United States
Environmental Protection
Agency
Corvallis Environmental
Research Laboratory
Corvallis, Oregon 97333
&EPA
THE
August 1987
CONCEPT OF TIME
-------
THE CONCEPT OF TIME
TEMPORAL INTEGRATED MONITORING OF ECOSYSTEMS (TIME)
Prepared by
Kent W. Thornton
Forrest E. Payne
FTN Associates, Ltd.
Little Rock, AR
J. Ford
NCASI
Corvallis, OR
D. H. Landers
U.S. Environmental Protection Agency
Corvallis, OR
Prepared for
U.S. Environmental Protection Agency
Corvallis Environmental Research Laboratory
200 SW 35th Street
Corvallis, OR 97333
31 August 1987
-------
TABLE OF CONTENTS
1.0 THE CONCEPT OF TIME 1-1
1.1 TIME 1-1
1.2 IMPORTANCE 1-2
2.0 NAPAPANDAERP 2-1
2.1 NAPAP 2-1
2.2 AERP 2-1
2.2.1 Program Approach 2-2
2.2.2 NSWS 2-2
2.2.2.1 NLS 2-4
2.2.2.2 NSS 2-4
2.2.2.3 Design Considerations 2-4
2.2.2.4 Statistical Frame 2-4
2.2.2.5 Index Sample 2-6
2.2.2.6 Standardized Protocol 2-6
2.2.2.7 QA/QC 2-6
2.4 OTHER AERP ELEMENTS 2-6
3.0 THE CONCEPT OF TIME 3-1
3.1 BACKGROUND 3-1
3.2 GOALS 3-2
3.3 OBJECTIVES 3-2
3.4 HIERARCHICAL DESIGN 3-3
3.4.1 Regional Tier (Tier 1) 3-3
3.4.2 Seasonal Tier (Tier 2) 3-3
3.4.3 Research Tier (Tier 3) 3-5
3.4.4 Special Studies Tier (Tier 4) 3-5
3.4.5 Adaptive Frame 3-5
4.0 PREVIOUS AND ON-GOING ANALYSES 4-1
4.1 OVERVIEW OF ISSUES AND QUESTIONS 4-1
4.2 LTM ANALYSES 4-4
4.2.1 Data Quality 4-5
4.2.2 Comparison of LTM Lakes to the NLS Population 4-6
4.2.3 Regional Sample Size 4-6
4.2.4 Sources of Variability 4-9
i
-------
TABLE OF CONTENTS (Continued)
4.3 MODEL BASED EXTRAPOLATION STUDIES 4-12
4.4 ROLE OF BIOLOGICAL DATA 4-13
4.5 BIOLOGICALLY RELEVANT CHEMISTRY 4-13
4.6 CALIBRATION OF INDEX SAMPLES 4-14
4.6.1 Lakes 4-14
4.6.2 Streams 4-16
4.7 TREND DETECTION - INDIVIDUAL SYSTEMS 4-19
4.8 REGIONAL TREND DETECTION 4-26
4.9 EXPLORATORY ANALYSES 4-26
4.10 LESS THAN DETECTION LIMIT DATA 4-26
4.11 QA/QC INTERPRETATION 4-28
4.11.1 Overview of QA/QC Data Analyses 4-28
4.11.2 Measurement Uncertainty (System Precision) 4-28
4.12 COLLATION OF CANDIDATE SITES 4-29
4.13 DEPOSITION NETWORK EVALUATION 4-29
4.14 ALTERNATIVES AND OPTIONS 4-30
5.0 ALTERNATIVES AND OPTIONS 5-1
5.1 MONITORING APPROACHES 5-1
5.2 ALTERNATIVES 5-1
5.2.1 Population Inference 5-1
5.2.1.1 Model-based Approach 5-1
5.2.1.2 Design-based Approach 5-11
5.2.1.3 Combined Approach 5-11
5.2.2 Site Type 5-11
5.2.2.1 Rapid Response Sites 5-11
5.2.2.2 Cross-Sectional Sites 5-12
5.2.2.3 Special Interest Sites 5-12
5.2.2.4 Holotypes 5-12
5.2.2.5 Cluster Sites 5-12
5.2.3 Sampling Schemes 5-13
5.23.1 Fixed Sites 5-13
5.2.3.2 Fixed + Randomly Selected Sites 5-13
5.2.3.3 Re-Survey 5-13
ii
-------
TABLE OF CONTENTS (Continued)
5.2.4 Monitoring Protocol 5-14
5.2.4.1 Sampling Interval 5-14
5.2.4.2 Number of Stations 5-14
5.2.4.3 Monitoring Variables 5-14
5.3 OTHER TOPICS 5-16
6.0 PROPOSED TIME FRAME 6-1
6.1 GENERAL CONSIDERATIONS 6-1
6.2 GENERAL TIME DESIGN 6-3
6.2.1 Northeast 6-3
6.2.1.1 Subregions 6-3
6.2.1.2 Potential Deposition Effects 6-3
6.2.1.3 Regional Concerns 6-3
6.2.1.4 General Design 6-3
6.2.1.5 Rationale 6-4
6.2.2 Mid-Atlantic and Southeast 6-4
6.2.2.1 Subregions 6-4
6.2.2.2 Potential Deposition Effects 6-5
6.2.23 Regional Concerns 6-5
6.2.2.4 General Design 6-6
6.2.2.5 Rationale 6-6
6.2.3 Florida 6-7
6.2.3.1 Potential Deposition Effects 6-7
6.2.3.2 Regional Concerns 6-7
6.2.3.3 General Design 6-7
6.2.3.4 Rationale 6-7
6.2.4 Upper Midwest 6-8
6.2.4.1 Potential Deposition Effects 6-8
6.2.4.2 Regional Concerns 6-8
6.2.4.3 General Design 6-8
6.2.4.4 Rationale 6-9
6.2.5 West 6-9
6.2.5.1 Subregions 6-9
6.5.2.2 Potential Deposition Effects 6-9
iii
-------
TABLE OF CONTENTS (Continued)
6.2.5.3 Regional Concerns . 6-9
6.2.5.4 General Design 6-9
6.2.5.5 Rationale 6-10
6.3 PROPOSED TIME FRAME 6-10
7.0 SITE SELECTION 7-1
7.1 OVERVIEW 7-1
7.2 GENERALIZED SITE SELECTION PROCESS 7-1
7.2.1 Define Inclusion Criteria 7-1
7.2.2 Classify/Stratify the Array of Potential Sites
Into Categories By Geographical Region 7-1
7.2.3 Define the Desired Total Number of
Sites in Each Category 7-3
7.2.4 Select the Appropriate Number of
Sites in Each Category 7-3
7.2.5 Define and Apply Exclusion Criteria 7-3
7.2.6 Compare the Number of Sites Remaining
in Each Unit to the Desired Number of
Sites in Each Unit and Adjust as Necessary 7-3
7.3 SITE SELECTION CRITERIA: PROBABILITY SAMPLES 7-4
7.3.1 Tier 1 7-4
7.3.1.1 Examples of Inclusion Criteria 7-4
7.3.1.1.1 Regional Distribution 7-4
7.3.1.1.2 Phase 1 7-4
7.3.1.1.3 ANC Levels 7-4
7.3.1.2 Classification 7-4
7.3.1.3 Define the Desired Number of Sites
in Each Category 7-5
7.3.1.4 Site Selection 7-5
7.3.1.5 Exclusion Criteria 7-5
7.3.1.5.1 Present or Likely Future Disturbance 7-5
7.3.1.5.2 Access Problems 7-5
7.3.1.5.3 Catchment Size 7-5
7.3.1.6 Adjustment of Number of Sites 7-5
iv
-------
TABLE OF CONTENTS (Continued)
7.3.2 Tier 2 7-6
7.4 SITE SELECTION CRITERIA: RAPID-RESPONSE SITES 7-6
7.4.1 Overview 7-6
7.4.2 Inclusion Criteria 7-6
7.4.3 Classification 7-7
7.4.4 Define the Desired Total Number of Sites in Each Unit 7-7
7.4.5 Site Identification 7-7
7.4.6 Exclusion Criteria 7-7
7.4.7 Adjustments of Number of Sites 7-8
7.5 SITE SELECTION CRITERIA: SPECIAL-INTEREST SITES 7-8
7.5.1 Overview 7-8
7.5.2 Inclusion Criteria 7-8
7.5.3 Classification 7-9
7.5.4 Define the Desired Number of Sites in Each Unit 7-9
7.5.5 Site Identification 7-9
7.5.6 Exclusion Criteria 7-9
7.5.7 Adjustment of Number of Sites 7-9
8.0 TIMELINE/TIME PROJECT AND REPORT FORMATS 8-1
8.1 TIMELINE 8-1
8.2 TIME PROJECT REPORT FORMATS 8-1
8.2.1 Annual Reports 8-1
8.2.1.1 Tier 1 8-1
8.2.1.2 Tier 2 8-1
8.2.1.3 Tier 3 8-6
8.2.1.4 Tier 4 8-6
8.2.1.5 QA/QC Results 8-6
8.2.1.6 Extended Analysis of Previous QA/QC Results 8-6
8.2.2 Biennial Reports 8-6
9.0 REFERENCES 9-1
v
-------
LIST OF FIGURES
Figure 2.1. Conceptual strategy of Regionalized
Integrative Studies within the Aquatic
Effects Research Program 2-3
Figure 2.2. Regions and subregions of the United States
used to define target populations for the
National Surface Water Survey 2-5
Figure 3.1. Hierarchical approach for TIME with increased
complexity of studies and decreased number of
sites as one moves up the hierarchy 3-4
Figure 4.1. Example output of cumulative distribution _
frequency curves (F(X)) for ANC, pH, and SO4
from the ELS data for Central New England (1C)
and Maine (IE). LTM lakes in these subregions
are indicated as open circles 4-7
Figure 4.2. Example of trilinear diagram output. Trilinear
diagrams for anions and cations or the
Adirondacks (1A), Central New England (1C) and
Maine (IE). Open circles represent LTM lakes,
while dashed and dotted lines indicate
percentiles of the ELS population density. The
smaller triangles to the right of the
trilinears indicate the percent of the ELS
population found in each of the subtriangles
(80th; 50th; »20th; 5th) 4-8
Figure 4.3. Estimated population distributions of
alkalinity for lakes with ANC < 400 ueq/1 in
the fall of 1984 and 1986 4-17
Figure 4.4. Estimated population distributions of
alkalinity for lakes with ANC < 400 ueq/1 in
the fall of 1984 and spring of 1986 4-17
Figure 4.5. Comparison of estimated population distri-
butions for pH (A), and ANC (B), based on
length of stream reaches, from the three spring
ana one summer sampling intervals in the
Southern Blue Ridge Subregion (Kauffmann,
unpublished data) 4-18
Figure 4.6. Detectable change in ANC vs number of equally
spaced sampled assuming a linear trend. Curves
labeled a,b,c,d,e correspond to standard
deviations (Loftis and Ward, 1987a) 4-21
VI
-------
LIST OF FIGURES (CONTINUED)
Figure 4.7. An example of a paired comparison plot showing
no significant differences (Overton 1987) 4-27
Figure 4.8. An example of a paired comparison plot showing
significant differences (Overton 1987) 4-27
Figure 6.1. Proposed general design frame for the TIME
project 6-2
Figure 7.1. Overview of the site selection process 7-2
Figure 8.1. A sample format for presenting data generated
in Tier 2 8-5
vii
-------
LIST OF TABLES
Table 4.1. Examples of issues being addressed during the
designed of the TIME project 4-2
Table 4.2. Percentage of variability contributed bv lake,
year, and seasonal effects (Newell et al.
1987) 4-10
Table 4.3. The percent of the total variance contributed
by the various sources or components (Payne
et al. 1987) 4-11
Table 4.4. Characteristics for spring to fall relation-
ships of ANC, pH, SO/ , and Ca (Newell et
al. 1987) .T. 4-15
Table 4.5. Detectable changes in ANC (ueq/1) vs number
of independent samples, assuming a linear
trend. Significance level = 20%; power = 80%
(Ward and Loftis, 1987a) 4-22
Table 4.6. Seasonal Kendall tau results from LTM lakes
in the Upper Midwest stratified by ANC
categories (Payne et al. 1987) 4-25
Table 5.1. Alternative Categories of TIME 5-2
Table 5.2. Alternative approaches for population inference
in Tier 1 - Regional Sampling 5-3
Table 5.3. Alternative approaches to site types 5-5
Table 5.4. Alternative approaches for sampling schemes 5-7
Table 5.5. Alternative approaches for monitoring protocol 5-8
Table 5.6. Variables monitored in the NSWS and Phase II of
the ELS 5-15
Table 8.1. Gantt Chart reflecting monthly schedule for
TIME 8-2
viii
-------
THE CONCEPT OF TIME
TEMPORAL INTEGRATED MONITORING OF ECOSYSTEMS (TIME)
BY
K.W. THORNTON1, F.E. PAYNE1, J. FORD2, AND D.H. LANDERS3
1.1 TIME
The Temporal Integrated Monitoring of Ecosystems (TIME) project is a
proposed long-term monitoring program to assess the future affects of acidic
deposition on aquatic ecosystems. The TIME project is intended to address the
following questions:
o What are the early and on-going regional trends in surface water
acidification or recovery?
o What are the relationships between the observed patterns and trends in
surface water chemistry and regional patterns and trends in atmospheric
deposition?
o Do these observed patterns and trends correspond with model forecasts
of future regional patterns in surface water chemistry (e.g., the EPA
Direct/Delayed Response Project)?
The TIME project currently is in the conceptual design phase. The purpose of
this document is to:
o Describe the current concept of the TIME project and its relation to the
EPA Aquatic Effects Research Project (AERP) and National Acid
Precipitation Assessment Program (NAPAP);
o Discuss on-going and proposed analyses to improve and refine the
Concept for preparation of a Research Plan;
1 FTN Associates, Ltd., Little Rock, AR
^ NCASI, Corvallis, OR
i Environmental Protection Agency, Environmental Research
Laboratory, Corvallis, OR
1-1
-------
o Discuss the possible options and alternatives for aquatic monitoring
programs;
o Describe the general TIME frame proposed for each region of concern;
and
o Elicit comments and constructive criticism of the proposed Concept of
TIME.
1.2 IMPORTANCE
Long-term monitoring is absolutely essential to determine if additional
aquatic systems will become acidic or will recover in the future. Evaluating the
effectiveness of any mandated emission control procedures must be based on high
quality, long-term records. Long-term, with respect to acidic deposition, is
measured in decades, not 2-4 years. Verifying model forecasts of future surface
water acidification or recovery can be accomplished only through comparisons with
long-term records of surface water chemistry. It is not possible to corroborate long-
term forecasts without having concommitant long-term data (Simons and Lam
1980). Likens (1983) stressed the importance of long-term monitoring in
understanding and detecting subtle environmental changes that may be occurring
and are difficult or impossible to detect from short-term or fragmented records^
Establishing long-term, high quality monitoring programs may represent the highest
priority in environmental research (Likens 1983).
In hearings before the U.S. House of Representatives Subcommittee on
Natural Resources, Agriculture Research and Environment (GPO, 1985), it was
noted that, in general, long-term monitoring programs in the U.S. are an ad hoc
collection of diverse public and private programs, many of which suffer from:
o Design and operation inadequacies;
o Parochialism in purpose and approach;
o Lack of comparable data from one system to another; and, above all,
o Lack of coordination.
In spite of the fact that environmental monitoring is essential to the
implementation of all major environmental statutes and is critical for the detection
of future environmental crises, consistent, reliable data to support environmental
policy-making in this country still do not exist (GPO, 1985). A major problem in
1-2
-------
assessing the effects of acidic deposition on aquatic systems has been the lack of
long-term records (NRC1986).
The purpose of the TIME project is to design and implement, through
NAPAP, a coordinated long-term monitoring effort that will obviate many of the
criticisms associated with environmental monitoring programs.
1-3
-------
2.0 NAPAP AND AERP
2.1 NAPAP
The Acid Precipitation Act of 1980 (PL 96-294) established an Interagency
Task Force to develop and implement a comprehensive National Acid Precipitation
Assessment Program (NAPAP). The purpose of NAPAP is to increase our
understanding of the causes and effects of acidic deposition on the environment.
The activities of various federal agencies engaged in acid deposition research are
collectively funded through this Program. The U.S. Environmental Protection
Agency (EPA) is one of the federal entities cooperating through NAPAP and is the
agency responsible for Task Group 6 - Aquatic Effects. The major EPA program in
Task Group 6 is the Aquatic Effects Research Program (AERP).
22 AERP
The AERP is focusing on four primary policy questions:
o What is the extent and magnitude of past damage attributable to acidic
deposition?;
o What damage is expected in the future under various deposition
scenarios?;
o What is the target loading of sulfate below which damage would not be
expected?; and
o What is the rate of recovery if sulfate deposition decreases?
To provide answers to these four policy questions, the AERP has focused on
biologically relevant changes in chemistry resulting from long-term and short-term
(i.e. episodic) acidification. The component projects, either ongoing or planned,
within AERP that are addressing these policy questions are the National Surface
Water Survey (NSWS), the Direct/Delayed Response Project (DDRP), the
Watershed Manipulation Project (WMP), the Episodic Response Project (ERP),
and the TIME project. These projects are addressing four major elements of the
policy and assessment questions:
o Quantification of the chemical status and extent of surface waters at risk
(NSWS,ERP);
2-1
-------
o Prediction of the future chemical and biological changes within aquatic
ecosystems (DDRP);
o Confirmation of these predictions and development of an improved
understanding of controlling mechanisms (WMP, ERP); and
o Corroboration or verification of these results and findings through long-
term monitoring (TIME).
