been addressed fully here need to be considered in further
development of the benthic index. The index developed here may not be the most cost-
effective. It is based on three replicate samples and incorporates biomass measures. It might
be possible to reduce sampling and laboratory processing effort with little loss in classification
efficiency. Additionally, the index was developed for a single region of the country. Although
it might be the optimal solution for classifying sites in that region, a more general solution that
is applicable nationally might be possible, with very little loss in effficiency at the regional
level. The index development approach above, once augmented with additional data for other
regions of the country and validation data, will allow these alternatives to be evaluated.
4.1.2.3 Sediment Profile Imaging Techniques
Traditionally, the condition of benthic assemblages is evaluated using quantitative measures of
species composition, abundance, and, in some cases, biomass. An alternative approach to
assessing benthic condition is sediment profile imaging, which involves still photography of the
sediment-water interface to quantify various physical, chemical, and biological characteristics
of sediments. These characteristics include sediment grain size, penetration of oxygen into
the sediments, and a qualitative description of benthic community composition (Rhoads and
Germano 1982, 1986). Profile imaging offers the advantage of rapid availability of data, as
automated image analysis can provide data in nearly real time. Eliminating laboratory
4-24
-------
processing also can reduce cost per sample, although capital costs associated with the
equipment are substantial.
Although sediment profile imaging has been applied widely in monitoring dredged material
disposal sites and organic enrichment gradients, it is not yet clear that the technology is
developed sufficiently for application to a regional monitoring program such as EMAP, which
involves sampling a wide variety of habitats. There have been only a few direct comparisons
between sediment profile imaging and traditional benthic assessment methods, and these
studies were conducted in fine-grained, polyhaline estuaries (Scott et al. 1987; Long et al.
1990; Grizzle and Penniman 1991). Although these studies support the use of sediment
profile imaging in high salinity (greater than 15 ppt) waters, the technique has not been tested
adequately in moderate or low salinity habitats (Holland et al. 1988, 1989), or in habitats with
coarse-grained sediments. Application of such techniques in EMAP requires a thorough
understanding of indicator responses across all habitat types.
During the Demonstration Project, sediment profile imaging was attempted at 20 sites where
benthos were collected using traditional techniques. The objectives of this sampling were to
determine the logistic feasibility of sediment profile image sampling across the broad range of
habitat types and to compare the Organism-Sediment Index (OSI) values produced from the
sediment profile images (see Section 3 for calculation of this index) with the benthic index (Bl)
values obtained from the same habitats.
Logistical difficulties with sediment profile imaging were encountered in sampling coarse
grained sediments, which resulted in being unable to gather data from 15% of the test sites
(see Section 4). At the remaining 17 sites, the OSI and benthic index were statistically
correlated (p<0.05), but with a low correlation coefficient (r2 = 0.36). Threshold values
distinguishing degraded from nondegraded benthic conditions are 3.4 for the benthic index
(this report) and 6 for the OSI (Valente et al. 1992). Ten of the 17 sites were classified in the
same condition by both indices (Table 4-8). Six of the seven sites that were classified
differently were classified as degraded by the OSI and nondegraded by the benthic index. Of
these, three exhibited OSI values at or just below the threshold. Over half (nine) of all OSI
values were at the threshold value, or one unit away from it, despite a total range for the index
of 22 units (-11 to 10). Small shifts in the OSI, therefore, can dramatically change the
classification analysis in this test.
The test was limited by the selection of sites at which profile imaging was conducted.
Although a primary purpose of the study was to examine the technology in low salinity and
coarse-grained environments, only three samples were collected in sites with salinity less than
15 ppt, and only three samples were collected from sites where silt-clay content was less than
20%. For two of the three low salinity sites, the OSI and benthic index classified the sites
similarly. Of the three coarse-grained sites, the two methods classified one site differently.
4-25
-------
Table 4-8. Comparison of benthic characterizations produced by sediment profile image
analysis and conventional macroinvertebrate sampling. The threshold value
for the benthic index (Bl) is 3.4; the threshold value for the Organism-
Sediment Index (OSI) is 6.
Station
98
215
95
177
199
71
76
24
106
25
26
21
74
73
70
96
7
Salinity
(PPt)
25
0
27
7
4
32
30
28
25
28
27
28
32
32
31
27
28
Sediment Type
(% silt/clay)
57
91
14
90
89
81
5
52
93
87
74
91
15
63
56
93
24
Sediment Profile
Image (OSI)
-8
5
6
-1
6
7
6
4
-3
6
9
4
10
5
7
5
9
Conventional
Sampling (Bl)
0
2.4
2.5
4.1
4.9
5.8
6
6.1
6.1
6.6
6.9
7.8
7.8
8.1
8.5
9.4
10
The test was limited further because sediment profile imaging was not conducted at sites of
known quality, such as those in the indicator testing data set; thus, it is not possible to
determine which of the two methods provided a more accurate assessment of local conditions.
The uncertainty associated with sediment profile imaging in selected habitats, combined with
logistical difficulties associated with collecting samples in those habitats, indicate that
replacing the conventional benthic invertebrate sampling with sediment profile imaging
presently is not an appropriate option for EMAP. Further testing of the technology appears
warranted since the available data are limited and insufficient to conclude that the technology
is not feasible; moreover, recent advancements in reducing the size of profile imaging
equipment may reduce the cost and eliminate some of the logistical obstacles associated with
sampling coarse-grained sediments.
4-26
-------
4.1.3 Fish
4.1.3.1 Development of a Fish Index of Estuarine Quality
Indices of environmental health based upon the fish assemblage occurring at a site have been
accepted widely for use in freshwater environments. The Index of Biotic Integrity (IBI; Karr
1981) has become a standard measure for defining environmental quality in several states
(Plafkin et al. 1989). To date, a validated index of estuarine environmental quality based on
information about fish assemblages that is applicable over broad geographic areas has not
been developed. During the Demonstration Project, attempts were made to develop a fish
assemblage indicator using the same procedure as for developing a benthic invertebrate
assemblage indicator. Each step in that procedure is outlined below.
Step 1: Identify candidate fish assemblage measures
Fourteen fish measures were identified as candidates for discriminating between sites of
differing environmental quality (Table 4-9). The list of candidate measures was developed
based on general knowledge of fish assemblages and their response to stress, and on historic
case studies examining the effects of local perturbations on fish assemblages. Most of the
case studies involved freshwater fisheries, particularly those considering application of the IBI
(Karr et al. 1985; Berkman et al. 1986; Leonard and Orth 1986). Measures suggested by
Miller et al. (1988) for estuarine application of an IBI and measures identified in a number of
estuarine case studies investigating the response of fish assemblages to site- specific
perturbations (Bechtel and Copeland 1970; White et al. 1977; Haedrich and Haedrich 1974;
Horn 1980; Allen et al. 1983) also were incorporated.
Identification of appropriate species composition measures presented the greatest challenge in
developing the list of candidates. Species composition can be quantified in two ways. The
first is the indicator species or the indicator taxa approach. In this approach, the pres-
ence/absence or dominance of a particular taxa is used to define the quality of the site.
Presence/absence is appropriate to sensitive or specialized species, such as endangered
organisms, whereas dominance might be more appropriate for abundant or ubiquitous
species. The second is the assemblage tolerance approach developed by Hilsenhoff (1987)
for freshwater benthic invertebrates. For this approach a tolerance value is assigned to each
taxon and used to calculate an abundance-weighted tolerance value for the assemblage. This
approach differs from the indicator taxa approach in two ways. First, it emphasizes the most
abundant taxa at the site; consequently, rare taxa, such as endangered species, that might be
accurate indicators of a high quality environment would contribute little to the ranking of the
site. Second, it requires ranking all taxa in the assemblage, not just those for which special-
ized knowledge about their preferences and tolerances is available.
4-27
-------
Table 4-9. Tests for significant differences among degraded and reference sites for each
candidate fish measure
Measures of Biodiversity
Shannon-Weiner index
Evenness
Number of species per haul
Measures of Community Condition
Total number of fish per haul
Measures of Individual Health
Percent of fish with gross pathology
Measures of Functional Groups
Percent top carnivores
Percent pelagic invertivores
Percent benthic invertivores
Percent planktivores
Measures of Taxonomic Composition
Composition index
Percent bullheads
Percent clupeids
Percent scianids
t-test
(p-value)
0.04
0.02
0.10
0.22
0.01
0.65
0.91
0.04
0.77
0.02
0.82
0.75
0.44
Direction
(+ = higher value
at reference
sites)
J
+
„,
+
t
The difficulty with applying these approaches to the 1990 EMAP-E data is that neither has
been elaborated for estuaries. Both require information on the relative tolerance of estuarine
fish to pollution exposure that is not readily available. Most studies to assess differences in
fish assemblages between degraded and nondegraded sites have been limited to site-specific,
discharge-related problems, rather than to more generalized habitat degradation problems.
Almost all have been conducted on local scales, usually within a single system, and most
have been limited to a narrow range of salinities.
Data from the 1990 Demonstration Project was used to develop a composition index, based
on the approach used by Hilsenhoff (1987). This approach requires tolerance values ranked
from 0 to 10 for each species encountered in the samples. No such values exist for estuarine
4-28
-------
fish, so they were generated based on presence/absence information at the test sites (Table
4-3). The tolerance value was set equal to the percent of reference sites at which the species
was found minus the percent of degraded sites at which it was found, normalized to a scale
from 0 to 10. These procedures were applied separately to three salinity zones (less than 5
ppt, 5-18 ppt, greater than 18 ppt) to account for the possibility that a fish's degree of
tolerance for degraded conditions may be a function of its proximity to the extreme boundary
of the natural habitat of the species. The composition index for each station was calculated
as the abundance-weighted average tolerance value.
Each of the candidate measures was tested for relationships with salinity, depth, and latitude;
substrate was not tested because a trawl covering several hundred meters typically encoun-
tered a wide range of substrate types. None of the measures were significantly correlated
with latitude. Abundance and species richness were the only measures that correlated
significantly with salinity. The relationship with salinity was greater for abundance than for
species richness, but the r2 value did not exceed 0.1 for either of these. The relationship with
depth was significant only for the composition index, but less than 5% of the variation was
explained by the relationship. Because the amount of variance accounted for by environmen-
tal factors was small, no adjustments were made to the data before they were used in the
discriminant analyses.
Step 2: Develop a test data set
The test sites used to evaluate the fish response were the same as those used to evaluate
the benthic response (Table 4-3). The original intent was to use data from sampling interval 2
to calibrate the fish indicators and from sampling interval 3 to validate the consistency of
responses; however, because fewer than 40 fish were caught at more than half of the stations
in interval 2, data from sampling intervals 2 and 3 had to be combined for fish analyses. Data
from interval 1 were not used because of the large number of sites that were not sampled in
that interval, and because it had been determined previously, based on analysis of the water
quality data, that sampling in future years will be concentrated between late July and early
September (see Section 5.1).
4-29
-------
Step 3: Identify combinations of candidate measures that discriminate between
degraded and nondegraded areas
When subjected to univariate examination, only five measures, prevalence of pathology,
evenness (number of species constituting 90% of the assemblage), the Shannon-Weiner
diversity index, the composition index, and percent of benthic-feeding fish in the assemblage,
distinguished significantly between degraded and nondegraded sites in the indicator testing
data set (Table 4-9). Visible pathology was the most effective measure. The prevalence of
visible pathology was less than 1% at all reference sites, and one-third of the polluted sites
exhibited prevalances higher than 1 %.
Three measures, prevalence of pathology, evenness, and the composition index, were
identified by stepwise discriminant analysis as the most appropriate combination of measures
for discriminating between degraded and reference sites (Table 4-10). When this group of
measures was weighted optimally, successful discrimination was achieved for 13 of the 14
(93%) reference sites, and 16 of the 19 (84%) degraded sites. All three degraded sites that
were misclassified were located in areas north of the Chesapeake Bay and were classified as
degraded on the basis of high concentrations of contaminants. No pattern or trends were
observed in the habitat variables among the misclassified sites. When the analysis was
repeated after eliminating station 106 from the calibration data set, as in the analysis for
benthic invertebrates, the same three candidate measures were selected in discriminant
analysis. Eliminating this station affected the parameter coefficients slightly but had no effect
on the number or identities of misclassified stations.
Step 4: Validating the model
Two potential fish indices were developed in step 3. The cross-validation efficiency for Index
1, which included station 106, was 85%; for Index 2, which did not include station 106, the
cross-validation efficiency was 83%. Although calibration of the fish index was conducted with
combined data from intervals 2 and 3, the index was recalculated for each period individually
to examine whether the response was consistent at each sampling site. Of the 19 indicator
testing sites for which data were available in both intervals, 16 (84%) were classified the same
between periods. At two of the three sites that were classified differently, calculation of the
index was based on a collection that included less than five fish in at least one of the periods;
an even higher performance might have been achieved by eliminating variability associated
with small sample sizes.
Validation against an independent data set is particularly important for the fish index because
the composition index was included in the final discriminant function. Inclusion of the
composition index involves some circular reasoning because the tolerance values on which it
is based were derived using the same data set used for examining discrimination
4-30
-------
1 II
1
..
CO
o
1
CO
£
3
8
03
£2
CO
!£=
3
To
5
1
D)
"c
la
8
p
M-^
"8
iant analyses conduct
E
•c
CO
"•o
*Q
1
d
•^
CO
O)
vP
o^
s
^
u)
"o
I
•Percent of fish with gi
•Eveness
•Composition index
1
I
CO
JS
c^
.£5
^5
!•
0
in
d
•-?
CO
0?
CO
^
O)
"o
I
•Percent of fish with g
•Eveness
•Composition index
ID "S
1 S
i|
-Q "
-li
|!!
$
d
.^>
^_
|v"
55
s
>,
O)
"o
f
CO
0 x
Ij
£ C
_W '(D
a> §
0 -C
"U
(D
•8
C
'E
'f
_c
.1
.*t=
i
"1
4-31
-------
between degraded and reference sites. Recognizing this shortcoming, stepwise discriminant
analysis was repeated excluding the composition index. Under this scenario, only percent
pathology and the Shannon-Weiner index were selected in the analysis (Table 4-9). The rate
of discrimination using just these two measures was 79%. The cross-validation efficiency also
was 79%. When period 2 and 3 data were compared, however, less than half of the sites
were classified the same (degraded or not degraded) between periods using this index. The
poor agreement between intervals 2 and 3 using this combination of measures may be due, in
part, to small sample size; however, it suggests the importance of the composition index to a
stable multi-metric fish index of environmental quality.
Step 5: Scaling the index
None of the potential fish indices developed above were validated sufficiently to warrant
placing them on a scale from 0 to 10 or incorporating them into the preliminary evaluation of
environmental condition presented in Section 2. Although index 1 appears promising, use of
the composition index requires further validation. Index 3, which did not include the composi-
tion index, had less than 80% discrimination and compared poorly between periods.
4.1.3.2 Discussion of the Fish Index
Developing an indicator of environmental quality based on fish assemblages may be more
difficult than developing one based upon benthic invertebrates for several reasons. First, the
mobility of fish makes them more difficult to collect; the efficiency of trawls is typically low and
species-specific (Kjelson and Colby 1977). Second, fish assemblages in an estuary are often
transient; thus, the assemblage at any one time may not accurately reflect conditions at a site.
Third, fish respond to habitat variables (e.g. structure) that are difficult to measure or incorpo-
rate in an index.
To assess whether the trawl sampling gear used in the Demonstration Project provided a
consistent description of the fish assemblage at a site, correlations were identified for each
measure between the first and second hauls at a site in interval 2 (replicates were collected at
a subset of sites in interval 2). The correlation was significant for every measure, and the r2
for most measures exceeded 0.75 (Table 4-11). Presumably the large gear combined with
efforts to standardize trawls resulted in this consistency.
To assess whether the assemblage at a site was transient over the sampling period, correla-
tions were conducted for each measure between the first trawls taken at a site in intervals 2
and 3. Correlations for the biodiversity measures, percent pathology, and the composition
index were significant, although none had an r2 as high as 0.5 (Table 4-11). The correlations
for measures relating to taxonomic composition and functional groups were not significant.
4-32
-------
Table 4-11. Correlations between the first and second trawl and between intervals 2 and
3 for each candidate fish measure
Trawl 1
vs.
Trawl 2
Sampling Interval 2
vs.
Sampling Interval 3
Measures of Biodiversity
Shannon-Weiner Index
Evenness
Number of Species
0.65
0.74
0.87
0.40
0.64
0.70
Measures of Community Condition
Total Number of Fish per Haul
0.84
0.37
Measures of Individual Health
Percent of Fish with Gross Pathology
0.89
0.87
Measures of Functional Groups
Percent Top Carnivores
Percent Pelagic Invertivores
Percent Benthic Invertivores
Percent Planktivores
0.63
0.86
0.72
0.82
0.03
0.38
-0.12
-0.02
Measures of Taxonomic Composition
Composition Index
Percent Bullheads
Percent Clupeids
Percent Seianids
0.81
0.95
0.80
0.93
0.62
-0.05
-0.04
0.18
The issue of transience is exacerbated by the fact that most fish populations respond to
environmental impacts on a cumulative, basinwide or regional level. Frequently, estuarine fish
are part of coastal stocks and complete their life cycles in portions of different estuaries;
consequently, the abundance and composition of fish may not respond on the same spatial
scale as the Demonstration Project sampling activities. While this would not eliminate fish as
useful indicators, it does present a difficulty for validating them, particularly given the site-
specific orientation of the validation approach used here.
Failure to account for all habitat variables at a site probably also hinders development of a fish
index that discriminates degraded sites from nondegraded reference sites. At each of the
sampling sites, habitat was characterized by two variables, salinity and depth. Fish abun-
dance is related to many additional habitat variables, such as bottom type, presence of
4-33
-------
structure, and distance from shore. Some measures will not distinguish well without normaliz-
ing for these additional habitat variables. For example, some fish species are attracted to
submerged structures for protection and food resources. Several of the degraded sites in the
indicator testing data set were located in small systems with substantial physical structure
(e.g. docks, debris such as cars, which serve as artificial reefs), whereas some of the
reference sites were located miles from shore in areas with a uniform bottom. Accounting for
these differences is difficult but may be required to develop a successful fish index.
Some of these difficulties could be reduced by conducting shore-zone sampling with a seine
net instead of offshore sampling with a trawl. Shore-zone sampling dramatically reduces the
amount of habitat variability because seine sampling is limited to low gradient beaches with
little structure. Additionally, the shore-zone assemblage typically includes smaller, less
mobile fish and is more stable over the summer period. Incorporation of shore-zone sampling
into EMAP, however, would be difficult. The percentage of shore- zone area that is sample-
able is limited, and sampling it in an unbiased manner would require a different sampling
design than is being used in the rest of the program.
Despite the apparent difficulties associated with developing a fish index based on trawl
sampling, the results of index 1 showed some promise for discriminating environmentally
degraded areas on a regional scale. Further testing, validation, and refinement should be
pursued. Three areas appear to be most appropriate for future effort: 1) refining tolerance
values used in the composition index, 2) developing a data set for validating the combination
of measures identified during the 1990 Demonstration Project, and 3) identifying new
measures that could be used in addition to those already identified. Several fish measures,
such as tissue contaminant levels and biomarkers, have not yet been incorporated into the
index development process.
4.1.3.3 Visible Pathology
Regardless of whether a fish index that successfully discriminates sites of high and low quality
can ever be validated, there are still a number of fish measures that convey meaningful
information for describing trends on a watershed basis. These measures include, the percent
of fish with visible pathological disorders, the percent of fish of commercial/recreational value,
the percent of introduced species, or the percent of estuarine area (or range) over which a
particular species of interest is found. These types of indicators cannot be used effectively in
some types of status evaluations because a threshold boundary for acceptable and degraded
condition cannot be defined easily; however, these indicators are useful for comparing status
among watersheds. They would be particularly useful for assessing trends in watershed or
regional condition because a change in indicators, such as an increase in the range or spatial
coverage of a nuisance fish (e.g. lampreys), or a decrease in the percent of fish with visible
pathological disorders, can be interpreted relatively easily.
4-34
-------
Estimates of the prevalence of visible pathological disorders provide a good example of how
data for which baseline conditions are not known can still be useful for assessment. It was
possible to demonstrate large differences in the prevalence of visible pathology among
different classes of estuaries using 1990 Demonstration Project Data (Section 2). Even
without knowing whether a specific rate of pathology is considered above background levels, it
was possible to identify that fish in small estuarine systems had considerably more pathologi-
cal problems than fish in large estuaries or large tidal rivers. This is important information
because the frequency of pathological disorders in fish is often used as an indicator of
environmental quality not only by scientists, but also by the general public. The public
associates fish pathologies with poor water quality. An increase in visible pathologies on
recreationally and commercially valuable species may result in adverse public reaction and
subsequent economic consequences. The relationship between environmental stress and
visible pathological conditions such as fin erosion, skin tumors, and ulcers in fish has been
established, and highly polluted habitats have greater frequencies of fish with visible pathologi-
cal disorders than similar, less polluted habitats (Sindermann 1979; Mearns and Sherwood
1974; O'Connor et al. 1987; Buhler and Williams 1988; Malins et al. 1984, 1988). In addition,
measuring visible pathological disorders is cost effective because fish can be screened rapidly
for abnormalities while being collected for other analyses, such as tissue chemistry or
assemblage information.
Although promising, the visible pathology indicator requires further development. Many types
of visible pathology like those encountered during the Demonstration Project are species- and
size-specific (Moller 1984; Wolthaus 1984); thus, status estimates that compare classes, or
trends analyses that compare years, may confound differences or changes in species or size
composition with differences or changes in environmental factors that lead to visible pathologi-
cal abnormalities. EMAP presently is developing normalization procedures that will account
for size and species effects on visible pathology in the assessment process.
4.1.3.4 Fish Tissue Contaminants
Several fish measures, such as tissue contaminant levels or biomarkers, have not yet been
measured in EMAP-E (tissue samples were collected, but not processed in 1990; see Section
4). The principal problem with using indicators at the level of individual fish is that making
regional estimates of environmental condition requires collecting fish from a small target group
of taxa at a large percentage of the sampling sites. The target group must be small to
minimize difficulties with species-specific response. Since the EMAP design is probability-
based, sites where fish from the target group can not be collected present an inference
problem: it is not possible to determine whether failure to capture target species resulted from
insufficient collection effort or from adverse (e.g. contaminated) conditions that caused fish to
avoid those sites. If the latter is true, failure to include those sites in the evaluation would
4-35
-------
result in underestimating the extent of estuarine area with subnominal environmental condi-
tions.
Data collected during the 1990 Demonstration Project were used to determine whether a small
number of target species could be captured at a sufficient number of sample sites to justify
generating estimates of condition for the Virginian Province. Ten target species were
identified as the most appropriate to provide provincewide coverage, based on a survey of
historic fish trawling information (Holland 1990). These species were retained for tissue
chemistry analysis during the Demonstration Project (Table 3-5). At least one of the target
species was captured at 75% of the sites sampled during the 1990 Demonstration Project.
Most of the sites where target species were not collected were too deep to trawl safely using
the hydraulic winch system available on sampling boats. Underwater obstructions prevented
successful trawling at less than 6% of the sites. With minor methodological modifications and
slight alterations in the list of target species, it is anticipated that target species will be
collected at more than 90% of the sites in future years. On this basis, collections for contami-
nant analysis will be continued in future years of the program, and biomarker research within
the program will be expanded.
4.2 EXPOSURE INDICATORS
Methods for measuring exposure indicators at specific sites are well-developed and have been
used in a number of monitoring programs throughout the province. Methods for measuring
indicators on a regional scale in a probability-based manner to develop regional status
estimates of the kind being made in EMAP are not well-developed.
4.2.1 Dissolved Oxygen
Dissolved oxygen (DO) is a fundamental requirement for the maintenance of fish, shellfish,
and other aquatic biota. Dissolved oxygen concentrations also reflect the integrated response
of complex natural systems to nutrient input. The increased prevalence of low dissolved
oxygen conditions in geographically dispersed estuaries is a central aspect of the declining
health of near coastal waters (Turner and Allen 1972; May 1973; Jorgensen 1980; Harper et
al. 1981; Rossignol-Strick 1985; Justic et al. 1987; Kuo and Neilson 1987; Rosenberg 1990).
Regulatory programs to improve dissolved oxygen conditions in estuaries have cost billions of
dollars, and assessing the regional effectiveness of these programs is one of the goals of
EMAP.
The major problem with using dissolved oxygen as an exposure indicator is its extreme
temporal variability in estuarine waters (Breitburg 1990; Sanford et al. 1990). Dissolved
4-36
-------
oxygen concentrations in bottom waters can fluctuate from supersaturation to severe hypoxia
within hours. Factors contributing to these fluctuations include tides, meteorological condi-
tions, and biological activity. In stratified estuaries, bottom waters move with tidal currents
and seiches, and if partially depleted of oxygen, bottom water DO measurements will fluctuate
in response to water movements. During the 1990 Demonstration Project, continuously-
recording DO meters were deployed at sites throughout the Virginian Province to document
the extent of these fluctuations in bottom waters. Examples of DO fluctuations in the Virginian
Province are shown in Figure 4-4.
The purpose of this section is to use continuously-recorded DO data collected during 1990 to
identify the most appropriate method for measuring DO within EMAP's sampling constraints.
These constraints included safety considerations, time and funding limitations, and equipment
performance limitations. For safety and logistical reasons, stations were sampled during
daylight hours only. Stations could be visited a maximum of three times during the summer
index period due to limited time and funds. Dissolved oxygen meters could be deployed
reliably for no more than three days at a time. The duration of deployment was limited
because fouling reduces the reliability of measurements after about three days. Deployments
beyond three days also exceeded the funding and logistical limitations of EMAP.
Given these constraints, three sampling strategies were evaluated:
• Single point-in-time samples collected on random days in the index period between
0800 and 1800 hours.
• Multiple point-in-time samples collected on random days in the index period, visiting
each site two or three times. The index period is divided into two or three equal
intervals, and sample dates are chosen randomly from each sub- period. Succes-
sive point samples at the same station are separated by at least nine days to avoid
autocorrelation problems.
4-37
-------
(VA90-070)
O
°
8
Month
(VA90-088)
i4
8
Month
Figure 4-4. Time series data obtained at two long-term dissolved oxygen monitoring sites
showing fluctuation of dissolved oxygen concentrations over the summer index period.
4-38
-------
• Short-term continuous records obtained by deploying a continuously-recording
meter at each site to measure DO at 30-minute intervals for one, two, or three
days. The starting date for the three-day period is selected randomly within the
index period.
