United States
Environmental Protection
Agency
Office of
Research and Development
Washington, DC 20460
EPA/620/R-93/003
February 1993
Surface Waters
1991 Pilot Report
Environmental Monitoring and
Assessment Program
-------
EMAP-Surface Waters 1991 Pilot Report
EDITORS:
David P. Larsen
U.S. EPA Environmental Research Laboratory
200 SW 35th Street
Corvallis, OR 97333
Susan J. Christie
ManTech Environmental Technology, Inc.
U.S. EPA Environmental Research Laboratory
200 SW 35th Street
Corvallis, OR 97333
AUTHORS AND CONTRIBUTORS:
Chapter 1: DP. Larsen1 and S.G. Paulsen2
Chapter 2: R.M. Hughes3, C. Burch Johnson3, S.S. Dixit4, AT. Herlihy5, P.R. Kaufmann5, W.L.
Kinney6, D.P. Larsen1, P.A. Lewis7, D.M. McMullen8, A.K. Moors9, R.J. O'Connor9,
S.G. Paulsen2, R.S. Stemberger10, S.A. Thiele3, T.R. Whittier3, and D.L. Kugler3
Chapter 3: D.P. Larsen1, K.W. Thornton11, N.S. Urquhart5, S.G. Paulsen2
Chapter 4: D.P. Larsen1 and N.S. Urquhart5
Chapter 5: S.G. Paulsen2, N.S. Urquhart5, D.P. Larsen1
Chapter 6: J.R. Baker12 and N. Tallent-Halsell12
10
11
12
U.S. EPA Environmental Research Laboratory, 200 SW 35th Street, Corvallis, OR 97333.
University of Las Vegas, c/o U.S. EPA Environmental Research Laboratory, 200 SW 35th Street, Corvallis, OR 97333.
ManTech Environmental Technology, Inc., c/o U.S. EPA Environmental Research Laboratory, 200 SW 35th Street,
Corvallis, OR 97333.
Department of Biology, Queen's University Kingston, Ontario K7L 3N6, Canada.
Oregon State University, c/o U.S. EPA Environmental Research Laboratory, 200 SW 35th Street, Corvallis, OR 97333.
U.S. EPA Environmental Monitoring Systems Laboratory, Las Vegas, NV 89193-3478.
U.S. EPA Environmental Monitoring Systems Laboratory, 3411 Church Street, Newtown, OH 45244.
Technology Applications, Inc., c/o U S. EPA Environmental Monitoring Systems Laboratory, 3411 Church Street,
Newton, OH 45244.
Department of Wildlife, University of Maine, 238 Nutting Hall, Orono, ME 04469-0125.
Biology Department, Room 101, Oilman Hall, Dartmouth College, Hanover, NH 03755.
FTN Associates, 3 Innwood Circle, Suite 220, Little Rock AR 72211.
Lockheed Engineering and Sciences Co., 980 Kelly Johnson Drive, Las Vegas, NV 89119.
-------
EPA/620/R-93/003
February 1993
EMAP-Surface Waters 1991 Pilot Report
David P. Larsen and Susan J. Christie, Editors
February 1993
U.S. Environmental Protection Agency
Region 5, Library (PL-1?J)
77 IVesi Jackson Boi.;3v;.rd, 12th Floor
Chicago, IL 60-504-3590
Printed on Recycled Paper
-------
ACKNOWLEDGEMENTS
In any undertaking with a scope as broad as that of EMAP-Surface Waters, numerous people
contribute in important ways not reflected by authorship on documents such as this. We hesitate
attempting to list all those who contributed, because we are sure to miss some. Instead, we will
identify the organizations whose staff generated important material included in this report:
• EMAP-Surface Waters and associated laboratory staff in Corvallis, Las Vegas, and
Cincinnati, including EPA and on-site contractor personnel (ManTech Environmental
Technology, Inc., Lockheed Engineering and Sciences Co., and Technology Applica-
tions, Inc.), and guest workers from Oregon State University and University of Nevada
at Las Vegas.
• Environmental Services Division of EPA Regions I and II.
• Personnel on cooperative agreements with Queens University, Dartmouth College, and
University of Maine.
• Members of the lake sampling crews of miscellaneous origin
• Members of the peer review panel.
We especially appreciate the members of our sampling crews for their diligent efforts in obtaining
data of outstanding quality. We also thank members of our peer review panel for their initial
evaluation of the results contained in this report. The following people provided official technical
reviews of this report: Steve Bartell, Charles Goldman, Steve Hedtke and Stephen Lozano,
Thomas LaPoint, Eugene Stoermer, Deborah Coffey, and Don King. Many others provided
informal but important review comments.
Financial support from U.S. EPA to the following institutions led to much of the work reported
here:
Oregon State University (CR818606 and CR816721)
Queens University (CR818707)
Dartmouth College (CR819689-01-0)
University of Maine (CR819659 and CR818179)
ManTech Environmental Technology, Inc. (68-C8-0006)
Lockheed Engineering and Sciences Co. (68-CO-0049)
Aquatech (1B0517NTSA)
The research described in this document has been funded by the U.S. Environmental Protection
Agency. This document has been subjected to the Agency's peer and administrative review and
approved for publication as an EPA document. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
-------
TABLE OF CONTENTS
Section Page
Acknowledgements ii
List of Illustrations vi
List of Tables ix
Abbreviations and Acronyms xi
Symbols xii
1. INTRODUCTION 1
1.1 Overview of the Role of Pilot Studies 1
1.2 FY 1991 Pilot Studies 2
1.3 Scope 4
2. DEVELOPMENT OF LAKE CONDITION INDICATORS FOR EMAP—1991 PILOT 7
2.1 Introduction 7
2.1.1 Societal Values and Their Response Indicators 9
2.1.2 Habitat, Exposure, and Stressor Indicators 12
2.1.3 Indexing 12
2.1.4 Indicator Evaluation 13
2.1.5 Determining Nominal/Subnominal Condition 13
2.1.6 Probability Lakes and Indicator Lakes 14
2.1.7 Chapter Layout 14
2.2 Evaluation of Indexing 15
2.2.1 The General Indexing Concept 15
2.2.2 Index Period, Index Location, and Sampling Gear 18
2.2.3 Indexing Results 22
2.3 Selection of Indicator Lakes and Their Physical and Chemical Characteristics 30
2.3.1 Rationale for Selecting the Indicator Lakes 30
2.3.2 Description of Indicator Lakes 34
2.4 Indicator Development and Evaluation 38
2.4.1 Selection of Candidate Assemblages 38
2.4.2 Indicator Evaluation Process 43
2.4.3 Comparison among Assemblages 47
2.4.4 Selection of Candidate Metrics 49
2.4.5 Results and Discussion of Indicator Evaluation on Indicator Lakes 54
2.5 Distinguishing Acceptable (Nominal) from Unacceptable (Subnominal)
Conditions 78
2.5.1 Describing the Reference Condition 79
2.5.2 Pilot Results 82
2.6 Summary and Conclusions 84
2.6.1 Indexing 84
2.6.2 Indicator Development 84
2.6.3 Acceptable/Unacceptable Condition 84
2.6.4 Concluding Remarks 85
Appendix 2A - Analytes to be Measured in Fish Tissue 87
Appendix 2B - Diatom, Zooplankton, Benthos, Fish, Bird, and Sediment Toxicity Methods ... 88
Appendix 2C - Physical Habitat Structure Field Methods 89
Appendix 2D - Water Quality Methods 90
-------
3. OVERVIEW OF SURVEY DESIGN AND LAKE SELECTION 91
3.1 Purpose of Survey Design 91
3.1.1 Probability Sampling Designs 92
3.1.2 Need for a Common Design 93
3.2 EMAP Sampling Grid 94
3.2.1 Base Grid 94
3.2.2 Hierarchical or Tier Structure 97
3.3 Frame and Tier 1 Sample Selection 98
3.3.1 Lake Sampling Frame and Tier 1 Sample 98
3.3.2 Frame Characteristics 100
3.3.3 Classification Strategies 101
3.4 Lake Selection 103
3.5 Temporal Sampling Schedule 103
3.6 Annual Revisit Sites 108
3.7 Summary and Lessons Learned 108
Appendix 3A - Details of the Lake Selection Process 111
4. A FRAMEWORK FOR EVALUATING THE SENSITIVITY OF THE EMAP DESIGN 119
4.1 Introduction 119
4.2 Important Components of Variance 120
4.2.1 Population Variance 120
4.2.2 Extraneous Variance 120
4.3 A Linear Statistical Model for Estimating Variance Components 124
4.4 Effects of Variance on Estimates of Status 129
4.5 Effects of Variance on Trend Detection 136
4.6 A Note on Available Databases 146
4.7 Implications for Pilot Surveys and Indicator Selection 148
4.8 Summary 150
Appendix 4A - Methods for Total P, Chlorophyll-a, and Secchi Depth
for EMAP-SW 1991 Data 151
Appendix 4B - A Flexible Linear Model for Status Estimation and Trend
Detection 152
5. PROTOTYPE ANNUAL REPORT 157
5.1 Introduction 157
5.2 Overview of EMAP 157
5.2.1 EMAP Goals and Objectives 158
5.2.2 EMAP Indicator Strategy 158
5.2.3 Reporting in EMAP 160
5.3 EMAP Surface Waters 160
5.3.1 Legislative Mandate 160
5.3.2 Issues and Problems 161
5.3.3 Surface Water Indicator Strategy 162
5.4 Extent of Surface Waters 164
5.4.1 Lakes 164
5.4.2 National Estimates 164
5.4.3 Regional Estimates 165
5.5 Condition of Resources 167
5.5.1 Indicator Data for 1991 167
IV
-------
5.5.2 Trophic Condition of Lakes 170
5.6 Conclusions 176
6. LOGISTICAL OPERATIONS SUMMARY 177
6.1 Introduction 177
6.2 Planning Activities 179
6.2.1 Lake Verification 179
6.2.2 Site Access 180
6.2.3 Protocol Development, Training, and Mobilization 181
6.2.4 Field Crew Personnel 182
6.3 Field Operations Overview 183
6.3.1 Daily Sampling Operations: Probability and Time Lakes 183
6.3.2 Support Operations 187
6.4 Results and Recommendations 188
7. LITERATURE CITED 191
-------
LIST OF ILLUSTRATIONS
Figure Page
2-1 Capture efficiency of fishing gear by (a) % species caught and (b) % individuals
caught, both by all gear in combination, for 19 lakes 24
2-2 Species/effort curve for fish caught by trap nets in (a) small and (b) large lakes ... 26
2-3 Locations of the 19 indicator lakes in New England 32
2-4 Thermocline temperatures and dissolved oxygen concentrations of 19 indicator
lakes 37
2-5 Sediment toxicity of 8 samples from 4 indicator lakes: (a) percent survival,
and (b) percent growth 39
2-6 Lake conceptual model showing macrohabitats and trophic levels 40
2-7 The indicator development process showing the five major phases 44
2-8 An example of a conceptual linkage diagram for fish and biological integrity 51
2-9 Diatom-inferred total phosphorus (TP) change in indicator lakes 56
2-10 Diatom-inferred chloride change in indicator lakes 57
2-11 Diatom-inferred pH change in indicator lakes 58
2-12 Detrended correspondence analysis of diatom assemblages 59
2-13 Zooplankton species richness versus lake area 63
2-14 Total fish species richness versus lake area 67
2-15 Species richness of native fishes versus lake area 68
2-16 Bird species richness versus lake area 72
2-17 Relationship between quality of physical habitat structure and (a) % intolerant
bird species, and (b) % native fish species 75
3-1 The base grid overlaid on North America 95
3-2 EMAP grid structure showing the relationship between the 635-km2 hexagons
and the embedded 40-km2 hexagon 96
3-3 Map illustrating the spatial distribution of the 1991 and 1992 Tier 2
samples of lakes selected from the Digital Line Graph frame 105
3-4 Distribution of selected 1992 Tier 2 lakes for the conterminous
United States 106
VI
-------
3-5 Temporal sampling schedule for all EMAP resources illustrating the uniform
spatial coverage for every year 107
4-1 A schematic illustrating the sources of variation described in Section 4.2 123
4-2 Cumulative distribution functions illustrating the effects of increased
extraneous variance on estimates of population status 130
4-3 Increasing extraneous variance spreads and flattens the normal distribution 131
4-4 Impact of variance components on the shape of cumulative distribution
functions 133
4-5 Illustration of the approximate distortion to population CDFs resulting from
observed extraneous variance estimated during the 1991 EMAP pilot survey .... 135
4-6 Trend detection sensitivity as a function of numbers of lakes visited and
numbers of revisits to individual lakes within an index period 139
4-7 An illustration of the relationship between power and magnitude of trend
detectable and years of monitoring 143
4-8 Power to detect a 1 %/year trend in Secchi Disk transparency
a2year varies; o2res is held constant 144
4-9 Power to detect a 1 %/year trend in Secchi Disk transparency
cr2 is held constant; <72res varies 145
5-1 A map illustrating three ecological regions on which statistical summaries
can be based for routine reporting 168
5-2 Cumulative distribution functions (CDFs) will be the primary format for
reporting the condition of ecological resources 169
5-3 Estimated chlorophyll-a and total phosphorus cumulative distribution
functions (CDF) for all lakes between 1 and 2,000 ha in the Northeast 171
5-4 Estimated chlorophyll-a and total phosphorus cumulative distribution
functions (CDF) for all lakes between 1 and 2,000 ha in the Adirondacks 172
5-5 Estimated chlorophyll-a and total phosphorus cumulative distribution
functions (CDF) for all lakes between 1 and 2,000 ha in the New England
Uplands 173
5-6 Estimated chlorophyll-a and total phosphorus cumulative distribution
functions (CDF) for all lakes between 1 and 2,000 ha in the Coastal/
Lowlands/Plateaus 174
5-7 Histogram of estimated proportion and number of eutrophic, mesotrophic,
and oligotrophic lakes between 1 and 2,000 ha in the Northeast 175
VII
-------
6-1 Daily water sampling activities at 1991 EMAP Northeast Lakes Pilot sites 184
6-2 Daily sediment sampling activities at 1991 EMAP Northeast Lakes Pilot sites .... 185
VIII
-------
LIST OF TABLES
Table Page
2-1 Characteristics of an Ideal Index Sample 15
2-2 Preferred Index Periods for Various Lake Attributes and Assemblages 19
2-3 Index Locations for Various Lake Attributes and Assemblages 19
2-4 Zooplankton Index and Lake Variance Estimated for Species Richness and Body
Size Ratio, Two Candidate Metrics for the Zooplankton Assemblage 22
2-5 Maximum Fish Catches per Net through Use of Stratified Random and Subjec-
tively Placed Trap Nets 27
2-6 Index and Lake Variance Estimated for 12 Candidate Richness and Tolerance
Metrics of the Bird Assemblage 28
2-7 Catchment Conditions of 19 Indicator Lakes 33
2-8 Selected Water Quality Characteristics of 19 Indicator Lakes 35
2-9 Habitat Structure of 19 Indicator Lakes 36
2-10 Response Indicator Selection Criteria 43
2-11 Typical Effects of Environmental Degradation on Biological Assemblages,
and Candidate Fish Metrics 50
2-12 Sedimentary Diatom Taxa Richness and Disturbance Index of 19 Indicator
Lakes 61
2-13 Ratio of Large to Small Zooplankton and Number of Trophic Links at 19
Indicator Lakes 64
2-14 Benthic Macroinvertebrate Metric Scores for 19 Indicator Lakes 66
2-15 Fish Metric Scores for 19 Indicator Lakes 70
2-16 Principal Components Analysis Scores and Number of Tolerant and Intolerant
Birds at 19 Indicator Lakes 73
2-17 Assemblage, Metric, and Lake Scores for 19 Indicator Evaluation Lakes 77
3-1 Numbers of Lakes in the Population and Tier 1 Sample from the Base Grid 100
3-2 Tier 2 Yearly Sample Sizes for the Northeast, with Corresponding Inclusion
Probabilities and Expansion Factors 104
4-1 Individual Index Observations Expressed in Terms of a Linear Statistical
Model 125
IX
-------
4-2 Analysis of Variance Table Summarizing Important Components of Variance
for EMAP-SW Lake Monitoring 126
4-3 Summary of Estimates of Components of Variance for Secchi Disk Transparency
Chlorophyll-a, and Total Phosphorus Derived from the Vermont Lake Monitoring
Database 128
4-4 Estimates of Variance Components Derived from the EMAP 1991 Pilot Survey ... 128
4-5 Least Significant Trend, as Defined in the Text, as Allocating Sampling
Effort to Revisits Versus Additional Lakes 140
4-6 Sensitivity of EMAP Design for Detecting Trends in Secchi Disk Transparency
Chlorophyll-a, and Total Phosphorus, Based on Variance Components Derived
from the Vermont Lake Monitoring Database 141
5-1 Regional and National Estimates of Lake Number and Area 165
5-2 Estimates of Lake Number and Lake Area for Lakes 1-2,000 ha for the
Northeast (EPA Region I and II) and Selected Subregions 166
6-1 Components of the Pilot Activities Planned for EMAP-Surface Waters in 1991 .... 178
-------
ABBREVIATIONS AND ACRONYMS
ANC — acid neutralizing capacity
ANSP — Academy of Natural Sciences at Philadelphia
APHA — American Public Health Association
AVHRR — advanced very high resolution radiometry (from satellite)
Bl — biotic index
CCA — canonical correspondence analysis
CDF — cumulative distribution function
Chl-a — chlorophyll-a
CV — coefficient of variation
DCA — detrended correspondence analysis
DIG — dissolved inorganic carbon
DIG — digital line graphs
DO — dissolved oxygen
DOC — dissolved organic carbon
EMAP — Environmental Monitoring and Assessment Program
EMAP-SW — EMAP-Surface Waters
EPA — Environmental Protection Agency
FIA — Forest Inventory Analysis, performed by U.S. Forest Service
FWS - Fish and Wildlife Service
GLM — General Linear Model (SAS procedure)
GPS — Global Positioning System
ha — hectare
km — kilometer
m — meter
ml — milliliter
NALMS — North American Lake Management Society
NASS — National Agricultural Statistics Survey
NES — National Eutrophication Survey
OEPA — Ohio Environmental Protection Agency
PCA — principal components analysis
QA/QC — quality assurance/quality control
REDOX — reduction-oxidation
RF3 — U.S. EPA River Reach File, Version 3
SAS — Statistical Analysis System
XI
-------
SD
SE
TIME
TP
TSI
USDA
USGS
Secchi Disk (transparency)
standard error
Temporally Integrated Monitoring of Ecosystems
total phosphorus
Trophic State Index
U.S. Department of Agriculture
U.S. Geological Survey
SYMBOLS
mg/L
°C
,weq/L
index
2
lake
2
lake*year
2
meas
2
res
2
year
milligrams per liter
degrees Celsius
microequivalents per liter
micro-grams per liter
index variance
population variance
lake-year interaction effects
measurement variance
residual variance
year variance
XII
-------
CHAPTER 1
INTRODUCTION
1.1 OVERVIEW OF THE ROLE OF PILOT STUDIES
In 1989, the U.S. Environmental Protection Agency (EPA) initiated the Environmental Monitoring
and Assessment Program (EMAP) to provide improved information on the current status of, and
long-term trends in, the condition of the nation's major ecological resources, including inland
surface waters. Before full-scale implementation of EMAP-Surface Waters (EMAP-SW), we antici-
pate conducting a series of lake monitoring pilot studies, followed by demonstration studies, to
show whether or not EMAP can produce the kinds of results intended.
It is often difficult to say exactly what constitutes a pilot study, except to describe it as the effort
expended to answer specific questions. Pilot studies can be specific field studies, desktop anal-
yses, or both. They can be conducted on a regional scale, using the EMAP probability-based
design as the basis for lake selection, or they can occur at specially selected sites, depending on
the nature of the questions to be addressed or the lake populations of interest. Pilot studies
sometimes address questions that can be answered without conducting field studies; for example,
they can evaluate the results of other studies or include workshops that seek to answer particular
questions.
The demonstration studies that follow the pilot studies are designed to produce regional-scale
estimates similar to the estimates expected from EMAP. In some cases, demonstration studies
can answer relevant questions just as well as pilot studies; in fact, a regional-scale implementa-
tion study may be the only way to answer some questions.
The FY 1991 pilot studies included elements of all three types of preliminary study (regional
surveys, special field studies, and desktop analyses). The fundamental role of a pilot study is to
focus on critical questions, the answers to which are necessary for effective implementation of
EMAP monitoring. We began by asking, "What critical pieces prevent us from implementing
regional or national monitoring?" or "Why can't we presently initiate an EMAP-type monitoring
program?" or "What will be the consequences of proceeding with monitoring without an answer to
this set of questions?" The EMAP-Surface Waters Northeast Lakes Pilot Implementation Plan
posed a set of questions to be addressed in a series of pilot studies (Pollard and Peres, 1991).
The questions, recast here to set the framework for this pilot report, focus on lake monitoring but
they apply equally well to stream monitoring.
1
-------
A. One basic set of questions focuses on what we will monitor and includes the following:
1. What indicators will we use as part of the basic monitoring program? What is a suffici-
ent number of indicators on which to base full-scale monitoring? In this report, we limit
the discussion primarily to selection of those indicators that will be used to address the
primary objectives of EMAP, the response indicators (see Chapter 2).
2. Where, when, and how will we measure the indicators? This question has two parts.
The first is the design-based, probability selection of the lakes and streams to be
sampled: How will we select the lakes for field visitation? (Survey Design Evaluation;
see Chapter 3.) The second is: Once the lakes have been selected, how and when will
we sample them? (Index Characterization; see Chapter 2.)
3. How will we use the indicators to make statements about condition, associations, and
probable cause of impaired and unimpaired conditions? (See Chapter 5.)
B. Indicator variability at various spatial and temporal scales affects our ability to characterize
status (condition) and detect trends in ecological condition. These topics are covered in
Chapter 4 of the report:
1. What are the important components of variance needing characterization?
2. How does the magnitude of these components affect our ability to describe status and
to detect trends?
3. What choices do we have to reduce variance (e.g., through methodological improve-
ments), to minimize its effects (e.g., through efficient design modifications), or to
remove its effects (e.g., through mathematical manipulation)?
C. Implementing a regional-scale monitoring program such as EMAP requires extensive
logistical planning and testing. These issues are addressed in Chapter 6. Basic questions
include:
1. Can we assemble and deploy sampling teams to obtain lake and stream samples within
the selected temporal index window?
2. Can we train different teams well enough to keep team-to-team differences minor and
not compromise program objectives?
3. Can we coordinate the work of multiple field crews well enough for smooth, consistent
sample collection and field recording? Can we ship and track samples to appropriate
analytical facilities or data processors in a timely manner and without loss of samples?
1.2 FY 1991 PILOT STUDIES
During the summer of 1991, EMAP-SW conducted its first series of field pilot studies, with lakes
as the resource of interest and the northeastern United States (EPA Regions I and II) as the area
of interest. A basic criterion for the pilot studies was that we would use the EMAP grid design
and probability methods to select a set of lakes for sampling to answer the questions outlined in
-------
Section 1.1, and that we would select other lakes only if we were unable to answer the questions
by sampling the probability lakes. For some indicators, a regional-scale pilot study would serve
to answer some questions. For other indicators, however, this was not so. In particular, it was
evident that we were not prepared to obtain index samples of the fish assemblage, the riparian
birds, the benthos assemblage, and the physical habitat within reasonable budget constraints.
Therefore, we selected a special set of lakes for use in developing methods to index lakes for
these attributes.
We also faced the need to begin data collection to answer some very specific questions raised by
reauthorization of the Clean Air Act, which mandates significant reduction of sulfate emissions (10
million metric tons) and nitrate emissions (2 million metric tons) over the next 10 years (Public
Law 101-549, Title IV, Section 401 b). The Clean Air Act also mandates an assessment of the
effect of this reduction on aquatic ecosystems: could we detect lake and stream responses to
this magnitude of reduction in sulfate emissions after 10 years? To some extent, the National
Surface Water Survey (Linthurst et al., 1986; Landers et al., 1987; Kaufmann et al., 1988),
including the Long-Term Monitoring Project (Newell and Hjort, 1991), were pilot studies that eval-
uated the ability of design-based probability surveys to collect and interpret limnological infor-
mation addressing critical policy questions. We felt confident that a regional-scale survey could
be implemented to address the mandates of the Clean Air Act. In addition, through regional-scale
pilot studies, we could begin data collection for some indicators (e.g., indicators of the trophic
condition of lakes and lake chemical characteristics) to answer questions about variance and
logistics not answerable any other way.
Therefore, the field-scale pilot studies had two parts. One part was designed to develop indexing
protocols for indicators with which we were not confident we could index lakes. We selected a
set of lakes for this part of the pilot activity and called them indicator lakes. Chapter 2 focuses on
questions about indicators. The second part of the pilot activity focused on logistics and variance
questions about those indicators for which we were confident we could obtain index samples
(trophic condition and chemical characteristics). In this part, we also began data collection on
indicators in order to address the Clean Air Act issues. We called the set of lakes selected for
this purpose the probability lakes, because we used the EMAP grid design and probability
methods to select lakes for field visitation.
In addition, a significant amount of thought has gone into refining the conceptual framework on
which EMAP is founded. We further explored the concepts of ecological values, response indica-
-------
tors, index samples, description of status, and nominal-subnominaJ. A major part of the 1991 pilot
addressed questions related to indicator development, particularly for the societal value of bio-
logical integrity. Furthermore, we have been identifying and evaluating the influence of important
components of variation on our ability to describe status and detect trends in condition. These
refinements will guide us in selecting available databases for a preview of EMAP capabilities.
1.3 SCOPE
This report is organized as follows:
• Chapter 2 focuses primarily on the process of selecting response indicators for EMAP
monitoring, drawing from numerous discussions on our conceptual framework for indicator
selection and from results of the 1991 pilot surveys.
• Chapter 3 describes how we selected the probability lakes by using the general EMAP
design guidelines and what we learned from that exercise. It covers questions related to
the Tier 1 and Tier 2 selection process, nontarget and noninterest lakes, and the setting of
inclusion probabilities (or expansion factors).
• Chapter 4 discusses the set of questions that address how well we think the proposed
design will describe status and detect trends in condition. We describe and illustrate
important components of variance with one set of indicators describing trophic condition.
We use available data, along with data derived from the EMAP pilot, to show the frame-
work for estimating the influence of variance. The results of this part of the pilot offer
guidance on the types of databases most useful for describing variance for the other indi-
cators.
• Chapter 5 briefly summarizes how annual statistical reports might be structured using data
collected on some indicators from the probability lakes.
• Chapter 6 covers the logistics component of the pilot study.
This report does not cover the following:
• Regional-scale monitoring conducted to answer questions about the response of lakes to
sulfate emissions reductions mandated by the Clean Air Act [the Temporally Integrated
Monitoring of Ecosystems (TIME) component of EMAP].
• Habitat, exposure, and stressor indicators, except as covered briefly in Chapter 2. We are
developing our diagnostic strategy.
• The process for identifying particular lake subpopulations, and those subpopulations at a
national scale, on which EMAP-SW is likely to target reporting.
• Information management.
A sound quality assurance/quality control (QA/QC) program is fundamental to any monitoring
program with the potential scope and duration of EMAP. A QA/QC plan has been developed that
-------
describes how the multitude of EPA and EMAP requirements will be addressed during the pilot,
demonstration, and implementation phases of EMAP-SW (e.g., Peck, 1992). In addition, periodic
reports (e.g., University of Maine, 1991) will focus on the QA/QC results of particular aspects of
EMAP-SW activities. This pilot report covers some aspects of QA/QC, including how to obtain a
representative, repeatable sample of various biological assemblages in lakes (Chapter 2); how to
obtain a representative sample of lakes on which to make the measurements (Chapter 3); how to
evaluate variance components and their influence on our ability to estimate status and detect
trends (Chapter 4). These chapters lay the foundation for how we currently address, and how we
expect to address, several fundamental aspects of assuring that data collected during EMAP-SW
monitoring are of known quality. The future reports will address these topics in more detail.
Chapter 2 discusses an issue of interest, and occasionally of concern, to many people. The issue
is how to set a criterion that can be used to evaluate whether the condition measured and
expressed with indicators is acceptable or unacceptable for a particular lake or stream type within
a given geographic location. Hunsaker and Carpenter (1990) and Messer et al. (1991) outline the
issue. Messer et al. state: "Operational criteria must be developed for each response indicator to
identify the transition from acceptable or desirable (nominal) to unacceptable or undesirable
(subnominal) condition...Criteria could be based on attainable conditions under 'best manage-
ment practices' as observed at regional reference sites..., or on theoretical grounds or manage-
ment targets." Clearly, this is not strictly a technical issue. We emphasize that our objective in
approaching this issue is not to make this decision about thresholds independent of others, but
rather to contribute to the process of deciding how to resolve these issues from a technical
perspective. We are not advocating any particular management decision or action or societal
decision. We do believe, however, that it is important to develop as sound a scientific basis as
possible for such decisions. We developed our discussions in Chapter 2 from this perspective.
-------
CHAPTER 2
DEVELOPMENT OF LAKE CONDITION INDICATORS FOR EMAP—1991 PILOT
2.1 INTRODUCTION
The primary objective of EMAP is to estimate, on a regional scale and with known confidence, the
status of, and changes and trends in, indicators of the condition of our nation's ecological
resources. Given this objective, it is critical that we clearly describe the EMAP-Surface Waters
(EMAP-SW) strategy for developing, evaluating, and selecting indicators of the condition of our
lakes and streams. Chapter 2 focuses on issues pertaining to indicator development and the
results of the 1991 pilot relative to these issues.
The selected indicators will be the foundation for the information presented to decision makers,
the scientific community, and the public about the condition of our nation's lakes and streams.
An effective indicator will inform decision makers and environmental managers, will be relevant to
one of a host of policies that impact these resources, and will provide information upon which
decision makers and managers might be willing to act (Hilborn, 1992; Ward, 1989). The indica-
tors ultimately selected for EMAP-SW must be ecologically relevant and scientifically credible and
must relate clearly to important biologically oriented characteristics of lakes and streams that are
valued by the public.
What then is an indicator? There has been some confusion about this term. Many ecologists
equate it with "indicator species," but we do not mean indicator species. We define an indicator
as an ecological measurement, metric, or index that quantifies physical, chemical, or biologi-
cal condition, habitat, or stress. Measurements conducted on a biological assemblage (e.g.,
fish, diatoms) must be converted into numerical metric or index scores that can be presented as a
distribution of characteristics found within the population of lakes and streams of interest. These
quantifiable forms, not the assemblage itself, constitute the indicators.
EMAP has identified four kinds of indicators: response, exposure, habitat, and stressor (Hunsaker
et al., 1990; Messer, 1990; Paulsen et al., 1991). These categories are meant to be functional
ones that describe the intended use of the indicator, not mutually exclusive pigeon holes for each
measurement.
-------
• Response indicators demonstrate the biological condition of the resource, at the organ-
ism, population, assemblage, or community level of organization. For surface waters,
the preferable organization levels are the assemblage and community, and these
should quantify the integrated response of the system to single, multiple, and cumula-
tive stressors and relate to the ecological values of the waterbody held by society. The
EPA now requires states to monitor response indicators in their biological criteria
programs (U.S. EPA, 1990).
• Exposure indicators show the occurrence or magnitude of a response indicator's con-
tact with a physical, chemical, or biological stressor. In surface waters, we focus on
stressors that are usually of human origin. These indicators are most useful for
diagnosing the probable causes of unacceptable condition and historically were the
focus of monitoring by state and federal water monitoring agencies.
• Habitat indicators are defined as physical attributes necessary to support an organism,
population, or community in the absence of pollutants (Hunsaker et al., 1990; Messer,
1990). However, this definition initially omitted chemical and biological habitat
components. For surface waters, we consider habitat indicators as characterizing the
typically natural chemical, physical, and biological conditions that support biological
assemblages in the absence of anthropogenic stressors. They are subjects of study by
many basic ecologists and by the U.S. Fish and Wildlife Service's HSI (Habitat Suita-
bility Index; U.S. FWS, 1981) and IFIM (Instream Flow Incremental Methodology; Bovee,
1982) groups. At various scales of resolution, habitat indicators are also useful for
classifying major lake and stream types (Brussock et al., 1985; Frissell et al., 1986;
Lewis, 1983) and ecological or biogeographic regions (Bailey, 1976; Omernik, 1987).
• Stressor indicators were defined by Messer as characterizing natural processes,
environmental hazards, or management actions that change exposure and habitat. We
propose that stressor indicators are those that quantify management actions, inactions,
and policies that ultimately change exposure.
All four types of indicators play a critical role in EMAP. However, because the major objective is
to describe the status of and trends in indicators of condition, our primary effort, and the main
focus of this chapter, is on the response indicators, as they will be used to describe the condition
of our lakes and streams. We have chosen to focus on biological measurements as the founda-
tion for the response indicators because we believe that they most effectively assess the cumula-
tive effects of the many physical, chemical, and biological stressors to which we expose our
aquatic resources. Before a response indicator can be effectively used within EMAP, we must
(1) link it to one or more of the selected societal values of concern, (2) describe how a lake or
stream will be effectively indexed for that indicator, (3) determine whether an indicator can be
used to distinguish acceptable from unacceptable conditions, (4) determine whether natural varia-
bility and variability within the measurement process will allow us to adequately describe mean-
ingful status or trends, and (5) identify a process for selecting among the number of candidate
indicators currently available for use.
8
-------
2.1.1 Societal Values and Their Response Indicators
In any monitoring program with a mandate as broad as that of EMAP, it is necessary to narrow
the focus from the broad goals and objectives to the specific attributes that will form the basis of
field monitoring and be suitable for making statements about the condition of resources. As a
first step in narrowing the focus, we attempted to identify the fundamental reasons for society's
concern about the condition of lakes from a biological perspective, which differs from a water
quality perspective, such as drinkability, or an industrial use. The selection of these societal
values drives the selection of appropriate indicators, and effective indicators must be clearly
related to growing societal and scientific concerns over the condition of aquatic resources. After
numerous discussions with managers of aquatic resources, individuals in decision making roles,
and members of the scientific community, we felt that a reasonable place to start was with bio-
logical integrity, trophic condition, and fishability as the initial societal values. This set of values is
not necessarily final or complete, but it represents a starting point. In addition, the Federal Water
Pollution Control Act and its various amendments specifically mandate that these three attributes
be protected, maintained or restored, and reported on periodically.
The societal value of greatest concern to EMAP-SW in the indicator pilot was biological integrity,
because we knew least about how to select indicators for it. Biological integrity can be defined as
the ability to support and maintain a balanced, integrated, adaptive community with a biological
diversity, composition, and functional organization comparable to those of natural lakes and
streams of the region (Frey, 1977; Karr and Dudley, 1981) and includes various levels of biologi-
cal, taxonomic, and ecological organization (Noss, 1990). Waters in which composition, structure,
and function have not been adversely impaired by human activities have biological integrity (Karr
et al., 1986). Karr et al. (1986) also defined a system as healthy "when its inherent potential is
realized...and minimal external support for management is needed." This value or ethic differs
considerably from values oriented toward human use or pollution that are traditionally assessed in
water quality and fisheries programs, in which production of a particular species of game fish is
the goal (e.g., Doudoroff and Warren, 1957), and may contradict them (Callicott, 1991; Hughes
and Noss, 1992; Pister, 1987). We focused our selection on candidate lake indicators that would
aid us in assessing the various structural and functional aspects of biological integrity and in
determining whether biological integrity was at an acceptable or unacceptable level.
We define trophic condition as the abundance or production of algae and macrophytes. Trophic
condition is the focus of an entire section (314) of the Federal Water Pollution Control Act; the
-------
federal government and local lake associations spend millions of dollars annually to control the
growth of algae and aquatic macrophytes. Trophic state involves both aesthetic (water clarity)
and fundamental ecological (production of plant biomass) components. It is, therefore, a key
aspect in determining both a lake's relative desirability to the public, its production of fish, and its
ecological character or classification by limnologists (e.g., eutrophic or oligotrophic). The trophic
state of a lake is largely a function of its geographic location and morphology, requiring us to
consider both in assessments of whether or not the lake's condition is acceptable. Trophic condi-
tion is nested somewhat within the concept of biological integrity; we separate it here because it
remains a primary concern for many lake managers. Like biological integrity, trophic condition
must be interpreted relative to some expected condition for the region and waterbody class.
Fishability is defined as the catchability and edibility of fish by humans and wildlife. Fish represent
a major human use of an aquatic ecosystem product, fishing is a multimillion dollar recreational
industry, and fishing quality and fish edibility are major concerns of millions of anglers. Therefore,
protecting fish is the goal of many water quality agencies, and fish drive their water quality stan-
dards. In addition, states have established agencies whose mission is to maintain and enhance
fisheries. Piscivorous mammals (mink, otter) and birds (eagle, osprey, heron, egret, tern, loon,
merganser), which are among our most desirable wildlife, are even more dependent on fish.
When their fisheries are depleted or contaminated by toxics, they develop abnormalities, are
extirpated locally, or become endangered regionally. Fishability may or may not be nested within
the concept of biological integrity, in that we have many systems with catchable stocks of
contaminant-free fish that have been managed at the expense of biological integrity. These
systems might be considered in good condition with respect to fishability, but in poor condition
from the perspective of biological integrity.
Before the pilot study, we concentrated on defining the values and attributes of concern and
selecting candidate indicators that the scientific literature suggests are ready for use in EMAP or
could be made ready with minimal effort. These candidates were the first to be evaluated. We
will continue to invest in the research community for the development of new indicators of con-
dition, but it was important to begin with those that appeared most ready to be evaluated.
The chief response indicators for trophic condition are chlorophyll-a and macrophyte coverage,
with total phosphorus, total nitrogen, and Secchi disk transparency as important exposure and
habitat indicators. Although macrophytes are poor indicators of water column trophic state, they
do indicate lakewide trophic state, especially in shallow lakes. Chlorophyll-a, total phosphorus,
10
-------
total nitrogen, and Secchi disk transparency are relatively standard and well-accepted measures
of trophic condition, thus our primary concern in the 1991 pilot was to evaluate survey logistics
and the variances associated with the lake population and indexing process (see Chapters 4-6).
Indexing methods for macrophytes were reserved for a later pilot.
In evaluating fishability, our major interests are in the catchability and edibility of fish; that is, are
they present and free of contaminants (Appendix 2A). Our 1991 pilot focused on the catchability
aspect, in other words, how to go about indexing a lake for its fish assemblage. In future years,
edibility will be based on concentrations of chemicals in fish tissue.
Developing indicators of biological integrity poses a much greater challenge than either trophic
condition or fishability, because biological integrity involves assessing several aspects of the
entire community (see Section 2.4.2). We focused indicator development on biological assem-
blages. Using a combination of workshops, literature reviews, and extended discussions among
EMAP-SW staff, members of various state and federal agencies, and the broader academic com-
munity, we reduced a long list of candidates down to those evaluated in the 1991 pilot: fish,
zooplankton, sedimentary diatoms, macrobenthos, and littoral/riparian breeding birds.
• Fish were chosen primarily for their societal value, their relationship to the fishability
value, and their role as a top consumer in lakes, and because they represent a long-
lived assemblage in lakes and therefore reflect the effects of a variety of stressors.
• Sedimentary diatoms are one indicator of primary production in lakes, have served as
an effective diagnostic indicator, and can be used to interpret earlier conditions in
lakes.
• Zooplankton and benthic macroinvertebrates represent the primary consumer level in
lakes and are intermediate lived organisms; they differ in their general habitat
requirements.
• Riparian/littoral birds indicate the condition of a lake's riparian and littoral zone, are
highly valued by society, and can serve as a top consumer in the absence of fish.
Our current position is that fish, zooplankton, and sediment diatoms form the probationary core
assemblages to be used for demonstration monitoring. We will continue to examine additional
pilot information to be derived from benthic macroinvertebrate and bird assemblages. The
questions we sought to answer about these candidate biological indicators during the 1991 pilot
and the results are examined in the remainder of Chapter 2. We focused primarily on (1) deter-
mining the sampling effort appropriate for each assemblage (how to obtain an index sample for
each, Section 2.2), (2) developing candidate metrics representative of each assemblage (Section
11
-------
2.4), and (3) evaluating the responsiveness of assemblages and metrics to a set of lake types and
stressor gradients (Section 2.4).
2.1.2 Habitat, Exposure, and Stressor Indicators
Although this chapter focuses primarily on the development of response indicators, physical,
chemical, and biological habitat play an important role in classifying lakes and in diagnosing
probable cause of impairment or unacceptable condition. Therefore, habitat, exposure, and
stressor indicators are also briefly discussed in this chapter. The primary objectives of the 1991
pilot were to quantify the necessary sampling (indexing) effort (Section 2.2) and to determine the
appropriateness of the habitat, exposure, and stressor metrics via their association with biological
responses (Section 2.4).
2.1.3 Indexing
Indexing is the consistent manner in which a waterbody is sampled, both spatially and temporally,
and the way in which its condition is numerically represented. Indexing is fundamental to all eco-
logical sampling, whether the objective is a site-specific assessment of the effects of a particular
perturbation, or a regional assessment of a population of lakes and streams. Indexing is neces-
sary in EMAP, as in most efforts, because we cannot measure all attributes in all parts of all
waterbodies at all times. We must select an appropriate time period for sampling, the sample
must adequately represent the waterbody's character, and the lakes and streams selected for
sampling must be representative of the population of waters from which they are drawn. We can
meet the last requirement through spatially balanced, probabilistic selection of lakes and streams.
Selection of the attributes sampled requires indicator development and evaluation (Section 2.4).
Selection of a sampling period and sampling locations involves several indexing considerations.
In the issue of indexing, the concepts of quality assurance (QA) also emerge. The role of QA is
to insure and document the repeatability and accuracy of a measurement process. Typically, we
think of this from the perspective of laboratory analysis, that is, the accuracy and precision of an
analytical number given the techniques employed in the analysis. In the context of EMAP, we
broaden this concept to incorporate the repeatability of the entire protocol, including the sampling
location, the sampling period, and the actual measurement of the indicator, whether by field or
laboratory analysis, identification, or enumeration.
12
-------
Acceptable lake indexing methods had already been determined for some attributes and assem-
blages before the 1991 pilot. These (chlorophyll-a, water quality, sedimentary diatoms, zooplank-
ton) were implemented on the probability lakes to answer questions about the logistics of con-
ducting regional surveys (Chapter 6) and to begin collecting data on population and index vari-
ance (Chapter 4). This process will require several more years to evaluate important variance
components in different regions of the country (Chapters 4 and 5). For other indicators, standard
methods of obtaining an acceptable index sample were not available in the literature or in meth-
odologies of the monitoring community. These indicators (macrobenthos, fish, riparian birds,
physical habitat structure) required focused pilot studies to develop efficient indexing protocols
appropriate for a single visit by a small crew. Section 2.2 describes the requirements for an
appropriate index sample, explains the indexing questions we wished to answer in the pilot, and
provides the results and current status of index sampling for each assemblage.
2.1.4 Indicator Evaluation
To ensure consistency among the various EMAP resource groups and comparable evaluation
among candidate indicators within a group, EMAP has developed a set of indicator evaluation
criteria (Olsen, 1992). We describe how these criteria, and a subset pertinent to evaluation of
biological integrity indicators, were used to select and evaluate candidate assemblages (Sections
2.4.1 and 2.4.2). In Sections 2.4.4 and 2.4.5, we explain the process for developing biological
integrity metrics from assemblage measurements and a likely process for culling candidate
metrics and assemblages. Indicator selection is a particularly critical process because of the role
indicators play in linking field measurements, waterbody condition, and ecological values. It is
made difficult by the differing sensitivities, discriminating power, and potential for contradictory
conclusions among any set of assemblages or metrics. We do not intend to select indicators
based on one summer's data, but rather to evaluate the data and present a process for eventually
choosing core indicators.
2.1.5 Determining Nominal/Subnominal Condition
Describing the range of conditions found and the trends in these conditions is EMAP's primary
objective. However, we suggest that the information will be useful to a wider range of decision
makers if we determine the proportion of the waterbodies that are in subnominal (unacceptable)
and nominal (acceptable) condition. This is actually nothing new. As a society, we engage in
this process routinely when we establish chemical criteria and standards or biological criteria and
13
-------
standards. We are not suggesting that EMAP do this alone, or that there is a single number
either nationally or regionally that should be used. The numbers may differ by ecological regions
or waterbody classes. We are suggesting, however, that EMAP enter into this activity in concert
with state and federal agencies as well as with the scientific community, and develop rational
methods for establishing these decision points.
Establishing acceptable or unacceptable condition requires some sort of benchmark or reference
condition against which the sample waterbodies can be compared. Several approaches have
been suggested for estimating reference condition: reference sites, pristine sites, historical data,
paleoecological data, ecological models, empirical distributions, experimental results, and
consensus of experts. Each has advantages and disadvantages, which we describe in Section
2.5; we applied a reference lake approach as a first step in assessing response indicators in this
pilot (Sections 2.3 and 2.4).