2.2.1 Program Approach
The NSWS, DDRP, WMP, ERP, and TIME collectively comprise the
Regionalized Integrative Studies (RIS). Each project in the hierarchy builds on the
findings of the previous projects (Figure 2.1). The RIS approach has a broad scale
perspective with emphasis on identifying and characterizing regional patterns in
surface waters. Acidic deposition occurs at regional and national scales but regional
and subregional differences in aquatic resources are critical to understanding the
effects of acidic deposition on these resources. The AERP is targeted to define the
subpopulations of aquatic resources at risk as a result of acidic deposition within
subregions. This approach allows a large number of systems to be described and
subsequently classified at regional scales. The resulting classification, then, permits
successively smaller subsets of systems to be selected for study. This also permits
extrapolation of in-depth, process-oriented research to better understand broad-
scale regional patterns. Site-specific, process-oriented research is essential and is
used to guide the broader-scale research program and to develop hypotheses for
testing on these scales. The mechanisms observed in lakes and streams typical of a
regional population can then be extrapolated with quantifiable/known confidence
to a regional or national scale.
Initial RIS activities used a large scale classification study (NSWS) to identify
regional patterns and characteristics of surface water chemistry. Subsequently,
more detailed characterization and process-oriented research on selected systems at
the subpopulation levels will provide an understanding of underlying mechanisms
responsible for the regionally observed patterns and effects.
2.2.2 NSWS
The NSWS represents the foundation of the AERP and has two major
components - the National Lake Survey (NLS) and the National Stream Survey
(NSS). The NSWS will be described briefly because of its importance in
2-2
-------
Conceptual Strategy
NSWS
Status and Extent
c
LU /
o
5
"5
h- 1
*o
MM
15
, >
CD
Q.
O
3
o
o
3D
D
Verification
WMP
Figure 2.1. Conceptual strategy of Regionalized Integrative Studies within the
Aquatic Effects Research Program.
2-3
-------
understanding several design aspects of the TIME project and several issues related
to long-term monitoring, in general.
2.2.2.1 NLS
The NLS is a two-phased project. Phase I consisted of fall index sampling in
lakes in the Eastern U.S. during 1984 and Western lakes in 1985 (Figure 2.2); Phase
II consisted of seasonal sampling (i.e., spring, summer and fall) in Northeastern
lakes during 1986 (Linthurst et al. 1986, Landers et al. 1987, Thornton et al. 1986).
2.2.2.2 NSS
The NSS also consisted of two components: A Pilot Survey conducted during
the spring and summer of 1985; and the NSS-Phase I conducted during the spring of
1986 in the Eastern U.S. (Figure 2.2), (Messer et al. 1986). Index sampling also was
used in the NSS.
2.2.2.3 Design Considerations
The NSWS incorporated several factors important in the design of a long-term
monitoring program. These factors were:
o A statistical frame with probability samples that permitted regional
population estimates of the status and extent of various surface water
chemistry attributes;
o An index sampling approach for surface water chemistry;
o A standardized sampling and analysis protocol; and
o Extensive QA/QC on sample collection, chemical analysis, data
management, and statistical analysis.
2.2.2.4 Statistic^ Frame
The NSWS was designed within a statistical frame with probability sampling.
Each lake or stream sample was selected with a known inclusion probability or
probability of being selected from the regional target population. This probability is
used to assign a weight to the physical or chemical attribute measured in the lake or
stream sample. Calculations can then be made to estimate population means,
medians, variances, or other descriptive statistics for the population. The important
consideration is that the lake or stream sample can be statistically related
2-4
-------
West
Pacific NW2
California*
Northern Rockies2
upper Midwest
NE Minnesota1
NC Wisconsin1
Adirondack*1
UP ol
Michigan1
Upper Great Lakes Area' 7 f poconos Catskllls
\ I \ /X«
Appalachian Plateau5
Central Rockies2
Ridge and Valley5
Southern Rockies2
Ozark Plateau*
Piedmont4
Florida1
Florida*
~ NLS
B NSS
¦ Overlap NLS/NSS
'ELS Phase-1
2WLS Phase-1
3NSS Pilot
4NSS Screening
5NSS Phase-1
Northeast
Maine1
Central and
Soulharn
New England1
Glaciated Highlands
of PA, NJ, and NY5
Chesapeake1
Mid-Atlantic
Southern Blue Ridge10
Southeast
Figure 22, Regions and subregions of the United States used to define target populations for the
National Surface Water Survey.
-------
back to the target population and the proportion of the population represented by
those lake or stream attributes can be estimated.
2.2.2.5 Index Sample
It is obvious that one sample, from one location, at one time of the day, in a
specific season of a particular year, cannot characterize the complex chemical
dynamics of a lake or stream. Such a sample is justified only in the sense that it is
an index to the essential characteristics of the system (Linthurst et al 1986). Even if
two or three samples were taken, these samples would still serve only as indices
because understanding the dynamics of a single system requires much more detailed
study (Linthurst et al 1986). The NSWS was designed to describe the characteristics
of populations of lakes and streams and estimate among lake/stream differences in
water chemistry rather than within lake/stream differences. The index concept is
appropriate for these types of comparisons.
2.2.2.6 Standardized Protocol
Standardized sampling procedures and analytical methods were used
throughout the NSWS. One of the major problems encountered in many earlier
studies was the use of different procedures and methods, which significantly reduced
or eliminated comparisons among data and studies. Data comparability is
important for regional studies but critical for long-term monitoring to detect subtle
trends in aquatic systems.
2.2.2.7 OA/PC
Quality assurance and quality control were an integral part of the NSWS.
QA/QC procedures were implemented prior to sample collection in the field and
followed throughout the program from field sampling and processing of samples to
laboratory spikes and splits to data validation and verification. Data of known
quality are important for any long-term monitoring program.
2.4 OTHER AERP ELEMENTS
Other AERP projects also are integrated with the TIME project, in addition
to the NSWS. Verification of the DDRP forecasts can occur only by collecting data
through TIME. The WMP and ERP process-oriented studies provide insight into
the causal mechanisms that might control changes in the regional patterns observed
2-6
-------
in aquatic systems through TIME. TIME provides a regional perspective for site-
specific studies and permits a better understanding of the relation between the
specific site attributes and the characteristics and possible response of other sites in
the region.
2-7
-------
3.0 THE CONCEPT OF TIME
3.1 BACKGROUND
In 1982, EPA initiated a program for long-term chemical monitoring of
surface waters under NAPAP. The EPA long-term monitoring (LTM) program was
initially designed to detect and measure chemical trends in low ANC surface waters
across atmospheric deposition gradients. In July 1982, the Aquatic Effects Task
Group (TG-E) organized an ad hoc committee to develop the framework for a
national network and a standardized sampling and analysis chemical monitoring
protocol to guide the LTM Program. In January 1983, a draft protocol was
prepared and subsequently used as a provisional guide for the Program. After
extensive peer review, this document was revised and approved as the basis for the
EPA Program. All EPA-supported cooperators were required to follow these
standardized protocols. However, some exceptions were made to allow continuity
with historical precedents, and QA/QC procedures were not finally standardized
until 1985. By 1985, EPA was supporting the monitoring of 121 lakes and reservoirs
and 23 streams in 11 states.
The NSWS, initiated in 1984, was designed in a statistical frame with
probability samples that permitted population estimates of the status and extent of
various surface water chemical attributes. Standardized sampling and analysis
protocols were an integral part of the NSWS program including extensive QA/QC
on sample collection, chemical analysis, data management, and statistical analysis.
Because the LTM program was initiated prior to the NSWS, the LTM was unable to
address some of the systematic problems associated with long-term monitoring
programs. Several of these systematic problems, therefore, were associated with the
LTM program.
Primarily, the LTM lakes and streams were chosen with an unknown
probability from an unknown population so quantifying regional trends is difficult.
Further, the sampling and analysis and QA/QC protocols were not totally
standardized among monitoring projects. Because of these concerns, the continuing
need for monitoring data to assess changes in surface water quality, and the
acquisition of an extensive regional surface water data base through the NSWS, a
re-evaluation of the scope and design of regional-scale, long-term monitoring has
been undertaken. The previous LTM program was concluded in 1987. The LTM
data, however, are being extensively analyzed and compared with the NSWS data
3-1
-------
for TIME planning purposes. Several of these analyses are discussed in the next
chapter, 4.0 PREVIOUS AND ON-GOING ANALYSES. Several of the LTM
sites will probably be incorporated in the TIME Project following characterization
of these sites with respect to specific regional subpopulations of interest.
To provide a regional-scale assessment of the effects of acidic deposition on
aquatic ecosystems, a long-term monitoring program needs to incorporate
representative site selection, measurement of biologically relevant chemical
variables, standardized analytical methods and quality assurance protocols, and a
sampling scheme that permits long-term changes in chemical response to be
differentiated from episodic changes and short-term daily, monthly, or annual
periodicities. The monitoring program must be predicated on a clear set of goals
and objectives.
32 GOALS
The TIME Project has as its goals to:
o Estimate the regional proportion and subpopulation physiochemical
characteristics of lakes and streams that exhibit early and on-going
trends of surface water acidification or recovery;
o Compare patterns and trends in observed surface water chemistry to
forecasts made using empirical or process-oriented procedures; and
o Determine the relationships between patterns and trends in atmospheric
deposition and trends in surface water chemistry for defined
subpopulations of aquatic resources in areas particularly susceptible to
acidification or recovery.
3.3 OBJECTIVES
In order to achieve these goals, the TIME project has the following objectives:
o Provide an early and ongoing indication of regional trends in surface
water acidification or recovery, using the most appropriate techniques to
detect such trends;
o Quantify, with known certainty, for defined subpopulations of lakes and
streams;
The rate at which changes in relevant chemistry are occurring;
3-2
-------
The subpopulation characteristics of the affected lakes and/or
streams; and
The regional or subregional extent of these systems.
o Compare trends in local and regional atmospheric deposition with
regional trends in surface water chemistry.
3.4 HIERARCHICAL DESIGN
To achieve the TIME objectives, it is proposed the TIME project be designed
within an integrated, hierarchical frame. This hierarchical frame would be
sufficiently flexible to accommodate investigations ranging from broad spatial
pattern identification for regional trend detection to specific process-oriented
research to identify causal mechanisms. This hierarchy or tiered approach is
illustrated in Figure 3-1.
3.4.1 Regional Tier (Tier 11
The purpose of this level is to describe broad regional patterns and trends in
ecosystem attributes such as water chemistry. A statistical frame with probability
samples might represent the design for this tier. In the base or bottom tier, there
might be a large number of regionally distributed ecosystems that could be sampled
during an index period. This tier would include some ecosystems that might serve as
early indicators (i.e., rapid response systems) of increased acidification or recovery
and ecosystems that reflect the subregional characteristics or range in characteristics
of ecosystems in the regional target population. These two categories of ecosystems
are probably not equivalent because the most appropriate early indicator systems
are probably atypical of many ecosystems in the subregion.
3.4.2 Seasonal Tier (Tier 2)
The second tier of the hierarchy could include a smaller number of ecosystems
in each region to be sampled seasonally in lakes or bi-monthly in streams to identify
seasonal patterns or trends. Ecosystems with similar attributes could be paired or
triplicated. If similar changes occur in both or all three systems, the significance
attributed to the change, and the confidence in subsequent decisions, might be
greater than if changes were observed in only one ecosystem. The seasonal samples
also could be used to "calibrate" the index samples monitored in the regional tier.
This could allow the validity of the more extensive index samples to be assessed
3-3
-------
Complexity
of
Studies
Frequency
of
Samplin3
A. . /MS V . . . '
Figure 3.1. Hierarchical approach for TIME with increased complexity of studies
and decreased number of sites as one moves up the hierarchy.
3-4
-------
on an on-going, annual basis. It could also permit more detail for systems of
particular interest such as those that indicate early recovery or increased
acidification.
3.4.3 Research Tier (Tier 31
The third tier would integrate process-oriented study sites or intensively
monitored sites such as the WMP, Regional Episodic and Acid Manipulation
(REAM), Forest Effects Program study sites, Long-term Ecological Reserve
(LTER), USGS Hydrologic Benchmark sites, soils monitoring activities, or other
similar sites with the TIME sampling regime. Hypotheses about causal mechanisms
or processes controlling surface water acidification can be tested in this tier. These
sites also provide information on short frequency occurrences such as storm events
and can improve the understanding of why various changes in site specific and
regional patterns might be occurring.
3.4.4 Special Studies Tier (Tier 4)
The fourth and final tier represents special studies. This tier would investigate
specific patterns of change noted within and among subpopulations or regions. For
example, resurveying NLS lakes and NSS streams could provide excellent
corroboration of any trends indicated in the TIME systems at the subpopulation or
regional level. The probability sampling frame and population estimates have been
established for the NLS and NSS. Such resurveys could determine if there has been
a statistically significant change in surface water chemistry within the subregions or
regions.
3.4.5 Adaptive Frame
TIME is envisioned as an evolving project with a flexible, adaptive frame. As
our understanding of acidic deposition effects increases, new questions and
hypotheses will arise that might require monitoring revisions or adaptations. The
hierarchical structure or frame provides this flexibility while retaining long-term
continuity among sites in the lower tiers.
The TIME project is presently in the conceptual/ preliminary design phase.
The goals and objectives of the TIME project, listed at the beginning of this section,
have been defined and will continue to guide the design. Two workshops were
conducted in 1986, one of which was a NAPAP Interagency Watershed
3-5
-------
Coordination Workshop, to discuss various considerations for an integrated
monitoring program for both lakes and streams in regions potentially susceptible to
atmospheric deposition. An hierarchical approach was agreed to be a useful
approach for long-term monitoring if at least the bottom three tiers were
implemented.
A number of issues were identified during the workshops and during the
preliminary conceputal phases of TIME that are important in designing an
efficacious long-term monitoring program. Several studies have been completed or
are on-going to address these issues. These are discussed in the next chapter.
3-6
-------
4.0 PREVIOUS AND ON-GOING ANALYSES
4.1 OVERVIEW OF ISSUES AND QUESTIONS
There are a number of issues and questions that must be addressed in
designing a long-term monitoring project. In TIME, the following five major
categories of issues have been identified:
o Statistical frame/Site selection;
o Regional estimates and trend detection;
o Appropriate biological/chemical measurements;
o Data analysis and interpretation of QA/QC data and the formulation of
appropriate QA/QC protocols; and
o Reporting.
Examples of issues that have been considered under these five major categories are
listed in Table 4.1.
A number of analyses have been, or are being, performed to address these
issues. For example, data from EPA's Long-Term Monitoring (LTM) program has
been and is being used to identify characteristics of the LTM sites, estimate the
components of variance in selected chemical constituents through an analysis of
variance, and estimate the number of lakes and samples required to detect
differences in constituent concentrations and/or trends in constituent
concentrations.
Other analyses being performed include:
o Model-based population extrapolation procedures (Section 4.3);
o Reviews and evaluations of possible biological indicators and indices for
detecting trends in acidification or recovery of aquatic systems (Section
4.4);
o Evaluation of biologically relevant chemical constituents, rates of
change, and constituent concentrations (Section 4.5);
o Comparison of fall index samples with other seasons (Section 4.6.1);
o Trend detection analyses for individual aquatic systems and the effect of
constituent variability on trend detection (Section 4.7);
o Specific trend detection (Section 4.7);
4-1
-------
Table 4.1. Examples of issues being addressed during the designed
of the TIME project.
A. Statistical Frame/Site Selection
o Appropriate regions, subregions, and subpopulations for sampling,
o The need and/or desirability to redefine regions,
o Prioritizing regions,
o Population inference.
o Stream sampling design for regions outside those sampled in National
Stream Survey.
o Desired confidence levels and estimates of precision for constituents and
regional population projections,
o Required number of sites, samples, and frequency to achieve the desired
confidence and precision,
o Constituents to monitor and constituent variability,
o Information gained from a fall index sample versus index samples from
other seasons.
o Characterization of streams based on two spring samples,
o Procedure for selecting regionally representative lakes and streams,
o Evaluation of data from special sites for possible site inclusion in TIME,
o Multivariate exploratory procedures to identify unique subpopulations of
systems.
B. Regional Estimation and Trend Detection
o Extrapolation procedures for non-randomly selected aquatic systems
(i.e., those outside the NSWS frame),
o Monitoring duration required to detect a trend in a constituent at a given
confidence and precision level,
o Expected annual rates of change for ANC, pH, and SO4.
o Relation between number of lakes, among lake variance and the
required time to detect trends in ANC, pH, and SO4.
o Number of samples needed to describe subpopulation versus subregional
characteristics,
o Trend detection at individual sites,
o Regional trend detection procedures,
o Influence of less than detection limit data on trend analyses.
4-2
-------
Table 4.1. Continued.
C. Appropriate Biological/Chemical Measurements
o Aquatic Organisms or associations of organisms that indicate
acidification and/or recovery,
o Relationships between chemical constituent concentrations or rate of
change and biological effects.
D. QA/QC Data
o Formulating rapid QA/QC feedback loops to cooperators.
o QA/QC assessment of centralized vs satellite laboratories,
o Procedures for transition of laboratories,
o QA/QC evaluation procedures for data outside TIME,
o Inclusion procedures for historical data.
E. Reporting
o Types of reports,
o Format of reports,
o Frequency of reports.
4-3
-------
o Estimates of duration to detect trends (Section 4.7);
o Regional trend detection analyses (Section 4.8);
o Multivariate analyses to identify characteristic subpopulations of lakes
and streams in various regions (Section 4.9);
o The influence of less than detection limit data on trend detection and
statistical procedures for using censored data (Section 4.10);
o Analysis and interpretation of QA/QC data collected during the NSWS
and proposed QA/QC procedures for TIME (Section 4.11); and
o Deposition network evaluation (Section 4.13).
Each of the previous and on-going analyses will be discussed briefly below.
The data source, analyses, and results will be presented for each topic. The
emphasis is on the interpretation of these results and the implications for the design
of the TIME project.
4.2 LTM ANALYSES
Analyses have been conducted on the LTM data sets to:
o Assess the quality of the data;
o Determine seasonal and annual variation;
o Evaluate the efficiency of the monitoring design;
o Describe the relationship of LTM lakes to the population of NSWS
lakes;
o Estimate the number of lakes to be sampled in the TIME Project;
o Examine data for temporal trends; and
o Evaluate sources of variability.