Making three-day continuous measurements was the most capital-intensive of the three altern-
atives, as it required purchasing a sufficient number of units for simultaneous deployment at
several stations, and extra units for calibration, repair, and replacement of lost units. Three-
day deployments, however, are not as labor intensive as three single-point measurements and
are approximately equal in labor to collecting two point-in-time samples per station. Lost data
may be a problem because of lost or malfunctioning meters, but 1990 data suggested that
such losses would occur less than 3% of the time (see Section 3); furthermore, point mea-
surements (CTD casts) taken upon meter deployment and retrieval serve as a backup
estimate if the continuously-recording meter fails.
Dissolved oxygen data were used in two ways: 1) for estimating status and trends, in which
the objective was to evaluate the condition of estuarine areas rather than individual sites; and
2) for identifying associations, which required a higher degree of certainty about conditions at
an individual site. Measurement methods were evaluated separately for each use.
4.2.1.1 Use of dissolved oxygen data for estimating status and trends
Developing a regional status estimate involves generating a population distribution estimate (a
cumulative distribution function, or CDF) from the sample data. In EMAP-Estuaries, samples
were collected over an index period lasting 30 to 60 days, and sample sites were visited in a
semi-random fashion (complete randomness is not possible due to logistical constraints). The
CDFs used to make EMAP-E status estimates were calculated from the sample data and
site-specific inclusion probabilities.
Point-in-time Measurements
Point-in-time measurements represent the least expensive and most logistically feasible
means of estimating the status of dissolved oxygen exposure. Given that it is not possible to
sample an entire province synoptically, the CDF developed from point-in-time samples
depended on the order and time period in which the samples were taken. This will especially
be true if DO fluctuates systematically between sites, resulting in an unstable distribution of
DO. Using single, point-in-time measurements to make status estimates requires stability in
the areal extent of low DO over the province throughout the index period.
4-39
-------
The stability of point-in-time measurements was tested by comparing the CDFs of bottom
dissolved oxygen concentrations between two successive sampling intervals (Fig. 5-5). These
comparisons demonstrated that the distribution of dissolved oxygen was stable between the
two periods; thus, single point-in-time measurements of dissolved oxygen are adequate for
estimates of status on a regional basis during the summer index period.
Multiple Point-in-time Measurements
Multiple point-in-time measurements changed neither estimates of status, nor the shape of the
distribution; however, the increased sample size of multiple point-in-time measurements
reduced sample variance and increased the power to estimate trends.
Continuous Oxygen Records
Continuous samples of one to three days duration were marginally better for obtaining status
estimates. The primary advantage of the short-term continuous samples was the ability to
remove the sampling bias due to collection of only daytime samples by allowing selection of a
single random sample from each continuous record. This results in the same sample size as
single point samples. The additional data points obtained by continuous samplers were of
little use in estimating status because the DO time series is strongly autocorrelated, and
samples within three days of each other are not independent (Fig. 4-6).
The effects of natural, daily fluctuations in DO concentration must be considered. EMAP
sampling was carried out during daylight hours for logistical and safety reasons, and stations
typically were sampled between the hours of 0800 and 1800. Water column dissolved oxygen
concentration is determined by photosynthesis, respiration, and atmospheric diffusion. In
productive shallow waters, there may be a circadian cycle of DO in which a minimum
concentration occurs around sunrise, and a maximum concentration occurs near sunset.
Diurnal periodicities in DO also occur. They were most pronounced in waters less than 3 m
deep, while tidal periodicities are most pronounced in waters more than 6 m deep (Fig. 4-7).
Deep waters are more likely to be stratified.and subpycnocline hypoxic water can move in and
out with the tide, producing a tidal signal in the DO record. The daytime sampling schedule of
EMAP may have resulted in biased estimates of dissolved oxygen exposure.
We tested the difference between daytime and completely random sampling, and found that
the advantage of complete temporal randomization is virtually undetectable, based on Monte
Carlo sampling of the long term DO (LTDO) time series. One hundred random observations
were taken from each of 13 time series, and 100 random observations were
4-40
-------
CO
•* I
in
0>
o»
tn
CO
CO
o
o>
GO
O
in
C3 O C3 O
CO CM ••-
D)
CO
,_c
OT
o
I
(D
O
O
O
ui CM
B9JV
4-41
-------
CD
O
C3
04
O
O
CO
CO
CM
•»- o
o
O
CO
o
o
CO
CM
CO
CO
CM
CM
O
CO
cp
CO
I
S
.-i 05
S.E
c o
8°
E .w
p c
J= CD
g-e
fc 0)
co >
w 5
c «
o co
c ^
O CD
'
ss
CO c
•si
||
CD OT
-1- c
o>.g
o "c
CO o
-i §
X O)-^
LU >; u
cc5 E
i ^3
CD .1 "
3 8 "
.O) CO D)
4-42
-------
co
CO
o oq ce •* evi
,-^0000
(I / 6W) 00 u! S90U9J9JJJQ
4
I
O)
4-43
-------
taken from observations between 0800 and 1800 hr in each time series. Samples from the
time series were combined to develop a CDF for each sampling strategy (Fig. 4-8). The two
sample CDFs were not significantly different from each other, and both adequately represent-
ed the population CDF.
Only a few of the LTDO sites displayed a relatively strong diurnal signal (Fig. 4-7). On a site
basis these patterns are very important. The Monte Carlo test suggested that on a regional
basis, the effect of those sites may be unimportant; however, this conclusion is dependent on
the representativeness of the LTDO sites. The importance of these cycles, hence, the
importance of randomizing with respect to them, depends on the relative frequency of sites
where diurnal cycles are dominant.
4.2.1.2 Use of the dissolved oxygen indicators for associations
The objective of examining associations is to identify relationships between ecological
response and environmental exposure. Although this can be done by comparing the coinci-
dence in trends of exposure and response in a population, the most powerful methods require
paired observations of DO and response indicators at each site, allowing regression or similar
analytical approaches.
There are two substantive problems in identifying associations with DO. One is that it is not
clear which DO parameters (e.g., mean, minimum, % of time below a critical value, duration
below critical value) are critical for survival of organisms and for controlling the responses of
ecological communities. Although most estuarine organisms can tolerate short exposures to
dissolved oxygen concentrations below saturation without apparent adverse effects, prolonged
exposures to less than 60% oxygen saturation (~ 5 ppm) may result in altered behavior,
reduced growth, reduced reproductive success, and mortality (Kramer 1987; Burton et al.
1980; Reish and Barnard 1960). Exposure to less than 30% saturation (~ 2 ppm) for one to
four days is lethal to most biota, especially during summer months, when metabolic rates are
high. Stresses that occur in conjunction with low dissolved oxygen (e.g., exposure to
hydrogen sulfide or ammonia) may cause more harm to aquatic biota than exposure to low
dissolved oxygen concentrations alone (Brongersma- Sanders 1957; Adelman and Smith
1972; Theede 1973). Finally, aquatic populations exposed to low dissolved oxygen may be
more susceptible to the adverse effects of other, unrelated stresses.
The second problem relates to our ability to measure DO parameters. Obviously, duration or
percent of time below a critical value cannot be estimated with a single point-in-time measure-
ment, or even a few such measurements. If the measured DO does not reflect "true"
exposure to low DO, then EMAP's ability to identify associations between low QO and
subnominal values of response indicators will be limited. As shown in Figure 4-4,
4-44
-------
JU90J8d
CO
'3T
1
s
£2
O>
CO
to
•O
I
CD
CO
0>
E
o
O)
"a.
CO
to
O>
11
fi
>. CO
CD E
g-g-
fl) t
CD w
73 E
CO O
o 2
II
U_ i
4-45
-------
bottom water dissolved oxygen concentrations can fluctuate between near saturation and
severe hypoxia in a matter of days (also see Breitburg 1990; Sanford et al. 1990). A single
moderate or high DO measurement may not represent hypoxia that occurred before or after
the sample was taken.
During the Demonstration Project, three sampling strategies were examined for their ability to
characterize DO at a site with sufficient precision for investigating associations. Once again,
Monte Carlo sampling of the LTDO data sets was used to simulate realistic methods of
sampling within EMAP constraints. Sampling strategies were one, two, and three point-in-time
samples and one, two, and three-day continuous sampling. For two and three point-in-time
samples, a random daytime (0800-1800 hours) sample was taken so that samples were
separated by at least nine days to avoid autocorrelation problems. Estimators of the mean
DO at a site were examined, as well as estimators that classified a site as above or below a
threshold DO value (e.g., 25% of the time above 2 ppm). It is not necessary to develop good
estimates of a parameter to identify associations; determining whether the parameter exceeds
a threshold is sufficient to produce a categorical variable, allowing categorical data analysis.
Performance of a sampling strategy in estimating the mean of a time series was evaluated by
determining the frequency of samples that were more than 1 ppm away from the true site
mean. Single point sampling performed the worst for estimating the site mean (52% error rate
above 1 ppm (Fig. 5-9). The one to three day continuous samples were not much better (39%
to 44% error rate). Three single points performed best for estimating the site mean, with a
22% error rate, followed by two single points at a 28% error rate. The reason for the poor
performance of the short term continuous samples is clearly the strong autocorrelation of
dissolved oxygen within a three-day sample period. A three-day continuous sample is slightly
better at estimating the site mean than a single sample, but not as good as two independent
samples.
The threshold used in the analysis was 25% of time below 2 ppm. Three of the 13 LTDO
sites were below this threshold (Fig. 4-10). Two measures were investigated, the mean and
the minimum of the observations. Alternative sampling strategies were tested by determining
their ability to discriminate between LTDO data sets above and below the threshold, and
calculating the misclassification rate for each threshold value for each sampling alternative
(Fig. 4-11). The overall misclassification rate for each strategy was the rate at which false
positive (labeling a good site as bad) and false negative (labeling a bad site as good)
classifications were the same.
Estimators of threshold exceedance performed far better than estimates of the site mean.
The single point-in-time sample was, again, the poorest, with a 26% overall misclassification
rate (Fig. 4-12). Measures based on the minimum of two or more point-in-time observations
performed better. The minimum of three point-in-time samples was the best
4-46
-------
99Jl|JL
9UQ
S
C3)
I
T3
0)
"o
CO
CO
C35
i
CO
-------
o u — < «
1
w WL!
CM t- fr) ?- 2
O" - Ql
• 0 D < m
t90-06VA
992-06VA
I-170-06VA
OSI.-06VA
OCH-06VA
OZO-06VA
90I--06VA
S20-06VA
Icqn-ORV/A
880-06VA
980-06VA
U5 Tfr CO CM T- O
Tuidd) ua6Axn oaAiossin
re 4-10. Dissolved oxygen quantities from minimum to 25%, for 13 LTDO data series.
g>
u.
4-48
-------
c.
O
tt
•*-«
'tn
o
Q.
CD
2
\
5
T3
O)
0
c
CD
"cB
U-
•
•
P
•
•
•
•
uoijeomsseiosiui jo
11
«-D
§
srjf
0) D-
§ CD
0 V)
OS
^: o
8
TJ «=
f
II
Jg W 03
111
511
£ a?i
3, 0 OT
.2s -c cc
U- 5 "o
4-49
-------
30 -i
Single Two Point Three Point
30 -\
CO
_«
O
E 10-
22
One Day
Two Day Three Day
Figure 4-12. Overall misclassification rates for point-in-time sampling (a) and short-term
continuous sampling (b) derived from plots like the one shown in in Figure 4-13.
4-50
-------
overall, with a misclassification rate of 15.6%. Two point samples also performed well,
misclassifying 16.2% of the time, which was slightly better than three-day continuous samples,
with 18% misclassification.
Multiple point-in-time samples performed better than short-term continuous samples because
of the autocorrelation problem in the continuous samples. It is crucial that the single samples
are separated by a minimum time period (in the Monte Carlo test case nine days minimum
separation was used) and that they are randomized. Sampling on a fixed interval, even as
long as nine or ten days, reintroduces the autocorrelation problem.
Optimum sampling of a serially autocorrelated process depends strongly on periodicities in the
process. Sampling should attempt to capture a large part of the variance of the process; if
most of the variance is expressed in a short-term diurnal cycle, then short-term continuous
sampling will be optimal. Dissolved oxygen at many sites in the Virginian Province appears to
have periodicities longer than several days; therefore, two or more point-in-time samples
performed slightly better for characterizing those sites than did short-term continuous samples.
The LTDO data sets used for the Monte Carlo sampling are not a random sample from the
Virginian Province. Few of the sample time series were characterized by diurnal periods, with
the result that multiple samples performed slightly better than short-term continuous samples.
A greater predominance of diurnal periods would make short-term continuous sampling the
better strategy, especially for accurate estimation of status. Three-day continuous samples
are being taken at all sites in the Virginian Province during 1991 and will be evaluated to
assess the importance of diurnal periodicity. The dissolved oxygen sampling strategy for the
Virginian Province will be refined following the 1991 sampling based on an analysis of these
data.
4.2.2 Sediment Toxicity
Sediment toxicity tests are the most direct measure available for estimating the potential for
contaminant-induced effects in benthic communities. These tests provide information that is
independent of chemical characterizations and ecological surveys (Chapman 1988). They
improve upon direct measures of contaminants because many chemicals are bound tightly to
sediment particles or are complexed chemically, making them biologically unavailable (USEPA
1989). Sediment toxicity tests have many applications in both marine and freshwater
environments (Swartz 1987; Chapman 1988). They have become an integral part of many
benthic assessment programs (Swartz 1989), where they are used to establish contaminant-s-
pecific effects. The solid-phase, amphipod, 10-day acute test has been employed in sediment
assessments in Puget Sound (Dinnel 1990), San Francisco Bay (Long et al. 1990), and New
York Harbor (Scott et al. 1990). It is being used currently to examine sediment toxicity in
three NOAA National Status and Trends studies in the Hudson/Raritan Estuary, Long Island
4-51
-------
Sound, and Tampa Bay (D. Wolfe, NOAA- Rockville, MD, pers. comm.). The success with
which this test identifies sediments that are toxic to indigenous biota led to its adoption in the
joint Environmental Protection Agency/Corps of Engineers guidance for dredged material
permitting (USEPA/ACE 1991).
Although amphipod toxicity test methods have gained general acceptance, a number of
factors that affect their application over the broad geographic and habitat range being
assessed by EMAP-E. The effects of sediment grain size must be considered. Most coastal
contamination is found in fine-grained, high organic carbon sediments, and certain test
species have been found to be sensitive to fine-grained sediments (Spies 1989). Similarly,
salinity effects are a potential concern. Sediment exposure tests for EMAP were all conduct-
ed at the same salinity using full strength seawater (30 ppt salt), regardless of the salinity at
the collection site; thus, low salinity sediments were adjusted to full strength salinity for the
test. This increase in salinity may have decreased bioavailability of metal contaminants. In
addition to the potential effects of habitat parameters such as sediment grain size and salinity,
potential effects due to different holding times are also of concern. In a program of this size,
processing samples may take as long as 30 days from the time of collection. The effects of
holding time on results of sediment toxicity tests, however, have not been well-established
(Becker and Ginn 1990).
Chi-square tests were conducted to test the null hypothesis that the relationship between
toxicity and contaminant concentration is consistent over a range of salinity, grain size, and
holding-time categories in the 1990 Demonstration Project data. For this test, the observed
distributions of stations over four conditions (i.e., high sediment toxicity-high contaminant
concentration, high sediment toxicity-low contaminant concentration, low sediment toxicity-high
contaminant concentration, and low sediment toxicity-low contaminant concentration) for each
salinity, grain size, and holding time category were compared to the distribution for the entire
province calculated as a percentage of the cases for each category (Table 4-12). Four salinity
categories were evaluated: less than 0.5, 0.5 to 5, 5 to 18, and greater than 18 ppt. Three
grain size classes were evaluated: less than 20%, 20% to 80%, and greater than 80% silt-
clay. Holding time was evaluated for three categories: less than 14, 15 to 28, and greater
than 28 days. Stations were characterized as having high sediment toxicity if survival was
less than 80% of that in the control. Stations were classified as having high contaminant
levels if the concentration of any one contaminant exceeded the median value for biological
effects (ER-M) given in Long and Morgan (1990).
Significant differences in the toxicity and contaminant distributions were detected for two
conditions: less than 20% silt-clay, and greater than 80% silt-clay (Table 4-12). The
differences in distribution between the grain size categories reflect more cases of low
sediment toxicity and low contaminant concentrations when the sediments are coarse, and low
toxicity with high concentrations of contaminant when the sediments are fine. The latter
indicates that high concentrations of contaminants are less toxic in fine sediments. These
4-52
-------
results are not surprising because contaminant concentrations are known to correlate well with
particle size (i.e., binding capacity is inversely related to particle size).
Table 4-12. Relationship of contaminant concentration to Long and Morgan (1990) ER-M
values and toxicity by site, as a function of salinity, grain size, and holding
time
Condition (N)
Province (141)
Salinity (ppt)
0-0.5 (17)
0.6-5 (16)
6-18 (29)
> 18 (69)
Grain Size (% silt-clay)
< 20 (57)
21-80 (22)
> 80 (55)
Holding Time (days)
< 14 (18)
15-28 (78)
> 28 (44)
High
Toxicity/
High
Contami-
nants
10
1
1
4
4
1
5
4
1
6
1
High
Toxicity/
Low
Contami-
nants
7
0
0
0
6
2
4
1
2
4
0
Low
Toxicity/
High
Contami-
nants
15
3
5
3
3
49
32
21
2
5
7
Low
Toxicity/
Low
Contami-
nants
109
13
10
22
56
49
32
21
13
63
36
Chi
Square
1.62
7.55
2.78
5.60
7.93?
2.75
12.80*
0.61
1 .63
2.28
* Statistically different from expected at p < 0.05.
Additional tests were conducted as part of the 1990 Demonstration Project to evaluate
whether conducting the bioassays at a constant salinity affected the sensitivity of the test.
Thirty-eight sediment samples from freshwater/oligohaline habitats were tested with both
Ampelisca and Hyalella azteca, an amphipod used frequently in freshwater sediment
bioassays (ASTM 1991). The tests with Hyalella were conducted using well water, whereas
the tests with Ampelisca were conducted using 30 ppt water. In only one of these sediment
samples was Hyalella survival significantly less than that in the control. For this same set of
sediment samples, significant toxicity to Ampelisca was found in eight samples. The original
concern was that the addition of salt to the low salinity sediments might cause inorganic
4-53
-------
contaminants to become less bioavailable, which would have caused the Hyalella test to be
more sensitive to toxic sediments than the Ampelisca test. Instead, the Ampelisca test was
found to be the more sensitive of the two, suggesting that the salinity adjustment should be of
limited concern; however, given the limited number of samples where toxicity to either species
was found in this comparison test, additional testing of Ampelisca at low salinities, possibly in
comparison with other freshwater species, is still warranted.
4.2.3 Sediment Contaminants
The presence of contaminants in estuaries has been identified in both the scientific and
popular press as a major problem contributing to degraded ecological resources and restricted
harvest of fish and shellfish resources due to human health concerns (Broutman and Leonard
1988; NOAA 1990; OTA 1987; O'Connor 1990). Reducing contaminant inputs and
concentrations, therefore, is often a major focus of regulatory programs for estuaries.
Contaminants include both inorganic (metals primarily) and organic forms originating from
many sources, including atmospheric deposition, freshwater inputs, land runoff, and point
sources. These sources are poorly characterized, except in the most well-studied estuaries.
Most contaminants that are potentially toxic to indigenous biological resources tend to bind to
particles, which ultimately are deposited at the bottom of estuaries (Santschi et al. 1980;
Santschi 1984). This binding changes the form of contaminants and removes them from the
water column; consequently, contaminants accumulate in estuarine sediments (Santschi et al.
1984; Nixon et al. 1986).
EMAP monitoring efforts have focused on sediment contaminants and have not included
measurement of contaminants in the water column (Holland 1990). Concentrations of
contaminants in sediments are less variable than those in the water column, and the sediment
integrates contaminant inputs to estuaries over months and years. The greater variability of
water column contaminant concentrations is due to interactions between contaminant inputs,
which are themselves variable, and natural processes in the estuary. Measurements of water
column contaminant concentrations made during the EMAP sampling period would
characterize estuarine contaminant concentrations poorly and, therefore, are inappropriate.
Sediment contaminant concentrations were measured to help interpret the spatial patterns
observed in the condition of biological resources in the estuaries of the Virginian Province.
Previous studies have suggested that the incidence of pathological disorders in fish is
generally higher in areas with high concentrations of contaminants in the sediment. The
species composition of benthic communities also may be affected by high sediment
contaminant concentrations; thus, sediment contaminant concentrations can be used as a
diagnostic variable for biological response indicators.
4-54
-------
A categorical approach can be used to interpret sediment contaminant concentration data as
both a diagnostic exposure indicator and an indicator of estuarine condition. Categorical data
analysis is appropriate to identify relationships between sediment contamination and biological
responses because such responses are unlikely to be linear. Categorical analysis is also
appropriate to the needs of environmental managers, whose primary concerns relate to
regulatory or target concentrations.
The typical approach for categorical data analysis involves establishing threshold values for
the concentration of each contaminant of interest; however, determining appropriate threshold
values depends upon the specific question of interest. There are at least three questions of
interest to environmental managers;
• Which contaminants are present due to human activities?
• Where are estuarine sediments most contaminated?
• Are sediment contaminants harmful to indigenous biological resources?
The threshold value for addressing each of these questions differs. Although EPA is in the
process of developing sediment criteria for estuaries, generally accepted approaches for
identifying appropriate threshold values to address these quesitons have not been established.
EMAP is in the process of developing analytical approaches for evaluating sediment
contaminant data to address each of the questions. Three potential approaches are
presented below, followed in Section 4.2.3.4 by a comparison of the results of applying these
methods to evaluate the extent of contamination in estuaries of the Virginian Province.
Presenting these methods now does not preclude using other methodologies in the program
eventually; evaluation of analytical methods is a continuing activity.
4.2.3.1 Identifying Anthropogenic Enrichment
Sites that have experienced anthropogenic enrichment with organic contaminants can be
defined operationally as any sites where there are organic contaminants. Synthetic organics
such as pesticides and PCBs only have human origins, and human activities are the main
sources of most PAHs. There are natural sources of PAHs, such as the oil seeps in the
Southern California Bight, but no such natural sources contribute significant amounts of
organic contaminants to the Virginian Province.
Metals in the sediment may be derived from anthropogenic sources or from the natural
geochemical processes of weathering and erosion of rocks, since metals occur naturally in the
4-55
-------
earth's crust. The difficulty arises in identifying which portion of the total metal content of the
sediment is due to natural processes and which is due to human activities. Several methods
are used to determine whether measured metal concentrations in estuarine sediments
represent natural or anthropogenically-enriched conditions, including analysis of specific grain
size fractions (Voutsinou-Taliodouri and Satsmadjis 1982; Ackermann et al. 1983; Schneider
and Weiler 1984), normalization by grain-size (NOAA 1990), normalization by organic carbon,
and normalization to a reference element (Klinkhammer and Bender 1981; Windom et al.
1989; Trefry et al. 1985; Kouadio and Trefy 1987). Most of these methods were not suitable
for EMAP-E. Use of a specific grain size fraction is not practical because EMAP's probability-
based sampling design requires using the entire population of samples to generate regional
estimates. Similarly, even normalization by grain size requires elimination of sandy samples
(NOAA 1990), and one-third of the samples from 1990 were from sandy sediments. Using
organic carbon concentrations to adjust sediment metal concentrations was rejected because
carbon concentrations are influenced strongly by human activities.
Normalizing concentrations of metals to conservative crustal materials, such as aluminum, has
been used to analyze sediment data from selected coastal regions (Bruland et al. 1974;
Schropp et al. 1990). The method involves establishing a relationship between the
concentration of a particular metal and aluminum. Aluminum-normalized metal concentrations
that are significantly greater than the concentrations established by the relationship are
interpreted to represent anthropogenically-enriched concentrations. Normalizing to aluminum
relies on several characteristics of the metal. Aluminum is highly refractory and does not
degrade or change forms in the environment. Metal to aluminum ratios are relatively constant
in estuarine sediments without human sources, and aluminum concentrations are not
influenced significantly by human activities. Although aluminum normalization techniques
have been used to identify enriched sediments over large geographic areas (Windom et al.
1989; Schropp et al. 1990), they have not been used to establish relationships over an area
as large as the Virginian Province or with the range of metal concentrations found there.
Simple linear regressions based on log-transformed data were used to determine whether
there was a baseline relationship between aluminum and the metals of concern (i.e.,
chromium, copper, lead, nickel, and zinc) in sediments that are not anthropogenically enriched
(Fig. 4-13). Identifying unenriched sediments in the estuaries of the Virginian Province is
challenging because of long-standing and widespread contamination. As a first step, sites
with significant sediment toxicity were considered enriched and removed from regression
calculations. The degree to which this criteria removed all sites already influenced by
anthropogenic sources of metals needs to be evaluated. Additional criteria
4-56
-------
_ o
- CO
00
CO
CM *-
I
1
EL
<
UJ
o
E
o
s*
0] C/J
E-o
2 o
o t
§ (D
"
'w'o
Q r
II
^ 03
W
O> o
1!
fl
« OT
2 S
'•a co
ro
•OUOQ
.
U_ T3
4-57
-------
most likely are needed because there may be other anthropogenically-enriched sites that are
not toxic. The slopes of regressions calculated with such sites may be steeper than if
calculated only with unenriched sites; consequently, the estuarine area with anthropogenically-
enriched metal concentrations may be underestimated. Metal to aluminum relationships and
95% confidence intervals were established for chromium, copper, iron, mercury, manganese,
nickel, lead, and zinc. A relationship was not attempted for silver because values above the
detection limit were measured at only three sites. Correlation coefficients, slopes, and
intercepts for each of the relationships are given in Table 4-13.
Table 4-13. Relationship between sediment metal concentration and aluminum.
Correlation coefficient (r2), slope (m), and intercept (b) values given for the
relationship: log (Me+) =m (log (Al)) + b. Alpha for all relationships except
mercury was < 0.0001 . Alpha for mercury was < 0.002.
Metal
Cr
Cu
Fe
Hg
Mn
Ni
Pb
Zn
Cd
Correlation
Coefficient
0.71
0.57
0.75
0.11
0.40
0.66
0.46
0.63
0.14
Slope
1.042
1.413
0.872
0.681
0.707
1 .095
0.861
1.046
0.613
Intercept
-7.389
-12.376
0.697
-9.440
-1.332
-8.818
-5.976
-6.528
-7.424
The relationships given in Table 4-13 were compared to those published by Schropp et al.
(1990) for uncontaminated sediments from a large area in Florida. Comparisons could be
made for chromium, copper, nickel, lead, and zinc. The metal to aluminum relationships were
most similar for lead, a heavy metal with a predominantly atmospheric input to estuarine
environments. For other metals, the EMAP relationships had greater slopes and smaller
intercept values, but the statistical significance of these differences could not be determined
without obtaining the Schropp et al. data. This finding suggests that some of the sites used in
the EMAP regressions were contaminated by anthropogenic sources of metals. EMAP will
4-58
-------
continue to develop aluminum relationships and means for excluding potentially contaminated
sites from subpopulations of data used to determine metal-aluminum relationships and for the
Virginian Province.