2.1.6 Probability Lakes and Indicator Lakes
The 1991 pilot study consisted of lakes selected using two different protocols, depending on the
state of readiness for a particular indicator. For indicators with suitable indexing protocols, we
used the proposed probability design. For fish, birds, benthos, and habitat structure, we hand-
picked lakes from a range of types, sizes, and apparent watershed impact. At these 19 "indicator
lakes," we developed plot sampling protocols. We also evaluated all assemblages together at
these lakes to develop and assess metrics. With a few exceptions, this chapter focuses on the
indicator development conducted at these handpicked indicator lakes. The selection process and
the characteristics of these lakes are described in Section 2.3.
2.1.7 Chapter Layout
This introduction has briefly touched upon the issues necessary to move ahead with the selection
of indicators to meet the first objective of EMAP. The results and progress of the 1991 pilot for
each of these topics are presented in the remainder of Chapter 2, as follows:
2.2 Evaluation of Indexing
2.3 Selection of Indicator Lakes and their Physical and Chemical Characteristics
2.4 Indicator Development and Evaluation
2,5 Distinguishing Nominal from Subnominal Condition
2.6. Summary and Conclusions
14
-------
2.2 EVALUATION OF INDEXING
In this section, we offer an overview of the considerations involved in index sampling for surveys
of this nature, drawn primarily from Stevens (pers. comm.). Following the overview is a descrip-
tion of how we addressed several key indexing issues relevant to lake surveys.
2.2.1 The General Indexing Concept
Selection of a sampling period and sampling sites involves 10 basic aspects (Table 2-1; Stevens,
pers. comm.). The first two involve precision or variance considerations, two are largely logistical
concerns, and the remainder are ecological concerns.
Table 2-1. Characteristics of an Ideal Index Sample3
1. Minimizes natural spatial and temporal gradients and periodicities.
2. Allows a single or a composite sample.
3. Has wide (regional) applicability.
4. Provides sampling ease and efficiency.
5. Distills essential aspects of the ecosystem or assemblage.
6. Establishes relative condition of the system (is sensitive to stressors).
7. Is responsive to stressors that occur at different seasons or in different places.
8. Can be sampled during period of maximum anthropogenic perturbation.
9. Has biological relevance (information rich).
10. Is accurate.
From Stevens, pers. comm.
2.2.1.1 Precision or Variance Considerations
(1) Indexing should minimize natural spatial and temporal gradients and periodicities within the lake
or stream. In other words, sampling should occur in places and at times that are relatively
15
-------
invariant, though not necessarily constant. If the system is changing rapidly for phenological or
seasonal reasons (case 1) or if the habitat structure undergoes cyclical changes, as in the riffles
and pools of streams (case 2), it will be difficult to distinguish anthropogenic effects from natural
noise. In case 1, hydrographs and climatographs are examined for gradients during potential
index periods; in case 2, littoral/profundal or riffle/pool samples are kept separate until data
analysis.
(2) Relative natural variation in the index period and in macrohabitats must be considered to
decide whether a single or composite sample is representative enough. If a single sample is likely
to differ greatly from another, as is expected for some attributes along the littoral zone of many
lakes or in many stream reaches, aggregates of multiple samples are advisable. If not, a single
sample is adequate. Similarly, if the index period occurs in a highly variable and unpredictable
season, as in the spring and fall in some regions, multiple sampling visits are required. However,
EMAP is oriented to single-visit sampling. As a result, we must measure the variation that occurs
during an index period and determine its influence on our ability to describe the status of and
detect trends in indicators of the condition of lakes. Addressing this issue will be a basic part of
the initial years of routine monitoring. We will revisit a subset of lakes (and streams when we
begin monitoring streams) within the index period to estimate index variance relative to other
variance components. See Chapter 4 for a detailed discussion of the variance components and
their relative importance.
2.2.1.2 Logistical Aspects of Indexing
(3) The selected methods should be applicable across a multistate region, both to reduce logis-
tical headaches and to increase the regional applicability of the data collected. We evaluate
regional applicability by considering the physical and chemical character of the waterbodies and
the distribution and abundance of the indicator assemblages. That is, are the major waterbody
types, macrohabitats, and assemblages regionally comparable, or do the salinities, lake level
fluctuations, lake morphologies, or obstructions require regional modifications in sampling
methods?
(4) Clearly the ease and efficiency of collecting a sample must be considered. In evaluating
sampling ease, we are concerned about the availability, work hours, health, safety, and comfort of
field crews. In the Northeast, the sampling window for crews was June to September. Chapter 6
details the logistical considerations of regional surveys.
16
-------
2.2.1.3 Ecological Considerations
(5) An index sample should distill the essential aspects of the system; that is, it should provide
important data about the condition of a waterbody, if not critical information about the fundamental
nature of the lake or stream. This aspect is evaluated by the ability of the various indicators, at
various times and places, to assess biological integrity or trophic state, diagnose unacceptable
levels of integrity or trophic state, or indicate different lake or stream types.
(6) An index sample must establish the relative condition of the waterbody. For example, it must
produce data that allow discrimination among lakes or streams of varying quality. This aspect is
initially evaluated by an indicator's responsiveness along known disturbance gradients and during
periods of greatest, or substantial, disturbance.
(7) Ideally, an index sample is responsive to stressors that occur in places and at times distant
from the time and place of sampling. A substantial perturbation of the system must be detectable
even if it occurs in another season or in a portion of the lake or stream distant from the index site.
This aspect of the sample is evaluated by the life cycle, lifespan, and mobility of the assemblage.
(8) An index sample is best collected with periods and places of maximum anthropogenic pertur-
bation in mind. Atmospheric deposition and agricultural chemical runoff are most stressful in early
spring when snow and ice meltwaters enter lakes and erosive rains occur, but late summer is
usually when eutrophication is of greatest concern because of in-lake processes, warming, and
human use. Places of maximum perturbation in lakes are the littoral and profundal zones and the
hypolimnion of stratified lakes. Littoral areas are most frequently the sites of anthropogenic
changes in physical habitat structure (snag removal, dock construction, beach development, trash
deposits) and water quality deterioration (septic field leaching, lawn runoff, acidification, stream
inlets). The profundal zone collects sediments, and in stratified lakes these can deoxygenate the
hypolimnion as they decay.
(9) We wish to sample biota at biologically relevant times and places, that is, when and where we
can most likely obtain an information-rich sample. We evaluate this aspect by examining pub-
lished life history characteristics of the assemblages sampled.
(10) Indexing must provide an accurate picture of the lake or stream population. Indicators that
produce excessive numbers of false positives or false negatives eventually will result in unneces-
17
-------
sary expenditures or unnecessary loss of ecological values. Overall condition accuracy may be
evaluated through the degree of agreement of multiple indicators and responses to known levels
of disturbance; taxonomic accuracy of voucher specimens is another component of accuracy.
The foregoing criteria serve to guide in the selection and evaluation of lake indexing.
2.2.2 Index Period, Index Location, and Sampling Gear
Before the lake pilot, we evaluated index period adequacy and sampling locations. Preferred
index periods for each attribute were tabulated (Table 2-2) and evaluated in combination. We
estimated that EMAP lake surveys would require a 2-month index period to sample approximately
60-80 lakes in this region. This estimate was based on each of 4-5 crews sampling 2 lakes per
week for 8 weeks. Late fall through early spring was rejected for safety and accessibility reasons.
Late spring and early fall were rejected because of the substantial temperature changes and con-
sequent biological changes that lakes experience over two months during those seasons. June
through early September was chosen as the index period. Birds were sampled by separate
crews in June because this is the peak breeding season, when birds are least mobile and easiest
to identify. July and August are months of substantial thermal, nutrient, and recreational stress on
lakes; also, fish, benthos, macrophytes, and zooplankton assemblages are relatively stable.
Index sites fell into two major groups, midlake and littoral (Table 2-3). Based on previous
experience and the literature (Herlihy et al., 1990; Smol and Glew, 1992; Tessier and Horwitz,
1991), we chose the midlake area at or near the greatest depth for water quality, chlorophyll-a,
sedimentary diatoms, sediment toxicity, and zooplankton.
The middle of the lake was selected for water quality and chlorophyll sampling because it best
represents the pelagic water volume of the whole lake, is well mixed in the surface layer, and is
least influenced by littoral perturbations. Herlihy et al. (1990) found insignificant differences in
water chemistry collected from 3 deep water sites at 41 randomly selected lakes in the north-
eastern United States. The midlake location offers the greatest rain or focusing of diatoms; the
sediments there include pelagic and benthic diatoms and are least disturbed by wind, currents,
and macroinvertebrates. Sediment samples from such locations are relatively invariant; standard
deviations in inferred pH were measured for diatom assemblages of 10 surface sediment samples
and 3 sediment cores from the same lake. The standard deviations were 0.21 and 0.10 pH unit,
respectively (Charles et al., 1991). For many of the same reasons, a midlake site yields the most
18
-------
Table 2-2. Preferred Index Periods for Various Lake Attributes and Assemblages
Attribute
Water Quality
Habitat Structure
Chlorophyll-a
Sedimentary Diatoms
Zooplankton
Benthos
Fish
Birds
Sediment Toxicity
Spring
•
•
•
•
Summer
•
•
•
•
•
•
•
•
•
Fall
•
•
Winter
Table 2-3. Index Locations for Various Lake Attributes and Assemblages8
Attribute
Water Quality
Chlorophyll-a
Sedimentary Diatoms
Zooplankton
Sediment Toxicity
Benthos
Fish
Birds
Habitat Structure
Midlake
1
1
1
1
3
3
1-10
Littoral
3
1-10
20-24
10
Comment
0.5 m below surface
0.5 m below surface
1 40-cm core
vertical tow from bottom
grabs
Stratified subjective
Stratified random, # depends
on lake size
Randomized systematic, #
depends on lake size
Randomized systematic
Numbers indicate number of samples taken. Single samples are taken at midlake, at or near maximum depth.
Multiple samples are taken according to protocols developed during the pilot study and described in the text.
19
-------
representative sample of sediment for sediment toxicity tests. The maximum depth provides the
longest vertical zooplankton tow, any potential hypolimnetic zooplankton assemblage, and the
most varied zooplankton pelagic habitats. Tessier and Horwrtz (1991) reported only 5% within-
lake variation at the index site for zooplankton in the same lakes that Herlihy et al. (1990) studied.
For benthos, birds, habitat structure, and fish, our general concern was to determine if we could
obtain an information-rich, discriminating, and repeatable sample with a minimum amount of
effort, conforming to our need to spend no more than two days on each lake visit; a single day
would be preferable. Our concerns included choosing the location and number of sampling sta-
tions, the number of samples per station, and the type of gear or combination of gear types.
Addressing these questions was a major thrust of our 1991 pilot survey.
2.2.2.1 Benthos
Our indexing objective for benthic macroinvertebrates was to compare the relative effort required
to assess lake condition consistently in 2 to 3 hours by sampling sublittoral, profundal, and littoral
habitats and to evaluate the relative information return in terms of numbers of species and indi-
viduals collected. Benthos were sampled in two habitat types by petite PONAR grab (Appendix
2B): the deepest area and the sublittoral zone (within which we selected two to three sites).
Coarse littoral substrates (cobble, gravel, sand, macrophytes, snags) were sampled at three to
four locations by sweep net or handpicking. Sampling times and necessary sampler expertise
were recorded.
2.2.2.2 Fish
After numerous discussions with lake ichthyologists and fishery biologists in universities and state
and federal agencies, we found no standard protocols for obtaining a lake index sample for the
fish assemblage. We planned to collect all the common and most rare species and to estimate
the proportional abundance of the abundant and common species, which meant sampling the
variety of fish macrohabitats within each lake with habitat-appropriate gear. Our primary means of
determining whether methods were adequate was to sample each habitat with the chosen gear to
estimate marginal return with increasing sampling effort. Our objective for fish sampling was to
determine the combination of gear types and the amount of each required to collect 90% of the
species caught by intensive sampling. We also were concerned about whether to use a sampling
design that resulted in a probability sample stratified by habitat, or one in which we subjectively
chose locations based on expert knowledge about where in lakes fish were likely to be found.
20
-------
Consequently, we compared catches from both stratified random and subjectively selected sites
within a lake.
Ultimately, we needed to develop an indexing methodology that would reduce subjectivity in site
selection among different crews to minimize crew variance, optimize sampling effort, ensure that
major habitat types were sampled, and produce data that could be related to the proportions of
those macrohabitats. These were not simple tasks because of the patchy distribution of fish
species, differential mobility of fish species and size classes, variable presence and extent of
macrohabitats, our inability to view the lake as a fish does, and variable gear efficiency among
different habitats and lakes.
Our pilot study evaluated various gear types to determine which was most cost effective relative
to labor and catch at the lake scale and what combinations of gear would be best. In all, seven
gear types were evaluated (Appendix 2B): gill net, trap net, minnow trap, eel pot, beach and
short seines, and boat electrofishers. The small indicator lakes were sampled for two days with
gill and trap nets at four midlake and four littoral locations, chosen using systematic random and
subjective methods, in those lakes with features that concentrate fish. Large lakes were sampled
for three days with the same gear types at eight locations stratified randomly, and occasionally
subjectively located. Electrofishing effort was evaluated at two to four transects on small lakes
and at four to seven transects on large lakes.
2.2.2.3 Birds
To our knowledge, riparian and littoral birds had not been quantitatively sampled in New England
lakes. Thus our pilot objective for birds was to determine if indexing methods developed for
terrestrial and stream bird surveys (Brooks et al., 1991; Bobbins et al., 1986) could be adapted to
lakes. We were especially concerned with the amount of time required for sampling and the
index period variance. We recorded the time it took to collect the data and sampled bird assem-
blages in each lake twice with an interval of approximately two weeks. An ornithologist indexed
birds at a set of 20-24 evenly spaced sites along the shoreline (Appendix 2B).
2.2.2.4 Physical Habitat Structure
Physical habitat structure, except for macrophyte density, has rarely been quantitatively studied in
lakes. Our methodological objectives for physical habitat in the pilot were to (1) quantify the
amount of time required to obtain the sample from lakes of various sizes, and (2) determine if the
21
-------
variables sampled and the sampling methodology were appropriate. We assessed the appropri-
ateness of the habitat structure data by comparing them with an annotated lake map and an over-
all assessment of lake character to determine if important lake characteristics were missed by the
10 samples; we also examined the data for associations with biological responses. We adapted
variables developed for streams (Plafkin et al., 1989; Platts et al., 1983) to lakes (Appendix 2C)
and measured them at 10 evenly spaced shoreline sites.
2.2.3 Indexing Results
2.2.3.1 Zooplankton
One of our objectives for zooplankton was to begin determining the magnitude of indexing varia-
tion relative to other variance components. (See Chapter 4 for details on the structure for esti-
mating variance components.) Developing a more complete picture of variance components will
require revisits across years as well as within years. The zooplankton estimates summarized in
Table 2-4 give us preliminary insight into the relative magnitude of index variance. Variability
across lakes is clearly higher than index variance; however, index variance for the body size ratio
is of some concern, because it represents about one-half of among-lake variance.
Table 2-4. Zooplankton Index and Lake Variance Estimated for Species Richness and
Body Size Ratio, Two Candidate Metrics for the Zooplankton Assemblage
ANALYSIS OF VARIANCE
Response Metric
Species richness
Body size ratio
Index Variance
12.3
6.0
Lake Variance
46.5
14.1
2.2.3.2 Benthos
Our preliminary decision about benthos indexing was to sample the oxygenated profundal zone
(sublittoral zone), although not all the samples from all the habitats were processed in time for the
information to be included in this report. A cursory analysis revealed that the taxa richness of the
22
-------
profundal deep sites was low, and therefore assemblages collected there contained very little
information compared to what we could obtain from the other habitats. Forbes (1925) described
a similar situation for midwestern lakes. The high microhabitat heterogeneity seen in the littoral
zone was of concern because the sampling effort required to obtain an adequate index sample
was likely to be great. As explained in Section 2.2.1, high microhabitat variability requires either
greater habitat stratification and more samples or a sizeable number of systematic random
samples. Stratification requires considerable experience sampling macroinvertebrates and both
choices result in greater field time than the desired 3-hour limit. However, the sublittoral zone
sites in the pilot produced a wide range of taxa (2-43) and individuals (7-82) among lakes, yet
sites in a lake were physically and biologically similar. In four lakes, triplicate sublittoral samples
yielded variances of 0.2 to 4.2 for a biological integrity index, with mean scores of 19 to 26.
Future benthos sampling, therefore, is expected to focus on the sublittoral zone, but a final deci-
sion must await analysis of the littoral samples and the remaining sublittoral and profundal sam-
ples. Pilot work in 1992 was designed to evaluate the effects of number of samples, sampling
device (petite PONAR, standard Ekman, K-B corer), and mesh size (589, 417, and 246 pm) on
species richness and number of individuals per sample.
2.2.3.4 Fish
The most effective single gear type was a boat electrofisher, in terms of proportion of individuals
and species caught (Figure 2-1), although it had serious logistical limits. Knight et al. (1991) also
found electrofishing was less selective and captured the most fish and species, and the majority
of total species, compared with gill and trap nets, although passive gear added species missed
by electrofishers. However, there is some danger of injuring large trout if electrofishers are
improperly operated (Reynolds and Kolz, 1988; Sharber and Carothers, 1988). Although we did
not kill or mark any fish electrofishing, neither did we examine them for internal injury. Electro-
fishing is clearly desirable because of its high sampling efficiency and because it harms relatively
few fish compared with the mortality for gill and trap nets (usually 100% and occasionally 96%)
and the occasional mortality of nontarget species (30 turtles, 1 cormorant, 2 muskrats, 1 beaver)
in trap nets. However, the logistical problems of generator weight and night sampling required
further pilot research before implementing electrofishing of all lakes, especially those inaccessible
by road. Therefore, a separate study comparing various lightweight electrofishers was conducted
in a subset of lakes in 1992. Because we could not use boat electrofishers in the 1992 pilot, we
examined combinations of the other gear types to see which one would produce an adequate
sample of the fish assemblage.
23
-------
1OO
9O
^ 8O
fro
TO
CD
"o so
CD
tfl" 4O
2O
1O
0
(a)
I
EP MT
SS BS GN
GEAR TYPE
TN
EF
1OO :
9O -
8O •
£ 70-
O)
3 60 -
CO :
0
j- 5O :
CO
il 40 •
xO
30 :
20 -
1O -
0_
(b)
I
I
c
3
a
C
: I
\
EP MT SS BS GN
GEAR TYPE
TN
EF
Figure 2-1. Capture efficiency of fishing gear by (a) % species caught and (b) % individ-
uals caught, both by all gear in combination, for 19 lakes. Box plots show
medians, quartiles, and ranges. EP = eel pot, MT = minnow trap, SS = min-
now seine, BS = beach seine, GN = gill net, TN = trap net, EF = boat electro-
fisher.
24
-------
Gill nets are the most appropriate gear type for sampling the deep water habitats; trap nets
effectively sample the shallow littoral habitats. Initially, EMAP limited the sampling time to two
days per lake for small lakes and three days for large lakes. This was not usually enough time for
a crew to set and tend any more than four trap nets and four gill nets per day, given other samp-
ling duties and travel time. Based on this four nets per day limit in the pilot, our data revealed
that 90% of the species caught in the trap nets could have been caught in three traps in small
lakes (7-28 ha) and in six traps in large lakes (100-564 ha) (Figure 2-2); similar results were
obtained for gill nets. This does not mean that we will always capture 90% of the species present
in a lake. The combination of gear types caught the same species as electrofishers in 11 of the
19 lakes; without electrofishing, 1 species would have been missed in 4 lakes, 2 species would
have been missed in 3 lakes, and 5 species would have been missed in 1 lake. Consequently, in
the 1992 pilot, gill and trap nets were emphasized, with seining and minnow traps in appropriate
habitats.
At only two lakes were more species caught using subjective placement than with stratified
random sampling (Table 2-5). At six lakes, more individuals were caught at subjectively selected
sites. However, stratified random net settings caught more species at four lakes and more
individuals at five lakes. We concluded that the occasionally greater catch does not warrant the
variance introduced by subjective gear placement by crews with different expertise. More impor-
tant, stratified random placement allows us to make lake-wide estimates of species and prevents
biasing the catch by placing gear in concentrating sites that do not represent a major portion of
the lake.
Based on the results of the 1991 pilot, we have developed tentative protocols for sampling fish in
lakes. We will use gill nets chiefly in pelagic macrohabitats. If the lake is unstratified with
> 4 mg/L dissolved oxygen at the bottom of the deep site, all gill nets will be set on the bottom.
The first gill net will be placed near the deepest site and the remainder will be placed midway
between it and the shore in randomly selected directions. If the lake is stratified with > 4 mg/L
profundal dissolved oxygen, gill nets will be set as just described, but at the bottom and at the
thermocline. On lakes > 100 ha, one gill net will be placed at mid-epilimnion and one in the
littoral zone, in addition to the profundal sets. If a lake lacks a well-oxygenated hypolimnion or
metalimnion, the gill nets will all be placed in the epilimnion and littoral zones.
The trap nets will be placed at systematically selected sites in the littoral zone stratified by major
habitat type with the following variables: human influenced or natural, cover or open, cover type,
25
-------
1.0
0.9
en 0.8
o
Q)
Q.
0.7
0.5
>
1 0.4
E 0.3 ^
° 0.2
0.1
0.0
0
(a)
234567
Number of Trapnets
8
1.0:
0.9:
0.8
0)
O 07
0) u-'
a
W 0.6
o^
O 0.5
1 0.4^
E 0.3
O
0.2-
0.1
0.0
0
(b)
234567
Number of Trapnets
8
Figure 2-2. Species/effort curve for fish caught by trap nets in (a) small and (b) large
lakes. Dashed lines are 90% confidence intervals.
26
-------
Table 2-5. Maximum Fish Catches per Net through Use of Stratified Random and
Subjectively Placed Trap Nets
Lake3
BL
Tl
MA
KE
TO
GR
WE
DE
HA
NU
NE
Random
Individuals
74
18
6
19
21
6
7
23
87
2
5
Species
3
3
4
4
4
4
5
5
6
2
2
Subjective
Individuals
4
9
0
150
25
9
8
9
9
5
116
Species
2
3
0
5
6
4
5
4
5
2
2
Two-letter code = the first two letters in a lake's name.
fine or coarse substrate. Some crew discretion will be allowed to ensure that the nets will fish
adequately and not be destroyed; that is, they will not be set on extreme drop-offs, snags, boat
landings, etc. However, the nets will not be placed at sites expected to be highly productive,
such as inlets, outlets, points, shoals, bay mouths, or areas of cover, unless these occur at the
systematically selected sites. Fish caught by the systematically placed gear can be extrapolated
to whole lake values by multiplying the net catch times the macrohabitat volume and then sum-
ming those volumes. Fish caught by beach seining will also be calibrated by beach volume or
area. In addition, crews will be expected to spend one additional hour (1) sampling habitats that
would otherwise not be sampled (such as the highly productive ones just listed) or (2) using non-
standard gear or methods (such as backpack electrofishers, angling, enclosure nets, dip nets, or
baited nets), but the catch will be recorded separately.
27
-------
2.2.3.5 Birds
The circular point count method for estimating bird species and proportionate abundances was
easily adaptable to lakes. Index variance between the two visits of the bird crews was relatively
low, compared with variability among lakes (Table 2-6). Ratios of index variance to among-lake
variance for birds also tended to be lower than ratios for zooplankton (Table 2-4), but the lake
populations were not identical. Apparently, a single visit during the breeding season provides an
appropriate index of the number of individuals and species per lake, as long as sampling occurs
within 4 hours after sunrise and avoids high winds and rain. However, variance components must
be evaluated further on a larger set of probability lakes.
Table 2-6. Index and Lake Variance Estimated for 12 Candidate Richness and Tolerance
Metrics of the Bird Assemblage
Analysis of Variance
Response Metric
Number tolerant species
Number intolerant species
Number tolerant individuals
Number intolerant individuals
Intolerant species/Tolerant species
Intolerant individuals/Tolerant individuals
% tolerant species
% intolerant species
% tolerant individuals
% intolerant individuals
Number specie^
Number individuals
Index Variance
1.3
1.4
242.7
17.4
0.1
0.3
.001
.001
.002
.002
16.1
1412.6
Lake Variance
5.3
3.5
1735.2
100.9
0.3
2.2
.004
.003
.02
.01
47.0
4801 .7
28
-------
2.2.3.6 Habitat Structure
The 10 systematic sampling sites allowed us to quantify the spatial extent of habitat measure-
ments or visual observations made at the lakes. When we compared results of this systematic
physical habitat survey with those from annotated lake maps and focused observations by field
biologists, we found no habitat structure concerns missed by the systematic approach. In fact,
the process of stopping at 10 sites and carefully investigating them produced information on
stressors, habitat conditions, and wildlife that would have been missed by a more cursory shore-
line cruise, especially on the large lakes. With 10 observation stops per lake, the physical habitat
survey should detect, on average, any habitat structure that comprises at least 10% of the lake
shoreline. Consequently, the spatial extent of habitat features must differ by at least 10 percen-
tage points among lakes to be distinguished by the shoreline survey.
The habitat structure protocol was modified slightly for use in 1992. To provide additional infor-
mation for interpreting bird assemblages, vegetation type will be classified as coniferous, decid-
uous, or mixed in future surveys. To aid in placing fish sampling gear, a section was added to
the field form to classify each of the 10 sites by fish macrohabitat type (natural/human, cover/
open/mixed, cover type, substrate type). We found that each observation stop required < 5 min-
utes to make and record observations; on small lakes, sites were 5-10 minutes apart, and on
large lakes with complex shorelines, these times were doubled. We concluded that habitat struc-
ture could be assessed in 1.5-3.5 hours on lakes of 7-560 ha. This method of indexing physical
habitat structure, therefore, seems appropriate to apply to grid lakes as long as the staff are
adequately trained on a diverse set of lakes.
2.2.3.7 Indexing Summary
In summary, we determined that existing indexing protocols were adequate for several lake attri-
butes, including water chemistry, chlorophyll-a, sediment diatoms, and zooplankton (Appendices
2B and 2D). For these attributes, the pilot survey focused on an evaluation of the logistical
requirements for conducting a regional-scale survey, on measuring the status of lakes in the
Northeast relative to these attributes (population descriptions), and on initial estimates of the
replicability of these protocols by repeat visits during the index period.
For other attributes (fish, macrobenthos, riparian birds, and physical habitat structure), the pilot
survey targeted development of protocols for indexing different types of northeastern lakes. For
29
-------
birds, the circular point count method developed for terrestrial systems could be used along the
lake shoreline with little modification. The systematic design for collecting physical habitat
structure data along a lake shoreline is an objective way of obtaining quantitative .information
about a lake's shoreline physical structure. For both these attributes, future research will focus on
quantifying the repeatability of making measurements within the index period and among different
sampling crews.
For fish, an indexing protocol that stratifies a lake's fish habitat, then samples each habitat type
by random placement of the appropriate habitat selective gear is an objective way of obtaining a
lake-wide estimate of the fish assemblage. Intensive sampling of each habitat type allowed us to
determine the amount of each gear type to deploy to reach our goal of sampling 90% of the
species obtained in the intensive sampling. The pilot survey revealed that boat electrofishing was
the most efficient method of obtaining fish assemblage data, but that the gear was too cumber-
some to transport to lakes without landings. Future surveys will use the fish indexing protocols
developed during the 1991 survey to evaluate repeatability offish sampling during the index
period. Also, resources will be devoted to evaluating electrofishing units that can be transported
into remote lakes.
Preliminary evaluation of macrobenthos protocols suggests that the sublittoral habitat is the most
favorable for indexing a lake's macroinvertebrate assemblages. A protocol that systematically
samples this habitat type is expected to emerge as an effective way of indexing benthos in lakes.
The 1992 pilot survey evaluated the effectiveness of several samplers and mesh sizes; evaluating
replicability must await the development of an operational protocol.
2.3 SELECTION OF INDICATOR LAKES AND THEIR PHYSICAL AND CHEMICAL CHARAC-
TERISTICS
The selection of indicator lakes served three primary functions: (1) developing site indexing pro-
tocols for assemblages lacking them, (2) developing metrics and evaluating their sensitivity, and
(3) evaluating the sensitivity of the assemblages. For all three purposes, we sought to represent
a variety of lake types and catchment disturbance types and intensities. This section explains the
selection process, and describes indicator lake physical and chemical habitat characteristics.
2.3.1 Rationale for Selecting the Indicator Lakes
We selected 19 lakes on which to develop and evaluate the use of diatoms, zooplankton,
benthos, fish, bird, and habitat structure indicators. We chose lakes that reflect the range of lake
30
-------
temperature and size throughout the northeastern United States (warm, cold, large, small), as well
as the range of disturbance types and intensities. The 19 lakes, represented in this chapter by
two-letter codes that equal the first two letters of their names, thus represent a broad spectrum of
lakes likely to be encountered in EMAP surveys of the Northeast. We assumed that if questions
about sampling gear, methods, effort, and index locations were answered for this set of lakes, the
protocols developed would be appropriate for most lakes throughout the region, including lakes
of intermediate size and temperature. We chose disturbance types for each lake class that
appeared to be fairly distinct and common to the Northeast. Also, by picking lakes along par-
ticular disturbance gradients, we were able to compare the effects on candidate assemblages and
metrics of suspected stressors and known stressor gradients. We emphasized temperature and
size because literature and data analyses revealed these as major factors differentiating biological
assemblages in New England lakes (Barbour and Brown, 1974; Dixit et al., 1992; Schmidt, 1985;
Tessier and Horwitz, 1991; Underhill, 1985). We selected lakes from Rhode Island to northern
Maine to maximize latitudinal, ecoregional, and biogeographic differences (Figure 2-3).
In selecting the 19 lakes, our concern was not in choosing lakes whose biota had already been
sampled; instead, we simply wished to locate lakes with different types and amounts of human
activity (agriculture, silviculture, residential development, fish stocking). Rawson (1939) consid-
ered human disturbance as important as geology, topography, and climate in determining the
biological character of lakes. We also selected lakes with boat landings to minimize initial access
problems and sought to maximize temperature and size differences, recognizing that the pres-
ence of boat landings indicates at least some minimal level of disturbance. Candidate lakes and
disturbance gradients were determined by consultation with state water quality and fishery bio-
logists and by limited field reconnaissance. Lake selection was made more problematic by the
lack of time before sampling to calibrate the subjective perceptions of different state biologists in
different ecoregions, creating different assumptions of what were highly and minimally disturbed
lakes of a class and stressor type.
More complete and comparable catchment data were obtained for each lake during and after the
index period, allowing a more precise depiction of disturbance levels, as summarized in Table
2-7. USGS land use/land cover data, revealed that the catchments of 2 of the 19 lakes (MA, BL)
were < 50% forest (Table 2-7). This database does not distinguish cut from uncut forest,
although the road density in forested catchments suggests historical cuts. Three lakes had urban
lands (MA, KE, WE) and four had agricultural lands (BL, Tl, FR, JO) in their catchments. Shore-
line disturbances and road and human population densities were also used as indicators of over-
31
-------
O O 9 *orm, small, agriculture/industry
Korm, large, residentiol/urbon
Cold, smoll, fish slocking
Cold, lorgt, silvicullure
scole 1:4,000,000
Figure 2-3. Locations of the 19 indicator lakes in New England.
32
-------
8
.*
2
k_
o
?
'-a
c
en
"^
"o
Conditions
CD
E
.c
o
to
0
CN
ra
1-
Sf—
5
SCO
cc
T3 *i:
« .§ i
""co — >,
LU 2 £
? ° c
LL- Q. 0)
z Q
"S'f'c
Q; fl> (?•
Q
73 OT^°
rr 5 f
Q ^-
2
C Z3 tJ
a a c
^ -^ r?
J w Q_
D)
^
C
ja
5
s.O
c£
Si
3
3
<(
S?
*J
1
£
~ co *$<:
0> -Q c
J 2 (T
S
1
(g O in CO O> ^CMTO
coyjcoinoi 2|s-°>
oj^in^o) m000
incooii^'^ in1"'"
^°cQ2* ^2^
"*~
^oocor^ro ^T-CM
T- T- T— T— •*—•!—
ooo-^P ocoo
000^5; d T^ d
o ^ P "^ o ocncM
dSS^d drioi
00 •* CO CO Q W CM CO
S3" CM O) *~ CM .— T ^ CO O} N-
O CJ>CM|-~inu *i CONCO
! !
h- CO
S. c
3 0)
•B OCD-^^ S wcor;
0) 1- ^- T- W -r- -r- -r-
03 S:
^ tt
§o
_ E,
¥ E
W -0 - « -o
^ CDp-JCCp< ^ ctO<
15 D)
E m
co .3
^ f^ CO
T— CO f^*
CM ^ CO
1^- O CO
•* in co
T- CO CO
co in •*
T— t— T*
^ O) I"
o o t
N. in CM
2 in d
in CM - CM N
O LU LU
_™-
T in J 5!
« —
CM ^ CO ^
^"
,- r "CM o
r- ^- T- °
^~
0000
d d d d
o o o i-
d d d •»'
o> m oo co
®" T-: CM O> CO
JJ O5 O) 00 O)
Q.
E
O)
1
01 T- in cvj oo
CO
t
2
^ "?L|D D LU
|f ^|i o: i-
1o
T- co w to
CO (M T- 00
i- CO CO 00
in co c~ ">.
d iri oo ^
1— 1- T- T-
O O O O
d d d d
o o o o
d d d d
o> m co o
N CO T^ is!
*f*** co co co r^
1
£
1
O CO CO °. N
2
t
2
03 '
0)
0)
J£
JS
j:
o
s
•5
CO
>.
>
3
s
!Q
,
w
c
_g
ra
c
E
LU
CD
D)
(0
1
b
|
'S
£
(0
g
15
_co
LU
CD
H
D
disturbance.
i_
B
(0
fl)
SI
D)
CJ
0)
1
2
5
E
c
v_
CD
c?
O t
in
"co
CD
O
C
CD
0)
2
^
"(0
C
E
CO
in
CD
^
Underlined
D
T3
C
D
O
n)
a
E
CO
o
CD
g
CO
t»
0)
b
LL
W
D
0)
33
-------
all disturbance. The disturbance rankings used for evaluating metrics is the mean rank tabulated
in Table 2-7. High altitude aerial photographs and AVHRR (advanced very high resolution radi-
ometry, from satellite) with 1-km2 resolution, will be examined in 1992 and 1993. The photo-
graphs are appropriate for assessing the land use and land cover of small catchments, but
AVHRR data are appropriate only for regional or large catchment assessments.
The lakes underlined in each group in Table 2-7, and throughout the text, represent minimally dis-
turbed reference lakes for each of the four lake classes. The disturbance gradient within each
lake class also offered a mechanism for developing and examining preliminary estimates of metric
and assemblage sensitivity. Section 2.5 describes the role of these minimally disturbed lakes as
reference lakes. The reference condition derived from a set of reference lakes is one way to iden-
tify lakes in acceptable or unacceptable condition. Section 2.5 elaborates on this concept.
2.3.2 Description of Indicator Lakes
The physical and chemical habitats of the 19 lakes are described in Tables 2-8 and 2-9 and
Figure 2-4. Sampling and analysis methods are tabulated in Appendices 2B, 2C, and 2D. The
21 °C, 5-mg/L dissolved oxygen (DO) habitat is considered the limit for salmonids (Ellis, 1937;
Scott and Grossman, 1973). All eight coldwater lakes had habitats at < 21 °C and > 5 mg/L DO,
although RU had a pH level < 6. Four of the warmwater lakes (FA, JO, BL, FR) had some waters
at these temperatures and oxygen concentrations, but only BL lacked a substantial DO deficit.
Among the coldwater lakes, only HA and TE had seasalt-corrected chloride levels > 30 ,ueq/L,
suggesting substantial catchment perturbation. The reverse was true among the warmwater
lakes, where all but TO and AB had excessive chloride. MA's chlorophyll and TP concentrations
indicated it was eutrophic. RU and NU showed a greater chlorophyll/phosphorus ratio (response)
than the others. In humid temperate regions, chloride concentration is a useful chemical indicator
of land use intensity, and unusually high chlorophyll concentrations for a region are often
associated with reduced biological integrity as described by Moyle (1956).
Physical habitat structure also differed among the lakes for natural and anthropogenic reasons.
Shoreline development, a measure of shape complexity (ratio of shoreline length to the circum-
ference of a circle with the same area as the lake), was > 1.9 for all the large lakes but KE, and
five lakes (UP, AB, BL, TO, GR) had littoral areas > 49% of total area. Six lakes (WE, FA, JO, TO,
AB, Tl) supported mean macrophyte densities that covered more than 29% of the nearshore area.
34
-------
Table 2-8. Selected Water Quality Characteristics of 19 Indicator Lakes
> 4 mg/L DOa
Lake Name < 20°C
Small Warm
ABC
BL +d
FR +d
Tl
MA
Large Warm
GRC
TO
FA +d
JO +d
KE
WE
Small Cold
UPC +
HU +
RU +
TE +
Large Cold
DEC +
HA +
NE +
NU +
a + indicates > 4 mg/L DO and s 20°C; -
Sea salt corrected.
Underlined lakes are minimally disturbed
A small lense of water a 4 mg/L DO and
Chlorideb
(neq/L)
24
78
195
296
1578
35
28
139
110
671
366
7
8
16
228
13
39
8
20
Mean Secchi
Depth (m)
Bottom (3.5)
Bottom (4.3)
2.0
1.9
1.3
2.8
5.3
7.5
4.1
2.5
3.9
2.7
4.6
13.3
6.8
5.8
4.1
2.9
7.6
indicates < 4 mg/L DO and > 20°C
reference lakes.
< 20°C existed in this lake near the
Total Chlorophyll
Phosphorus -a
8.8
11.0
14.0
14.0
30.0
7.8
4.5
5.0
7.9
13.0
8.6
11.0
7.5
0.9
1.8
3.0
6.8
7.0
0.9
thermocline or
4.2
3.6
11.3
6.3
21.4
6.5
3.2
2.1
4.0
6.5
6.1
6.8
5.5
2.0
1.3
3.1
4.9
3.3
2.2
near springs.
35
-------
0)
(Q
Jj
^
o
03
o
c
O)
•5
Q)
.W
3
**
3
*-»
"to
3
n
I
o>
0)
3
ra
H
* $
c cz
.S <0
Oj (Q t—
2 .9-£
o>
"(5 r- CO
>-- -5 t
•5 IT co
n3 "- >
Z O
a>
|£|
.e o |
i|s
2
Q_
CO
SO)
Q
< £
3? co
V
•£
BJ fl»
.c E
k_ O
O d)
CO 5
Q
E E"
c
'if
ro 0
2 Q
0) "ffi"
CO O -C
si S3
00 <
0)
NCOCOCOCO (Mcninmoo o^-CMin coo^in
OOOO-*— T-O-^C\jT-CO OOOO OOOO
OOOQO OOO.-.OO OOOO OOOO
I^-^N-^" ^"^"com^" 0)^0)^ O>COCMCO
o^r — CMO^CO r — r^cor^T— co COT-O)IO h^mcor^
CMCOCMT-T- CNJCOCM-r-CMr- COCOCOCM T-CO-^-CM
CMCMCMOO cocMOomco coinNCO OCJoco
Qtocor-cn ocoocncocn - (ONCO5
NCOT-(^.C\J T— O CM in h*- CO CVJtOCM^1 OOO5CO
' (~ I .-' f* i j^l " rr\ ...- '^j'.^;
^ •«— i— tM T— tM
in co ^ ^ o inmcMcoocM [nOinoi cot-mco
m^^^^f ° ™ * & <2 ^ •* ^ CM ? 5 n - CM
00 « Is- °> o S £! Eo S 8 co N o <° » SSS§
VXJl-xl -i-coT-t-1-in V.MT-T- i-^j-incu
E E 22
CO 05 O O
~- Q) /rjj | — LL ~3 J^ ^ ™ "~^l T" Q^ l~ Si. Q -C 2 21
CO CD ^ L?
Em ^ CO
tn _3 w _i
0)
o
£;
(0
."
tn
Ui
• c.
° 's
"c
II 0
•S ^
fifi
o> -Q
•^ <«
1 -=
!3 °
J T-
1=5
'5 !»
4§
(D (0
T"S
"B C"
J5 ro
V o
•" O
a.i
(0 »
s °
0 13
o>
C CO
£ o>
(0 o
to c
a> 5
4= ^
I 'o
^ 0
2 „
*o $
° fe
» i5
« o
i 'c
fe £
°-l
o> 1=
5 "°
o o .
*- i_ W
S 13 ^
» S B
•o^-o
g §g
S^l
£ .a c
CO LL D
« n o
36
-------
11
NU
RU
10
TE
DE
en
E
AB
FA
JP
BL
O)
X 7
X '
o
T3
6
o
-------
Natural forest canopy was dominant along nine lake shorelines (UP, RU, TE, DE, HA, NU, KE, FR,
AB); the average number of disturbances/sampling site was > 1 at WE, MA, GR, FA, and JO; and
all four lake classes showed fish cover gradients. Clearly, the littoral area and macrophyte den-
sity were not always directly related, as one might predict. Increased shoreline disturbance is
generally associated with lower biological integrity (see Section 2.4), as suggested by Plafkin et
at. (1989).
Sediment toxicity (< 80% of control value) was not found in the four indicator lakes tested (Figure
2-5). In fact, all the indicator lake samples showed greater growth rates than the controls (Figure
2-5b), probably because of more nutritious sediments or the reduced metal solubility of the aero-
bic, high pH, overlying test water. This result suggests that toxics were unlikely causes of the
perturbations observed in these lakes, or that the standard tests used were not sensitive enough.
A more nutritious control sediment and modified REDOX (reduction-oxidation) conditions are also
warranted.
2.4 INDICATOR DEVELOPMENT AND EVALUATION
2.4.1 Selection of Candidate Assemblages
Clearly we cannot sample all aspects of lakes or even all biological attributes of lakes. Instead,
we must implement a process for selecting a set of indicators that will adequately convey the con-
dition of lakes relative to the biological integrity value as efficiently as possible. We chose
assemblages, rather than indicator species or community processes, chiefly because we per-
ceived them to be more directly linked to the ecological value of biological integrity. Assemblage
information is usually of greater concern to the public and to decision makers than the other
options. Also, we understand assemblage ecology and taxonomy better than community proc-
esses, and aquatic assemblages appear to have greater diagnostic power, sensitivity, and respon-
siveness to perturbation than do processes (Schindler, 1987; Ford, 1989; Fausch et al., 1990) or
indicator species (Karr, 1987; Landres et al., 1988). Our conceptual model of a lake ecosystem
(Figure 2-6) incorporates a basic view of lake macrohabitats (riparian, littoral, pelagic, profundal),
as often depicted in limnology and ecology textbooks (Cole, 1975; Smith, 1977; Ruttner, 1963).
Each macrohabitat in the model has assemblages representing the four major trophic levels
(producers, primary consumers, secondary consumers, decomposers). Although we are not cur-
rently evaluating decomposition in lakes, we are evaluating sediment metabolism in streams. If
38
-------
100 -
90 -
80-
70 -
< 60 -
> 50-
W 40-
N?
^ 30 -
20-
10 -
n -
(a)
MA1 MA2 MA3 AB1 AB2 AB3 WE
Lake Sample
KE Control
DU -
50 -
40 -
I
0 30-
f£
CD
55 20 -
10 -
n -
(b)
MA1 MA2 MAS AB1 AB2 AB3
Lake Sample
WE
KE Control
Figure 2-5. Sediment toxicity of 8 samples from 4 indicator lakes: (a) percent survival,
and (b) percent growth.
39
-------
RIPARIAN
Producer
trees
1° Consumer
birds
2° Consumer
birds
Decomposer
LITTORAL -
Producer
macrophytes
1° Consumer
benthos
2° Consumer
birds, fish
Decomposer
BOD
PROFUNDAL
Producer
1° Consumer
benthos
2° Consumer
fish
Decomposer
-^ PELAGIC
Producer
diatoms
1° Consumer
zooplankton
2° Consumer
fish
Decomposer
Figure 2-6. Lake conceptual model showing macrohabitats and trophic levels. Only candi-
date indicators are listed for each trophic level. Note: profundal aquatic
insects may serve as prey for littoral and pelagic fishes, and they are an
important food of riparian birds when they emerge as adults.
40
-------
that research proves fruitful, we can use sediment metabolism in lake pilot studies with minor
modifications of current methods.