Lake monitoring sites for EPA's LTM Program were selected in areas where
annual wet sulfate deposition ranged from 0-10,10-20, and 20-30 kg/ha and average
volume-weighted precipitation pH ranged from about 5.5 to 4.3. Specific site
selection criteria for LTM sites included:
o ANC < 200 ueq/L"l;
o No recent land use changes or prospects of future changes;
o Absence of local atmospheric pollutant sources;
4-4
-------
o Accessibility;
o No recent history of forest fires or logging in the watershed; and
o No chemical manipulation of lake or watershed.
Lakes were selected for monitoring in the following areas:
o The Adirondack Mountains in New York;
o Vermont;
o Maine;
o The Upper Midwest (i.e., Minnesota, Wisconsin, and Michigan);
o The Southern Appalachians (TVA reservoirs in the Southern Blue Ridge
Province of North Carolina, Tennessee, and Georgia); and
o The Rocky Mountains.
4.2.1 Data Quality
Through an analysis of quality assurance samples, Newell et al. (1987)
estimated the quality of the LTM data, including precision and accuracy for selected
constituents. The available audit data indicated a relative overall bias of 0.3%
+.38.4% but were too variable to estimate between-laboratory bias. Blank and
duplicate data were insufficient to estimate possible contamination or precision.
LTM cooperators, however, did comply with EPA guidelines on the number of
blanks and duplicate samples. Analysis of the QA data did indicate three important
considerations for future long-term monitoring programs. First, a more rigorous
QA/QC program could identify and quantify inter-laboratoiy bias, and would
enable smaller confidence bounds to be placed about the data collected. Such a
program should require an increased number of blank and duplicate samples and
include periodic analyses of stable audit samples so that inter-laboratory bias could
be assessed. Second, future QA/QC designs should include minimum numbers of
blank and duplicate samples, so that the number of samples collected will be large
enough to permit estimates of precision with known confidence. Third, to insure
that accuracy is maintained over time, an audit system should be designed that is
capable of detecting changes in bias from one year to the next, as well as detecting
bias among laboratories at any point in time.
4-5
-------
4.2.2 Comparison of LTM Lakes to the NLS Population
Analyses performed by Newell et al. (1987) to compare LTM lakes to the
ELS-Phase I lakes included cumulative distribution frequency curves and trilinear
plots (Figures 4.1 and 4.2, respectively).
Generally, LTM lakes occupied the lower portion of the ANC cumulative
frequency distribution for ELS lakes (Figure 4.1) in the various ELS subregions.
The trilinear diagrams indicated the occurrence of many sulfate dominated lakes in
the LTM population. Cumulative frequency distribution curves indicated sulfate
was moderately higher and dissolved organic carbon was lower in the LTM lakes
than the median values in ELS lakes.
Although the LTM lakes represented low ANC systems, three problems
precluded their use for regional extrapolation or estimation. First, the lakes in the
LTM program were not chosen from a defined population, so the inclusion
probability for the LTM lakes is unknown. Second, although one could argue that
LTM samples were chosen from lakes with ANC <200 ueq/1 and thus represent the
population of low ANC lakes, the strict statistical representation cannot be
quantified. Third, LTM lakes were clustered within subregions and, therefore, may
not exhibit spatial variability found across the subregion.
Based on their physical-chemical characteristics with respect to the ELS lakes,
the LTM lakes might represent rapid response lakes. Rapid response lakes might
provide an early indication of increased surface water acidification or recovery in
the subregion or region. This is an important class of lakes to incorporate in TIME.
As will be discussed later, the strict statistical representation of rapid response lakes
in the population is not critical in the current concept of TIME. Further analyses
are being conducted to characterize rapid response and LTM systems.
4.2.3 Regional Sample Size
Newell et al. (1987) and Payne et al. (1987) estimated the number of lakes
(i.e., samples) in a subregion to detect a given change in the subregional mean
concentration for selected constituents. In the Northeast, Newell et al. (1987)
estimated from 13 to 88 lakes per state were required to detect a 10 ueq/1 change in
ANC (mean lake ANC concentrations <. 10 ueq/1) and from 5 to 64 lakes per state
to detect a 10 ueq/1 change in ANC (mean lake concentrations 20 < ANC <. 100
ueq/1). The number of lakes estimated to detect a 0.20 change in pH or 10%
change in sulfate was similar to the ANC estimates. As the change to be detected
4-6
-------
IC
0.0
-100 0
200
400 600
UK tuEO/U
BOO
1000
IE
400 600
INC luEO/U
1000
A
t
vl
9.0
9 0
1?0 180
S04 luEO/ll
¦<00
120
?40
300
S04 fuEO/ll
Figure 4.1. Example output of cumulative distribution frequency curves (F(X)) for ANC, pH, and SO^ from the
ELS data for Central New England (1C) and Maine (IE). LTM lakes in these subregions are
indicated as open circles.
-------
50
,16,
34
CI
>
kio.
88
1A
Ca
<
S3
33,
< Ca
i«Sr. ^ -
CI >
Figure 42. Example of trilinear diagram output. Trilinear diagrams for anions and cations for
the Adirondacks (1A), Central New England (1C), and Maine (IE). Open circles
represent LTM lakes, while dashed and dotted lines indicate percentiles of the ELS
population density. The smaller triangles to the right of the trilinears indicate the
percent of the ELS population found in each of the subtriangles
( ~ 80th; 50th; 20th; 5th).
4-8
-------
became larger, the estimated number of lakes to be sampled decreased. The above
estimates were calculated with an and level of 0.10.
In the Upper Midwest, estimates of lakes to be sampled to detect a 10 ueq/1
change ranged from 20 to 412 lakes per state (mean lake concentrations <20 ueq/1)
and from 30 to 78 lakes per state (mean lake concentrations 20 < ANC <. 100
ueq/1). As above, lake estimates were similar for SO4 and pH and decreased as the
desired change to be detected increased. The and levels used in these estimates
were 0.10.
Payne et al. (1987) estimated lake sample sizes for the Upper Midwest region
and the Adirondack subregion, rather than by states. Estimates of the number of
lakes to detect a 10 ueq/1 ANC change with mean ANC concentrations <, 20 ueq/1
was 10 for the Upper Midwest and 24 for the Adirondacks at an level of 0.10. The
estimated number of lakes in the Upper Midwest for lakes with mean ANC
concentration 20 < ANC <_ 100 uq/1 was 30 and 83 for the Adirondacks at level of
0.10.
Although Newell et al. (1987) and Payne et al. (1987) used different
techniques, the results are not grossly different. About 30-100 lakes should allow
detection of changes if they occur in nearly all subregions. Variance estimates are
expected to decrease in the TIME project.
4.2.4 Sources of Variability
It was recognized that variability associated with the LTM data sets could be
contributed by a variety of sources. Both Newell et al. (1987) and Payne et al.
(1987) used nested analysis of variance methods to partition the total variance into
the variance components explained by each source. The results of these analyses
are presented in Table 4.2 and 4.3. In general, among-lake variance was the largest
single component of constituent variability, generally explaining greater than 80% of
the variability.
Although Newell et al. (1987) and Payne et al. (1987) analyzed the data in
slightly different ways, both investigators determined spatial variability generally
accounted for 70 percent of the variance in ANC, pH, and SO4 compared with
temporal variability (i.e., among seasons and among years). This implies it might be
more important to have more lakes to detect regional changes than a few lakes with
more samples throughout the year.
4-9
-------
Table 42. Percentage of variability contributed by lake, year, and
seasonal effects (Newell et al., 1987).
4
Lake b
Year
SeasfYearV
Year
ANC
NY
88.39
1.08
3.93
0.98
VT
86.65
0.86
4.55
1.65
ME
9236
1.09
2.10
2.07
UPMW
89.70
5.12
2.02
0.35
MN
83.00
2.43
2.74
6.82
MI
90.59
1.99
1.30
1.61
WI
57.95
19.49
7.92
3.98
SBR
95.07
3.66
0.64
0
EH
NY
86.93
1.44
6.36
0.95
VT
86.66
1.16
4.96
1.15
ME
9630
1.23
135
0.38
UPMW
93.80
5.62
4.86
0
MN
72.98
2.69
7.86
3.93
MI
98.10
2.08
0.23
0
WI
7739
8.85
3.18
4.72
SBR
45.00
15.57
4.07
14.75
so4"2
NY
82.16
1.29
6.00
3.54
VT
57.54
10.69
12.39
6.56
ME
96.26
0.35
1.37
1.04
UPMW
87.49
3.51
2.46
4.87
MN
69.48
15.45
4.21
7.67
MI
91.89
4.42
1.75
0.71
WI
86.78
10.03
1.46
0
SBR
93.56
1.15
1.70
1.58
Lake by
5.63
6.29
2.39
2.80
5.01
4.51
10.66
1.43
4.34
6.07
0.74
0
12.54
1.04
5.86
20.62
7.01
12.83
.097
1.67
3.18
1.24
3.66
2.01
* Seas(Year) refers to a nested effect of season within each year.
4-10
-------
Table 43. The percent of the total variance contributed by the various sources or
components (Payne et al., 1987).
I. H+
% of Variance
Source of
Variation
Upper Midwest
20 < ANC <.100
<20uea/l uea/1
Adirondacks
20 < ANC <.100
<20 uea/1 uea/1
Investigator
0.0
0.0
Subregions
59.1
17.7
--
Year
2.0
0.5
0.0
0.0
Season
0.0
25.9
2.6
5.4
Lakes
37.6
12.0
78.4
67.3
Error
1.3
44.0
18.9
27.3
II. ANC
% of Variance
Source of
Variation
Upper Midwest
20 < ANC <.100
<20 uea/1 uea/1
Adirondacks
20 < ANC <.100
<20 uea/1 uea/1
Investigator
0.0
0.0
--
Subregions
41.1
43.2
Year
13
6.6
0.3
0.0
Season
12.3
4.7
19.1
23.4
Lakes
27.4
6.7
38.1
46.0
Error
18.0
38.9
42.5
30.6
III. S04
% of Variance
Source of
Variation
Upper Midwest
20 < ANC <.100
<20 uea/1 uea/1
Adirondacks
20 < ANC <.100
<20 uea/1 uea/1
Investigator
4.8
0.0
Subregions
9.6
70.8
Year
2.7
1.4
4.4
11.9
Season
0.0
0.0
0.0
2.1
Lakes
80.7
17.8
86.7
57.8
Error
2.2
10.0
4-11
8.9
28.2
-------
4.3 MODEL BASED EXTRAPOLATION STUDIES
One of the issues being considered in TIME is the appropriate statistical
frame. The statistical frame needs to incorporate regionally representative lakes
including rapid response and special interest systems so that regional trends can be
detected, if regional changes are occurring. Regional estimates can be obtained
using either a design-based or a model-based approach.
Regional population estimates, using a design-based approach, are based on a
set of regionally representative probability samples collected within a statistical
sampling frame (i.e., statistical design or design-based). Because the inclusion
probability for each lake and stream is known, these sample systems can be
weighted to provide unbiased regional estimates. This approach was successfully
used in the NSWS.
Regional population estimates also can be provided for non-randomly
selected lakes and streams using a model based approach. A model based approach
relates non-randomly selected systems with a subpopulation of randomly selected
systems assuming these systems have similar system characteristics and attributes.
The model-based approach assumes the statistical frame relating the probability
sample to the target population is an adequate model for relating non-random
systems to the target population. This approach, for example, might be used to
relate results from the LTM lakes to the subregional or regional target population.
The general form of a model-based prediction equation might be:
y = g(x)
With the probability or design-based analysis, y is determined directly. With
the model-based analyses, the relationship between the x's (the vector of attributes
that help to predict y) and the y's (the attributes of interest) is determined. Then y
is predicted by substituting x's for the systems not randomly sampled and
distributions are generated.
Both design-based (direct estimates) and model-based (indirect estimates)
approaches to determine population estimates with the ELS-Phase II data are
being investigated (Overton, personal communication). For the ELS-Phase II
sample, direct estimates gave better precision than originally thought and a
relatively small sample size per region (i.e., 50 to 100 lakes) could be used to
4-12
-------
provide estimates with acceptable precision. Both approaches are being considered
for inclusion in TIME.
4.4 ROLE OF BIOLOGICAL DATA
Biological indicators or indices may be able to provide early indications of the
onset of acidification or recovery processes of aquatic systems. Reorganization of
biological communities begins early in the acidification process and can actually
precede the subtle changes in surface water chemistry involved with acidification or
recovery. Observed biological changes are the clearest signal that changes in
chemical parameters are biologically significant.
Several literature reviews to determine the usefulness of various organismal
groups as early warning indicators have been initiated. These reviews will be
followed by a workshop to resolve divergent views and come to a state-of-the-art
evaluation of possible useful approaches for integrating cost effective biological
information into long-term monitoring program.
Indices are being investigated for all resources (lakes and streams) and
regimes. The range of organisms being considered include phytoplankton,
periphyton, zooplankton, benthic invertebrates, and fish. Different approaches may
be utilized in different areas depending upon the nature and composition of the
local communities. The focus will be on indices that can be demonstrated to
complement and or extend information already contributed by the chemical
monitoring program.
4.5 BIOLOGICALLY RELEVANT CHEMISTRY
Efforts relating changes in water chemistry to the biological community have
been restricted primarily to fish (Haines and Baker 1986). The reasons for the
emphasis on fish are:
o General ecological processes and functions appear to be relatively
robust, with significant ecosystem impacts only at acidity levels above
those that affect major fish species (Altshuller and Linthurst 1984;
Schindler et al. 1985);
o Effects of acidification on fish, for the most part, appear to be direct
rather than mediated through changes in food availability or quality
(Rosseland 1985; Baker 1986);
4-13
-------
o The number of studies directed at quantifying the dose response for
affects of acidification is substantially greater for fish than other aquatic
organisms (Haines and Baker 1986); and
o Effects on fish and declines in the fishery resource can more readily be
expressed in terms directly relevant to public interest and resource
utilization.
The key chemical variables that influence fish response to acidification are
pH, inorganic aluminum and calcium (Altshuller and Linthurst 1984). Because lake
pH and levels of inorganic aluminum and calcium are often highly correlated within
any given region, estimates of effects may be reasonably predicted by pH. The most
data, and most reliable estimates, are available for populations of fish in lakes, and
for four species in particular: brook trout, lake trout, white sucker, and brown
bullhead (Baker and Harvey 1984).
Although studies relating changes in water chemistry to the biological
community have been primarily concerned with fish, the literature review and
workshop discussed in Section 4.4 will expand the consideration to other organisms.
These reviews will be used, in part, to define the chemical variables that should be
included in the TIME project.
4.6 CALIBRATION OF INDEX SAMPLES
4.6.1 Lakes
During the NLS, lakes were sampled once during fall overturn. The
assumption was that within lake and among lake variability would be minimised
during this period of time. Fall overturn was selected because it is a relatively stable
period and is a broader, more predictable period than spring ice-out.
Newell et al. (1987) investigated the spring to fall relationships for the primary
variables, ANC, pH, SO4 and Ca using LTM data to determine if fall samples might
serve as predictors of lake chemistry in other seasons of the year. General linear
models were used to investigate the relationship. The results are presented in Table
4.4. In general, coefficients of multiple determination were high (R^ > 0.8).
Although a strong relationship between spring and fall samples was generally
observed, the analyses in Table 4.4 do not indicate how well fall data reflects spring
chemistry (Newell et al. 1987). When spring to fall relationships are parallel from
one year to the next, fall values alone may be used to indicate long term trends. If
4-14
-------
Table 4.4. Characteristics for spring to fall relationships of ANC, pH, SO4*2, and Ca (Newell et al, 1987).
ANC pH S04*2 Ca
R2
n
Slope
IV
R2
n
Slope
IV
R2
n
Slope
IV
R2
n
Slope
rv
NY
0.90
66
m< 1
yr
0.92
64
m=l
yr
0.83
66
m< 1
yr
0.91
66
m< 1
yr
VT
0.85
91
m< 1
0.79
98
m< 1
yr
0.55
90
m< 1
yr
0.93
89
a
yr
ME
0.97
12
m< 1
0.95
12
m=l
0.97
11
m> 1
0.89
12
m=l
MN
1.00
20
b
yr
0.88
23
m=l
0.90
23
m=l
yr
0.92
23
m=l
yr
MI
0.97
32
m< 1
yr
0.97
36
m< 1
yr
0.96
36
m 1
yr
SBR
0.95
23
m=1
o.r^
23
m< 1
0.88
23
m=l
0.97
22
m=1
Legend:
R2 = the coefficient of determination for the model,
n = the sample size used in the model.
IV = the indicator variable used in the model,
m = the slope as determined by the model.
? slope not different from 1 for one year and less than 1 for all others.
" slope varies from < 1 to > 1 across years.
S slope > 1 two out of three years, and not different from 1 in the remaining year.
d p = 0.292
-------
the slopes are different, however, as they were for Ca in Vermont (VT) and
Michigan (MI), then trends in fall values might not be good predictors of trends
occurring among spring values.
Preliminary analyses also have been conducted using the ELS-Phase II data.
Seasonal Phase II population distributions were compared to the Phase I population
distributions for various constituents. The estimated population distributions for
lakes with ANC < 400 ueq/1 in the Fall of 1984 (Phase I) and the Fall of 1986
(Phase II) are shown in Figure 4.3. The estimated population distributions for these
same lakes in the Fall of 1984 and the spring of 1986 are shown in Figure 4.4. The
consistency of the Fall index over a two-year period and the relationship of Fall to
Spring suggest that using fall index chemistry to monitor and detect seasonal and
annual trends is feasible.
4.6.2 Streams
As in the NLS-Phase I, the National Stream Survey (NSS) relied on samples
taken during an appropriate season from a regionally representative sample of
water bodies to provide an "index" of the chemical characteristics of the regional
population. The choice of the index sampling period was a compromise between
minimizing season chemical variability and maximizing the expected probability of
sampling during chemical conditions potentially limiting for aquatic organisms.
Ford et al. (1986) summarized the results of four recent studies of seasonal
and short term variability in six second and third order streams in the Catskill
Mountains of New York (Murdoch, 1986), the Laurel Hills of Pennsylvania (Witt
and Barker, 1986), the Southern Blue Ridge Province of North Carolina and
Tennessee (Olem, 1986) and the Ouachita Mountains of Arkansas (Nix et al. 1986).