Aluminum normalization relationships also were attempted for mercury and cadmium. These
relationships were very weak, though statistically significant, similar to the observation of
Schropp et al. (1990). For the EMAP data, aluminum explained only about 10% of the
variation in the mercury and cadmium concentrations (Table 4-13). Since the relationship was
so poor, any presence of these elements in the sediment was assumed to result from
anthropogenic sources. This assumption may be valid for mercury relative to the
measurement detection limits of the laboratory procedures used, but it most likely
overestimates the anthropogenic contribution of cadmium.
4.2.3.2 Identifying "Above Background" Concentrations
Some contaminants are present throughout the province at relatively low levels. Although the
concentrations of these contaminants are high enough to suggest anthropogenic enrichment,
the source of enrichment may be atmospheric or diffuse rather than discrete. One interest of
managers is to identify "hot spots" resulting from local sources or habitat conditions that
concentrate selected chemicals.
Contaminant concentration data for most analyses are log-normally distributed; a high
percentage of sample sites has low values, and a lesser percentage has very high values.
Selecting an inflexion point to define background in this distribution is a difficult and somewhat
arbitrary decision. The most widely used approach for selecting this inflexion point is that of
NOAA's NS&T program (NOAA 1990), which uses one standard deviation above the
geometric mean.
Applying this approach to EMAP data is problematic; the same data set can not be used to
define both the "background," or threshold, concentrations for contaminants and to estimate
estuarine condition. An independent data set is required to define threshold concentrations so
that the choice of thresholds from the distribution will not directly determine the estimate of
estuarine condition. For example, if the EMAP Virginian Province data set was used to define
a threshold as the concentration of the 90th percentile of the distribution, then approximately
10% of the sites and 10% of the estuarine area would have high contaminant concentrations.
To resolve this difficulty, sediment data from NOAA's National Status and Trends (NS&T)
program were used to define the threshold values. Following the lead of previous analyses
4-59
-------
of these data (NOAA 1990), the mean and standard deviation were determined for each
contaminant after first transforming the data using a logarithmic function:
transformed value = ^(concentration + 1).
Logarithmic data transformations made the characteristically log-normal distribution of the
contaminant data more gaussian. The "background" value was defined as the mean plus one
standard deviation, and the antilog of this sum was used as the threshold value. EMAP
sampling sites with contaminant concentrations exceeding these values were classified as
having high contaminant concentrations.
4.2.3.3 Determining Biologically Significant Concentrations
Environmental resource managers are concerned with the effects of contaminant concen-
trations on the condition of indigenous biological resources, regardless of whether the source
of the contaminant is natural or anthropogenic. There are no generally accepted sediment
quality criteria, despite an extensive literature linking ecological changes from the cellular level
to the community level to sediment contaminant concentrations. Various approaches have
been used to determine threshold contaminant concentrations for biological effects (Hinga
1992, Long and Morgan 1990; Ginn and Pastorok 1992). These approaches have included
relating contaminant concentrations to sediment toxicity test results, benthic infaunal
community responses, or both (apparent effects threshold and sediment triad approaches), or
have used water quality criteria to determine the toxicity of theoretical pore water contaminant
concentrations at equilibrium with sediment contaminant concentrations (equilibrium
partitioning approach, USEPA 1990).
Sediment toxicity, benthic infaunal community abundance, and biomass data from the
Demonstration Project have not been used to define threshold values for sediment
contaminants because the covariance between multiple contaminants was high and the
potential effects of individual contaminants could not be separated from the effects of multiple
contaminants. Threshold values have been determined from other studies using the various
approaches mentioned above. These have merit and will be considered further for application
to the EMAP data; however, most examples of threshold values were developed in restricted
geographic regions without the range of sediment types, contaminants, or contaminant
concentrations represented in the data set for the Virginian Province.
The databases used by Long and Morgan (1990) to compile derived effects values are
exceptions. In this investigation, all studies that have demonstrated some type of bjplqgical
effect were ranked by the concentrations of individual contaminants. Two values were
identified for each contaminant: an effects range-low (ER-L) value corresponding to the
concentration of the 10th percentile, and an effects range-median value corresponding to the
4-60
-------
median (ER-M) concentration, or mid-point, of the ranking. The Long and Morgan ER-L
values were interpreted to correspond to threshold contaminant concentrations above which
biological effects begin to appear. Except for sensitive species and life stages, these
biological effects are likely to be chronic, sublethal effects; therefore, evaluations of sediment
contaminant concentrations using the ER-L values provide early warning indicators of pollutant
exposure before contaminant concentrations cause widespread acute toxicity. The ER-L
values were used in the evaluation of environmental condition of the Virginian Province
(Section 2) to estimate the area with contaminants at concentrations having the potential to
cause sublethal biological effects; however, as Long and morgan (1990) have noted, acute
effects are also possible at ER-L concentrations, particularly to sensitive species and life-
stages.
ER-M values were interpreted to correspond to contaminant concentrations above which bio-
logical effects are not only possible, but probable. Acute biological effects are more likely to
occur at concentrations exceeding ER-M values than at concentrations between ER-L and ER-
M values. These interpretations of the Long and Morgan ER-L and ER-M values are similar to
those made by Klapow and Lewis (1979), who used a similar analytical approach to assess
contaminant concentrations in marine waters. The potential limitation of these interpretations
is that the actual effect on biota will depend on the specific taxa present at a site and habitat
characteristics of the site that might affect bioavailability of the contaminants.
ER-L and ER-M values are not available for all of the contaminants measured by EMAP
(Table 4-14). Values were available for all of the measured metals, 6 of the 15 pesticides, 12
of 23 PAHs, and total PCBs. Contaminants for which there were no ER-L and ER-M values
were excluded from the analyses presented in Section 6. This approach may underestimate
the number of sites with concentrations of contaminants high enough to cause biological
effects. It may also underestimate the prevalence of high organic contamination relative to
metals because none of the contaminants for which threshold values are unavailable were
metals measured by EMAP. The magnitude of these underestimates appears to be slight,
however, since several contaminants in a class for which at least one ER-L value was
available generally exceeded that value at a site.
4.2.3.4 Comparison of Analytical Approaches
The threshold values for each contaminant derived by several different methods are compared
in Table 4-14, Threshold values are not presented for the anthropogenic enrichment
methodologies, since this approach does not employ a single value. For organic
contaminants, however, the reader is referred to the median detection limit in Table 4-4. *
4-61
-------
Table 4-14. Comparison of threshold contaminant levels based upon the "above
background", ER-L, and ER-M values. Above background values were
defined from NOAA National Status and Trends Sediment Data Base. ER-L
and ER-M values taken from Long and Morgan (1 990). Values not available
for contaminants marked with an asterisk (*).
Metals
Cadmium
Chromium
Copper
Lead
Mercury
Nickel
Zinc
Pesticides
Aldrin
Alpha-Chlordane
Trans-Nonachlor
Dieldrin
DDT - Total
o.p'-DDD
p.p'-DDD
o,p'-DDE
p,p'-DDE
o.p'-DDT
p.p'-DDT
Heptachlor
Heptachlor epoxide
Hexachlorobenzene
Lindane
Mirex
Polychlorinated Biphenyls
"Above
Background"
Values
(ppm)
0.7
140
45
46
0.4
40
150
(PPb)
0.4
1.0
1.0
1.1
14
1.4
4.7
1.5
6.1
0.4
1.9
0.2
0.3
1.1
0.5
0.4
28
ER-L
Values
(ppm)
5
80
70
35
0.15
30
120
(PPb)
*
0.5
*
0.02
3
*
2
*
2
*
1
*
*
*
*
ft
50
ER-M
Values
(ppm)
9
145
390
110
1.3
50
270
(PPb)
*
6.0
*
8.0
350
*
20
- *
15
*
7
*
*
*
*
*
400
4-62
-------
Table 4-1 4. Continued
Polycyclic Aromatic Hydrocarbons
Acenaphthene
Acenaphthylene
Anthracene
Benzanthacene
Benzo(a)pyrene
Benzo(e)pyrene
Biphenl
Chrysene
Dibenz(a,b,)anthracene
2,6,-Dimethylnaphthalene
Fluoranthene
Fluorene
ldeno(1 ,2,3-c,d)pyrene
1 -Methylnaphthalene
2-MethylnaphthaIene
1 -Methylphenanthrene
Naphthalene
Perylene
Phenanthrene
Pyrene
Benzo(k)fluoranthene
Benzo(g, h, i)perylene
2,3,5-Trimethylnaphthalene
"Above
Background"
Values
(ppb)
4.5
10
26
94
100
98
5.3
133
12
10
229
11
42
10
19
22
29
97
113
222
59
47
4
ER-L
Values
(PPb)
150
*
85
230
400
*
*
400
60
*
600
35
*
*
65
*
340
*
225
350
*
*
*
ER-M
Values
(PPb)
650
* ;
960
1600
2500
*
*
2800
260
-*
3600
640
*
*
670
*
2100
; ' *
1380
2200
*
*
*
Estimates of the estuarine area with contaminant concentrations of concern were generated
using each of the approaches discussed above. Based upon data from the Demonstration
Project, almost all of the estuarine area in the Virginian Province showed some evidence of
chemical contamination due to anthropogenic activities (Table 4-15). Contamination from
organics was most widespread; PAHs were found in nearly all areas (98%). Contaminant
concentrations exceeding "background" values were found in 40% of the estuarine area in the
Virginian Province. If estuarine areas with high contaminant concentrations were distributed
equally around the country, then the expected value, based upon the distribution of the NOAA
NS&T data, would have been approximately 17%. The observation that the actual number
was over two times larger suggests that the estuaries in the Virginian Province generally have
higher contaminant concentrations than estuaries in other parts of the country.
4-63
-------
As the question changes from presence or presence above background levels to biological
significance, the proportion of estuarine area in the Virginian Province with contaminant
concentrations of concern decreases, and metals become of greater concern than organics
(Table 4-15). The area with contaminant concentrations having the potential to cause
sublethal biological effects (ER-L values) was 39%. In this area, biologically relevant
concentrations were due mainly to elevated concentrations of metals (36%). Biologically
significant concentrations of organics were more restricted spatially. This finding and
comparisons to "background" values suggest that, although contamination from organic
compounds may be more pervasive, contamination from metals may have proportionately
greater potential for biological effects. This finding also is reflected in analyses using ER-M
values as assessment thresholds. The proportion of estuarine area with contaminant
concentrations with the potential to cause more acute biological effects (ER-M values) was
7%; most of the elevated concentrations were due to metals.
The extent of contaminant conditions for each of the three classes of estuarine resources also
was estimated using the assessment methods outlined above (Figure 4-14). Estimates of
contaminated areas varied greatly between the methods. Regardless of method, however, a
greater proportion of the estuarine area within large tidal rivers and small estuarine systems
had contaminant concentrations of concern when compared to the estuarine area in large
estuarine systems.
Following the pattern observed for the province as a whole, nearly all of the estuarine area in
all three classes of estuarine resources showed some evidence of chemical contamination
due to anthropogenic activities. Sediment contaminant concentrations above background
levels were found in 73% of the area in large tidal rivers and 61 % of the area in small
estuarine systems. Contaminant concentrations above background were less prevalent in
large estuaries. A similar pattern was observed using ER-L values as assessment thresholds.
A slightly different pattern was apparent using ER-M values. Sediments with contaminant
concentrations that are potentially acutely toxic to biota were found in 24% of the estuarine
area in small estuarine systems, but were present in less than 5% of the area in large tidal
rivers or large estuarine systems (Figure 4-14). These findings suggest that significant
sources of contaminants may be present in both large tidal rivers and small estuarine
systems. The short residence times of water in large tidal rivers generally minimizes
accumulation of extremely high contaminant concentrations in these systems. The longer
residence times in small estuarine systems may increase the opportunity for contaminants to
become trapped in the sediments, in some cases reaching concentrations sufficient to initiate
significant biological effects. The proportionately higher occurrence of sediments with
contaminant concentrations exceeding ER-M values is consistent with the finding that
sediments that were toxic to laboratory test organisms
4-64
-------
Table 4-15. Estimated percent of the area in the Virginan Province exceeding each of
four threshold sediment contaminant concentrations.
Contaminants
All
Metals
All
Cadmium
Chromium
Copper
Lead
Mercury
Nickel
Zinc
Pesticides
All
Aldrin
Alpha-Chlordane
Trans-Nonachlor
Dieldrin
DDT - Total
o,p'-DDD
p,p'-DDD
o.p'-DDE
p,p'-DDE
o,p'-DDT
p,p'-DDT
Heptachlor
Heptachlor epoxide
Hexachlorobenzene
Lindane
Mi rex
Polychlorinated
Biphenyls
Enriched
97.7
96.3
96.3
4.2
4.7
3.3
65.3
3.5
1.4
22.3
0.7
13.0
0.1
5.7
5.7
1.0
11.8
1.7
14.4
0.1
9.9
2.0
0.5
5.7
0.3
5.8
39.8
"Above
Background"
Values
40.3
30.3
13.3
3.1
14.1
16.9
4.3
8.9
19.7
10.1
0
4.1
0.1
0.6
0.5
0.1
0.9
0.4
0.9
0.1
3.2
1.0
0.1
0.1
0.3
5.8
6.2
Exceeding
ER-L
Values
39.1
36.5
3.1
15.2
10.2
27.2
21.7
15.7
24.9
12.2
*
5.9
*
5.7
5.7
*
2.1
*
4.5
*
6.2
*
*
*
*
*
1.0
Exceeding
ER-M
Values
8.0
6.9
0
3.2
0.2
5.2
0.7
4.3
5
1.3
*
0
*
0
0
*
<0.1
*
0.4
*
0.9
*
*
*
*
*
<0.1
4-65
-------
Table 4-1 5. Continued
Polycyclic Aromatic
Hydrocarbons
All
Acenaphthene
Acenaphthylene
Anthracene
Benzanthacene
Benzo(a)pyrene
Benzo(e)pyrene
Biphenyl
Chrysene
Dibenz(a,b,)anthracene
2,6,-
Dimethylnaphthalene
Fluoranthene
Fluorene
Ideno{1 ,2,3-c,d)pyrene
1 -Methylnaphthalene
2-Methylnaphthalene
1 -Methylphenanthrene
Naphthalene
Perylene
Phenanthrene
Pyrene
Benzo(b,k)fluoranthene
Benzo{g,h,i)perylene
2,3,5-
Trimethylnaphthalene
Enriched
98.1
30.8
19.7
33.9
85.4
37.7
51.8
24.3
81.9
6.1
21.1
69.3
30.5
9.3
45.9
62.8
41.2
96.5
55.1
91.2
71.9
53.1
19.5
6.5
"Above
Background"
Values
27.4
15.4
1.1
2.8
4.1
2.5
3.0
6.3
8.4
4.6
2.5
8.3
7.9
6.7
8.5
6.4
3.7
15.6
1.9
8.3
3.7
5.7
1.9
1.8
Exceeding
ER-L
Values
3.7
0.2
*
2.1
3.2
0.9
*
*
2.9
1.1
*
3.1
2.5
*
*
3.2
*-
1.3
*
3.3
3.5
*
*
*
Exceeding
ER-M
Values
0.2
0.2
*
0.2
0.2
0.2
*
*
0.2
0.2
*
0.2
0.2
*
*
0
*
0
*
0.2
0.2
*
*
*
were more prevalent in small estuaries than in either large tidal rivers or large estuarine
systems.
The discussion above illustrates that analytical approaches to assessing sediment contami-
nant concentrations depend greatly on the primary question of interest. Further, assessment
results are strongly influenced by method. EMAP intends to work closely with the potential
4-66
-------
users of program results to refine assessment methodologies, aiming to identify
methodologies that produce the information most relevant to the needs of users.
4-67
-------
SSBJO u| Bejv jo lueoied
I
I
UJ
5
8 §
a:
UJ
i
g
CO
§88-98
SSBJO U| B8JV JO JU90J9d
T3
I
^
0)
i
CO
o
o
10
S?
^ a
s«
(D 2
CO Cfi
s £
CO £
03
Q. -
§8
4-68
-------
4.3 HABITAT INDICATORS
Habitat indicators were not used to distinguish degraded from nondegraded environmental
conditions. Instead, habitat indicators were used to describe the physical estuarine
environment and to normalize selected response indicators. The three principal habitat
indicators were depth, sediment grain size, and salinity.
4.3.1 Depth
Depth was considered to be a stable indicator over the index period and over multiple EMAP
sampling cycles. This stability was used to evaluate how well field crews could return to the
same sampling site and repeat the same measurement. Separate CDFs were constructed for
depth data from sampling intervals 2 and 3 using sampling sites visited during both intervals.
The CDFs for these two intervals, which are shown in Figure 4-15, were not statistically
different. The degree of similarity between the two CDFs strongly suggests that the methods
employed in the 1990 Demonstration Project are adequate to ensure reproducible results.
4.3.2 Salinity
Salinity can vary dramatically at a site within hours as a function of tidal stage, and within
weeks as a function of rainfall; consequently, point measurements taken during a single
EMAP visit will be inadequate to describe the range of salinity conditions occurring at that site
throughout the summer. Similar to dissolved oxygen, however, the salinity profile for the
province may remain stable over the summer sampling period. To determine the stability of
salinity on a province level, CDF's for intervals 2 and 3 were compared, and no significant
differences were found between these two periods (Fig. 4-16). Presumably this results
because salinity varies with tide stage at a site, whereas the total salt within the province as a
whole remains relatively stable over the summer period. The randomized sampling design,
therefore, is an appropriate means for describing conditions in the province, even if it is
inadequate for describing the range of conditions at a particular site.
4-69
-------
£
"a.
o
•o
CO
o
-------
\
x^* ••-••.,
o
•JO
"o
CM
JO
CO
CD
4—*
O)
"a.
CO
CO
o
c
"
I
=3
E
o
.co
CO
o _^^
E ^
o .£
4 -a
u. co
B9JV
4-71
-------
-------
SECTION 5
EVALUATION OF THE SAMPLING DESIGN
One of the objectives of the 1990 Demonstration Project was to evaluate whether the EMAP
sampling approach is adequate to assess status and trends in estuarine resources with an
acceptable degree of confidence. An additional objective was to identify which attributes of
the design should be altered to improve the confidence in assessments of status. The
Demonstration Project included study elements to address the following questions about the
sampling design:
• What is the appropriate time frame for sampling?
• Should sampling be stratified based on salinity, substrate type, or other
system features?
• What is the precision of the status estimates?
• Can the precision of status estimates be improved? If yes, how?
• Can the estimation of precision for status be improved? If yes, how?
• What is the power of the design for detecting changes in status (i.e.,
trends) over time?
• Should index sites be included as part of the design?
This section discusses each of these questions with regard to results of the 1990
Demonstration Project.
5.1 DETERMINATION OF THE APPROPRIATE TIME FRAME FOR SAMPLING
EMAP does not have the resources to sample adequately on all time scales of interest for
understanding biological responses to stress in estuaries; therefore, sampling was confined to
a portion of the year when living resources and ecological processes were expected to show
the greatest response to anthropogenic stress (i.e. an index period). In the Near Coastal
Program Plan for 1990: Estuaries (Holland 1990), summer was identified as the most
appropriate index period for the Virginian Province because it is the time when 1) dissolved
oxygen levels are most likely to approach critical low values; 2) contaminant exposure is likely
to be greatest because of low dilution flows and peak metabolic activity; and 3) living
resources are most abundant, maximizing the probability of collecting organisms required for
5-1
-------
assessments. Sampling too early in the summer, before response indicators have been
exposed to the lowest dissolved oxygen concentrations, would result in underestimating the
extent of degraded ecological conditions. Sampling too late in the summer, after recovery
processes have begun, also would result in underestimating degraded ecological condition.
To define the boundaries of the index period for the Virginian Province more precisely, the
stability of two indicators, dissolved oxygen concentration (exposure indicator) and the benthic
Index (response indicator), was examined across three sampling intervals (20 June to 18 July,
19 July to 31 August, 1 September to 22 September). These indicators were selected for
examination because they represent 1) both biological and physical processes, and 2) two of
the major categories of indicators measured in EMAP monitoring.
The stability of bottom dissolved oxygen measurements and the benthic index was evaluated
using a paired t-test to compare indicator responses between sampling intervals.
Comparisons between intervals 1 and 2 were conducted separately from those between
Intervals 2 and 3. Comparisons were limited to sites sampled in both intervals. The
comparisons were made for the province as a whole and, separately, for stations north and
south of the Hudson River. The latter analysis was conducted to identify any latitudinal
differences in the boundaries of the index period. In addition, the semi-continuous data on
dissolved oxygen concentration were used to calculate weekly mean and minimum dissolved
oxygen values for 16 stations that had relatively complete data records. These results were
examined for temporal patterns that might not be detected using only the point measurements
taken at each station in each sampling interval.
Point measurements of bottom dissolved oxygen concentration differed significantly between
intervals 1 and 2 (Table 5-1). The lowest weekly average and weekly minimum bottom
dissolved oxygen values at most stations occurred in early August (Tables 5-2 and 5-3).
Average bottom dissolved oxygen concentration did not differ significantly between intervals 2
and 3 (Fig. 5-1, Table 5-1). These findings are based only on point measurements, since
continuous bottom dissolved oxygen measurements were not made during interval 3. The late
summer continuous bottom dissolved oxygen records, however, suggest that bottom dissolved
oxygen concentrations were beginning to increase at many stations, especially those north of
the Hudson, toward the end of August.
The benthic index did not differ significantly between the sampling intervals (Table 5-1).
Although bottom dissolved oxygen concentrations decreased from interval 1 to interval 2, the
sites with low bottom dissolved oxygen and poor benthic conditions presumably had sufficient
exposure to low dissolved oxygen stress by June for benthos to respond, or the degraded
benthos at these sites reflected previous exposure to low dissolved oxygen stress (i.e., the
previous year).
5-2
-------
Table 5-1. T-test comparison of dissolved oxygen concentrations and the benthic index
among periods. Values in the table are p-values.
Dissolved Oxygen
Benthic Index
Interval 1 vs. 2
0.04
0.48
Interval 2 vs. 3
0.71
0.28
The stability of the benthic index over much of the summer period permits future field efforts
to be scheduled to characterize the temporally variable dissolved oxygen exposure accurately.
Based on the dissolved oxygen patterns, future sampling in the Virginian Province will not be
initiated until late July and will be scheduled for completion by the end of August; however,
given the similarity in exposure between intervals 2 and 3, samples that cannot be obtained in
August can be collected through the third week of September, if necessary.
It is inappropriate to define precise boundaries for the index period in the Virginian Province
based on a single year of data. Year-to-year variation in meteorological patterns may cause
between-year differences in index period boundaries; therefore, deployments to measure
dissolved oxygen continuously were conducted from June to September of 1991 and may be
conducted in future years. In 1991, meters were placed at four locations that were determined
to be representative of the Virginian Province: two in the northern half of the province and two
in the southern half. These data will be used to estimate year-to-year variation in the stability
of patterns of bottom dissolved oxygen concentration. In addition, the index period boundaries
will be reviewed periodically to determine whether the sampling period needs to be adjusted
as new indicators are incorporated into the program. The index period boundaries defined by
this study are applicable only to the Virginian Province. EMAP-E will need to define province-
specific index period boundaries before initiating sampling in other provinces.
5.2 STRATIFIED SAMPLING
Estuaries vary in size, shape, and ecological characteristics. Many, like the Chesapeake Bay,
are large, continuously distributed resources that consist of expansive regions of relatively
homogeneous habitats; others (e.g., small bays, inlets, and salt ponds) are relatively discrete
resources composed predominantly of one habitat type (e.g., high salinity muds). It would not
be cost-effective to sample such vastly different estuarine types with a sampling design that
treated all estuaries the same. Excessive numbers of
5-3
-------
"el
'•*5
>.-—
> £
i §
S m
% ID
CO "8
3 2
to co
«— c
42 o
.j •*-£
«5 $
CD OT
sJ* CO
-O CD
"£* rt
O **^
O co
>^^ CD
C —'
0 13
!~
£5 CO
C CD
CD 5
o -J
O
C ^
CD 12
C3) O
Is
^a
> 8
to .2* *5
&£ c?
2 ^--i
CD E to
£ 81
C\I
10
CD
s
•ft
E
o
Q
O
c
Ul
^
Ul
£
.>•
-3
"S
m
c
c
8s
m
CD
CO
It)
m
CM
T-
1
E
c
o
•s
System and Sti
5
o
CM
"*
CD
CO
'«*
T~.
,_
CM
,_
CM
O)
CM
CM
CO
T—
"*
CM
in
^^
&
•o
o
to
•g
CO
CO
D)
o
_)
UJ
"jfi
CM
CM
m
CO
r--
co
^
CO
^^
CM
•o
C
=}
o
CO
TJ
]ro
OJ
CD
T3
^
CM
"*
CO
•*
n
^
00
CO
CM
m
CO
in
CM
m
CM
in
o"
li
is
m
i
CO
1
CO
CD
l^
CM
CO
^
;tn,
'•s <
o
in
fs*
r--
0
CO
••*
CO
CO
o"
o
§
be
13
SI
CD
to
*i
<9
o>
•*
CD
"*
o>
•*
o
*
CO
h--
CM
•*
in
in
m
CD
CD"
o
T—
1
DC
o
V
•<•£
in
in
CM
CO
CM
I"-
co
in
h-
in
^
00
CM
t"-
I
I
i
c
o
&
^.
CM
N
CO
CO
CO
<9
w»
*
m
CO
^*
^*
o
m
CM
in
12.
jd
0
^
apeake Bay-Mainstem
CO
CD
_c
O
'•Ir
'f;
'**
00
in
C35
•*
CM
N-
00
CD
m
N
^
CD
05
CO
tD
^
CO
J3
O
S
(0
CD
1
m
.i
CO
CD
Q.
m
TO
CD
6
"fa*
l^'
CM
CM
m
T~
o
CM
O
CM
^<
CM
h-
CM
•*
CM
CO
co"
00
CD
>
rr
s:
0)
_Q
.s
UJ
^
m
&
CO
CD
Q.
O
*
,"*
^.
in
^_
10
in
in
^.
in
i^-
in
0
CO
O5
in
in
CO
^P
^ — *
T3
ZI
o
apeake Bay-Tangier S
1
6
00
CM
in
CM
*S
H
CO
CM
00
CO
(T)
f-
CO
in
cp_
1
O
S
apeake Bay-Mainstem
8
6
CO
in
o
in
CM
r--
in
•*
-9
m
CO
^
Tf
CO
^f
CO
05
TT-
c
o
D.