Our initial efforts at assemblage selection included listing possible assemblages and lake attri-
butes, evaluating them through workshops (Halsell, 1990; Whittier, pers. comm.), and conducting
literature reviews. We decided nof to use the following in the pilot: (1) lake periphyton and
phytoplankton (except as they are represented by chlorophyll-a measurements and sedimentary
diatom assemblages), because they were considered redundant with, and more variable than,
sedimentary diatoms, (2) lake amphibian assemblages, because they tend to be species depau-
perate in many lakes, (3) macrophyte assemblages (except as areal cover in trophic state esti-
mates), because they too were believed to be species depauperate in lakes, (4) protozoa assem-
blages, because they were assumed to be redundant with diatoms and microzooplankton, but
less important to the lake, and (5) biomarkers, water column bacteria, fish growth rates, and
primary production, because they are not assemblage-level indicators and are highly variable.
The candidate assemblages (riparian/littoral birds, fish, benthic macroinvertebrates, pelagic
zooplankton, sedimentary diatoms) from which we will derive indicators, and the reasons for their
selection, are discussed in the following paragraphs.
Sedimentary diatoms reflect pelagic algal assemblages as well as benthic and littoral assem-
blages. Although they comprise only a part of the algae in a lake, sedimentary diatoms give us a
way to sample the lake's evolution and its condition over recent years. This is impossible with a
single phytoplankton sample. Sedimentary diatoms thus have the potential to provide information
about the presettlement and preindustrial condition of lakes. Also, the high species diversity of
lentic diatoms, and the narrow habitat requirements of many, facilitate their use in diagnosing
acidification (Charles et al., 1990), eutrophication (Hall and Smol, 1992), salinization (Lowe, 1974),
and general pollution (Lowe, 1974; Stoermer et al., 1985), as well for evaluating trends (Charles et
al., 1990).
Zooplankton are the dominant pelagic consumer of most lakes in terms of numbers and biomass.
They are sensitive to fish predation (Brooks and Dodson, 1965), acidification (Sprules, 1975;
Tessier and Horwitz, 1991), toxics (Keller and Van, 1991), nutrient enrichment (Siegfried et al.,
1989), and thermal condition (Patalas, 1990). Often they are the primary prey of fish.
Like zooplankton, benthos are a trophic link between primary producers and fish. They also are
important prey for birds. As the name implies, benthos are the dominant macrofauna of lake
41
-------
bottoms. They have long been used in biomonitoring and they are sensitive to acidification (Lien
et al., 1992), nutrient enrichment (Brinkhurst, 1974), toxics (Chapman and Brinkhurst, 1984), and
organic enrichment (Hilsenhoff, 1987).
Fish represent the end of the aquatic food chain in many lakes and they are important sources of
food and recreation for humans. Unlike the above assemblages, reduced fish diversity has been
reported at the stock, species, assemblage, and faunal levels (Hughes and Noss, 1992). Fish are
sensitive to migration barriers (Ebel et al., 1989), exploitation (Smith, 1968), toxics (Gilbertson,
1992), acidification (Lien et al., 1992), habitat loss (Miller et al., 1989), and introduced species
(Miller et al., 1989).
Traditional aquatic biologists have questioned our inclusion of birds, more than any other assem-
blage, as a candidate indicator. However, lakes and streams cannot be separated from their
catchments (Frey, 1977; Hynes, 1975; Likens and Bormann, 1974; Rawson, 1939), and birds pro-
vide a response indicator linking catchment and surface water condition. The riparian zone is an
integral part of our conceptual model of a lake ecosystem. Just as we identify habitat specific
indicators of other parts of the lake, it is necessary to select indicators of the condition of the
riparian zone. Birds are our candidate assemblage for a riparian indicator. Birds are occasionally
the top consumer in the littoral and pelagic zones and they are important primary and secondary
consumers in the riparian zone. It is useful to monitor birds in the riparian zone because that
macrohabitat is often the first and most disturbed lake habitat and the conduit for much water
from the catchment to the lake. Furthermore, the riparian zone is the major buffer between a lake
and its land use stressors. This is because riparian zones are often wetlands and always active
interfaces between terrestrial and purely aquatic systems; consequently, riparian areas themselves
demonstrate stress and shape the physical, chemical, and biological character of lakes and
streams (Naiman et al., 1992; Gregory et al., 1991). Birds often respond to disturbances before
vegetation itself (Redford, 1992), or before the aquatic biota (Brooks et al., 1991; Carson, 1962),
and they are sensitive to changes in land use, such as agricultural practices (O'Connor and
Shrubb, 1986) and wetland (riparian) damage (Sharrock, 1974). The British Trust for Ornithology
has been surveying riparian birds since 1974 (Spellerberg, 1991). Birds provide an enormously
popular recreational resource (Payne and DeGraaf, 1975) and the bird fauna continues to decline
rapidly. For example, more than 70% of the migrant species in the eastern United States have
declined since 1978 and the number of individuals is half what it was in the 1960s (Ackerman,
1992; Senner, 1986; Terborgh, 1989). Birds are sensitive to habitat stressors (Brooks et al., 1991;
42
-------
Moors and O'Connor, 1992), thermal and recreational stressors (Moors and O'Connor, 1992), and
toxic stressors (Carson, 1962) in aquatic ecosystems.
2.4.2 Indicator Evaluation Process
Throughout the evaluation process, we examine how well each candidate meets the 22 criteria
listed in Table 2-10. Although many of the criteria are appropriate for selecting both assemblages
and metrics, our examples are mostly of assemblages because we are between the assemblage
and metric phases of indicator development (Figure 2-7). When we select metrics instead of
assemblages, we can substitute "metric" for "assemblage" in many of the following paragraphs.
The 22 criteria differ in their importance for selecting assemblages; the most critical ones are so
indicated in Table 2-10.
Table 2-10. Response Indicator Selection Criteria (modified from Olsen, 1992)a
Richness of Biological Information
1. Knowledge of species' trophic and
habitat guildsb
2. Incorporates multiple macrohabitats
3. Includes multiple organismic levels
4. Spatially and temporally integrative
5. High species richness
6. Easily identifiable species
Impact Assessment Ability
7. Broadly sensitive and responsiveb
8. Anticipatory
9. Retrospective
10. Diagnostic
11. High signal/noiseb
12. High index period stability
13. Low measurement error
14. High among-lake/within-lake
variance13
15. Low among-year variance13
Comprehensiveness
16. Applicable to temporary systems
17. Suite of indicators of varying size and
guilds
Societal Relevance
18. High societal valueb
19. Linked with ecological value
Cost and Logistics
20, Inexpensively implemented13
21. Ease of sampling and analysis
22. Rapidly available data
These criteria guide the choices of assemblages and metrics for routine monitoring.
Most critical criteria.
43
-------
SELECT SOCIETAL VALUES
SELECT CANDIDATE ASSEMBLAGES
REVIEW LITERATURE
DEVELOP CONCEPTUAL MODEL
EVALUATE HISTORICAL DATA
WRITE SAMPLING PROTOCOL
COLLECT DATA
ASSESS VARIABILITY
I
DEVELOP CANDIDATE METRICS
ASSESS VARIABILITY & RESPONSIVENESS
DEVELOP INDEX
ASSESS VARIABILITY & RESPONSIVENESS
ASSEMBLAGE INDICATOR
OTHER ASSEMBLAGE INDICATORS
f
DETERMINE
REFERENCE CONDITION
ASSESS VARIABILITY & RESPONSIVENESS
SELECT IND CATORS
ASSESS SOCIETAL VALUE
Figure 2-7. The indicator development process showing the five major phases.
phase involves reevaluation of previous steps.
Each
44
-------
The first set of criteria is oriented towards the biological information available from the index
sample: (1) An assemblage should provide information that allows us to deduce biological, phys-
ical, and chemical habitat quality through knowledge of species habitat, trophic, reproductive, and
general tolerance guilds. Birds and fish offer more of such guild information than the other
assemblages, but sedimentary diatoms are most useful for inferring changes in water quality.
(2) It is desirable for an assemblage to represent a number of macrohabitats of the lake (e.g.,
pelagic, profundal, littoral, and riparian zones). Fish and diatoms rate highest for this criterion.
(3) An assemblage should contain information about a number of organismic levels (individual,
population, assemblage, community). Such assemblages facilitate analyses at multiple levels of
biological organization. Fish provide the most information about multiple organismic levels.
(4) An ideal assemblage integrates the spatial and temporal variability in habitat and stressors.
Birds, fish, and sedimentary diatoms all seem highly integrative. (5) Assemblages should have
high species richness, thereby having the potential to provide greater information about habitat
changes. Diatoms are the most speciose of our candidate lake indicators. (6) Assemblages
should be relatively easy for trained individuals to identify. Bird species are easiest to identify and
most can be identified in the field by trained biologists.
Several criteria are concerned with our ability to assess an impact by means of an assemblage:
(7) Assemblages must be responsive and sensitive to a wide range of direct and indirect bio-
logical, physical, and chemical stressors on biological integrity, such as land use type and
intensity, stocking and harvesting, introduced species, increased turbidity and temperature,
decreased snags and macrophytes, and increased concentrations of nutrients and toxics. Fish
assemblages appear most responsive to the broad array of stressors (Hughes and Noss, 1992;
Karr et al., 1986; Miller et al., 1989; Nehlsen et al., 1991; Williams et al., 1989), but other assem-
blages are often much more sensitive to individual stressors. Typically, large-bodied species and
local endemics are more susceptible to ecological, or actual, extinction (Bedford, 1992; Miller et
al., 1991; Williams et al., 1989) and their loss can greatly distort some assemblages without
outwardly affecting the ecosystem's appearance (Goulding et al., 1988; Paine, 1966; Power et al.,
1985; Redford, 1992). (8) The assemblage should be anticipatory; that is, it should suggest early
warning of future, more substantial changes. An ideal assemblage is (9) retrospective, indicating
past conditions in the population of lakes, and also (10) diagnostic, providing information about
probable stressors. Sedimentary diatoms are the most anticipatory, retrospective, and diagnostic
indicator, at least to water quality changes. (11) The assemblage must have a high signal/noise
ratio so that natural variability does not hinder detection of an impact. The spatial and temporal
integration by sedimentary diatoms gives them extremely low noise. A good assemblage (12) is
45
-------
fairly stable during the index period and (13) has low measurement error, including all aspects of
field sampling and laboratory analyses. Although birds and fish may be more stable, they are
sampled with greater variability, so diatoms reflect the lowest measurement error. (14) For status
estimates, an assemblage should exhibit a high ratio of lake population variance/extraneous vari-
ance (consisting of index, interaction, and year components). (15) For trend detection, the year
component of variance (operating on all lakes together) should be small. These variance compo-
nents and their effects on status and trends estimation are described in Chapter 4. Diatoms,
zooplankton, fish, and birds are believed to have the highest among-lake/within-lake variance;
diatoms, zooplankton, benthos, and fish are estimated to have low among-year variance.
It is also useful to examine the comprehensiveness of the suite of assemblages when making
selections. For example, small lakes are likely to be temporary, becoming wetlands in some
years and lakes in others, or even dry in some years. Thus, an assemblage must be (16) appli-
cable to temporary systems. Birds and sedimentary diatoms are most applicable to ephemeral
lakes if EMAP chooses to assess the integrity of such systems. (17) It is wise to limit redundancy
yet ensure that a diversity of organism sizes, life spans, taxonomic/functional groups, and macro-
habitats are sampled. This suggests monitoring small-sized/short-lived individuals (diatoms or
zooplankton), medium-sized/longer lived organisms (benthos), and large-sized/long-lived individ-
uals (fish or birds). Key taxonomic/functional guilds include producers (diatoms or macrophytes),
invertebrate consumers (zooplankton or benthos), vertebrate consumers (fish or birds), and
decomposers (fungi or bacteria). Important lake macrohabitats are riparian (birds), littoral (birds,
fish, benthos, diatoms, microbes, or macrophytes), profundal (benthos, fish), and pelagic (fish,
zooplankton, or diatoms).
Five criteria assess societal relevance, cost, and logistics: (18) A satisfactory assemblage has
high societal value, that is, people care about these organisms for their own sakes. Vertebrates
rank higher than plants and invertebrates for this criterion. (19) All response indicators for bio-
logical integrity must be linked directly with some ecological value. In other words, a response
indicator of biological integrity should incorporate some aspect of species composition and rich-
ness, guild structure, and life history characteristics. Such metrics are best developed for fish and
birds. (20) An assemblage must be fairly inexpensive to implement, since sampling and analysis
costs limit the number of sites monitored when budgets are fixed. Fish are currently the most
expensive lake assemblage; the others appear to have comparable costs. (21) An index sample
should be relatively easy to obtain from the field and, if necessary, field samples should be
relatively easy to process in the laboratory. (22) Data need to be available for analyses three or
46
-------
four months after the field season. Time-consuming identifications and chemical analyses hinder
our ability to plan for the following summer and to meet the EMAP goal of publishing annual
statistical summaries nine months after the end of the field season. Riparian bird data are the
most easily obtained and most rapidly available.
Two selection criteria or considerations lead to a dilemma. Typically we know more about the
guild memberships of species in species-poor assemblages (fish) than of those in species-rich
assemblages (zooplankton), but both species richness and guild knowledge are desirable. Pre-
sumably this dichotomy results from there being fewer species to know and study in the former
group, as well as from disproportionate societal concern and research funding.
2.4.3 Comparison among Assemblages
Although metrics, not assemblages, ultimately will yield the indicators, it is necessary to choose at
the assemblage level because assemblages are the functional sampling units. Cost savings
come from dropping assemblages, rather than metrics, which usually are only different ways of
summarizing assemblage composition data. It will be necessary to choose a smaller suite of core
assemblages for national implementation because EMAP lacks the resources to monitor all five
assemblages at all lakes annually. The core assemblages could differ from one region to
another.
We have reached the following conclusions regarding culling the five assemblages to yield the
tentative core assemblages for evaluating biological integrity in lakes:
• Fish, zooplankton, and sedimentary diatoms are tentative core assemblages for at least
one cycle of probability lake monitoring.
• The primary reasons for choosing fish are their societal relevance, their representation
of many guilds and levels of biological organization, their sensitivity to multiple stres-
sors, their ability to integrate those stressors through long time spans and across multi-
ple habitats, and their role in evaluating the fishability value, as well as the biological
integrity value.
• Sediment diatoms were selected because they represent the primary producer compo-
nent of lake ecosystems, they have low measurement error, they represent several
chemical habitat guilds and occupy all lake habitats, they anticipate chemical changes,
they provide a historical perspective on lake condition, and they have proven useful for
diagnosing many kinds of lake condition. Furthermore, they are easily collected and
provide an added perspective to evaluation of lake trophic condition as well as to the
biological integrity value.
47
-------
• Zooplankton are the most abundant primary consumers in lake ecosystems, are prey
for aquatic vertebrates, and are relatively easy to collect.
We will continue to evaluate macrobenthos and birds, at a minimum, in northeastern lakes.
Macrobenthos, like zooplankton, are primary consumers and prey for vertebrates, but they
occupy different macrohabitats and are likely to respond to different stressors. Indexing protocols
(sampling a lake for macroinvertebrates) have not yet been fully developed. Our future pilot
research on this indicator assemblage will address indexing protocols, variance components,
metric responsiveness compared with zooplankton, and redundancy with other indicator assem-
blages.
There are strong arguments for including the riparian zone in routine lake monitoring and for
using birds as a core assemblage. They are represented by many guilds, they are spatially and
temporally integrative, and they have very high societal value. Birds are sensitive to physical,
chemical, and biological changes in the lake and riparian zone, and they are much easier to
sample than vegetation or other animals. In fishless lakes, they often represent the primary
vertebrate consumer, although, since most birds are riparian rather than aquatic, they would not
be used to assess the same habitat as fish. Bird research will focus on variance components and
relative sensitivity to disturbances.
Although we have made tentative decisions about the core and additional assemblages, we will
continue to evaluate evidence to support or refute their inclusion. Pilot studies in contrasting
areas of the country will be necessary to weigh the regional strengths and weaknesses of various
indicators and determine whether the probationary assemblages should be modified. For exam-
ple, biological integrity measurements may be meaningless in some reservoirs. Bird surveys may
be redundant with fish surveys in many lakes, but essential in alpine, prairie pothole, and desert
lakes that are naturally fishless and therefore lack another vertebrate consumer. Other indicators,
fish tissue contamination, for example, may be sampled initially for baseline information and then
sampled infrequently thereafter. Although we are unlikely to monitor all indicator assemblages
nationwide, primarily because of cost, a single core collection of indicators that work well for the
northeastern United States may be less applicable in other regions.
The seven critical criteria listed in Table 2-10 for selecting assemblages will be essential in this
process. We have already assessed the candidate assemblages' societal values and their exis-
ting guild information. We cannot accurately estimate their relative implementation costs until
after finalizing the field and laboratory protocols for all the candidate assemblages. Meaningful
48
-------
variance estimates require several years of sampling with the established protocols. Similarly,
satisfactory estimates of assemblage responsiveness and signal/noise ratio require a greater
sample size than the 19 indicator lakes. However, the 1991 pilot is useful as an example of the
process of using assemblage sensitivity and signal clarity in assemblage selection. Thus, we
sought comparative assemblage information to answer the following questions:
• What is the sensitivity of an entire assemblage to the selected disturbance gradients?
This criterion was evaluated by examining the response of the aggregate set of candi-
date metrics across different types and intensities of catchment disturbance. A par-
ticular concern was the total percent of occasions that the assemblage performed as
predicted.
• How clear is an assemblage's signal? We evaluated this criterion by comparing the
aggregate metric responses for all four minimally disturbed lakes to the responses for
the 15 lakes with more highly disturbed catchments.
• Are reference lakes useful for assessing the condition of an aggregate of assemblages?
We measured this criterion by comparing the total number of hits (metric score > 1.5X
the minimum in a lake class) for each lake class.
As already stated, mere collections of assemblages do not adequately express lake condition; we
need a way to extract information from collections of species and their proportionate abundances.
The next section describes this process and the role of the surveys of the indicator lakes.
2.4.4 Selection of Candidate Metrics
In order to extract useful information from the assemblage collections that describes the condition
of lakes, we have adopted our primary approach from Fausch et al. (1990). In this approach, we
select a series of metrics to represent various aspects of an assemblage under consideration.
The process involves converting assemblage richness and composition measurements to a set of
intrinsically important metrics (e.g., the number of native species), as well as to metrics known or
likely to be responsive to the expected array of anthropogenic disturbances to lakes. Fausch et
al. (1990) summarize a series of changes that occur when aquatic ecosystems are subjected to
various kinds of anthropogenic stress. This list (as modified in Table 2-11) served to guide the
selection of many candidate metrics. We based our selection of others on the literature or on
evaluated historical databases available in our laboratories. We also used results of the pilot
survey to identify metrics apparently responsive to the lake types and gradients we chose—an
essential step for assemblages such as zooplankton that are not typically used to assess
disturbance. For each assemblage, we developed a conceptual model of how candidate metrics
might respond to major stressors (see Table 2-11 and Figure 2-8 for examples) as a way of
synthesizing the set of candidate metrics.
49
-------
Table 2-11. Typical Effects of Environmental Degradation on Biological Assemblages, and
Candidate Fish Metrics
A. Typical effects of environmental degradation on biological assemblages (from Fausch et al.,
1990; Hughes and Noss. 1992; Margalef 1963).
• The number of native species, and of those in specialized taxa or guilds, declines.
• The percentage of exotic or introduced species or stocks increases.
• The number of generally intolerant or sensitive species declines.
• The percentage of the assemblage comprising generally tolerant or insensitive species
increases.
• The percentage of trophic and habitat specialists declines.
• The percentage of trophic and habitat generalists increases.
• The abundance of the total number of individuals declines.
• The incidence of disease and anomalies increases.
• The percentage of large, mature, or old-growth individuals declines.
• Reproduction of generally sensitive species decreases.
• The number of size- and age-classes decreases.
• Spatial or temporal fluctuations are more pronounced.
Metrics may be based on number or percent of individuals or
B. Candidate fish metrics.
species.
Assemblage Composition
• Ordination score
• Species richness
• Family richness
Trophic Guilds
• Omnivores
• Invertivores
• Planktivores
• Piscivores
Tolerance Guilds
• Intolerant species
• Sensitive species
• Tolerant species
Habitat Guilds
• Pelagic species
• Benthic species
• Littoral species
Reproductive and Life History Characteristics
Nonguarding lithophils
Young of year species richness
Juvenile species richness
Native adult game species
Trophy fish
Introduced species
50
-------
*?
1.".
Q- (0
Q 0)
S28
O
CO
(0
0)
.Q
_0>
•a
0)
TJ
ft
n
T3
I
0)
+*
c
i
Q
te <0
O 0)
II
*l
'C 0)
^2
*i (0
•Es
« E
•§,2
o o
_ *->
II
?I
^li
s^s
(0
0)
o
a>
3
O)
51
-------
Our goal is to derive a single indicator (from a series of carefully selected assemblages and
metrics) to serve as the indicator of biological integrity. However, for the short term, it is more
reasonable to choose a series of indicators, with each one representing a lake's condition relative
to a particular assemblage. The two options are not mutually exclusive. Thus, a major challenge
is to develop and aggregate the set of candidate metrics into an efficient set that extracts enough
information from each assemblage to adequately describe the biological integrity of lakes. This
process involves evaluating the metrics against the same screening criteria identified in Table 2-10
and evaluating the performance of the metrics in pilot field surveys.
The general process of metric development and evaluation is described in Section 2.4.2, using
assemblages as examples. In the 1991 indicator pilot, we focused on two critical criteria for
metrics: sensitivity or responsiveness to the disturbance gradients chosen and signal/noise ratio.
Future research on metrics should include lakes experiencing various levels of eutrophication,
toxics, acidification, level alterations, and shoreline and riparian modifications. In all these
studies, a major concern is in evaluating a metric's sensitivity to disturbance and the clarity of its
signal. Such information will aid in the development and selection of a useful set of metrics for
assessing condition and diagnosing probable stressors, although it is not conclusive. If a metric
does respond as expected, we can be more confident that it will respond to proven stressor gra-
dients. Lack of metric response is one reason for rejecting a candidate metric, unless it is sen-
sitive to other stressors. A large part of a metric's sensitivity and noisiness may be a function of
its score interpretation. By rating ranges of metric scores as 5, 3, or 1, Karr et al. (1986), for
example, increased the signal clarity of 12 fish assemblage metrics. As with assemblages, further
study of metric responsiveness, signal/noise ratio, and the other criteria, particularly costs and
variance components, will continue on larger numbers of probability and handpicked lakes in
future years.
Some researchers have suggested using only a multimetric index approach; others have encour-
aged the use oi' multivariate analysis alone. We plan to use both. It is too early in the indicator
development process to choose either path, and each has value. Multivariate analysis is a useful
tool for evaluating species or guild structure through use of a single figure and for classifying lake
or stream types. Trophic, tolerance, and habitat guilds have proven useful for analyzing assem-
blages of species whose autecology is fairly well developed (e.g., diatoms, fish, and birds), and
species richness and life history characteristics are of intrinsic interest. Each approach also has
drawbacks. Multivariate analysis involves considerable subjectivity in its interpretation and axes
52
-------
cannot be interpreted without being associated with other indicators. Multimetric indices require
subjective species assignments and considerable calibration; also, metrics are poorly developed
for some assemblages. Therefore, we plan to evaluate both multivariate analysis and multimetric
indices in the search for indicators that are sensitive to anthropogenic disturbances yet relatively
unaffected by natural fluctuations.
Our process for selecting and evaluating candidate metrics for each assemblage initially allowed
considerable latitude to the scientists responsible for developing each indicator. General gui-
dance, as described in the previous paragraphs, defined the goals of the process, but the details
were left to individual investigators. This process, from which we could all learn, allowed metrics
to be developed and defined in diverse ways, within general constraints. The time taken to make
data available for each assemblage varied considerably, thus the time for data analysis did, as
well. As a result, a variety of metric types and analytical procedures were used, reflecting the
investigators' interpretation of the goals as well as their choices of appropriate analytical tools for
the task.
As individual investigators developed candidate metrics, they shared the information to assure
that individuals didn't deviate too far from the goals, and to propose possible metrics based on
those that served other assemblages well. This section describes the metrics proposed for each
assemblage and illustrates their sensitivity to the different lake types and to the catchment
disturbance gradients.
In the future, we will develop a logically consistent set of metrics across all assemblages, as well
as assemblage-specific metrics, and assess various ways of metric selection and evaluation. We
expect to do this after a more thorough evaluation of assemblage theory, existing databases, and
additional years of EMAP data.
In evaluating the sensitivity of various metrics, the expected condition was a function of the
desired direction of change. If low metric scores were considered desirable (e.g., % tolerant
species), they were considered unacceptable if they were > 2X the minimum for a lake class;
marginal scores were those that were 1.5-2X the lowest score. If high scores were desirable
(e.g., number of native species), they were rated as unacceptable if they were <0.5X the highest
score in a lake class; marginal ratings were given metrics 0.5-0.75X the high score. These score
ratings were considered appropriate by participants in an indicator workshop (Thornton, pers.
comm.), but we recognize that they are arbitrary and that metric responses may be nonlinear.
53
-------
Choosing any arbitrary metric score implicitly suggests that we know what a desirable condition
is, regardless of disturbance. This is probably incorrect, but it removes the circularity of evalu-
ating the condition of reference lakes using metrics scored from them. In addition, using arbitrary
scores to assess effects reflects our incomplete knowledge about all impacts and the actual or
true disturbances in the lake/watershed. At the outset, we also recognized that the sample size
was small, hence we could not meaningfully describe the variance associated with these descrip-
tions of the expected condition. However, we use the results from the indicator lakes to illustrate
these methods of identifying an expected condition and examining metric sensitivity.
Metric scores were evaluated against each of the four stressor (land and resource use) gradients
and lake classes we chose (Section 2.3). The most and least disturbed lakes in each class were
examined for substantially different scores, and moderately disturbed lakes were examined for
intermediate scores. Making comparisons by lake class avoids developing expectations for con-
siderably different sizes and types of lakes.
2.4.5 Results and Discussion of Indicator Evaluation on Indicator Lakes
This subsection first discusses the results of evaluating each metric and lake class, beginning
with diatoms and ending with birds, then compares these results by assemblage.
2.4.5.1 Metric Results and Discussion
In this subsection, each metric is evaluated by lake class, and as before, the four reference lakes
are underlined to distinguish them from the others.
Sedimentary Diatoms. Six response metrics (changes in position on a detrended correspon-
dence analysis (DCA) plot, taxa richness, total phosphorus, pH, chloride, and a disturbance
index) were evaluated for diatoms. The DCA and richness metrics evaluate the fundamental
structure of the assemblage, the chemically oriented metrics capture the habitat tolerance and
specialist aspects, and the disturbance index integrates all the metrics. Canonical correspon-
dence analysis (CCA) and weighted averaging regression, based on water quality measurements
and diatoms from surface sediments of 66 lakes (19 indicator lakes and 47 probability lakes),
were used to calibrate the chemical requirements of 245 diatom taxa. Taxa representing the
range in total phosphorus (TP), chloride, and pH were used to develop diatom-inferred chemical
values. The r2 between inferred and measured values was 0.77 for TP, 0.66 for chloride, and 0.86
54
-------
for pH. After further calibrations, the EMAP database and the one generated from other
northeastern lakes will be combined to provide more accurate predictions.
Changes in chemical habitat were suggested by several metrics. Diatom-inferred TP indicates an
increase for four small warm lakes (AB, MA, Tl, and BL), relative to FR, and two large warm lakes
(KE and JO), relative to WE, but no cold lakes (Figure 2-9); thus 3 reference lakes and only 5 of
15 disturbed lakes were distinguished. Diatom-inferred chloride change suggests marked
increases in BL, Tl, MA, and FR relative to AB (small warm lakes), in KE, TO, GR. FA, and JO
relative to WE (large warm lakes), in UP and TE relative to RU (small cold lakes), and in DE and
HA relative to NU (large cold lakes). Chloride change thus distinguished only 1 reference lake
and 10 disturbed lakes. In general, TP and chloride concentrations have increased most in those
lakes with high catchment disturbance (Table 2-7; Figure 2-10) and high measured TP and chlor-
ide values (Table 2-8). Only one of the indicator lakes (RU) revealed notable acidification,
although pH has declined slightly in several of the low TP lakes (Figure 2-11). In general, the pH
of lakes with high catchment disturbance has increased, similar to the findings of Cumming et al.
(1992) in the Adirondacks. Only 5 disturbed lakes appeared to experience substantial enrich-
ment, despite considerable levels of disturbance in the catchments of 13 lakes. This suggests
that (1) the effects of catchment disturbances were not as great as expected, (2) the metrics were
not sufficiently sensitive to the disturbances, (3) the diatom assemblages were not affected by the
disturbances, or (4) the core was not long enough to provide a presettlement layer, possibly
because nutrient enrichment stimulates macrophyte growth and greater sediment accumulation.
However, Tables 2-7 to 2-9 indicate considerable catchment disturbance in several of the lakes,
and diatoms were affected by nutrient enrichment in 6 lakes and by chloride increases in 13
lakes. Dixit et al. (1992) summarize the sensitivity of diatoms to the nutrient enrichment typically
resulting from such disturbances; disturbances that yield more phosphorus and chloride affect
diatom assemblages. Because our results clearly indicated that the catchments were disturbed,
we assumed that the cores were too short; core length was increased for the 1992 pilot.
A detrended correspondence analysis (DCA) of historical and present diatom flora from the indi-
cator lakes (Figure 2-12) suggests several have experienced anthropogenic or natural changes;
these changes in DCA values indicate intrinsic change in assemblage composition. All the small
warm lakes (AB, BL, Tl, FR, MA) changed considerably between historical and present times, but
only BL changed significantly relative to AB. UP. RU, and TE changed more than HU (small cold
lakes); HA and NU experienced markedly more change than DE (large cold lakes); and TO FA,
JO, KE, and WE changed more than GR (large warm lakes). These changes may have resulted
from previously explained responses to chemical changes. As demonstrated in Figures 2-9, 2-10,
55
-------
20
3 10
Q)
D)
C
CO
jr
O
CL
O
i_
i_
0)
s—
C
0
-10
-20
0 1 2 3 4 5 6 7 8 9 10111213141516171819 20
DE HA NU NE UP HI) RU GR AB FR TE TO FA JO KE WE Tl BL MA
Lakes Arranged with Increasing Agriculture & Urban Activity
Figure 2-9. Diatom-inferred total phosphorus (TP) change in indicator lakes. Underlined
lakes are reference lakes. Agricultural and urban activities were determined
from a land use database.
56
-------
~ 200
0)
D)
C
co
JC
O
150
100
| 50
O
Q)
i_
v_
CD
>+-
C
0
J I
• 9" • ^^gg> «^^^
I I I I I I
j I
0 1 2 3 4 5 6 7 8 9 10111213141516171819 20
DE UP HA HU GR NE NU AB TE FA TO FR RU JO KE WE Tl BL MA
Lakes Arranged with Increasing sum of Agriculture, Urban Activity, & Road Density
Figure 2-10. Diatom-inferred chloride change in indicator lakes. Underlined lakes are
reference lakes. Agricultural and urban activities and road densities were
determined from a land use database.
57
-------
0.50
S) 0.25
c
CD
O
X
Q.
OJ
i_
CD
0.00
-0.25
-0.50
I
0 1 2 3 4 5 6 7 8 9 10111213141516171819 20
DE HA NU NE UP HU RU GR AB FR TE TO FA JO KE WE Tl BL MA
Lakes Arranged with Increasing Agriculture & Urban Activity
Figure 2-11. Diatom-inferred pH change in indicator lakes. Underlined lakes are reference
lakes. Agricultural and urban activities were determined from a land use
database.
58
-------
350 -
300
250
CM
.£ 200
X
150
100
50 -
RU
RU D
KE
UP
• Top
• Bottom
HU>
,HU
BL
" . 1 . 1
1 50 200
i , AB / i
250 300
Axis 1
,
350
i
400
Figure 2-12. Detrended correspondence analysis of diatom assemblages. The line lengths
represent the amount of change between the assemblages present at the
surface and bottom sections of a 40-cm sediment core.
59
-------
and 2-11, RU was acidified, and TE, FA, FR, JO, KE, Tl, BL, and MA showed considerable catch-
ment disturbance. MA and BL were enriched considerably, and AB. Tl, KE, and JO were slightly
enriched; although, given their current TP concentrations, it is unlikely these increases substanti-
ally altered the flora. This metric distinguished 3 of 4 reference lakes and 10 of 15 disturbed
lakes. An even clearer picture may have resulted had longer cores been used. It is also possible
that diatom taxa were replaced by others as a result of interspecific competition unrelated to
catchment disturbance.
The number of taxa in the top and bottom sections of the sediment cores showed little relation-
ship to lake area or lake volume, presumably because of the small habitat space required by
individuals, or what is known as the paradox of the plankton (Hutchinson, 1961). Small habitat
requirements are easily met by small or large lakes because both volumes are many orders of
magnitude greater than those required by a microscopic organism (Smith, 1950). The top sec-
tions of sediment cores from BL, Tl, and MA (small warm lakes), TO, JO, KE, and FA, (large warm
lakes), and NU and NE (large cold lakes) all contained markedly fewer taxa than the maxima for
their lake classes (Table 2-12). Therefore, taxa richness distinguished all 4 reference sites and 9
of 15 disturbed sites. The change in taxa richness between bottom and top sections was much
lower than the maximum in each class for all lakes but AB (small warm lake), JO (large warm
lake), RU and TE (small cold lakes), and DE (large cold lake). Richness change distinguished 2
of 4 minimally disturbed reference lakes and 12 of 15 disturbed lakes, and was therefore slightly
more affected by the level of watershed disturbance than was present species richness.
Change in the diatom disturbance index revealed a number of disturbances similar to the number
for richness change (Table 2-12). Index change was substantially greater in BL, FR, Tl, and MA
than in AB (small warm lakes), in GR, FA, JO, and KE than in TO (large warm lakes), in UP. TE,
and HU than in RU, and in HA, NU, and NE compared to DE. The change in index values was
able to distinguish 2 of 4 reference lakes and 12 of 15 lakes with disturbed catchments.
The diatom results indicate the need to (1) ensure that the cores are long enough to collect pre-
settlement diatom assemblages and (2) date the bottom layers to substantiate presettlement
times. The lengths of our cores ranged from 30 to 40 cm, which was long enough to recover pre-
settlement sediments from oligotrophic lakes, but not from all eutrophic systems. In areas of the
country like New England that were deforested in the 18th and 19th centuries (Thompson, 1853;
Marsh 1864), the major enrichment of lakes may have occurred at times earlier than those repre-
sented by the bottom layers of our cores (some as recent as 125 years ago). If this is the case,
60
-------
Table 2-12. Sedimentary Diatom Taxa Richness and Disturbance Index of 19 Indicator
Lakes
Lake
Small Warm
ABC
BL
FR
Tl
MA
Large Warm
GRC
TO
FA
JO
KE
WE
Small Cold
UPC
HU
RU
TE
Large Cold
DE°
HA
NE
NU
Top
Taxaa
64
38*
53
42 *
29 •
96
72 *
48 •
72 *
64*
75
82
64
84
81
91
76
36 •
44 •
lndexb
2.5
4.5*
3.0
3.7*
4.8*
1.7
1.8
2.7*
3.0 *
3.2 *
1.8
1.8*
2.7 •
1.0
1.8*
1.7
1.8
2.3
2.2
Bottom
Taxaa
52*
56
59
69
34 •
85
79
57*
55 *
60*
69
89
63*
67*
64*
79
73
60*
48*
lndexb
2.3
2.0
2.2
2.3
3.7*
1.5
1.7
2.0
2.0
1.7
1.7
1.7
2.8 •
1.3
1.8
1.7
1.7
2.0
2.0
Change
Taxaa lndexb
+ 12
-18 •
-6 •
-27 •
-5 •
+ 11 *
-7 •
-9 •
+ 17
+4»
+6 •
-7 •
+ 1 •
+ 17
+ 17
+ 12
+3 •
-24 •
-4 •
0.2
2.5 •
0.8 •
1.4 •
1.1 •
0.2 •
0.1
0.7 •
1.0 •
1.5 •
0.1
0.1 •
-0.1 •
-0.3
0.0 •
0.0
0.1 •
0.3 •
0.2 •
= 0.5-0.75X the maximum for a lake class; • = < 0.5X the maximum.
*
c
* = 1.5-2X the minimum for a lake class; • = > 2X the minimum.
Underlined lakes are minimally disturbed reference lakes.
the bottom layers we collected may depict disturbed, rather than reference, conditions where
lakes, watersheds, and riparian zones have been recovering and are now largely forested.
Enriched lakes may require cores 60 to 100 cm long.
Although well established as a diagnostic indicator and often used to evaluate historical change,
this assemblage requires more research before indicators can be chosen. Future work will
involve additional literature review and data analyses to develop and evaluate metrics such as
61
-------
metal tolerance/intolerance, benthic/pelagic, turbid/clearwater, coldwater/ warmwater, oligotrophic/
eutrophic, generally sensitive/insensitive species, and organic and toxic tolerance/intolerance. In
place of DCA, we will examine assemblage similarity indices such as Pinkham and Pearson's B,
which Boyle et al. (1990) found to be the most sensitive, stable, and consistent among 16 com-
monly used numerical indices. Work will also continue on refining the diatom disturbance index.
Zooplankton. Three metrics (species richness, the ratio of large to small zooplankton, and the
number of trophic links) were evaluated for this assemblage, and 2 new rotifer species and 12
new range records for New England were revealed in this pilot. Species richness was of interest
for its own sake, the large/small ratio was developed to assess size classes, and trophic links
were examined to evaluate trophic specialization and complexity. Species richness tended to
increase with lake area and often with temperature and disturbance (Figure 2-13). Because of
this relationship, expected scores for zooplankton (and later, fish and bird) species richness were
estimated by extending a line through the top scores of the small and large lakes, similar to the
method recommended by Karr et al. (1986). AB, Tl, MA, and BL supported fewer species than FR
(small warm lakes); GR had markedly fewer taxa than FA (large warm lakes), RU had markedly
fewer taxa than TE (small cold lakes), and NE had markedly fewer taxa than DE (large cold
lakes). This metric was affected by the disturbances at 5 of 15 disturbed lakes and distinguished
2 of 4 reference lakes. This lack of sensitivity to watershed disturbances and fish stocking may
partly explain why zooplankton species richness is not commonly used as an indicator.
Ordinations of zooplankton assemblages showed that individual body size was correlated with
much of the variation among lakes. A preliminary metric was developed based on the ratio of
macro- to micro-zooplankton (Table 2-13). Substantially lower values were obtained for AB, Tl,
FR, and MA compared with BL (small warm lakes). Among large warm lakes, KE, TO, GR. JO,
and WE had substantially lower values than FA. RU and UP had substantially lower ratios than
TE (small cold lake). The ratios for DE and NU were markedly lower than for NE (large cold
lakes). Zooplankton size ratio detected perturbation in 9 of 15 disturbed lakes, but distinguished
none of the minimally disturbed lakes from the others. Although zooplankton size structure is
known to be responsive to nutrient enrichment (Tessier and Horwitz, 1990) and fish predation
(Brooks and Dodson, 1965), it was not sensitive to watershed development or fish stocking. This
may have occurred because watershed changes did not sufficiently increase nutrient levels or
because stocking did not change planktivory rates.
The number of trophic links in the crustacean food web (Table 2-13) was found to be no more
revealing of the effects of catchment disturbance on zooplankton. AB and MA both had markedly
62
-------
45 -
40 -
LLJ
0 35
LU °°
Q_
V)
LJ_
0
CE
LLJ 30
CO
^;
•2
D
_J
^ 25
O
20
15
• Small Warm
+ Large Warm + FA
* Small Cold
• Large Cold +J0
+ TO
+ KE
• FR
• NU +WE
•BE
*TE +GR
• HA
#HU
• NE
• MA
*UP «TI
RU
ABt 9BL
0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8
LogiQ LAKE AREA (ha)
Figure 2-13. Zooplankton species richness versus lake area. Underlined lakes are
reference lakes.
63
-------
Table 2-13. Ratio of Large to Small Zooplankton and Number of Trophic Links at 19
Indicator Lakes
Lake
Small Warm
ABb
BL
FR
Tl
MA
Large Warm
GRb
TO
FA
JO
KE
WE
Small Cold
UPb
HU
RU
TE
Large Cold
DEb
HA
NE
NU
Ratio3
1.94 •
7.89
1.97 •
1.46 •
0.79 •
3.30 •
2.38 •
12.71
3.98 •
0.23 •
0.58 •
3.60*
5.27
2.45 *
6.80
6.08*
8.01
10.21
2.68*
Trophic Links3
27*
39
45
37
24*
45
53
51
50
28*
33 *
46
21 •
12 •
32 *
63
65
35 *
69
* = 0.5-0.75X the maximum per lake class; • = s 0.5X the lake class maximum.
Underlined lakes are minimally disturbed reference lakes.
fewer links than FR (small warm lakes), whereas WE and KE supported markedly fewer than TO
(large warm lakes). All three disturbed small cold lakes (HU, RU, TE) had substantially fewer links
than UP, but only NE among the large cold lakes had a marked reduction in links. The trophic
link metric was affected by the disturbances at 7 of 15 disturbed lakes and distinguished 3 of 4
reference lakes, thus it was the most sensitive zooplankton metric. The zooplankton assemblage
clearly requires considerably more metric development before it will be useful for assessing
acceptable biological integrity. Continued research on trophic and thermal guilds have potential
and the stimulatory effects of both fish predation and nutrient enrichment on microzooplankton
require further study.
64
-------
Benthos. A set of metrics based on those in Plafkin et al. (1989) for streams was selected,
evaluated separately, and combined into a Biotic Index (Bl). The metrics included taxa richness
and two redundancy measures: mean number of individuals/taxon and % of individuals con-
tributed by the dominant taxon. Higher scores for the redundancy metrics and Bl indicate
disturbance (Table 2-14). No relationship was evident between taxonomic richness and lake area,
probably because the greater habitat complexity of larger lakes is not reflected in profundal
sediments, which tend to be homogeneous (Forbes, 1925). Among small warm lakes, AB, Tl MA,
and FR had substantially lower richness than BL. KE, FA, GR, JO, and WE all supported fewer
than half the taxa of TO (large warm lakes). HU, RU, and TE had fewer species than UP (small
cold lake). DE. HA, and NU contained fewer taxa than NE (large cold lake), possibly because the
large amounts of sunken logging debris provided additional habitat in NE. This metric showed
the effects of disturbance in 12 of 15 disturbed lakes, but distinguished only 1 minimally disturbed
lake out of 4.
Among the small warm lakes, Bl scores were substantially higher for FR, Tl, and MA, relative to
AB; there was no marked difference in scores among the large warm lakes. HU and TE had
higher Bl scores than UP (small cold lakes), and NU and NE had markedly higher scores than
DE. Although the Biotic Index distinguished 4 of 4 minimally disturbed reference lakes, it
indicated the effects of disturbance in only 7 of 15 disturbed lakes.
Both redundancy metrics demonstrated a similar pattern. The mean number of individuals/taxon
was substantially higher for BL, FR, Tl, and MA than for AB (small warm lakes), for TO, FA, KE,
and WE than for JO (large warm lakes), for HU compared with TE (small cold lakes), and for DE,
NE, and NU compared with HA (large cold lakes). The % of individuals as the dominant taxon
was greater in BL, FR, Tl, and MA than in AB (small warm lakes), in TO, FA, KE, and WE than in
JO (large warm lakes), in HU compared with RU (small cold lakes), and in DE. NE, and NU com-
pared with HA (large cold lakes). Both metrics distinguished 3 of 4 reference lakes and 11 of 15
disturbed lakes.
Thus, species richness and the two redundancy measures distinguished the expected effects of
lake disturbance levels more often than the more complex biological index. Further work will
include evaluation of other index components, such as % of particular taxonomic groups and %
intolerants. We will also examine various options for interpreting scores for the metrics and
indices to increase their usefulness.
65
-------
Table 2-14. Benthic Macroinvertebrate Metric Scores for 19 Indicator Lakes
Lake
Small Warm
ABC
BL
FR
Tl
MA
Large Warm
GRC
TO
FA
JO
KE
WE
Small Cold
UPC
HU
RU
TE
Large Cold
DEC
HA
NE
NU
Taxa
Richness3
15 •
41
8 •
13 •
3 •
24 •
67
29 •
24 •
20 •
18 •
43
16 •
25*
12 •
18*
13*
25
15*
Biotic
lndexb
23
31
48 •
38*
49 •
25
33
30
26
25
36
23
36*
29
37*
19
23
28*
29 *
Mean Number
of Individuals
per Taxonb
2.5
23.8 •
13.4 •
8.5 •
6.0 •
4.9
19.2 •
8.6 •
3.5
26.3 •
54.3 •
9.4
13.4*
7.3
7.2
3.0*
1.9
8.6 •
9.7 •
% Individuals
as Dominant
Taxonb
16
54 •
49 •
28*
56 •
16
59 •
32 •
14
59 •
80 •
30
36 *
21
28
26 •
12
41 •
46 •
* = 0.5-0.75X the maximum for a lake class; • = < 0.5X the lake class maximum.