Minimum flow-weighted pH values and concentrations of base cations and ANC
occurred during the spring at most sites. Therefore, spring appeared to be the most
appropriate index sampling period because streamwater ANC was typically low and
life stages of aquatic biota, sensitive to low pH, were likely to be present. The index
sampling period for NSS was chosen as the time period following snowmelt or
winter rains but prior to leaf out. A further restriction was to avoid sampling within
24 hours of a significant rain event.
A comparison of year to year data is not available, but Figure 4.5 illustrates
comparisons of estimated population distributions of pH and ANC from three
spring and one summer sampling intervals used in the NSS - Pilot Study in the
4-16
-------
1.0
0.8
0.6
0.4
Fall 1984
Fall 1986
(Raw ELS II data)
0.2
0.0
100
-100
200
300
400
0
500
Alkalinity (ueq/l)
Figure 4.3. Estimated population distributions of alkalinity for lakes with ANC <400 ueq/l
in the fall of 1984 and 1986.
1.0
0.8
0.6
0.4
Fall 1984
Spring 1986
(Raw ELS II data)
0.2
0.0
-100
0
100
200
300
400
500
Alkalinity (ueq/l)
Figure 4.4. Estimated population distributions of alkalinity for lakes with ANC <400 ueq/l
in the fall of 1984 and spring of 1986.
4-17
-------
o
o.
o
l_
Q.
o
>
1.0
0.8
0.6
| 0.4
3
E
o 0.2
0.0
1
i
"
A
LJ^
-
W
-
-
/
V
-
_
Spring
Summer
i -*¦»
'
7
pH
8
o 0-6
o 0.2
Spring
Summer
300 400
ANC (ueq/l)
600
Figure 4.5. Comparison of estimated population distributions for pH (A), and ANC (B), based
on length of stream reaches, from the three spring and one summer sampling intervals
in the Southern Blue Ridge Subregion (Kauflmann, unpublished data).
4-18
-------
Southern Blue Ridge Subregion (Kaufmann, unpublished data). In this subregion,
there was little difference in population distributions taken approximately three
weeks apart between March 15 and May 15. Preliminary results indicate that a
spring sampling index is appropriate for estimating population distributions.
Additional analysis regarding the feasibility of spring sampling are ongoing.
Based on the NSS - Pilot Survey, streams were sampled twice in the base flow
period as opposed to three times in the NSS. There was little difference in
population distributions for important variables during the spring sampling window.
Currently the NSS scientists are analyzing among season and within season
variability to assess and re-evaluate the utility of the spring baseflow chemical index
in:
o Evaluating chemical conditions most limiting to aquatic organisms;
o Predicting chemical conditions at different times of the year; and
o Representing the variation of chemistry within the spring chemistry.
Historical data from special interest sites are being used for the above analyses.
A number of other analyses are currently being conducted. The NSS scientists
are analyzing historical data from special interest sites and selected NSS sites
revisited in 1987 to assess the year to year stability of the spring index sample. This
analysis is especially important to the NSS because sample year 1986 was a drought
year in many regions of the Southeast. Hydrologic data (i.e., flows and
precipitation) are being examined to refine knowledge of regional patterns in
seasonal flow such as elevated spring flow, summer low flow and seasonal
frequency-duration analyses of stormflow. This information will be used in the
design of the stream monitoring program.
4.7 TREND DETECTION - INDIVIDUAL SYSTEMS
Loftis and Ward (1987a) examined trend detection procedures using quarterly
and annual samples assuming linear trends, normal distributions and independent
samples. Using LTM data for five lake regions and one stream region, they
determined a range of standard deviations appropriate for ANC, pH, and SO4
values in the eastern United States.
Approximate cumulative changes in mean ANC, SO4, and pH, which would
be detectable at a significance level of <* = 0.2 and power of ft = 0.2 for each region
4-19
-------
and for several group sizes and length of record, were determined from generic
curves as illustrated in Figure 4.6. Figure 4.6 presents the change detectable for
ANC as a function of sample size and a wide range of standard deviations, which
might be appropriate for each variable.
Using regional standard deviations and the curves presented in Figure 4.6 the
detectable change in a given constituent versus the number of independent samples
can be estimated (e.g., Table 4.5) For example, the regional standard deviation for
ANC in Region 1A was 25.7. This regional standard deviation lies between curves c
and d in Figure 4.6. For one lake and 5 samples (annually, quarterly, etc.) it would
take a change of approximately 110 ueq/1 to be detectable. However, if 5 samples
were collected from 4 lakes then the standard deviation of the sample mean, x, over
n lakes would be o^/n or, in our example, 25.7/ 4 = 12.85. This regional standard
deviation now lies between curves d and e and the detectable change would be
approximately 55 ueq/1. A number of similar curves were generated by Loftis and
Ward (1987a) for ANC, pH and SO4 and can be found in their report.
The results indicate that changes can be detected sooner and with fewer
samples as the number of lakes sampled increases. In addition, change could be
detected earlier if more samples are collected per year. However, it is important to
remember these analyses were conducted under the assumptions of linear trends,
normal distributions and independent samples. Seasonality and autocorrelation
were not considered in these initial analyses. Therefore the time to detect change
may increase.
Analyses on trend detection in lake water quality also were performed by
Loftis et al. (1987b). Using a range of statistical characteristics for selected LTM
lakes and Twin Lakes in Colorado (long-term data provided by U.S. Bureau of
Reclamation), they compared alternative trend detection techniques considering the
effects of seasonal variation, non-normality, and serial correlation.
Seasonal variation in Adirondack and Vermont Lake pH, ANC, and SO4
values ranged from minimal seasonal differences to the case where the maximum
quarterly mean and/or standard deviation was two to five times the minimum
quarterly mean/standard deviation. Because of the small record length, no attempt
was made to show seasonality was statistically significant. However, all three
variables showed obvious seasonality in at least one region and ANC was seasonally
more variable than sulfate.
4-20
-------
600
500 -
ANC
alpha ¦ 0.20, beta >0.20
400 -
\
cr
a>
3
«
9
C
O
300 -
200 -
100 -
Standard Deviation
a- 160
b- 80
c- 40
0L-
0
20 30
Number of Observations
Figure 4.6 Detectable change in ANC vs number of equally spaced sampled assuming a linear trend.
Curves labeled a,b,c,d,e correspond to standard deviations (Loftis and Ward, 1987a).
-------
Table 4.5. Detectable changes in ANC (ueq/1) vs number of independent
samples, assuming a linear trend. Significance level = 20%;
power = 80% (Ward and Loftis, 1987a).
Region
(Std.Dev.^
NLS 1A
(25.7)
NLS 1C
(17.7)
NLS3A
(26.5)
NLS IE
(7.1)
NLS 2
(22.4)
STREAMS
(39.8)
No. in
Number of Samples
Group
5
10
20
30
40
1
110
70
45
38
32
4
55
33
22
18
15
16
27
17
10
8
6
64
12
8
5
4
3
1
80
45
30
25
22
4
40
24
17
12
10
16
20
12
8
7
6
64
10
6
4
3
3
1
110
70
45
38
32
4
55
33
22
18
15
16
27
17
13
9
7
64
12
8
5
4
3
1
32
20
12
9
8
4
17
10
6
4
3
16
8
4
3
2
2
64
<8
<4
<3
<2
<2
1
100
60
40
35
30
4
50
31
21
18
15
16
23
16
12
9
7
64
11
8
5
4
3
1
170
105
68
57
48
4
95
51
33
28
26
16
42
27
18
15
12
64
20
12
9
8
7
4-22
-------
Most of the records studied by Loftis et al. (1987b) appeared to be normally
distributed and, in general, log transformations or removal of quarterly means did
not increase or decreased the number of data records that appeared to be normal.
However, Loftis et al. (1987b) warn against assuming normality for TIME
monitoring.
Loftis et al. (1987b) also investigated autocorrelation of ANC, pH, and sulfate
values in several lakes. They found significant seasonal and serial autocorrelation
for all three constituents.
Although Loftis et al. (1987b) did not attempt to analyze LTM data records
for trend, they did set a range of trend magnitudes (i.e., 0.2%, to 2.0% of the
standard deviation per quarter) based on U.S. Geological Survey benchmark
streams (NRC, 1986) and long-term forecasts of lake quality using the MAGIC
model. These trend magnitudes were then used in Monte Carlo simulations under a
large array of alternative conditions.
The trend testing procedures used by Ward and Loftis (1987b) included:
o Seasonal Kendall-tau with correction for serial dependence (Hirsch and
Slack, 1984)
o Kendall-tau with quarterly means removed (Snedecor and Cochran,
1980).
o Analysis of covariance for simulataneous estimation of quarterly means
and trends, assuming normal distribution.
o Modified t test procedure assuming normal distribution.
o Analysis of covariance using ranks of data.
o Modified t test procedure using ranks.
Preliminary examination of the results indicated an analysis of covariance on
ranks was the best procedure when the errors were normally distributed. For larger
data records, the seasonal Kendall-tau test performance was similar to the analysis
of covariance on the ranked data and, in about 7 percent of the simulations, had
greater power than the analysis of covariance in ranked data. The above results
should be considered preliminary since additional checks and analyses are on-going.
Miah (personnel communication) is investigating the use of Bayesian
approaches to characterize system error and to detect trends. In this approach,
4-23
-------
characteristics of the system error are used to detect a significant change in SO4,
ANC, or pH lake values over time.
The methods Miah is using (posterior trend probability and regional trend
analyses) are useful in a long term monitoring program where historical data over a
long period of time is not available and the research objectives are to estimate
regional and subregional trends by measuring a limited number of lake samples
from a regionally representative number of lakes. Another advantage of the
proposed technique is that it provides a means of relating the trend phenomena by
using a quadratic regional trend with other external factors such as atmospheric
deposition.
Payne et al. (1987) conducted preliminary trend detection analyses and slope
estimations with the Seasonal Kendall-tau (Hirsch et al. 1982; Smith et al. 1982) in
the LTM data set for the Upper Midwest. These analyses were conducted both on
individual lakes and combined lakes in a given season. In general, very few trends
(P
-------
Table 4.6. Seasonal Kendall tau results from LTM lakes in the Upper Midwest stratified
by ANC categories (Payne et al. 1987).
Probability
Slope
Trend
Fall
0.154
1.002
ANC (ueq/1)
ANC <20
Spring
0.113
2.512
Summer
<0.001
2.810
Fall
20 < ANC <.100
Spring Summer
Probability 0.845
Slope -0.383
Trend
0.518
-0.675
0.043
1.877
Probability
Slope
Trend
Fall
<0.0001
-3.954
SO4 (ueq/1)
ANC <20
Spring
0.087
-1.873
Summer
0.0006
-3.438
Probability
Slope
Trend
20 < ANC <.100
Fall Spring Summer
0.190 0.114 0.003
-1.457 -2.081 -2.160
Significant
4-25
-------
4.8 REGIONAL TREND DETECTION
Preliminary analyses to detect regional trends is on-going. During the June
1987 Data Analysis Workshop, Overton discussed regional trend detection. Using
the 1984 NSWS data as a reference frame, a description of change can be based on
comparisons to this reference. A change on a regional basis could be determined by
using either a paired comparison or a repeated measures test on individual lakes. A
chi-square test could be used to determine if the distributions were different or if
the medians were different. Figure 4.7 and 4.8 are examples of a generic plot
comparing data collected at two different time periods (e.g., 1984 and 1986). If
there has been no change in the parameter distribution, the points should cluster
around the main diagonal, which represents a 1:1 slope (Figure 4.7). If a change in
the parameter distribution has occurred, then a shift in the points above or below
the diagonal would be observed (Figure 4.8).
Further refinements in these analyses could include testing for changes in
distributions above and below the median or above and below quartiles.
Comparisons can also be made between distributions of slopes, ranks, or sign and
the percent of the population that showed a change.
4.9 EXPLORATORY ANALYSES
Multivariate exploratory analyses have been used to delineate subpopulations
of lakes or streams of interest in various regions and subregions. Cluster analyses,
for example, were used to identify subpopulations of lakes and their characteristics
during the ELS-Phase II design. These analyses were important in developing
appropriate criteria for stratifying the population of lakes prior to the random
selection process.
Multivariate analyses of NSS stream chemistry are being used to identify acid
mine drainage impacts on streams. These analyses also are being used to classify
NSS sample stream and aid in the selection of study sites for AERP, ERP, and
TIME projects. In conjunction with the multivariate analyses, ion-ratios, field
reconnaissance, aerial photos and maps are being used.
4.10 LESS THAN DETECTION LIMIT DATA
Magoun and Malcolm (1987) investigated techniques to analyze samples that
contain below detection limit data or samples that are censored. The techniques
used were grouped into three basic categories:
4-26
-------
Total number
of points plotted ¦ 50
chi-square ¦ .72
DF - I
22
28
1984 Fall Constituent Values
Figure 4.7. An example of a paired comparison plot showing no
significant differences (Overton, 1987).
Total number
of points plotted ¦ 50
chi -square ¦ 50
DF I
0
50
1984 Fall Constituent Values
Figure 4.8. An example of a paired comparison plot showing
significant differences (Overton, 1987).
4-27
-------
o Subjective analyses;
o Data removal; and
o Statistical estimation.
Subjective analytical procedures are convenient but biased. Data removal
techniques result in the loss of valid data when the measured values are less than or
equal to the maximum detection limit value.
Magoun and Malcolm (1987) recommended the maximum likelihood
technique for the less than detection limit data. Maximum likelihood estimators are
statistically based and all the data are used in the estimation procedure. Although
not necessarily unbiased, the maximum likelihood have properties of asympotically
efficiency, squared-error consistency, and invariance. All are highly desirable
properties when unbiased estimators cannot be obtained. This procedure will be
further evaluated for use in the TIME project.
4.11 QA/QC INTERPRETATION
The QA/QC interpretation of NSWS data is on going. Activities related to
the development of a QA plan for TIME; based on these results, are presented
below.
4.11.1 Overview of OA/OC Data Analyses
At the present time, QA data from NSWS are being analyzed for precision,
accuracy, interlaboratory bias, and system detectability in order to determine the
range of values expected for various analytes measured using similar methodologies
in TIME. Also, the QA procedures that were the most useful in the NSWS (i.e.,
provided the most information) are being determined with reference to quality
control, evaluation of system performance, and planning for future studies.
A QA workshop is planned for early November 1987. The goal of the
workshop is to obtain a sufficient review of past QA experience to develop a QA
plan for TIME that will be rigorous but flexible.
4.11.2 Measurement Uncertainty (System Precision)
Mericus et al. (1986) investigated the nature of measurement uncertainty in
the ELS-I duplicate sample data and presented a means of estimating and
expressing precision as a continuous linear function. Because uncertainty is
4-28
-------
described as a continuous function, an estimate of the expected precision is
available for observations throughout the range of analyte concentrations. In
addition, each precision estimate reflects the true nature of the analyte
concentration to precision response relationship. Expected precision for the TIME
project might be estimated from these relationships and concentrations measured in
the NSWS or from other records.
4.12 COLLATION OF CANDIDATE SITES
There are at least three sources from which candidate sites will be identified:
the NSWS; meetings with local experts; and surveys. A task has been initiated to
identify sites at which aquatic monitoring has been and is continuing to be
conducted as well as other sites with high quality data in the geographic areas of
interest. A previous NAPAP watershed survey identified over 700 watersheds
where some type of monitoring and/or research activity was occurring. This survey
is now being expanded to include monitoring of aquatic systems by federal, state,
local and private agencies or organizations. The results of these surveys will be used
to develop a candidate list of sites that might be included in TIME.
4.13 DEPOSITION NETWORK EVALUATION
There is concern that the density of the current deposition network may be not
adequate to provide data suitable for relating increased acidification or recovery of
lakes and streams with changes in deposition. The density of the deposition network
is generally sparse compared with precipitation monitoring stations. The location of
NADP/NTN sites will be compared with the location of TIME sites, once the TIME
sites are selected, to determine if the distribution is satisfactory to correlate trends
between atmospheric deposition and surface water quality. A similar analysis will
be conducted for precipitation monitoring stations. Of particular concern is the
elevation of precipitation collectors in relation to the watershed elevation.
Differences in elevation between the lake/stream and precipitation collector can
confound correlation and regression analyses between precipitation inputs and
system responses.
4-29
-------
4.14 ALTERNATIVES AND OPTIONS
The studies reported above, workshop discussions and coordination activities
have identified numerous alternatives and options for designing a long-term
program. The next section list many of these alternatives and discusses these
advantages and disadvantages.
4-30
-------
5.0 ALTERNATIVES AND OPTIONS
5.1 MONITORING APPROACHES
Designing a long-term monitoring project is a fluid process. The design must
reflect the project objectives, regional geography, climate, resource characteristics,
and other factors. There are no standard designs suitable for all situations. Various
alternatives for designing a long-term monitoring program were identified during
workshops, previous and on-going analyses, and the preparation of the conceptual
plan.
Six general alternative categories have been identified. These categories and
several specific approaches in each category are listed in Table 5.1. Each
alternative has advantages and disadvantages with respect to:
o The TIME objectives;
o Precision and confidence in the alternative; and
o Relative cost.
The advantages, disadvantages, relative precision, and relative cost of each
alternative are listed in Tables 5.2 - 5.5 and discussed briefly in this chapter.
The design of a flexible, evolving long-term monitoring program should
incorporate those alternatives best suited for a particular region or subregion. A
uniform design across all the regions of concern is not a pre-requisite for the TIME
Project. The goal is to provide an adequate design that will achieve the TIME
objectives in each region and subregion.
5.2 ALTERNATIVES
5.2.1 Population Inference
Three alternative approaches might be considered for making inferences
about trends in the regional population or extrapolating from the sampled lakes and
streams to the target population. These alternatives include a model-based, design-
based or combined approach for obtaining regional estimates.
5.2.1.1 Model-based Approach
The model-based approach was discussed in Section 4.3. The advantages and
disadvantages of this approach are listed in Table 5.2. A model-based approach
5-1
-------
Table 5.1 Alternative Categories of TIME.