-a
_co
CO
DC
o
o
o
DL
0
CO
CM
co
i*0
'•>#&
CO
CO
CM
00
o
o>
c\r
05
ahannock River-TFO (
DL
8-
tr
CM
*~
in
CO
CO
CM
00
CO
0
"*
h-
CM
'(W:
'•''^
T—
'S'
ccT
oo_
1
DC
.«
S
c
*»
',•&',
CD
in
0
tn
co
CO
r-
CD
(O
m
o
in
o
CO
in
co'
i River Bay (150)
.2
in
in
^.
in
n
nil
^
in
CD
in
r--
in
CO
m
o
05
CVl
^1
CD
tr.
CD
"55
Q
5-4
-------
S
25.1
la
"ti T*1
rS
.c
> 0)
0 g
o> >
X — 1
•o co
O tJ
Table 5-3. Weekly minimum di
complete Hydrolab
missing data.
f
o
LJ
:•=
3
s
;=
>j
\j>
Ul
ll
Ul <
Jjjj
Ul
^
>,
3
0)
'E
'5>
en
Ui
O>
CO
r-
to
in
**
CO
L.
E
3
z:
,o
1
a
I
CO
|| Northern
CM
CO
CO
f •
K£^
t-
^
^j
^j.
T-
CM
CM
^
CM
CO
CO
^
^
S
1
TO
*4
^
f
fmi
,_
CM
CM
CO
CO
CO
in
CO
CM^
•o
o
CO
CO
_w
_JO
00
CM
CM
f
Ci^
^
in
0
•t
K
"*
CM
^
CO
Narragansett Bay (70)
CO
CO
CO
'<¥
''$'
f f
V>.i
O)
^
CO
^
•*
CO
T-
co
T—
o
1
DC
k_
CO
o
CM
CO
CD
'//,':
T~
4-
^.
0
in
o
O)
*
CO
CM
^.
CM
CM
CO
s
T—
1
oc
o
'lo
^
T—
CO
CO
T~*
^
CO
Tj.
s
at
t-
^
O)
1~
CM
h-
co
co
,—
^
^
' f
ff f
CO
CM
CO
in
CO
o
^~
gv
in.
1 Chesapeake Bay-Mainstem Mouth
CO
c\i
0
CO
m
'-r~~
f *
"'A
O)
CM
^
00
CM
CO
O5
"*'
co
i~-
51
s?
Chesapeake Bay-Near Mobjack Be
o
0
CD
/Ji"'
#
CM
0
^>
O
CM
O
00
O
^.
O
T_
CO
co
m
1 Chesapeake Bay- Elizabeth River (
^i
»•;
in
CO
CD
CM
co
CM
CO
"*
CM
CO
^
CO
CO
"3"
T+-
Chesapeake Bay-Tangier Sound (*
^.
0
CO
CD
O
0
O
0
O
- "''„
1
O
,_
o
(Chesapeake Bay-Mainstem Near
Choptank (65)
CO
CO
in
o"
CO
CO
0
CO
*>*
0
f '.
CO
0
CM
f-
co
•^
s
o>
T-
1
I
CO
1
DC
O_
00
CO
m
f S
<8'
ff t
\~,
•*
00
(O
00
CO
CO
h-
Rappahannock River-TFO (192)
m
o
CD
iv-
?
O)
O
1—
CM
CM
T—
O
0
,_
CM
|| Anacostia River (88)
»j
*~_:
co'
T—
CM
O
m
0
m
in
CO
N-
CM
0
"•"
CM
^"
1
m
if
.§
1
m
CO
in
CM
*t
%L
0>
Tj-
^>
in
^
in
in
m
CM
00
1 Delaware River (223)
5-5
-------
CO
Q.
"'o
TJ
-------
samples would be collected from extensive and abundant resources, and rare resources
would not be represented adequately. The present EMAP sampling design for estuaries strati-
fies them into three size classes (i.e., large estuaries, tidal rivers, and small estuaries) that
have similar physical features and are likely to respond to stress in similar ways. Stratification
allowed the sampling design to be customized for the specific geographic features of each
class so that sampling effort was sufficient to assess the extent of degradation for specific
subpopulations with small total areas.
The preliminary assessment results presented in Section 6 show that the extent of degraded
area differs substantially among the three estuarine classes. Small estuaries and tidal rivers
have a higher proportion of degraded area than large estuaries. These differences are
sufficient to justify allocation of a disproportionate number of samples to small estuaries and
tidal rivers, because regulators and resource managers may want to target these types of
systems for further research and remediation activities.
Supplemental stratification schemes considered for EMAP sampling in estuaries have included
stratifying by salinity, substrate type, or specific estuarine system (e.g., Chesapeake Bay,
Long Island Sound). The rationale for further stratification by substrate and salinity is that
estuarine biota respond differentially to habitat parameters, as supported by the analyses
presented in Section 6. In addition, pollution exposure is affected by many of these habitat
parameters. For example, contaminants tend to accumulate in fine grained sediments, and
low dissolved oxygen concentrations are more likely to develop in mesohaline habitats that
have two-layered circulation. Stratification based on salinity and substrate characteristics
would enable sample sizes in each stratum to be based on precision requirements.
The stratification process requires establishing a sampling frame prior to sampling; misclas-
sification of sites within a class should be minimal. This was a major reason for choosing a
stratification scheme based on estuary size. Size data were readily available, and the
probability of misclassification was low. Stratification by substrate was considered to be
difficult because detailed maps of sediment characteristics are not available for much of the
Virginian Province. Even less information is likely to be available about the regional
distribution of substrate characteristics in other regions of the country. A priori stratification by
salinity also was considered problematic because the spatial extent of salinity classes can
vary substantially from year to year, depending on rainfall patterns and, in some estuaries
such as San Francisco Bay, water management practices.
To estimate the degree of misclassification that would be expected if strata were defined by
substrate type, the NOAA database of sediment characteristics for estuaries of the East
Coast, the most comprehensive database in existence on substrate distribution for the
Virginian Province, was used to project the silt-clay content for each of the EMAP sampling
sites visited in 1990. The estimates of silt-clay content were organized into three classes: 1)
sand - less than 20% silts and clays; 2) mixed sediments - 20% to 80% silts and clays, and
5-7
-------
3) mud - greater than 80% silts and clays. The projected data was compared to the actual
substrate type identified by sampling in 1990 to estimate the classification efficiency of the
historical data.
No prediction of sediment type was possible for about one-third of the EMAP sites sampled in
1990. Most of these sites were located in small estuaries that are poorly represented in the
NOAA historical database. The predicted substrate type differed from actual measurements
for more than 50% of the remaining sites (Table 5-4). Stratification by substrate type,
therefore, would be highly impractical given existing knowledge of sediment distribution in the
Virginian Province.
Table 5-4. Comparison of sediment type predicted from NOAA database with that
measured at EMAP sampling sites. Numbers in the boxes represent number
of sampling sites. Correct classifications occur on the diagonal.
PREDICTED
Sand Mixed Mud
MEASURED
Sand
Mixed
Mud
24
20
6
8
28
10
2
25
9
Marginal Misclassification Rate 52%
39%
75%
Overall Misclassification Rate = 54%
There are several possible reasons for the high misclassification rate for substrate type:
1) the available substrate data for the Virginian Province are not sufficiently detailed relative to
sediment patch size; 2) the extrapolation techniques are not sufficiently advanced to project
substrate type in the Virginian Province adequately; or 3) the imprecision in EMAP
measurements was large. To determine whether measurement error for EMAP samples was
large, substrate information was compared at sites that were sampled in both intervals 2 and
3. Less than 10% of the sites changed sediment classes between the periods. These
findings and QA/QC checks indicate that the measurement error in EMAP sediment grain size
analysis was low; therefore, explanations 1 and 2 above are more likely to be responsible for
the high misclassification rates.
5-8
-------
To estimate the potential misclassification rate associated with stratification by salinity, salinity
data from Maryland's Benthic Monitoring Program in the Chesapeake Bay (Holland et al.
1989) were examined for the period 1984 to 1988, and the percentage of sites that switched
salinity classes among years was calculated. Three salinity classes were defined for this
analysis: 1) brackish (less than 5 ppt), 2) transitional (5 to 18 ppt), and 3) high salinity
(greater than 18 ppt). Comparisons were limited to August data, since this is the period
targeted for future sampling in the Virginian Province. More than one-third of the sites
switched salinity classes over the five-year period. Since the analysis was conducted using
only five years of data, it probably underrepresents the rate of misclassification that might
occur over several decades (a more appropriate period for EMAP planning). If Long Island
Sound data were used, the percent of sites switching salinity class presumably would be
much lower because the Sound is characterized by much less year-to-year salinity variation
than the Chesapeake Bay.
Stevens and Olsen (1992) investigated the impact of misclassification on the precision of an
estimated CDF when the sample was allocated to optimize the estimate of the proportion of a
population that is below a critical value xc. They concluded that the potential benefit from
optimal allocation is small, and even under ideal circumstances, precision is improved only in
the immediate neighborhood of xc. Under moderate misclassification (greater than 5%), the
attempted optimal allocation was worse than proportional allocation.
The approach of Stevens and Olsen was applied to hypothetical distributions of the benthic
index (Bl) to illustrate these points. The population of benthic index values was assumed to
consist of a mixture of two normally distributed subpopulations, one with mean of 4.5 and
standard deviation of 0.5, the other with mean of 7 and standard deviation of 1.5. The first
subpopulation was centered on the critical value between degraded and nondegraded and
could be considered degraded. The second subpopulation approximated the observed
distribution of the benthic index for the entire Virginian Province (Fig. 5-2). The behavior of
optimal and proportional allocation was compared by calculating and plotting the width of 90%,
one-sided confidence intervals over the range of the response for a sample size of 50.
Allocation was optimized at the critical value x,, using the equations given by Cochran (1977).
The entire population was assumed to be a 50/50 mix of the two subpopulations.
This scenario was selected for close to maximal potential benefit from optimal allocation when
complete, error-free information on population distribution is available. The critical value was
set at xc = 4.5, the mean of population 1, so that F^xJ = 0.5, F^x,.) = 0.047, and stratum
proportions differed by a factor of 10. Optimal allocation split the sample 70/30, compared to
proportional allocation of 50/50. Under this scenario, optimal allocation did give slightly more
precise (smaller confidence interval) estimates of the composite population proportion below
xc, but at the price of a larger confidence interval over most of the range (Fig. 5-3a).
5-9
-------
1.00 -i
0.20 -
0.00
Population 1
Population 2
4 ' 5 ' 4
BenUite Index (x)
Figure 5-2. Hypothetical distribution functions used in
analysis of stratum misclassification effects
When a 20% misclassification rate was assumed for both populations, however, proportional
allocation became the superior approach (Fig. 5-3b). "Optimal" allocation was no longer
beneficial, even at the critical value, and was substantially worse than proportional allocation
over much of the range of the indicator. This degree of misclassification is to be expected in
environmental applications and is less than or comparable to the observed rates using salinity
or sediment type.
Resource management and environmental agencies for seven of the estuaries in the Virginian
Province are in the process of developing comprehensive management and restoration plans.
The current sampling design for estuaries is adequate for providing baseline estimates of
status against which to evaluate the effectiveness of management plans for three of these
estuaries: Chesapeake Bay, Long Island Sound, and Delaware Bay. The EMAP sampling
grid can be adapted easily to increase sample density in geographical areas of interest. For
example, during the 1990 Demonstration Project, the sampling density in the Delaware
Estuary was intensified by a factor of four from a 280 km2 grid-frame to a 70 km2 grid-frame,
allowing population estimates for Delaware Bay to be made with greater precision (see
Section 5.6). Similar intensification could be accomplished in each of the remaining systems
in the province that are developing management plans (i.e., Narragansett Bay, Buzzards Bay,
New York Harbor Complex, and
5-10
-------
(a)
0.15-1
0.10-
0.05 -
0.00
Optimal \
Proportional
4'i'i'ib
Banthfe Index
(b)
0.15 -i
0.10 -
i
•8
0.05 -
0.00
Optimal
Proportional
4'I'i'ib
Banthte Index
Figure 5-3. Comparison of optimal vs proportional allocation, (a) No misclassification, xc
4.5, 50/50 population mix. (b) 20% misclassification, xc = 4.5, 50/50 population mix.
5-11
-------
Delaware Inland Bays) to obtain baseline estimates of their status. The number of samples
required to generate system-specific status estimates for these four systems with
reasonable confidence exceeds the existing EMAP budget. EMAP is exploring opportunities
to develop partnerships with one or more of these programs to demonstrate the value of
information obtained by intensifying EMAP monitoring in individual systems.
5.3 PRECISION OF STATUS ESTIMATES
One of the goals of EMAP is to describe status with known and acceptable levels of con-
fidence. Defining "acceptable" confidence levels is an iterative process in which the need for
additional precision must be weighed against the cost of obtaining it. A first step in this
process is to identify the precision with which the extent of degraded area can be estimated
for various estuarine populations of interest (e.g., province as a whole, size classes of
estuaries, specific estuaries) with the existing sampling effort.
In this section, precision estimates are presented for each of the response and exposure
indicators used in the preliminary evaluation of environmental condition of Virginian Province
estuaries. The discussion of precision estimates is centered around two representative
indicators, the benthic index and the acute toxicity of sediments to indigenous biota. For
these indicators the precision estimates are presented as confidence intervals around the
entire cumulative distribution function for the province as a whole (Fig. 5-4); for each estuarine
class (Fig. 5-5); and for three large estuaries, Chesapeake Bay, Delaware Estuary, and Long
Island Sound (Fig. 5-6). Although precision estimates surrounding the entire CDF are
interesting, the most relevant aspect is the confidence interval at the critical levels used to
delineate degraded and nondegraded areas, which is how precision estimates for the
remaining indicators are presented (Table 5-5). Specific methods used in calculating precision
estimates were provided in Section 2.
For the benthic index, the critical value used to distinguish degraded from nondegraded areas
was 3.4; for sediment toxicity, that value was 80% of control survival. Based on the current
sampling effort, there is a 90% chance that the limits of 16% to 30% for the benthic index
shown in Fig. 5-4 include the actual value for percent of area characterized by degraded
biological resources. For sediment toxicity, the estimated value is 8%, and there is a 90%
chance that the limits of 3% to 13% include the actual value for the area of the province
having sediments that are toxic to the test organism, Ampelisca abdita.
Estimates of the percent of degraded area for the three classes of estuaries were less precise
than for the province as a whole because fewer samples were taken in each class. For the
benthic index, the confidence intervals surrounding the estimates of percent
5-12
-------
100
80
60
40
20
CD
CD
O
o3 100
Q.
80
60
40
20
2 4 68
Benthic Index
10
0 20 40 60 80 100 120
Sediment Toxicity
(percent control survival)
Figure 5-4. Confidence intervals (90%) for estimates of benthic index and sediment toxicity
over the entire Virginian Province
5-13
-------
Large Estuaries
100
80
60
40
20
100
80
60
40
20
6 8 10
0 20 40 60 80 100 120
Cti
O)
a
-------
Chesapeake Bay
100
80
60
40
20
0
100
80
60
40
20
6 8 10
0 20 40 60 80 100 120
cd
-------
13
i
to
CO
o
Q.
CO
s>
T5
CO
CD
°c3
>
18
:•§
o
•o
=»
g
CO oj
CO ^
li
||
8 §.
CD "•
£ ~
1 8.
T3 CL
m CO
li
CO CO
^"co
o •"£
O5 C_^
CO •
7R £2
IU fc^»
CD CO
"e .S?
•— S
8~
F 2
CD =;
S OT
*£= O
§ X
O 8
IO
in
CD
§
m"0 "9
o CO 3
„ j to |5
— CO
CO
jS m
o
a
CD
•K '
(Q
O. §*
i m
£
O
0
«— £• c
"tn "*••
E § ^
Ui W
O — • CO
m -"2 5
*| !"• M*
CO
m
CM
li
CO
CO
^3-
T™
+1
CO
CM
CM
CO
+1
CD
"*
CO
+1
^5
^1
f*«»
+1
CO
CM
CO.
T3
.O
CD
m
CM
+1
in
CO
+1
CO
^.
4-1
CO
CO
T—
+1
CO
CO
+1
CM
CO
+1
CM
in
+1
CO
<£
1 Sediment toxicity (<8C
survival)
CO
CO
+i
t—
in
^j.
-fl
T—
+|
O
00
in
-H
"^
CO
CM
+1
CD
"~
CD
+1
CO
CM
Is-
+1
CM
^^
Q.
Q.
Dissolved oxygen (<5
CO
+i
in
,_
*
rf
+1
0
1~
CO
T—
+1
CM
CM
+1
00
T~
CM
+1
^j.
+|
O)
Iciostridium
(> 250 cells/1 ,000 gm
8
-H
00
m
CM
*'
T~
o
1 1
r--
co
00
1—
+1
CM
in
m
+i
^-
"^
o
+l
00
CM
CD
+1
CO
CO
.co "a?
*= "ffl
•4-* >
OJ -J
=5 LU
O)
T—
+1
CO
r--
5
CO
in
+i
m
^j-
T~
+1
CM
CM
+1
CO
co
+1
m
^i"
+1
CM
T7-
Sediment Pesticides
(> ER-L values)
o
+i
o
CO
+1
00
CO
+1
CO
CM
O
CM
+1
8
CO
+1
CO
+1
CD
00
+1
h-
03
i
(D
•0 .E
7>£
o> o
O ^3
"0 T-~
"5 CD
Q. Q.
Si
+1
CO
CM
o
T"
+1
CO
h-
+1
CO
1
CO
+1
CO
CM
T—
+1
CM
CO
CO
+1
o
in
+1
5:
If
1
£
&
•c
CD
1
o
+1
o
00
+1
$
m
-fl
in
1
T—
I —
-(-(
£:
o
CO
+1
"rf*
CO
m
rf
+l
CO
f.
CD
•n
1 Water clarity (Secchi <
<1 m)
5-16
-------
degraded area in the small estuarine system and large tidal river classes exceed 50% of the
mean (Table 5-5). The EMAP Design and Statistics Team is developing appropriate statistical
methods for comparing estimates of degraded area among classes, but as a rough
approximation, non-overlapping confidence intervals are required for differences to be
significant. Using this criteria and 90% confidence intervals, the differences in the benthic
index between the large tidal river and small estuarine system classes were not significant
based on the 1990 data, even though the estimate of percent degraded area for large tidal
rivers was twice that for small estuarine systems (Table 5-5). For sediment toxicity, the
percent of area in small systems that exhibited unacceptable toxicity to test organisms could
be distinguished statistically from that in large estuaries and tidal rivers, but only because
these differences were 20-fold. Estimated CDFs for two large estuaries, the Delaware Bay
system and Chesapeake Bay system, appeared quite different. With only one year of data,
however, the confidence limits are too wide to consider the differences statistically significant
(Fig. 5-6), even though the sampling density in the large estuary and large tidal river portions
of the Delaware system was enhanced to approximately three times (four times normal in the
Bay; two times normal in the river) the number of sites that would be sampled in a typical
year.
These confidence intervals indicate a high degree of uncertainty associated with estimates
based on a single year of data; however, EMAP's goal was never to make single-year
estimates with great confidence. Instead, the intention was to derive status estimates on the
basis of four-year running averages (Stevens et al. 1991; Holland 1990). The four-year
running average increases sample size and expands the spatial scale of the systematic grid.
It also allows for averaging over interannual variability so that status estimates are not based
on a single year of potentially anomalous conditions.
Assuming that variance in 1990 was representative of that in other years, and that interannual
variability is small compared to within-year variability, estimates of status for the Virginian
Province using a full, four-year base period (i.e., a fourfold increase in sample size) will be
approximately twice as precise as those presented in Table 5-5. Based on this assumption,
the existing sampling design should be able to detect differences in the extent of degraded
area as small as twofold between major subpopulations after four years. The magnitude of
interannual variability and the specific procedure for combining four years of data into a single
estimate of status will affect precision estimates, but these factors are presently unknown or
undecided.
5.3.1 Improving the Precision of Estimates
Levels of precision varied among the three estuarine classes and among indicators (Table 5-
5). Estimates for large tidal rivers generally had wider confidence intervals than large
estuaries and small estuarine systems, particularly for the integrated indices. The higher
5-17
-------
degree of uncertainty for tidal rivers compared to other classes suggests that ways to increase
precision within the tidal river class need to be investigated. Precision in the tidal river class
could be increased simply by allocating a greater number of samples to the sampling stratum;
however, unless the total amount of sampling effort for the program is increased, such
reallocation would require reduced precision in another class.
Another means for increasing precision is to reconfigure the allocation scheme so that the
areas of river segments are more equal in size. The size of the confidence limits produced by
the Yates-Grundy estimator of variance is strongly influenced by the inclusion probabilities for
tidal river segments and the indicator response values. Large differences in segment areas
result in large differences in inclusion probabilities and, generally, in a large estimate of
variance, regardless of the variability in indicator values. Under the present configuration, the
tidal rivers are partitioned into 25 km segments. Defining segments of equal length facilitates
making status estimates on the basis of river length; however, the primary basis for making
status estimates is area because area represents the only practical way to integrate findings
across strata of different sizes. At present, there is more than an order of magnitude
difference in area among the river segments within the tidal river class. This results, in part,
from inherent size differences among the river systems, but mainly from the disparity in width
between the upper and lower portions of these tidal rivers.
Reconfiguring the allocation scheme within the tidal river class so that all inclusion proba-
bilities (i.e., areas) are of similar magnitude would result in higher precision for status
estimates. There are several alternatives for accomplishing this within the context of the
existing design. One is to divide the river into segments of equal area, rather than equal
length, and then sample each segment in each year using the existing procedure. Another
alternative is to intensify the systematic grid used for large estuaries until a sufficient sample
size is achieved in tidal rivers. These possibilities and other alternative allocation strategies
for tidal rivers currently are being evaluated by the EMAP Estuaries Resource Group and the
EMAP Statistics and Design Team.
5.3.2 Using Replicates to Improve Precision
In small estuarine systems, the precision of the estimated CDFs is a function of the number of
systems sampled and the number of replicate samples obtained in each system. The
precision of the estimates can be increased by allocating sampling effort optimally between
systems and replicates. For example, if the largest source of variation is between systems,
effort is most appropriately targeted at sampling more systems to increase precision.
Conversely, if the largest source of variation occurs within systems, effort should be targeted
at obtaining replicate samples within systems. An analysis was conducted to assess
potential improvements in precision from reallocating sampling effort for estimating the
percentage of small systems that have a benthic index value below the critical threshold (i.e.,
5-18
-------
less than 3.4). In particular, the analysis addressed whether the precision associated with the
benthic index response could be increased more effectively by increasing the number of
sampled systems, or by increasing the number of replicates within systems.
During the Demonstration Project, benthic samples were collected from only a single location
in 26 small estuarine systems, and at least two randomly selected locations were sampled in
the remaining six small systems (Back River, Mystic River, Mullica River, Mattaponi River,
Elizabeth River, and Indian River Bay). An estimate of the variation in the value of the benthic
index between small systems was obtained from the data for all 32 sampled systems; an
estimate of the average within-system variation was obtained from the six systems with
replicate samples.
The approximate mean squared error (assuming that the within system variation in the benthic
index response is constant across all systems) of the CDF estimate (Cochran 1977; eq.
11.27) was calculated as:
MSE(m,ri) =
NS*
nA2
rn
where
n
N
A =
Yt =
y =
A =
number of small systems sampled
number of small systems in the Virginia Province (137)
n/N
12 /TT ~I\
EACH-/)
-t]
AM
= between system variation
area of system /'
probability of the benthic index being less than 3.4 in system /
N _
EAy/
^i
A
N
E A,= total area of small systems in the Virginia Province (4875.2 km2)
within system variation
5-19
-------
m = number of replicate samples obtained in each system
Unbiased estimates of the between system variation (3,2) and the average within-system
variation (£,2) were obtained from the 1990 small system data. Estimates of the approximate
mean squared error for various values of the number of systems sampled (n) and the number
of replicates per system (m) were calculated by inserting these values and the variance
estimates into the above formula. Assuming a baseline sampling strategy of 32 systems and
one sample per system, the relative efficiency of adding systems and/or replicates was
calculated as:
This analysis indicated that the precision of the estimates was affected more by the number of
systems sampled than by the number of replicates per system. If replicates were added at
randomly selected locations in each of the 32 small systems sampled in 1990 (i.e., 64 total
samples), the confidence intervals would shrink by only 14%. In comparison, if the 32
additional samples were allocated to different small systems (i.e., 64 systems sampled without
replication), the confidence intervals would be reduced by 43%. Only eight samples in new
systems would be required to achieve the 14% improvement in precision obtained by adding
replicates in all 32 small systems. Sampling additional systems provides a greater increase in
precision than obtaining additional replicates within systems because the variation across
systems was much greater than the variation within systems. This finding is consistent with a
comparison between the benthic indices at index and random sites within systems in which no
significant differences were detected (see Section 5.5); however, the difference in benthic
Indices among small systems was large.
Although replicate samples do not appear to be the most efficient allocation of effort,
discontinuing them at this time may not be warranted. Replicates provide estimates of within-
system variation that are necessary for estimating the mean squared error of CDFs. Although
these could be estimated from data collected during the 1990 Demonstration Project, replicate
samples were obtained at only 6 of the 32 sampled stations, and only in a single year. The
representativeness of conclusions drawn from these data needs to be evaluated after
gathering the same type of information from additional systems and in additional years;
moreover, these data need to be gathered in other provinces because the patterns observed
in small estuarine systems of the mid-Atlantic may not be repeated in other areas of the
country. Finally, randomly-located replicate samples were collected only in small systems
during the 1990 Demonstration Project, further limiting the ability to generalize conclusions
based on these data. Additional analyses are needed to determine the validity of these
conclusions for large tidal rivers and large estuaries. As discussed above, replicate samples
5-20
-------
within grid cells of the large estuaries collected in 1991 can be used to conduct such
analyses.
5.3.3 Improving Estimation of Precision
EMAP-E and the EMAP Statistics and Design Team are developing data analysis methods
and evaluating alternative sampling designs to improve the program's ability to characterize
the precision of estimates. Three issues identified during the 1990 Demonstration Project are
being addressed. The first is that current estimates of variance assume a constant sample
size and do not incorporate the variance due to random sample size. Using the randomized
grid to select sampling sites in the large estuaries, and possibly tidal rivers in the future,
results in random sample sizes; therefore, assuming constant sample sizes may result in
underestimating variance. This shortcoming is being addressed by developing variance
estimators that account for random sample sizes. Systematic spatial separation of samples in
large estuaries is the second important issue in estimating variance. The separation is
expected to reduce variance because only small-scale spatial variability will be included in the
variance of estimates. The current method of variance estimation, however, represents large-
scale spatial variability (i.e., differences between sites that are no closer than the spacing
between adjacent points on the systematic grid). If spatial autocorrelation is substantial, the
small-scale variability will be lower than the large-scale variability. Accordingly, the current
method of variance estimation will produce overestimates of variance. The third issue
concerns the presence of joint inclusion probabilities that are zero and the use of
approximations for the joint inclusion probabilities in the Horvitz-Thompson estimator for
variance. The Horvitz-Thompson variance estimator is unbiased if all joint inclusion
probabilities are non-zero, and the joint inclusion probabilities for the pairs of stations actually
sampled are known. The consequences of not satisfying these assumptions are not known.