* = 1.5-2X the minimum for a lake class; • = > 2X the lake class minimum.
Underlined lakes are minimally disturbed reference lakes.
Fish. Five metrics were examined for this assemblage: total and native species richness, %
generally intolerant species, % omnivores, and % piscivores. All five metrics were suggested by
Fausch et al. (1990) as generally appropriate indicators and they have been used successfully in
large rivers (Hughes and Gammon, 1987; Oberdorff and Hughes, 1992).
Fish species richness and number of native species were directly related to lake area (Figures
2-14 and 2-15), as predicted from the theory of island biogeography (MacArthur and Wilson,
66
-------
20
15
UJ
O
ill
Q_
(/)
U.
O
£ 10
CO
s
D
Z
O
t Small Warm
+ Large Warm
* Small Cold
• Large Cold
Tl
#TE
*yp
DE
HA
WE
+
NU
FA
KE
'MA
+ GR
+ JO
+TO
• FR
•AB
BL • * HU
NE
0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8
LogiQLAKEAREA(ha)
Figure 2-14. Total fish species richness versus lake area. Underlined lakes are reference
lakes.
67
-------
LL
O
CC
LLJ
CO
20
o 15
LLJ
QL
if)
LU
10
5 5
0
• Small Warm
+ Large Warm
* Small Cold
• Large Cold
DE
HA
NU
uFA
I r
**MA
KE +
JCW
WE
GR
BL
NE
*UP
RU
*•
FR
0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8
LoglQ LAKE AREA (ha)
Figure 2-15. Species richness of native fishes versus lake area. Underlined lakes are
reference lakes.
68
-------
1967; Barbour and Brown, 1974), in which lakes function as islands in the terrestrial landscape.
NU, HA, and NE had substantially fewer species than DE (large cold lakes). Among small cold
lakes, UP, HU, and RU had fewer species than TE. AB, FR, and BL supported markedly fewer
species than Tl (small warm lakes). GR, TO, and JO held fewer fish species than expected (large
warm lakes). The species richness metric detected 9 of the 15 disturbed lakes, but only 1 of the
4 minimally disturbed lakes. The number of native species metric performed no better. HA, NE,
and NU had fewer natives than DE (large cold lakes); UP and RU had fewer native species than
TE (small cold lakes); AB. BL, FR, and MA supported fewer native fish species than Tl (small
warm lakes); and GR. JO, and TO held fewer natives than FA (large warm lakes). The number of
native species metric distinguished 10 of 15 disturbed lakes but only 1 of 4 reference lakes.
These results differ from those found in large rivers, where species richness (Gammon, 1991;
Oberdorff and Hughes, 1992) and native species richness (Hughes and Gammon, 1987) were
sensitive to disturbance.
Fish species tolerances were determined largely from descriptions in Scott and Grossman (1973).
The % of fish species intolerant of warming, turbidity, and low DO was substantially lower in AB.
Tl, FR, and MA than in BL (small warm lakes) (Table 2-15). In large warm lakes, the % of
intolerants was markedly lower in KE, TO, GR. JO, and WE than in FA. In the small coldwater
lakes, HU, RU, and TE had a substantially lower % of intolerant species than UP. NE had higher
% intolerant scores than DE and NU (large cold lakes). The % of intolerant fish species dis-
tinguished 11 of 15 disturbed lakes, but only 1 of 4 reference lakes. Hughes and Gammon (1987)
and Oberdorff and Hughes (1992) found the number of intolerant fish species much more
sensitive than this to chemical and physical habitat disturbances in large rivers.
Trophic guilds of fishes were determined largely from species descriptions in Scott and Grossman
(1973). The % of omnivorous species was substantially higher in BL, FR, Tl, and MA than in AB
(small warm lakes), in GR. TO, JO, and KE than in WE (large warm lakes), in HU and TE than in
UP (small cold lakes), and in DE, HA, and NU compared to NE (large cold lakes). The % omni-
vores metric thus distinguished 11 of 15 disturbed lakes, but only 1 of 4 reference lakes.
The % of piscivorous fish species was markedly lower in BL and Tl than in AB (small warm lakes),
in HU, RU, and TE than in UP (small cold lakes), and in HA and NE than in NU (large cold lakes).
There were no marked differences in % piscivores among large warm lakes. The % piscivores
metric distinguished only 7 of 15 disturbed lakes, but all 4 reference lakes. Calculating trophic
guild percentages is admittedly less sensitive when species are used rather than individuals;
69
-------
Table 2-15. Fish Metric Scores for 19 Indicator Lakes
Lake
Small Warm
ABf
BL
FR
Tl
MA
Large Warm
GRf
TO
FA
JO
KE
WE
Small Cold
UPf
HU
RU
TE
Large Cold
DEf
HA
NE
NU
% Intolerant
Speciesa'b
0 •
50
0 •
9 •
0 •
11 *
10 *
17
13*
0 •
8 •
100
50 •
0 •
13 •
22 •
40
50
31 •
% Omnivorous
Species°id
0
20 •
33 •
28 •
22 •
22*
20*
17
25*
20*
13
0
17 •
0
25 •
16 •
25 •
0
14 •
% Piscivorous
Species3'6
50
20 •
50
27*
44
44
50
50
50
50
47
100
16 •
33 •
38 •
32
25 *
25*
36
a * = 0.5-0.75X the maximum for a lake class; • = < 0.5X the lake class maximum.
Intolerant fish: brook trout, fallfish, lake chub, lake trout, lake whitefish, pearl dace, slimy sculpin.
° * = 1.5-2X the minimum for a lake class; • = a 2X the lake class minimum.
Omnivores: northern redbelly dace, brown bullhead, yellow bullhead, common carp, white sucker, fathead minnow.
e Piscivores: brook trout, brown trout, lake trout, rainbow trout, black crappie, largemouth bass, smallmouth bass, yellow
perch, chain pickerel, burbot, white perch, American eel.
Underlined lakes are minimally disturbed reference lakes.
however, Hughes and Gammon (1987) and Oberdorff and Hughes (1992) found that the % of pis-
civorous and omnivorous individuals was also relatively insensitive to disturbance in large rivers.
Although the fish results are preliminary, it appears that all five metrics are roughly comparable in
their ability to distinguish the effects of watershed disturbance and fishery management practices.
70
-------
Karr et al. (1986) predicted for streams that an intolerance metric would be most effective at the
highest levels of biological integrity, but that trophic guilds would be sensitive from the middle to
high levels. We did not find either case true. Additional metric development will focus on habitat
guilds, reproduction metrics, and various options for interpreting metric scores in order to develop
a multimetric fish assemblage index. Improved indexing methods will also allow the use of
metrics based on % and number of individuals, as well as % and number of species.
Birds. Four bird assemblage metrics were examined through use of the pilot data: species rich-
ness, number of individuals generally tolerant and intolerant, and principal components analysis
factor 1 derived from an evaluation of 65 potential metrics. The number of species was correlated
with lake area (r2 = 0.744) and warmwater lakes supported more species than cold (Figure 2-16),
except for MA (small warm lakes) and WE (large warm lakes). There were no markedly lower
numbers of species among lakes in the small or large coldwater classes. Bird species richness
distinguished all 4 reference lakes, but only 2 of 15 disturbed lakes. In forested areas, an
increase in species number may actually indicate disturbance, because forest openings and
development make new habitats available to a collection of species preferring brushy, grassland,
or agricultural habitats.
Principal components analysis was used to investigate how the foraging, dietary, and migratory
guild metrics interacted. Factor 1 of a PCA accounted for 67% of total variance and was corre-
lated with % seed eating species (-0.90), % seed eating/omnivorous/ground gleaning species
(-0.90), % ground gleaning species (-0.89), % foliage gleaning species (0.78), % foliage gleaning
individuals (0.76), % insectivorous species (0.76), and % insectivorous individuals (0.70). Factor 1
scores were negatively correlated with riparian disturbance and showed great discriminating
power for all but the large coldwater lakes (Table 2-16). BL, Tl, FR, and MA had lower scores
than AB (small warm lake); TO, FA JO, KE, and WE had lower scores than GR (large warm lakes);
HA and NU had lower scores than NE (large cold lakes); and HU, TE, and RU had lower scores
than UP (small cold lakes). PCA factor 1 distinguished 14 of 15 disturbed sites and all 4
reference sites. Croonquist and Brooks (1991) also found avian guilds useful indicators of
anthropogenic impacts in riparian areas.
To develop another metric describing avian response to human impacts, we reviewed the litera-
ture and compiled a list of New England bird species that were either tolerant or intolerant of
human disturbance on their nesting territories. To ensure a clear signal, we avoided using
wetland dependent species and adaptable or ubiquitous species. Interior forest birds that were
71
-------
w
o
UJ
Q.
D
Z
65
6O
55
o 50
DC
HI
CD
45 •
4b
O 40
H
35
• BL
• Tl
tAB
*HU
• FR
*TE
*UP
*
RU
• MA
+ FA
+ GR
+ KE
+TO
+ JO
• NU
DE
NE
• HA
+ WE
• Small Warm
+ Large Warm
* Small Cold
• Large Cold
1OO 2OO 3OO 4OO
LAKE AREA (ha)
5OO
GOO
Figure 2-16. Bird species richness versus lake area. Underlined lakes are reference
lakes.
72
-------
Table 2-16. Principal Components Analysis Scores and Number of Tolerant and Intolerant
Birds at 19 Indicator Lakes
Lake
Small Warm
ABd
BL
FR
Tl
MA
Large Warm
GRd
TO
FA
JO
KE
WE
Small Cold
ypd
HU
RU
TE
Large Cold
DEd
HA
NE
NU
Factor 1a
0.4
0.0 •
-0.7 •
-0.8 •
-2.6 •
0.3
0.1 •
-0.2 •
-0.9 •
-1.1 •
-1.5 •
1.5
0.9*
0.7 •
0.6 •
0.8
0.6 *
0.9
0.2 •
Number
Tolerantb
50
97*
126 •
105 •
155 •
86 •
32
74 •
105 •
94 •
189 •
16
43 •
15
64 •
8
13 *
71 •
27 •
of Birds
Intolerant6
6 •
34
10 •
2 •
0 •
9 *
12
8*
6 •
3 •
0 •
50
25 •
12 •
4 •
44
10 •
15 •
25*
PCA factor 1 was positively correlated with foliage gleaning and insectivorous species and individuals; it was negatively
correlated with ground gleaning, seed eating, and omnivorous species. * = 0.5-0.75X the lake class maximum; • =
< 0.5X the lake class maximum.
Tolerant birds: American goldfinch, American robin, brownheaded cowbird, bobolink, bam swallow, chipping sparrow,
common grackle, chestnut-sided warbler, European starling, field sparrow, house finch, house sparrow, house wren,
magnolia warbler, Nashville warbler, northern mockingbird, rock dove, Savannah sparrow, white-throated sparrow,
yellow-throated vireo. * = 1.5-2X the lake class minimum; • = > 2X the lake class minimum.
Intolerant birds: blackburnian warbler, boreal chickadee, black-throated green warbler, Cape May warbler, common
loon, evening grosbeak, golden-crowned kinglet, gray jay, hairy woodpecker, hermit thrush, pine siskin, pileated
woodpecker, ruby-crowned kinglet, solitary vireo. * = 0.5-0.75X the lake class maximum; • = < 0.5X the
lake class maximum.
Underlined lakes are minimally disturbed reference lakes.
73
-------
intolerant of habitat changes and serai birds preferring clearcut or developed habitats were
identified in the database.
Birds, unlike fish, could be sampled quantitatively and with one consistent method; thus, we
could use data based on numbers of individuals in place of numbers of species. The numbers of
individuals/lake that were generally intolerant of physical and biological disturbance, or forest
interior birds common to old growth forest, were lower than expected for AB. Tl, FR and MA
compared with BL (small warm lakes), for KE, FA, GR. JO, and WE compared with TO (large
warm lakes), for HU, RU, and TE compared with UP (small cold lakes), and for HA, NE, and NU
compared with DE (large cold lakes) (Table 2-16). The number of tolerant birds was higher than
expected at BL, Tl, FR, and MA than at AB (small warm lakes), at KE, GR, FA, JO, and WE than at
TO (large warm lakes), at HU and TE than at RU (small cold lakes), and at HA, NU, and NE than
at DE (large cold lakes). Number of intolerant individuals distinguished 2 of 4 reference lakes and
13 of 15 disturbed lakes, whereas number of tolerants distinguished 3 reference and 13 disturbed
lakes.
Bird tolerance and ordination (foraging guilds) metrics better distinguished lake disturbance and
reference sites than did species richness. Evidently, knowledge of species' habitat and foraging
requirements is more useful than simple taxonomic identity. Patrick (1972) also described several
stream assemblages with consistent species richness, but variable species composition, through
time. Future work on the bird data will focus on evaluating the PCA factor 1 component metrics
separately and on developing an assemblage index composed of trophic, habitat, and species
richness metrics.
Physical Habitat Structure. We examined the relationships between the physical habitat struc-
ture and vertebrate assemblages. Data on canopy complexity and extent and intensity of anthro-
pogenic shoreline disturbances were aggregated into a lake ranking of 1 to 19 and plotted
against the % of intolerant bird species (Figure 2-17a). A similar plot was developed for fish,
including fish cover complexity in the habitat ranking, and substituting the % native fish species
for birds (Figure 2-17b). Both plots reveal a direct relationship between habitat quality and the
metric ratios. We propose to further analyze habitat/exposure data in order to evaluate response
indicator metrics, to develop diagnostic models for explaining unacceptable condition, and to
distinguish probable cause as chemical, physical, or both, as suggested by Plafkin et al. (1989).
74
-------
20
18 -J
16
1 14
12
10
CD
CL
-o
S
03
_o5 8 •
o
5 6
o^
4 -
2 -
0 -
-*- up
DE
(a)
* HU
* RU
HA
BL
NE
NU
* TE
•TO
•+-QR * JO
•+• FA
• Small Warm
+ Large Warm
* Small Cold
• Large Cold
KE
FR
MA WE
0
4 6 8 10 12
Disturbance Rank
14
16
18
20
100 -
90 -
80 -
CD 70 -
0
CD
fn 60 •
I 50-
CD
I 4°-
-5 30 -
20 -
10 -
0 -
UP * • BL * HU
• HA
+T0 • AS
-i- GR
•4- KE
* TE
* RU
• FR
(b)
ME + JO
+ FA
+ WE * MA
• Small Warm
+ Large Warm
* Small Cold
• Large Cold
8 10 12
Disturbance Rank
14
16
18
20
Figure 2-17. Relationship between quality of physical habitat structure and (a) % intolerant
bird species, and (b) % native fish species. Physical habitat quality was
ranked 1 through 19 by level of canopy complexity and extent and intensity of
shoreline disturbance in (a) and by all three plus fish cover complexity in (b).
Underlined lakes are reference lakes.
75
-------
2.4.5.2 Metric and Assemblage Comparisons from Indicator Lake Results
The relative scores for the reported metrics of the biological response indicators are summarized
in Table 2-17. Metric scores were rated as explained in Section 2.4.4 and given a black dot (•) if
unacceptable and an asterisk (*) if marginal. The dots and asterisks were tallied by metric and
assemblage to evaluate the responsiveness of the candidate metrics and assemblages to catch-
ment perturbation. When metrics indicated different perturbation of the reference and disturbed
lakes than predicted, it may have been due to erroneous classification of lake type or disturbance
level, to inappropriate response indicator metrics or score interpretation, or to the fact that no
single metric is equally sensitive to all perturbations. The last explanation has been reported
elsewhere (Angermeier and Karr, 1986; Karr et al., 1987; Oberdorff and Hughes, 1992; Figure 2-9)
and justifies using a multimetric index. The misclassification of disturbance level and lake type
are discussed in Section 2.5.2, along with the ability of the combined set of metrics to assess
predicted lake condition. Our objective in this section of the report is to evaluate individual
metrics and assemblages for their responsiveness.
None of the metrics or assemblages consistently scored the minimally disturbed and more highly
disturbed lakes, but some were affected more by the disturbances than others. Nine metrics
(diatom DCA change, diatom disturbance index change, diatom richness change, benthos rich-
ness, benthos individuals/taxon, benthos % as dominant taxon, bird PCA factor 1 scores, and bird
number of tolerant and intolerant individuals) scored as expected in 14-18 of the 19 lakes, sug-
gesting that they could potentially detect the effects of such disturbances in other lakes at least
75% of the time. The diatom enrichment metric, all three zooplankton metrics, and the fish and
bird richness metrics all scored as expected in only 10 or fewer cases out of 19, indicating a
success rate of approximately 50%, too low to be useful for detecting the effects of such distur-
bances as studied in this pilot. This result suggests that when used alone, without further inter-
pretation, these metrics are unlikely to respond to land use and fish stocking activities as much as
to other habitat or exposure conditions. It also suggests that they should be used with care when
combined with other metrics. Clearly, additional research is needed to further develop and refine
metrics and their scoring interpretation.
We assessed assemblage responsiveness by examining how well the set of metrics for an
assemblage distinguished disturbed from reference lakes (Table 2-17). We carried out this pro-
cedure for disturbed lakes by tallying the lakes at which half the metrics or more for an
assemblage showed the effects of disturbance. For reference lakes, we tallied the lakes at which
76
-------
ra
UJ
o
'•5
c
s!
o
o
C/)
T3
C
03
o"
0)
D)
TO
3
,
"co
H
j Indicator
CM
CM
I Diatoms
*
•
•
•
•
*
•
•
•
*
o
rr
0)
N
w
«
*
•
•
«
«
-«
«
r^
n
w
jc
c
o
'si
a
o
H
CM
CO
|| Benthos
*
*
-*
•
«
•
•
•
•
•
•
•
•
•
•
CM
'~
|| Richness
*
*
«
«
•
*
•
r-
'i-
5
•
•
*
«
•
•
•
•
•
•
•
-
CO
I] Individuals/taxon
•
•
•
*
•
•
•
•
•
-K
•
-
CO
% as dominant
|| taxon
CM
CM
«
il
it
•
«
•
•
•
#
#
«
«
•
O)
T~
|| Richness
«
•
«
•
•X
#
«
•K
#
«
•
*
0
*-
|| Natives
*
•
•
•
•
•
•
«
*
*
•
•
•
•
^_
'-
|| % Intolerant
•
•
•
•
•
*
-1C
-K
#
•
•
•
-
CM
|| % Omnivore
•
•
•
•
•
«
r~
^i-
|| % Piscivore
•>r
CO
If
•a
15
«
•k
CM
1-
Richness
•
*
•
•
*
•
•
•
•
•
•
•
•
<*
i-
<
O
Q.
*
•
•
•
•
•
•
•
•
*
*
•
•
•
•
CO
CM
# Intolerant
•
•
«
•
•
•
•
•
•
•
•
•
•
CO
CO
1 # Tolerant
CM
-
O
CD
£
O
C\J
(O
£
"fr
CM
O
O)
Cl
CD
in
in
CO
CO
1
a
E
5
o
D
S
2
<
5
2
D
<
<
5
<
<
<
<
13
5
2
<
1 Acceptability
-
"
D »T3
77
-------
fewer than half the metrics showed the effects of disturbance. Based on this method of interpre-
tation, birds, benthos, fish, and diatom assemblages distinguished reference from disturbed lakes
in 14-17 of the cases, or at least 75% of the time. The zooplankton assemblage distinguished the
lakes only 42% of the time. This difference in sensitivity among assemblages may have resulted
from the relative level of metric development, the method by which the assemblages were evalu-
ated, or the types of lake disturbance. Further metric and index development is expected to
improve the sensitivity. Development of a multimetric index for each assemblage may improve
score interpretation. Also, evaluation of different types of disturbance in future pilots may reveal
different assemblage sensitivities.
2.4.5.3 Indicator Evaluation Conclusions
These results indicate the need for five types of additional research: (1) Continued metric devel-
opment is needed for all five assemblages to complete the evaluation of listed metrics and to
suggest and evaluate additional metrics. (2) Additional reference lake selection and study is
necessary to examine the amount of assemblage variation expected in minimally disturbed condi-
tions. This information is needed both for indicator development and for defining acceptable lake
conditions. (3) Disturbance gradients and lake classes more likely to affect zooplankton should
be evaluated to test their sensitivity, because the 1991 disturbances may have been biased
toward the other assemblages. (4) It is also probable that apparent disturbance levels of the
indicator lakes may differ from actual disturbance and that there are multiple disturbances. We
plan additional data assessments this year to examine these possibilities. (5) Candidate metrics
must be evaluated for indexing variance, which is a combination of crew, measurement, index
period, and index station variances. This was the major focus of the 1992 pilot, which will facil-
itate evaluation of among-lake versus within-lake variance for diatoms, zooplankton, fish, fish
tissue contamination, and habitat structure. Indexing variance for benthos and birds will be
conducted in the 1993 demonstration study.
2.5 DISTINGUISHING ACCEPTABLE (NOMINAL) FROM UNACCEPTABLE (SUBNOMINAL)
CONDITIONS
As mentioned earlier, EMAP's primary objective is to describe the condition (as measured with the
selected indicators) and trends in condition of ecological resources. However, the information will
be useful to a wider audience if we can set criteria to distinguish between acceptable (nominal)
and unacceptable (subnominal) conditions for particular lake or stream types within selected
78
-------
geographic locations. Hunsaker and Carpenter (1990) and Messer et at. (1991) outline the issue.
Messer et al. state: "Operational criteria must be developed for each response indicator to
identify the transition from acceptable or desirable (nominal) to unacceptable or undesirable
(subnominal) condition...Criteria could be based on attainable conditions under 'best manage-
ment practices' as observed at regional reference sites..., or on theoretical grounds or manage-
ment targets." Clearly, this is not strictly a technical issue. We emphasize that our objective in
approaching this issue is not to make this decision about thresholds independent of others, but
rather to contribute to the process of deciding how to resolve these issues from a technical
perspective. We are not advocating any particular management decision or action or societal
decision. We do believe, however, that it is important to develop as sound a scientific basis as
possible for such decisions.
This section describes options for assessing acceptable and unacceptable condition,
demonstrates how reference lakes could be used for such purposes, and discusses some
concerns regarding that approach.
2.5.1 Describing the Reference Condition
We will report the condition of populations of surface waters by plotting cumulative distribution
functions (CDFs) and histograms from metric scores or from indices that combine multiple metrics
and assemblages. However, EMAP is also charged with estimating the proportion of waterbodies
that are in acceptable or unacceptable condition (Hunsaker and Carpenter, 1990; Messer et al.,
1991). This task requires some sort of benchmark or reference condition against which the
sample waterbodies can be compared. Several approaches have been suggested for estimating
reference condition: reference sites, pristine sites, historical data, paleoecological data,
experimental results, ecological models, empirical distributions, and expert consensus. We
describe the advantages and disadvantages of each in the following paragraphs.
Regional reference sites (Hughes et al., 1986; Karr et al., 1986) incorporate historical information,
assemblage-habitat knowledge, and expert consensus in selecting waters of major waterbody
types that are minimally disturbed by humans. Allan et al. (1992) called for increased monitoring
and preservation of minimally disturbed waters in order to distinguish various sources of varia-
bility, particularly the effects of climate change. The assumption is that such lakes and streams
represent conditions that incorporate the natural limits of the region and that lower levels of
watershed, riparian, and in-lake disturbance support less impaired or healthier communities.
79
-------
Major disadvantages of reference site monitoring are the additional monitoring costs required, the
effort needed to locate minimally disturbed sites dispersed across a region, and questions of their
representativeness and acceptable level of disturbance in extensively disturbed regions. The
representativeness of reference sites for populations of diverse surface waters is also of concern.
Representativeness can be examined by using ordination analyses to test whether the habitat
indicators for grid and reference sites are similar when both have similar response indicator
values. Such comparisons are more problematic for the many sites at which response indicator
values differ from those of reference sites. In such cases, habitat indicator values plotted against
response indicator values for sites in the same ecoregion are useful (Plafkin et al., 1989). Despite
these shortcomings, reference sites remain EMAP's primary method for estimating reference
conditions (Paulsen et al., 1991).
Pristine sites are a special case of reference site with the same advantages and disadvantages,
except they are presumably undisturbed. However, they do not exist in most regions and are not
representative of waters in markedly different regions. Given the global extent of atmospheric
pollution, the prospects of continued global climate change (Ehrlich et al., 1977), and the preva-
lence of fish stocking, there may be no truly pristine sites.
Historical data offer a perspective of species once present across a region. This approach
requires no additional sampling and can occasionally provide data on waters at the time of
settlement. Such data are available from some natural history museums or the grey literature.
There are several problems with historical data. For many regions of the country and many
assemblages, the sites selected are sparse and the data are incomplete or were collected with
markedly different methods and during various seasons. Information about the level of distur-
bance at the collection sites is rare. Typically, historical data sets offer little information on
species abundances and many collections focus on particular species, rather than entire assem-
blages. Occasionally, museum curators are reluctant to release such data and the grey literature
is particularly troublesome to locate and examine. Although EMAP may not be able to use such
data everywhere to determine unacceptable condition, historical data will be very useful in range
finding and classification, and should be acquired and analyzed wherever possible.
Paleoecological information preserved in lake sediments is another form of historical data. If
cores are long enough, they offer estimates of presettlement or preindustrial water quality condi-
tions (Gumming et al., 1992). We propose to use sedimentary diatoms for this purpose. A key
assumption in this approach is that historical sediments represent the desired or equilibrium state,
80
-------
but additional retrospective research, in addition to studying diatom assemblages preserved at
the top and bottom of the sediment cores, is required to determine this.
Experimental results from laboratory studies are useful for assessing toxic levels and mechan-
isms; however, they are unreliable for determining reference conditions for entire communities.
The strength of experimental tests is in reducing stressor and response variables, but this limits
their applicability in systems composed of multiple stressors, multiple assemblages with indi-
viduals of multiple ages, and various types of habitats.
Models developed from historical, paleoecological, experimental, and EMAP-Surface Waters grid
data could be very useful for developing standards for determining acceptability. Such models
should be increasingly useful as EMAP databases increase in size. However, models developed
from incomplete data or from data collected at disturbed sites can yield only incomplete or
undesirably low acceptability standards.
Empirical distributions of EMAP results may show substantial discontinuities, or we could simply
select an index score arbitrarily. This approach would apply just the data from probability sites
and is least expensive, but its arbitrary nature is politically troublesome; also, discontinuities may
not appear or they may simply indicate different lake and stream types. There is a high proba-
bility that even the highest quality probability sites may be impaired where diffuse pollution is
extensive (e.g., in intensively farmed regions such as the Corn Belt, Central California, or the
Mississippi Alluvial Plain).
Expert panels could be formed to seek consensus on index scores that would be acceptable or
unacceptable. Such reference conditions would be the product of their professional expertise
and regional experience, the quality of the available data, and personal biases. Panelists would
require information on reference condition from the other approaches. A similar process has
recently been used to estimate the effects of varying forest harvest levels on threatened and
endangered species of Pacific Northwest forests.
Regardless of how reference data are acquired, there will be a range of scores, so we will be left
with the dilemma of determining the indicator score at which a waterbody is acceptable or
unacceptable. Box plots of reference versus grid data and nonparametric tests such as the
Wilcoxon rank test can be used to test for significant differences from reference conditions. We
could also set an arbitrary percentile of the reference data. For example, Ohio EPA (1988) uses
81
-------
the 25th percentile, focusing attention on the most impaired waters, while Plafkin et al. (1989)
delimit impairment at the 80th percentile, where waters are slightly impaired. Hughes and Noss
(1992) demonstrate how these differences may result in stream impairment rates of 64% versus
95% in highly disturbed regions. Because EMAP is also a trends monitoring program, we can
also simply track trends in a direction we consider undesirable (e.g., loss of species, increase of
pest species, increased anomalies).
2.5.2 Pilot Results
A reference lake or gradient approach to comparing assemblage (and the derivative metric)
responsiveness is based on the assumptions that (1) lakes most lacking human disturbances
represent systems with greater biological integrity than those with moderate or substantial
disturbances and (2) subtle natural differences in lake morphology and chemistry among lake
classes in an ecoregion are of less consequence than the level of human disturbance. A similar
approach has been successfully used for streams (Hughes et al., 1986; Karr et al., 1986; Ohio
EPA, 1988; Plafkin et al., 1989). We also chose to compare assemblage and metric responsive-
ness along a disturbance gradient, to determine how this process might work for lakes. We
examined the results to see if all metrics together distinguished between lakes with disturbed and
undisturbed catchments. We emphasize that this is one process by which we intend to develop
and select indicators, and that we will need much more information than that resulting from this
pilot before we can do so. The stressor gradient and reference lake design enabled us to
(1) examine the process for determining acceptable condition from the combined set of assem-
blages and metrics and (2) evaluate difficulties in selecting reference lakes.
The acceptable/unacceptable status of the indicator lakes was evaluated by adding the metric
scores from Table 2-17 for each lake (dot = 1, asterisk = 0.5) and ranking a lake as unaccep-
table or marginal if its tally was > 2X or 1.5-2X that of the lake with the lowest total in the same
lake class, respectively. These criteria represent the uncertainty in assessment given the single
reference lake in each class, and were selected as being tallies that several workshop participants
agreed were substantially different from expected values (Thornton, pers. comm.). Given these
criteria, lakes with substantially impaired biological integrity are MA, NU, and HU. Marginal lakes
are BL, FR, Tl, KE, RU, TE, HA, and NE. The only lakes in acceptable condition are AS, GR, TO,
FA, JO, WE, UP, and DE. These lake evaluations demonstrate the manner by which a set of care-
fully selected reference lakes can be used to determine whether or not a probability lake is in
82
-------
acceptable condition. However, a lake that is unacceptable or marginal may be so because of
natural events or anthropogenic activities.
The lake tallies for entire communities, if not individual assemblages, generally confirmed the
state biologists' preclassifications of lake disturbance and the land use databases. Reference
lakes had low sums, highly disturbed lakes in a lake class had high sums, and moderately dis-
turbed lakes usually had intermediate values. Exceptions were KE, WE, and HU. We also had
difficulty selecting a large warm reference lake that was truly minimally disturbed, as shown by the
high level of shoreline disturbance at GR (Table 2-7). This problem indicates the need to select
several reference lakes, both to obtain a measure of reference lake variance and to be able to
reject candidate reference lakes disturbed to such a degree that they are unsuitable. In eco-
regions with high levels of human disturbance, the search effort for reference lakes must be
increased. In addition, we must develop a method of ranking various catchment and riparian
disturbances differently and obtain more accurate land use data. Clearly, before reference lakes
are finalized, they must be more completely evaluated than was the case for this pilot.
The pilot indicates a process for using preliminary sets of reference lakes and assemblages to
assess whether similar types of lakes are in acceptable or unacceptable condition. However, lake
types other than those evaluated in this pilot may need examination. For example, EMAP will
sample shallow 1-5 ha ponds, for which other metrics may be needed, and some lakes will not
fall neatly into distinct cold/warm lake classes. For example, three warmwater lakes (FA, JO, and
FR) supported a small coldwater zooplankton fauna, but apparently their water was not cold and
oxygenated enough for trout reproduction. We might also develop lake classes from their com-
munity characteristics, as Tonn et al. (1983) and Schneider (1981) did for fish assemblages. This
would have to be done for the entire community, not a single assemblage; it is, therefore, unlikely
to be accomplished without considerably more data and interpretation. Finally, numerous refer-
ence lakes are needed for confident estimates of the reference condition for all the grid lakes and
lake classes examined in the northeastern United States. Walters et al. (1988) showed that for
several management questions, agencies need one reference site for every four to five modified
sites to distinguish spatial and temporal trends. This would mean sampling 10-15 reference sites
annually in the northeastern United States. During the 4-year EMAP rotation, the 40-60 carefully
selected reference lakes sampled could equate to 8-20 for 3-5 ecoregions or 8-12 for 5 lake
classes. Of course, some of the reference lakes will come from the probability sample, as long as
their catchments are minimally disturbed. Preliminary estimates suggest that about half can be
located this way.
83
-------
In summary, we found that we could locate minimally disturbed lakes to serve as references for
estimating acceptable and unacceptable condition. The process requires numerous conversa-
tions with local biologists, examination of remote sensing data, and site visits. Minimally
disturbed warm water lakes will require more effort to locate than cold water lakes because of the
greater rarity of the former in the Northeast.
2.6 SUMMARY AND CONCLUSIONS
2.6.1 Indexing
The indicator pilot fulfilled several research objectives. We were successful in developing an
indexing protocol in evaluating sampling gear for all assemblages except benthos. We deter-
mined that indexing protocols based on a systematic sample of the littoral and riparian zones
were appropriate for fish, birds, and physical habitat structure. No single gear type or index
location was adequate for sampling fish; sampling was stratified by littoral and pelagic habitat,
and different types of gear were needed in each. Benthos indexing methods require further data
analyses and pilot studies to determine number and locations of samples and appropriate gear.
Additional research is also needed on the effectiveness of lightweight electrofishers.
2.6.2 Indicator Development
Preliminary metrics enabled us to assess the effects of four different types of common distur-
bance (silviculture, agriculture, residential development, fish stocking) along an intensity gradient.
Several metrics were responsive to the above disturbances, particularly changes in diatom
species richness, a diatom disturbance index, diatom DCA change, benthos richness, two
benthos redundancy measures, the bird PCA factor 1 score and the numbers of tolerant and intol-
erant birds. When multiple metrics of an entire assemblage were assessed, birds, benthos, fish,
and diatoms were most affected by the disturbances studied. Bird data were the most quickly
available for analysis, and bird and fish assemblages were responsive to changes in physical
habitat structure.
2.6.3 Acceptable/Unacceptable Condition
The aggregated metrics appeared sensitive to the disturbance gradients of agriculture, residential
development, fish stocking rates, and silviculture. In other words, they were able to distinguish
acceptable, marginal, and unacceptable conditions for four lake classes and types of disturbance
84
-------
through use of a single reference lake in each class. The assemblages in the large warmwater
reference lake appeared more affected by disturbances than those in another lake in the same
class with greater catchment disturbance. This suggests a need to improve the accuracy of the
landscape database or the response indicators or both.
2.6.4 Concluding Remarks
The process of selecting, developing, evaluating, and rejecting indicators is a continuum with
numerous iterations and few clear breaks (Figure 2-8). We plan to determine the most appropri-
ate assemblages, given the EMAP objectives, and to develop indicators from measurements on
those assemblages. These assemblage indicators in turn must be integrated into a multi-
assemblage or community assessment, which requires indicators to be complementary, without
obvious contradiction, and to have minimal redundancy. However, some redundancy is useful in
diagnosing cause and effect relationships when the changes are subtle. In the case of RU, for
example, fish, benthos, zooplankton, and diatoms all indicated acidification, but any single
assemblage would have been unconvincing. Insufficient biological evidence of impact to counter
apparently acceptable chemical concentrations is a concern for low ANC lakes, because midlake
summer chemistry underestimates acidification pulses that occur in the littoral area during spring
snowmelt (Eshleman, 1988; Gubala et a!., 1991).
The analyses described in this chapter focused on calibrating metric response to known distur-
bance gradients, but EMAP must also be able to assess unknown disturbances. To do this, we
must know more about the basic nature of assemblage change, particularly changes in species
richness, guilds, and processes that indicate fundamental shifts in assemblage integrity. We
should also be tracking taxa likely to be sensitive to climate change and continued land use
changes. We must develop a clearer understanding of what we believe is unacceptable change
or else set an arbitrary index score. This task requires identifying the most important changes in
biological integrity, trophic condition, and fishability about which enlightened citizens have
concerns. In addition, our reports must clearly state that unacceptable lake integrity does not
mean that such lakes are biological wastelands, they are simply unacceptable deviations from
reference conditions.
Further metric development is necessary, particularly for zooplankton. A promising area for
zooplankton is trophic complexity. Benthos metric development will focus on score interpretation.
Bird and fish metric development will continue to focus on various guilds, and we are analyzing
three large fish databases to further estimate sampling variability.
85
-------
Further evaluation of six essential criteria for indicator selection (responsiveness, high signal/
noise, high information/cost, high societal value, high among-site/within-site variability, low
among-year variability) will be necessary in the Northeast and elsewhere before we decide which
assemblages are best for national monitoring. Indicator assemblages will be evaluated against
these criteria for several years in the northeastern United States before any final decisions are
made for that region. Additional pilot studies in other regions are needed before the final set of
assemblages is selected for nationwide monitoring.
Future demonstration projects will expand our determination of variance components (Chapter 4).
Reasonable estimates of the year component require about 5 years of sampling. These variance
studies will also provide data for quantifying signal/noise and cost/information. Cost estimates will
be refined once the indexing protocols for all assemblages are decided, probably in 1994, after
the results of the methodological research for benthos and fish are available.
Another critical aspect of indicator development involves coordination with other federal and state
agencies. The U.S. Geological Survey and U.S. Fish and Wildlife Service are both developing
new national monitoring initiatives. The interest of the U.S. Forest Service, Bureau of Land
Management, Bureau of Reclamation, and Army Corps of Engineers in monitoring ecological con-
ditions continues to expand. Each agency has different information needs that sound similar
when expressed at a general level. State agencies are required by the U.S. EPA to develop
biological criteria based on aquatic assemblages for surface waters. Although each program has
a different set of objectives, and consequently a different design, it would be valuable to each
agency if a common set of core indicators and sampling methods were adopted. An intergovern-
mental task force on water quality monitoring is currently convening to promote just such
coordination.
86
-------
APPENDIX 2A
ANALYTES TO BE MEASURED IN FISH TISSUE
Analyte (CAS Number)
Detection Limits (ppm)
Aldrin (309-00-2)
Aluminum (7429-90-5)
Arsenic (7440-38-2)
Cadmium (7440-43-9)
Chlordane-cis (5103-71-9)
Chromium (7440-47-3)
Copper (7440-50-8)
2,4'-ODD (53-19-0)
4.4'-DDD (72-54-8)
2,4'-DDE (3424-82-6)
4,4'-DDE (72-55-9)
2,4'-DDT (789-02-6)
4,4'-DDT (50-29-3)
Dieldrin (60-57-1)
Endrin (72-20-8)
Heptachlor (76-44-8)
Heptachlor Epoxide (1024-57-3)
Hexachlorobenzene (118-74-1)
Hexachlorocyclohexane [Gamma-bhc/Lindane] (58-89-9)
Lead (7439-92-1)
Mercury (7439-97-6)
Mirex (2385-85-5)
Nickel (7440-02-0)
trans-Nonachlor (3765-80-5)
PCB Isomers
2,4-Dichlorobiphenyl (34883-43-7)
2,2',5-Trichlorobiphenyl (37680-65-2)
2,4,4'-Trichlorobiphenyl (7012-37-5)
2,2',5,5'-Tetrachlorobiphenyl (35693-99-3)
2,2',3,5'-Tetrachlorobiphenyl
2,3',4,4'-Tetrachlorobiphenyl
2,2',4,5,5'-Pentachlorobiphenyl (37680-73-2)
2,3',4,4',5-PentachIorobiphenyl (31508-00-6)
2,2',4.4',5,5'-HexachlorobiphenyI (35065-27-1)
2,3,3',4,4'-Pentachlorobiphenyl
2,2',3,4,4',5-Hexachlorobiphenyl (35065-28-2)
2,2',3,4',5,5',6-Heptachlorobiphenyl (52663-68-0)
2,2',3,3',4,4'-Hexachlorobiphenyl (38380-07-3)
2,2',3,4,4',5,5'-Heptachlorobiphenyl (35065-29-3)
2,2',3,3',4,4',5-Heptachlorobiphenyl (35065-30-6)
2,2',3,3',4,4',5,6-Octachlorobiphenyl (52663-78-2)
2,2',3.3',4,4'.5,5',6-Nonachlorobiphenyl (40186-72-9)
Decachlorobiphenyl (2051-24-3)
Silver (7440-22-4)
Tin (7440-31-5)
Zinc (7440-66-6)
0.00025
10.0
2.0
0.2
0.00025
0.1
5.0
0.00025
0.00025
0.00025
0.00025
0.00025
0.00025
0.00025
0.00025
0.00025
0.00025
0.00025
0.00025
0.1
0.01
0.00025
0.5
0.00025
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.01
0.05
50.0
87
-------
APPENDIX 2B
DIATOM, ZOOPLANKTON, BENTHOS, FISH, BIRD AND SEDIMENT TOXICITY METHODS
Variable
Diatoms
Zooplankton
Benthos
Fish
Birds
Sediment toxic ity
Method
K-B corer, 20-50-cm core 210Pb dating of top, bottom,
and 10 and 20 cm depths in eutrophic lakes. Identify
500 valves to species at 1 250 X magnification.
Paired 48-fim and 202-^m mesh nets. Vertical tow from
1/2 m off bottom of deepest site to surface at 5-6 s/m.
Samples anesthetized and preserved, and both macro-
and microzooplankton identified to species.
Petite PONAR grab, K-B corer in profundal sites. 595 ^m
sieve to remove sediments from grab samples. Corer
sediments will be cleared by digestion. Sweep nets at
littoral sites. Specimens identified to genus or species.
Experimental gill nets and eel pots in pelagic and
profundal zones. Indiana trap nets, boat electrofishing,
beach and short seines, minnow traps, and eel pots in
littoral zone. Passive gear fished overnight and active
gear fished after sunset. Specimens identified to species
and size class and examined for anomalies. Voucher
specimens sent to Harvard Museum of Comparative
Zoology for confirmation.
Circular plot count with 100-m radius 10 m from shore.
20-24 plots. Identified to species visually and aurally.
Petite PONAR Grab. 7-D Hyallela azteca test using 20
animals 2-3 days old with 100 mL sediment and 400 mL
reconstituted water with daily replacement. Measure
survival and dry weight.
Reference
Glew, 1989;
Dixitetal., 1992;
Smol and Glew, 1992.
McCauley, 1984.
Weber, 1973.
Klemm et al., 1990
Nielsen and Johnson, 1983;
Hocuttand Stauffer, 1980;
Kolz and Reynolds, 1991.
Reynolds et al., 1980;
Moors and O'Connor, 1992
Klemm and Lazorchak, 1992
88
-------
APPENDIX 2C
PHYSICAL HABITAT STRUCTURE FIELD METHODS
Variable
Method
Reference
Canopy layer
Mid layer
Ground cover
Shore substrate
Human influence
Bank type
Fish cover
Aquatic Macrophytes
Bottom substrate
General observations
(sediment color, odor,
surface scum, mats)
From boats anchored 10 m offshore, crews made
observations of riparian vegetation structure and human
disturbances on plots 15 m wide extending 15 m back
from the shore. Littoral measurements or observations of
shoreline substrate, bottom substrate, near-shore fish
cover, and human disturbances were made along the 15-
m shoreline plots and in the water between the shore and
the boat. Field forms standardized most observations as
absent, present (< 10% areal cover), subdominant
(10-40% cover), or dominant (> 40% cover).
Tallent-Halsell and Merritt,
1991
89
-------
APPENDIX 2D
WATER QUALITY METHODS
Variable
Depth
Temperature, in situ
Dissolved oxygen, in situ
Transparency
pH, closed system
pH, equilibrated
Acid Neutralizing
Capacity (ANC)
Carbon, dissolved
inorganic (DIG), closed
system
Carbon, dissolved
organic (DOC)
Conductivity
Aluminum, total dissolved
Aluminum, monomeric
Calcium
Magnesium
Potassium
Sodium
Ammonium
Chloride
Nitrate
Sulfate
Phosphorus, total
Nitrogen, total
Silica, dissolved
Turbidity
Total Suspended Solids
True Color
Chlorophyll-a
Summary of Method
Sonar measurement or calibrated line.
Measured at various depths using thermistor probe.
Measured at various depths using membrane electrode and meter.
20-cm diameter black and white Secchi disk with calibrated line;
estimated as depth of disappearance plus depth of reappearance,
divided by 2.
Sample collected and analyzed without exposure to atmosphere;
electrometric determination (pH meter and glass combination
electrode).
Equilibration with 300 ppmV CO2 for 1 hr prior to analysis;
electrometric determination (pH meter and glass combination
electrode).
Acidim,etric titration to pH £ 3.5, with modified Gran plot
analysis.
Sample collected and analyzed without exposure to atmosphere;
acid-promoted oxidation to CO2 with detection by infrared spectro-
photometry.
UV-promoted persulfate oxidation, detection by infrared spectro-
photometry.
Electrolytic (conductance cell and meter).
Atomic absorption spectroscopy (graphite furnace).
Collection and analysis without exposure to atmosphere.
Colorimetric analysis (automated pyrocatechol violet).
Atomic absorption spectroscopy (flame).
Atomic absorption spectroscopy (flame).
Atomic absorption spectroscopy (flame).
Atomic absorption spectroscopy (flame).
Colorimetric (automated phenate).
Ion Chromatography.
Ion Chromatography.
Ion Chromatography.