A. POPULATION INFERENCE
1. Model-based Regional Estimation
2. Design-based Regional Estimation
3. Combined Approach
B. SITE TYPE
1. Rapid Response
2. Cross-Sectional
3. Special Interest
4. Holotype
5. Cluster
C. SAMPLING SCHEMES
1. Fixed Sites
2. Fixed + Random Selection Annually
3. Re-Survey
D. MONITORING PROTOCOL
1. Index Sampling 7. Phase I Chemical Parameters
2. Seasonal Sampling 8. Biologically Relevant
3. Continuous Sampling Chemical Parameters
4. Single Station/System 9. Biological Parameters/
5. Multiple Stations/System Indices
6. Limited Chemical 10. Other Parameters - Soils,
Parameters Veget., etc.
E. DATA MANAGEMENT AND ANALYSIS
1. STORET 6. Indiv. Sys. Trend Detection
2. Standard Data Mgt System 7. Regional Trend Detection
3. Customized Data Mgt. 8. Early Indication Analyses
4. Descriptive Statistics 9. Laboratory Reporting
5. Multivariate Statistics 10. User Distribution and Access
F. LOGISTICS
1. Contracts
2. Field Sampling
3. Laboratory Analyses
4. QA/QC
5. Database Mgmt
5-2
-------
Table 5.2. Alternative approaches for population inference in Tier 1 -
Regional Sampling.
A. MODEL-BASED SAMPLING
Advantages
1. Permits detection of regional
patterns and processes.
2. Permits the inclusion of
special interest and early
warning lakes
3. Can include non-randomly
selected systems.
4. Permits regional estimates of
various population attributes.
5. Can be used to estimate portion
of population exhibiting change.
6. Precision and confidence in
population estimates can be
calculated for any region,
subregion, or subpopulation.
Disadvantages
1. Distribution of lakes/streams
has to be similar to tne
distribution obtained through
probability sample (Phase I).
2. Data bases for a particular subregion
or resource type might not exist to
characterize subpopulation of interest.
3. Might exclude systems exhibiting
the fastest response to acidic
deposition.
4. Special interest and early warning
lakes may be monitored using
different field and analytical
techniques.
5. Existing data from special interest
and early warning lakes may
be of unknown quality.
B. DESIGNED-BASED
Advantages
1. Permits regional estimates of
various population attributes
2. Can be used to estimate pro-
portion of population exhibiting
change.
3. Can use existing NSWS frame
for regional estimation.
4. Various criteria can be used
to stratify the population
of lakes prior to sampling.
5. Precision and confidence in
population estimates can be
calculated for any region,
subregion, or subpopulation.
6. Trends associated with individual
lakes can be related directly
to regional proportions.
Disadvantages
5-3
1. Special interest lakes will likely
not be randomly selected
for inclusion in the program.
2. Might exclude systems exhibiting
the fastest response to acidic
deposition.
3. The Phase II lakes might
not represent the target
population of interest.
4. There might be insufficient
information to characterize or
identify the appropriate
subpopulation for long-term
monitoring.
5. To attain a desired precision and
confidence level might require too
large a sample or have different
precision estimates for
subpopulations.
6. Increased costs and logistical
problems.
-------
Table 52. Continued.
C. COMBINED APPROACH
Advantages
1. Permits regional estimate
of various population
attributes.
2. Can use existing NSWS frame
to estimate regional propor-
tion of population exhibiting
change.
3. Permits detection of regional
Patterns and processes,
ermits inclusion of special
interest and rapid response systems.
5. Various criteria can be used
to stratify the population
of systems prior to sampling.
6. Precision and confidence in
population estimates can be
calculated for any region,
subregion, or subpopulation.
7. Precision and confidence in
population estimates can be
calculated for any region,
subregion, or subpopulation.
Disadvantages
1. Limit special interest systems
for inclusion in the program.
2. Might limit systems exhibiting
the fastest response to acidic
deposition.
3. The Phase II lakes might not
represent the target population
of interest.
4. To attain a desired precision and
confidence level might require
too large a sample or have
different precision estimates
for subpopulations.
5. Distribution of special interest
lakes/streams has to be similar
to the distribution obtained
through the probability sample.
6. Data bases for a particular
subregion might not exist to
characterize subpopulation of interest.
7. Different field and analytical
techniques might result in data
of unknown quality.
8. Cost and logistic problems might
be excessive.
5-4
-------
Table S3. Alternative approaches to site types.
A. RAPID RESPONSE SYSTEMS
Advantages
1. Provides early indication
of increased acidification/
recovery.
2. Regional estimates possible through
model based approach.
3. Limited number of systems,
more measurements/better
QA/QC possible.
4. Any subpopulation of interest
can be used to stratify
prior to sampling.
5. Minimal cost.
Disadvantages
1. Presumes characteristics of
fast response systems known.
Important subpopulations might be
missed.
2. Limited number of systems,
low precision/confidence estimates.
3. Regional coverage inadequate
for all subpopulations, subregions,
so lower confidence in regional
patterns.
4. Stratification by other factors,
e.g., deposition might not be
practical due to low number of
systems in some areas.
B. CROSS SECTIONAL
Advantages
1. Permits regional, subregional
subpopulation estimates.
2. Permits testing (e.g., paired
comparisons or repeated measure
test) to determine if distribution
is changing.
3. A statistical frame for selecting
lakes/streams is already in place.
4. Good regional coverage, high
confidence in regional patterns.
Disadvantages
1. Regional patterns/changes may not
be applicable to individual lakes.
2. Autocorrelation among measurement
error could be a problem.
3. Greater number of systems
required to represent various
subpopulations.
4. Greater cost associated with
more sites.
C. SPECIAL INTEREST
Advantages
1. Significant information
already available.
2. Can be identified in any
region or subregion even if
no other class of site
occurs in the subregion.
3. Permits testing of ancillary
hypotheses.
4. Lower costs and logistical
problems.
Disadvantages
1.
2.
3.
4.
5.
Quality of historical records,
may not be documented.
Limited number of systems,
low precision/confidence
estimates.
Different analytical methods
probably used.
QA/QC might be unknown.
Inclusion probability unknown.
5-5
-------
Table 5.3. Continued.
D. HOLOTYPIC
Advantages
1. Permits specification of
lake types of interest.
2. Can provide early indication
of increased acidification/recovery.
3. Reduces number of lakes required
to characterize various
subpopulation responses.
4. Minimal cost.
Disadvantages
1. Presumes characteristics of
interest are known.
2. Limited number of sites for
regional estimation.
3. Holotype response might not
represent subpopulation response.
4. Confirmation of holotypic nature of
sample site is expensive and time
consuming.
5. Reduces precision because of small
sample size.
E. CLUSTER
Advantages
1. Concentrate efforts around
few calibrated watersheds
with satellite waterbodies
sampled less intensively.
2. Can compare sites in different
regions.
3. Reduced number of systems
required to detect subpopulation
changes.
4. Lower costs than cross-
sectional monitoring.
5. Estimates of system variances.
Disadvantages
1. Difficult to detect regional
trends.
2. Presumes the type of system
around which tne cluster is
developed is representative
of subpopulation.
3. Limited number of systems
for regional estimation.
4. Reduces precision because of small
sample size.
5. Extensive regional distribution of
rapid response systems unlikely.
5-6
-------
Table 5.4. Alternative approaches for sampling schemes.
A. FIXED SITES
1.
3.
4.
5.
Advantages
Continuous reference frame
for trend detection.
2. Permanent installation of
equipment
Increases logistical efficiency.
Data analyses procedures
standardized.
Cost-effective, budget can
be estimated accurately.
Disadvantages
1. Some unique/important resources
might never be sampled.
2. Might be biased by ease of
access, logistics or other factors.
3. Rapid response sites might
not be included.
4. Decreased flexibility in
modifying monitoring program.
B. FIXED PLUS RANDOM SELECTION ANNUALLY
Advantages
Disadvantages
1.
Continuous reference frame for
trend detection at fixed sites.
1.
2.
Permanent installation of
equipment for some sites.
2.
3.
Potential reduction in annual
costs per sampling effort.
3.
4.
Eventually obtain information
on all sites in the target
4.
population.
5.
6.
Reduction in data available
per lake/stream.
Comparisons between fixed sites
and randomly selected sites
tenuous.
Potential increase in costs because of
increased logistical difficulties.
Decreased logistical efficiency.
Increased QA/QC difficulties.
Requires annual site selection.
C. RE-SURVEY
Advantages
1. Reduction in number of
sites monitored annually.
2. Reduction in annual costs.
3. Reduction in annual logistics.
4. Continuous reference frame
for trend detection.
Disadvantages
1. Presumes representative sites to
detect change are selected.
2. Periodic substantial cost increases
to resurvey when changes
are detected.
3. Increases logistical difficulties
during re-survey.
4. Trend detection based on
limited temporal data.
5-7
-------
Table 5.5. Alternative approaches for monitoring protocol.
A. SAMPLING INTERVAL
Advantages
1. Index Sampling
o Permits adequate characterization
of regions, subregions, and
subpopulations based on one
sample per year during fall
overturn in lakes, two samples
per year during the spring m streams,
o Permits comparisons among
systems and over time,
o Reduction in costs,
o Reduction in human resources
demands (i.e., person hours).
2. Seasonal Sampling
o Permits adequate seasonal
characterization of regions,
subregions, and subpopulations.
o Potential to reference conditions
of lakes/streams in other places,
o Potential to reduce the time
necessary to detect a change,
o Permits characterization of
seasonal conditions and to
calibrate assumptions of index
sampling.
3. Continuous Monitoring
o Increased resolution in charac-
terizing lake/stream on a site
specific basis,
o Potential to increase resolution
in characterizing lake/stream
on regional, subregional, and
subpopulation basis,
o Episodic monitoring possible.
Disadvantages
1. Index Sampling
o Presumes index is
a reference condition of
lakes/streams in other
places and other times,
o Presumes fall overturn is
the "best" time to
index lakes,
o Presumes spring is "best"
time to index streams.
2. Seasonal Sampling
o Reduction m coverage of
regions, subregions, and
subpopulations.
o Potential to significantly increase
costs if tradeoff between number
of samples/year and monitoring
duration is not realistic,
o Inaccessibility of regions
(i.e., the West) during late fall,
winter, and spring.
3. Continuous Monitoring
o Reduction in number of
lakes and streams monitored,
o Increased cost,
o Potential to increase complexity
of data analyses without
concurrent gain in information,
o Increased human resource
demands (i.e., person hours).
5-8
-------
Table 5.5. Continued.
B. NUMBER OF SITES
Advantages
1. Single Station/System
o Reduction in analytical cost
o Increase efficiency of human
resource utilization,
o Index general characteristics
of lakes and streams.
2. Multiple Stations/System
o Permits evaluation of within
and among system variability,
o Increases precision of
estimates.
Disadvantages
1. Single Station/System
o Presumes limnetic spatial
variation insignificant,
o Does not permit examination
of patterns within lakes/streams.
2. Multiple Stations/System
o Increases demands on human
resources.
Increases analytical costs.
Increased complexity of data
analysis.
Information gains may not be
necessary to meet TIME objectives.
C. MONITORING VARIABLES
Advantages
1. Limited Chemical Parameters
o Eliminates contractual laboratory
expenses,
o Reduces field time,
o Reduces time committed to
data analyses.
2. Phase I Chemical Parameters
o Permits cohesiveness between
TIME and NSWS.
o Variables are related to
atmospheric deposition, and
surface water chemistry,
o Increase number of variables to
permit QA/QC checks and balances,
o Certain variables are biologically
relevant.
3. Biologically Relevant Chemicals
Parameters
o Relate directly to underlying
goals of TIME,
o Reduces analytical costs,
o Reduces costs.
Disadvantages
Limited Chemical Parameters
o Significantly reduces information,
o Reduces ability to address
cause and effect relationships,
o Increases subjectivity of
impacts of acidic deposition.
Reduces QA/QC checks and
balances.
Phase I Chemical Parameters
o Increases costs,
o Increases field time,
o Requires analytical laboratories,
o Presumes all appropriate
variables have oeen
identified.
Biologically Relevant Chemicals
Parameters
o Presumes all important biologically
relevant chemicals are known,
o QA/QC variables may not be
adequate for QA needs,
o Assumes parameters exert
comparable affect in all regions.
5-9
-------
Table 5.5. Continued.
Advantages
Disadvantages
4.
Biological Parameters/Indices
o May permit detection of subtle
shifts in biological community
structure before chemical
changes are detected,
o Permits concommittent comparisons
between biology and chemistry
o May reduce time to detect recovery
or acidification.
5.
Other Parameters - Soils,
Vegetation, etc.
o Increases understanding of
Patterns and processes,
rovides linkages between
atmospheric, watershed, water
quality.
o Provides ecosystem understanding.
Biological Parameters/Indices
o Increases human resource
requirements (i.e., specific
technical expertise),
o Increases costs
o Presumes appropriate biolog-
ical organisms for monitoring
are known,
o Presumes organisms can be
collected with a known
precision and confidence level,
o Presumes potential biological
interactions (i.e., differences in
competition and predation) can be
accounted for over a regional
distribution of lakes.
Other Parameters - Soils,
Vegetation, etc.
o Few sites can be supported,
o Substantial resource commitments
on a site specific basis.
5-10
-------
relates non-randomly selected systems to a sample of randomly selected systems
with similar characteristics and attributes. This approach assumes the statistical
frame relating probability samples to the target population is an adequate model to
relate the non-randomly selected systems to the target population.
5.2.1.2 Design-based Approach
The design-based (Probability Sampling) approach also was discussed in
Section 4.3. Its advantages and disadvantages are listed in Table 5.2. Regional
population estimates using a design-based approach are based on probability
samples (i.e., lakes or streams) collected within a statistical sampling frame.
Because the inclusion probability for each lake or stream is known, these sample
systems can be weighted to provide unbiased regional estimates. This was the
approach used in the NSWS to provide regional estimates.
5.2.1.3 Combined Approach
The design and model-based approaches also can be combined to provide
greater precision about the estimates. Various subregions or subpopulations might
have different sample numbers, which influence the precision of the estimate. For
some of these subregions, the model-based approach might provide more precise
estimates while the design-based approach might be more precise in other
subregions. Therefore, greater precision about the regional estimates can be
obtained by combining both approaches. The advantages and disadvantages of the
combined approach are listed in Table 5.2.
5.2.2 Site Type
Several types of sites or site combinations could be incorporated in a long-
term monitoring program. Several of these alternative site types are assessed in
Table 5.3.
5.2.2.1 Rapid Response Sites
Rapid response sites are systems expected to respond rapidly to acidic
deposition. This rapid response might be exhibited through changes in surface
water chemistry, aquatic biota or other indicators such as increased seasonal
variance. These sites would be considered to be the most sensitive (i.e., exhibiting
the fastest change) sites in a region.
5-11
-------
5.2.2.2 Cross-Sectional Sites
Cross-sectional sites represent the range or cross-section of the site types in
the region such as drainage lakes, seepage lakes, Clearwater systems, darkwater
systems, shallow lakes, etc. The sites would range from those sites expected to
respond rapidly to acidic deposition to sites with higher ANC and other attributes
expected to delay the response to acidic deposition. The range of site types and
attributes included in the NSWS represents the category of cross-sectional sites.
5.2.2.3 Special Interest Sites
Special interest sites generally are considered to be sites that were not
included in the NSWS but have special or unique attributes of interest. These
special attributes might be a long monitoring record, sites known to respond rapidly
to acidic deposition, good biological data on fish or other biota, or other important
characteristics. The inclusion probability or regional representation is generally
assumed to be unknown for special interest sites.
5.2.2.4 Holotvpes
Holotypic sites are sites selected to represent a particular subpopulation or
type of site. The subpopulation of lakes or streams might be perched lake (seepage)
systems, small headwater mountain streams, or a category of systems identified
through various exploratory analyses (i.e., cluster analysis). One or two sites
considered typical of sites in the subpopulation might be selected for monitoring
with any changes observed in these sites viewed as typical responses expected for
other sites in this subpopulation. The holotypic site is analogous to a holotypic
biological species but not with the implicit definition that it is the "type species"
against which other sites are to be compared.
5.2.2.5 Cluster Sites
The cluster site concept has a main site that is "typical" and several peripheral
or satellite sites in the area. The main site might be considered a holotypic site
while the peripheral sites provide an indication of variability in system response
within the area or subregion. The peripheral sites could be sampled on a less
frequent basis.
5-12
-------
5.2.3 Sampling Schemes
Three alternatives for sampling schemes include: sampling a fixed number of
sites during each sampling period; sampling a smaller number of sites and
supplementing sampling with randomly selected sites; and re-surveying
lakes/streams in the region. The advantages and disadvantages of each approach
are listed in Table 5.3.
5.2.3.1 Fixed Sites
A fixed number of sites could be selected to achieve a desired precision and
confidence level in each region with each site sampled during each sampling period.
Sampling a fixed number of sites provides continuity in the data and minimizes
logistical problems. This represents the typical approach to monitoring.
5.2.3.2 Fixed + Randomly Selected Sites
In this approach, a fixed number of sites (generally a smaller number than
above) are sampled during each sampling period to maintain continuity in the
monitoring program. Additional sites are randomly selected from the target
population for sampling on an annual or biannual basis. The sample selection can
be stratified to ensure good cross-sectional representation of system types in the
population. This approach provides an on-going continuous record on some sites
and, eventually, some information on all sites of interest in the region. Some of the
randomly selected sites might provide an earlier indication of regional change than
the fixed sites.
5.2.3.3 Re-Survev
A small number of sites expected to exhibit rapid change, because of acidic
deposition, might be monitored in this approach. Sampling other sites or a re-
survey of sites might occur only when a change is indicated in the rapid response
sites or at some fixed interval such as every five to ten years. This approach is
compatible with either of the two approaches discussed above. This approach can
indicate regional changes that might be occurring but at minimal cost on an annual
basis.