To address these issues, two changes were made in the sampling conducted in the Virginian
Province in 1991. A study that sampled along transects between sampling sites in the
systematic grid was conducted to provide information on the magnitude of small scale
sampling variability. These data will be used to evaluate the reduction in variance gained by
spatial separation of samples. In addition, large estuarine systems in the Louisianian Province
were sampled by randomizing the position of the overlaid grid, and then randomly selecting a
sample point within the 280 km2 area around each sampling site. Sampling in this manner will
result in unbiased variance estimation (Robson et al. 1991). The two alternatives being tested
in the Virginian and Louisianian provinces will provide data for determining which is most
robust for long-term use in EMAP.
5-21
-------
5.4 EVALUATION OF POWER FOR TREND DETECTION
Two classes of sampling design have been suggested to be most appropriate for long-term
ecological monitoring: rotating panel designs (Duncan and Kalton 1987) and serially-
alternating designs (Overton et al. 1990), like the one used in EMAP-E. The rotating panel
design prescribes that a set of sites will be visited for several consecutive years, and then
eliminated from future consideration. As a set is eliminated from the monitoring program, it is
replaced with a new set of equal size, which then remains in the survey for a number of years.
In a rotating panel design the total sample is divided evenly into a number of sets equal to the
number of years a set remains in the sample. For example, in a four-year rotation, one-fourth
of the sample is replaced every year. The serially-alternating design also splits the total
sample into several equal-sized sets, but only one set is visited each year. A set is not
revisited until all other sets have been visited, and serial revisitation is continued indefinitely.
The basic serially-alternating design does not prescribe any replacement or annual revisits.
Both designs can be augmented by adding a set of sites that are visited annually for the
duration of the monitoring program.
The serially-alternating design was selected for use in EMAP-E based on a general linear
model that compared the relative efficiency of these two designs (Urquhart et al. 1991). The
linear model allowed Urquhart et al. to consider the estimation of both status and trends under
various levels of population variation, measurement error, and interannual variation, and to
incorporate some correlation between years and between sites measured at different times.
The serially-alternating design was almost always more efficient than the rotating panel and
was never less than 99% as efficient, despite the wide range of possibilities investigated.
Urquhart et al. (1991) concluded that the augmented serially-alternating design offers a
substantial advantage in ability to make estimates for subpopulations because more sites are
visited sooner than with the rotating panel design.
Their exploration of the two designs also included an investigation of the optimal fraction of
resources devoted to augmentation (annual sampling). They found that the fraction of
resources devoted to annual sampling in the augmented serially-alternating design depended
primarily on the two correlations and the number of years of sampling. The dependence on
the variance components was substantially less. The optimal fraction increased as the site
and year correlations increased and declined substantially after the first cycle of visits.
The correlation of sites over years models the identity of sites over time. If the characteristics
of a site change very slowly, then the site correlation would be close to 1. If, however, the
sites exhibit little identity over time, the site correlation would be closer to 0. One might
expect terrestrial sites to have site correlations near 1 and estuarine sites to have somewhat
lower correlations. The correlation between years reflects persistence of an effect and
persistence of the source of the effect. For example, a drought might persist for several
years, but once normal rainfall is reestablished, the effect of the drought may disappear
5-22
-------
quickly. Table 5-6 gives the optimal fraction after four and eight years of sampling for various
values of annual and site correlation for a four-year, augmented, serially-alternating design.
The optimal fractions are never more than 25% and are that large only for correlations that are
probably unrealistically high for EMAP-E. Realistic considerations suggest that 10% to 15% of
the total sample size allocated to annual sampling should be nearly optimal. Based on these
analyses, EMAP-E initially will revisit 10% of its sites annually and will re-evaluate this
proportion after several years of data have been collected from which to assess the site
correlation.
Table 5-6. Optimal percent of sample size devoted to annual sampling (adapted from
Urquhartetal. 1991)
Site
Correlation
0.50
0.75
0.90
0.95
1.00
After 4 Years
After 8 Years
Correlation Between Year Effects
0.00
0
0
8
13
13
0.05
0
0
10
13
13
0.10
0
0
10
15
13
0.25
0
2
15
17
15
0.50
0
10
23
25
23
0.00
0
0
0
2
2
0.05
0
0
0
2
2
0.10
0
0
0
2
2
0.25
0
0
0
2
4
0.50
0
0
2
4
6
EMAP's goal is to detect changes of 2% per year in the amount of degraded area over a ten
year period, using selected response and exposure indicators. A model was developed, and
a power analysis was performed to ascertain the effectiveness of the present program for
accomplishing this goal and to provide a tool for determining optimal allocation of sampling
effort among resource classes. The model-was based upon the assumption of a linear trend
in the proportion of degraded area over three EMAP sampling cycles and was applied to the
benthic index data obtained during the 1990 Demonstration Project. The benthic index was
selected for the first application of the model because of the importance the program places
on assessment of environmental condition based upon biological response indicators.
5.4.1 Formulation of the Model
The model used to test for trend detection was a two-factor analysis of variance (ANOVA)
model with an additional term for time trend. The two factors, sites and years, involve random
5-23
-------
effects, since the specific sites selected can be viewed as a sample from the population of all
possible sites, and the years observed can be viewed the same way. Interaction effects
between sites and years also were included in the model. Site effects, year effects, and site-
by-year interaction effects were assumed to be three independent random vectors, each
multivariate-normally distributed with expectation zero and a suitable variance-covariance
matrix.
The model was formulated by letting Yg denote the observed value of the response or
exposure indicator at site / in year j. The basic idea underlying the variance-covariance
structure is that the degree of correlation between a pair of observations, Yg and Y,,]h
increases as the geographic distance between sites / and /', d(!,lf), decreases and as the
time difference between years / and /, \]-]'\, decreases.
In addition to the site and year terms, the model included a vector of time trend terms for the
years observed. These terms are not linear themselves, but were constrained to correspond
to a linear trend across years in the proportion of area that is classified as degraded. This is
because the main concern is not Yf, but the binary indicator variable, Xf, each value of which
tells whether or not the corresponding Y9 is degraded.
Temporal autocorrelation in both year and site effects was incorporated into the model,
following and extending the approach of Urquhart et al. (1991). Additionally, because of the
extensive character of estuarine resources, spatial autocorrelation was incorporated as a
power function of distance between sample sites. The model for the response indicator data
Yf (e.g., benthic index) is given by the equation:
where
e,=
(8T), =
for site M,...,/?, year/=1 ,...,
observed value of the indicator at site / in year J
population mean (over all sites and years)
random effect of site /
random effect of year J
random interaction effect of site / and year J
time trend term for year j
5-24
-------
e» = random error term for site / in year J
The covariance structure for the random effects assumed in this model was:
fi - /lfV7V(0,Se) where Se(n x fl)=[p?*^o5l
i - MVW(0,ST) where sT(f x 0=feJ"V]
fi,£, and fl i are uncorrelated with each other
The parameters of the model are: \i, & ojj, are defined to correspond to a linear trend in the proportion of degraded area.
The random effects 8; (/=1,...,n), Ty (/=1,...,J), and (e-c), are random variables.
For the purpose of power calculations, multivariate normality, hence statistical independence
among the three vectors of random effects, was assumed. Note that the model of Urquhart et
al. (1991) can be obtained from the above model as a special case by dropping fi and £, and
assuming no spatial covariance.
5.4.2 The Trend Detection Procedure
Investigating the question of whether a time trend is present was begun by estimating the
parameters of the model from the data by maximum likelihood, then testing for linear trends
using hypotheses H0: p=0 vs H,: p#0. Supplementing this analysis informally by plotting Y.{
against year J (for /=1,...,f) can be informative. The rest of this subsection provides a more
detailed discussion of the procedure for testing for trends.
The problem was transformed into degraded/nondegraded terms because the quantity of
primary interest was not Yt, but the binary indicator variable Xg corresponding to Yy, which is
defined by
X9 = Indicator {V^ <; y0} * Indicator {V, is degraded} for all ij
0 otherwise
For example, for the benthic index (Bl) with y0 = 3.4,
if fl/,^3.4,
x -
X~
0 otherwise
5-25
-------
The proportion of area in year J that is degraded is py = E^X); furthermore, by definition,
Pj a F/t/Ofc) m Prl Yt randomlychosen at time J satisfies: Yg * y0]
In this analysis, the issue of deconvolution was ignored, thus retaining extraneous variation
(e.g., measurement error) in the distributions.
The values of Xf were used to estimate the cumulative distribution function of the benthic
index for the province at time J , Fy^ , via
P/"£»>*ffory=W
where w, is the known sampling weight (see Sec. 2.4.2) given by
Wf H fly/2} j/
The trend in the true population value py over time, as the year y varies from 1 to t, was
modeled by assuming that it is a linear function of y:
/^=v+T}ry fory=l,...f
The test procedure was to fit the generalized least squares (GLS) regression model
ty=v+yJ+Kj fory=i t
to the data, where the departure from linearity in the trend is modeled as
&~Afl/W(0,V) withV(fxf) = i
and then to test H0: y =0 vs H,:
For purposes of power calculation, the form of the cumulative distribution function F.was
assumed to be normal for each value of y=l t:
jVo-(n+PO!))l
=
-------
so that
(n+p(j!)) =y0 - o c (critical value), then reject «„.
The calculations needed for this test include the following:
E(X,) =Pr[ Yjzy0\, which comes right from the cumulative distribution function
and
(i)
(ii)
(iii) These expectations give Vai(X), Cov(XtlX9), et. These, in turn, give the variance-
covariance matrix of the values of
(iv) Solve for $, v, o|, of,
(v) The test is then: if |(ilr-0)/6t | >c (critical value), then reject
The power of the test is given approximately by:
| < 4=1 -Pr{-c
=1 -
4> (-c-
5-27
-------
where 6+ is taken from the variance-covariance matrix of the parameter estimates P and
and X is the design matrix for a test for difference between the mean value of P from the first
four years of study and the last four years of study (assuming a four year sampling cycle).
5.4.3 Component Parameter Estimates
Variance and covariance components (i.e., oe, OT, oflT, oe, Pe, pt, Pee, pxe ) of the model
were estimated using an ANOVA model with random effects applied to two data sets The
first data set was that collected during the Demonstration Project, supplemented by data from
10 sites that were revisited in 1991 to estimate interannual variability. The second data set
was that of Dauer et al. (1989), which contains five years (1985-1989) of benthic monitoring
data collected from the Virginia portion of the Chesapeake Bay. The EMAP-E samples from
1990 provided estimates of the spatial variance component (Table 5-7) based on the multiple
stations that were sampled within each resource class (large estuary, large tidal river, and
small estuarine systems). The stations that were revisited in 1991 provided the basis for
estimates of the temporal variance component (interannual variability). Because estimates of
interannual variability based on only two years of data are likely to be imprecise, a second set
of interannual variability estimates was computed for comparison using the data from Dauer et
al. (1989). Neither of these data sets provided sufficient information to partition interaction
variance from residual variance; therefore, these two variance components were treated
together as residual variance in the power analyses.
For large systems, the temporal variance component in the EMAP-E data, which was based
on five stations, was not significantly different from zero; however, a significant temporal
variance component was found in the Dauer et al. data (Table 5-7), which is also based on
five stations. Because of the relatively small number of years contained in these data sets,
neither estimate is likely to be very reliable. Unusually high temporal variability was expected
in the Dauer et al. data due ,to the location of the stations. For these reasons, both the
estimate based on the Dauer et al. data and the zero estimate from the EMAP-E data were
used in the power analysis; using variability estimates from both data sets allowed
examination of the upper and lower extremes of interannual variability. The EMAP-E data
were used to estimate the residual variance component for large systems (Table 5-7) because
this component represents variability specific to the program's sampling protocols.
5-28
-------
Table 5-7. Estimates of variance components used in the power analysis based on
EMAP data. Value in parentheses represents temporal variance estimate
from the Dauer et al. (1989) Chesapeake Bay data set.
Variance
Component
Spatial
Temporal
Residual
System Class
Large Estuaries
2.694
0(1.869)
1 .491
Large Tidal Rivers
3.564
0
2.602
Small Estuarine
Systems
2.850
0
1.228
Only two stations in tidal rivers were sampled in both 1990 and 1991, and these stations were
in two different rivers; therefore, the ANOVA could not be applied to this data set, and the
residual variance component for tidal rivers was estimated from the Dauer et al. data set of
seven stations in two rivers. Analysis of the Dauer et al. data set produced a non-significant
temporal variance component; therefore, a variance estimate of zero (Table 5-7) was used in
the power analysis.
For small estuarine systems, the temporal variance component in the EMAP-E data, which
was based on three stations, was not significantly different from zero. The temporal variance
component in the Dauer et al. data for four stations within the small estuarine system class
also was not significantly different from zero. Based on these two results, the temporal
variance component was set to zero for power analyses for small estuarine systems. As for
large systems, the residual variance component was estimated from the EMAP-E data (Table
5-7). •
Analyses of the EMAP-E and Dauer et al. data revealed no evidence of spatial
autocorrelation. Visual inspection of plots of estimated covariances between responses
obtained at fixed distances produced no discernable pattern with respect to distance.
Temporal and interaction autocorrelations could not be estimated uniquely from the available
data. Accordingly, all autocorrelation values were set to zero in the power analyses.
Predicted values of the mean benthic index value (\i + p(/)) in any year, J, (e.g., 1990 to
2000) under the hypothesis of a 2% change in area per year were based on estimates of the
proportion of degraded area in 1990. An estimate of the proportion of degraded area in each
of the three resource classes in 1990 was computed as described in Section 2.4.2. Each of
5-29
-------
these estimates then was used to predict the mean benthic index value in year j as a
function of the variance components:
where Zj is defined such that $(Z) = Pf
P0 = estimated proportion of degraded area in 1990
Pj = P0 + 0.02/
and
" " "
-------
Table 5-8. Power for trend detection in the benthic index over three EMAP sampling
cycles. Values in parentheses represent power based on temporal variance
estimates from the Virginia Chesapeake Bay Monitoring Program (Dauer et
al. 1989).
Province
Large Estuaries Class
Large Tidal Rivers Class
Small Estuarine System Class
Trend (% Per Year)
1%
.95 (.22)
.83 (.17)
.41
.60
2%
.99 (.29)
.99 (.32)
.88
.98
3%
.99 (.73)
.99 (.52)
.99
.99
-1%
.99 (.36)
.95 (.25)
.44
.68
Neither the EMAP nor the Dauer et al. data sets are ideal for developing provincewide
estimates of interannual variability. The EMAP data set has the appropriate spatial resolution
but is based on only two years of data. The Dauer et al. data set has better temporal
resolution but is limited spatially and may not represent the whole province. The area
sampled in the Virginia program is characterized by unusually large interannual variation in
dissolved oxygen conditions, which would exaggerate interannual variability in the quality of
the benthic assemblage. The interannual variability estimated using the EMAP data set is
undoubtedly an underestimate, and the Virginia value is probably an overestimate. The true
number probably falls between the two.
The model development work conducted to date indicates that the present design and
allocation of effort could be sufficient to detect trends at the desired level under a select set of
model assumptions and parameter estimates; however, additional effort is needed to
document that these assumptions are valid. Four activities are planned to improve upon the
existing analysis. First, EMAP will continue to gather additional years of data to improve the
parameter estimates for the trend detection model. This will be done in both the Virginian and
Louisianian provinces. Parameter estimates will be supplemented by additional data sets from
other areas of the country, but obtaining such data sets is difficult because so few long-term
data sets exist, most do not include all the parameters (e.g. biomass) that are necessary to
calculate the benthic index, and some use methods that are significantly different from those
used by EMAP. Second, the power analysis will be extended to other indicators, such as
sediment chemistry and dissolved oxygen, for which there are several available data sets that
would be appropriate for developing estimates of interannual variability. Third, efforts will be
made to improve the sensitivity of the trend detection model. The present model was based
on a single type of test, a test for differences. Other tests, such as one based on a
5-31
-------
regression model, might prove to be more powerful. Fourth, the model will be applied to other
types of trends. The present test examines linear trends in percent of degraded area; future
model runs might examine power for trend detection in parameters such as average condition
and will consider the effect of nonlinear trends.
5.5 INDEX SITES
During the 1990 Demonstration Project, nonrandom index sites were sampled in large tidal
river segments and in small estuarine systems. Index sites are the locations most likely to be
exposed to pollution insults (e.g., low dissolved oxygen stress or the effects of contaminated
sediments), if such conditions occur. Index sites for the 1990 Demonstration Project were
located in depositional, muddy environments, where fine-grained sediments accumulate
(Holland 1990). Whereas randomly sampled base sites are designed to provide unbiased
estimates of the areal extent of pollution effects (i.e., to provide data for estimating the extent
of degraded condition), index sites are designed to determine the number of river segments or
small systems that have degraded condition in habitats that are particularly vulnerable to
pollution, without having to conduct intensive surveys. Together, information from base and
index sites can be used to distinguish between degradation occurring in a small number of
systems but a large proportion of the area of those systems from degradation manifested in a
large number of systems but a small proportion of each.
One of the goals of the 1990 Demonstration Project was to assess the value of sampling at
index sites. For this assessment, benthic index values were compared between index and
random sites. No statistical difference (paired t-test, p < 0.05) was found in either the large
tidal rivers or the small estuarine systems (Fig. 5-7). The similarity of the benthic index
between index and random sites indicates that there is no value added by sampling at these
sites. This could be because
• degradation in small estuarine systems and large tidal rivers is widespread,
affecting most of the small systems or tidal river segments; or
• depositional, muddy environments are the appropriate location to use as index
sites, but identify depositional areas were not identified accurately in the 1990
sampling program (i.e., we sampled in the wrong places).
An evaluation of the sediment data from index and random sites indicated that 50% of the
random sites for both small estuarine systems and large tidal rivers had finer-grained sedi-
ments than their corresponding index sites, suggesting that the criteria used in 1990 to select
index sites did not identify depositional, muddy environments accurately. With the r
5-32
-------
Small Estuarine Systems
100
90
80
70
| 60
2
™ 50
§
| 40
30
20
10
0
4567
Benthic Index
10
100
90
80
70
I 60
| 50
I 40
30
20
10
0
Large Tidal Rivers
456
Benthic Index
8
9
10
Figure 5-7. Comparisons of CDFs for the benthic index between index (solid line) and random
sites (broken lines) in small estuarine systems and large tidal rivers
5-33
-------
available data, therefore, we cannot determine if any value would be added by sampling in
"true" depositional habitats. As a result, EMAP-E used available sediment maps and the
scientific knowledge of local experts to define index sites in small estuarine systems for the
1991 sampling program (index sites were discontinued in large tidal rivers in 1991). The
physical characteristics of sediments, sediment contaminant concentrations, sediment toxicity
to indigenous biota, and the benthic index will be measured at these sites. If an evaluation of
the 1991 data continues to reveal difficulty in identifying index sites that are at least as
degraded as random sites, sampling at index sites will be discontinued in 1992.
5.6 DETERMINATION OF THE APPROPRIATE SPATIAL SAMPLING SCALE
The EMAP-E sampling design in the large estuary and large tidal river classes contains both
systematic and random elements. One of the questions investigated in the 1990 Demon-
stration Project was whether the spatial scale (i.e., grid dimensions) of the systematic element
was appropriate to the spatial patterns of the resources being sampled. To address this
question, supplemental samples were taken in the Delaware Estuary to augment the spatial
scale in that system, and the response at the higher sampling density was compared to that
obtained by the base monitoring effort. The Delaware Estuary was selected for this study
because it is composed primarily of the large tidal river and large estuary classes (ie., it
contains proportionately little area in its tidal tributaries). In the Delaware Bay, the supplemen-
tal samples represented a fourfold increase in spatial coverage, intensifying the scale from
grid cells of 280 km2 to cells of 70 km2. In the Delaware River, the increase was twofold, and
the supplemental sites were placed half way between the base sites. The basic design
provided 10 samples in the Delaware Estuary. Spatial supplements resulted in 24 additional
sites.
Status estimates for four parameters (depth, substrate type, sediment toxicity, and the benthic
index) calculated from the base samples were compared to those obtained when the supple-
mental samples were added. For all of these parameters, the CDFs produced by the base
stations alone were similar to those produced when the supplemental stations were included
(Fig. 5-8). The similarity of the response was greatest for the benthic index, for which the
percent of area classified as degraded (i.e., an index value less than 3.4) differed by less than
10% between the two data sets. The similarity for the depth CDF was also striking, although
the CDF did differ at the upper end of the range because two supplemental sites occurred at
depths several meters greater than any of the base sites. For sediment toxicity, there was no
difference in the range of values observed, and the difference at the threshold value used in
the preliminary assessment (80% mortality) was small. For substrate type, differences in the
CDFs between the two data sets were most pronounced (71% vs. 92%) at the boundary used
for identifying mud sediment (less than 20% silt/clay content). Both data sets identified the
system to contain a very high percentage of mud, and the difference between the two curves
was well within the confidence limits of the estimates. Together, these analyses indicate that
5-34
-------
-------
the spatial scale of sampling in estuarine systems such as the Delaware is appropriate to the
pattern of the resources being described and that, if among year variability is not extensive,
the patterns observed after one year of sampling are the same as can be expected after the
four-year assessment cycle is completed.
5-36
-------
Section 6
Preliminary Estimate of Ecological
Status of Virginian Province Estuaries
Introduction
The EMAP Demonstration Project in the estuaries of the
Virginian Province was conducted during the summer of 1990
(June through September). A probability-based sampling
design was used so that estuarine resources and character-
istics were sampled in proportion to their areal distribution
(Overton et al. 1990; Stevens et al. 1991). This sampling
design makes it possible to estimate, with known confidence,
the proportion or amount of area having defined environmental
characteristics.
Five hundred sampling visits were completed at 217 sites in
estuaries between Cape Cod and the mouth of the
Chesapeake Bay (Fig. 6-1). A series of indicators that are
representative of the overall health of estuarine resources was
measured at each site. These indicators were designed to
address three major attributes of concern to estuarine scien-
tists, environmental managers, and the public: 1) biotic
integrity, or the existence of healthy, diverse, and sustainable
biological communities; 2) pollutant exposure, or the condition
of the physico-chemical environment in which biota live; and
3) societal values, or indicators related to public use of
estuarine resources. The specific methods used to measure
these indicators and the calculations used to produce
preliminary estimates of condition based on these
measurements are presented in Section 3.
This section presents a preliminary evaluation of the condition
of estuaries in the Virginian Province based upon the data col-
lected during the 1990 Demonstration Project. Status
statements of the ecological condition of the estuaries in the
Virginian Province that will meet all program requirements are
intended to be based on four consecutive years of monitoring
information. However, preliminary status estimates can be
calculated based on a single year of monitoring information.
6-1
-------
SAMPLING TEAM 1
ANNAPOUS, MD
SAMPLING TEAM 2
SAMPLING TEAM 3
Figure 6-1. Location of the EWIAP 1990 Demonstration Project sample sites in the estuaries of the
Virginian Province.
6-2
-------
This single year estimate is representative of only that
particular year and will have more uncertainty associated with
it than the 4-year composite estimate. The 1990 ecological
status estimate represents a first attempt at presenting
information from a rich and unique data set to a varied
audience. The EMAP audience includes specialists with
intimate knowledge of estuaries and the specific resources of
the Virginian Province, and others with limited knowledge of
estuarine ecology and the environmental perturbations
affecting estuarine resources. Development of the approach
used in this evaluation was aided by completing an example
assessment report (Frithsen et al. 1991) and by a series of
workshops with potential users of EMAP data. This report
continues the process of identifying the questions of greatest
interest to users of EMAP data and of experimenting with
meaningful ways to present data and information.
Preliminary evaluations of
biotic integrity are based
on two indicators: assem-
blages of bottom dwelling
animals and the health of
individual fish.
Biotic Integrity
The condition of biological
resources in the Virginian
Province was evaluated
using two indicators: one
that measured the condi-
tion of bottom dwelling
(benthic) animal assem-
blages, and one that mea-
sured the health of fish.
The benthic indicator,
which is discussed first,
uses measures of species
composition, abundance,
and biomass to evaluate
the condition of benthic
assemblages. The fish
indicator is based upon
measures of visible patho-
logical abnormalities (e.g.,
lesions and tumors) and
reflects the response of
Benthic
- A.>6tttKfc index based tipon several
structufai and funclbiuat properties of
benthic assemblages was used as part
of Ihe preliminary evateatfoa. This
index represents a- first attempt at
'redttaing & comptex' set of maasar&-
ments to a simple, teterprelable value.
It is consistent with the EPA directive
to integrate biological criteria into-
assessments of ecoiogicat condition.
The index was developed by using
rffSorfminartf analysis to identify a. com.'
binatian of characteristics Of benfhtc
assemblages that distinguishes reliably
between regional reference $ite$
dies with known pdiutioft
Th&tndex; has been parttalty validated,
but several attditldnal years of data will
be required for complete validation;
therefore, assessments of cQnditim
based on the fridex should be con-
sidered preliminary. Detaff& or> the
methodology used ft> develop Ihe
benthic fedex and the validation steps
eompTeted to date are given in Seetfen
4,
6-3
-------
The estuarine area in the
Virginian Province having
benthic resources with
poor community structure
is estimated to be between
16-30%.
individual fish to pollutants and contaminants. Both of these
indicators provide integrated measures of environmental
condition in estuaries.
Benthic assemblages were used as an indicator because pre-
vious studies suggested that they are sensitive to pollutant
exposure (Pearson and Rosenberg 1978; Boesch and Rosen-
berg 1981). They also integrate responses to exposure over
relatively long periods of time (months to years). One reason
for their sensitivity to pollutant exposure is that benthic organ-
isms live in the sediment, a medium that accumulates environ-
mental contaminants overtime (Nixon et al. 1986; Schubel
and Carter 1984). Their relative immobility also restricts ben-
thic organisms from avoiding pollutant exposure and environ-
mental disturbances.
Preliminary estimates based on the 1990 Demonstration
Project indicate that 23% (± 7%) of the estuarine area in the
Virginian Province had benthic resources characterized by low
species richness, low abundance of selected indicator spe-
cies, and low mean weight _
for selected indicator spe-
cies. This benthic com-
munity type is referred to
herein as "degraded";
however, "degraded" does
not imply anthropogenic
causes of the observed
condition. The "degraded"
ecological condition may
be due to natural environ-
mental conditions. Of the approximately 23,574 km2 (9,102
mi2) of estuaries in the Virginian Province, about 5,422 km2
were characterized by benthic communities with low species
numbers and low abundance of indicator species. The total
degraded area was about one and a half times the size of
Rhode Island.