Acid-persulfate digestion with automated Colorimetric determinaton
(molybdate blue).
Alkaline persulfate digestion with determination of nitrate by
cadmium reduction and determination of nitrite by automated
colorimetry (EDTA/sulfanilimide).
Automated Colorimetric (molybdate blue).
Nephelometric
Gravimetric
Visual comparison to calibrated glass color disks.
Filtration (glass fiber) in field; extraction of filter into acetone;
analysis byjspectrophotometry (trichromatic).
Reference
Chaloud etal., 1989
Chaloud etal., 1989
Chaloud etal., 1989
Lind, 1979;
Chaloud etal., 1989
U.S. EPA, 1987
U.S. EPA, 1987
U.S. EPA, 1987
U.S., EPA, 1987
U.S. EPA, 1987
U.S. EPA, 1987
U.S. EPA, 1987
APHA, 1989;
U.S. EPA, 1987
U.S. EPA, 1987
U.S. EPA, 1987
U.S. EPA, 1987
U.S. EPA, 1987
U.S. EPA, 1987
U.S. EPA, 1987
U.S. EPA, 1987
U.S. EPA, 1987
Skougstad etal., 1979
U.S. EPA, 1987
U.S. EPA, 1987
U.S. EPA, 1987
U.S. EPA, 1987
U.S. EPA, 1983
APHA, 1989
APHA, 1989
U.S. EPA, 1987
APHA, 1989
90
-------
CHAPTER 3
OVERVIEW OF SURVEY DESIGN AND LAKE SELECTION
This chapter presents the rationale for the overall EMAP design and its specific application to
lakes. Information the reader should derive from this chapter includes:
Rationale for the EMAP design strategy
Lake selection process (Tiers 1 and 2)
Differences in 1991 and 1992 lake selection procedures
Inclusion probabilities and expansion factors—their importance and use
Lessons learned
3.1 PURPOSE OF SURVEY DESIGN
A primary EMAP objective is to monitor the condition of the nation's ecological resources and
provide unbiased estimates, with known confidence, of the status of, and trends in, indicators of
condition. To carry out this task, EMAP relies on a probability-based survey of natural resources,
from which inferences about the condition of natural resources can be drawn. Without a statistic-
ally sound survey design for monitoring, there is no assurance of obtaining unbiased estimates of
the ecological condition of targeted resources with known confidence or to detect trends in eco-
logical indicators for this target resource population (Overton et al., 1991; Stevens, pers. comm.).
A probability-based survey establishes a firm foundation for estimating resource (i.e., lakes,
streams, forests, wetlands) characteristics or attributes, just as an experimental design provides a
basis for testing hypotheses in comparative scientific research studies.
As in any field, some of the terminology specific to the field of survey design may be unfamiliar to
those in other fields, or may have somewhat different meanings. The terminology used here will
be most familiar to those trained in survey theory and applications and may be unfamiliar to those
trained solely in experimental design. Terms such as populations and probability samples are
used extensively, but often have different connotations in the ecological and limnological disci-
plines. The population, in the survey sense, is the collection of units or items to be described
(e.g., all lakes, streams, or wetlands in a defined area, such as the conterminous United States).
Samples are drawn, with known probability (thus a probability sample), from the population.
Measurements made on the samples are used to infer the properties of the population. Popula-
tions can be defined inclusively, such as all lakes in the United States, or the focus can be on
subsets. The terms target population and subpopulation are sometimes used to represent a sub-
set of particular interest about which inferences or conclusions will be drawn (e.g., all publicly
91
-------
owned lakes between 1 and 1,000 ha in area; blue-line streams on 1:100,000-scale maps; all
wetlands > 0.4 ha).
Survey design and probability sampling theory have been rigorously defined, developed, and
documented in the statistical literature. The EMAP survey design is similar in concept to the
approaches used by the USDA National Agricultural Statistics Survey (NASS) (Cotter and Nealon,
1987) and the Forest Service Inventory Analysis (FIA) (Bickford et al., 1963; Hazard and Law,
1989) in monitoring the condition and productivity of agricultural and forest resources. Similar
approaches are used by the Census Bureau and in Gallup polls to determine household charac-
teristics and the opinions (i.e., attributes) of a representative cross-section of the U.S. population.
3.1.1 Probability Sampling Designs
Probability sampling is the general term applied to sampling plans in which:
• Every member of the population (i.e., the total assemblage from which individual
sample units can be selected) has a known probability (> 0) of being included in the
sample.
• The sample is drawn by some method of random selection, or systematic selection
with random start, consistent with these probabilities.
• The probabilities of selection are used in making inferences from the sample to the
target population (Snedecor and Cochran, 1967).
An advantage of probability-based surveys is their minimal reliance on assumptions about the
underlying structure of the population (e.g., normal distribution). In fact, one of the goals of
probability-based surveys is to describe the underlying population structure. Randomization is an
aspect of probability-based surveys worth emphasizing. Probability sampling designs use ran-
domization in the sample selection process as a way to select unbiased samples with known
inclusion probabilities. If randomization is not incorporated into the design, it is impossible to
know how well the sample represents the population. What are the biases? What proportion of
the target population does the sample represent? A common problem in many water resource
surveys is identifying what the samples represent, because randomization procedures have not
been used in selecting the samples. Without probability sampling, each sample often is assumed
to have equal representation in the target population, even though selection criteria clearly
indicate this is not the case. Without the underlying statistical design and probability samples, the
representativeness of an individual sample is unknown.
92
-------
Drawing inferences from samples selected without randomization and without incorporating inclu-
sion probabilities (expansion factors) can yield misleading conclusions. To provide policy-
relevant information, not only is the ecological condition of the target population important but
also the proportion of the resource that is in a particular state or condition such as good or
degraded. Very different policy and management alternatives might be evaluated if 25% rather
than < 1% of the target resource was found to be in a degraded ecological condition.
3.1.2 Need for a Common Design
EMAP is an integrated program, including all ecological resources—forests, arid lands, agroeco-
systems, wetlands, lakes, streams, Great Lakes, estuaries, and near-coastal systems—which
means we must be able to combine and compare information within and across ecological
resources and geographic regions. A common survey design among ecological resources
contributes to this integration. The design, therefore, applies to more than just aquatic resources.
Design goals include:
• Consistent representation of the environment and ecological resources by use of
probability samples.
• The ability to include all ecological resources and environmental entities.
• Provision for the capacity to respond quickly to emerging environmental questions and
issues.
• Representation of the spatial distribution of the ecological resource across the United
States.
• Provision for linking the results with results from other probability-based national
resource surveys.
• Ability to revisit sampling sites for trend detection.
The common design selected for EMAP requires explicitly defined target populations from which
samples are drawn with known probability to estimate population attributes rigorously and without
bias. This design is capable of sampling any spatially distributed and well-defined ecological
resource and can accommodate sampling resources that are discrete, such as lakes, or exten-
sive, such as forests. A common design may not be the optimal design for a focused, resource-
specific question, but it is an adequate design for all ecological resources. In addition, results
from it can be combined with results from other probability-based monitoring networks, if the
other monitoring programs meet the basic criteria for design-based surveys. The common design
93
-------
also has the flexibility to accommodate multiple resources and multiple environmental problems at
different scales, including adaptations to address emerging environmental issues and problems.
3.2 EMAP SAMPLING GRID
EMAP has selected and designed a triangular grid structure as the basis for selecting sampling
locations from which to estimate the status of and temporal trends in indicators of ecological
resource condition. This grid structure also permits estimates of the spatial extent or area of
target ecological resources. Grid requirements that satisfy the EMAP objectives include the
following:
• An equal-area sampling structure using a regular placement of sampling locations.
• A hierarchical structure for enhancing or decreasing the grid density.
• A realization of the grid on a single planar surface for the entire conterminous United
States.
The grid serves primarily as a tool for achieving spatial coverage. It does not imply a belief that
ecological resources are distributed uniformly.
3.2.1 Base Grid
The baseline grid used in EMAP, including EMAP-Surface Waters (EMAP-SW), has about 12,600
grid points distributed over the conterminous United States (Figure 3-1). Grid points are in a
triangular array about 27 km equidistant. Large, contiguous hexagons can be scribed around
each grid point, each with an area of 635 km2 (e.g., the large hexagon in Figure 3-2). The large
hexagons completely cover the areas of interest but are too large for cost-effective sampling.
Therefore, the requirement for equal area sampling is further accommodated by ascribing a
smaller hexagonal area centered around each grid point; each small hexagon has an area of
40 km2 (e.g., the small hexagon in Figure 3-2). Preliminary evaluations indicate that the 40-km2
hexagon around the grid point, combined with the number of grid points, probably is adequate
for characterizing the extent and distribution of the different ecological resources (e.g., forests,
arid lands, lakes, streams, estuaries, wetlands, agricultural systems) in the United States. These
12,600 small hexagons uniformly cover 1/16, or about 6%, of the area of the conterminous United
States; sometimes this is called a 1/16 area sample of the United States.
94
-------
'Hfpj
^^^'''ypi?V
Figure 3-1. The base grid overlaid on North America. There are about 12,600 points in the
conterminous United States.
95
-------
Grid Density Enhancement
27km
base density
3-fold
4-fold
7-fold
Figure 3-2. EMAP grid structure showing the relationship between the 635 km hexagons
and the embedded 40-km2 hexagon. The figure also illustrates the three-,
four-, and sevenfold grid enhancements for increased sampling density.
96
-------
Initial randomization of the grid on the United States establishes the systematic sample (i.e.,
uniform and regular grid points and small hexagons) as a probability sample. This randomization
occurs when we place the grid on a map of the United States, then shift it random distances in
east-west and north-south directions. The grid structure reflects the importance of achieving
geographic coverage of ecological resources. The uniformity of spatial coverage provided by a
grid (Figure 3-1) ensures that each ecological resource can be sampled in proportion to its
geographic presence in the United States and that all ecological resources can be included in the
monitoring program (e.g., lakes, streams, prairie pothole wetlands, forests, grasslands, deserts,
estuaries, managed agricultural ecosystems, and the Great Lakes). We can also place the same
grid structure over maps of Hawaii, Alaska, and the Caribbean and use it as a basis for a consis-
tent international survey design.
The grid density also can be increased to sample rare resources such as the California redwoods,
additional resource classes, or smaller areas such as individual states or ecoregions. The grid
arrangement also makes it easy to either increase or decrease the grid density (Figure 3-2) and
retain the basic triangular structure important for consistent spatial coverage. Specific multiple
factors (e.g., three-, four-, and sevenfold) are available to increase or decrease the base grid
density and still maintain the sampling design requirements (Figure 3-2). The triangular nature of
the grid allows greater flexibility in these enhancement factors than a square or rectangular grid.
3.2.2 Hierarchical or Tier Structure
The sampling structure is hierarchical, presently defined with four levels or tiers, two of which are
used for routine monitoring. The different stages in the hierarchy reflect the desire to incorporate
measurements that can be obtained for different costs. For example, we can obtain information
on the distribution and extent of resources rather inexpensively with remote imagery, so many
sites can be assessed. On the other hand, we can obtain detailed, time intensive data for only a
few resources; the cost of such monitoring prevents sampling very many sites.
In EMAP, a Tier 1 sample of lakes consists of all lakes in the 12,600 40-km2 hexagons. This Tier
1 sample provides good spatial coverage of the U.S. lake resource, and some measurements,
such as lake size, can be made easily on the entire resource. But it is not necessary to sample
every Tier 1 lake to achieve robust estimates of condition, and it would be prohibitively expensive
to do so. Therefore, we select a subsample of the Tier 1 sample, called the Tier 2 sample. How-
ever, the greater Tier 1 density permits more precise estimates of resource extent, distribution,
97
-------
and characterization than can be obtained using only Tier 2 sites. Tier 1 characterization also
decreases the time required for EMAP to respond to emerging environmental problems, because
information about ecological resources acquired at that level can be used to refine or focus the
Tier 2 sample. The Tier 2 sites are selected so they also satisfy the design criteria of good spatial
and geographic distribution with known inclusion probabilities.
EMAP also establishes a four-year sampling cycle, with assignment of a particular year in the four-
year cycle to each hexagon. This four-year rotational cycle is more fully described in Section 3.5.
3.3 FRAME AND TIER 1 SAMPLE SELECTION
In the following sections, we incorporate specific information about using the design for the
selection of lakes. We include information on how we selected Tier 2 samples (the set of lakes to
be field visited) for both the 1991 and 1992 field seasons. In general, we describe the process
used during the 1991 selection and note how it was altered for the 1992 selection. Some differ-
ences arose, related both to what we learned during the 1991 selection process and to a slightly
different focus during the 1992 field season.
3.3.1 Lake Sampling Frame and Tier 1 Sample
A sampling frame is an explicit representation of a population from which a sample can be
selected. As the representation, or frame, of the population of lakes, we used the USGS
1:100,000-scale map series in digital format (DLGs) and the modification of the DLGs represented
by the U.S. EPA River Reach File, Version 3 (RF3), which established edge matching and direc-
tionality in the DLG files. The DLG (or RF3) files contain a representation of the spatial
distribution of all lakes and streams as recorded on the USGS 1:100,000-scale topographic map
series. The lake population of interest to EMAP-SW in the current series of pilot studies consists
of all lakes in the conterminous United States with a surface area > 1 ha, excluding the
Laurentian Great Lakes.
One difficult issue has been to provide an operational definition of a lake that field crews could
use in deciding whether to sample a waterbody chosen via the lake selection process. After
examining Cowardin's wetland classification system (Cowardin et al., 1979), which mainly dis-
tinguishes deep water from shallow water habitats, and after discussing the issue at length,
EMAP-SW staff settled on an operational definition that combines information from the 1:100,000-
scale maps and information available only from field visits.
98
-------
The first level of selection included waterbodies identified as lakes > 1 ha in surface area in the
DIG database (also as recorded in the DLGs). The second level of selection specified a criterion
for field visitation. If the waterbody had > = 1,000 m2 of open water and an estimated maximum
depth > 1 m, it would be sampled for lake indicators. If not, the waterbody would be recorded as
a noninterest waterbody for the purpose of measuring lake indicators.
A Tier 1 sample consists of all lakes captured by the 40-km2 hexagons; this 6% area sample
therefore captures approximately 6% of the lakes across the region of interest. Operationally,
each lake is uniquely represented by a point with specific coordinates in the DLGs or in RF3,
generally at the centroid of the lake. Each point has associated attributes such as latitude/
longitude, unique identification, lake surface area, and other available information. Because the
DLGs and RF3 are in electronic form, it is also feasible to create an inventory of lakes in specific
regions of interest.
For the 1991 field sampling season, we confined our selection process to lakes in the northeast,
the region of the country selected for the initial pilot studies. We also restricted the selection to
lakes < 2,000 ha for the 1991 sample, because we were not confident that we could effectively
apply the index sampling process to large lakes. However, since subsequent refinements to RF3
allowed us to draw a national Tier 1 sample (lower 48 states) for the 1992 field season, we did so
in view of the possibility of conducting a national survey on a small set of lake indicators. We
also included larger lakes. Table 3-1 summarizes the inventory and Tier 1 sample of lakes for
both years. The summaries are organized by size class; different size classes were selected for
each year, as described in Section 3.3.3.
Evaluation of the 1992 Tier 1 national sample revealed that the number of lakes > 500 ha cap-
tured by the 40-km2 hexagons was too small to form an adequate Tier 2 sample. Any lakes
< 500 ha that fell within the 40-km2 hexagons were selected as Tier 1 lakes (each lake is
uniquely represented by a point with specific coordinates in RF3). The complete list of lakes
> 500 ha was drawn from the frame and the Tier 2 sample was drawn from this list. It was more
efficient to select a probability sample directly from the lake list than to increase the density of the
grid to obtain an adequate sample of these larger lakes. The spatial distribution of these larger
lakes was preserved by associating each large lake with its nearest hexagon and then using the
normal selection procedure described in Section 3.4 and Appendix 3A.
99
-------
Table 3-1. Numbers of Lakes in the Population and Tier 1 Sample from the Base Grid3
1991 (Northeast)
Size Class (ha)
1-5
5-20
20-500
500-2,000
Total
Population
10,791
5,969
3,444
174
20,378
Tier 1
(4 yrs)
662
371
197
9
1,239
Tier 1
(1991)
158
92
57
4
311
Tier 1
Rejects
35
10
3
0
48
Tier 1
Screened
123
82
54
4
263
1992 (Lower 48 states)
Size Class (ha)
1-5
5-10
10-50
50-500
500-5,000
> 5,000
Total
Population
161,616
43,744
41 ,648
11,712
1,661
257
260,638
Tier 1 (4 yrs)
10,101
2,734
2,603
732
NAb
NAb
Population numbers come from the Digital Line Graph summaries; 1991 covers the northeast pilot and 1992 represents
the conterminous United States.
The Tier 2 lakes were selected directly from the target population of lakes, so no Tier 1 sample was drawn.
3.3.2 Frame Characterization
The frame is not a completely accurate representation of the population of lakes of interest.
Various errors creep in. Also, a waterbody identified by the USGS as a lake for mapping
purposes does nbt always correspond to a limnologically defensible definition of a lake.
Therefore, the lake frame requires characterization to identify a combination of mapping errors,
such as (1) parts of lakes (e.g., coves and bays), estuaries, and wide sections of rivers
designated as lakes, (2) changes in the landscape since the time the maps were compiled (some
100
-------
lakes and reservoirs have been drained, others created) and (3) waterbodies that will be tracked
but not included in the sample to be visited. Thus the characterization activity identifies two cate-
gories: frame errors and noninterest (or nontarget) waterbodies. We use the term noninterest
solely in the context of identifying types of waterbodies that will not be monitored in the field and
that will be excluded from initial population characterization in our pilot studies. In keeping with
EMAP's goal of leaving no orphan ecosystems, we do not presume to exclude these waterbodies
from ever being considered for EMAP monitoring. One of the advantages of survey designs such
as EMAP is that the proportion or number of waterbodies not chosen for consideration can be
estimated and tracked.
We evaluated frame errors and noninterest waterbodies for the 1991 Northeast Pilot by examining
larger scale maps (7.5-minute topographic and larger scale county maps) and talking with local
experts. The majority of the frame errors and noninterest waterbodies were associated with lakes
< 20 ha. About 20% of the Tier 1 lakes from 1 to 5 ha represented on the DLGs and 10% of the
lakes from 5 to 20 ha were determined to be either frame errors or noninterest waterbodies such
as cranberry bog reservoirs, wide spots in rivers, or waterbodies that were arms or bays of larger
lakes. No (0) noninterest lakes were found among lakes with surface areas > 50 ha. Chapter 5
(Annual Statistical Summary) contains additional details. For the 1992 pilot, resources needed to
perform a Tier 1 screen on the lower 48 states were unavailable, so we drew a Tier 2 sample
directly from the unscreened Tier 1 sample, while accounting for the expected low-interest lakes.
Identifying lakes that are of interest but not represented in the frame (i.e., not in the DIG file or on
USGS topographic maps) is more difficult. Methods considered for future identification of lakes
not represented in the frame include using remote aerial imagery/photography and relying on
local experts to provide detailed area knowledge. Results of these efforts can then be compared
to the lake frame, either in conjunction or separately. Our initial impression for lakes in the
Northeast is that the frame over-represents the target population; that is, it is more likely to
classify noninterest waterbodies as lakes than to exclude interest lakes. Evaluating the accuracy
of the Tier 1 and Tier 2 lake selection process represents a major activity that will occur over the
next several years.
3.3.3 Classification Strategies
To ensure an adequate representation of various subpopulations in the sample, lakes could be
classified by subpopulations as part of the Tier 1 activity. The question is whether or not to
101
-------
develop a lake classification scheme, other than one based on size, to stratify the Tier 1 sample,
then select the Tier 2 samples from each stratum. For example, the lakes could be classified as
oligotrophic, mesotrophic or eutrophic; warmwater or coldwater; deep or shallow; impoundment
or natural lake, etc. Because of the variety of overlapping lake classification approaches and the
potential loss of precision in case of misclassification, we decided to minimize stratification and
rely on post-classification as a more appropriate part of the evaluation of results. However, as
part of the lake frame evaluation activities, we will analyze the Tier 1 and Tier 2 lake character-
istics that can readily be obtained from maps, atlases, or similar references in order to initially
determine the different subpopulations of lakes in the target population.
A second question, however, was whether we could afford not to classify lakes on the basis of
surface area. A nonstratified, random sample would yield lakes in proportion to their abundance,
so most of the lakes selected would be small. Over 75% of lakes in the frame have surface areas
between 1 and 10 ha (Table 3-1). Equal probability allocation would produce Tier 2 samples with
proportional numbers of small lakes. The importance of lakes to society is related both to
numbers of lakes and to their sizes. If we strictly apportioned Tier 2 selection on the basis of
numbers of lakes, we would impair our ability to describe the condition of the larger lakes. There
is considerable public, policy, and regulatory interest in lakes > 10 ha. But if we strictly appor-
tioned sampling according to area, we would impair our ability to describe the condition of small
lakes. Thus, we sought to balance the selection of Tier 2 lakes. One approach to obtaining a
more equitable distribution of lakes by size would be to allocate equal numbers of samples along
a logarithmic or square root transformation of surface area, so that more large lakes would be
selected. However, allocating samples along a continuous scale requires the use of variable
inclusion probabilities, substantially complicating variance estimation (Overton, pers. comm.). The
issue of whether the advantages of using variable inclusion probabilities outweigh the disadvan-
tages of complicated variance estimation has not been completely resolved, but we decided on
the approach of classifying by lake area. We used an iterative process, varying size classes and
sample sizes among classes, to select samples about equally among lake size classes, approxi-
mating what would have been achieved by logarithmic or square root transformations. For the
1991 pilot, we based our sample selection only on the population of lakes in the Northeast and
restricted the size strata to lakes between 1 and 2,000 ha. For the 1992 pilot, we used estimates
for the lower 48 states and included lakes > 2,000 ha. As a result, slightly different size classes
emerged from the process, summarized in Table 3-1, along with the numbers of lakes in each
size class. Over the next several years, we will evaluate the relative allocation between size
classes. These size classes are not intended to convey any ecological information, but were
102
-------
selected as a way to spread the Tier 2 sample more evenly among lake sizes than would have
been done by equal probability sampling.
3.4 LAKE SELECTION
Evaluation of the Tier 1 sample allows us to refine the selection of the Tier 2 sample. Character-
izing the Tier 1 sample by lake size (Section 3.3.3) focuses the Tier 2 sample so that it better
characterizes larger lakes than could be done otherwise. It also prevents oversampling the small
lakes beyond what is necessary to characterize their condition. The selection of Tier 2 lakes met
two criteria: (1) we used probability methods to select the lakes from the Tier 1 sample, and (2)
we preserved the spatial distribution of the lakes in the Tier 2 sample.
The details of the Tier 2 lake selection process appear in Appendix 3A. The outcome of the Tier 2
selection process is a set of lakes on which measurements of lake condition will be taken during
the appropriate index period. Table 3-2 summarizes the sets of Tier 2 lakes selected for the 1991
and 1992 sampling seasons. The sample of lakes represents the larger population of lakes of
interest; each lake is selected with a known inclusion probability reflecting the probability of
selecting any particular lake at the Tier 1 level and at the Tier 2 level. Reciprocals of inclusion
probabilities are the multipliers by which the sample attribute is expanded to its portion of the
population. This multiplier is called an expansion factor. For example, if the attribute surface area
for a particular lake in the Tier 2 sample were 10 ha, and that lake's expansion factor were 100
(inclusion probability = 0.01), the lake would have a weight of 1,000 ha. Summation of the
weights across the sample of lakes produces an estimate of the total surface area of lakes in the
population. Inclusion probabilities and expansion factors for each size class are included as part
of Table 3-2, and the spatial distribution of these lakes is illustrated in Figures 3-3 and 3-4.
3.5 TEMPORAL SAMPLING SCHEDULE
EMAP is designed both to describe the current status of and to detect trends in the condition of
ecological resources. Status is best estimated by allocating sampling effort among as many
different sample units (lakes) as possible. However, trend is best detected by incorporating
repeat visits but not necessarily annual revisits. EMAP's design balances both objectives by
establishing a four-year rotational cycle in which each grid point is assigned a year. This
rotational design increases the number of lakes visited over the four-year cycle to a higher
number than would be the case under annual revisits. The increased effective sample size
103
-------
Table 3-2. Tier 2 Yearly Sample Sizes for the Northeast, with Corresponding Inclusion
Probabilities and Expansion Factors
Size Class (ha)
Tier 2 Sample
Overall
Inclusion
Probability
1991
1-5
5-20
20-500
500-2,000
Total
16
20
24
4
64
.001953
.003906
.007422
.01562
Expansion
Factor
512.000
256.000
134.737
64.000
1992
1-5
5-10
10-50
50-500
500-5,000
> 5,000
Total
9
31
20
6
10
1
77
0.00125
0.00938
0.00469
0.00859
0.0625
0.0500
800.000
106.667
213.333
116.364
16.000
20.000
improves estimation of status (over the four-year interval) and trend detection, compared to
annual revisits. This assignment preserves the triangular grid structure, so that in each year
consistent spatial coverage is maintained. During the first year, one quarter of the hexagons are
selected (schematically represented by "f' in Figure 3-5). Only lakes in these hexagons are
candidates for field sampling during the first year. Recall that the lakes captured by the 40-km2
hexagons comprise the Tier 1 sample and that only a portion of the Tier 1 sample will be visited.
During the next year, the second quarter of the lakes, those captured by hexagons designated as
"second-year" ("+"), are available for sampling, and so on. In this manner, all Tier 2 lakes are
sampled during a four-year period. Any individual lake, therefore, is sampled once every four
104
-------
EMAP Lakes
Tier 2 Samp Ie
EPA Regions 1 and 2
EMAP-Stf 8Sept92
Figure 3-3. Map illustrating the spatial distribution of the 1991 and 1992 Tier 2 samples of
lakes selected from the Digital Line Graph frame.
105
-------
I
.0
I
£
•8
T3
(0
*
•o
I
§
0
O
o
CM
o
73
« c
« o
•R c
O i_
•5 «
I?
** k
.2 5
Q o
2
D>
106
-------
year
+ year 2
O year 3
year 4
Figure 3-5. Temporal sampling schedule for all EMAP resources illustrating the uniform
spatial coverage for every year.
107
-------
years. A second monitoring cycle begins in the fifth year by revisiting the first year (" f") lakes;
this process could continue indefinitely. The temporal sequence of hexagons has been fixed for
EMAP. Therefore, the lakes associated with each hexagon are also associated with a particular
sampling year. This ensures that all the EMAP resources sampled in that hexagon will be
sampled during the same year. To permit comparisons among ecological resource condition in
the same year, the hexagon sequence for any year is fixed and cannot be substituted for that of
another year.
The temporal sampling schedule and design have several advantages. Each year's sample
provides, in itself, both national and regional estimates of ecological condition, with uniform
spatial coverage from year to year. Annual estimates of population parameters are provided for
every geographic region and every identifiable population, no matter how dispersed. Revisiting
sites on a four-year cycle provides sufficient time for recovery from measurement stress and
allows time for subtle trends to be expressed. The design is well adapted for detecting persistent,
gradual change on disperse subpopulations and representing spatial patterns in ecological
resources.
3.6 ANNUAL REVISIT SITES
The EMAP design foresees that lakes will be revisited every four years. Over eight or more years,
this gives good sensitivity to trend detection. The sensitivity to trends during the initial years of
monitoring can be increased substantially by annual revisits to a few lakes (Urquhart et al., 1991).
Ten of the lakes visited in 1991 were randomly selected, subject to spatial coverage, to be
revisited in 1992 and subsequent years. The process used for achieving random selection,
subject to spatial balance, is described in Appendix 3A.
3.7 SUMMARY AND LESSONS LEARNED
During this past year of design work, we were actually able to complete two design cycles. The
first allowed us to put our design concepts into practice in selecting the 1991 lakes for sampling.
The process, although not completely automated the first time, worked reasonably well. The
selection resulted in an acceptable spatial distribution of lakes in the Northeast. We gathered
valuable information from the Tier 1 screening on the lake frame errors and the noninterest lakes.
This permitted a more efficient Tier 2 selection that contained relatively few errors and noninterest
lakes. Because of the initial grid density, the size of the selected Tier 2 sample, and the natural
108
-------
distribution of lakes, two or more Tier 2 lakes occurred in the same hexagon many times in our
1991 sample. This has both positive and negative aspects. On the positive side, we may be able
to estimate some of the spatial autocorrelation effects sometimes observed among lakes close to
one another. The negative aspect is that we do not achieve as broad a spatial coverage as might
otherwise be possible, a desired characteristic in Tier 2. The only ways to decrease the
frequency of this multiple occurrence of lakes are to reduce the Tier 2 sample size or to increase
the Tier 1 density. We are in the process of evaluating the consequences of these alternatives.
During the second design cycle, which resulted in the selection of the 1992 samples, we were
able to apply our design concepts and experience to a national (conterminous U.S.) sample. This
was an important step in improving our understanding of the overall design and spatial distri-
bution of the resulting Tier 2 sample. If the relatively simple approach applied to the Northeast
resulted in an acceptable sample (i.e., adequate coverage of multiple subpopulations) nationwide,
then we would continue along the same course. That is, if the same inclusion probabilities (within
each size class) could be applied nationally, the analyses resulting from the data and comparison
across regions would be greatly simplified. Given the magnitude of the national Tier 1 sample,
we chose not to do the initial Tier 1 screening for frame errors and noninterest waterbodies the
second year. Instead, we increased the Tier 2 sample size to account for the expected losses
and achieve the desired Tier 2 sample size. Table 3-2 and Figures 3-3 and 3-4 show 1992 size
categories, inclusion probabilities, and the spatial pattern of selected Tier 2 lakes. In general, the
selection produces few surprises. There is greater density of lakes in the east than in the west.
In the east, there is a greater density in the north than in the south. However, the distribution of
samples raises some concerns and questions. For example, the density of lakes in the upper
midwest reflects the greater density of lakes in this region; we are not sure, however, that a
sample size that large is necessary in the upper midwest to adequately characterize the region.
Conversely, the sparser sample selection of lakes in the west clearly reflects their lower density;
but we are concerned that we may not adequately characterize some lake subpopulations of
interest, such as the alpine lakes. Alternatives for dealing with these situations are currently under
discussion.
Perhaps the final lesson learned from the 1992 sample selection process has been the need to
perform the Tier 1 screening. An effective evaluation will in fact make the selection of Tier 2 a
much more efficient process both in terms of what we are able to say about lakes simply from the
Tier 1 sample and also from the perspective of a more efficient use of time by those doing the
109
-------
field reconnaisance on the Tier 2 samples. This Tier 1 screening will be incorporated in
upcoming years.
Our experience with the lake selection process indicates that, given a lead time of 4 to 6 months,
we can select a set of lakes for monitoring that meets the general design criteria for a probability
sample with spatial balance. Refinements will be necessary to define important lake subpopula-
tions and to improve descriptions of the small lakes (< 10 ha).
110
-------
APPENDIX 3A
DETAILS OF THE LAKE SELECTION PROCESS
The lake selection process is illustrated schematically as Figure 3A-1; details of the selection
process are as they occurred for selecting the 1992 national lake sample. Where appropriate, the
text that follows identifies the differences in the process used for the 1991 sample. The first three
steps are straightforward and produce the Tier 1 sample, summarized earlier in Table 3-1.
The selection of the Tier 2 sample from the Tier 1 sample recognizes the same basic survey
criteria met by the Tier 1 sample selection:
• The sample is drawn with known probability.
• The sample is drawn to preserve the spatial pattern of the population of interest.
In order to meet these criteria, conceptually, we start the Tier 2 selection process by dividing the
region of interest into smaller compact clusters of hexagons, then randomly selecting lakes within
each cluster. For the 1991 Northeast Pilot, the clusters were developed solely for the Northeast
(New York, New Jersey, and the New England states) and were subjectively drawn. For the 1992
pilot, the clusters were developed nationally using a computer algorithm. The clustering met the
following criteria:
• The sums of the inclusion probabilities in each cluster for each year were as similar as
possible across clusters.
• Compact, "round" clusters were preferable to long, narrow clusters.
• All lakes in a 40-km2 hexagon were in the same cluster.
The clusters were constructed so that they contained a total inclusion probability of at least 2 in
each year of the 4-year cycle. Delineating the clusters and selecting lakes within each cluster
randomly also assured that the spatial distribution of the lakes would be preserved. Because the
primary function of the clusters was to distribute the sampling effort in proportion to the spatial
distribution of lakes, the actual dimensions and boundaries of the clusters were not critical for the
next steps. It is desirable for at least one lake to be selected from each cluster, but the advan-
tages of the clusters are defeated if more than three or four lakes are selected from any one
cluster. Therefore, the target was to define clusters such that at least two lakes from each cluster
would be selected. Step 4 in Figure 3A-1 shows an example of this clustering.
111
-------
1. Common EMAP grid frame for all U.S. ecological
resources.
2. Four-year sampling cycle established for EMAP.
Hexagons designated by year. After random
displacement of the base grid.
635km2
40km
3. EPA River Reach File (RF3) used to develop
sampling frame.
a. Identify all lakes in 40-km2 hexagons based
on unique point locators for each lake.
b. Extract lakes >: 1 ha using GIS.
c. For lakes ^ 500 ha, keep grid
structure.
d. For lakes > 500 ha, inventory or list the
lakes.
e. These are Tier 1 lakes.
Figure 3A-1. Schematic representation of the lake selection process (page 1 of 2).
112
-------
4. Classify lakes by surface area into 6 size
categories: 1-5, 5-10, 10-50, 50-500, 500-5,000,
and > 5,000 ha.
a. Determine inclusion probabilities for lakes by
size class.
b. Establish lake clusters so that sum of
inclusion probabilities in each cluster is about
the same, regardless of size class.
5. Separate the hexagons, and associated lakes, by
sampling year in each cluster. Start with Year 2
hexagons - for 1992.
a. Randomly order clusters, hexagons within
clusters, and lakes within hexagons.
b. Randomly order list of lakes > 500 ha.
c. Assign length proportional to inclusion
probability to each lake in array.
d. Select random number to initiate process and
incrementally increase by length that results
in selection of desired number of sample
lakes.
e. Follow similar process for inventory of lakes
> 500 ha.
r =*•
r+1 =
r+2=>
r+3=».
r+5
r+6
Lake 18,3,2
Lake 18,3,1
Lake 18,3,4
Lake 18,3,3
Lake 18,1,7
Lake 18,1,1
Lake 18,1,4
Lake 18,1,3
Lake 18, 1,2
Lake 18,1,5
Lake 18,1,6
Lake 18,2,2
Lake 18,2,1
Lake 4,7,6
•
•
•
Hex 18,3
Hex 18,1
Hex 18,2
Hex 4,7
•
•
•
C
L
U
S
T
E
R
1
8
C
L
U
S
T
E
R
4
6. Repeat process for hexagons and lakes in the
other three years of the sampling cycle.
7. Perform QA on lake list to identify nontarget lakes.
Selected Lakes
Lake 18,3,1
Lake 18,1,7
Lake 18,1,1
Lake 18,1,3
Lake 18,1,6
Lake 18,2,1
QA
Lake
Lake
Wetland
Wide spot in river
Lake
Lake drained
Figure 3A-1. Schematic representation of the lake selection process (page 2 of 2).
113
-------
Drawing the Tier 2 Sample
For the 1991 Tier 2 sample, the Tier 1 sample was screened for frame error and noninterest lakes
before the Tier 2 sample was drawn (See Table 3-1). For the 1992 Tier 2 sample, no Tier 1
screening was performed because the time required to perform a national screen exceeded the
time available to produce the Tier 2 sample. For future work, efforts to estimate Tier 1 errors will
be factored into the process.
A tentative national yearly Tier 2 sample of 800 for each resource unit has been identified as a
target for each EMAP resource group in each year for the lower 48 states. We translated this into
800 lakes and 800 streams per year as design targets for EMAP-SW. Based on our Tier 1 screen-
ing of the 1991 sample and subsequent field visits that identified other nontarget lakes in our
1991 sample, we overselected lakes in the 1-5 and 5-10 ha classes. Our best estimate, based
on the 1991 experience, was that about half of the 1-5 ha and about one-third of the 5-10 ha Tier
1 sample (unscreened) would be errors or nontarget lakes. Therefore, target goals were estab-
lished to select 200 lakes in the 1-5 ha size class and 375 lakes in the 5-10 ha size class with the
expectation of achieving about 100 lakes in the 1-5 ha size class and 200 lakes in the 5-10 ha
size class. If our projections were correct, the resulting sample of interest lakes should approxi-
mate 800 each year, distributed among size classes as in Table 3A-1.
The process outlined here was followed in selecting the 1992 Tier 2 sample from the Tier 1 set of
lakes:
1. A conditional inclusion probability representing the fraction of the Tier 1 sample to be
selected as the Tier 2 sample for each size class was assigned to each Tier 1 lake.
2. Clusters were delineated based on their total conditional inclusion probability.
3. The hexagons for the first, second, third, and fourth years of sampling were identified
along with their corresponding lakes. A four-year rotation sampling cycle has been
established for EMAP, so the hexagons to be sampled, if the resource exists in that
hexagon, have been designated as year 1, 2, 3, 4 (Figure 3A-1).
4. Three arrays were established by randomly ordering the clusters, randomly ordering the
40-km2 hexagons for a specific sampling year (e.g., year 1, year 2) within each cluster,
and randomly ordering the lakes within each 40-km2 hexagon. This resulted in an order
for all the lakes in the sample; the lakes within each 40-km2 hexagon were adjacent to
each other, and the 40-km hexagons within each cluster were also adjacent to one
another (See Figure 3A-1). Each cluster consisted of a contiguous group of hexagons, for
which the total conditional inclusion probability was at least 2 for each of the four years of
hexagons in the cluster. In order to achieve this condition, some cluster/year
combinations substantially exceeded the target value of 2.
114
-------
Table 3A-1. Target Sample Sizes, Actual Sample Sizes, Inclusion Probabilities, and
Expansion Factors for the 1992 National Lake Sample
1992 National Lake Sample
Size Class
1-5
5-10
10-50
50-500
500-5,000
> 5,000
Totals
Target
Sample Size
200 (100)a
375 (250)a
200
100
100
50
1,025
Actual
Sample Size
188
413
197
97
111
10
1,016
Inclusion
Probability
0.00125
0.00938
0.00469
0.00859
0.0625
0.0500
Expansion
Factor
800.000
100.667
213.333
116.364
16.000
20.000
Because of frame errors and noninterest waterbodies in these two groups, we selected a higher target sample size
than is normally required. Desired numbers of target lakes are in parentheses.
5. Within this array, each lake was assigned a length equal to its conditional (Tier 2)
inclusion probability [e.g., lakes 50-500 ha in area (with conditional inclusion probabil-
ity = 0.55) would have an array length about equal to the lakes 5-10 ha in area (with
conditional inclusion probability = 0.6), about twice the length of the lakes 10-50 ha in
area (with conditional inclusion probability = 0.3), and about seven times as long as
the lakes 1-5 ha in area (with conditional inclusion probability = 0.08)]. This is
illustrated in Figure 3A-1.
6. A random number, r, between 0 and 1.0 was chosen, and the lake located or over-
lapping the corresponding distance on the line was selected as the first Tier 2 lake.
This random number then was incrementally increased by 1 and each corresponding
lake encountered along the array was selected as belonging to the Tier 2 sample
(Figure 3A-1).
7. A similar procedure was followed for the lakes > 500 ha. The list of lakes was ordered
geographically so four lists were developed based on the sampling year of the hexa-
gon with which the larger lake was associated. Each lake was assigned a length pro-
portional to its inclusion probability, and the selection process (step 6) was repeated.
8. The actual number of Tier 2 samples selected annually is shown in Table 3-2 for both
years.
115
-------
The Tier 1 and Tier 2 lakes selected, with coordinate locations and the corresponding EPA
Region, were distributed to the EPA Regional Environmental Services Division Offices for quality
assurance to identify frame errors and noninterest lakes. Figure 3-4 shows the national
distribution of Tier 2 lakes for all years. This basic procedure was used both years, with slight
differences. For example, clusters were manually determined in 1991, whereas they were drawn
with a computer algorithm in 1992. Also, 1991 sample lakes were drawn from lakes of 1-2,000
ha, with none above 2,000 ha, whereas lakes with areas of 1-500 ha were drawn separately from
lakes > 500 ha in area in 1992.
Sample Expansion Factors
We have asserted several times that a fundamental feature of the design is the selection of each
sample unit (lake in this case) with known probability. Why is this important? What is done with
the information? In the end, the scores or attributes for each lake are weighted; the weights are
expansion factors allowing us to make inferences to the entire population based on the sampled
lakes.
Sample expansion factors can be interpreted as the number of lakes in the population repre-
sented by each sample lake. Therefore, the sum of the expansion factors estimates the total
number of lakes in the population, or in a defined subpopulation if the summation of expansion
factors is limited to sample lakes in the subpopulation. Expansion factors are calculated as the
reciprocal of the lake's probability for inclusion. For example, if there are 1,000 lakes in the
population and each is selected with probability of 0.05, then the sample expansion factors all
equal 20 (i.e., 1/0.05). Under this scheme, the number of lakes actually selected for sampling
varies around an expected value of 50 (= 0.05*1000).
If instead, 100 of the 1,000 lakes are selected with probability 0.14, while 900 are selected with
probability 0.04, the expected number of lakes would still be 50 = 0.14 x 100 + 0.04 x 900. The
expansion factors for the two groups of lakes would be 7.14 = 1/0.14 and 25 = 1/0.04; each
sample lake in the first group would represent about 7 lakes in the population, whereas each
sample lake in the second set would represent 25 lakes in the population.
Tier 1 Probability
In practice, how are inclusion probabilities determined? The probability of selecting a lake in Tier
1 is given by the proportion of the land area captured by the EMAP sample areas (usually 40-km2
116
-------
hexagons) relative to the area of the base tessellation (usually 635-km2 hexagons). Thus, by the
initial randomization of the grid:
Tier 1 inclusion probability, (1X hexagons) = PT11 1X = 1/16
Because we sample on a four-year cycle, each year one-fourth of the Tier 1 sample is selected.
Thus:
Tier 1 Year 1 inclusion probability (1X hexagons) = PT1Y11X = (1/16)/4 = 0.015625
This applies for any one of the four years.
Tier 2 Probability for 1991
Tier 2 lakes are selected for field visitation as described in Section 3.4. The probability of
selecting a lake in the Tier 2 sample from the Tier 1 sample reflects the proportion of lakes from
the Tier 1 sample that will be selected for the Tier 2 sample. For example, if the Tier 1 sample
consisted of 1,000 lakes and our target Tier 2 sample size were 50, the conditional Tier 2 selec-
tion probability would be 0.05. Overall Tier 1/Tier 2 inclusion probabilities are the product of the
Tier 1 inclusion probability and the conditional Tier 2 inclusion probability. The sample expansion
factor is then calculated as the reciprocal of this overall selection probability:
1X expansion factor =
(PT1Y1,1X *
For example, the probability of selecting a lake of 20-500 ha is (1/64)*(0.475) = 0.007421875,
and its associated expansion factor is 134.737. Thus, that lake represents 134.737 lakes in the
population and any indicator scores for that lake will be expanded to the population using this
weighting factor.
Because we selected different proportions of lakes from the Tier 1 sample for the Tier 2 sample in
each of the size classes, the conditional probabilities differ among size classes, as illustrated here
for 1991:
117
-------
Lake Size Class (ha) Tier 2 Probability (PT2) Expansion Factors
1-5 0.125 512.000
5-20 0.25 256.000
20-500 0.475 134.737
500-2,000 1.00 64.000
Tier 2 Probability for 1992
For the 1992 sample, the inclusion probabilities reflect a design goal of 800 Tier 2 lakes across
the conterminous United States, with overselection in the two smallest size categories, as noted
earlier. All lakes > 500 ha were selected for sampling at the Tier 1 level, so the Tier 1 probability
for that size class is 1.00 over the four-year cycle, or 0.25 for each year. Table 3A-1 summarizes
the 1992 National Lake Sample.
118
-------
CHAPTER 4
A FRAMEWORK FOR EVALUATING THE SENSITIVITY
OF THE EMAP DESIGN
4.1 INTRODUCTION
We continue to refine our ability to evaluate the sensitivity of the proposed EMAP design for
estimating the condition, and changes in condition, of lakes and streams. Preliminary descrip-
tions of design capabilities appear in the EMAP-Surface Waters Research Plan (Paulsen et al.,
1991). Since that time, we have developed a general framework for investigators (indicator leads)
to use in evaluating how well candidate indicators can be expected to perform. A variety of
important components of variance affect our ability to describe status and estimate trends. The
framework, based on a description and estimation of these components of variance, can be used
to analyze descriptions of population and subpopulation status and to evaluate how extraneous
variance might distort those descriptions. Structured as a sensitivity analysis, it also allows
investigators to explore the effects of components of variance, sample sizes, years of monitoring,
and levels of Type I and Type II errors on the ability to detect trends. Combined with an analysis
of variance components on carefully selected available datasets and on data derived from EMAP
pilot studies during the first several years, this evolving framework will provide insight into how
well we can expect to describe status and to detect trends. It will also guide us in selecting
indicators and estimating their sensitivity.