5-13
-------
5.2.4 Monitoring Protocol
There are at least three considerations for the monitoring protocol, and each
consideration includes a number of alternatives or options. These three
considerations are the sampling interval, number of stations/site, and monitoring
variables. These alternative approaches are compared in Table 5.5. The number of
samples to be collected at a station is part of the QA/QC program and is not
discussed here.
5.2.4.1 Sampling Interval
There are several alternative sampling concepts ranging from an index
concept (See Section 2.2.2.5 Index Sample) to continuous monitoring. An index
sample is collected on an infrequent but systematic interval (e.g., once per year in
lakes or 2 samples/year in streams) during a critical period (e.g., fall overturn in
lakes, spring elevated flow in streams) that serves as an index or indicator of the
essential characteristics of the system. The index concept was used in the NSWS.
Seasonal samples are collected once during each season in lakes or bimonthly in
streams to characterize seasonal patterns occurring in these systems. This approach
was followed in the ELS - Phase n. Continuous sampling might be monthly
sampling in lakes and monthly or biweekly sampling in streams to characterize the
seasonal dynamics in these systems.
5.2.4.2 Number of Stations
The number of stations per site or system depends on the program emphasis.
Evaluating changes in patterns among lakes or streams might be achieved by
sampling only one station per site. This station could indicate or index the general
lake or stream characteristics. This approach was used in the NSWS. Evaluating
changes in patterns both within the system and among systems requires multiple
stations per system.
5.2.4.3 Monitoring Variables
The monitoring variables can vary widely both in the number of different
variables and the type of variables. Some monitoring programs focus on in situ
variables that can be continuously monitored such as pH, conductivity and
temperature. The variables monitored in the NSWS are shown in Table 5.6. These
variables were selected because of the relation between atmospheric deposition and
5-14
-------
Table 5.6. Variables monitored in the NSWS and Phase II of the ELS.
In Situ
PH
Conductance
Lake Temperature
Secchi Disc Transparency
Field Laboratory
Laboratory pH, closed system
Dissolved Inorganic Carbon, closed system
True Color
Turbidity
Analytical Laboratoiy
pH, air-equilabrated
pH, open system
Acid Neutralizing Capacity (ANC)
Extractable Aluminum
Total Aluminum
Calcium
Chloride
Dissolved Inorganic Carbon, air equilabrated
Dissolved Inorganic Carbon, initial ANC
Dissolved Organic Carbon
Fluoride, total dissolved
Iron
Potassium
Analytical Laboratory (Cont.)
Magnesium
Manganese
Sodium
Ammonium
Nitrate
Phosphorus
Silica
Sulfate
Conductance
Additional Parameters Monitored
in Phase II
Acidity
Inorganic Monomelic Aluminum
Total Nitrogen
Chlorophyll a
5-15
-------
surface water chemistry, QA/QC checks and balances, and the relevance of certain
constituent concentrations to biological effects. Biologically relevant chemical
parameters such as calcium, pH, inorganic monomelic aluminum and dissolved
organic carbon, for example, have been identified through laboratory and field
bioassays as being important in biotic responses, particularly fish, to acidic
deposition. The biotic response represents the integrated effect of surface water
chemistry on biological organisms. Monitoring biological parameters or indices,
therefore, might provide earlier indications of biotic effects than monitoring the
change in chemical parameters. Monitoring watershed variables such as soil,
vegetation, or other factors also might contribute to achieving the TIME objectives
and aid in regional interpretation of results.
5.3 OTHER TOPICS
Data management and analysis and logistics are important considerations in
the design and implementation of a long-term monitoring program. As indicated in
Chapter 4.0, PREVIOUS AND ON-GOING ANALYSES, studies have or will be
initiated to address the topics listed in Table 5.1 under each of these categories.
The Draft Research Plan scheduled for May 1988 will discuss each of these topics in
detail.
5-16
-------
6.0 PROPOSED TIME FRAME
6.1 GENERAL CONSIDERATIONS
The design of any project or program must be predicated on the objectives.
The TIME objectives focus on the subregion and regional scales. The regions of
concern, however, span broad geographic boundaries from the East to the West,
Upper Midwest to the Southeast with very different deposition patterns, climate,
geology, soils, and aquatic resources. As discussed in the previous chapter, however,
there are a number of alternative approaches that can be incorporated in the TIME
design to accommodate these regional differences, provide the flexibility to achieve
the TIME objectives and adaptable to address future environmental concerns.
Several factors to consider in the selection and integration of appropriate
regional alternatives include:
o Regional characteristics (e.g., climate, topography, soils, geology, etc.);
o Proposed resources for monitoring (e.g., drainage lakes, seepage lakes,
streams, etc.);
o Anticipated change in the resource quality under current level of
deposition;
o Differences in the primary regional concerns; and
o Anticipated changes in future deposition rates.
The general design frame proposed for each region that incorporates these
factors is discussed below and summarized in Figure 6.1. This general design frame:
o Satisfies the TIME objectives;
o Provides a flexible, adaptive frame that can be readily modified to
incorporate other tiers, options, or other components such as watershed
soils, forests, etc.;
o Recognizes and incorporates the unique characteristics and concerns of
each region; and
o Provides a cost effective approach for achieving the TIME objectives.
6-1
-------
CT\
I
to
REGION/SUBREGIPN
OF INTEREST
PROPOSED
RESOURCE
FOR MONITORING
ANTICIPATED CHANGE
IN RESOURCE QUALITY-
CURRENT DEPOSITION
PRIMARY
REGIONAL CONCERNS
GENERAL
DESIGN FRAME
ANTICIPATED
FUTURE CHANGE
IN DEPOSITION
NE
Adlrondacks
Poconos/Catskills
S New England
C New England
Lakes
»
~
~
~
Expect decrease In S04
Expect Increase In N03
NOg "breakthrough* may be of concern
Need more deposition monitoring sites
Subreglonal emphasis important
- DDRP forecasts available
4
~
Maine
~
Mid-Atlantic
N Appalachians
Valley and Ridge
Chesapeake Area
Southeast
S Appalachians*
Piedmont
Blue Ridge
Streams
Streams
1
I
1
- Presently receMng some of the highest
deposition In the U S.
- Coastal and inland streams of concern
- DDRP forecasts available
- Some systems In S04 steady state
NO3 appears to be a developing concern
S04 concentration Increase expected even
IF deposition decreases
- Select 2 subreglons to study the year to year
variability In the Tj and relationship to Tg*
- SO4 breakthrough Is perhaps now
beginning
1 T! | 2 SUI>_ *
1 1 1 1 n«llons
*
\
~
-*
Ouachitas
~
*
Florida
Lakes
#(?)
- Projected Increase In emissions - down wind
are highest In U S.
- Panhandle lakes are most sensitive
- Prima rfly seepage lakes
W//////A
" ! T1
*
Upper Midwest
Lakes
-
- Seepage lakes dominate
- Declines In S04 expected
- Large lake chemistry changes not expected
- East-West S deposition gradient exists
t2
1 ! T1
-
West
S Rockies
Sierras
Cascades
Lakes
4
~
- Logistics prohibit probability and wide
spread seasonal sampling
- Changes In deposition possible
- II lakes receive Increased deposition they
wBI acidify quickly
- NOg Is a developing problem
t
v/ss/ssm/A t,
Figure 6.1. Proposed general design frame for the TIME project.
R«f*J Reipom* A SpecW tntfcil Symmj
(nut nacsiMily PiotMbfly Stn^b)
RtglonaJ PfobabUy Sam|M
u ^ Docr*aM/D«t«rtonit>on
(laiQa, *moD)
^ Incmw/linJiwiTOrt
Cbtq*. smefi)
-------
6.2 GENERAL TIME DESIGN
The general TIME design frame will be discussed for five regions: the
Northeast, Mid-Atlantic/Southeast, Florida, Upper Midwest and the West.
6.2.1 Northeast
6.2.1.1 Subregions
Subregions of interest in the Northeast include the Adirondacks,
Poconos/Catskills, Southern New England, Central New England, and Maine.
These subregions correspond with the subregional designations used in the ELS-
Phase I and II. Lakes represent the proposed resource for monitoring in this region.
6.2.1.2 Potential Deposition Effects
Lake watersheds in the Northeast are assumed to be near sulfate steady-state
(Rochelle and Church 1987). Under current levels of deposition, lake quality might
continue to deteriorate but at a relatively slow rate. The Northeast has received
considerable attention with respect to the current levels of deposition and proposed
reductions in deposition on aquatic systems. Future sulfate deposition rates are
anticipated to decrease because of possible emission control strategies.
6.2.1.3 Regional Concerns
Although sulfate deposition is anticipated to decrease, nitrate might become a
concern in surface water acidification. There is some indication nitrate saturation
might be occurring in some Northeastern watersheds, resulting in increased acid
loading to surface waters (Driscoll, personal communication). DDRP forecasts for
future changes in the Northeast will be available in late 1988 and will indicate
whether additional lakes might become acidic within the next 50 years and their
relative location. Lakes that are currently acidic also might recover if acidic
deposition decreases.
6.2.1.4 General Design
A regional probability sampling approach is proposed for the regional tier
(Tier 1) and seasonal tier (Tier 2) in the Northeast. The ELS-Phase II statistical
frame permits regional estimates and provides broad spatial coverage of lakes
across the deposition gradient that exists in the Northeast. The ELS-Phase II lakes
were sampled with consistent, comparable methods and known QA/QC in the fall
6-3
-------
of 1984 and the spring, summer, and fall of 1986. The ELS-Phase II lakes (or a
specific subset) are proposed as the Tier 1 lakes with a fall index sampling approach
on an annual basis.
A subset of these Tier 1 lakes are proposed as probability samples for Tier 2
or seasonal sampling. These Tier 2 lakes should be supplemented by non-randomly
selecting special interest and/or rapid response lakes. The Tier 2 lakes would be
sampled during winter, spring overturn, summer and fall overturn.
The selection of Tier 3 lakes, with possible supplemental sampling and
funding, will be coordinated with existing research sites (i.e., WMP site) and/or
intensively monitored systems.
Chemical variables measured in the Tier 1 and 2 samples would include those
variables measured in the ELS-Phase II Survey (Table 5.6). The use of various
biotic measurements or indices are currently being evaluated and will be
summarized in the Research Plan.
6.2.1.5 Rationale
A probability frame currently exists for the Northeast with regional estimates
for two years of fall index samples and a year of seasonal samples. The power of
this statistical frame in describing regional patterns in surface water chemistry was
exemplified through analysis of both the ELS-Phase I and II Survey data. This
frame can be used effectively for evaluating regional trends in lake quality.
Emission control strategies and target loading scenarios have focused on deposition
reductions in the Northeast. Evaluating the efficacy of these strategies, if
implemented, can be effectively accomplished through regional and subregional
estimates of changes in lake quality. The spatial coverage of the ELS-Phase II lakes
will permit precise subregional estimates based on the Tier 1 index samples.
Specific inclusion of special interest or rapid response systems can provide an early
indication of changes that might be occurring in the region and subregions. These
rapid response systems are most appropriately associated with Tier 2 to account for
seasonal differences.
6.2.2 Mid-Atlantic and Southeast
6.2.2.1 Subregions
Subregions of interest in the Mid-Atlantic region include the Northern
Appalachians, Valley and Ridge, and Chesapeake Area. Subregions of interest in
6-4
-------
the Southeast include the Southern Appalachians, Piedmont, Blue Ridge and
Ouachita Mountains. These subregions correspond with the subregional
designations used in the NSS-Phase I. Streams represent the proposed resource for
monitoring in these regions.
6.2.2.2 Potential Deposition Effects
The Mid-Atlantic region is currently receiving some of the highest deposition
in the U.S. Stream watersheds in this region appear to be in transition with some
watersheds at or near sulfate steady-state and others retaining sulfate (Rochelle and
Church 1987). Future sulfate deposition rates in this region are anticipated to
decrease.
The Southeast generally has lower precipitation sulfate concentrations but wet
deposition rates are nearly comparable to the Northeast because of higher
precipitation inputs. Watersheds in the Southeast are generally retaining sulfate but
there is concern about the time to sulfate steady-state (Rochelle and Church 1987).
Future deposition in the Southeast is anticipated to increase as the Southeast
becomes more industrialized.
6.2.2.3 Regional Concerns
Nitrate concentrations in Mid-Atlantic streams also appear to be increasing
and there is concern that nitrate saturation might be occurring now in some
watersheds (Corbeti and Lynch 1987, Helvey and Edwards 1987). Acidic stream
reaches currently exist in the Mid-Atlantic region. DDRP forecasts for future
changes in Mid-Atlantic stream chemistry will be available in 1989 and will indicate
if additional stream reaches might become acidic over the next 50 years. Currently
acidic stream reaches might recover if acidic deposition rates decrease.
Stream watersheds in the Southeast are currently retaining sulfate but there is
indication that sulfate is replacing bicarbonate as the dominant ion in high elevation
southeastern streams (Waide and Swank 1987). Preliminary DDRP forecasts
indicated a significant potential for southeastern streams to become acidic in the
next 50 years at current levels of deposition (Church et al. In preparation). DDRP
forecasts for the Southern Blue Ridge Province will be available in late 1988 and
will indicate the number of stream reaches that might become acidic within the next
50 years. If deposition increases, the potential number of acidic streams also would
be expected to increase.
6-5
-------
6.2.2.4 General Design
A regional probability sampling approach is proposed for two subregions: the
Chesapeake Area of the Mid-Atlantic and the Southern Appalachians in the
Southeast. These two subregions bound the geographic extent of these two regions.
Streams will be selected from the NSS frame for the regional tier, Tier 1, in these
two regions. A subset of streams from Tier 1 will be selected for Tier 2 seasonal
sampling and supplemented with special interest streams or streams expected to
exhibit a rapid response to acidic deposition. The probability samples selected from
the NSS frame will permit the use of design-based approaches for regional trend
estimation. An index approach similar to the 1986 NSS sampling protocol is
proposed for Tier 1 streams with bimonthly sampling proposed for Tier 2 streams.
The general design for the remaining five southeastern subregions is to sample
on a bimonthly basis only a select number of special interest streams or streams
anticipated to respond rapidly to deposition (i.e., Tier 2 streams). There would be
no streams selected for Tier 1 in these five subregions. If the Tier 2 streams or
these rapid response streams indicate regional or subregional changes are occurring,
a regional or sub-regional re-survey could be conducted within the NSS frame.
Each of the Tier 2 streams would be gaged. Flow measurements also will be
made at the time of sampling in each of the Tier 1 streams. Chemical variables in
Tier 1 and 2 samples would include those variables measured in the NSS-Phase I.
The use of various biotic measurements or indices are currently being evaluated and
will be summarized in the Research Plan.
The selection of Tier 3 streams, with possible supplemental sampling and
funding, will be coordinated with existing research sites and/or intensively
monitored systems.
6.2.2.5 Rationale
A probability frame is proposed for one subregion in the Mid-Atlantic and
one subregion in the Southeast to bound this geographic area and provide regional
estimates of changes that might be occurring in stream quality. Specific inclusion of
special interest or rapid response systems can provide an early indication of changes
that might be occurring in these regions. These rapid response systems are most
appropriately associated with Tier 2 to account for seasonal differences.
An early indication of changes that might be occurring in the region could be
used to trigger a re-survey. This re-survey would be based, in part, on an early
6-6
-------
indication of change and, in part, on the detection of regional trends based on the
probability samples of two of the seven subregions.
6.2.3 Florida
6.2.3.1 Potential Deposition Effects
Current levels of deposition in Florida are relatively high but the projected
increases in emissions down-wind are the highest in the U.S. Future deposition, as
with the Southeast region, is anticipated to increase.
6.2.3.2 Regional Concerns
Florida currently has the highest proportion of acidic lakes in the U.S. Many
of these systems are seepage lakes that are particularly susceptible to acidic
deposition. The Panhandle subregion and Northern Florida ridge area are two
areas that have a relatively large number of seepage lakes.
6.2.3.3 General Design
Lakes represent the proposed resource for monitoring in Florida. A select
number of special interest lakes and lakes anticipated to be rapid response systems
are proposed for seasonal sampling (i.e., Tier 2). There would be no lakes selected
for Tier 1. If the Tier 2 lakes indicate a regional change is occurring in the lake
systems, a regional re-survey could be conducted within the ELS-I frame. Re-
surveys also might be conducted at a fixed interval such as every 10 years.
Chemical variables in Tier 2 samples would include those measured in the
ELS-Phase II (Table 5.6). The use of various biotic measurements or indices are
currently being evaluated and will be summarized in the Research Plan.
The selection of Tier 3 lakes, with possible supplemental sampling and
funding, will be coordinated with existing research sites and/or intensively
monitored systems.
6.2.3.4 Rationale
Although Florida has the greatest proportion of acidic lakes, the susceptible
lakes are generally restricted to a relatively small geographic area in northern
Florida. For this area, a cost-effective monitoring approach is to conduct seasonal
sampling on a limited number of special interest lakes with a relatively long period
of record and/or lakes expected to respond rapidly. Many of the potentially
6-7
-------
susceptible lakes are seepage systems. When a response is identified in these select
lakes, a re-survey of ELS-I lakes can be used to evaluate regional trends that might
be occurring.
6.2.4 Upper Midwest
6.2.4.1 Potential Deposition Effects
Current levels of deposition in the Upper Midwest are slowly decreasing.
Emission reductions have been proposed or promulgated by the States of Minnesota
and Wisconsin, which should reduce emissions further. The highest wet sulfate
deposition rates in this region occur over the upper peninsula of Michigan with a
distinct East-West gradient in deposition. Future deposition rates in this region are
anticipated to decrease.
6.2.4.2 Regional Concerns
The Upper Midwest is dominated by seepage lakes that are potentially
susceptible to acidic deposition. Acidic lakes were measured in this region during
the ELS-Phase I and there is a concern additional lakes might become acidic in the
future. If deposition decreases, however, recovery of currently acidic lakes might
occur. Monitoring this recovery rate is an important element of a monitoring
program in this region.
6.2.4.3 General Pesign
Lakes represent the proposed resource for monitoring in the Upper Midwest.