Although EMAP's primary objective is to describe status and
trends at the province level, estimates can also be generated
for subpopulations. The EMAP sampling design defined three
classes of estuarine resources according to surface area and
For the benlftte index, "degraded"
-------
Large tidal rivers had a
greater estimated propor-
tion of "degraded" benthic
resources than other types
of estuaries; however, the
absolute area of "degrada-
tion" was greatest in large
estuaries.
shape: large estuaries, large tidal rivers, and small estuarine
systems. These classes were defined because estuaries of
different sizes may respond differently to anthropogenic
impacts. Large estuaries like Chesapeake Bay, Delaware
Estuary, and Long Island Sound typically have large, complex
watersheds and are affected by multiple environmental
stresses and contaminant sources. Large tidal rivers, such as
the Delaware and Potomac rivers, also drain complex
watersheds, but their geometry (length-to-width ratio)
enhances interactions with bordering terrestrial systems,
including urbanized areas. In addition, flushing rates in large
tidal rivers are typically faster than in large estuaries. Small
estuarine systems such as New Bedford Harbor and Barnegat
Bay typically have small watersheds and less diverse contam-
inant sources than large estuaries.
The prevalence of "degraded" benthic resources was dissimi-
lar among the three classes of estuaries sampled during the
1990 Demonstration Project. Proportionately, large tidal rivers
were the most "degraded"; 46% (+ 32%) of the area of large
tidal rivers in the Virginian Province was estimated to have
"degraded" benthic resources (Fig. 6-2). The proportion of
Degraded Benthic Assemblages
50
1 40
O
c
f30
2 20
§
i 10
0-
46
20
*•
23
*mamum
,_
„-„
Urge Large Tidal Small
Estuaries Rivers Systems
Figure 6-2. Estimated percent of estuarine area in the three
classes of estuarine resources having degraded benthic com-
munities (see Table 5-5 for levels of uncertainty)
6-5
-------
area within large estuaries containing degraded benthic re-
sources was less than half that of tidal rivers; however, owing
to the larger size of this class, the actual area of degradation
was greatest in large estuaries. In large estuaries, the total
area with degraded benthic resources was about 3,365 km ,
compared to only 839 km2 in large tidal rivers and 1,119 km
jn small estuarine systems.
Less than 0.1% percent of
the fish examined that are
commercially or recrea-
tionally harvested had
visible pathological disor-
ders.
The second indicator used to assess the biological integrity of
estuaries was the occurrence of visible pathological problems,
such as tumors and lesions, in fish. Although several factors
may contribute to the occurrence of such disorders, they most
frequently occur in response to high contaminant
concentrations in the environment. Studies have shown that
conditions such as fin erosion, skin tumors, and ulcers are
most prevalent in polluted habitats (O'Connor et al. 1987;
Buhler and Williams 1988). Unlike benthic organisms, fish are
highly mobile and can move out of contaminated estuarine
areas; therefore, ascribing an areal extent to estimates of
pathological abnormalities is difficult. Instead, pathological
disorders are expressed in terms of prevalence within a
region, water body, or class of water bodies.
Fish examined that are
closely associated with
bottom sediment had a
greater prevalence of
pathological disorders.
Less than 1% (4 fish in 1,000) of the fish examined during the
1990 Demonstration Project had visible pathological disorders
(Fig. 6-3) establishing an estimated background level for gross
pathologies at about 0.2%-0.7%. The prevalence of
abnormalities was even lower (less than 1 in 1,000) among
commercially and recreationally harvested species. Patho-
logical disorders were more prevalent in fish that feed on or
are associated with sediments (demersal fish). The
prevalence of pathological disorders in demersal fish was 17
(± 8) fish per 1,000.
6-6
-------
Visible pathological disor-
ders in fish examined were
most prevalent in small
estuarine systems.
The prevalence of visible
pathological disorders dif-
fered substantially among
the three classes of estu-
arine resources. Thirty-six
(± 20) of every 1,000
demersal fish in small
estuarine systems had
pathological disorders; this
rate was significantly
greater than in large estu-
aries or large tidal rivers (Fig.
One of the goals of EtoAP is id fwavfck
vt states with a known m$
bted^rafrofeorifidSKcje, Con-
Intervals tor tfee description t>1
estuarine sfatuspteserttedhere appear
in Section &, Data collected during tfte
1980 Demonstration, l*rojeei are betng
'used, to avatuale the allocation atid
meet ihe preeisioR needs of the pro- .
pram, %
6-4).
OT
V)
10
| 5
Pathological Disorders in Fish
17
Bottom-Dwelling
Fish
Harvested
Fish
Figure 6-3. Prevalence of fish with pathological disorders in
the Virginian Province. Presented for all fish, bottom dwelling,
and commercially or recreationally harvested fish
Pollutant exposure was
measured with three
indicators: dissolved
oxygen, sediment
contaminants, and
sediment toxicity.
Pollutant Exposure
Although EMAP's major objective is to describe the status of
estuaries using indicators of ecological condition, envi-
ronmental managers are also interested in descriptions of the
extent and magnitude of pollutant exposure. Measures of pol-
lutant exposure historically have been the mainstay of
6-7
-------
Bottom dissolved oxygen
concentration below
5 ppm, the water quality
standard for many states
in the province, was
estimated to occur in 14-
28% of the province.
Two to sixteen percent of
the estuarine area in the
province was estimated to
have oxygen
concentrations below
2 ppm, which is con-
sidered extremely stressful
to biota.
Pathological Disorders in Demersal Fish
-a 50
| 40
§.30
.w
il 20
j§ 10
36
9
Large
Estuaries
Large Tidal
Rivers
Small
Systems
Figure 6-4. Prevalence of pathological disorders in demersal
fish in the three classes of estuarine resources.
environmental monitoring programs. Indicators of pollutant
exposure measured during the 1990 Demonstration Project
were water column dissolved oxygen concentrations, sediment
toxicity, and the concentration of contaminants in sediments.
Two typed of dissolved
stiK^nte Were made during the
Demonstration Project; point mea-
continuously-recorded
Dissolved oxygen is a fun-
damental requirement for
estuarine organisms. A
threshold concentration of
5 ppm is used by many
states to set water quality
standards. Bottom waters
in 21% (±7%) of the
estuarine area of the
Virginian Province had dis-
solved oxygen concen-
trations that failed to meet
this criterion (Fig. 6-5). A
concentration of approxi-
mately 2 ppm often is used
as a threshold for oxygen
concentrations thought to be extremely stressful to most
estuarine biota. Results from the 1990 Demonstration Project
conditions frt this report &t&
based oh the point measurements.
The continuously-' recorded data were
collected at a subset of sites to
determina whether EMAP can eost-
effectivefy gather additional information
for estimating measures that cannot be
estimated from a single, point
measurement, such as percent of time
below a criticaf ValUe-i An$ly$e$ of the
QorttmUQu$ty~r<&cjrd
-------
indicate that bottom water dissolved oxygen concentrations
below this threshold were found in 9% (± 7%) of the Virginian
Province (Fig. 6-5).
Dissolved Oxygen Concentrations
2 to 5 ppm Oxygen (12%)
2 ppm Oxygen (9%)
5 ppm Oxygen (79%)
Figure 6-5. Estimated percent of estuarine area in the Virginian
Province with bottom dissolved oxygen concentrations below 2
ppm and 5 ppm.
Previous studies have shown
dissolved oxygen
concentrations is greater in
areas where there is
density stratification of the
water column. This occurs
because stratification
reduces exchange between
bottom waters and
productive surface waters,
which are generally more
oxygen-rich due to phyto-
plankton production and
diffusion from the atmos-
phere. Results from the
1990 Demonstration
Project are consistent with
these findings. More than
half of the area having
oxygen concentrations below
that the probability of finding low
Natural vs.
Anthropogenic Degradation
Ite distinction is macfe in this report
between degraded conditions brought-
on &y anthropogenfo activities anct
those due to nattffai catisss. F0r Indi*
pafer^ Bitch ts dissotod
-------
Between three and thirteen
percent of the area in the
Virginian Province were
estimated to have sedi-
ments that were toxic to
estuarine organisms.
Two indicators of the potential effects of contaminants on
biota were measured during the Demonstration Project. The
first was a bioassay for acute toxicity in which estuarine biota
were exposed to the sediments under controlled laboratory
conditions. The second was direct measurement of contami-
nant concentrations in the sediment.
Sediment bioassays are the most direct measure of
contaminant-induced effects on biological communities
(Chapman 1988). Mortality in these laboratory exposure tests
provides evidence of toxic contamination without requiring
interpretation of how complex mixtures might interact to affect
biota. Based upon results of bioassays, 8% (± 5%) of the
Virginian Province were estimated to contain sediment that
was toxic to estuarine organisms.
Direct measurement of contaminants complements the
bioassay by identifying the compounds most likely to have
produced the toxicity. The
most important contributors
to acute toxicity were three
metals, lead, mercury, and
zinc. The concentrations
of these metals exceeded
the threshold concentration
for biological effects (ER-L)
defined by Long and
Morgan (1990) at 90% of
the sites with acute toxicity.
In contrast, only 17% of
the Demonstration Project
sites where there was mor-
tality in bioassays had
concentrations of an
organic contaminant that exceeded the ER-L value.
Ten-day acute toxicity tests wem
conducted using fine ampfupotf Amp&-
iisca dMteu' This was the most abtnv
fcenthtc species identified in the. f
ft
for prolong^ pwfotte; flietefow,
toxMty taste vvete run HA & stent
salinity of 30 ppC, to evaluate
fhjs test
water, seteetad
also te&jetf using *hs
amphipod Hyattetaazteoa, These tests
confirmed Amp&lisca^ sensitivity to
contaminants frt brackish water,. These
results are presented to Seetfon 4.
6-10
-------
Sediments in 30-48% of
the province were
estimated to have
contaminants at
concentrations that could
potentially cause sublethal
effects in biota.
Direct measurement of sediment contaminant concentrations
also provides an early warning of pollutant exposure.
Whereas sediment bioassays measure acute toxicity, direct
measurement of contaminant concentrations can be used to
evaluate the potential for sublethal biological effects.
Currently, however, contaminant concentrations that are likely
to produce sublethal effects are not well-defined. For the
Demonstration Project data, the potential for sublethal effects
from sediment contamination was evaluated by comparing
contaminant concentrations from all sites to ER-L values
(Fig. 6-6). ER-L values represent concentrations at which any
Sediment Contaminants
None (61%)
Organics (3%)
Both (9%)
Metals (27%)
Figure 6-6. Estimated percent of area in the Virginian Province
with sediment contaminant concentrations greater than Long
and Morgan (1990) ER-L values.
type of biological effect (sublethal and some lethal) was noted
in at least 10% of contaminant exposure studies. Based on
this approach, 39% (± 9%) of the Virginian Province had
concentrations of contaminants that have the potential to
cause at least sublethal effects in biota. The appearance of
no contaminants in 61% of the sediments of the Virginian
Province (Fig. 6-6) refers to the absence of measurable
quantities of the contaminants listed in Table 2-7. Other
unmeasured contaminants may exist in these areas.
Metals were the most prevalent contaminants at concen-
trations potentially leading to biological effects. Thirty-six
percent of the area contained elevated concentrations
6-11
-------
Estimated exposure to low
dissolved oxygen was
greatest in large tidal
rivers; toxic sediments
were most prevalent in
small estuarine systems.
primarily of lead, nickel,
and zinc. Organic contami-
nants at concentrations of
potential biological concern
were found in 12% (± 4%)
of the province. Of the
organic contaminants,
chlorinated pesticides were
the most prevalent; poly-
aromatic hydrocarbons
(PAH) and polychlorinated
biphenyls (PCB) were
found at biologically rele-
vant levels in less than 1 %
of the province. Estimates
of areal extent of the indi-
vidual contaminants mea-
sured in the program are
provided in Section 4.
Metats vs Organic Contaminants
The relative prevalence of cfifferent
Classes of cont aminants in 81© environ-
ment is an important ecological que£-
Son since many of these/ classes come
torn
fetevattf dbrteentrflttons than <*t«
ddntamfnanfe, tiis
should be considered
Biologically
-------
100
60
Pollutant Exposure Conditions
n Large Estuaries
• Large Tidal Rivers
HI Small Systems
Oxygen Sediment
<; 2 ppm Toxicity
Metals Organics
Figure 6-7. Pollutant exposure conditions for the three classes
of estuarine resources
with that goal. Based
upon present sample allo-
cation, EMAP will also
have a sufficient number
of samples to evaluate the
condition of three
individual estuarine
systems in the Virginian
Province - Chesapeake
Bay, Delaware Estuary,
and Long Island Sound -
after completing a four-
year sampling cycle. To
meet the needs of other
regional and state
programs, the EMAP
sampling design may be
enhanced to evaluate
estuarine resources of any
size. EMAP is currently
developing partnerships
with regional and state
EMAP Sampling Cycle
As currently envisioned* EMAP will
make assessments of environmen-
tal status using a tour-year eycfe of
monitoring data. Multiple years of
data afford a greater number of
sample points and minimize the
effects of natural InterannUaf vari-
ation due to climate and other ftl-
fluancea. The number of sample
points after four years will result fit
greater confidence In the estimates
than is possible after a single year.
This fs particularly relevant when
only a portion of the sampling
frame te used for estimation, such
as In the 'examination of Individual
estuaries. Complete assessments
will use results from the 1990
Demonstration Prefect and the
three years that follow to establish
baseline conditions for the
estuaries of the Vifgfnfan Province,
6-13
-------
resource managers to design and implement enhanced
sampling in specific estuaries.
The three systems for which estimates can be made differ
substantially in their physical characteristics. Long Island
Sound is the deepest; over 50% of its area was estimated to
be 20 m or deeper compared to estimates of less than 5% of
the Chesapeake and Delaware systems. The Delaware
Estuary had the greatest proportion of estimated sand sub-
strate and the least proportion of mud habitat. Less than 10%
of the Delaware was estimated to contain mud substrate,
whereas mud accounted for 32% of Long Island Sound
sediments (Fig. 6-8). The Delaware was also the most turbid.
Whereas water with visibility of greater than 1 m constituted
an estimated 85% (+ 5%) of the area in Chesapeake Bay and
nearly 100% of Long Island Sound, it occurred in less than
65% (± 18%) of the Delaware.
Sediment Type
100
Chesapeake
Bay
Delaware
Bay
Long Island
Sound
Figure 6-8. Sediment types in the three major estuarine sys-
tems in the Virginian Province
Long Island Sound was characterized by the relative absence
of brackish and transitional salinity waters; an estimated
95% of its area contained marine waters (Fig. 6-9). Brackish
and transitional waters each accounted for at least 10% of the
area in the other two estuaries; transitional waters constituted
more than 50% of the area in Chesapeake Bay. The degree
6-14
-------
of water column stratification was determined from delta
sigma-t, a measure used in physical oceanography to
describe density differences between surface and bottom
waters. Long Island Sound was the least stratified of the
three estuaries. One-fifth of the area in the Chesapeake and
Delaware systems had a delta sigma-t of four or greater,
indicating strong density stratification. Delta sigma-t was not
100
Salinity
-E 60
20
• Brackish
D Transitional
ill Marine
Chesapeake
Bay
Delaware
Bay
Long Island
Sound
Figure 6-9. Salinity habitats in the three major estuarine
systems in the Virginian Province: brackish (0 to 5 ppt),
transitional (5 to 18 ppt), and marine (> 18 ppt).
greater than three anywhere in the Long Island Sound.
Presumably, the absence of a large transitional salinity zone
and the lesser degree of water column stratification reflect
fewer major tributaries providing freshwater input .to Long
Island Sound than in the other two major systems in the
Virginian Province.
6-15
-------
Of the largest estuarine
systems In the Virginian
Province, Long Island
Sound had the highest
estimated proportion of
area with oxygen concen-
trations less than 5 ppm;
Chesapeake Bay had the
highest proportion below 2
ppm.
Long Island Sound had the largest estimated percent of area
(51% ± 36%) with bottom dissolved oxygen concentration less
than 5 ppm (Fig. 6-10). These data support previous findings
of degraded conditions in the western basin of this estuary
(Parker and O'Reilly 1991; Welsh and Eller 1991). Although
the Chesapeake Bay had a smaller estimated percentage of
Dissolved Oxygen Concentrations
100
80
1
| 20 ^J
nJ—HHi—
<; 2 ppm
12 to 5 ppm
I > 5 ppm
Chesapeake Delaware Long Island
Bay Bay Sound
Figure 6-10. Bottom dissolved oxygen conditions in the three
major estuarine systems in the Virginian Province.
area with bottom DO
concentrations less than
5 ppm, 19% (±11%) of
its area had concentra-
tions less than 2 ppm
compared to less than
1%(+7%)fortherest
of the province. The
lower concentrations in
the Chesapeake Bay
are due, in part, to the
greater degree of water
column stratification
especially in deeper,
central portions.
Critical
oxygert owaitfott Were
described by carrtpai'teari vyith tvto critical
values, 2 pprfl and S ppffl, The flr$ value
was selected because erf jte perceived
biologist importance, as ctetsf^jned j«
tabarateiy exposure studies, \Nt\4tet$s Un&
second was selected because It is use<|
as a water qualify standard In many
states. The biologically important value is
species-dependent, and the regulatory
value variesamon^stetes In the province.
One of the strengths of Eft/iAP b thaj: an
unbiased estimate for the amount of area
feetow any critical value can be generated
Wite Know 00nfid«w}$< EMAP Wiff
most
critical vatos* for
6-16
-------
Of the largest systems,
sediments that are
estimated to be toxic to
biota were most prevalent
in Chesapeake Bay;
contaminants at
concentrations likely to
cause sublethal effects
were estimated to be most
prevalent in Long Island
Sound.
Sediment toxicity was
most widespread in
Chesapeake Bay. Sedi-
ment from an estimated
8% (+ 5%) of its area
was toxic to amphipods
in bioassays. This was
about double the percent
of area with toxic sedi-
ments found in the
Delaware Estuary and in
Long Island Sound.
Sediment contaminants
at concentrations that
potentially cause at least sublethal biological effects (i.e.,
above ER-L values) were most widespread in Long Island
Sound, covering an estimated 58% (± 36%) of the area. In
contrast, only 39% (± 19%) of the Delaware estuary and 37%
(± 10%) of Chesapeake Bay had similarly elevated sediment
Sediment Contaminants
Sediment Contaminants
Sediment quality criteria for con-
taminant concenttatrans are being ctevef-
oped by EPA but are not yet
available. The ER-L values of Long
and Morgan
-------
Between one-quarter and
one half of the area in
Chesapeake Bay had
benthic resources
characterized by low
number of species and low
abundance of indicator
species.
trations of chlorinated pesticides, notably dieldrin, were above
ER-L values over an estimated 34% (± 19%) of the area and
were more prevalent than high concentrations of metals.
Thirty-six percent (± 12%) of the area of Chesapeake Bay had
"degraded" benthic resources characterized by low number of
species and low number of indicator species (Fig. 6-12). Only
15% (± 9%) of the area in the Delaware and 5% (± 2%) in
Long Island Sound had similar "degradation". Much of the
area in Chesapeake Bay with poor benthic assemblages
occurred in locations where dissolved oxygen stress was most
severe. Several authors have suggested that some of this
degradation may be natural rather than anthropogenic,
resulting from dissolved oxygen depletion induced by
stratification (Officer et al. 1984).
50
40
Degraded Benthic Resources
36
VJ
rt 30
2
S 2°
I 10
Q
CL
i^ - ,
>
15
n rn
Chesapeake Delaware Long Island
Bay Bay Sound
Figure 6-12. "Degraded" benthic resources in the three largest
estuarine systems in the Virginian Province. (See Table 5-5 for
uncertainty).
6-18
-------
Toxicity and elevated
contaminant
concentrations were most
prevalent in fine-grained
sediments.
Specialized Habitats
EMAP's sampling design makes it possible to post-classify
monitoring data to make unbiased estimates of environmental
condition for specific estuarine areas or habitats of interest to
estuarine resource managers and scientists. Examples of
specific subpopulations of interest may include those defined
by salinity or substrate characteristics. The ability to
post-classify is limited only by the number of samples
available in each subpopulation.
Muddy sediments represent a specialized habitat of interest.
Previous studies have shown that both inorganic and organic
contaminants have a greater affinity for binding to fine-grained
sediments (NOAA 1988, 1991). Because of their tendency to
accumulate contaminants, these sediments generally have the
greatest potential to be toxic to estuarine organisms. Fine-
grained sediments (those with a silt-clay content greater than
20%) constituted an estimated 58% of the estuarine area in
the Virginian Province.
Demonstration Project results confirm that elevated sediment
contaminants and sediment toxicity are associated with fine-
grained sediments. Sixteen percent of the area with fine-
grained sediments was toxic, compared to less than 1 % of the
area with sandy sediments. Ninety-nine percent of all toxic
sediments and 91% of sediments having elevated concen-
trations of contaminants were fine-grained.
Degraded benthic assemblages also were more prevalent in
fine-grained sediment habitats, possibly because of the
greater potential for increased contaminant concentrations in
this type of habitat. Seventy-four percent of the estuarine
area with degraded benthic assemblages had fine-grained
sediments. Only 14% of all habitats having coarse sediments
(those with a silt-clay content less than 20%) contained
degraded benthic communities.
Salinity describes another specialized habitat of interest
because it is an important factor controlling ecological
processes and the distribution of organisms (Remane and
Schlieper 1971). Sixty-seven percent of the estuarine area in
6-19
-------
Salinity
(15%)
(28%)
(52%)
• 0 to 0.5 ppt
H 0.5 to 5 ppt
II5 to 18 ppt
El 18 to 25 ppt
D> 25 ppt
Tidal freshwater habitats,
which occur in only 2% of
the province, were propor-
tionately more degraded
than other salinity
habitats.
Ninety percent of the tran-
sitional salinity area in
the province, which is
critical habitat for oyster
production, occurred in
Chesapeake Bay.
Figure 6-13. Salinity habitats in the Virginian Province as a
percent of total estuarine area.
the province had bottom salinities typical of marine
environments (greater than 18 ppt) (Fig. 6-13). Twenty-eight
percent of the province had salinities characteristic of transi-
tional regions between fresh and marine waters (5 to 18 ppt
salt), and 5% of the estuarine area in the province had
brackish water (0 to 5 ppt salt).
Tidal freshwater is a particularly important habitat because it
is the spawning grounds for anadromous fishes such as
striped bass. Although tidal freshwater composed a small per-
centage of the area of the province, the benthic index
suggested that it was proportionately more degraded than
other salinity habitats. Forty-one percent of the tidal
freshwaters in the province contained degraded benthic
assemblages compared to less than 2% of the remaining
area.
Management decisions can be influenced not only by the
proportion of a specific habitat type that is degraded, but also
by the distribution of that type of habitat. Approximately 28%
of the area in the province had transitional salinity (5-18 ppt).
Of this transitional area, approximately 90% is in the
Chesapeake Bay system. Transitional habitats support
spawning and nursery activities of many important estuarine
6-20
-------
species, such as blue
crab and oysters.
Fifty percent of the area
estimated to have
"degraded" benthic
resources was associated
with low dissolved oxygen,
whereas only 12% was
associated with toxic sedi-
ments.
Associations
Jtt flii$ K$ort, s^BopMiOns We identi-
fied. toiw$$i pollution pxpa$Mt£ $«J
biological response indicators. In th£
future, EMAP Intends to examTrm
associations between environmental
condition and external stresses to the
environment Measures of external'
stress include parameters such as
human population deneily, land trse,
industrial activity, and pollution dis-
charge, many of which are presently
monilored as part of other programs,
SWAP, in partnership with other a-
g©neis8,,vw)l compile this information
for me tn idenitfying associations and
make it avatteWe in an integrated,
data b3$& to i
Associations
One strength of the
EMAP sampling design is
the co-location of many
types of measurements
at each site. Co-location
of data facilitates identi-
fication of associations
between biological
responses and measures
of pollutant exposure. Although these associations do not
define cause and effect, they can be used to gauge how bio-
logical response indicators reflect measures of pollutant
exposure and to formulate hypotheses concerning causal
relationships.
Based on association, low dissolved oxygen concentration
appears to contribute more to the degradation of benthic
resources than does sediment toxicity. Fifty percent of the
area with degraded benthic assemblages had concentrations
of dissolved oxygen less than 5 ppm in bottom waters
(Fig. 6-14). In contrast, toxic sediments were found in only
11% of the area with degraded benthos.
Thirty-nine percent of the area with degraded biological re-
sources could not be associated with either low dissolved oxy-
gen or toxic sediments. Some of this area may have had low
dissolved oxygen exposure, but not at the time samples were
collected. About half of this area contained sediment con-
taminant concentrations with the potential to cause sublethal
biological effects. Degradation may have resulted from pollut-
ant exposure from a contaminant that was not measured
during the 1990 Demonstration Project. As EMAP evolves, .,
these areas will be the subject of special studies to identify
any new or emerging types of environmental exposure.
6-21
-------
Degraded Benthic Assemblages
Associations with Pollution Exposure
(39%)
I Unknown Exposure
I Oxygen <; 5 ppm
I Both
I Toxic Sediments
(48%)
Figure 6-14. Association between degraded benthic
assemblages and the dissolved oxygen and sediment toxicity
pollutant exposure indicators.
Data from the 1990 Demonstration Project also were used to
identify associations between nondegraded biological
resources and pollutant exposure. Ninety percent of sites
having nondegraded benthos were sites where dissolved
oxygen measurements were above 5 ppm, and sediments
were not toxic to test biota. Only 5% of the area with nonde-
graded benthos corresponded to area with low dissolved
oxygen conditions, another 5% corresponded to area with
toxic sediments.
Societal Values
Although a major objective of EMAP is to describe the status
of estuarine resources using indicators of ecological condition,
certain characteristics of estuaries that are valued by society
may not be reflected by these indicators. The EMAP
Estuaries Resource Group intends to provide data and
information to address questions commonly asked by the gen-
6-22
-------
Anthropogenic marine
debris (trash) was
estimated to be present in
9-19% of the estuarine
area of the province.
Marine debris was most
prevalent in tidal rivers
and small estuaries.
eral public, including: Are estuaries aesthetically acceptable,
with relatively clear waters and little floating algal scum and
trash? Is the water safe for swimming? Are the fish safe to
eat?