This chapter addresses the following:
• Components of variance that are important to consider when evaluating design
sensitivity
• Estimation of components of variance using a general linear model as the framework
• Influence of extraneous variance on descriptions of population status
• Influence of extraneous variance on sensitivity to trend detection
Good, long-term data records from which we can estimate all the variance components needed
are generally not available. For indicators of trophic condition, some statewide datasets can be
used to illustrate the important components of variance. Additional effort will be devoted to
describing these variance components for other biological indicators derived from EMAP pilots
and a search for other data sets. This chapter provides examples, primarily of Secchi disk trans-
119
-------
parency (SD), total phosphorus (TP), and chlorophyll-a (chl-a). As outlined here, we can now
describe the kinds of data sets that would be most beneficial for evaluating candidate indicators.
We can also design appropriate studies into the pilot programs that will supply the necessary
variance estimates during the first few years of the program (at least four years for estimation of
some variance components derived solely from the pilots).
4.2 IMPORTANT COMPONENTS OF VARIANCE
It is critical that we evaluate the magnitude and effects of different components of variance to
determine the utility of particular indicators in estimating status and trends. Furthermore, the
choices for minimizing variance components, or their effects on estimates of status and trends,
depend upon their relative magnitudes. In some cases, it might be cost effective to reduce the
variance components associated with sample collection and processing; in other cases, it might
be effective to increase the number of visits to individual lakes, or to increase the number of lakes
visited at the expense of revisiting individual lakes. The following sections outline components of
variance that could have important effects on status description and trend detection.
4.2.1 Population Variance (
Population variance describes the measured differences among lakes in a regional population or
subpopulation during the index period, that is, the status of the lake population. We use cumula-
tive distribution functions (CDFs), histograms, or other population descriptors to express it. If no
other forms of variation interfere with the sampling process, index snapshots derived from the
randomly chosen sample of lakes would unambiguously express population variance.
4.2.2 Extraneous Variance
Extraneous variance comprises all variance components that are not a part of population vari-
ance, but that reduce the precision of estimates of status and trends. Some of these variance
components are of interest themselves as possible indicators, but for the most part, they inhibit
our ability to describe the population characteristics. Extraneous variance can be decomposed
into several pieces as follows (components estimated here are symbolically identified):
4.2.2.1 Year Variance (<^ar\ year effects)
Year variance measures the amount by which all lakes in a population or subpopulation are high
or low in a particular year. The condition of regional populations of lakes fluctuates around a
120
-------
central value in the absence of a trend. In the presence of a trend, this variance component
measures the year-to-year variation from the trend line or curve. Trend detection is fundamental
to EMAP, and trend detection capability is very sensitive to the relative magnitude of this compo-
nent of variance. This population level variation is sometimes called a year effect, as it is a
measure of the amount by which all lakes in a particular year are above or below a long-term
central value.
This common pattern of variation among lakes is caused by regional-scale factors affecting the
population in a consistent way, such as regional-scale climatological conditions (e.g., wet years or
dry years). We are unaware of any regional-scale databases on indicators of interest to EMAP-
Surface Waters (EMAP-SW) from which we could obtain estimates of this component of variance.
To our knowledge, no consistent biological monitoring programs on lakes and streams have been
conducted over broad regions for many years. Some state-level sampling programs give insight
into the magnitude of this component of variance at a statewide scale of resolution, that is, on a
scale more local than the scales at which EMAP intends to monitor.
4.2.2.2 Lake-Year Interaction (o^ake.eap interaction effects)
The condition of an individual lake fluctuates from year to year around its central value or around
a trend for that lake. These fluctuations are responses to local effects operating at the individual
lake level and are independent among lakes. This component of variance specifically describes
that part of a lake's year-to-year variation not already accounted for by the year component. This
interaction variance is a natural feature of the populations under study and may be of interest in
itself as an indicator of stress or change in an ecosystem. It can be estimated by repeat visits to
lakes across years. However, if lakes are sampled each year without revisiting during the index
period, the variance estimated from the year-to-year differences confounds two components of
variance (<72|ai
-------
subcomponents described here can be reduced. We used this strategy in our pilot studies:
First, identify the magnitude of o2^^ and then decide if additional information is needed. The
following is a partial list of components of index variance that might be targets for reduction in
overall index variance.
• Measurement variance (
-------
A BASIC STRUCTURE Fl
Lake 1
Lake 2
Lake 3
Lake 4
Lake 5
Lake B
Lake (last)
Overall
(like Average)
1
2
A BASIC STRUCTURE
Lake 1
Lake 2
Lake 3
Lake 4
Lake 5
Lake B
Lake (last)
Overall
(like Average)
i
.
— '
_
2
— r
3
FOR
- YE,
3
•
twr-nm
W D
EAR
4
Panel *
ISCUSSING EHAP
DISCUSSING
W_
4
= 0
Overall
Panel C
EHAP
Overall
— ,
— **
Trend + tr*reta
A BASIC STRUCTURE FOR DISCUSSING
Lake 1
Lake 2
Lake 3
Lake 4
Lake 5
Lake 6
Lake (last)
Overall
(like Average)
i
----
—---
—--
—
— -1
2
— — ~-~,
"""" ""*!
~~~ J
~~—*~-
' --s
~- — -.
"i
,---~-^,
(7
A BASIC STRUCTURE
Lake i
Lake 2
Lake 3
Lake 4
Lake 5
Lake 6
Lake (last)
Overall
(like Average)
i
.— —
-^
— •—,
2
, — — —
—- — -
**
^
ffg
3
"-— -
^-_-
~^^—
"--_-
•^.
-~-^_-
ff'fEM
•--. •
P
ff A R
FOR
~ YE,
3
•*.
,-•-— "™ ~
.
0
IS
. . .
__--—"
— ^_
— — ~_
Panel D
EHAP
Overall
-*•
—
-------
(a2 ) is seen as the amount of variation exhibited by the lakes in concert (panels B and C); the
lake-by-year interaction arises from the independent excursions of each lake (compare panel D
with panels B and C). The EMAP-SW pilot studies are intended to characterize one or more of
these sources of extraneous variation. The EMAP-SW staff would like to know about any data-
bases derived from similar sampling programs for physical, chemical, and, most important, bio-
logical indicators of lake and stream condition (discussed in Chapter 2).
In the remainder of this chapter, we use indicators of trophic condition to illustrate the design,
components of variance of interest, and status and trends sensitivity. We have begun to obtain
information about the broader range of indicators of the condition of lakes and streams and will
continue to do so.
4.3 A LINEAR STATISTICAL MODEL FOR ESTIMATING VARIANCE COMPONENTS
A convenient way to summarize the description of components of variance and to set the stage
for estimating the magnitude of these components is to use a linear statistical model:
Y|j=/. + Li + Tj + Eij (1)
where / indexes lakes, / = 1,2, - f, j indexes years, / = 1,2, - 1\ Y,, is the condition of a lake
at any particular time; /* is the grand average condition of lakes across the region of interest
during the time interval of interest; L( is the average difference of a particular lake from the grand
average during the interval of interest [L| ~ (0, <^\a^}\ "T is the difference between the regional
average lake condition during any year and the grand average [T= ~ (0, o2 )]; Ey is the residual
term [Ey ~ (0, o2^; o*res = ozlake^ear + a2index].
In terms of the model components, individual index observations on each lake can be expressed
as illustrated in Table 4-1 . The year means, expressing the regional snapshot each year, are:
Y.j = fi + L + TJ + E.j
and the lake means, expressing the average condition of an individual lake across years, are:
Y,. = p + L; + T. + B,
124
-------
0>
0)
u
CM
HI
UJ
U?
0)
•o
o
8
«
CO
«
o>
c
(S
0)
a
Q)
O
0)
0)
XI
u
CM"
UJ
UJ
«J
•5
.2
"8
in
o>
a
x
UJ
^ (/> .y
5 If
0)
o>
(0
•5-
0)
0 fj
c ..
o g
o?
« S
E|
0) 0)
l§
0)
-c 2
(o
LU
Q)
« £
Q) O
5 O
•S^
"S'S
C! Q
1 g
•a -o
£.£
(Q Qj
J2 £
11
>» ^
CO "O
_JW
1
CM
ul
CM
CM
UJ
CM
111
UJ
CM
u
UJ
uT
U?
0)
X
a
CM
125
-------
Variance of (lake means) is given by: o2|ake + o2year/«; + cP^f, variance of (year means) is given
bV'- ^lake/' + ^year +
-------
Evaluating the potential sensitivity of the EMAP design requires estimates of these variance
components. Few databases contain the spatial and temporal coverage needed for useful esti-
mates, but occasionally, we find databases from which most components of variance can be esti-
mated. In the near term, we may need to estimate some components from independent data
sets. The Vermont Department of Environmental Conservation adopted a lake monitoring pro-
gram that has developed a database on total phosphorus (TP), chlorophyll-a (chl-a), and Secchi
disk transparency (SD) that covers many years (Smeltzer et al., 1989). In general, TP measure-
ments are obtained once each year during the spring; chl-a and SD are obtained weekly or
biweekly during the summer months. Site replicates are also collected, except for TP. Not all
lakes are visited each year, which produces an unbalanced design, but for many years a sub-
stantial number of lakes was visited continuously, so that most cells in a design such as the one
shown in Figure 4-1 can be filled.
We used this database to isolate as many components of variance as we could estimate. We
selected sets of lakes for which data were available over many consecutive years (at least four)
and then selected data from periods corresponding to our July-August index period. We trans-
formed both chl-a and TP to loge, because variances increased in direct proportion to levels of
chl-a and TP. We estimate and report on the resulting variance components, with one exception.
In Vermont, TP was measured during a spring index period and no repeat measurements were
made during this index period, preventing separation of (^\a^ei^Qar and o2index. Many state
programs monitor TP in springtime to characterize trophic condition. Because so many lakes
were monitored over many years, we considered it important to estimate the variance compo-
nents we could and compare them with future summer estimates. We estimated variance compo-
nents using results provided by SAS's General Linear Model (GLM) procedure, summarized in
Table 4-3.
For the EMAP-SW 1991 pilot survey, the only major components of variance that could be esti-
mated were the aggregate index variance and the population variance (confounded by lake-year
interaction). Other components will be estimated as the monitoring program progresses across
years (Table 4-4). Appendix 4A summarizes sampling and analytical methodologies for the indi-
cators of trophic condition.
127
-------
Table 4-3. Summary of Estimates of Components of Variance for Secchi Disk Transpar-
ency (SD), Chlorophyll-a (Chl-a), and Total Phosphorus (TP) Derived from the
Vermont Lake Monitoring Database3
VARIANCE SOURCE
^lake (Population Variance)
o2 (Year Variance
°2|ake*year (Lake-Year Interaction)
^index (lndex Variance)
o2meas (Measurement Variance, a component of
Index Variance
VARIANCE ESTIMATES
SD
4.203
0.007
0.628
0.660
0.191
Chl-a
0.229
0.002
0.178
0.230
0.111
TP
0.412
0.018
0.089b
a Grand means: SD = 4.82 m, chl-a = 10.0^g/L, and TP = 11.7/ig/L Both chl-a and TP were loge transformed before
calculating variance components.
Only the aggregate of <72|ake*yearand ^index could be calculated because single samples were obtained each year.
Table 4-4. Estimates of Variance Components Derived from the EMAP 1991 Pilot Survey3
Variable
Secchi disk transparency
Chlorophyll-a
Total Phosphorus6
No. of
Lakes
75
75
74
No. of
Observations
99
99
97
J^lake+
lake*vear
4.61
1.23
1.084
2
°index
0.092
0.046
0.214
Lakes have been monitored only one year, so only two variance components can be estimated. Chlorophyll-a and
total phosphorus were loge transformed before calculating variance components.
One lake had an extraordinarily high TP value of 8,740 ,ug/L; this lake was excluded from the analysis.
128
-------
4.4 EFFECTS OF VARIANCE ON ESTIMATES OF STATUS
There are two types of uncertainty in our ability to describe the status of a lake population. The
first occurs as a result of drawing a sample from a population, through the sampling process.
This type of uncertainty is characterized by confidence bounds (e.g., 95% upper and lower
confidence bounds around the CDF) calculated with estimates of variance associated with the
Horwitz-Thompson algorithm as the primary method. A description of this type of uncertainty
estimate is not covered in this chapter, because the general approach has been documented
elsewhere (Kaufmann et al., 1988; Kaufmann et al., 1991). Modifications pertinent to EMAP-SW
will be documented in a future publication. The second type of uncertainty arises from the effects
of extraneous variance on estimation of status. These effects are described here.
One way to illustrate the effects of extraneous variance on estimates of status is through a series
of graphs. EMAP will use CDFs (cumulative distribution functions) to characterize populations of
lakes. Under ideal conditions, the CDFs display population variation unencumbered by the
effects of extraneous variation (curve E of Figure 4-2) but with uncertainty associated with the
sampling process, not shown here. An alternative expression of this distribution, with which most
are familiar, is shown in Figure 4-3. The initial illustrations show normal distributions. One of the
goals of the sampling program is to describe the shape of the distribution functions; a normal
distribution might be unusual.
For the yearly snapshot of the condition of a regional population of lakes, the effects of
extraneous variation can be seen as a distortion of the variance structure of the regional
population; larger and larger amounts of extraneous variation spread the CDFs out further,
corresponding to increasing (o2^^ + °2year + CT2res)' illustrated as curves A through D in Figures
4-2 and 4-3. Observe that the mean values remain the same (located where the curves cross in
Figure 4-2, although this occurs only for symmetric base distributions). Also, the precision of
estimating the mean decreases as extraneous variance increases:
e.g., var(mean) = c?l\Jn .
The effects of this distortion are most prominent at one to two standard deviations away from the
mean illustrated by the maximum vertical differences between the distorted curves and the
undistorted curves [see Overton (1989), from which this illustration is derived; more detail is also
given in the EMAP-SW Research Strategy (Paulsen et al., 1991)].
129
-------
ZONE
OF
MAXIMUM
DISTORTION '
-3
BIASED: 29% of the
lakes are in poor
condition
UNBIASED: 16% of
the lakes are in
poor condition
-2
-1
Figure 4-2. Cumulative distribution functions illustrating the effects of increased
extraneous variance on estimates of population status, where d2 = (o2lake
(from Overton, 1989).
130
-------
-3
Figure 4-3. Increasing extraneous variance spreads and flattens the normal distribution.
Curves as in Figure 4-2.
131
-------
One way the effects of the distortion appear is by overestimates of the proportion of lakes with a
condition less than a specified indicator score for scores below the mean, and by underestimates
for scores above the mean. For example, consider the proportion of lakes falling below the con-
dition specified by -1. In the undistorted situation, 16% of the lakes fall below this condition, but
for the distorted situation (curve A, high extraneous variation), this percentage increases to 29%, a
significant overestimate of lakes in hypothetically poor condition. This distortion of the CDFs
means that greater proportions of the lakes appear in the extremes than in the undistorted situa-
tion.
The effects of the distortion on a CDF are a function of the relative magnitudes of population and
extraneous variance. A particular amount of extraneous variance might have a negligible effect on
the CDF representing a subpopulation with large population variance, but the same amount of
extraneous variance might impart significant distortion to the CDF for a subpopulation with low
population variance. We will not know the relative magnitudes of these components of variance
until we begin to obtain information on lakes selected in the probability sample and on important
subpopulations of these lakes.
More complex scenarios can be modeled (see Appendix 4B), with the advantage that we do not
have to make assumptions about the variance structure of the population or the extraneous
variance. For illustration, we use an arbitrary, imaginary population of lakes consisting of three
discrete subpopulations, one with relatively low indicator scores but with high extraneous
variance, one with moderate indicator scores and moderate extraneous variance, and a third with
high indicator scores but low extraneous variance. Population variance overlaps among the sub-
populations. A series of panels (Figure 4-4) illustrates the effects of the components of variance
on population descriptions (CDFs).
• Case 1. Year effects (o2™,.) and residual variation (o2|ake,year + ^jndex) are constant
(or zero); we use these types of simulations to illustrate the effect of increasing
population variance on the shape of the CDFs, which are the descriptors of the popula-
tion characteristics to be depicted by EMAP-SW status summaries. The consequence
of increasing population variance is stretching of the CDFs, as expected (Figure 4-4,
Case 1). If the population exhibited no variation (all indicator scores identical), a
vertical line would characterize the population.
• Case 2. Year effects (d2'^^) operating on all lakes of a subpopulation shifts the entire
CDF identically from its central location, for example, as might be observed after
several years of monitoring (assuming no trends are present; Figure 4-4, Case 2). The
greater the cPyear value, the broader the band within which the CDFs migrate from year
to year. A particular yearly snapshot locates the CDF somewhere in this band; only
after years of monitoring can we develop a reasonable picture of the status of the
population.
132
-------
Values of
the cdf
1.00
0.80
0.60
0.40
0.20
0.00
123456789 10
VALUES OF AN INDICATOR VARIABLE
Values of
the cdf
1.00
123456789 10
VALUES OF AN INDICATOR VARIABLE
Values of
the cdf
1.00
O.BO
0.60
0.40
0.20
0.00
CASE 3
1234567S910
VALUES OF AN INDICATOR VARIABLE
CASE
1
2
3
POPULATim
0.11
1.78
7.11
28.44
2.13
2.13
C
0
0
0.25
1.00
4.00
16.00
0
"RESIOJJU.
0
0
0
0.07
0.3
1.16
T ot;
Figure 4-4. Impact of variance components on the shape of cumulative distribution
functions. Case 1 varies o2,^; Case 2 varies o2year; Case 3 varies o2res.
Parameter values used are summarized in the lower righthand panel.
133
-------
• Case 3. As illustrated for the case of normal distributions, increases in residual
variation (o2^ flatten the CDFs in a pinwheel fashion, and also smooth the sub-
population effects. However, if normal distributions are poor descriptors of the
population characteristics, the pinwheel effect is not necessarily centered at the
population mean (Figure 4-4, Case 3).
The design of the pilot survey allows calculation of the approximate distortion to the population
CDFs for the indicators of trophic condition. A subset of lakes was resampled during the index
period; the results from the revisits are used to estimate (£-mdm, a part of extraneous variance
distorting the CDFs. However, a2^^^ cannot be separated from o2^^ from a single year's
survey, so estimates of distortion are conservative; distortion is likely to be larger than illustrated
here. Variance components are summarized in Table 4-4. For SD, index differences remained
constant across the range of SD measured and the distribution of SD among lakes was
approximately normal, so comparisons with Figure 4-2 are appropriate; estimated d2 = 1.1,
indicating only a slight amount of distortion.
For chl-a and TP, index differences increased with the magnitude of the measurements and
population distributions were more closely lognormal than normal; thus, in both cases data were
log (natural) transformed before variances were estimated. Results cannot be compared with
Figure 4-2, which is based on normal distributions. Instead, distorted and undistorted curves
were simulated with population and variance structure like that estimated from the pilot survey
(Table 4-4 and Figure 4-5). For both TP and chl-a, distortion appears to be relatively minor.
What options are available if extraneous variance is too large? One is to examine the compo-
nents of index variance to determine whether improved sampling or measurement protocols allow
significant reduction in index variance. A second option is to routinely increase the frequency of
revisits to the lake during the index period to refine the estimate of lake condition during the index
period by averaging revisits, but this is only useful if o2.^^ is large. A third option is to estimate
the extraneous variance by revisiting a portion of the lakes during the index period for several
years, until an adequate estimate of the components of extraneous variance is obtained. Then
this estimate is used to deconvolute the distorted CDFs, estimating the true shape of the popula-
tion distribution in absence of extraneous variance. Until we know the relative magnitude of
extraneous variance and its influence, we would be unwise to implement a study on the individual
components of extraneous variance or to select a strategy to reduce it or account for its effects.
During pilot studies, repeat visits to lakes will be routinely included to estimate this variance
component.
134
-------
0
0
Total Phosphorus in New England Lakes
20 40
Total Phosphorus
60
80
Chlorophyll_a in New England Lakes
20
40
Chlorophyll_a
60
80
Figure 4-5. Illustration of the approximate distortion to population CDFs resulting from
observed extraneous variance estimated during the 1991 EMAP pilot survey.
Solid line represents the approximate true CDF; dotted line represents the
distorted CDF.
135
-------
4.5 EFFECTS OF VARIANCE ON TREND DETECTION
Extraneous variance also affects our ability to detect trends. Trends of interest to EMAP may not
be evident until after at least 8 years of EMAP monitoring, although some trends might be detec-
table within 5 years, after the first set of lakes is revisited; trends of major magnitude that could be
detected sooner do not need a design such as EMAP for detection. The following discussion
focuses on a time frame of 8 to 1 2 years. After 8 years, two EMAP cycles will have been com-
pleted, and after 12 years, three cycles. For trend detection, population variance is of no con-
cern; it is removed by the repeat visits to lakes in the 4-year cycles. This is analogous to the
improved efficiency gained from the use of paired difference tests in experimental protocols,
compared to an unpaired protocol. The following explanation first shows the basis of trend detec-
tion in EMAP and then makes several points about the effects of the variance components. At
this point, it ignores some of the technical details, but these are described in Appendix 4B and
incorporated into the explanation and the illustrations later. The version of the model used here
includes annual revisits until year 5 to a subset of year 1 lakes.
To approximate the sensitivity of the design to trend detection, we assume the possibility of a
linear trend and wish to estimate its slope, ft. Refer to Section 4.3 for a description of the
symbols. Consider the EMAP design after 8 years. Each set of lakes would have been visited
twice. For any particular set of lakes, the yearly difference between the first and second visits
(year 5) can be expressed as [symbols defined with Eq. (1)]:
Averaging these differences across lakes in the first lake set (to be visited in years 1 and 5)
produces an estimate of/?:
(Y.5-Y.1)/4=(T5-T1 +E.5-E.1)/4
with variance of:
var(Y.5 - Y.^/4 = (2a2year
Combining years 2 and 6, 3 and 7, and 4 and 8 similarly and averaging those results yields:
136
-------
= [(Y.5 + Y.6 + Y.7 + Y.8) - (Y.., + Y.2 + Y.3 + Y.4)]/4 X 4
4 (2)
In this illustration, the denominator in var(/J) was 32 = 8*4. The factor 4 comes from averaging
the four estimates derived from each set of repeat visits. The factor 8 comes from 2(Xj - X)2,
where X| expresses years relative to some starting year, as in the variance of an estimated slope
in simple linear regression. In the case of a 4-year interval, that becomes (-2)2 + 22 = 8. For 12
years, namely 3 complete cycles of revisits, it becomes 32 = (-4)2 + 42, and after 16 years, it
becomes 80 = (-6)2 + (-2)2 + 22 + 62. The denominator of the variance term for trend estima-
tion clearly increases dramatically over time; this is the source of the power, in designs such as
EMAP, to detect progressively smaller trends the longer the monitoring program is in place.
Inspection of Eq. (2) reveals the relative importance of year effects (o2^^) and residual effects
(^res) f°r describing the potential sensitivity of the EMAP design for trend detection. It allows
rapid approximation of the effectiveness of choices for reducing or accommodating these variance
components. Year effects cannot be controlled by any methodological improvements or by
increased sample sizes. If trend detection sensitivity is controlled primarily by the magnitude of
year effects, trend detection sensitivity increases only as the period of record increases for that
indicator. In this case, expending effort reducing or accommodating o2res would not be cost
effective, because it would not substantially influence the outcome on trend detection. Another
choice, of course, is to select other indicators less influenced by year effects, or develop
explanations of the year effects, such as dependence on precipitation, which can be estimated
and removed.
On the other hand, if residual effects are large relative to year effects, one option for increasing
trend detection sensitivity is to increase the number of lakes monitored; however, a point of dimin-
ishing returns is reached as (d*rej() declines with increasing numbers of lakes sampled. Another
option is to evaluate the components of variance comprising <^-lndex to determine the extent to
which methodological improvements will decrease its magnitude. Recall that o2res = (^y&art\ake +
^index- ^index ls tne onlv component subject to methodological improvements. The effects of
(T2 r*iake can be overcome only by increasing sample size or years of monitoring.
If o^jndex is relatively large, one possible option is to increase the number of times during an index
period that individual lakes are sampled in order to estimate more precisely the lake mean condi-
tion during the index period. This option is worth considering only if the costs of repeat visits to
137
-------
lakes is substantially lower than the cost of sampling additional lakes. Eq. (2) may be slightly
rewritten as:
x 4 (3)
to illustrate the relative merits of revisits compared to sampling additional lakes. Here, r is the
number of visits to individual lakes; r = 1 in Eq. (2). The sensitivity of var(J3) to several reason-
able scenarios of ( and r is illustrated in Figure 4-6. In this example, for situations in which the
cost of revisits is about the same as the cost for sampling additional lakes, the choice clearly is to
sample additional lakes. If the cost of revisits is 1/3 or less than the cost of additional lakes, it
may be worth considering revisits instead of adding more lakes for trend detection purposes. In
any case, with reasonable estimates of variance components, trend sensitivity can be estimated
and options for increasing sensitivity can be evaluated. Tradeoffs such as these are also dis-
cussed in the EMAP-Surface Waters Research Strategy (Paulsen et al., 1991).
We can use the estimates of variance components derived from the Vermont dataset to illustrate
the sensitivity of the EMAP design under a base set of conditions (e.g., 50 lakes per year over a
period of 8 years with a variance structure like that of the Vermont lakes) and to evaluate sensi-
tivity to some possible design choices for reduction in variance components. Eq. (2) yields an
estimate of the variance of the slope. This can be translated into an estimate of the confidence
interval for the slope, having a half-length of 2x the standard error of the slope for a 95% confi-
dence interval. Cast as a null hypothesis, the test is whether the slope = 0. For the null hypothe-
sis to be rejected, the slope must be > 2x its standard error to be detected at a = 0.05; we
should be reasonably certain of detecting trends of greater magnitude. In other words, reject
H: = 0 in favor of H: + 0 if
P - o
s.e. (p)
> 2
with a = 0.05; s.e. is the standard error.
Table 4-5 presents 2 s.e.(J3), a measure of minimum observed trend, which implies real trend, or
least significant trend. Consequently, it summarizes alternatives for increasing trend sensitivity by
increasing the number of lakes visited or by revisiting lakes during the index period to better
estimate individual lake condition. Assuming that the cost of revisiting is about the same as the
138
-------
- I
ll
5
m
I
CM
5
Kl
'> Q) 0)
w * -o
2. £ 0)
*. 2 w
O -H 3
« o> «
™ fll
1 §
-o *
£ g
w —
•c *«
> o
o| gg
O
II
2
UJ u^
1 I
in
2
I
I K'l
*lsl
_ . C\J C
5 *-^o .S
P eo .-: L.
ki
w - o> Q.
BJ
±i ~O > *=
WO - ^»
S 'C 1<.<2
I g«vg
••- "O . O
0 c +
* •- .
^"«
Q) C
TJ (0
•o .S
o £
-
Ic
2
o>
il
139
-------
Table 4-5. Least Significant Trend (m/yr), as Defined in the Text, as Allocating Sampling
Effort to Revisits Versus Additional Lakes3
r
1
2
3
(,
50
0.064
0.057
0.055
100
0.050
0.046
0.044
150
0.044
0.041
0.040
An 8-year time frame is used; trends are Secchi Disk Transparency (SD). This example uses the variance components
derived from the Vermont Secchi Disk transparency measurements. Units are: m/yr = meters per year; l = number of
lakes visited annually; r = number of revisits within a year.
cost of visiting additional lakes and that the variance components are as summarized in Table 4-3
for Vermont Secchi Disk transparency, revisiting lakes clearly would be an inefficient way of
increasing trend detection sensitivity (compare sensitivity at t = 50, r = 2 with sensitivity at 6 =
100, r = 1 (Table 4-5).
Table 4-5 gives least significant trend, but cannot incorporate the ability of the design to detect
trend. This quantity depends on the actual trend present in reality. Table 4-6 summarizes trend
detection power for Secchi disk transparency, chlorophyll-a, and total phosphorus for 8 and 12
years of monitoring and for 50 and 100 lakes visited. EMAP posted a general goal to be able to
detect trends of 1-2%/yr within 10 years, if such a trend were present, with a = 0.2 and /3 = 0.3
(power = 0.7; McKenzie memo dated 7/13/92). Consequently, the reference values for amount of
trend were chosen to be whole or half percentage values as close to 2%/yr as possible and for
which power was not very small or extremely close to 1 (power = probability of observing a sig-
nificant trend when there is a real trend). In contrast, the values in Table 4-5 are based on a
useful approximation, associated with a consistent and fairly good estimator of trend. However,
this approximation is (statistically) inefficient because it does not incorporate information about
annual revisits to lakes, nor does it completely account for the effect of years. The model
described in Appendix 4A provides a smaller s.e. (/?) than that given by Eq. 3.
These results imply that if variance structure is like that seen for the VT data sets, trends of the
desired magnitude with the desired power should be easily detectable for SD; for chl-a, the target
appears feasible, but it may take slightly longer than 10 years. For TP, although trends of the
140
-------
Table 4-6. Sensitivity of EMAP Design for Detecting Trends (%/yr) in Secchi Disk
Transparency, Chlorophyll-a, and Total Phosphorus, Based on Variance
Components Derived from the Vermont Lake Monitoring Database8
Attribute
Secchi disk
transparency
Chlorophyll-a
Total phosphorus
Number of
Years
8
12
8
12
8
12
Number of Lakes Sampled
50 100
2.0
(0.95)
1.0
(0.94)
2.5
(0.44)
1.5
(0.62)
3.0
(0.53)
2.0
(0.65)
1.5
(0.94)
0.5
(0.72)
2.0
(0.58)
1.0
(0.54)
3.0
(0.55)
2.0
(0.67)
Variance components are used to estimate trend detection capabilities are from Table 4-3. Results are summarized as
a %/yr change; power is included in parentheses.
desired magnitude may be detectable within 12 years, power is substantially lower than the
desired level; more years of monitoring will be required to confirm such trends should they be
present. A similar evaluation will be required for each indicator proposed for routine monitoring.
TP demonstrates a case we must be alert for: a2 s = 0.2o2res, so sensitivity increases only
slightly with more lakes visited.
Another choice for improving trend detection is to execute our studies to minimize residual
variance (a2^. About half of o2^ (for the Vermont SD data) consists of o2^^^, a natural
attribute of the lakes; of o2^^, > 70% comes from variation during the index period, also a
natural attribute of the lakes under study. Relatively little variance is associated with the
measurement process compared with these other components; thus, in this case, refining the
measuring process will yield little return (Table 4-3). It may be possible to reduce <^-indeK if within-
lake spatial and temporal patterns can be consistently detected. Then the effects of variance
associated with these patterns can be removed.
141
-------
The improvement in trend detection derived from sampling more lakes must be evaluated against
the improvement in estimation of status derived from repeat visits conducted to deconvolute the
CDFs. During initial phases of EMAP, it will be essential to allocate some effort to revisiting lakes
to estimate this component of variance. It would be reasonable to drop revisits after enough
years have passed to be confident that this variance component is stable; or, it may be important
enough to continue to monitor this variance component as an indicator in itself.
The model described in Appendix 4B can be used to more thoroughly analyze the sensitivity of
the design for trend detection. Sensitivity can be illustrated by describing changes in standard
errors of estimates; the resulting curves typically approximate decaying exponentials. However, of
greater utility in communicating the sensitivity for trend detection is a set of power curves,
generalizing the values in Table 4-6, that demonstrate the probability of detecting a trend of a
specified magnitude under specified levels of a, variance, and sample size. To illustrate trend
sensitivity, we again use the variance components estimated for Vermont Secchi depth
transparency but cast them as if they were representative of a regional population or
subpopulation of lakes sampled in an EMAP context.
As was the case for illustrating the influence of components of variance on descriptions of status,
we use a series of illustrations that vary different elements of the model individually (Figures 4-7,
4-8, and 4-9). The population characteristics and variance structure used here are derived from
Vermont's SD database and from the variance components we have described. Population vari-
ance has no effect on trend detection because its effects are removed by the repeat visits in the
design; a is set at 0.05 for the illustrations.
• Case 1. Power clearly increases as the magnitude of the trend to be detected
increases (Figure 4-7). These kinds of curves can tell us how many years may be
required to detect specified trends for variance structures characteristic of selected
indicators and subpopulations of interest. We can also use these curves to
demonstrate the magnitude of trend to be detected within a specified number of years
at a selected power. In some cases, trends detected at low power might be enough to
trigger management actions because the risk associated with delay in action if such
trends are present might be high. On the other hand, if risks are low, it would be
reasonable to continue monitoring to confirm the weakly detected trend.
• Case 2. Year effects (o2^^) have a pronounced influence on the ability to detect
trends (Figure 4-8), and few design alternatives allow us to minimize these effects. If
we select indicators for their monitoring importance, then high year effects mean that
we either monitor for longer periods before making decisions about trend detection or
accept lower power associated with any potential trends detected. It is likely that some
indicators will be more sensitive to year effects than others, in which case, other factors
being equal, indicators exhibiting lower year effects are desirable. It is also possible
that within populations of interest, there may be subpopulations whose variance
structure minimizes year effects. For trend detection purposes, it will be important to
determine these subpopulations.
142
-------
POWER
1.00
0.80
0.60
0.40
2 4 6 8 10 12 14 IB 18 20
YEARS
Figure 4-7. An illustration of the relationship between power and magnitude of trend
detectable and years of monitoring. From left to right, curves are for trends of
approximately 2%/yr, 1.5%/yr, 1%/yr, and 0.5%/yr.
143
-------
II
1
2 -r-
$1
10 n:
IB
O OL
.(3 S
5
S in
It
10 K
Sffi
5
w
o
o
.S2
(A
£
n
v
>.
01
TO
a
«
ca
?£
II UJ
(A
Q
o
8
CO
•o
0)
(0
1
"S
•o
I
a
2
3
O)
144
-------
w
£
1
o
u
2
o>
.c
(0
(8
a
w
a
o
u
a>
CO
T3
0)
o
0)
fl)
•o
2
I
a.
145
-------
• Case 3. The power of trend detection is not as sensitive to the magnitude of residual
variance as it is to year effects. For example, compare YEARERR = 0.5 in Figure 4-8
with RESIDUAL = 0.5 in Figure 4-9. Also compare YEARERR = 0.1 in Figure 4-8 with
RESIDUAL = 2 in Figure 4-9. These cases, as well as other scenarios, indicate the
extent to which power is dramatically more sensitive to o2&a{ than to o^res.
The basic features of the trend detection model used in the foregoing scenarios have been
assembled, as described in Appendix B. The next step is to convert the model into a form that
indicator leads can use to explore the sensitivity of the design and design options for particular
candidate indicators. A future step will then be to evaluate the responsiveness of the various
candidate indicators together as a component in the determination of the set of indicators that
would be used for routine monitoring.
4.6 A NOTE ON AVAILABLE DATABASES
Throughout the development of EMAP-SW, we have been aware of the need to identify databases
that could be used to assist us in evaluating the potential for various biological indicators to serve
as candidates for routine monitoring. Up to now, we have not performed a broad search for such
databases because we had not described the kinds of data most useful for our EMAP studies and
were unwilling to expend much effort along these lines until we could more explicitly define what
kinds of databases would be best and how we were likely to use them. Now we are closer to
specifying those types of databases, as described earlier in this chapter. We must also be cogni-
zant of the costs associated with acquiring databases that require considerable modification and
clean-up; for example, electronic databases are desirable; data contained in file cabinets or mis-
cellaneous reports are undesirable; databases derived from consistent monitoring protocols over
many years are desirable; databases compiled from many separate studies with differing pur-
poses and unknown or undocumented quality are less desirable.
Typical of the kinds of databases available and that we have begun to acquire are those
developed by state agencies, or other institutions with long-term monitoring interests, such as the
Academy of Natural Sciences at Philadelphia, or individual research scientists who have been
able to maintain long-term monitoring programs. A brief description of the kinds of databases we
have acquired follows.
In the northeast, most states monitor lakes for indicators of trophic condition. For this report, we
relied on a database acquired from the Vermont Department of Environmental Conservation.
Other states in the northeast have also sent us copies of their databases. These are being
146
-------
compiled and evaluated for their utility in estimating the important variance components. Typical
of these monitoring programs is the lack of data on a consistent set of lakes over many years;
often state agencies target a selected set of lakes for one year or a few years, then switch lakes
as interests and priorities change. Our strategy is to cull these databases for sets of lakes that
have been monitored consistently over many years, then conduct the variance analyses on these
subsets. The number of lakes and years over which data are available varies by indicator and by
state. It would be ideal if we could extract a set of lakes from the various state databases that
could be analyzed across the entire northeast.
During the early 1970s, the U.S. EPA conducted the National Eutrophication Survey (NES). The
survey was limited to one year in any region, preventing its use for estimating some components
of variance; however, multiple sites within a lake were sampled and most lakes were sampled
three times during the year. This database is being examined to estimate site scale variance and
seasonal variance. In general, the repeat visits within the year occurred over a longer time
window than that anticipated for conducting EMAP lake monitoring, so variance estimates derived
from an analysis of the NES data should exceed the estimates we encounter for EMAP.
A few states have ongoing monitoring programs that target biological indicators. To our
knowledge, the Ohio Environmental Protection Agency (OEPA) has the most complete and
extensive biological monitoring program among states. In combination with routine physical and
chemical monitoring of streams throughout the state, OEPA also collects macroinvertebrate and
fish assemblage data. These biological data are compiled into various indices of condition of the
streams. This monitoring program has been in place for more than 10 years, covering more than
2,000 stream sites. We have worked with this database for many years for various purposes.
With regard to its utility for EMAP, this database has one major shortcoming. The majority of sites
are monitored only during a single year, so we are unable to estimate o2 , an important com-
ponent for trend detection. Furthermore, population variance is confounded by an unknown
contribution by the stream*year interaction effects, separable only by revisits to streams across
years. However, the database allows us to estimate index variance because repeat visits during
the summer months are part of the routine fish monitoring. We have encouraged OEPA to begin
collecting data on a subset of streams across years; data from such studies will be valuable for
the states as well as for EMAP-SW.
Many other state agencies are beginning to monitor biological indicators of stream condition on a
routine basis, stimulated by EPA's recent policy on the development of biological criteria. A few
147
-------
other states, besides OEPA, have been monitoring biological indicators of stream condition
routinely, for example, Maine and North Carolina, but with the same shortcomings, for EMAP
purposes, as the OEPA databases. As we begin our pilot studies on streams, we will acquire
more of these databases.
The Academy of Natural Sciences at Philadelphia (ANSP) conducts biological monitoring pro-
grams on a variety of aquatic ecosystems (David Hart, pers. comm.). Some individual systems
have been monitored routinely for decades. We recently completed an inventory of the moni-
toring programs ANSP has conducted and distributed an electronic database to appropriate indi-
cator leads for further investigation of its utility for individual indicators.
Individual investigators often compile impressive databases for particular indicators on particular
systems. These databases tend to be limited in geographic extent, thus they do not allow estima-
tion of expected population variation or calculation of variance components across a broad range
of sites. One such database that we are now compiling pertains to the Wabash River (primarily
mainstem) in Indiana (Dr. James Gammon, pers. comm.). In this case, fish assemblages at many
sites along the Wabash mainstem were monitored for more than 10 years. In addition, a variety
of metrics have been calculated based on the fish assemblage raw data and are part of the data-
base. This type of database will be quite useful for estimating all the variance components; the
survey design is like the one described in Figure 4-1. For EMAP-SW purposes, the only limitation
is the relatively limited geographic scope of the survey. We will use this database as an example
of the kinds of databases most useful for our purposes.
We will continue to identify and acquire databases that appear useful for EMAP-SW purposes.
Our experience so far has been that state agencies and individual investigators are quite willing to
make the data sets available and encourage us to use them. We recognize the extensive amount
of effort that has gone into obtaining the data and compiling the databases and we respect the
proprietary interests of the principal investigators.
4.7 IMPLICATIONS FOR PILOT SURVEYS AND INDICATOR SELECTION
The preceding discussion develops a framework that identifies and describes components of
variance that are important for estimating status and detecting trends in indicators of condition of
ecosystems measured on sites selected with a probability-based survey design such as the one
for EMAP. This framework can be used by indicator leads and others as a guide to aid in
148
-------
efficiently targeting research and pilot activities on elements important for surveys such as EMAP.
In particular, the framework:
• Illustrates the kinds of data sets most useful for estimating the important variance
components. Therefore, indicator leads can search the literature and cull their contacts
for information on such data sets. It is possible that surveys of the type needed have
been conducted by scientists and agencies in other countries.
• Suggests that available data sets will give only general insight into the magnitude of the
components of variance of importance to EMAP monitoring. The magnitude of some
variance components may be specific to region or lake type, so estimates derived from
other regions or other types of lakes and streams might be misleading. In addition, the
use of different methodologies for data collection and analytical procedures may result
in misleading variance estimates, especially in attempts to combine data collected by
various investigators into combined data sets. There is a risk of unknown magnitude if
decisions are made based on data obtained from studies that are not like EMAP.
• Establishes a set of tools that indicator leads can use to evaluate how well particular
indicators can be expected to perform for status estimation and trend detection. These
tools allow the dissection of variance components with an objective of allocating
sampling effort most efficiently, and can be used to evaluate current pilot results and
make necessary modifications as needed. These tools will also assist in the design of
particular studies, as for example, a focused study on important subpopulations of
lakes and streams for detection of trends in acid neutralizing capacity associated with
reduction in sulfate emissions mandated by the Clean Air Act Amendments.
• Indicates the need to factor into the routine pilot and demonstration studies collection
of data to estimate variance magnitudes. It will be necessary to estimate the major
components of variance as a fundamental part of the ongoing monitoring activities until
enough data have been gathered to estimate the stability of the variance components.
Some variance components can be estimated only as part of ongoing monitoring activities based
on a consistent survey design run over many years. These variance components are natural
features of the lakes and streams, and, although they interfere with status estimation and trend
detection, it is important to estimate them in their own right. Some will be estimated by virtue of
the routine sampling design (year effects; lake-year interactions). It will be necessary to factor
special studies into the pilot surveys to answer specific variance questions not answered as part
of the basic survey structure. As the routine monitoring unfolds, and variance estimation stabil-
izes, only then will we have a sound basis for determining how well status can be characterized
and trends detected. It is therefore important that pilot studies proceed to the monitoring of
probability-selected lakes and streams with all due expediency. Results of evaluating available
data, such as the Vermont data set, give us insight into what to expect. We have performed
some preliminary analyses on other state databases that contain survey results of indicators of
trophic condition. These results will be available in future reports.
149
-------
4.8 SUMMARY
This chapter describes the variance components that impart important influences on EMAP's
ability to describe status and detect trends. A general linear model is used as a framework within
which we characterize and estimate the variance components; illustrations show how different
variance components distort estimates of status and affect trend detection. A model for evalu-
ating trend detection capability (expressed as power to detect trends of specified magnitudes),
described in this chapter, can be used to examine the potential of the EMAP design to detect
status and trends under different combinations of variance structure, years of monitoring,
magnitude of trend, and Type I and Type II errors. A database on indicators of trophic condition
available from the Vermont Department of Environmental Conservation is used throughout the
chapter to illustrate the estimates of variance and their influence on status estimation and trend
detection as a real world example.
The analyses presented suggest that the EMAP objective of detecting a 2%/yr change in 10 years
(with a = 0.2 and power = 0.7) is feasible for both Secchi disk transparency and chlorophyll-a,
but several more years will be required before such trends can be detected in total phosphorus, if
the variance structure of data collected in regions of interest to EMAP is similar to that of the
Vermont database used in the illustrations.
150
-------
APPENDIX 4A
Methods for Total P, Chlorophyll-a, and Secchi Depth for EMAP-SW 1991 Data
Analyte
Total Pb
Chlorophyll-ad
Secchi Depth
Methoda
Acid-persulfate digestion with
colorimetric (molybdate blue)
determination
Field filtration of 250 to 1000 mL
of water; spectrophotometric
analysis
20-cm black and white Secchi
disk with calibrated line.
Estimated as depth of
disappearance plus depth of
reappearance, divided by two.
Reference
Skougstad et al.
(1979); U.S.EPA
(1987)
APHA Standard
Methods (1989),
Method 10200 H
Lind (1979);
Chaloud et al.