A select number of special interest lakes and lakes anticipated to be rapid response
systems are proposed for seasonal sampling (i.e., Tier 2). There would be no lakes
selected for Tier 1. If the Tier 2 lakes indicate a regional change is occurring in the
lake systems, a regional re-survey could be conducted within the ELS-I frame. Re-
surveys also could be conducted at some fixed interval.
The selection of Tier 3 lakes, with possible supplemental sampling and
funding, will be coordinated with existing research sites and/or intensively
monitored systems.
Chemical variables in Tier 2 samples would include those measured in the
ELS - Phase II (Table 5.6). The use of various biotic measurements or indices are
currently being evaluated and will be summarized in the Research Plan.
6-8
-------
6.2.4.4 Rationale
The proportion of acidic lakes in the Upper Midwest is relatively small and
there is a trend for slowly decreasing deposition rates in the region. The majority of
the lakes are seepage systems that might be expected to respond relatively slowly to
deposition. A cost-effective monitoring approach for this region also is to conduct
seasonal sampling on a limited number of special interest lakes with historical
records and/or rapid response lakes. A re-survey of the region using the ELS-I
frame can be conducted when a response is detected in these early indicator lakes.
6.2.5 West
6.2.5.1 Subregions
Subregions of interest in the West are the Southern Rockies, the Sierras, and
the Cascades. These are three of five subregions surveyed in the 1985 Western
Lake Survey (WLS) - Phase I. Lakes represent the proposed resource for
monitoring in this region.
6.2.5.2 Potential Deposition Effects
Current sulfate deposition is low in the West and there is little anticipated
change in the resource quality at current deposition rates. Deposition in the West,
however, could increase dramatically with proposed mining and smelting activities
and increased industrialization. There is an increasing concern about deposition in
the Sierras and Southern Rockies.
6.2.5.3 Regional Concerns
Acidic episodes have been measured in lakes associated with summer storms
in the Sierras (Melack et al. 1987). Nitrate deposition also is a growing concern in
the Sierras. The high elevation western lakes have exceedingly low conductivities
and ANC concentrations and might be expected to become acidic quickly if
deposition increased. Many of the western lakes are located in wilderness areas and
are considered unique aquatic resources.
6.2.5.4 General Design
A select number of special interest lakes and lakes expected to be rapid
response systems are proposed for annual sampling in Tier 1. The index sampling
approach would be used in these Tier 1 lakes. If the Tier 1 lakes indicate a change
6-9
-------
in resource quality might be occurring, a resurvey could be conducted within the
WLS frame. Because of the potential susceptibility of these high elevation lakes, a
resurvey every 10 years might be warranted. There would be no lakes selected for
seasonal or Tier 2 sampling because of access problems during the late fall, winter
and spring. A selection procedure similar to that for other regions would be
proposed for Tier 3 lakes where seasonal sampling would occur.
Chemical variables in Tier 2 samples would include those measured in the
ELS-Phase II (Table 5.6). The use of various biotic measurements or indices are
currently being evaluated and will be summarized in the Research Plan.
6.2.5.5 Rationale
Western lakes are expected to become acidic quickly if deposition increases so
rapid response lakes or lakes of special interest that might provide an early warning
of increased acidification are particularly appropriate. Because many of the lakes
are remote and difficult, if not impossible, to sample, a probability sampling frame is
not feasible. In general, seasonal sampling also is not feasible. These lakes are
difficult to reach in the spring during snowmelt. Early fall snowstorms and
inclement weather make fall sampling treacherous. Summer sampling is logistically
feasible and can provide an index of lake quality for trend detection and was,
therefore, selected as appropriate for Western lakes.
6.3 PROPOSED TIME FRAME
The proposed TIME frame is provided to stimulate discussion. While there
are certain objectives that must be satisfied in the TIME project, there might be
alternative designs that also satisfy the TIME objectives but are more compatible
with on-going monitoring programs conducted by other federal or state agencies.
Review comments will be considered and evaluated from inclusion in the TIME
Research Plan.
6-10
-------
7.0 SITE SELECTION
7.1 OVERVIEW
Three types of sites will be selected for the TIME project:
o Probability samples;
o Rapid-response sites; and
o Special interest sites.
This chapter outlines the steps involved in the site selection process, and sets
forth preliminary criteria to be used in selecting each of the three classes of sites.
The steps are illustrated in Figure 7.1.
7.2 GENERALIZED SITE SELECTION PROCESS
7.2.1 Define Inclusion Criteria
This step identifies the desired target populations for each class, and may
apply to any level of the classification process (e.g., individual sites, subpopulations,
subregions, regions, national distribution). Inclusion criteria will vary for each of
the three classes of sites but might include:
o Regions in which changes in sulfate are likely to occur in the near future;
o Sites sampled in the NSWS for lakes and streams;
o ANC levels;
o Bedrock/surficial geology;
o Soil depth; and/or
o Special interest sites.
7.2.2 Classify/Stratify the Array of Potential Sites Into Categories
Bv Geographical Region
This process will vary for each of the three site types. The purpose of this
activity is to ensure good coverage of the range of issues that are being addressed. It
will also serve as an aid to hypothesis testing in each category of sites. Sample
categories are lakes/streams, ANC classes, ranges of sulfate deposition, groups of
sites defined by cluster analyses or other statistical procedures, etc.
7-1
-------
(A) Define Inclusion Criteria.
t
-~(B) Classify into Categories.
_Q)
n
'ot
tn
o
Q.
*-»
o
c
t
(C) Determine Total Number of Sites in Each Category.
t
(D) Overselect in Each Category.
t
(E) Apply Exclusion Criteria to Eliminate Non-Useful Sites.
r
i
i
If number of sites now < C,
If number of sites still > C,
eliminate sites:
I
Apply ancillary inclusion
criteria to prioritize;
Eliminate sites from bottom up.
Figure 7.1. Overview of the site selection process.
7-2
-------
7.2.3 Define the Desired Total Number of Sites in Each Category
The desired number of sites in each category will be estimated using standard
statistical sampling formula. Desired precision and confidence levels will be
specified for each category.
7.2.4 Select the Appropriate Number of Sites in Each Category
Initial site selection will be random for the probability samples, and non-
random (i.e., hand-picked) for rapid-response and special interest sites.
Overselection will be necessary for probability samples, because inclusion of some
sites might not be appropriate even though they belong to the target populations
(see Section 7.2.5 below). Over-selection also will probably occur for at least some
categories of special interest and rapid-response sites because more sites will be
identified than can be accommodated (as defined in Section 7.2.3).
7.2.5 Define and Apply Exclusion Criteria
There might be site characteristics that will compromise the usefulness of
individual sites even though they belong to the target population. The application
of exclusion criteria will eliminate some sites from further consideration. Each
individual site identified for sampling must be evaluated against the list of exclusion
criteria generated for that class of sites. There may be compelling reasons why
individual exclusion criteria are not considered for rapid-response or special-interest
sites, but if a site in either of these two classes violates several criteria, it should be
eliminated from further consideration.
7.2.6 Compare the Number of Sites Remaining in Each Unit to the Desired
Number of Sites in Each Unit and Adjust as Necessary
If there are too few sites, more sites should be selected, if possible (i.e. Section
7.2.4). If this is not possible, then Sections 7.2.2 and 7.2.3 need to be re-evaluated
and adjusted, if necessary, (i.e., Section 7.2.2) until a balanced selection is achieved.
If there are too many sites in the unit, some sites will be eliminated. The first
step in this process should be to prioritize remaining sites by applying ancillary
inclusion criteria such as proximity to other types of studies, inclusion in existing
monitoring programs, etc. These inclusion criteria will be specific to the class of site
being selected. Once the list of sites has been prioritized by these ancillary inclusion
7-3
-------
criteria, sites can be eliminated from the lowest priority upwards until the correct
number is achieved.
7.3 SITE SELECTION CRITERIA: PROBABILITY SAMPLES
7.3.1 Tier 1
7.3.1.1 Examples of Inclusion Criteria
7.3.1.1.1 Regional Distribution - Only regions in which: Changes in sulfate
deposition are likely to occur in the near future; and Annual probability sampling is
logistically feasible are included in the probability sample. Currently, this restricts
the probability sample to 2 of the 4 NSWS regions (Northeast and Mid-
Atlantic/Southeast).
7.3.1.1.2 Phase I - All sites in the probability sample will be drawn from the
sampled NSWS sites (i.e., all sites sampled in Phase I lakes and streams). One
consequence of this approach is that regions without a NSWS statistical frame
cannot be included in the probability sample. This means, for example, that streams
can only potentially be selected in two regions (Northeast and Mid-
Atlantic/Southeast). This, however, is consistent with Inclusion Criteria, Section
7.3.1.1.1, which only requires probability sampling in these two regions.
A second consequence of this approach is that hand-picked (rapid-response
and special-interest) sites will not be used in the probability sample. In regions
without a proposed TIME probability sample (the Upper Midwest and the West) it
might be possible to create regional approximations using a model-based approach.
7.3.1.1.3 ANC Levels - Because the goal of the TIME project is to track
acidification and recovery, ANC should neither be very high nor very low (i.e.,
approximately -10 to 100 ueq/1). In either situation, substantial buffering against
changes in pH can occur (carbonate/bicarbonate buffering at high ANC, buffering
by aluminum and organic complexes at very low ANC). Desirable ranges of ANC
will be specified. ANC is of interest as a surrogate for sensitivity and for stratifying
sampling.
7.3.1.2 Classification
Classification procedures will be conducted on a regional level. Statistical
techniques such as multivariate techniques (e.g., cluster analyses) will be used for
this classification. Because streams have actually been sampled only in the
Southeast and Mid-Atlantic, classification (Section 7.2.2) can only be achieved for
7-4
-------
these streams. A statistical frame exists for the Northeast streams but they have not
been sampled. Classification for the probability sample will include consideration
of deposition gradients in order to help ensure that, if deposition-related patterns of
acidification and recovery occur, they will be detected.
7.3.1.3 Define the Desired Number of Sites in Each Category
This activity will take advantage of ELS-Phase II analyses where available
(i.e., the Northeast) as well as results from the several analyses of the LTM data set
that have been made (Newell 1987; Payne et al. 1987; Loftis and Ward, 1987a,
1987b; Overton 1987)
7.3.1.4 Site Selection
Sites will be (over) selected using random selection techniques applied to the
categories defined in Section 7.2.4.
7.3.1.5 Exclusion Criteria
7.3.1.5.1 Present or Likely Future Disturbance - Because the purpose of the
TIME project is to monitor changes in surface water resources due to atmospheric
deposition, it is important that other confounding factors that could induce chemical
and/or biological change be avoided as much as possible.
7.3.1.5.2 Access Problems - If there are physical or administrative (e.g.,
inaccessibility, ownership, etc.) obstacles that constrain sampling on a regular
schedule, these should be weighed against the overall desirability of sampling that
particular site. Maximum allowable sample holding times will be defined as part of
this exclusion criteria.
7.3.1.5.3 Catchment Size - Maximum catchment size or catchment/surface
water area ratio will be defined for both lakes and streams, as appropriate.
7.3.1.6 Adjustment of Number of Sites
This process should follow that outlined in Section 7.2.6.
For the probability sample, ancillary inclusion criteria should include:
o Proximity to existing deposition stations, hydrometeorological stations,
or other monitored watersheds;
7-5
-------
o Inclusion in an existing program (e.g., LTM, state monitoring programs,
etc);
o Availability of other historical data (e.g., paleolimnological analyses);
o Availability of other types of data (e.g., hydrological regime, vegetation
and soils mapping, studies of soil and in-lake processes relevant to
surface water acidification and recovery); and
o The existence of continuous discharge stations, or automated event
sampling stations (streams or inlets to lakes).
7.3.2 Tier 2
A subset of Tier 1 probability sites will be chosen for Tier 2 probability
sampling in each region in order to examine seasonal properties of the index
concept. The site selection process for this activity will follow that for Tier 1 and
will be based on desired precision and accuracy and available resources.
7.4 SITE SELECTION CRITERIA: RAPID-RESPONSE SITES
7.4.1 Overview
Rapid-response sites are primarily identified for Tier 2, because a seasonal
sampling schedule might be able to shorten the time frame within which trends can
be detected. Because seasonal sampling will include samples in the index period
(fall for lakes, spring for streams), there will be Tier 1 data available for most rapid-
response sites. These will not be used as probability samples, however, unless these
sites are coincidentally identified during the Tier 1 probability sample site selection
process. Exceptions to this generalization occur in the West, where seasonal
sampling is not possible for many subregions.
7.4.2 Inclusion Criteria
Inclusion criteria include:
o Bedrock surficial geology;
o ANC range;
o Quantified response to both acid and base additions;
o First-order stream or headwater lake (if a drainage lake);
o (Lake) surface area;
o Catchment area;
7-6
-------
o Overlap with DDRP systems for rapid response;
o Soil depth; and/or
o Length and quality of historical record.
Rapid-response sites will be identified for all regions (Northeast, Southeast,
Upper Midwest, and West), and will form the backbone of the monitoring program
in regions in which probability samples will not be taken (the Upper Midwest and
West, as well as the Florida subregion of the Southeast).
7.4.3 Classification
Units of interest will include hydrological regimes, lithology, and current or
anticipated sulfate deposition range (Upper Midwest, Southern Rockies). For
example, although all rapid response sites will be located on highly sensitive
bedrock/surficial geology, it might be appropriate to select sites on different
substrates and/or with different hydrological characteristics within any one region
and deposition range. This activity is an important component of hypothesis testing.
7.4.4 Define the Desired Total Number of Sites in Each Unit
Good replication within units will be necessary to make strong tests of
alternate hypotheses.
7.4.5 Site Identification
Site selection for rapid-response sites will be done in close association with
local scientists who have been involved in similar activities in the past. Local
resources will include individuals in state agencies as well as LTM cooperators and
other appropriate personnel in the public and private sectors. In this phase,
candidate lists will be developed for each region, which will later be shortened. In
this phase, however, the number of good rapid-response sites that can be used to
test particular hypotheses will be maximized.
7.4.6 Exclusion Criteria
The exclusion criteria will be the same as those listed in Section 7.3.1.5. In
addition, no rapid-response sites will be identified in subregions in which current
deposition levels are low and changes in deposition have not occurred and are not
expected (e.g., the Northern and Central Rockies, the Ouachita Mountains).
7-7
-------
7.4.7 Adjustments of Number of Sites
This process should follow that outlined in Section 7.2.6. For the rapid-
response sites, ancillary inclusion criteria should be as in Section 7.3.1.6.
o Proximity to existing deposition stations, hydrometeorology stations, or
other monitored watershed;
o Inclusion in existing monitoring programs (e.g., LTM, state monitoring
programs, etc.;
o Availability of other historical data (e.g., paleolimnological analyses);
o Availability of other types of data (e.g., hydrological regime, vegetation
and soils mapping, studies of soils and in-lake processes relevant to
surface water acidification and recovery); and
o The existence of continuous discharge stations, or automated event
sampling stations (streams or inlets to lakes).
7.5 SITE SELECTION CRITERIA: SPECIAL-INTEREST SITES
7.5.1 Overview
Special-interest sites are primarily identified for Tiers 2 and 3. These include
sites for which there is significant information already available, or significant
ancillary information from other sources that will become available. Most special-
interest sites will also be the object of studies outside EPA's TIME project.
7.5.2 Inclusion Criteria
Inclusion criteria include:
o Inclusion in other existing programs (e.g., LTM, other agency programs,
state monitoring programs);
o Length and quality of historical records (including paleolimnological
analyses);
o Extent and quality of independently funded ancillary studies (especially
hydrology, vegetation and soils mapping, and studies of soil processes
and in-lake processes relevant to lake acidification and recovery). Sites
that already have continuous discharge monitoring, or that have
7-8
-------
automated event sampling (streams or inlets to lakes) are of particular
interest;
o Usefulness for hypothesis testing. These need not be rapid-response
sites, but could be added for testing hypothesis generated during the
rapid-response site selection process. An example of such sites might be
those with high nitrate loadings over a range of current or anticipated
sulfate deposition loadings;
o Availability of cost-sharing or in-kind services; and
o Special-interest sites may be identified in any region or subregion, even if
no other class of site (probability or rapid-response sites) occurs for this
subregion. For example, special-interest sites might occur in the
Northern or Central Rockies or the Ouachita Mountains or may be
streams in areas other than the Southeast or lakes or reservoirs in
subregions of the Southeast other than Florida.
7.5.3 Classification
Units of interest are ANC classes, bedrock/surficial geology, hydrological
regime, current or anticipated sulfate deposition range.
7.5.4 Define the Desired Number of Sites in Each Unit
Aside from considerations of hypothesis testing, this will depend on budget
constraints and the availability of cost-sharing or in-kind services for particular sites.
7.5.5 Site Identification
Site identification for special-interest sites will be done in close association
with other programs, agencies, LTM cooperators, and others who have been
involved in similar activities in the past. As with rapid-response sites, candidate lists
will be developed for each region, which will be shortened later.
7.5.6 Exclusion Criteria
As in Section 7.3.1.5. and 7.4.6.
7.5.7 Adjustment of Number of Sites
This process should follow that outlined in Section 7.2.6. For the special-
interest sites, ancillary criteria should include:
7-9
-------
o ANC class;
o Relationship to deposition gradient; and
o Proximity to existing deposition stations, hydrometeorological stations,
or other monitored catchments.
7-10
-------
8.0 TIMELINE/TIME PROJECT AND REPORT FORMATS
8.1 TIMELINE
In order to coordinate the TIME project and to ensure tasks are completed in
a timely fashion, Project Manager Workbench (PMW) was used to establish
timelines through a Gantt chart (Table 8.1) and to determine the critical path using
the Critical Path Network. The Gantt chart (Table 8.1) reflects the monthly
schedule.
8.2 TIME PROJECT REPORT FORMATS
Two tentative report formats are presented. One format is for annual reports
and the other biennial reports.
8.2.1 Annual Reports
The annual report is scheduled to appear approximately six months after
sampling is completed for the year. In the report, a standard set of analyses will
appear in an appropriate format. The focus will be on "highlights" of what was
observed during the previous year. Analytes of interest in major policy issues will be
analyzed with verified (peer reviewed) data from Tiers 1 and 2.