Data collected during the 1990 Demonstration Project allowed
description of some of these attributes. Aesthetics was
addressed by estimating the areal extent of trash and turbid
waters in estuaries of the Virginian Province. The question of
swimmability Was not addressed directly, but the abundance
of Clostridium perfringens, a bacterium indicative of sewage
pollution, was measured to approach the question. Fish
tissue samples were collected for contaminant analysis but
were not processed (see Section 3). Tissue samples will be
processed in future years of the program.
Observations concerning marine litter and debris are important
because debris has multiple deleterious effects on animals
(entanglement and ingestion), impacts fisheries (decreased
market potential for fish and damaged vessels and gear), can
economically affect tourist areas (loss of tourists, beach clean-
up costs), and contributes to public perception of the general
environmental condition of estuaries (Ross et al. 1991). It is
estimated that trash was present over 14% (± 5%) of the
estuarine area in the Virginian Province (Fig. 6-15). Paper
and plastic wastes were found most frequently, followed by
cans and glassware. No trash that could be specifically
identified as medical or hospital waste was found.
Trash was most prevalent in tidal rivers and small estuarine
systems. Trash was found in 32% (± 17%) of the area in tidal
rivers and 23% (± 16%) of the area in small estuarine systems
(Fig. 6-15). In contrast, trash was found in only 10% (± 6%)
of the area in large estuaries. Presumably, proximity to the
shore and urban areas influences the distribution of trash in
estuaries.
Clear waters are valued by society and contribute to the main-
tenance of healthy, productive biological communities. Water
clarity in estuaries is influenced by biological processes (phy-
toplankton blooms, for example), as well as by inputs of
sediment and detritus from streams, rivers, and nonpoint
6-23
-------
High concentrations of
Ciostridium, a bacterial
tracer of sewage pollution,
were estimated to be in 5-
13% of the province
sediments.
source runoff. Although the geomorphology of some portions
of estuaries causes natural turbidity, data collected during the
Demonstration Project allow establishment of a baseline
against which future changes in water clarity can be
assessed. Less than 1% (± 1%) of the estuarine area in the
province had waters with visibility less than 0.3 m; however,
13% (± 4%) of the province had water with visibility less than
1 m, the depth of one's feet when wading in waist-deep water.
Ciostridium perfringens is an obligate-anaerobic bacterium
that is present in the feces of warm-blooded animals. Its
spores survive longer in the sediments than other indicators of
Marine Debris
1
o 30
c
§
| 20
•5
10
0.
32
10
|~1
i :r;j
immmmm
.,':'..''' '
,"' •
f ;
/ , :
' :
23
•4WMMMJ
I „;
''
, f
' ,
Large Large Tidal Small
Estuaries Rivers Systems
Figure 6-15. Percent of area having anthropogenic marine
debris (trash) in the three classes of estuarine resources.
fecal matter (such as coliform bacteria) and, thus, provide a
conservative tracer of sewage pollution (Bisson and Cabelli
1980; Duncanson et al. 1986). Spores of this bacterium were
above background levels in 9% (± 4%) of the province.
Similar to the pattern observed for contaminants, Ciostridium
was most prevalent in small estuarine systems and tidal
rivers. Twenty-one percent of the area in small estuaries
(± 13%) had high levels of Ciostridium compared to 4%
(±2%) in large estuaries (Fig. 6-16).
6-24
-------
Clostridium
40
O 30
c
21
| 20
•5
§ 10
09
D_
0
10
^^^^^^
4
jmnmmm
\ 9
•
•
*mmmmmmm
Large Large Tidal Small
Estuaries Rivers Systems
Figure 6-16. Percent of area in the three classes of estuarlne
resources where high sediment concentrations of Clostridium
perfringens spores were found.
Integration of Estuarine Conditions
As EMAP reaches full implementation, a single index probably
will be developed to summarize the overall condition of
estuaries in the Virginian Province. That index may incorpo-
rate measures of fishability, swimmability, and aesthetics,
combined with measures of biotic integrity based on benthic
and fish assemblages (Fig. 6-17). Data to evaluate fishability
are not yet available, but it is possible to begin constructing
an overall index based upon several of the indicators
measured in the Demonstration Project. The methods for
combining these indicators are presented in Section 2.
6-25
-------
About one-third of the
estuarine area in the
Virginian Province was
estimated to exhibit some
form of anthropogenic or
natural environmental
"degradation".
Large tidal rivers exhibited
the most environmental
"degradation".
Indicators relating to biotic integrity and societal values were
used to estimate overall environmental conditions in the
estuaries of the Virginian Province. Thirty-six percent (± 7%)
of the estuarine area in the province showed evidence of
degraded biological resources or was impaired with respect to
its ability to support activities valued by society (Fig. 6-18).
Expressed on an areal basis, an estimated 8,436 km2 of the
23,574 km2 total area in estuaries of the Virginian Province
were potentially degraded.
The locations of degraded biological resources were different
from those impaired with respect to societal values. Both sets
of conditions were found in an estimated 7% of the estuarine
area, whereas degraded biological conditions alone were
found in an estimated 17% of the province, and loss of
societal value was associated with an estimated 12% (Fig.
6-18). This suggests that the visual symptoms of
environmental degradation (trash, water clarity) or factors that
might limit human contact with the water do not necessarily
indicate biologicaj degradation.
An estimated 64% of the area in large tidal rivers showed
evidence of either degraded biological resources or impaired
societal value. In contrast, only an estimated 30% of small
estuarine systems and 35% of large estuaries showed such
evidence. Although tidal rivers were proportionately more
degraded than large estuaries or small estuarine systems, the
largest area of degraded resources occurred in the large estu-
aries. Approximately 70% of all degraded estuarine
resources, representing 5,835 km2, were found in large
estuaries. In contrast, 1,437 km2 of degraded resources were
found in small estuarine systems and 1,164 km2 in tidal rivers.
Program Direction
Using a probability-based sampling design to develop regional
estimates of ecological status is a new approach to monitoring
estuarine health. The preliminary evaluation of environmental
status presented in this section was the first attempt by
EMAP-E to convey the types of information that can be
6-26
-------
egrated
c
«s
s
jc
Q.
LU
JB1
o
w
D)
I
(0
O 0>
"*i
II
E
«
s
Si
£ 0>
rag
u. to
6-27
-------
Integrated Conditions
Poor Biotic Integrity (17%)
Both (7%)
Impaired Societal Values (12%)
Undegraded Resources (64%)
Figure 6-18. Summary of environmental conditions in the
estuaries of the Virginian Province.
Preliminary Nature of Estimates
generated using this
approach. The topics ad-
dressed were defined in
partnership with potential
users of the program's
data. This partnership is
crucial to ensure that
EMAP monitoring and
assessment activities
produce information that is
relevant to users' needs.
EMAP-E will continue to
expand the list of ques-
tions towards which future
assessment activities are
directed. That expansion
will be based, in part, on
continuing interactions
between the Estuaries
DResource Group and its data users. The list will also
expand as EMAP obtains additional data. During 1991, sam-
pling continued in the Virginian Province, allowing examination
and subject to
revision, VaTfdatfon Of
[ntfoatoNS fog* ftdt beert
and will require monitoring date* trow
at least' two years (see section 4j<
the estimates are based upon a,
single year of date £MA|>^
assessments typieally wilt b& eon-
ducted ttsing a tour-year running
average to reduoe the effects of
niatural Interannuaf variation in
climate and other influences. Esti-
mates based upon one year of data
are intended primarily to provide th&
reader with examples of the kinds of
EMAP is
after font y
-------
of interannual variability in ecological condition. A demonstra-
tion project was conducted in the Gulf of Mexico (Louisianian
Province) in 1991, which will permit comparison of relative
status of estuaries in divergent portions of the country.
Similar comparisons and aggregation will be used to complete
a national assessment of estuarine condition after EMAP-
Estuaries is fully implemented. Finally, EMAP is also intended
as a trend detection program; trends questions will be empha-
sized after EMAP has completed two full sampling cycles.
Throughout all EMAP monitoring and reporting activities, there
is continued commitment to building partnerships with
potential users and to sharing data and information to meet
the heeds of decision makers. Comments concerning this
report, monitoring data, and the activities of the estuarine
component of EMAP should be directed to:
Dr. Kevin Summers
Technical Director of EMAP-Estuaries,
Environmental Research Laboratory,
Gulf Breeze, FL 32561,
Telephone (904)934-9244,
EPAEMAIL: SUMMERS.KEVIN
6-29
-------
-------
SECTION 7
SUMMARY AND CONCLUSIONS
The 1990 EMAP-Estuaries Demonstration Project in the Virginian Province collected data and
information to address six objectives (Holland 1990). The degree to which these objectives
were met is evaluated in this report. These evaluations will be used to refine the program in
future years. Although the results of the Demonstration Project are now of most use to those
within the program, results are also relevant to users of its products. The ultimate users of
the products of EMAP provided comments during the design, implementation, and assessment
phases of the 1990 Demonstration Project and are important partners in the future progress of
the program. The utility of the results of the Demonstration Project to the users of EMAP data
and information is summarized below for each of the six objectives.
Objective 1: Demonstrate the value of regional monitoring using an unbiased sampling
approach as a basis for assessing the condition of estuarine resources.
A preliminary evaluation of the environmental condition of estuaries in the Virginian Province
was presented in Section 6. The full value of this type of assessment will not be realized until
users of EMAP products become familiar with and begin to use the results presented in the
report; nonetheless, the utility of the results is already becoming evident. The assessment
provided information about estuaries in the mid-Atlantic region that, prior to EMAP, was
difficult to obtain in all but the most well-studied of estuaries. This information is unique in its
ability to define regional estimates of the areal extent of specific physical habitats, biological
conditions, and pollutant exposure with known confidence. The assessment results identified
specific habitats and classes of estuaries that should be of greater interest because of the
extent or magnitude of environmental impact identified within them. Although based upon only
one year of EMAP monitoring data, these results are already of interest to environmental
resource managers establishing estuarine research directions and management priorities.
Objective 2: Evaluate the ability of a suite of indicators of environmental quality to
discriminate between polluted and unpolluted sites on a regional scale.
The development and testing of estuarine indicators of biological response, pollutant
exposure, and habitat was described in Section 4. Based on the information presented in that
section, a suite of indicators that will form the core of future EMAP monitoring activities in
estuaries was identified. One of the most important accomplishments was the development
and application of a framework for calibrating and validating biological indicators. This is
particularly noteworthy because development of biological criteria is an Agency priority, and
biological indicators that are applicable over a range of latitudes and habitat types have not
7-1
-------
been identified previously. Several years of additional data will be required to fully validate
the biological indicators developed here, and the framework identified for the 1990
Demonstration Project provides a basis for conducting that validation.
Objective 3: Establish standardized methods for monitoring indicators of ecological
status and trends in estuaries.
The collection and sample processing methods used during the 1990 Demonstration Project
(Strobel 1990; USEPA 1991) were summarized in Section 2. These methods have been
tested and evaluated and are now ready to be used directly by other estuarine monitoring
programs. Development and use of standardized methods is important if EMAP is to be able
to assess ecological conditions over a wide range of systems. Standardized methods will
form the basis for comparisons across provinces in the future. EMAP is promoting these
methods and the associated QA protocols to EPA Regions, states, and local institutions
responsible for monitoring to facilitate achieving comparability and integration of information
from multiple monitoring programs and estuaries. Presently, such integration is limited due to
the variety of methods being used (NRG 1990).,
Objective 4: Obtain data on regional-scale variability in ecological parameters to
evaluate and refine the sampling design.
Results from the Demonstration Project were used to evaluate the EMAP estuarine sampling
design for its ability to define environmental status and trends (Section 5). This is the
beginning of a process that requires multiple years of EMAP monitoring data as well as
environmental monitoring data from other, long-term monitoring programs. The process
depends upon the involvement of potential clients to help define the power for detecting
changes (i.e., trends) required by EPA programs. Results thus far suggest that no major
modifications of the EMAP sampling design are required. The present design produces
unbiased estimates with an acceptable degree of precision; however, minor modifications may
improve the precision of estimates of environmental condition for particular subclasses of
estuaries or habitats of interest, and these modifications need to be evaluated further.
Objective 5: Develop analytical procedures for using regional monitoring data to
assess the ecological status of estuaries, and apply the procedures to
establish baseline conditions in the Virginian Province
The assessment of estuarine condition presented in Section 6 begins to identify baseline
conditions for the estuaries of the Virginian Province and provides examples of the kind of
environmental baseline that can be defined using monitoring data collected in EMAP. This
7-2
-------
baseline condition is relative to regional environmental reference conditions defined using
Demonstration Project data. Greater precision for this estimate will be possible when a full
cycle (four years) of monitoring data becomes available. The baseline will then be used to
evaluate changes in environmental condition over time (i.e., trends). The developers of the
program realize that the influence of previous anthropogenic stress and disturbance may
already be expressed at the regional reference sites selected for the 1990 Demonstration
Project in subtle ways that are not evident in the measures of pollution used for the project.
For that reason, EMAP is working with regional clients to identify historical data that can be
used to evaluate present reference conditions.
Objective 6: Identify and resolve logistical problems associated with conducting a
regional monitoring program in estuaries.
The logistical considerations involved in conducting a large-scale monitoring program such as
EMAP were evaluated in Section 3. The lessons learned from completing the 1990
Demonstration Project are relevant to EMAP data users and to others involved in designing
and implementing large monitoring programs. These lessons include the importance of
intensive field-crew training, standardization of sampling equipment and procedures, well-
defined field and laboratory quality assurance protocols, state-of-the-art communications and
information management systems, and advanced preparation of laboratories for sample
processing.
Although each of the objectives of the Demonstration Project was addressed in this report, the
evaluations conducted for each objective represent only an initial step. Evaluations are
iterative and depend upon the availability of monitoring data from subsequent years,
comparisons with historical and existing monitoring programs, and continued participation by
users of products of the program.
7-3
-------
-------
SECTION 8
REFERENCES
Ackermann, F., H. Bergmann, and U. Schielchert. 1983. Monitoring of heavy metals in
coastal and estuarine sediments - A question of grain size: <20 |j,m versus <60 jxm. Environ.
Technol. Lett. 4:317-328.
Adelman, I.R., and L.L. Smith, Jr. 1972. Toxicity of hydrogen sulfide to goldfish (Carassius
auratus) as influenced by temperature, oxygen, and bioassay techniques. J. Fish. Res. Bd.
Canada 29:1309-1317.
Albert, R.C. 1988. The historical context of water quality management for the Delaware
estuary. Estuaries 11:99-107.
Allen, L.G., M.H. Horn, F.A. Edmands II, and C.A. Usui. 1983. Structure and seasonal
dynamics of the fish assemblage in the Cabrillo Beach area of Los Angeles Harbor, California.
Bull. Southern California Acad. Sci. 82(2):47-70.
Aller, R.C. 1982. The effects of macrobenthos on chemical properties of marine sediments
and overlying waters. Pages 53-102. In: P.L McCall and M.J.S. Tevesz, eds. Animal-
Sediment Relations. Plenum Press, New York.
APHA. 1985. Standard Methods for the Examination of Water and Wastewater. 17th Edition
American Public Health Association, Washington, DC.
ASTM (American Society of Testing and Materials). 1991. Standard guide for conducting
10-day static sediment toxicity tests with marine and estuarine amphipods. Annual Book of
ASTM Standards Volume 11.04:1052-1075.
Bailey, R.G. 1983. Delineation of ecosystem regions. Environ. Manage. 7:365-373.
Bechtel, T.J., and B.J. Copeland. 1970. Fish species diversity indices as indicators of
pollution in Galveston Bay, Texas. Contrib. in Mar. Sci. 15:103-132.
Becker, D.S., G.R. Bilyard, and T.C. Ginn. 1990. Comparisons between sediment bioassays
and alterations of benthic macroinvertebrate assemblages at a marine Superfund site:
Commencement Bay, Washington. Environ. Toxicol. Chem. 9:669-685.
Bell, S.S. and B.C. Coull. 1978. Field evidence that shrimp predation regulates meiofauna.
Oecologia 35:141-148.
8-1
-------
Berkman. H.E., C.F. Rabeni, and T.P. Boyle. 1986. Biomonitors of stream quality in
agricultural areas: Fish versus invertebrates. Environ. Manage. 10:413-419.
Bilyard, G.R. 1987. The value of benthic infauna in marine pollution monitoring studies. Mar.
Pollut Bull. 18:581-585.
Bisson, J.W and V.J. Cabelli. 1980. Clostridium perfringens as a water quality indicator. J.
Water Pollut. Cont. Fed. 52(2):241-248.
Boesch, D.F. 1973. Classification and community structure of macrobenthos in the Hampton
Roads area, Virginia. Mar. Biol. 21:226-244.
Boesch, D.F. and R. Rosenberg. 1981. Response to stress in marine benthic communities.
Pages 179-200. In: G.W. Barret and R. Rosenberg, eds. Stress Effects on Natural
Ecosystems. John Wiley and Sons, New York.
Bousfield, E.L. 1973. Shallow-water Gammaridean amphipoda of New England. Cornell
University Press, Ithaca, NY. 312 pp.
Breitburg, D.L. 1990. Nearshore hypoxia in the Chesapeake Bay: Patterns and relationships
among physical factors. Estuar. Coast. Shelf Sci. 30:593-609.
Brongersma-Sanders, M. 1957. Mass mortality in the sea. Pages 941-1010. In: J.W.
Hedgpeth, ed. Treatise on Marine Ecology and Paleoecology. Mem. Geol. Soc. Am. 67.
Broutman, M.A. and D.L. Leonard. 1988. National estuarine inventory: The quality of the
shellfish growing waters in the Gulf of Mexico. NOAA, Strategic Assessment Branch,
Rockville, MD. 43 pp.
Bruland, K.W., K. Bertine, and E.D. Goldberg. 1974. Metal pollution in southern California
coastal zone. Environ. Sci. Technol. 8:425-432.
Buhler, D.R. and D.E. Williams. 1988. The role of biotransformation in toxicity. Fish. Aquat.
Toxicol. 11:303-311.
Burton, D.T., LB. Richardson, and C.J. Moore. 1980. Effect of oxygen reduction rate and
constant low dissolved oxygen concentrations on two estuarine fish. Trans. Amer. Fish. Soc.
109:552-557.
Cabelli, V.J. and D. Pederson. 1982. The movement of sewage sludge from the New York
Bight dumpsite as seen from Clostridium perfringens spore densities. Pages 995-999. In:
IEEE Council on Oceanic Engineering, Oceans '82 Conference Record. September 20-22,
1982, Washington, DC.
8-2
-------
Capuzzo, J.M., M.N. Moore, and J. Widdows. 1988. Effects of toxic chemicals in the marine
environment: Predictions of impacts from laboratory studies. Aquat. Toxicol. 11:19-28.
Carriker, M.R. 1967. Ecology of estuarine benthic invertebrates: A perspective. Pages 442-
487. In: G.H. Lauff, ed. Estuaries, Publ. No. 83. American Association for the Advancement
of Science, Washington, DC.
CEQ (Council on Environmental Quality). 1990. Environmental quality. The twenty-first
annual report of the Council on Environmental Quality together with the President's message
to Congress. U.S. Government Printing Office, Washington, DC. 494 pp.
Chao, LIM. and J.A. Mustek. 1977. Life history, feeding habitats, and functional morphology
of juvenile scianid fishes in the York River estuary. Fish. Bull. 75:657-702.
Chapman, P.M. 1988. Marine sediment toxicity tests. Pages 391-402. In: J.J. Lichtenberg,
F.A. Winter, C.I. Weber and L. Fredkin, eds. Chemical and Biological Characterization of
Sludges, Sediments, Dredge Spoils, and Drilling Muds. STP 976. ASTM, Philadelphia, PA.
Cloern, J.E. 1982. Does the benthos control phytoplankton biomass in South San Francisco
Bay? Mar. Ecol. Prog. Ser. 9:191-202
Cochran, W.G. 1977. Sampling Techniques. 3rd Edition. John Wiley, New York.
The Conservation Foundation. 1988. Protecting America's Wetlands: An Action Agenda.
The Conservation Foundation, Washington, DC.
Culliton, T.J., M.A. Warren, T.R. Goodspeed, D.G. Remer, C.M. Blackwell, and J.J.
McDonough. 1990. 50 years of population change along the nation's coasts, 1960-2010.
Second Report of Coastal Trends Service. U.S. Department of Commerce, NOAA, Strategic
Assessment Branch, Rockville, MD. 41 pp.
Dauer, D.M., R.M. Ewing, J.A. Ranasinghe, and A.J. Rodi, Jr. 1989. Macrobenthic
communities of the lower Chesapeake Bay, Chesapeake Bay Program, March 1985-June
1988. Prepared for the Virginia Water Control Board, Richmond, VA. Dept. of Biological
Sciences, Old Dominion University, Norfolk, VA.
Dinnel, P.A.. 1990. Annotated bibliography of bioassays related to sediment toxicity testing in
Washington State. FRI-UW-9017. Submitted to Seattle District, U.S. ACOE, University of
Washington, Fisheries Research Institute.
Duncan, G.J., and G. Kalton. 1987. Issues of design and analysis of surveys across time.
Int. Stat. Rev. 55:97-117.
8-3
-------
Duncanson, R.A., D.L. Saad, H. Namdari, M.R. Rodgers, and V.J. Cabelli. 1986. Monitoring
fecal contamination of marine waters by measuring C. perfringens spores in the sediments.
Pages 560-565. In: Conference Record, Oceans '86, September 23-25, 1986, Washington,
DC.
Emerson, D.J. and V.J. Cabelli. 1982. Extraction of Clostridium perfringens spores from
bottom sediment samples. Appl. Environ. Microbiol. 44:1144.
Fauchald, K. and P.A. Jumars. 1979. The diet of worms. A study of polychaete feeding
guilds. Oceanogr. Mar. Biol. Ann. Rev. 17:193-284
FDA (Food and Drug Administration), Shellfish Sanitation Branch. 1971. National shellfish
register of classified estuarine waters. Northeast Technical Services Unit, Davisville, Rl.
FDA, Shellfish Sanitation Branch, National Oceanic and Atmospheric Administration, Strategic
Assessment Branch, and National Marine Fisheries Service. 1985. National shellfish register
of classified estuarine waters. Rockville, MD.
Fein, G.G., J.L Jacobson, S.W. Jacobson, P.W. Schwartz, and J.K. Dowler. 1984. Prenatal
exposure to polychlorinated biphenyls: Effects on birth size and gestational age. Pediatr.
105:315-320.
Forstner, U. and G.T.W. Wittmann. 1981. Metal Pollution in the Aquatic Environment. 2nd
Revised Edition. Springer-Verlag, New York.
Frithsen, J.B. 1989. The benthic communities within Narragansett Bay. An assessment
completed for the Narragansett Bay Project. University of Rhode Island, Kingston, Rl.
Frithsen, J.B., M.C. Fabrizio, J. Gerritsen, A.F. Holland, G.E. Saul, and S.B. Weisberg. 1991.
Example environmental assessment report for estuaries. EPA/600/04-91/026. U.S.
Environmental Protection Agency, Atmospheric Research and Exposure Assessment
Laboratory, Research Triangle Park, NC.
Ginn, T.G., and R.A. Pastorok. 1992. Assessment and management of contaminated
sediments in Puget Sound. Pages 371-401. In: G.A. Burton, Jr., ed. Sediment Toxicity
Assessment. Lewis Publishers, Ann Arbor, Ml.
Grizzle, R.E., and C.A. Penniman. 1991. Effects of organic enrichment on estuarine
macrofaunal benthos: a comparison of sediment profile imaging and traditional methods.
Mar. Ecol. Prog. Ser. 74:249-262.
8-4
-------
Haedfieh, R.L., and S.O. Haedrich. 1974. A seasonal survey of the fishes in the Mystic
River, a polluted estuary in downtown Boston, Massachusetts. Est. Coastal Mar. Sci. 2:59-
73.
Hargis, W.J., Jr. and D.S. Haven. 1988. Rehabilitation of the troubled oyster industry of the
lower Chesapeake Bay. J.Shellfish Res. 7:271-279.
Harper, D.E., D.L. McKinney, R.R. Saltzer, and R.J. Case. 1981. The occurrence of hypoxic
bottom water off the upper Texas coast and its effect on benthic biota. Contrib. in Mar. Sci.
24:53-79.
Hilsenhoff, W.L. 1987. An improved biotic index of organic stream pollution. Great Lakes
Entomol. 20:31-39.
Hinga, K.R. 1988. Seasonal predictions for pollutant scavenging in two coastal environments
using a model calibration based upon thorium scavenging. Mar. Environ. Res. 26:97-112.
Hinga, K.R. 1992. Contaminated sediments in Narragansett Bay: Severity and extent.
Prepared for the Narragansett Bay Project. Graduate School of Oceanography, University of
Rhode Island, Narrangansett, Rl.
Holland, A.F., ed. 1990. Near coastal program plan for 1990: Estuaries. EPA/600/4-90/033.
U.S. Environmental Protection Agency, Environmental Research Laboratory, Narragansett, Rl.
Holland, A.F., A. Shaughnessy, and M.H. Hiegel. 1987. Long-term variation in mesohaline
Chesapeake Bay macrobenthos: Spatial and temporal patterns. Estuaries 10:277-245.
Holland, A.F., AT. Shaughnessy, LC. Scott, V.A. Dickens,. J.A. Ranasinghe, and J.K.
Summers. 1988. Progress report: Long-term benthic monitoring and assessment program
for the Maryland portion of Chesapeake Bay (July 1986-October 1987). PPRP-LTB/EST-88-1.
Prepared for Maryland Power Plant Research Program and Maryland Department of the
Environment, Office of Environmental Programs. Versar, Inc., ESM Operations, Columbia,
MD.
Holland, A.F., A.T. Shaughnessy, LC. Scott, V.A. Dickens, J. Gerritsen, J.A. Ranasinghe.
1989. Long-term benthic monitoring and assessment program for the Maryland portion of
Chesapeake Bay: Interpretive report. CBRM-LTB/EST-2. Prepared for Maryland Department
of Natural Resources, Power Plant Research Program. Versar, Inc., ESM Operations,
Columbia, MD.
8-5
-------
Honeyman, B.D. and P.M. Santschi. 1988. Metals in aquatic systems: Predicting their
scavenging residence times from laboratory data remains a challenge. Environ. Sci. Technol.
22:862-871.
Horn, M.H. 1980. Diel and seasonal variation in abundance and diversity in shallow-water
fish populations in Morro Bay, California. U.S. Nat. Mar. Fish. Ser. Fish. Bull. 78:759-770.
Hunsaker, C.T. and D.E. Carpenter, eds. 1990. Environmental Monitoring and Assessment
Program: ecological indicators. EPA/600/3-90/060 U.S. Environmental Protection Agency,
Office of Research and Development, Washington, DC.
Jorgensen, C.B. 1966. Biology of Suspension Feeding. Pergamon, Oxford. 357 pp.
Jorgensen, B.B. 1980. Seasonal oxygen depletion in the bottom waters of a Danish fjord and
its effects on the benthic community. Oikos 34:68-76.