(1989)
Method
Detection Limit
0.51 ng/Lc
2/zg/Le
—
a Field methods are described in Peck (1991).
b Precision of total P data was estimated from field performance evaluation samples and labora-
tory replicates as either the standard deviation (for concentrations < 20 fig/L) or as the
coefficient of variation (CV) (standard deviation divided by the mean). For 1991 data, the mean
CV of laboratory replicates was 2.1% and the mean standard deviation of field performance
evaluation samples was 1.4 fig/L. Bias was estimated from quality control check samples as
the difference between the mean value of repeated measurements and the target value of the
QC sample, divided by the target value. Bias from the 1991 data for total P was -0.45%.
0 Method detection limits were estimated from low concentration level quality control check
samples and are calculated as a Students' t value at a significance level of 0.01 and n-1
degrees of freedom times the standard deviation of the repeated measures of the low-level QC
sample (equation 5-1 in Peck, 1991).
Precision of chlorophyll-a data was estimated with field performance evaluation samples and
laboratory quality control samples as the CV; CV for the 1991 data ranged from 10% to 20%.
e Method detection limit was based on the minimum absorbance for a 250-mL sample.
151
-------
APPENDIX 4B
A FLEXIBLE LINEAR MODEL FOR STATUS ESTIMATION
AND TREND DETECTION
In this discussion, we use a flexible linear model and the tools of experimental design to illustrate
the effects of various sources of variability on descriptions of the status of a population of a
resource type, such as a population of lakes, and on trend detection capability. In doing so, we
do not suggest that the approach exactly depicts every detail of the sampling situation. Neverthe-
less, the flexibility of the model allows us to explore the importance of various assumptions of the
basic EMAP sampling design and characteristics of the indicators and resource populations of
interest that influence status description and trend detection.
Consider an indicator of ecological condition (Y) evaluated at a lake in a particular year. Because
the lakes have been randomly selected, we assume their effects are random in the sense of a
random effect in a linear model. This random contribution of each individual lake may remain
constant across years, but in other situations, part of a lake's effect may decay across years. For
example, consider a deep water lake in which only 5% of the water volume is replaced annually.
If such a lake is above the regional average one year, it is very likely to remain above the regional
average the next year and for several following years. Likewise, because the years selected are
of no intrinsic interest, we also assume this factor has a random effect, but years are consecutive.
Part of the effect of years will be a consequence of atmospheric conditions, which may have
some (auto)correlation through time. For both lakes and years, the correlation between
responses will diminish, the farther apart in time the responses are. We model their covariance
structure as a stationary Markov process. Machin (1975) used this covariance structure; Morrison
(1970) investigated more general aspects of repeated measurements using the same covariance
structure; Rao and Graham (1964) even used this structure in a study of sampling designs.
If lakes are revisited in some organized manner, as we anticipate here, sets of lakes will be visited
in the same pattern of years. A reasonable plan, and the one assumed here, is that lakes to be
visited will be partitioned into disjoint sets for which the visit pattern will be the same. Thus, if any
lake in a lake set is visited in a year, all lakes in that set will be visited that year. Let / index these
sets of lakes, / = 1, 2, - , f, let/ index years, / = 1,2, - , t; and let k index lakes within a lake
set, k = 1, 2, - , HJ. If the lake effects remain constant across years, then the random contribu-
tions to an indicator of ecological condition (Y|jk) at a particular lake (/, k) in a particular year (/)
can be modeled as:
152
-------
Yiik ~ Lik + Ti + Eiik'
0)
The three terms on the right in Eq. (1) represent the lake, time (year), and residual effects,
respectively. The residual term includes all effects not otherwise modeled, including any lake by
year interaction; more generally, the residual effect includes all random contributions specific to a
lake during a particular year, including measurement error. If the effect of lakes decays with
years, the lake term in Eq. (1) must be indexed also by/; this subscript will be enclosed in paren-
theses, however, to preserve the identification of lakes by their two subscripts, / and k, yet allow a
correlation of lake effects across years of less than the unity implied by Eq. (1):
Yijk ~
T
(2)
Suppose the random components in Eqs. 1 and 2 have variances a,^, a, and o^, and are
mutually uncorrelated except for those correlations among the Sj/«k and among the T= explicitly
discussed later. For now, we assume the lake components are uncorrelated between lakes;
possible effects of an anticipated spatial correlation between lakes is the topic of a subsequent
consideration. Define the matrix function Fn(p) by:
1
p
P2
P3
n-1
P
1
P
P2
P"-2
P
P
1
P
P"-3
... p-'
-. P"-2
... p-3
... p-4
... 1
i
If we denote the autocorrelation between year effects by pT, then the covariance among the year
contributions (Tv T2, - , Tt) can be modeled as Zyear = c^yeaft(p-^; likewise, the covariance
among the responses at the same lake in consecutive years (namely U^, U^, - Lj,^) can
be modeled as 2|ake = (^lakorfafafa). This assumed structure is equivalent to requiring that:
153
-------
i 2
= PsLi(j-1)k + p-Ps Fijk< (3)
where the Fj:k represent new independent parts of yearly random contributions to the lake effects
satisfying var(Li/1jk) = var(Fj-k) = <^\a but tne ^ijk are otherwise uncorrelated. This correlation
structure provides a way of investigating the impacts of possible correlations under our expecta-
tion that correlations between L^'s (or Tj's), if present, will decline as the L^ (or Tp become
farther apart in time. Our use of this structure does not imply that we believe it exactly models
the correlation structure that exists in reality; it provides a suitable approximation that will support
an investigation of possible effects of correlation. The case /OT = 0 and ps = 1 deserves special
notice because it corresponds to a randomized complete block experimental design with additive
treatment (l_ik) and block (or year, Tj) effects. Thus, in practice, we expect pT to be fairly small,
say less than 0.25, and P|ake to be fairly close to 1 .
If all lakes in a lake set are visited in the same year, the overall behavior of status and trend can
be evaluated from the means in a table indexed by lake set and year. The averaging described
here could be taken over the entire population, or over a relevant subpopulation. Thus, consider
the vector of lake set-year means organized by lake set within year:
Y' = 0^11' Y21- •"' Y/j, Y12, Y22, -, Y/2, -, Y/f)'.
The general variance-covariance among these means can be represented in scalar form, using
the Kronecker 6, as:
(4)
cov(Y,y, Y/T) = 6,ro*akePl^ ' In, + o2y
or in matrix form, using the Kronecker product of matrices, as:
(5)
cov(Y) = <£ =
where 1n denotes a column of n ones and D(-) denotes a diagonal matrix having the indicated
values on its diagonal. This variance-covariance matrix clearly depends on all of
^res' Piake- /^year- and ni' n2> '" ' "lake- and the Pattern of cells containing means.
154
-------
Now consider the estimation of status and trend as if the trend were the same at all lakes and use
generalized least squares incorporating the variance-covariance matrix given by Eq. (5). If we
assume that the E(Yijk) are finite, then Y are linearly sufficient in the sense of Barnard (1963).
Consequently, the generalized least squares estimator, based on all of the observationsYyk,
actually can be expressed as a linear combination of the lake set-year means Yy. Let X denote a
(regression) matrix having as many rows as Y has means, a column of ones, and a column of the
year index associated with each mean. Estimates of the regression coefficients are given by:
0 = (X'*'1X)'1X'*'1Y, (6)
and their variances and covariances are given by:
(X'*'1X)'1. (7)
The second element of /?, the slope, clearly gives a measure of the trend. Status can be
estimated by the value of the fitted line at the most recent time, namely, at time t. If the 2x2
matrix C0 has a first row containing 1 and t, and a second row consisting of 0 and 1 , then C0/)
estimates status and trend and has a variance-covariance matrix given by:
C0 (X'*-1X)-1C6. (8)
When a sampling plan is imposed on this context, it specifies the n^'s and the years in which
specific lake sets are visited. Specifically, n^-, the number of observations taken in each cell of the
conceptual mean table, is either 0 or n-t, as specified by the sampling design. Consequently, a
proposed sampling design can be defined in terms of the number of lake sets, the n-t, and the
years in which each lake set is to be visited. If only some of the n-{- > 0, all of Y, X, and need
to be replaced by Y*, X*, and *, where these represent merely that subset of rows (and
columns, if appropriate) of the respective vector and matrices corresponding to cells where
Given values for a^, o^^, o^, Piake1 Pyear vanances of the estimate of the status and trend
can be obtained from Eq. (8). We used values from the Vermont set to choose values for the
variance components: P|ake = 1 and pyear = 0, reasonable limiting values.
155
-------
CHAPTER 5
PROTOTYPE ANNUAL REPORT
5.1 INTRODUCTION
The primary objective of the Environmental Monitoring and Assessment Program (EMAP) is to
describe the condition of our nation's ecological resources on a regional scale. The Program
provides data on the current status and extent of resources, and on changes and trends in
condition. The production of monitoring data is not an end in itself, however; the data must be
available as useful information in a timely manner. EMAP has set a goal of reporting the results
of each year's field efforts, within 9-12 months. These annual statistical summaries must be more
than extensive data tables. The information must be reported in a way that is readily understand-
able by policy makers, scientific audiences, and the public, including Congress. In order to pro-
duce reports within the specified time constraints, however, detailed interpretation of the results is
not possible. A balance must be struck between summaries that can be produced in a relatively
automated manner, yet be useful over time, and detailed interpretive reports that often take years
to produce.
This chapter of the pilot report is intended to be a prototype of an annual statistical summary. It
is limited to a few response indicators sampled at lakes selected by EMAP's probability design
during 1991. It also contains only one year's data. A fairly complete description of condition will
be based on data from the full four-year cycle. We have included the introductory material that
we expect to produce every year with the annual statistical summary. Comments on the format
and clarity of the presentation from various user groups will be most welcome and of tremendous
value in improving these summaries1.
5.2 OVERVIEW OF EMAP
The design and implementation of EMAP resulted from a recommendation by EPA's Science
Advisory Board that EPA (1) implement a program to monitor ecological status and trends and
(2) develop innovative methods for anticipating emerging problems before they reach crisis pro-
portions. EMAP is therefore being designed to estimate the location, extent, and magnitude of
Send comments to Dr. Steven G. Paulsen, U.S. EPA Environmental Research Laboratory, 200 SW 35th Street, Corvallis,
OR 97333.
157
-------
degradation or improvement in indicators of the condition of ecological resources at regional and
national scales. This integrated monitoring program is addressing the following types of critical
questions:
• What is the current status and extent of our ecological resources (e.g., estuaries, lakes,
streams, forests, arid lands, wetlands), and how are they distributed geographically?
• What percentages of the resources appear to be adversely affected by pollutants or
other human-induced environmental stresses?
• Which resources are degrading, where, and at what rate?
• What are the relative magnitudes of the most likely causes of adverse effects?
• Are adversely affected ecosystems improving as expected in response to control and
mitigation programs?
5.2.1 EMAP Goals and Objectives
To provide the information necessary to answer these questions, EMAP has three major
objectives:
1. To estimate the current status, changes, and trends in the extent (e.g., kilometers of
streams, hectares of wetlands) and in indicators of the condition of the nation's
ecological resources on a regional basis with known confidence.
2. To monitor indicators of pollutant exposure and habitat condition and seek associations
between human-induced stresses and ecological condition that identify possible causes
of adverse effects.
3. To provide periodic statistical summaries and interpretive reports on ecological status
and trends to the Administrator and the public.
The EMAP monitoring design allows researchers to make statistically unbiased estimates of the
status of ecological resources, trends in ecological condition, and associations among indicators
of ecological condition. The probability design allows quantification of uncertainty over regional
and national scales for periods of years to decades. The sampling approach offers a snapshot of
conditions during an index period, rather than a detailed view of cycles within a year. This
approach places some constraints on indicator selection and requires clear definition of assess-
ment objectives and strategies for using information from multiple indicators.
5.2.2 EMAP Indicator Strategy
Society values several aspects of our ecological resources. These societal interests are some-
times described as environmental goods and services, or beneficial uses of ecological resources.
158
-------
Biodiversity, sustainability, aesthetics, and production of food and fiber are examples of these
societal interests. Society wishes to protect, preserve, and restore these kinds of biologically
important, often conflicting interests. However, they are often difficult to quantify (e.g., sus-
tainability, aesthetics, biodiversity). Indicators, as used in EMAP, are objective, well-defined, and
quantifiable surrogates for important environmental and ecological values.
A long-term goal in EMAP is to measure biological indicators to determine whether resources of
interest are in nominal or subnominal (acceptable or unacceptable) condition relative to a set of
environmental or ecological values. Specific values of EMAP-Surface Waters (EMAP-SW) are
described later. As an end result, EMAP should estimate the extent (i.e., number, length, surface
area) of the resource of interest (e.g., lakes, streams) and the proportion of the resource in
acceptable or unacceptable condition. In addition, the probable or possible cause of the poor
conditions should be identified. Within EMAP, we define four categories of indicators to accom-
plish this goal—response, exposure, habitat, and stressor indicators.
1. Response indicators quantify the overall biological conditions of ecological resources
by measuring either organisms, populations, communities, or ecosystem processes as
they relate to values of concern.
2. Exposure indicators quantify the levels of pollutants to which the biota might be
exposed, such as toxics, nutrients, or acidity.
3. Habitat indicators represent conditions on a local or landscape scale that are necessary
to support a population or community (e.g., availability of snags, rocky stream bottoms,
or adequate acreage or connectivity of wooded patches).
4. Stressor indicators reflect activities or occurrences that cause changes in exposure or
habitat conditions and include pollutant export, management activity, and natural
process indicators.
EMAP staff will use the response indicators to determine the condition of the systems, as
described by the terms nominal and subnominal. Subnominal implies an unacceptable condition,
and if a certain proportion (as yet undefined) of the systems are in subnominal condition, reme-
dial action might be initiated, such as a specific management action focused on the types of
lakes or streams affected. We will rely upon the exposure and habitat indicators to suggest
possible cause for subnominal conditions on a regional scale. Stressor indicators can then be
used to confirm or support the suggestion of probable cause, or they can be used alone as
causal evidence for poor condition. It has been clear from the outset that defining nominal/
subnominal conditions for most systems is a long-term process that will require several years to
achieve.
159
-------
5.2.3 Reporting in EMAP
An important objective of EMAP is to publish annual statistical summaries of indicators of the
conditions of ecological resources and the exposure to stressors, as well as periodic interpretive
reports. EMAP will produce data and information at varying levels of detail. Often monitoring
programs or surveys report only the most detailed level of data possible. The reader is then left
with massive amounts of data, but little real information, and must condense and synthesize the
data into information independently. There are distinct advantages to this approach, but it is
useful to a limited audience. On the other hand, as we summarize data into various levels of
information, increasing amounts of interpretation are required. This condensation requires careful
analyses and evaluation of the data in order to prevent development of spurious associations and
erroneous conclusions. This level of care requires several years of effort and is not conducive to
providing the users of the information with timely results.
EMAP will balance these conflicting needs by producing annual statistical summaries, timely, yet
with enough data reduction to be useful to a range of audiences, and by producing periodic
(probably every four years) interpretive reports with the full synthesis and interpretation needed
and completed in a reasonable time frame.
5.3 EMAP-SURFACE WATERS
5.3.1 Legislative Mandate
Recognizing the extensive degradation of surface waters, Congress established their protection in
1972, as a priority with the passage of the Federal Water Pollution Control Act (P.L. 92-500). The
primary objective of P.L. 92-500 was to restore and maintain the physical, chemical, and biologi-
cal integrity of the nation's waters. An interim goal was to provide for the protection and propa-
gation of fish, shellfish, and wildlife, and for recreation in and on the water [Section 101 (a)].
Several other sections of the Act relate to EMAP objectives. Section 105(d)(3) requires EPA to
conduct, on a priority basis, an accelerated effort to develop and apply improved methods of
measuring the effects of pollutants on the chemical, physical, and biological integrity of water.
Section 304(a)(1) states that EPA shall develop and publish criteria on the effects of pollutants on
biological community diversity, productivity, stability, and eutrophication. Section 305(b) man-
dates biennial reports that assess the extent to which all waters provide for the protection and
propagation of a balanced community of aquatic life.
160
-------
Since 1972, the Act has been further strengthened. The Water Quality Standards Regulation (U.S.
EPA, 1983) requires that states designate aquatic life uses consistent with the goals of the Act,
provide criteria sufficient to protect those uses, and establish an antidegradation policy protective
of waters of outstanding quality.
The Water Quality Act of 1987 amends P.L. 92-500 and emphasizes ambient standards and
assessments as the driving forces behind further pollution abatement. Section 303(c)(2)(B) allows
states to adopt criteria based on biomonitoring (a key EMAP component). Section 304(1)(A)
requires listing of waters not expected to attain protection and propagation of balanced biological
communities. Section 304(m)(2)(g) requires EPA to study the effectiveness of applying best
available pollution controls for protecting balanced communities. Section 314(a) requires trophic
classification of all publicly owned lakes and an assessment of the status and trends of water
quality in those lakes. Section 319 mandates identification of waters that cannot protect balanced
aquatic communities without nonpoint source pollution controls.
Other legislation requires assessment of environmental risk to aquatic communities. Of particular
importance are stressors with potential regional impacts that fall under the Clean Air Act Amend-
ments, the Federal Lands Policy and Management Act, the National Environmental Policy Act, the
Resource Conservation and Recovery Act, and the Federal Insecticide, Fungicide, and Rodenti-
cide Act. In addition, the Endangered Species Act mandates assessments and protection of rare
and threatened species and the National Forest Management Act of 1976 requires conservation of
animal diversity.
There is certainly a need to evaluate the effectiveness of each of the laws and regulations, but
there is also a need to determine if all of these regulations in the aggregate are resulting in the
desired effect of protection of our aquatic resources. By describing the condition of our lakes and
streams, EMAP-SW will determine the cumulative effectiveness of our protection efforts.
5.3.2 Issues and Problems
The variety of hazards to inland surface waters can be grouped into four broad categories for
ease of summary and interpretation; several common types of hazards fall within each of these
groups.
161
-------
• Physical-Chemical alterations
Nutrient enrichment
Contamination (both point and nonpoint, as well as toxic and nontoxic)
- Atmospheric pollutants (e.g., acidic deposition, other air toxics)
Thermal alteration
Global warming
• Physical habitat alterations
• Hydrologic alterations
Flow modification
• Biological alterations
Introduced species
Harvest imbalance (overstocking or overharvesting)
Each of these general categories of perturbations presents a potential threat to surface water
condition, in the context both of traditional concepts of beneficial uses and of ecological integrity.
5.3.3 Surface Water Indicator Strategy
To be effective, monitoring data must be related to perceptions regarding aquatic condition and
represent issues of concern to aquatic scientists and the public. In ecological risk assessment,
these perceptions are called endpoints. One approach to determining these endpoints is to iden-
tify designated beneficial uses of aquatic resources. A designated use is attained if the system is
being used for its intended purpose, such as habitat for aquatic life, fishing, water sports, aes-
thetics, navigation, or water supply. This approach has been ineffective for protecting biological
integrity, because many designated uses, and the indicators for assessing use impairment, have
little relationship to biological condition or integrity. Often the uses are poorly defined. For
example, "warmwater fish" or "fishable" may be considered attained whether the water supports a
few carp or abundant smallmouth bass, and "aquatic life" may likewise mean bluegreen algae or
arctic char.
A host of perspectives and opinions exists concerning the appropriate environmental or ecologi-
cal values on which a monitoring program such as EMAP should be based. Instead of selecting
"beneficial uses" as the endpoints or values on which to base assessments and indicator selec-
tion, primary consideration was given to those ecological attributes or environmental values for
which we manage surface waters. These were perceived to be:
162
-------
• Biological integrity
• Trophic condition
• Fishability
The protection and restoration of the value "biological integrity" is mandated in P.L 92-500, but
biological integrity was not defined by the legislation. However, biological integrity has been
described as "a balanced, integrated, adaptive community of organisms having a species compo-
sition, diversity, and functional organization comparable to that of natural habitat in the region"
(Karr and Dudley, 1981). Fishability and trophic condition also appear as important values
expressed in P.L. 92-500. These three values, though not parallel in perspective, are identified
separately because managing for one is often to the detriment of the others; this is otherwise
known as conflicting beneficial uses.
Ideally, a single indicator of each of the listed endpoints would form the basis for the annual
statistical reports on the condition of aquatic resources. There would be one indicator for the
overall condition of biotic integrity, one indicator for trophic condition, and one indicator for
fishability. However, at this time, the statistical summaries report on some subindicators, such as
chlorophyll-a (as an indicator of trophic condition). Much work has to be done over the coming
years to integrate the various proposed measures that describe biological integrity and fishability.
By adopting the perspective that we are striving for a single indicator or index (aggregation of
other indicators) for each of the ecological values, EMAP-SW concentrates on major values for
which we manage systems and recognizes that they often compete. We may eventually be able
to demonstrate with more consistent data the ways in which they conflict. We have recognized
from the outset that we are not in a position to adequately describe condition with respect to
these three endpoints, especially the nominal/subnominal aspects, or to compartmentalize the
categories of possible cause as outlined here. It is, however, important to keep these long-term
goals in mind while we evaluate the results of individual indicators presented in the annual
summaries.
For fishability and biological integrity, the general categories of perturbations that have been
discussed (e.g., chemical, hydrologic, biological, and habitat alterations) are the categories for
probable cause of subnominal condition. Development of habitat, exposure, and stressor indica-
tors is progressing in concert with the development of response indicators.
163
-------
5.4 EXTENT OF SURFACE WATERS
One of the objectives of EMAP is to estimate the extent and spatial distribution of the ecological
resources of interest such as lakes or streams. Estimates of the number, length, or surface area
of these resources will be generated along with estimates of uncertainty. One advantage of a
probability design such as EMAP is that maximum flexibility is retained for post-stratification to
evaluate the results. In EMAP-SW, we will consistently report the results by both geopolitical
regions (i.e., EPA regions) and some form of more meaningful ecological regions.
5.4.1 Lakes
The lake component of EMAP-SW will be the first to be implemented. For extent, the aspects of
interest are the number, total area, and spatial distribution of lakes. As a lower limit of resolution,
the program has chosen lakes of one hectare in surface area. Thus any estimates of extent or
condition will apply only to lakes greater than one hectare.
Whenever a particular resource type is described, complications in definition always arise. The
definition should be consistent across the United States; however, we recognize that the issue is
not clear cut. There are regional differences in what local residents and legal authorities consider
a lake. The definitions are somewhat dependent upon the abundance, or lack thereof, of lakes in
that region. For example, a waterbody of 1-2 ha may be of great significance to local residents
and wildlife in arid regions of the west, but barely recognized in areas of the northeast or upper
midwest where there are many lakes. Generally, lakes are easily recognized, although small
shallow lakes might also be considered wetlands. For operational purposes, the lake population
targeted by EMAP-SW includes standing bodies of water > 1 ha, with 1,000 m2 of open water
and a maximum depth > 1 m.
5.4.2 National Estimates
Table 5-1 contains one estimate of the number of lakes in the lower 48 states, presented by each
EPA region and as an aggregate by the entire area. This summary represents lakes drawn on the
1:100,000-scale U.S. Geological Survey (USGS) topographic series maps. These maps have
been digitized and incorporated into EPA's River Reach File, version RF3. The waterbodies iden-
tified as lakes in these digital files comprise the estimates. These national estimates have not
been verified other than by the USGS in its original mapping. Thus the estimates are based
164
-------
Table 5-1. Regional and National Estimates of Lake Number and Area3
EPA Region
Region I and llb
Region III
Region IV
Region V
Region VI
Region VII
Region VIII
Region IXC
Region Xd
NATIONAL TOTALS
Lake Number
21,725
10,739
63,829
64,863
46,034
24,350
41 ,385
8,224
9,060
290,209
Lake Area (ha)
963,330
149,761
1,756,055
2,323,056
1,635,451
360,955
1 ,497,363
812,491
567,385
10,065,847
Data derived from U.S. EPA Office of Water, River Reach File 3 (U.S. EPA, 1991). These data represent estimates of
waterbodies found on the USGS 1:100,000-scale map series and have received the quality assurance provided in the
original cartography. These estimates will be evaluated during the EMAP-SW lake selection process (see Table 5-2)
and will change as improvements are incorporated into River Reach File 3.
Data for Regions I and II combined; does not include Puerto Rico or the Virgin Islands.
Data for Region IX does not include Guam or Hawaii.
Data for Region X does not include Alaska.
solely on the digital information available. They have been compiled and reported by the U.S.
EPA (1991) Office of Water in an effort to provide consistency across the states in reporting on
water quality. An extensive national effort to verify these national estimates, similar to that
described in the next section for EPA Regions I and II, will be developed by the States and the
U.S. EPA (Office of Water, Office of Research and Development, and Regions). Refined estimates
can be reported when verification of the maps and routine monitoring of the extent of lakes is
included as part of the basic landscape characterization of the spatial distribution of ecological
resources.
5.4.3 Regional Estimates
During the 1991 pilot study, EMAP-SW considered only lakes between 1 and 2,000 ha in EPA
Regions I and II. By taking a probability sample of all lakes, we can estimate the actual number
165
-------
of lakes present for comparison with the estimates from RF3. During the first stage of screening,
we evaluated 311 lakes to identify map errors and waterbodies of noninterest for this phase of the
monitoring program. We accomplished this by evaluating maps of different resolution and dis-
cussing our concerns with local managers. Forty-eight lakes were identified. The probability
design allows inferring the size of the population of lakes of interest for the region of interest
(Northeast). Further screening of the second-stage sample by field visits indicated that another
18 did not meet the definitional criteria. Table 5-2 lists the estimated number and area of lakes
meeting the definitional criteria, based on one-fourth of the expected sample size over the four-
year cycle; 11,455 waterbodies in the northeast between 1 and 2,000 ha meet our operational
definition of a lake. The majority of the frame errors and noninterest lakes were 1-20 ha in size.
Table 5-2. Lake Number and Lake Area for Lakes 1-2,000 ha for the Northeast (EPA
Region I and II) and Selected Subregions, Estimated from the Tier 2 Sample3
Class
Tier 2 sample lakesb
Tier 2 non-lakes
Tier 2 waterbodies that
were too shallow, or
wetlands
Tier 2, not visited
Tier 2 target lakes
Adirondacks
New England Uplands
Coastal/Lowlands/Plateau
Sample
Size
(# Lakes)
92
3
12
3
74
26
29
19
Estimated
Number
of Lakes
16,795
1,109
3,840
391
11,455
1,506
5,669
4,280
SE of Lake
Number
Estimate
563
723
1,197
230
1,251
285
1,206
1,048
Estimated
Lake Area
(km2)
4,221
27
116
49
4,030
1,082
2,099
850
SE of Lake
Area
Estimate
830
17
37
32
814
395
758
254
Estimates of population size (for both area and number) are accompanied by the standard error (SE) of the estimates.
Includes additional lakes selected lor acidification studies.
166
-------
It is also important to describe the extent of the lake resource in specific subregions of the North-
east. One way is to identify ecologically important regions, such as those delineated by Omernik
(1987), within which extent and condition can be described. Due to a limited sample size this first
year, only three subregions were delineated: the Adirondacks, the New England Uplands, and an
aggregation of the coastal, lowland, and plateau areas (Figure 5-1). Estimates for the extent of
lake resources in these three regions are also displayed in Table 5-2.
5.5 CONDITION OF RESOURCES
5.5.1 Indicator Data for 1991
The following section contains results of data collected on specific response and exposure indi-
cators during EMAP-SW field sampling in lakes of the northeast during July-August 1991. As out-
lined in Section 5.2, the long-term assessment goal for EMAP is to describe the proportion of
populations in nominal and subnominal condition. The decision of what constitutes nominal and
subnominal for specific indicators varies by region and requires extensive evaluations of "refer-
ence conditions" in a region along with a variety of subjective judgments. The statistical sum-
maries will present data in a manner that minimizes the extent of subjective interpretation that
precedes presentation of the data. Again, recall that these results, while based on 1991 data, are
presented as a prototype; expect modifications as we refine the information contained in the
annual statistical summaries.
One of the key tools for presenting information in the summaries will be the cumulative distribu-
tion function (Figure 5-2). The CDF is a snapshot of the complete population variation and allows
us to estimate the proportion of the population above or below a particular indicator score. It
provides the complete data for the population, with uncertainty estimates, with few if any value
judgments imposed. Different readers can evaluate the same data with different criteria (Figure
5-2a) and come to their own conclusions.
If, as in Figure 5-2b, we select a score of 40 or less to distinguish poor condition for a particular
indicator, we can estimate that 45% of the total resource (by number, area, or length) has a value
of 40 or less and thus would be considered in poor condition. The uncertainty about the esti-
mate, in this case, ± 15%, can also be estimated. Although this may appear complicated to
many readers, it is a relatively simple way of presenting the complete data set with no judgment
about particular indicator scores of concern. The reader can determine which scores are of
167
-------
Adirondacks
New England
Upland
Coastal / Lowland / Plateau
EMAP-SW 8Sept92
Figure 5-1. A map illustrating three ecological regions on which statistical summaries can
be based for routine reporting.
168
-------
1.0
0.75 -
O
Is
3
Q.
O 0.50 -
O
O
Q.
O
0.25 -
Upper Confidence
Interval
15 30 45
Indicator Score
60
1.0
c 0.75 -
O
'•^
2
a.
o
a.
O 0.50 -
c
O
o
Q.
O
0.25 —
Upper Confidence
Interval
I I
co
O
O
C/J
15 30
Indicator Score
45
60
Figure 5-2. Cumulative distribution functions (CDFs) will be the primary format for
reporting the condition of ecological resources. CDFs can be interpreted by
reading the proportion of the resource (number of lakes, kilometers of
streams) below any selected score as in A. If a particular score is chosen to
delimit poor condition, the proportion (with confidence limits) of the resource
in poor condition can be displayed as in B.
169
-------
interest and evaluate the population from that perspective. In some instances, the CDF sum-
maries will be accompanied by a histogram that reduces the information into some relatively
"value free" classes, such as various trophic designations for lakes. After criteria are developed
to classify lakes with respect to their acceptable/unacceptable condition, these criteria will be
used to classify lakes into nominal and subnominal classes.
5.5.2 Trophic Condition of Lakes
Trophic condition in lakes has been one of the dominant concerns in surface water monitoring
and the subject of extensive research for the last 25 years. Concerns range from the more
aesthetic, which generally center around water clarity or nuisance floating algal mats, to the more
biological, for example, excessive algal production that leads to oxygen depletion and subsequent
fish kills during the summer or winter. Although lakes undergo a natural aging process that leads
to increasingly eutrophic conditions, it is clear that in many instances this process has been
accelerated by anthropogenic disturbances such as point and nonpoint discharges of organic
and inorganic nutrients. A variety of indices have been developed, for example, Carlson's Trophic
State Index (TSI) (Carlson, 1977), based on a combination of nutrient concentrations such as total
phosphorus and water transparency taken from Secchi disk readings, and algal biomass as esti-
mated by concentrations of algal chlorophyll-a. Until we have final concurrence on a trophic state
index appropriate for all or most regions of the country, we will present the fundamental informa-
tion that goes into indices of this type. Figure 5-3 shows the data for chlorophyll-a and total
phosphorus for EPA Regions I and II. The data are presented as cumulative proportion of lake
number [F(x)] and lake area [G(x)J. In this figure, the quartiles for the data are also presented.
For example, the 25th percentile for chlorophyll-a is 2.41 ,ag/L. This means that 25% of the lakes
in Regions I and II have a chlorophyll-a concentration of 2.4 /zg/L or less. Figures 5-4, 5-5, and
5-6 present similar data, but broken down by the three regions outlined in Figure 5-1, the
Adirondacks, New England Uplands, and the Coastal/ Lowland/Plateau.
Using criteria developed by NALMS (1988), we can summarize these data and estimate the pro-
portion of lakes in various trophic categories (Figure 5-7). Using the NALMS criteria for oligo-
trophic (< 10 /ig/L TP) and eutrophic (> 30 ,ug/L TP), we estimate that 21% of the lakes between
1 and 2,000 ha in Regions I and II are eutrophic. Based on one year's sample of 74 lakes, this
estimate may be as high as 33% or as low as 9%. The regional differences (Figure 5-7) are quite
distinct, although the confidence intervals are much larger given the smaller sample sizes. Figure
5-7 also presents the results as number of lakes rather than proportion of lakes, allowing the
reader to see the actual number of lakes affected.
170
-------
e
2,
3;
c
d
Hi
a «J
6 d
5 1
Mo
•? » E
.E CD
0)
8 o to
> C 0)
•s « £
SJ -o o
^s ~ o
-
.* O O
J5 -s «
= u> .2
TO 0) "D
. i_ C
o o» ~
•C Q. O
1
«
£ Q.
•*- (0
« i*
f
O
0) W
• 2
o
U
(0
3* c
-•00)
4rf C Jt2
|.«
CO ® .£ « Q
fi*B
^ Q) I- «
•C -C . "
o *; u. "x"
^Sa
4-> Si
n -1- a> _(«
-1§ § fe S
LJ CM £ a.
S
a>
o>
171
-------
o 9 co (• c
r) d d d o
§ 3
3 §
d
Mo
O)
il
172
-------
o'd
od
Mo
can
A 6 d
>• e
d d
M3
r
o
f1
0)
i
I
D)
iZ
173
-------
Q JO
a «j (• c » <
d d d d d o
* 1 5 *•
d o o tf
0 o'
g t| H o
o o o o
U)
0)
3
D)
174
-------
1.0 1
Oligotrophic
Mesotrophic
Eutrophic
95% Cl
Regions 1-2 Adiron. NE Uplands C/L/P
Total
Region
12000-1
10000-
w 8000
o>
.*
CO
Q>
J3
E
D
# Oligotrophic
ii # Mesotrophic
• # Eutrophic
95% Cl
6000-
4000-
2000-
Reg. 1&2 Adiron. NE Upland C/L/P
Region
Figure 5-7. Histogram of estimated proportion and number of eutrophic, mesotrophic, and
Oligotrophic lakes between 1 and 2,000 ha in the Northeast. Black bars
represent the 95% confidence interval for the estimates.
175
-------
5.6 CONCLUSIONS
The primary objective of the Annual Statistical Summary is to report significant results annually.
Given this objective, the data presented in the Annual Summaries will focus on the EMAP
response indicators that can be currently interpreted. In 1991, trophic state measurements and
zooplankton were the only response indicators collected at all probability lakes. The zooplankton
indicator will be undergoing development over the next several years and thus is not ready for
effective summarization in the annual statistical summaries. During the 1992 pilot, regional data
on fish and riparian birds, as well as zooplankton and diatoms are being collected. We anticipate
presenting the results of these additional indicators in later years.
In addition, much debate has occurred on the subject of presenting the exposure, habitat, and
stressor indicator summaries in the annual statistical summaries. Given the focus of EMAP on its
first objective, that is, describing status and trends in conditions, and the role of the exposure,
habitat, and stressor indicators in diagnostics, we believe it is premature to commit to extensive
inclusion of these additional indicators in the annual summaries. Thus, the extensive data on
water chemistry and physical habitat will not be presented in the annual summaries, but will be
used in the interpretive reports to explain patterns seen in response indicators. However,
thoughtful comments and perspectives on this subject from users would be appreciated.
176
-------
CHAPTER 6
LOGISTICAL OPERATIONS SUMMARY
6.1 INTRODUCTION
The complexity of EMAP leads to several major logistical issues, such as personnel identification,
procurement of equipment, and site access. One purpose of the 1991 Northeast Lakes Pilot
Survey was to assess the difficulty of assembling, training, and deploying sampling teams and
required equipment to obtain lake measurements and samples within the necessary time peri&d.
We also needed to know if it is possible to coordinate the work of multiple field crews well
enough for smooth, consistent sample collection and field recording. Can samples and field data
collection forms be shipped and tracked to appropriate analytical facilities or data processors in a
timely manner and without loss of samples or data? The logistical planning, implementation, and
post-season reporting activities for the 1991 pilot focused on providing answers to these
questions for specific indicators and for the total field efforts.
The 1991 pilot survey had two major components (Table 6-1). The probability lakes component
focused on logistical and variance questions related to sampling lakes at a regional scale. This
component had two subcomponents—one to test whether or not selected indicators can be effec-
tively sampled in an index mode and the second, in response to the mandates of the Clean Air
Act, to incorporate the Temporally Integrated Monitoring of Ecosystems (TIME) Program.
EMAP-Surface Waters (EMAP-SW) staff selected 64 regional probability lakes from the grid design
to evaluate EMAP's ability to efficiently collect samples within the index period for indicators for
which there is consensus about methodology and usefulness. Field plans included 1 visit to each
of these lakes, a revisit to 32 of the lakes to collect data for estimating index period variability, and
1 visit to each of the additional 28 lakes selected for TIME. Each of the 3 field crews visited 4 of
the original 64 lakes to provide data for crew comparability studies. Of these 12 visits, 4 were
counted as initial visits. Thus, the 64 initial visits, 32 repeat visits, 28 additional TIME lakes, and 8
crew comparability visits resulted in 132 possible lake site visits. Table 6-1 lists the kinds of
samples taken during each of these visits.
The indicator lakes component (Table 6-1) addressed the program's ability to obtain a cost-effec-
tive index sample for fish, benthic macroinvertebrates, and riparian birds, and a more intensive
evaluation of physical habitat. This component of the pilot activity focused primarily on obtaining
enough information to select adequate sampling protocols to be used in later surveys, identify
177
-------
o>
CD
_c
2
I
8
|
CO
UJ
o
•o
0)
c
>
'&
o
0)
1
0>
o
a
o
O
(O
0)
3
a
UJ
DC
D
CO ^
DC j-
UJO
^ Q-
UJ UJ
o *
<
CO
UJ
^f
^"j
"^
DC
£
O
Q
Z
CO
111
3%
^~ "*""*
f— !s2
1 >
3 co
CQ ^
O
DC
Q_
(/)
fl)
O
CM
I 1
15 co
c: CM
.2 "-'
^ o
< .5
TJ
cl
0)
15
o
'5)
DC
UJ
•5
-n *W*
? 0)
^ v
"D ^
3 JO,
CO
"2
(0
(5
TI
C
(C
to
o
'•55
?
i/r
00
CM
."tl
1
II"
1'
o S,
1
o
CM .-e
*- .S2
!S2 >
CO,
"5T
I- '«
« '>
CO
o
If , 11
_Q to c co X
C (Q "re -^
^5 ^ C r^
^ I ^ll g,l«o Si
•— 0) ,-^5x"c°n"o R^TI »\
CO
O
ts
2
(0
CL
Ego"0-2to«£Q)I-NCOCOCLCLU.U.U.DCCQ
s.
CO
E
(0
ro
CL
'? -Si
t (0
^ M
0 O
to o
CO
O
£
(0
Q.
" §
£ o -°
o> -e —
.C -;£ (0
O ^j
0 "CLD
| 82
> N Q-
CO
(0
O
"S
aram
CL
^ £
| o c §
c nj o -Q
1o S —
0 0 m "°
•— (U i—
S> "o "Q- P
4— ' "- *™» ^~
£ 2 8 2
^ 1- N CL
co (o
^» Zl ^ ^J* .
^" •P O 'o •"
•= w c * !§ 'S S
meters
£
<5
0_
5^1 «^5f
O O S "O O Q) <0
•2 ^- O ^ ^ ^ is
-5 2 O 2 Q) Q) -C
> h- N D_ CO CO 0_
•n w
1-8
•5! 3
Ji
.£'<»
a> <«
11
iS
2 P
•a „,
|i£
'f
§W
S±l
T3 .3 <§
T3 S —
(0 C a?
5sS
J2 O
' c a>
O _C
•£Z •*-•
:g 5
-E5
o>
« £
1-5
-o
>, a)
111
| o -S
- '^
. •-
1 I "•
TI
c
0.12 c)
111
178
-------
major habitat types, and estimate and evaluate the logistical and time requirements for imple-
menting the complete suite of indicators in the field. Twenty indicator lakes were handpicked
from outside the EMAP grid design to ensure that the variety of lake sizes and impacts expected
to occur during annual surveys could be evaluated for sampling feasibility and variations in
sampling protocols.
Three different types of field crews collected data for the indicator lakes component of the survey.
One fisheries crew collected data and samples for indicators for fish, water chemistry, trophic
state, zooplankton, sediment diatoms, sediment toxicity, and physical habitat. A second team
conducted the riparian bird survey to address specific concerns inherent in bird survey studies.
The third team collected littoral and profundal samples to assess benthic sampling protocols.
Each of these crews traveled and worked independently of the others, and their logistical plan-
ning, implementation, and reporting activities will be discussed in other reports; some aspects are
described in Chapter 2.
6.2 PLANNING ACTIVITIES
A logistics team was established early in the planning stages of the study to ensure that all field
activities were conducted efficiently and that the data collected were of high quality. The logistics
team was concerned primarily with the limnological field activities for the probability lakes. In
some cases, simultaneous planning and preparation activities included the fisheries and bentho-
logical field operations. Members of the logistics team developed plans related to quality assur-
ance (Peck, 1991), information management, and afield operations and training manual. Specific
planning steps and activities are described in the following subsections.
6.2.1 Lake Verification
In October 1990, the design team provided the logistics team with a list of the 311 lakes in the
Tier 1 frame for the northeast United States. In order to assess the quality of the selection
process initiated by the design team, the logistics team manually plotted the point coordinates on
7.5' or 15' USGS topographic maps. This lake verification process revealed several potential
selection errors. For example, lakes and rivers with transecting bridges may appear in the base
frame as multiple lakes. The digitization process that electronically stores the map information
cannot differentiate between a natural barrier and a human-made structure, such as a bridge that
does not restrict cross flow of a waterbody. Other errors included identifying water bodies as
lakes when actually they were ephemeral pools found in flood plains, marshes, and coastal bays,
or pools in rivers generated by beaver dams or human-made dams later destroyed. Once this
179
-------
plotting effort was completed, the logistics team was able to evaluate the sample points selected
in terms of whether or not a site met the definition for an EMAP lake. After examining the maps
and telephoning the communities in which the lakes were located when necessary, the logistics
team stored the resulting information in a database that was transmitted to the design team.
From the original 311 lakes drawn from the lake frame, the logistics team identified 48 sites that
did not meet the criteria for EMAP lakes (a body of standing water < 1 ha in surface area, > 1 m
in depth, and containing 1,000 m2 of continuous open water) and 51 sites lacking sufficient data.
Using this information, the design team removed the 48 nontarget sites from the list of 311 and
then selected 64 probability lakes for the pilot study. Another 28 lakes were selected from an
enhanced grid for the Temporally Integrated Monitoring of Ecosystems (TIME) study (for a total of
92 TIME lakes).
Time and weather restricted the reconnaissance conducted to investigate apparent access prob-
lems and to determine if these sites met the EMAP definition of lakes. Attempts to obtain
reconnaissance information by telephone had limited success and information obtained from
some local sources was not always accurate. At least one crew hiked to a site, only to discover
that a road was present that provided much easier access.
In order to investigate apparent access problems, the limnological field coordinator traveled to
Maine in May to participate in flyovers of about eight lakes that appeared to present access
problems. The flyover effort was successful in verifying the presence and size of the lakes and
clarifying some access questions. The flyover observations indicated that five of the lakes did not
meet EMAP criteria and summer field crew visits confirmed these conclusions.
6.2.2 Site Access
Local government officials were contacted by telephone to identify landowners and to collect as
much information as possible about a lake site. In addition, representatives of several state
government agencies furnished lake information. Most of the sites selected for sampling were
privately owned and thus the number of sites on public lands was limited. There were no diffi-
culties in obtaining permission to access public lands during the 1991 survey. In some cases,
especially in Maine, it was not necessary to acquire site access permission because of great
pond laws which specify, for example, that ponds containing more than 10 acres lie in common
for public use.
180
-------
Before the logistics team contacted private landowners by telephone, EPA Regions I and II sent a
letter, approved by EPA headquarters, requesting permission to access the property. It took a
significant amount of time to obtain access permission and accurate details about the status of a
site by telephone or mail to state and local contacts and landowners. In addition, the information
obtained was often derived from subjective evaluation of a site. At times, the state or local
contact would identify the wrong lake or consider the selected site to be a marsh or wetland,
although the site fell within the EMAP definition of a lake. In one case, a landowner reported in a
telephone conversation that his lake was greater than a hectare in surface area and greater than a
meter in depth. In fact, when the site was visited for sampling, no lake was present, only a
stream running through a field. Because some of the USGS topographic maps used to determine
locations had not been updated recently, they sometimes misrepresented roads or did not show
more recently developed roads. In some cases, access permission letters were not received
before the crews reached the field, but permission letters were always in hand before the field
team visited the site.
6.2.3 Protocol Development, Training, and Mobilization
The EMAP-SW team adopted sampling protocols for limnological, fishery, and benthological
sampling that were originally selected by individual indicator leads. The indicator leads forwarded
the methods to the logistics team for inclusion in a field operations and methods manual. Some
of the methods were fully developed; others were in the formative stage, with the pilot survey
serving as the means to field test the proposed methodology.
Personnel at the U.S. EPA laboratory in Las Vegas conducted protocol development and training
sessions, called dry runs, at Lake Mead, Nevada, and Pahranagat National Wildlife Refuge in
Alamo, Nevada. These sessions enabled individuals to work in small groups, identify problems
and resolutions, and concentrate on issues pertinent to the pilot study.