8,2,Tier 1
Each sample variable will be discussed by region and cumulative frequency
distributions (CFD) will be presented for about sue key variables. Comparisons will
be made between present CFD and previous CFD. Additional comparisons may be
made by pair wise and/or repeated measure tests. Shifts in overall distributions as
well as shifts in specific parts of the distributions will be investigated. Bivariate
comparisons (i.e., ANC vs. SO4) have been considered but are as yet undefined.
8.2.1.2 Tier 2
This section of the report will present the data gathered by region, by variable
and by lake. An example of the type of presentation envisioned is illustrated in
Figure 8.1. Subregional averages and confidence limits will be calculated and
presented in a modified form of Figure 8.1. Some narrative will be provided, and
8-1
-------
Table 8.1. Gantt Chart reflecting monthly schedule for TIME.
AUG 31 87
PROJECT: TIME PROJECT
PAGE 1
FILE: TIME2
TIME PROJECT
sre
CONCEPTUAL PLAN
CRITICAL PATH ANALYSIS
ASSESSMENT MATRICES LTM
EXAM TREND AMALSIS RPT
ENVIRON CONCERNS*EPA HO
REDEFINE REGIONS
PRIORITIZE REGIONS?
DRAFT CONCEPTUAL PLAN
LTM C00PERAT08 MEETING
MONITORING FRAME
STREAM SAMPLING 0ES1GN
7 FOR MONITORING NETWORKS
BIOLOGY
DEVELOP BIO. STRATEGY
BIO LIT. REVIEWS
BIOLOGY WORKSHOP
FINAL CONCEPTUAL PLAN
FINAL CONCEPTUAL PLAN
RESEARCH PLAN
DESIGN ELEMENTS
CRITERIA ID PARAMETERS
RAPID RESPON i REG CHAR
OA1A ANALYSIS WORKSHOP
REAL CHANGES S04.ANC.PH
KlUlM ft HAJCOH VAfl LSI
AUTOCORRELATION (FALL)
EVALUATE SITE COVERAGE
CLUSTER ANALYSIS
ALT. CLUSTER ANALYSIS
PLOT CSU CURVES BY XSD4
ASSESSMENT MATRICES OTHER
DDRP SEN ANALY. (RAPID)
REGIONALY REPRES CRITERIA
RAPID RESPONSE CRITERIA
CHANGES FROM OORP OUTPUTS
COOS RTP (HYDROMET/DEP)
COMPONENETS (FOREST/SOIL)
WO. NSVS DESIGN (T I,ID
STREAM ANALYSES
ASSESS REG 01F/SAMPL DES
INTEGRATION W/ AGENCIES
LOGISTICS
COOP VS CENTRAL SAMP SYS
ANALT LAB STRATEGY
MEET WITH STATES
OA PLANNING
PRECISION VS CONCENTRAT
VARIABLE VARIANCES
SYSTEM BIAS
EVALUATE OA FEEDBACK LOOP
OA WORKSHOP
13
FP
MM
65
JL
RW
1
XT
2
KT
DL
10
AK
4
DL
JF
35
FP
KT
9
OL
KT
3
PK
DL
4
DL
KT
3
JF
67
JF
X
12
B JF
25
FP
KT
3
KT
JF
4
KT
SO
11
JF
KT
1
KT
*
X
6
JB
AN
6
SO
AK
5
JB
5
JB
3
JL
RU
14
JL
RW
6
KT
JF
10
AN
JB
11
AN
JF
3
KT
3
KT
JF
6
KT
JF
7
OL
JF
0
PK
KT
6
AN
KM
10
JF
4
DL
JE
7
OL
JF
16
DL
JF
10
JP
10
JP
10
MO
3
JP
21
JP
MM
1987 |1900
Jiri Jul Aug It? Oct Nov Dec Jen Feb FUr Apr May Jun Jut Aug Sep Oct
g Sep Oct Nov :«c
III
Cmiliiiiiiii
'*
ill
lllll
II
Cumin*
.mil
¦¦¦¦a
I
!
Mill
I
I
I
.III
111
.III .
. ¦
ilium
in
.mil
¦i ¦
|
iiimi
in
8-2
-------
AUG 31 87
Table 8.1. Continued.
PAGE 2
PROJECT: TIME PROJECT
TIME PROJECT |Do| Who
DATA MANAGEMENT
I0EHTIFY DBM ADVISORS
3
JB
CRITERIA DATA INCLUSION
S
an m
DATA BASE SPECIAL SITES
k
AN
FIRST DRAFTS
D SITE SELECT. CRITERIA
8
JF KT
DRAFT 000 DOCUMENT
30
KT MM
000 INTERNAL REVIEW
10
DL JF
REVISE ORAFT DOOS
6
MM KT
SUBMIT DOOS TO HQ
5
DL X
S1RAMFT QT GEOLOGY?
6
AN PK
ANC RELAT IN SE STREAMS
3
PK
F. SITE SELECT. CRITERIA
21
AN KT
DEFINE LAKES OUTSIDE NSUS
5
DL JF
FINAL 000 DOCUMENT
20
MM KT
IDENTIFY CHEN PARAMETERS
7
JF JE
IDENTIFY BIO. PARAMETERS
7
JF B
ROUGH ORAFT RESEARCH PLAN
15
FP KT
COORDINATE W/OORP.WMP.ERP
r
JF DL
01 OA PLAN
20
JP KM
DEFINE OA OBJ.- CHEN.
10
MM JP
DEFINE OA 09J.-B10
8
MM e
NETKOOS MEET CHEN OA 08J.
6
x m
NETHflOS MEET BIO. OA 06J
6
B JF
CHANCES ANALT. METHODS
10
an m
PROC FOR LOSS OF COOP.
4
jp m
INTEREST ft AVAIL. OF COOP
15
JF OL
D1 SITE SELECTION PLAN
16
AN JF
PLAN AERIAL RECON. TR 1
5
DL JF
CIS HYDROMET/DEP/USHED
4
AN KM
OA OF DATA OUTSIDE TINE
5
MM
ID REPORT CONTENTS/FORMAT
6
DL JF
01 DATA ANALYSIS PLAN
20
JL RU
HIRE DBM CONSULTANT
6
JB OL
01 DATA MANAGEMENT PLAN
25
JB X
D1 RESEARCH PLAN
10
FP KT
SU0N1T RESEAR. PLAN 10 MO
6
DL X
ORAFTS FOR INTERNAL REVIEW
02 OA PLAN
15
JP MM
02 SITE SELECTION PLAN
15
AN JF
D2 DATA ANALYSIS PLAN
10
JL RU
D2 DATA MANAGEMENT PLAN
15
JB X
01 ANALY. NETKOOS MANUAL
K
MS JP
01 FIELD MANUAL
10
JR
02 RESEARCH PLAN
6
FP KT
COORDINATE W/AERP PROJ.
9
OL JF
CONT. STRM MONITORING?
6
AN PK
ORAFTS FOR PEER REVIEW
03 OA PLAN
10
JP MM
03 SITE SELECTION PLAN
10
AN JF
D3 RESEARCH PLAN
7
FP KT
FILE: TIME2
1987i1988 11989
Dtcl Jan Ftb Mar Apr May Jvi Jul Aug Sep Oct Nov D*cI Jan Feb Nar Apr Hay Jixi
¦¦¦¦ I---- .....j 1 | |
.IIIM
Ciiiiii
ata
.¦aaiaaiaaaa
.C**«a
aaaa
aaaa ...
aaaa
aaaa
aaaaaaaaa
. .
. C**»a
¦ a*
aaaaa
a
aaaa
.C«
mmihmhiii
Ci*i
C*a«
8-3
-------
Table8.1. -Continued.
AUG 31 87 PAGE 3
PROJECT: TIME PROJECT FILE: TIME2
1987
1968
1989
11 ME PROJECT jD^WhO
O.C
J,ft
Feb Mar Apr Nay
Jui Jul Aug Sep Oct fccv Dec
Jan Feb Mtr Apr M«y Jin
03 DATA ANALYSIS PLAN
7
JL
RW
03 DATA MANAGEMENT PLAN
e
JB
X
02 ANALT METHODS MANUAL
10
JP
MS
02 FIELD MANUAL
5
JR
PEER REVIEW
REVIEW ORAFTS
10
X
PEER REVIEW
5
X
PEER REVIEW REPORT
7
X
PREPARE FINAL DRAFTS
FINAL OA PLAN
15
JP
MM
FINAL SITE SELECTION PLAN
20
AH
JF
FINAL RESEARCH PLAN
e
FP
K1
FINAL OATA ANALYSIS PLAN
e
JL
RW
FINAL DATA MANAGEMENT
is
JB
X
FINAL AMALY. METHODS MAN.
10
JP
MS
FINAL FIELD MANUAL
10
JR
IMPLEMENTATION
ORAFT SAMPLING SCHEtJLE
15
AN
MM
SITE EVALUATIONS
40
X
FINAL SAMPLING SCHEDULE
15
AN
MM
FUND COOPERATES
20
DL
MM
TRAIN COOPERATORS
10
JR
SAMPLE COLLECTION
0
X
8-4
-------
Region 2
ANC (ueq/l)
LAKE
GRAPH
SIGNIFICANT TREND
2AI-
Statistical Statistical
Test 1 Test 2
P <
2BI-
P <
2CI -
P <
2DI -
P <
etc.
P <
"eg. Seasonal Kendall tau
Figure 8-1. A sample format for presenting data generated in Her 2.
8-5
-------
part of this narrative will focus on rapid-response sites and continued evaluation of
fall and spring index samples, if appropriate.
8.2.1.3 Tier 3
The preliminary plan for Tier 3 is to present a 1 to 2 page narrative of
activities, and findings of each cooperator or intensive site. This summary will be
based only on written material contractually required from the funded principal
investigator. This section of the annual report, which may represent diverse types of
information (soil, forests, etc.), will be customized rather than follow a specified
format.
8.2.1.4 Tier 4
This section of the annual report will concentrate on what was done that
previous year, what has been resolved and what is being proposed as special studies.
8.2.1.5 OA/OC Results
This section will address the analytical DQO's.
8.2.1.6 Extended Analysis of Previous OA/OC Results
The QA/QC results discussed in this section will be concerned with how the
extended analysis of previous QA/QC results impact previous annual report results.
This section is a trade-off between the desirability of getting the annual report out in
a timely manner and allowing adequate time to complete the analyses of the
QA/QC results. The annual reports will deal only with verified data but not
necessarily validated data.
8.2.2 Biennial Reports
This report, which will appear every two years, is not well defined at this time.
Major elements will be concerned with verified and validated data, interpretative
data analyses, responses to policy maker's feedback on annual reports, and the
effect of system error on observed and reported trends.
8-6
-------
9.0 REFERENCES
Altshuller, A.P. and R. A. Linthurst, eds. 1984. The acidic deposition phenomenon
and its effects: Critical assessment review papers. EPA 600/8-83-016BF, U.S.
Environmental Protection Agency, Washington, DC.
Baker, J. 1986. Effects of acidification on fish. State of the Science. Final report to
U.S. Environmental Protection Agency.
Baker, J. P. and H. Harvey. 1984. Critique of acid lakes and fish population studies
in the Adirondack region of New York State. Report to the Environmental
Research Laboratory, U. S. Environmental Protection Agency, Corvallis, OR.
Corbett, E. S. and J. A. Lynch. 1987. Long-term trend analysis of streamflow pH on
a forested watershed. (Supplement). In Aquatics Effects Task Group VI Peer
Review: Summaries Vol. II. The National Acid Precipitation Assessment Program,
Washington, DC.
Ford, D. E., K. W. Thornton, J. F. Nix, J. T. Malcolm, and F. E. Payne. 1986.
Acidic episodes and surface water chemistry: A comparison of Northeast and
Southeast study sites. Report submitted to EPA Environmental Research
Laboratory - Corvallis, OR. Submitted by: Ouachita Baptist University,
Arkadelphia, AR; and FTN Associates, Ltd, Little Rock, AR.
Haines, T. A. and J. P. Baker. 1986. Evidence of fish population responses to
acidification in the Eastern United States. Water Air and Soil Pollut. 31:605-629.
Helvey, J. D. and P. J. Edwards. 1987. Time trends of precipitation and streamflow
chemistry at the Fenow Experimental Forest. Pages 413-419 in Aquatics Effects
Task Group VI Peer Review: Summaries Vol. II. The National Acid Precipitation
Assessment Program, Washington, DC.
Hirsch, R. M., J. R. Slack, and R. M. Smith. 1982. Techniques of trend analysis for
monthly water quality data. Water Resources Research 18:107-121.
Hirsch, R. M. and J. R. Slack. 1984. A nonparametric trend test for seasonal data
with serial dependence. Water Resources Research. 20(6): 727-732.
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.
Teague, and E. P. Meier. 1987. Volume I. Characteristics of lakes in the western
United States. U.S. Environmental Protection Agency, EPA/600/3-86/054a.
Washington, D.C.
Likens, G.E. 1983. A priority for ecological research. Bull. Ecol. Soc. Am. 64:234-
243.
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 1. Population descriptions and physico-chemical relationships.
EPA-600/4-86/007a. Washington, D.C.: U.S. Environmental Protection Agency.
9-1
-------
Loftis, J. C. and R.G Ward. 1987a. Long-term monitoring project: Statistical tests
for trend analysis. (Progress Report - Corvallis Environmental Research
Laboratory).
Loftis, J. G, R. G Ward, R. D. Phillips and G H. Taylor. 1987b. Time project:
Statistical tests for trend analysis. (Progress Report - Corvallis Environmental
Research Laboratory).
Magoun, A. D. and J. Malcolm. 1987. Statistical estimation techniques for
detection limit samples (Draft).
Mericus, G, M. Miah, and R. D. Schonorod. 1986. Measurement uncertainty in the
NSWS. Lake and Reservoir Management (In press).
Messer, J. J., K. N. Eshleman, S. M. Stambaugh, and P. R. Kaufmann. 1986.
National Surface Water Survey: National Stream Survey, Phase I-Pilot Survey.
EPA-600/4-86-026. Environmental Research Laboratory, U.S. Environmental
Protection Agency. Corvallis, OR.
Murdoch, P.S. 1986. Chemical input-output budgets and steam chemistry dynamics
during a two year period in Biscuit Brook, Catskifi Mountains, NY. U.S. Geological
Survey Water-Resources Investigations Report 86 (Draft). Prepared in cooperation
with U.S. Environmental Protection Agency.
National Research Council. 1986. Acid deposition: Long-term trends National
Academy Press, Washington, DC. 506 pp.
Newell, A, D., C. F. Powers, and S. J. Christie. 1987. Analysis of data from long-
term monitoring of lakes. EPA 600/4-87/014. Environmental Research
Laboratory, Office of Research and Development, U.S. Environmental Protection
Agency, Corvallis, OR (Pre-print).
Nix, J. F., K. W. Thornton, D. E. Ford, and J. T. Malcolm. 1986. Storm event
sampling of southwestern Arkansas streams susceptible to acid precipitation EPA
Cooperative Agreement 811863-01-1 (Draft).
Olem, H. 1986. Episodic changes in stream water quality in five watersheds in the
Southern Blue Ridge Province (Draft). Interagency Agreement No. DW64930283-
01, TV-61968A.
Overton, S. 1987. National Lake Survey, Phase II analysis plan working draft.
Technical Report 115. Department of Statistics, Oregon State University, Corvallis.
Parsons, D. J., T. J. Stahlgren, D. M. Graber, and J. Melack. 1987. Effects of acid
deposition on selected ecosystems of Sequoia National Park. Pages 247-254 in
Aquatics Effects Task Group VI Peer Review: Summaries Vol. II. The National
Acid Precipitation Assessment Program, Washington, DC.
Payne, F. E., A. D. Magoun, and K. W. Thornton. 1987. Constituent variability:
Estimates of the number of lakes to be sampled, estimates of slope, and estimates of
time monitoring to detect changes (Draft).
9-2
-------
Rochelle, B. P. and M. R. Church. 1987. Regional patterns of sulfur retention in
watersheds of the Eastern U.S. Water, Air, and Soil Pollut. (In press).
Rosseland, B. O. 1985. Ecological effects of acidification on tertiary consumers:
Fish population response. In abstracts. International Symposium on Acidic
Precipitation, Muskoha, Ontario, pp 47-48.
Schindler, D. W., K. H. Mills, D. F. Mulley, D. L Findlay, J. A. Shearer, I. J. Davies,
M. A. Turner, G. A. Lindsey, and D. F. Cruikshunk. 1985. Long-term ecosystem
stress: The efforts of years of experimental acidification on a small lake. Science '
228:1395-1401.
Simons, T. J. and D. C. L. Lam. 1980. Some limitations of water quality models for
large lakes: A case study of Lake Ontario. Wat. Resour. Res. 16:105-116.
Smith, R. A., R. M. Hirsch, and J. R. Slack. 1982. A study of trends in total
phosphorus measurements at NASQAN stations. U.S. Geological Survey Water-
supply Paper 2190.
Snedecor, G. W. and W. G. Cochran. 1980. Statistical methods. Iowa State
University Press. 507 pp.
U.S. Government Printing Office. 1985. Hearing before the Subcommittee on
Natural Resources, Agricultural Research and Environmental of the Committee on
Science and Technology, U.S. House of Representatives - Ninety-Eighth Congress
(2nd Session), March 28,1984.
Waide, J. B. and W. T. Swank. 1987. Patterns and trends in precipitation and
stream chemistiy at the Coweeta Hydrologic Laboratory. Pages 421-430 in Aquatics
Effects Task Group VI Peer Review: Summaries Vol. fl. The National Acid
Precipitation Assessment Program, Washington, DC.
Witt, E. C. and J. L. Barker. 1986. Stream chemistiy response during episodes of
acidic rainfall and snowmelt runoff on Laurel Hill, Somerset County, Pennsylvania,
November 1983 to July 1985. U.S. Geological Survey Open File Report 86 (Draft).
Prepared in cooperation with the U.S. Environmental Protection Agency.
9-3
------- |