Justic, D., T. Legovic and L Rottini-Sanori. 1987. Trends in oxygen content 1911-1984 and
occurrence of benthic mortality in the Northern Adriatic Sea. Estuar. Coast. Shelf Sci. 25:435-
445.
Karr, J.R., and D.R. Dudley. 1981. Ecological perspective on water quality goals. Environ.
Manage. 5:55-68.
Karr, J.R., R.C. Heidinger, and E.H. Helmer. 1985. Effects of chlorine and ammonia from
wastewater treatment facilities on biotic integrity. J. Water Pollut. Control Fed. 57:912-915.
Karr, J.R., K.D. Fausch, P.L Angermeier, P.R. Yant, and I.J. Schlosser. 1986. Assessing
biological integrity in running waters: a method and its rationale. Special Publication 5, Illinois
Natural History Survey, Champaign, IL
Keith, LH. 1991. Environmental Sampling and Analysis: A Practical Guide. Lewis
Publishers, Chelsea, Ml. 143 pp.
Kemp. W.M. and W.R. Boynton. 1980. Influence of biological and physical processes on
dissolved oxygen dynamics in an estuarine system: Implications for measurements of
community metabolism. Estuar. Coast. Mar. Sci. 11: 407-431.
Kemp, W.M., and W.R. Boynton. 1981. External and internal factors regulating metabolic
rates of an estuarine benthic community. Oecologia 51:19-27.
8-6
-------
Kjelson, M.A., and D.R. Colby. 1977. The evaluation and use of gear efficiencies in the
estimation of estuarine fish abundance. Pages 415-424. In: M. Wiley, ed. Estuarine
Processes, Vol. 2. Academic Press, New York.
Klapow, L.A. and R.H. Lewis. 1979. Analysis of toxicity data for California marine water
quality standards. J. Water Pollu. Cont. Fed. 51 (8):2051-2070.
Klinkhammer, G.P. and M.L. Bender. 1981. Trace metal distributions in the Hudson River
estuary. Estuar. Coast. Shelf Sci. 12: 629-643.
Knapp, C.M., D.R.Marmorek, J.P. Baker, K.W. Thornton, J.M. Klopatek, and D.F. Charles.
1990. The indicator development strategy for the Environmental Monitoring Assessment
Program. EPA 600/3-91/023. U.S. Environmental Protection Agency, Office of Research and
Development, Washington, DC.
Kouadio. I. and J.H. Trefry. 1987. Sediment trace metal contamination in the Ivory Coast,
West Africa. Water Air Soil Pollut. 32: 145-154.
Kramer D.L 1987. Dissolved oxygen and fish behavior. Environ. Biol. Fish. 18(2):81-92.
Kuo, A.Y., and B.J. Neilson. 1987. Hypoxia and salinity in Virginia estuaries. Estuaries
10:277-283.
Lachenbruch, P.A. 1975. Discriminant Analysis. Hafner, New York.
Leonard, P.M., and D.J. Orth. 1986. Application and testing of an index of biotic integrity in
small coolwater streams. Trans. Am. Fish. Soc. 115:401-414.
Leonard, D.L, M.A. Broutman, and K.E. Harkness. 1989. The quality of shellfish growing
areas on the east coast of the United States. National Oceanic and Atmospheric
Administration OAD. Rockville, MD.
Lippson, A.J., M.S. Haire, A.F. Holland, F. Jacobs, J. Jensen, R.L. Moran-Johnson, T.T.
Polgar, and W.R. Richkus. 1979. Environmental Atlas of the Potomac Estuary. Prepared for
Maryland Department of Natural Resources, Power Plant Siting Program. Martin Marietta
Corporation, Baltimore, MD.
Long, E.R. and L.G. Morgan. 1990. The potential for biological effects of sediment-sorbed
contaminants tested in the National Status and Trends Program. NOAA Technical
Memorandum NOS OMA 52. US Department of Commerce, National Oceanic and
Atmospheric Administration, National Ocean Service, Rockville, MD.
8-7
-------
Long, E.R., M.F. Buchman, S.M. Bay, R.J. Breteler, R.S. Carr, P.M. Chapman, J.E. Hose, A.L
Lissner, J. Scott, and D.A. Wolfe. 1990. Comparative evaluation of five toxicity tests with
sediments from San Francisco Bay and Tomales Bay, California. Environ. Toxicol. Chem.
9:1193-1214.
Malins, D.C. 1989. Sediment contamination and marine ecosystems: Potential risks to
humans. Pages 155-164. In: Contaminated Marine Sediments -- Assessment and
Remediation. National Academy of Sciences. National Academy Press, Washington, DC.
Malins, D.C., B.B. McCain, D.W. Brown, S.L Chan, M.S. Myers, J.T. Landahl, P.G. Prohaska,
A.J. Friedman, L.D. Rhodes, D.G. Burrows, W.D. Gronlund, and H.O. Hodgins. 1984.
Chemical pollutants in sediments and diseases in bottom-dwelling fish in Puget Sound,
Washington. Environ. Sci. Technol. 18:705-713.
Malins, D.C., B.B. McCain, J.T. Landahl, M.S. Myers, M.M. Krahn, D.W. Brown, S.L. Chan,
and W.T. Roubal. 1988. Neoplastic and other diseases in fish in relation to toxic chemicals:
An overview. Pages 43-67. In: D.C. Malins and A. Jensen, eds. Aquatic Toxicology, Toxic
Chemicals, and Aquatic Life: Research and Management. Elsevier Science, Amsterdam.
Mare, M.F. 1942. A study of the marine benthic community with special reference to the
microorganisms. J. Mar. Biol. Assoc. U.K. 25:517-545.
May, E.B. 1973. Extensive oxygen depletion in Mobile Bay, Alabama. Limn.ol. Oceanogr.
18:353-366.
Mearns, A.J., and M. Sherwood. 1974. Environmental aspects of fin erosion and tumors in
southern California Dover sole. Trans. Amer. Fish. Soc. 103:799-810.
Messer, J.J. 1990. EMAP indicator concepts. Chapter 2. In: C.T. Hunsaker and D.E.
Carpenter, eds. Environmental Monitoring and Assessment Program: ecological indicators.
EPA 600/3-90/060. U.S. Environmental Protection Agency, Office of Research and
Development, Washington, DC.
Miller, H. 1984. Dynamics of fish diseases in the lower Elbe River. Helgolander
Meeresunters. 37:389-413.
Miller, D.L., and thirteen coauthors. 1988. Regional applications of an index of biotic integrity
for use in water resource management. Fisheries 13(5):12-20.
Millero, F.J., and A. Poisson. 1981. International one-atmosphere equation of state of
seawater. Deep-Sea Research 28A:625-629.
8-8
-------
Morganthau, T. 1988. Don't go near the water: Is it too late to save our dying coasts?
Newsweek, August 1, 42-47. . -
Nacci, D., R. Walsh, and E. Jackim. 1987. Guidance manual for conducting sperm cell tests
with the sea urchin, Arbacia punctulata, for use in testing complex effluents. Contribution No.
X105. In: S.C. Schimmel, ed. Users Guide to the Conduct and Interpretation of Complex
Effluent Toxicity Tests at Estuarine/Marine Sites. U.S. Environmental Protection Agency,
Environmental Research Laboratory, Narragansett, Rl.
Nixon, S.W., C.D. Hunt and B.L Nowicki. 1986. The retention of nutrients (C,N,P), heavy
metals (Mn, Cd, Pb, Cu), and petroleum hydrocarbons in Narragansett Bay. Pages 99-122.
In: P. Lasserre and J.M. Martin, eds. Biogeochernical Processes at the Land-sea Boundary.
Elsevier, NY.
NOAA (National Oceanic and Atmospheric Administration). 1987. NOAA estuarine and
coastal ocean science framework. U.S. Department of Commerce, National Oceanic and
Atmospheric Administration, Estuarine Program Office, Washington, DC,
NOAA. 1988. Federal plan for ocean pollution research development, and monitoring: Fiscal
years 1988-1991. Prepared by the National Ocean Pollution Program Office for the National
Ocean Pollution Policy Board, Rockville, MD.
NOAA. 1990. Estuaries of the United States: Vital statistics of a national resource base. A
special NOAA 20th Anniversary report. National Oceanic and Atmospheric Administration,
Rockville, MD.
NOAA. 1991. Second summary of data on chemical concentrations in sediments from the
National Status and Trends Program. NOAA Technical Memorandum NOS OMA 59. US
Department of Commerce, National Ocean Service, Rockville, MD.
NRC (National Research Council). 1989. Contaminated marine sediments -assessment and
remediation. National Academy Press, Washington, DC. 493 pp.
NRC. 1990. Managing Troubled Waters: The Role of Marine Environmental Monitoring.
National Academy Press, Washington, DC.
O'Connor, T.P. 1990. Coastal Environmetnal Quality in the United States, 1990: Chemical
contamination in sediment and tissues. A special NOAA 20th Anniversary Report. National
Oceanic and Atmospheric Administration, Rockville, MD.
8-9
-------
O'Connor, J.S., J.J. Ziskowski, and R.A. Murchelano. 1987. Index of pollutant-induced fish
and shellfish disease. Special Report. National Oceanic and Atmospheric Administration,
NDS, Rockville, MD.
O'Connor, B.D.S., J. Costelloe, B.F. Keegan, and D.C. Rhoads. 1989. The use of REMOTS
technology in monitoring coastal enrichment resulting from mariculture. Mar Pollut Bull
8:384-390.
Officer, C.B., T.J. Smayda, and R. Mann. 1982. Benthic filter feeding: A natural
eutrophication control. Mar. Ecol. Prog. Ser. 9:203-210.
Officer, C.B., R.B. Biggs, J.L Taft, LE. Cornin, M.A. Tyler, and W.R. Boynton. 1984.
Chesapeake Bay anoxia: Origin, development, and significance. Science 223:22-27.
Olsen, S., D.D. Robadue, and V. Lee. 1980. An interpretive atlas of Narragansett Bay.
Coastal Resources Center, University of Rhode Island, Mar. Bull. 40.
OTA (Office of Technology Assessment). 1987. Wastes in Marine Environments.
Washington, DC.
Overton, W.S. 1989. Design report of the environmental monitoring and assessment
program. U.S. EPA, Environmental Research Laboratory, Corvallis, OR.
Overton, W.S., and S.V. Stehman. 1987. An empirical investigation of sampling and other
errors in the national stream survey: analysis of a replicated sample of streams. Technical
Report 119, Dept. of Statistics, Oregon State University, Corvallis, OR.
Overton, W.S., D. White, and D.L. Stevens. 1990. Design report for EMAP: Environmental
Monitoring Assessment Program. EPA/600/3-91/053. U.S. Environmental Protection Agency,
Office of Research and Development, Washington, DC.
Parker, C.A. and J.E. O'Reilly. 1991. Oxygen depletion in Long Island Sound: a historical
perspective. Estuaries 14:76-120.
Parker, C.A., J.E. O'Reilly, and R.B. Gerzoff. 1986. Oxygen depletion in Long Island Sound.
Historical Trends Assessment Program, National Oceanic and Atmospheric Administration,
OAD, Rockville, MD.
Parsons, T.R., M. Takahashi and B. Margrave. 1977. Biological Oceanographic Processes.
Second Edition. Pergamon, New York.
8-10
-------
Pearl, H.W. 1988. Nuisance phytoplankton blooms in coastal, estuarine, and inland waters.
Limnol. Oceanogr. 33:823-847.
Pearson, T.H. and R. Rosenberg. 1978. Macrobenthic succession in relation to organic
enrichment and pollution of the marine environment. Oceanogr. Mar. Biol. Ann. Rev. 16:229-
311.
Plafkin, J.L., M.T. Barbour, K.D. Porter, S.K. Gross, and R.M. Hughes. 1989. Rapid
bioassessment protocols for use in streams and rivers: Benthic macroinvertebrates and fish.
EPA/444/4-89/001. U.S. Environmental Protection Agency, Assessment and Watershed
Protection Division, Washington, DC.
Plumb, R.H. 1981. Procedure for handling and chemical analysis of sediment and water
samples. Technical Report EPA/CE-81-1. Prepared for the U.S. Environmental Protection
Agency/ Corps of Engineers Technical Committee on Criteria for Dredge and Fill Material.
Environmental Laboratory, U.S. Army Waterways Experiment Station, Vicksburg, MS.
Poole, H.H., and W.R.G. Atkins. 1929. Photo-electric measurements of submarine
illumination throughout the year. J. Mar. Biol. Assn. U.K. 16:297-324.
Rabalais, N.N., M.J. Dagg, and D.F. Boesch. 1985. Nationwide review of oxygen depletion
and eutrophication in estuarine and coastal waters: Gulf of Mexico. Louisiana Universities
Marine Consortium.
Reish, D.J., and J.L. Barnard. 1960. Field toxicity tests in marine water utilizing the
polychaetous annelid Capitella capitata (Fabricius). Pacif. Natur. 1:1-8.
Remane, A., and C. Schlieper. 1971. Biology of Brackish Water. John Wiley and Sons, New
York.
Rhoads, D.C. 1974. Organism-sediment relations on the muddy sea floor. Oceanogr. Mar.
Biol. Ann. Rev. 12:263-300.
Rhoads, D.C., and LF. Boyer. 1982. The effects of marine benthos on physical properties of
sediments. Pages 3-52. In: P.L. McCall and MJ.S. Tevesz, eds. Animal-Sediment Relations.
Plenum Press, New York.
Rhoads, D.C., and J.D. Germano. 1982. Characterization of organism-sediment relations
using profile imaging: an efficient method of remote ecological monitoring of the seafloor
(REMOTS system). Mar. Ecol. Prog. Ser. 8:115-128.
8-11
-------
Rhoads, D.C., and J.D. Germane. 1986. Interpreting long-term changes in benthic
community structure: a new protocol. Hydrobiologia 142:291-308.
Rhoads, D.C., P.L McCall, and J.Y. Yingst. 1978. Disturbance and production on the
estuarine sea floor. Amer. Scient. 66:577-586.
Robson, D.S., D.G. Heimbuch, and H.T. Wilson. 1991. Random point sampling from a
randomized grid. Paper presented at the annual meeting of the Canadian Statistical Society,
1990.
Rosen, J.S., J. Beaulieu, M. Hughes, H. Buffum, J. Copeland, R. Valente, J. Paul, F. Holland,
S. Schimmel, C. Strobel, K. Summers, K.J. Scott, J. Parker. 1990. Data base management
system for coastal demonstration project. EPA/600/X-90/207. EPA Office of Research and
Development, Washington, DC. (Internal report)
Rosenberg, R. 1976. Benthic faunal dynamics during succession following pollution
abatement in a Swedish estuary. Oikos 27: 414-427.
Rosenberg, R. 1990. Negative oxygen trends in Swedish coastal bottom waters. Mar. Pollut.
Bull. 21:335-339.
Ross, J.B., R. Parker and M. Strickland. 1991. A survey of shoreline litter in Halifax Harbour
1989. Mar. Pollut. Bull. 22: 245-248.
Rossignol-Strick, M.A. 1985. A marine anoxic event on the Brittany Coast, July 1982. J.
Coast. Res 1:11-20
Sanders, H.L 1956. Oceanography of Long Island Sound, 1952-1954. X: The biology of
marine bottom communities. Bull, of the Bingham Oceanog. Coll. 15:345-414.
Sanders, H.L. 1960. Benthic studies in Buzzards Bay. III. The structure of the soft bottom
community. Limnol. Oceanogr 5:138-153.
Sanders, J.E. 1989. PCB pollution in the upper Hudson River. Pages 365-400. In: National
Academy of Sciences, Contaminated Marine Sediments - Assessment and Remediation.
National Academy Press, Washington, DC.
Sanford, L.P., K.G. Sellner, and D.L. Breitburg. 1990. Covariability of dissolved oxygen with
physical processes in the summertime Chesapeake Bay. J. Mar. Res. 48:567-590.
8-12
-------
Santschi, P.M. 1984. Particle flux and trace metal residence time in natural waters. Limnol.
Oceanogr. 29:1100-1108.
Santschi, P.M., Y.H. Li, and S. Carson. 1980. The fate of trace metals in Narragansett Bay,
Rhode Island: Radiotracer experiments in microcosms. Estuar. Coast. Mar. Sci. 10:635-654.
Santschi, P.M., S. Nixon, M. Pilson, and C. Hunt. 1984. Accumulation of sediments, trace
metals (Pb, Cu) and total hydrocarbons in Narrgansett Bay, Rhode Island. Estuar. Coast.
Shelf Sci. 19:427-449.
Schneider, B. and K. Weiler. 1984. A quick grain size correction procedure for trace metal
contents of sediments. Environ. Technol. Lett. 5:245-256.
Schropp, S.J., F. G. Lewis, H.L Windom, J.D. Ryan, F.D. Calder and LC. Burney. 1990:
Interpretation of metal concentrations in estuarine sediments of Florida using aluminum as a
reference element. Estuaries 13: 227-235.
Schubel, J.R., and H.H. Carter. 1984. The.estuary as a filter for fine-grained suspended
sediment. Pages 81-104. In: V.S. Kennedy, ed. The Estuary as a Filter. Academic Press,
Orlando, FL.
Scott, J., D. Rhoads, J. Rosen, S. Pratt, and J. Gentile. 1987. The impact of open-water
disposal of Black Rock Harbor dredged material on benthic recolonization at the FVP site.
Technical Report D-87-4. U.S. Army Engineer Waterways Experiment Station, Vicksburg, MS.
Scott, J., W. Berry, D. Cobb, K. Keith, G. Tracey, and N. Rubinstein. 1990. The application
of the amphipod ten-day sediment toxicity test for dredged material evaluations. Tech. Rept.
Contr. No. 1181. Submitted to U.S. EPA Region II, New York. U.S. Environmental Protection
Agency, Environmental Research Laboratory, Narragansett, Rl. 33pp.
Sharp, J.H., J.R. Pennock, T.M. Church, T.M. Tramontano, and LA. Cifuentes. 1984. The
estuarine interaction of nutrients, organjcs, and metals: A case study in the Delaware
Estuary. Pages 241-258. In: V.S. Kennedy, ed. The Estuary as a Filter. Academic Press,
Orlando, FL.
Sinderman, C.J. 1979. Pollution-associated diseases and abnormalities of fish and shellfish:
A review. Fish. Bull. 76:717-741.
Smart, T., E. Smith, T. Vogel, C. Brown, and K. Wolman. 1987. Troubled Waters. Business
Week, October 12, 88-104.
8-13
-------
Smayda, T.J., and T.A. Villareal. 1989. The 1985 "brown-tide" and the open phytoplankton
niche in Narragansett Bay during summer. Pages 159-187. In: E.M. Cosper, V.M. Bricelj,
and EJ. Carpenter, eds. Novel Phytoplankton Blooms. Coast. Estuar. Stud. 35. Springer-
Verlag, NY.
Spies, R.B. 1989. Sediment bioassays, chemical contaminants and benthic ecology: New
insights or just muddy water? Mar. Environ. Res. 27:73-75.
Stehman, S.V., and W.S. Overton. 1987a. Estimating the variance of the Horvitz-Thompson
estimator in variable probability systematic samples. Proceedings of the Section on Survey
Research Methods of the American Statistical Association.
Stehman, S.V., and W.S. Overton. 1987b. An empirical investigation of the variance
estimation methodology prescribed for the national stream survey: simulated sampling from
stream data sets. Technical Report 118, Dept. of Statistics, Oregon State University,
Corvallis, OR.
Stehman, S.V., and W.S. Overton. 1989. Pain/vise inclusion probability formulas in random-
order, variable probability, systematic sampling. Technical Report 131. Dept. of Statistics,
Oregon State University, Corvallis, OR.
Stevens, D.L., Jr., and A.R. Olsen. 1992. Statistical issues in environmental monitoring and
assessment. Proceedings of the American Statistical Association Section on Statistics and
the Environment-1991. American Statistical Assocation, Alexandria, VA.
Stevens, D.L, Jr., A.R. Olsen, D. White. 1991. Environmental Monitoring and Assessment
Program: integrated sampling design. Draft report. Environmental Research Laboratory, U.S.
Environmental Protection Agency, Corvallis, OR.
Strobel, C.J. 1990. Environmental Monitoring and Assessment Program Near Coastal
Component 1990 Demonstration Project Field Operations Manual. U.S. Environmental
Protection Agency, Office of Research and Development, Washington, DC.
Swartz, R.C. 1987. Toxicological methods for determining the effects of contaminated
sediment on marine organisms. Pages 183-198. In: K.L. Dickson, A.W. Maki, and W.A.
Brungs, eds. Fate and Effects of Sediment Bound Chemicals in Aquatic Systems. Pergamon
Press, New York.
Swartz, R.C. 1989. Marine sediment toxicity tests. Pages 115-129. In: National Academy
of Sciences. Contaminated sediments - assessment and remediation. National Academy
Press, Washington, DC.
8-14
-------
Swartz, R.C., W.A. DeBen, J.K. Jones, J.O. Lamberson, and F.A. Cole. 1985.
Phoxocephalid amphipod bioassay for marine sediment toxicity. Pages 284-307. In: R. D.
Cardwell, R. Purdy, and R.C. Bahner, eds. Aquatic Toxicology and Hazard Assessment:
Seventh Symposium. American Society for Testing and Materials, Philadelphia, PA.
Terrell, T.T. 1979. Physical regionalization of coastal ecosystems of the United States and its
territories. FWS/OBS-79/80. U.S. Fish, and Wildlife Service, Office of Biological Services.
Theede, H. 1973. Comparative studies on the resistance of marine bottom invertebrates to
oxygen deficiency and hydrogen sulphide. Mar. Biol. 2:325-337.
Thursby, G.B., and R.L. Steele. 1987. Guidance manual for conducting sexual reproduction
tests with the marine macroalga Champia parvula for use in testing complex effluents.
Chapter 3. In: S.C. Schimmel, ed. Users Guide to the Conduct and Interpretation of
Complex Effluent Toxicity Tests at Estuarine/Marine Sites. U.S. Environmental Protection
Agency, Environmental Research Laboratory, Narragansett, Rhode Island.
Toufexis, A. 1988. Our filthy seas: The world's oceans face a growing threat from manmade
pollution. Time, August 1.
Trefry, J.H., S. Metz and R.P. Trocine. 1985. The decline in lead transport by the Mississippi
River. Science 230: 439-441.
Turekian, K.K.
1144.
1977. The fate of metals in the oceans. Geochim. Cosmbchim. Acta 41:1139-
Turner, R.E., and R.L Allen. 1972. Bottom water oxygen concentration in the Mississippi
River Delta bight. Contrib. in Mar. Sci.25:161-172.
Urquhart, N.S., W.S. Overton, and D.S. Birkes. 1991. Comparing sampling designs for
monitoring ecological status and trends: Impact of temporal patterns. Technical Report #49.
Department of Statistics, Oregon State University, Corvallis, OR.
USEPA. 1984. Chesapeake Bay: A framework for action. Prepared for the U.S. Congress
by the Environmental Protection Agency, Region 3, Philadelphia, PA.
USEPA. 1989. Briefing Report to the EPA Science Advisory Board on the equilibrium
partioning approach to generating sediment quality criteria. EPA/440/5-89-002. USEPA,
Criteria and Standards Division, Washington, DC.
8-15
-------
USEPA. 1990. Evaluation of the equilibrium partitioning (EqP) approach for assessing
sediment quality. Report of the Sediment Criteria Subcommittee of the Ecological Processes
and Effects Committee of the EPA Science Advisory Board. EPA-SAB-EPEC-90-006. U.S.
Environmental Protection Agency, Washington, DC. -
USEPA. 1991. The near coastal laboratory procedures manual. Environmental Monitoring
and Assessment Program. US Environmental Protection Agency, Environmental Monitoring
Systems Laboratory, Cincinnati, OH.
USEPA. 1992. Safeguarding the Future: Credible science, credible decisions. Report of the
Expert Panel on the Role of Science at EPA. EPA/600/9-91/050. U.S. Environmental
Protection Agency. 38 pp. plus Appendices.
USEPA/ACE. 1991. Evaluation of dredged material proposed for ocean disposal (Testing
manual). EPA/503/8-91/001. U.S. Environmental Protection Agency, Office of Marine and
Estuarine Protection and Department of the Army, United States Army Corps of Engineers.
Valente, P.M., CJ. Strobel, J.E. Pollard, K.M. Peres, T.C. Chiang, and J. Rosen. 1990.
Quality assurance project plan for EMAP Near Coastal: 1990 Demonstration Project. U.S.
EPA, Environmental Research Laboratory, Narrangansett, Rl.
Valente, R.M., D.C. Rhoads, J.D. Germane, and V.J. Cabelli. 1992. Mapping of benthic
enrichment patterns in Narrangansett Bay, Rhode Island. Estuaries 15:1-17.
Voutsinou-Talliodouri, F. and J. Satsmadjis. 1982. Trace metals in the Pagassitikos Gulf.
Greece. Estuar. Coast. Shelf Sci. 15: 221-228.
Warwick, R.M. 1980. Population dynamics and secondary production in benthos. Pages 1-
24. In: K.R. Tenore and B.C. Coull, eds. Marine Benthic Dynamics, Belle W. Baruch Library in
Science Number 11. University of South Carolina Press, Columbia, SC.
Weaver, G. 1984. PCB contamination in and around New Bedford, Mass. Environ. Sci.
Technol. 18:22A-27A.
Welsh, B.L and F. C. Eller. 1991. Mechanisms controlling summertime oxygen depletion in
Western Long island Sound. Estuaries 14: 265-278,
White, J.W., W.S. Woolcott, and W.L. Kirk. 1977. A study of the fish community in the vicinity
of a thermal discharge in the James River, Virginia. Chesapeake Sci. 18:161-171.
8-16
-------
Whitledge, T.E. 1985. Nationwide review of oxygen depletion and eutrophication in estuarine
and coastal waters: Executive Summary. Brookhaven National Laboratory, Upton, NY.
Windom, H.L, S.J. Schropp, F.D. Calder, J.D. Ryan. R. G. Smith, LC. Bumey, F.G. Lewis
and C.H. Ralinson. 1989. Natural trace metal concentrations in esturine and coastal marine
sediments of the southeastern United States. Environ. Sci. Technol. 23: 314-320.
Wolthaus, B.G. 1984. Seasonal changes in frequency of diseases in dab Limanda limanda
from the southern North Sea. Helgolander Meeresunters. 37:375-387.
8-17
*U.S. G.P.O.:1993-752-614
-------
-------
-------
m
CO
CD
O c
00
m c:
§5 — 3 iS 5 =i,
IS- 2. 3, ^- CD
CD
w ~ -»,
33 >
CD (Q
cn CD
CD 3
Q> O
8s£
=! a- s
S-m
------- |