Training for limnological, benthological, and fisheries sampling was held at the Pahranagat
National Wildlife Refuge in Nevada from June 17 to 24. Trainers had extensive experience in
training scientific personnel in collection techniques and field operations. During this week-long
training period, groups of two field samplers worked in hands-on sessions with trainers who intro-
duced and demonstrated all aspects of the field activities. Indicator leads participated in the
training. The sessions also covered activities such as shipping, sample tracking, and equipment
maintenance. Each sampler had to demonstrate competence in the practical aspects of field
181
-------
activities, such as handling boats, operating global positioning systems (GPS) and dissolved
oxygen (DO) meters, locating sampling sites, completing data forms, understanding and
demonstrating safety procedures, and performing general administrative tasks.
Additional practice runs were conducted on July 8 in Nashua, New Hampshire. At the conclusion
of training, crews began field work by participating in a crew comparability study. Each of the
three crews sampled two lakes at different times over a three-day period.
Some indicator leads visited the field teams early in the sampling period. These visits were
extremely useful, providing opportunities for the leads to work with the field crews to improve their
understanding of the need for particular protocol elements, solve problems with implementing the
protocols, and correct minor errors in the protocols.
Limnological, benthological, and fisheries crew members moved to New Hampshire from Las
Vegas with their equipment, leaving June 29 with two 14-foot rental trucks. Four-wheel drive
vehicles were leased in the northeast for the field work. The crews assembled equipment and
began full implementation on July 10. The fisheries and benthological teams operated indepen-
dently of the limnological team. The crews used equipment that had been obtained for previous
surveys and stored in Las Vegas. Although this equipment served the purpose during the 1991
survey, by the end of the field work it was obvious that most of it would need to be replaced
before the next field season.
6.2.4 Field Crew Personnel
Field crew members for the limnological, fisheries, and benthological crews were drawn from U.S.
EPA laboratories and regions and from contract personnel. Scientists from the U.S. EPA
Environmental Monitoring Systems Laboratories in Las Vegas and Cincinnati served as field
samplers for the full field season. Scientists from U.S. EPA Regions I and II also served as
limnological crew members for the entire season. Participation by Region I included providing
GPS units, as well as training the crew members in the use of these units. For the entire field
season period, at least one U.S. EPA regional and one laboratory participant was in the field.
Thirteen individuals filled the six limnological crew positions (three crews, two people each) at
different times. The length of time a sampler stayed in the field varied from one to nine weeks.
One person served as field coordinator for the eight-week sample collection effort.
182
-------
Early efforts were directed at getting the states involved in the planning and field activities.
Representatives of each of the New England states and New York and New Jersey participated in
the 1991 survey at different levels. A majority of the states provided ownership and access
information. New Hampshire, Vermont, and New York personnel accompanied field crews to
learn more about EMAP and about the EMAP-SW effort. In general, though, the widespread
depressed economic circumstances prevented the states from assuming a larger role in field
activities.
6.3 FIELD OPERATIONS OVERVIEW
Three field crews of two people each and a base coordinator conducted field operations between
July 8 and August 28, 1991. The crews used 4-wheel drive vehicles pulling trailers carrying 12-
foot boats to travel between base sites and sampling sites. It took approximately 2 to 4 hours to
sample the EMAP probability lakes and the TIME lakes. To reach 21 lakes located in remote
areas, crews had to hike in (15 sites; 6 were not EMAP target sites) or use a float plane (6 sites).
The crews worked out of 14 centralized base sites. The following subsections describe daily
sampling and support operations.
6.3.1 Daily Sampling Operations: Probability and TIME Lakes
The field crews were scheduled to collect measurements [watershed characteristics, site location
(GPS), site depth, temperature profile, dissolved oxygen profile, Secchi disk transparency] and
samples (water chemistry, chlorophyll-a, zooplankton tow, sediment core, sediment toxicity, pro-
fundal benthos) during 132 visits to probability and TIME lakes (Table 6-1). Specific water and
sediment sampling activities carried out at the lakes are shown in Figures 6-1 and 6-2.
The three crews began activities each morning by calibrating instruments and using equipment
checklists to ensure that all necessary equipment and supplies were loaded into the vehicles
before departing for the sample site. The field coordinator remained at the motel base site to
organize logistics support (e.g., picking up equipment and supplies, shipping and tracking
samples, transmitting data, contacting landowners) and maintain a communications link with
program management. The field coordinator also served as a backup sampler. Using a central
base site sometimes required long drives by the field crews to the lake site and then resulted in
long work days.
183
-------
WATER SAMPLING ACTIVITIES
ANCHOR BOAT AT
INDEX SITE
INITIATE D.O.
AIR EQUILIBRATION
INITIATE
PROFILE FORM
SAMPLER #1
(DATA RECORDED ON
SAMPLE COLLECTION FORM)
CALIBRATE D.O.
METER
MEASURE
SECCHIDEPTH
D.O./TEMPERATURE
PROFILE
SAMPLER #2
(SAMPLE COLLECTION)
I
COLLECT
VAN DORN SAMPLE #1
-FILL 4-50 mL SYRINGES
-FILL1 -4-LCUBITAINER
- FILL IN SAMPLE COLLECTION FORM I
COLLECT
VAN DORN SAMPLE #2
- FILTER 250 mL CHLOROPHYLL
- FILL IN SAMPLE COLLECTION FORM
- FILL 500 mL BOTTLE FOR
CONDUCTIVITY
COMPLETE SITE
PROFILE FORM
ZOOPLANKTON TOW
- 2 SAMPLES COLLECTED
SIMULTANEOUSLY
- FILL IN SAMPLE
COLLECTION FORM
VERIFY FORM
COMPLETION
I
1
RETURN TO TRUCK
- TRANSFER EQUIPMENT
-STORE SAMPLES
Figure 6-1. Daily water sampling activities at 1991 EMAP Northeast Lakes Pilot sites.
184
-------
SEDIMENT SAMPLING ACTIVITIES
RETURN TO INDEX SITE
ANCHOR BOAT
COLLECT SEDIMENT CORE SAMPLE
- MEASURE LENGTH OF CORE
- SIPHON OVERRIDING
CORE WATER
- EXTRUDE & COLLECT TOP
1-CM OF SEDIMENT CORE
- EXTRUDE REMAINING
SEDIMENT
- COLLECT 1-CM OF SEDIMENT
CORE - 2-CM FROM BOTTOM
RECORD DATA ON SAMPLE |
COLLECTION FORM
COLLECT TWO PONAR DREDGE SAMPLES
- HALF FILL TWO -1- GAL SAMPLE BAGS
RECORD DATA ON SAMPLE
COLLECTION FORM
COLLECT THREE PONAR DREDGE SAMPLES
- SIEVE INDIVIDUAL SAMPLES
- TRANSFER TO INDIVIDUAL CONTAINER
- PRESERVE SAMPLES
COMPLETE SAMPLE COLLECTION FORM
- VERIFY FORM COMPLETION
RETURN TO TRUCK
- TRANSFER EQUIPMENT
Figure 6-2. Daily sediment sampling activities at 1991 EMAP Northeast Lakes Pilot sites.
185
-------
Upon arrival at the site, the field crew verified the location based on landscape features, topo-
graphic maps, and GPS information (if coverage was available at that time). After loading the
equipment and launching the boat, the crew used sonar to determine the index site for sample
collection at the deepest part of the lake. All samples were preserved (when appropriate),
labeled, and packed for return to the base site, and the boat was returned to shore. The boat
was then re-outfitted for collection of sediment samples. All samples and forms were checked for
completeness against a check list and then packed for return to the base site.
The GPS units proved to be successful locational devices when connections to the satellite
system were available. This system will be expanded in the future and the GPS units should
become more useful.
At the base site, the field coordinator conducted a debriefing with the sample crews and checked
the data forms, sampling labels, and condition of the samples. Selected data forms were trans-
mitted by FAX to the Las Vegas communications center for data entry after they were reviewed by
the field coordinator. The sampling crews and the field coordinator then cleaned and prepared
equipment and supplies for the next day.
Water chemistry samples were shipped to the analytical laboratory on the morning after sampling,
whereas sediment toxicity and chlorophyll-a samples were held for shipment for up to one week.
The preserved benthic invertebrate, zooplankton, and sediment diatom samples were stored for
biweekly shipment.
Measurements and samples were collected on 105 of the scheduled 132 site visits. The 27
remaining sites were not sampled for the following reasons:
1 access was denied when crew reached site
1 access was restricted (a locked gate)
1 site was misidentified (map typographic error; a lake has never existed at that site)
1 site visit was not accomplished during crew comparability lake visits
14 sites did not meet the definition of a lake (e.g., too shallow, wetland)
9 site revisits were originally scheduled to the 14 non-EMAP lakes
In total, 1,321 samples were collected and shipped to analytical support sites. Sample collection
and shipment of the samples to the laboratory was 100% complete.
186
-------
6.3.2 Support Operations
The communications center, located in Las Vegas, Nevada, was monitored by two people
working in shifts. Answering and facsimile machines and a toll-free telephone number were
available for communications throughout the field season. The field coordinator also had a
facsimile machine to facilitate communications. The communications center monitored all aspects
of field sampling, including coordinating and tracking support requests for supplies and main-
taining a central contact point for information exchange among the limnological, benthological,
and fisheries field crews; the management team; the information management team; the analytical
laboratory; and the public. An information pamphlet was also available for distribution in the field
to interested parties. The field crews did not have communication equipment and, when neces-
sary, had to locate public telephones or return to the base site.
Electronic entry of field data was not implemented during the 1991 pilot survey and all field data
were manually entered on paper forms. There were budget and time constraints and most data
were derived from sample analyses. A portable data recorder was evaluated at the end of the
field season. Several disadvantages were noted, including difficulty in entering comments and
the need to download the unit on a daily basis. Further investigations into other kinds of
recording units were recommended for future field seasons.
Development of the field forms was a combined effort involving the indicator leads, information
management, logistics, and quality assurance team members. Thoroughly reviewed before the
field season, these forms provided the means to record accurate data and site information. The
forms were checked by the field coordinator before they were transmitted to the communications
center; they were checked again at the communications center. Information on the data forms
was entered into the database in Las Vegas.
The sample tracking system was enhanced by the use of barcoded labels. These labels, applied
to all samples at the base site, enabled the sample identification to be electronically entered into
the sample tracking system. The site identification, data, crew identification, and type of sample
were stored in a database that could provide sample-specific information on sample shipping and
tracking records. Data transfer was restricted to overnight shipment of diskettes and field forms.
Use of a modem to transmit data electronically was curtailed by inadequate telephone systems in
remote regions of the Northeast.
187
-------
6.4 RESULTS AND RECOMMENDATIONS
In general, the 1991 Northeast Lakes Pilot Survey was successful in assembling, training, and
deploying sampling teams and equipment to obtain lake measurements and samples within the
required index period. The work of multiple field crews was coordinated well enough to result in
efficient and consistent sample collection and field records. Field visits determined that 24 of the
132 planned site visits were to nontarget sites, leaving a possible 108 sites from which EMAP data
could be collected. Of these, field crews completed operations during 105 visits for a completion
rate of over 97%. In addition samples and data collection forms were successfully shipped,
tracked, and received at the appropriate destination without any loss.
Problems encountered in the planning and implementation process were solved in a way that did
not prevent the successful completion of the field operations. Recommendations for future
surveys include:
• Site Access
Develop coordination and cooperation with the regional offices for site verification
and access activities.
Identify possible sites earlier in the season to provide more time to conduct
verification and access investigations.
• Reconnaissance
Conduct additional reconnaissance before the field operations begin to reduce the
number of field crew trips to non-EMAP sites.
Consider conducting reconnaissance for 1993 sites during the 1992 field season.
• Field Crew Personnel
Cooperate with other agencies in field efforts already in place to identify crew
members.
• Training
Conduct training in the area where field operations will take place.
Develop short courses, to be completed before training, for requirements such as
boat handling, map orienteering, and survival skills.
• Mobilization
Develop warehouse facilities in the regions where field operations will take place.
Use existing motor pools.
188
-------
• Field Activities
Identify base coordinator on each field crew and do not establish central base
sites.
Schedule field crew visits to sites within small geographical area.
• Field Quality Assurance
Estimate precision and accuracy of measurements.
Improve field quality control measurements.
Make minor modifications in forms and labels.
Conduct additional field audits.
Improve procedures for preventive maintenance and calibration.
Refine field logbooks.
• Communications
Provide field crews with field communications capability.
• Information Management
- Evaluate grid pad portable data recorder.
Use bar codes for field samples.
Investigate using modem to track samples.
In summary, the logistical element of the 1991 Northeast Lakes Pilot Survey was able to plan,
implement, and successfully complete field operations and to identify recommendations for future
surveys.
189
-------
CHAPTER 7
LITERATURE CITED
Ackerman, J. 1992. Tracking a river of birds. Nature Conservancy 42:23-27.
Allan, D. and 21 coauthors. 1992. Report of the NABS FWI (Freshwater Imperative) working
group. Bull. N. Am. Benthol. Soc. 9:266-271.
American Public Health Association (APHA). 1989. Standard Methods for the Examination of
Water and Wastewater. Seventeenth Edition. American Public Health Association,
Washington, D.C.
Angermeier, P.L., and J.R. Karr. 1986. Applying an index of biotic integrity based on stream fish
communities: considerations in sampling and interpretation. N. Am. J. Fish. Manage.
6:418-429.
Bailey, R.G. 1976. Ecoregions of the United States. U.S. Forest Service, Ogden, UT.
Barbour, C.D., and J.H. Brown. 1974. Fish species diversity in lakes. Am. Nat. 108:473-489.
Bickford, C.A., C.E. Mayer, and K.D. Ware. 1963. An efficient sampling design for forest
inventory: the Northeast Forest Resurvey. J. For. 61:826-833.
Bovee, K.D. 1982. A Guide to Stream Analysis Using the Instream Flow Incremental Methodology.
Instream Flow Information Paper 12. FWS/OBS 82/26. U.S. Fish & Wildlife Service, Fort
Collins, CO.
Boyle, T.P., G.M. Smillie, J.C. Anderson, and D.R. Beeson. 1990. A sensitivity analysis of nine
diversity and seven similarity indices. J. Water Pollut. Control Fed. 62:749-762.
Brinkhurst, R.O. 1974, The Benthos of Lakes. MacMillan, London.
Brooks, J.L., and S.I. Dodson. 1965. Predation, body size, and composition of plankton.
Science 150:28-35.
Brooks, R.P., M.J. Croonquist, E.T. Silva, J.E. Gallagher, and D.E. Arnold. 1991. Selection of
biological indicators for integrating assessments of wetland, stream and riparian habitats.
Pages 81-89 in Proceedings Symposium on Biological Criteria. U.S. Environmental
Protection Agency, Washington, D.C.
Brussock, P.P., A.V. Brown, and J.C. Dixon. 1985. Channel form and stream ecosystem models.
Water Resour. Bull. 21:859-866.
Callicott, J.B. 1991. Conservation ethics and fishery management. Fisheries 16:22-28.
Carlson, R.E. 1977. A trophic state index for lakes. Limnol. Oceanogr. 22:361-368.
Carson, R. 1962. Silent Spring. Fawcett, Greenwich, CN.
191
-------
Chaloud, D.C., J.M. Nicholson, B.P. Baldigo, C.A. Hagley, and D.W. Sutton. 1989. Handbook of
Methods for Acid Deposition Studies: Field Methods for Surface Water Chemistry.
EPA/600/4-89/020. U.S. Environmental Protection Agency, Washington, D.C.
Chapman, P.M., and R.O. Brinkhurst. 1984. Lethal and sublethal tolerances of aquatic
oligochaetes with reference to their use as a biotic index of pollution. Hydrobiologia
115:139-144.
Charles, D.F., and 12 coauthors. 1990. Paleoecological investigation of recent lake acidification
in the Adirondack Mountains, NY. J. Paleolimnol. 3:195-241.
Charles, D.F., S.S. Dixit, B.F. Gumming, and J.P. Smol. 1991. Variability in diatom and
chrysophyte assemblages and inferred pH: paleolimnological studies of Big Moose Lake,
New York, USA. J. Paleolimnol. 5:267-284.
Cole, G.A. 1975. Textbook of Limnology, C.V. Mosby, Saint Louis, MO.
Cotter, J., and J. Nealon. 1987. Area Frame Design for Agricultural Surveys. Area Frame Section,
Research and Applications Division, National Agricultural Statistics Service, U.S. Department
of Agriculture.
Cowardin, L.M., V. Carter, F.C. Golet, and E.T. LaRoe. 1979. Classification of Wetlands and
Deepwater Habitats of the United States. FWS/OBS-79/31. U.S. Fish and Wildlife Service,
Washington, D.C. 131 pp.
Croonquist, M.J., and R.P. Brooks. 1991. Use of avian and mammalian guilds as indicators of
cumulative impacts in riparian-wetland areas. Environ. Manage. 15:701-714.
Gumming, B.F., J.P. Smol, J.C. Kingston, D.F. Charles, H.J.B. Birks, K.E. Camburn, S.S. Dixit, A.J.
Uutala, and A.R. Selle. 1992. How much acidification has occurred in Adirondack lakes
(New York, USA) since preindustrial times? Can. J. Fish. Aquat. Sci. 49:128-141.
Dixit, S.S., J.P. Smol, J.C. Kingston, and D.F. Charles. 1992. Diatoms: powerful indicators of
environmental change. Environ. Sci. Techno/. 26:23-33.
Doudoroff, P., and C.E. Warren. 1957. Biological indices of water pollution with special reference
to fish populations. Pages 144-163 in Biological Problems in Water Pollution. U.S. Public
Health Service, Cincinnati, OH.
Ebel, W.J., C.D. Becker, J.W. Mullan, and H.L Raymond. 1989. The Columbia River: toward a
holistic understanding. Pages 205-219 in D.P. Dodge, ed. Proceedings of the international
large rivers symposium. Can. Spec. Pub/. Fish. Aquat. Sci. 106.
Ehrlich, P.R., A.H. Ehrlich, and J.P. Holdren. 1977. Ecoscience: Population, Resources,
Environment. Freeman, San Francisco, CA.
Ellis, M.M. 1937. Detection and measurement of stream pollution. Bull. Bureau Fish. 48:365-437.
Eshleman, K.N. 1988. Predicting regional episodic acidification of surface waters using empirical
models. Water Resour. Res. 24:1118-1126.
192
-------
Fausch, K.D., J. Lyons, J.R. Karr, and P.L Angermeier. 1990. Fish communities as indicators of
environmental degradation. American Fisheries Society Symposium 8:123-144.
Forbes, S.A. 1925. The lake as a microcosm. Illinois Natural History Survey Bulletin 15:537-550.
Ford, J. 1989. The effects of chemical stress on aquatic species composition and community
structure. Pages 99-144 in S.A. Levin, M.A. Harwell, J.R. Kelly, and K.D. Kimball, eds.
Ecotoxicology: Problems and Approaches. Springer-Verlag, New York.
Frey, D.G. 1977. Biological integrity of water—a historical approach. Pages 127-140 in R.K.
Ballentine and L.J. Guarria, eds. The Integrity of Water. U.S. Environmental Protection
Agency, Washington, D.C.
Frissell, C.A., W.J. Liss, C.E. Warren, and M.D. Hurley. 1986. A hierarchical framework for stream
habitat classification: viewing streams in a watershed context. Environ. Manage.
10:199-214.
Gammon, J.R. 1991. The Environment and Fish Communities of the Middle Wabash River.
Annual Report. Dept. of Biology, DePauw University, Greencastle, IN.
Gammon, James R., DePauw University, Greencastle, Indiana (personal communication).
Gilbertson, M. 1992. PCB and dioxin research and implications for fisheries research and
resource management. Can. J. Fish. Aquat. Sci. 49:1078-1079.
Glew, J.R. 1989. A new trigger mechanism for sediment samplers. J. Paleolimnol. 2:241-243.
Goulding, M., M.L. Carvalho, and E.G. Ferreira. 1988. Rio Negro: Rich Life in Poor Water. SPB
, Academic Publ., The Hague, The Netherlands.
Gregory, S.V., F.J. Swanson, W.A. McKee, and K.W. Cummins. 1991. An ecosystem perspective
of riparian zones. BioScience 41:540-551.
Gubala, C.P., C.T. Driscoll, R.M. Newton, and C.L. Schofield. 1991. Chemistry of a near-shore
lake region during spring snowmelt. Environ. Sci. Technol. 25:2024-2030.
Hall, R.I., and J.P. Smol. 1992. A weighted-averaging regression and calibration model for
inferring total phosphorus concentration from diatoms in British Columbia (Canada) lakes.
Freshw. Biol. 27:417-434.
Halsell, N. 1990. Lake indicator workshop. U.S. Environmental Protection Agency, Las Vegas,
Nevada.
Hart, David D., Academy of Natural Sciences, Philadelphia, Pennsylvania (personal communica-
tion).
Hazard, J.W., and B.E. Law. 1989. Forest Survey Methods Used in the USDA Forest Service.
EPA/600/3-89/065. NTIS PB89-220-594/AS. U.S. EPA Environmental Research Laboratory,
Corvallis, OR.
Herlihy, AT., D.H. Landers, R.F. Cusimano, W.S. Overton, P.J. Wigington, A.K. Pollack, and T.E.
Mitchell-Hall. 1990. Temporal Variability in Lakewater Chemistry in the Northeastern United
193
-------
States: Results of Phase II of the Eastern Lake Survey. EPA/600/3-91/012. U.S. Environ-
mental Protection Agency, Corvallis, OR.
Hilborn, R. 1992. Can fisheries agencies learn from experience? Fisheries 17:6-14.
Hilsenhoff, W.L. 1987. An improved biotic index of organic stream pollution. Greaf Lakes
Entomol. 20:31-39.
Hocutt, C.H., and J.R. Stauffer, Jr. (eds) 1980. Biological Monitoring of Fish. D.C. Heath,
Lexington, MA.
Hughes, R.M., and J.R. Gammon. 1987. Longitudinal changes in fish assemblages and water
quality in the Willamette River, Oregon. Trans. Am. Fish. Soc. 116:196-209.
Hughes, R.M., and R.F. Noss. 1992. Biological diversity and biological integrity: current
concerns for lakes and streams. Fisheries 17:11-19.
Hughes, R.M., D.P. Larsen, and J.M. Omernik. 1986. Regional reference sites: a method for
assessing stream potentials. Environ. Manage. 10:629-635.
Hunsaker, C.T., and D.E. Carpenter. 1990.Environmental Monitoring and Assessment Program
Ecological Indicators. EPA/600/3-90/060. U.S. Environmental Protection Agency, Research
Triangle Park, NC.
Hunsaker, C., D. Carpenter, and J. Messer. 1990. Ecological indicators for regional monitoring.
Bull. Ecol. Soc. Am. 71 (2): 165-172.
Hutchinson, G.E. 1961. The paradox of the plankton. Am. Nat. 95:137-146.
Hynes, H.B.N. 1975. The stream and its valley. Verh. Internal. Verein. Limnol. 19:1-15.
Karr, J.R. 1987. Biological monitoring and environmental assessment: a conceptual framework.
Environ. Manage. 11:249-256.
Karr, J.R., and D.R. Dudley. 1981. Ecological perspective on water quality goals. Environ.
Manage. 5:55-68.
Karr, J.R., K.D. Fausch, P.L. Angermeier, P.R. Yant, and I.J. Schlosser. 1986. Assessing Bio-
logical Integrity in Running Waters: A Method and Its Rationale. Illinois Natural History
Survey Spec. Publ. No.5., Champaign, IL.
Karr, J.R., P.R. Yant, K.D. Fausch, and I.J. Schlosser. 1987. Spatial and temporal variability of
the index of biotic integrity in three midwestern streams. Trans. Am. Fish. Soc. 116:1-11.
Kaufmann, P.R., A.T Herlihy, J.W. Elwood, M.E Mitch, W.S. Overton, M.J. Sale, J.J. Messer, K.A.
Cougan, D.V. Peck, K.H. Reckhow, A.J. Kinney, S.J. Christie, D.D. Brown, C.A. Hagley, and
H.I. Jager. 1988. Chemical Characteristics of Streams in the Mid-Atlantic and Southeastern
United States. Volume I: Population Descriptions and Physico-Chemical Relationships.
EPA/600/3-88/021 a. U.S. Environmental Protection Agency, Washington, D.C.
Kaufmann, P.R., A.T. Herlihy, M.E. Mitch, and W.S. Overton. 1991. Chemical characteristics of
streams in the Eastern United States: I. Synoptic survey design, acid-base status, and
regional chemical patterns. Water Resour. Res. 27:611 -621.
194
-------
Keller, W., and N.D. Van. 1991. Recovery of crustacean zooplankton species richness in
Sudbury area lakes following water quality improvements. Can. J. Fish. Aquat. Sci.
48:1635-1644.
Klemm, D.J., and J.M. Lazorchak. 1992. Environmental Monitoring and Assessment Program
(EMAP) Field and Laboratory Methods Manual Used on Probability Freshwater Lakes. U.S.
Environmental Protection Agency, Cincinnati, OH.
Klemm, D.J., P.A. Lewis, F. Folk, and J.M. Lazorchak. 1990. Macroinvertebrate Field and Labora-
tory Methods for Evaluating the Biological Integrity of Surface Waters. EPA/600/4-90/030.
U.S. Environmental Protection Agency, Cincinnati, OH.
Knight, J.G., M.B. Bain, and K.J. Scheidegger. 1991. Ecological Characteristics of Fish Assem-
blages in Two Seasonally Inundated Palustrine Wetlands. Project Report. U.S. Fish and
Wildlife Service, Auburn University, Auburn, AL.
Kolz, A.L., and J.B. Reynolds. 1991. Principles and Techniques of Electrofishing. U.S. Fish and
Wildlife Service, Kearneysville, WV.
Landers, D.H., J.M. Eilers, D.F. Brakke, W.S. Overton, P.E. Kellar, M.E. Silverstein, R.D.
Schonbrod, R.E. Crowe, R.A. Linthurst, J.M. Omernik, S.A. league, and E.P. Meier. 1987.
Characteristics of Lakes in the Western United States. Volume I: Population Descriptions
and Physico-Chemical Relationships. EPA/600/3-86/054a. U.S. Environmental Protection
Agency, Washington, D.C.
Landres, P.B., J. Verner, J.W. Thomas. 1988. Ecological uses of vertebrate indicator species: a
critique. Conserv. Biol. 2:316-328.
Larsen, D.P., D.L. Stevens, A.R. Selle, and S.G. Paulsen. 1991. Environmental Monitoring and
Assessment Program, EMAP-Surface Waters: A Northeast Lakes Pilot. Lake Reserv.
Manage. 7(1): 1-11.
Lewis, W.M. 1983. A revised classification of lakes based on mixing. Can. J. Fish. Aquat. Sci.
40:1779-1787.
Lien, L., G.G. Raddum, and A. Fjellheim. 1992. Critical Loads of Acidity to Freshwater Fish and
Invertebrates. Report No. 0-89185. Norwegian Institute for Water Research, Oslo, Norway.
Likens, G.E., and F.H. Bormann. 1974. Linkages between terrestrial and aquatic ecosystems.
Bioscience 24:447-456.
Lind, O.T. 1979. Handbook of Common Methods in Limnology. Second edition. C.V. Mosby
Co., St. Louis. 199 pp.
Linthurst, R.A., D.H. Landers, J.M. Eilers, D.F. Brakke, W.S. Overton, E.P. Meier, and R.E. Crowe.
1986. Characteristics of Lakes in the Eastern United States. Volume I: Population Descrip-
tions and Physico-Chemical Relationships. EPA/600/4-86/007a. U.S. Environmental Protec-
tion Agency, Washington, D.C.
Lowe, R.L. 1974. Environmental Requirements and Pollution Tolerance of Freshwater Diatoms.
EPA/670/4-74/005. U.S. Environmental Protection Agency, Cincinnati, OH.
195
-------
MacArthur, R.H., and E.O. Wilson. 1967. The Theory of Island Biogeography. Princeton Univer-
sity Press, Princeton, NJ.
Machin, D. 1975. Note 385: On a design problem in growth studies. Biometrics 31:749-753.
Magnuson, J.J., B.J. Benson, and T.K. Kratz. 1990. Temporal coherence in the limnology of a
suite of lakes in Wisconsin, U.S.A. Freshwater Biol. 23:145-159.
Margalef, R. 1963. On certain unifying principles in ecology. Am. Nat. 97:357-374.
Marsh, G.P. 1864. Man and Nature. Harvard University Press, Cambridge, Massachusetts.
McCauley, E. 1984. The estimation of the abundance and biomass of zooplankton in samples.
Pages 228-265 in J.A. Downing and F.H. Rigler, eds. A Manual on Methods for Secondary
Productivity in Fresh Waters. Blackwell, Oxford, U.K.
Messer, J.J. 1990. EMAP indicator concepts. Pages 2.1-2.26 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, Research Triangle Park, NC.
Messer, J.J., R.A. Linthurst, and W.S. Overton. 1991. An EPA program for monitoring ecological
status and trends. Environ. Monit. Assess. 17:67-78.
Miller, R.R., J.D. Williams, and J.E. Williams. 1989. Extinctions of North American fishes during
the past century. Fisheries 13:22-38.
Miller R.R., J.D. Williams, and J.E. Williams. 1991. Extinctions of North American fishes during
the past century. Fisheries 14:22-38.
Moors, A., and R.J. O'Connor. 1992. Development of an Avian Index of Biotic Integrity from
Selected New England Lakes. Final Report on CR 818179-01 -0. Department of Wildlife,
University of Maine, Orono.
Morrison, D.F. 1970. The optimal spacing of repeated measurements. Biometrics 26:281-290.
Moyle, J.B. 1956. Relationships between the chemistry of Minnesota surface waters and wildlife
management. J. Wildl. Manage. 20:303-320.
Naiman, R.J., D.G. Lonzarich, T.J. Beechie, and S.C. Ralph. 1992. General principles of classifi-
cation and the assessment of conservation potential in rivers. Pages 93-123 in P.J. Boon,
P. Calow, and G.E. Petts, eds. River Conservation and Management. John Wiley & Sons,
London.
Nehlsen, W., J.E. Williams, and J.A. Lichatowich. 1991. Pacific salmon at the crossroads: stocks
at risk from California, Oregon, Idaho, and Washington. Fisheries 16:4-21.
Newell, A.D., and R. Hjort. 1991. Part II: Long-Term Monitoring Project Data Dictionary. In Date
User's Guide for the U.S. EPA Long-Term Monitoring Project: Quality Assurance Plan and
Data Dictionary. EPA/600/3-91/072. U.S. EPA Enviornmental Research Laboratory,
Corvallis, Oregon.
Nielsen, L.A., and D.L. Johnson, eds. 1983. Fisheries Techniques. American Fisheries Society,
Bethesda, MD.
196
-------
Noss, R.F. 1990. Indicators for monitoring biodiversity: a hierarchical approach. Conserv. Biol.
4:355-364.
Oberdorff, T., and R.M. Hughes. 1992. Modification of an index of biotic integrity based on fish
assemblages to characterize rivers of the Seine Basin, France. Hydrobiologia 228:117-130.
O'Connor, R.J., and M. Shrubb. 1986. Farming and Birds. Cambridge University Press,
Cambridge, New York. 290 pp.
Ohio EPA. 1988. Biological Criteria for the Protection of Aquatic Life. Ohio Environmental
Protection Agency, Columbus, OH.
Olsen, A.R. (ed). 1992. The Indicator Development Strategy for the Environmental Monitoring and
Assessment Program. EPA/600/3-91/023. U.S. Environmental Protection Agency, Corvallis,
OR.
Omernik, J.M. 1987. Ecoregions of the conterminous United States. Ann. Assoc. Am. Geogr.
77:118-125.
Overton, W.S. 1989. Effects of Measurement and Other Extraneous Errors on Estimated
Distribution Functions in the National Surface Water Surveys. Technical Report 129.
Department of Statistics, Oregon State University, Corvallis, OR.
Overton, W. Scott, Department of Statistics, Oregon State University, Corvallis, Oregon (personal
communications).
Overton, W.S., D. White, and D.L. Stevens, Jr. 1991. Design Report for EMAP, the Environmental
Monitoring and Assessment Program. EPA/600/3-91/053. U.S. Environmental Protection
Agency, Washington, D.C.
Paine, R.T. 1966. Food web complexity and species diversity. Am. Nat. 100:65-75.
Patalas, K. 1990. Diversity of the zooplankton communities in Canadian lakes as a function of
climate. Verh. Internat. Verein. Limnol. 24:360-368.
Patrick, R. 1972. Aquatic communities as indices of pollution. Pages 93-100 in W.A. Thomas,
ed. Indicators of Environmental Quality. Plenum, New York.
Paulsen, S.B., D.P. Larsen, P.R. Kaufmann, T.R. Whittier, J.R. Baker, D.V. Peck, J. McGue, R.M.
Hughes, D. McMullen, D.L. Stevens, J.L. Stoddard, J. Lazorchak, W. Kinney, A.R. Selle, and
R. Hjort. 1991. The Environmental Monitoring and Assessment Program (EMAP) - Surface
Waters Monitoring Research Plan, Fiscal year 1991. EPA/600/3-91/022. U.S. Environmental
Protection Agency, Corvallis, OR.
Payne, B.R., and R.M. Degraaf. 1975. Economic values and recreational trends associated with
human enjoyment of mongame birds. Pages 6-10 in Proceedings Symposium on Manage-
ment of Forest and Range Habitats for Nongame Birds. Gen. Tech. Rept. WO-1. U.S. Forest
Service.
Peck, D.V. (ed). 1992. Environmental Monitoring and Assessment Program: Integrated Quality
Assurance Project Plan for the Surface Waters Resource Group, Fiscal Year 1991.
EPA/600/X-91/080. U.S. Environmental Protection Agency, Las Vegas, Nevada.
197
-------
Pister, E.P. 1987. A pilgrim's progress from group a to group b. Pages 221-232 in J.B. Callicott,
ed. Companion to a Sand County Almanac. University of Wisconsin Press, Madison.
Plafkin, J.L., M.T. Barbour, K.D. Porter, S.K. Gross, and R.M. Hughes. 1989. Rapid Bioassess-
ment Protocols for Use in Streams and Rivers. EPA/444/4-89-001. U.S. Environmental
Protection Agency, Washington, D.C.
Platts, W.S., W.F. Megahan, and G.W. Minshall. 1983. Methods for Evaluating Stream, Riparian,
and Biotic Conditions. Gen. Tech. Rep. INT-138. U.S. Forest Service, Ogden, UT.
Pollard, J.E., and K.M. Peres, eds. 1991. Environmental Monitoring and Assessment Program:
Surface Waters Implementation Plan - Northeast Pilot Lake Survey, Summer 1991. U.S.
Environmental Protection Agency, Las Vegas, NV.
Power, M.E., W.J. Matthews, and A.J. Stewart. 1985. Grazing minnows, piscivorous bass, and
stream algae: dynamics of a strong interaction. Ecology 66:1448-1456.
Rao, J.N.K., and J.E. Graham. 1964. Rotation designs for sampling on repeated occasions. J.
Am. Stat. Assoc. 59:492-509.
Rawson, D.S. 1939. Some physical and chemical factors in the metabolism of lakes. Pages 9-26
in E.R. Moulton, ed. Problems of Lake Biology. Publ. 10. Amer. Assoc. Advance. Sci.
Washington, D.C.
Redford, K.H. 1992. The empty forest. Bioscience 42:412-422.
Reynolds, J.B., and A.L Kolz. 1988. Electrofishing injury to large rainbow trout. N. Amer. J. Fish.
Manage. 8:516-518.
Reynolds, R.T., J.M. Scott, and R.A. Nussbaum. 1980. A variable circular-plot method for
estimating bird numbers. Condor 82:309-313.
Robbins, C.S., D. Bystrak, and P.M. Geissler. 1986. The Breeding Bird Survey: Its First Fifteen
Years, 1965-1979. U.S. Fish and Wildlife Service Resource Publ. 157. Washington, D.C.
Ruttner, F. 1963. Fundamentals of Limnology. Translated by D.G. Frey and F.E.J. Fry. University
of Toronto Press, Toronto, Ontario.
Schindler, D.W. 1987. Detecting ecosystem responses to anthropogenic stress. Can. J. Fish.
Aquat. Sci. 44:6-25.
Schmidt, R.E. 1985. Zoogeography of the northern Appalachians. Pages 137-160 in C.H. Hocutt
and E.O. Wiley, eds. The Zoogeography of North American Freshwater Fishes. John Wiley
& Sons, New York.
Schneider, J.C. 1981. Fish Communities in Warmwater Lakes. Michigan Department of Natural
Resources, "Fisheries Research Report 1890. Lansing, Ml.
Scott, W.B., and E.J. Grossman. 1973. Freshwater fishes of Canada. Fish. Res. Bd. Can. Bull.
184. Ottawa, Ontario.
198
-------
Senner, S.E. 1986. Federal research on migratory nongame birds: is the U.S. Fish and Wildlife
Service doing its job? Am. Birds 40:413-417.
Sharber, N.G., and S.W. Carothers. 1988. Influence of electrofishing pulse shape on spinal
injuries in adult rainbow trout. N. Amer. J. Fish. Manage. 8:117-122.
Sharrock, J.T.R. 1974. The changing status of breeding birds in Britain and Ireland. Pages 204-
220 in D.L. Hawksworth, ed. The Changing Flora and Fauna of Britain. Academic Press,
New York.
Siegfried, C.A., J.A. Bloomfield, and J.W. Sutherland. 1989. Planktonic rotifer community
structure in Adirondack, New York, U.S.A. lakes in relation to acidity, trophic status and
related water quality characteristics. Hydrobiologia 175:33-48.
Skougstad, M.W., MJ. Fishman, L.C. Friedman, D.E. Erdman, and S.S. Duncan (eds.). 1979.
Method I-4600-78, Automated phosphomolybdate colorimetric method for total phosphorus.
In: Methods for Determination of Inorganic Substances in Water and Fluvial Sediments:
Techniques of Water Resources Investigations of the United States Geological Survey. Book
5, Chapter A1. U.S. Government Printing Office, Washington, D.C.
Smeltzer, E., V. Garrison, and W.W. Walker, Jr. 1989. Eleven years of lake eutrophication
monitoring in Vermont: a critical evaluation. Enhancing States' Lake Management
Programs: 53-62.
Smith, G.M. 1950. The Freshwater Algae of the United States. McGraw-Hill, New York.
Smith, R.L. 1977. Elements of Ecology and Field Biology. Harper & Row, New York.
Smith, S.H. 1968. Species succession and fishery exploitation in the Great Lakes. J. Fish. Res.
Board Can. 25:667-693.
Smol, J.P., and J.R. Glew. 1992. Paleolimnology. Pages 551-564 in W.A. Nierenberg, ed.
Encyclopedia of Earth System Science. Academic Press, San Diego, CA.
Snedecor, G.W., and W.G. Cochran. 1967. Statistical Methods. 6th ed. Iowa State University
Press, Ames.
Spellerberg, I.F. 1991. Monitoring Ecological Change. Cambridge University Press, New York.
Sprules, W.G. 1975. Midsummer crustacean zooplankton communities in acid-stressed lakes. J.
Fish. Res. Bd. Can. 32:389-395.
Stevens, Donald L, Jr., ManTech Environmental Technology, Inc., U.S. EPA Environmental
Research Laboratory, 200 SW 35th Street, Corvallis, OR 97333 (personal communication).
Stoermer, E.F., J.A. Wolin, C.L. Schelske, and D.J. Conley. 1985. An assessment of ecological
changes during the recent history of Lake Ontario based on siliceous algal microfossils
preserved in the sediments. J. Phycol. 21:257-276.
Tallent-Halsell, N.G., and G.D. Merritt. 1991. Environmental Monitoring and Assessment Program:
Surface Waters Component Northeast Lakes Pilot Study Field Training and Operations
Manual. U.S. Environmental Protection Agency, Las Vegas, NV.
199
-------
Terborgh, J. 1989. Where Have All the Birds Gone? Princeton University Press.
Tessier, A.J., and R.J. Horwitz. 1990. Influence of water chemistry on size structure of
zooplankton assemblages. Can. J. Fish. Aquat Sci. 47:1937-1943.
Tessier, A.J., and R.J. Horwitz. 1991. Analysis and Interpretation of Zooplankton Samples
Collected during Phase II of the National Lakes Survey. U.S. EPA/600/R-92/012.
Environmental Protection Agency, Corvallis, OR.
Thompson, Z. 1853. Natural History of Vermont. C.E. Tuttle, Rutland, VT.
Thornton, Kent W. FTN Associates, 3 Innwood Circle, Wuite 220, Little Rock, AR, 72211 (personal
communication).
Tonn, W.M., J.J. Magnuson, and A.M. Forbes. 1983. Community analysis in fishery manage-
ment: an application with northern Wisconsin lakes. Trans. Am. Fish. Soc. 112:368-377.
Underhill, J.C. 1985. The fish fauna of the Laurentian Great Lakes, the St. Lawrence lowlands,
Newfoundland and Labrador. Pages 105-136 in C.H. Hocutt and E.O. Wiley, eds. The Zoo-
geography of North American Freshwater Fishes. John Wiley & Sons, New York.
University of Maine. 1991. Laboratory Operations and Quality Assurance Report: EMAP/TIME
Summer 1991. Draft report. Environmental Chemistry Laboratory, Sawyer Environmental
Research Center, University of Maine, Orono.
Urquhart, N.S., W.S. Overton, and D.S. Birkes. 1991. Comparing Sampling Designs for Moni-
toring Ecological Status and Trends: Impact of Temporal Patterns. Technical Report 149.
Department of Statistics, Oregon State University, Corvallis.
U.S. EPA. 1983. Handbook for Chemical Analysis of Water and Wastes. EPA 600/4-79-020. U.S.
Environmental Protection Agency, Cincinnati, Ohio.
U.S. EPA. 1987. Handbook of Methods for Acid Deposition Studies: Laboratory Analyses for
Surface Water Chemistry. EPA/600/4-87/026. U.S. Environmental Protection Agency, Office
of Research and Development, Washington, D.C.
U.S. EPA. 1989. Handbook of Methods for Acid Deposition Studies: Field Methods for Surface
Water Chemistry. EPA/600/4-89/020. U.S. Environmnental Protection Agency, Washington,
D.C.
U.S. EPA. 1990. Biological Criteria: National Program Guidance for Surface Waters. EPA-440/5-
90-004. U.S. Environmental Protection Agency, Washington, D.C.
U.S. EPA. 1991. Total State Waters: Estimating River Miles and Lake Acreages for the 1992 Water
Quality Assessments 305(b) Reports. Draft.
U.S. FWS. 1981. Standards for the Development of Habitat Suitability Index Models. 103 ESM.
U.S. Fish & Wildlife Service, Washington, D.C.
Walters, C.J., J.S. Collie, and T. Webb. 1988. Experimental designs for estimating transient
responses to management disturbances. Can. J. Fish. Aquat. Sci. 45:530-538.
200
-------
Ward, R.C. 1989. A systems approach to design. Pages 37-46 in R.C. Ward, J.C. Loftis, and
G.B. McBride, eds. Design of Water Quality Information Systems. Information Series No. 61.
Colorado Water Resources Research Institute, Colorado State University, Fort Collins.
Weber, C.I. (ed). 1973. Biological Field and Laboratory Methods. EPA-670/4-73-001. U.S.
Environmental Protection Agency, Cincinnati, OH.
Whittier, Thomas R., ManTech Environmental Technology, Inc., U.S. EPA Environmental Research
Laboratory, 200 SW 35th Street, Corvallis, OR 97333 (personal communication).
Williams, J.E., J.E. Johnson, D.A. Hendrickson, S. Contreras-Balderas, J.D. Williams, M. Navarro-
Mendoza, D.E. McAllister, and J.E. Deacon. 1989. Fishes of North America endangered,
threatened, or of special concern: 1989. Fisheries 14:2-20.
201
* U.S. GOVERNMENT PRINTING OFFICE: 1993 — 750-002 / 60147
-------
EPA/620/R-93/0
o
CO
f»-D O
g§ 3
o j5 Q.
-SL
0?
~* c/>
"0 5'
=- (D
< en
JD yj
5T
C
in
(D
Cincinnati, OH
A
CJ1
ro
O>
oo
0
CD
:3
CD
-^
o"
m
3
3'
3
CD
~i
— )
5
33
CD
CO
CD
U
O
:r
5|
o"
3
Environmental
Tl
^
O
c?
o
o'
^
>
CQ
CD
3
O
^
-D
o R
3- "
«>f H.
QJ
as
S <"
CD t/>
?S
= •3
So
3-M
Cf =3
. "°
S ">
§:»
5§
» 3
5 ®
below label,
the upper
------- |