MA
Ditrittfsiontng Uncertainty in Estimates of
Regional Fish Population Damage Caused by
Acidification in Adirondack Ponded Waters
A Final Report Prepared For:
The Office of Policy Analysis
U.S. Environmental Protection Agency
Contract #68-01-4596
By:
Donald C. Peterson
Daniel M. Violette
Energy and Resource Consultants, Inc.
P.O. Drawer O
Boulder, CO 80306
(303) M9-5515
3une 10,1983
"Although the information described in this report has been funded wholly
or in part by the United States Environmental Protection Agency under
assistance agreement #68-01-6596 to Energy and Resource Consultants,
Inc., it has not been subjected to the Agency's required peer and adminis-
trative review and, therefore, does not necessarily reflect the views of the
Agency and no official endorsement should be inferred."
Our Reference: ACID2
-------
-------
Dimensioning Uncertainty in Estimates of
Regional Fish Population Damage Caused by
Acidification in Adirondack Ponded Waters
A Final Report Prepared For:
The Office of Policy Analysis
U.S. Environmental Protection Agency
Contract //6S-01-6596
By:
Donald C. Peterson
Daniel M. Violette
Energy and Resource Consultants, Inc.
P.O. Drawer O
Boulder, CO 80306
(303)^9-5515
October 14, 1985
"Although the information described in this report has been funded wholly
or in part by the United States Environmental Protection Agency under
assistance agreement #68-01-6596 to Energy and Resource Consultants,
Inc., it has not been subjected to the Agency's required peer and adminis-
trative review and, therefore, does not necessarily reflect the views of the
Agency and no official endorsement should be inferred."
Our Reference: ACID2
-------
TABLE OF CONTENTS
Page
1.0 Introduction 1-1
2.0 Approach Background 2-1
2.1 Dimensioning Scientific Uncertainty for Policy Assessment 2-1
2.2 Background to the Approach Used to Dimension Uncertainty 2-4
2.3 A Policy Assessment Approach 2-5
3.0 Development and Structure of the Damage
Function Estimation Procedure 3-1
3.1 Scope and Purpose of the Estimates 3-1
3.2 Development of the Process .. 3-6
3.2.1 Literature Review 3-7
3.2.2 The Initial Elicitation Structure 3-9
3.3 Evolution of the Elicitation Structure 3-22
3.* The Final Elicitation Structure 3-28
4.0 implementation and Results 4-1
4.1 Preparation for the Elicitation 4-1
4.1.1 Selection of Participating Scientists 4-1
4.1.2 Preparatory Orientation and Review Material 4-2
4.1.3 Compilation of Background Data for Use
in the Elicitation 4-2
4.2 Performance of the Eiicitation 4-4
4.2.1 Introductory Briefing 4-7
4.2.2 Variable Specification 4-9
4.2 J The Implemention of the Final Elicitation Structure 4-12
n
-------
TABLE OF CONTENTS
(Continued)
Page
4.2.4 Review and Assessment of the Elicited
Damage Functions 4-17
4.3 Results of the Elicitation 4-18
5.0 Interpretation and Use of the Elicitation Data
5.1 Steps Needed to Obtain Probabilistic Effects Estimates 5-1
5.1.1 Assumptions 5-2
5.1.2 Procedure for Estimating Damages from
Acid Deposition 5-7
5.2 Description of Adirondack Data Used in the Analysis 5-12
5.3 Calculation oi Incremental Damages Caused by Changes
in Fresh Water Chemistry 5-29
5.4 Estimates of Damages/Effects Due to Changes
in Acidification 5-33
5.4.1 Increase in Acidification - Predicted Effects Upon
Adirondack Fish Habitats 5-33
5.4.2 Decrease in Acidification - Predicted Effects Upon
Brook Trout Habitat 5-49
5.4.3 Alternative Damage Calculation 5-54
6.0 Conclusions 6-1
6.1 Project Overview 6-1
6.2 Conclusions Regarding the Specific Procedures Used
in this Project 6-4
6.3 Areas for Future Research 6-6
6.4 Concluding Remarks 6-9
Appendix A: Test Case Sensitivity Analysis of the Weight and Confidence
Interval Assumptions A-l
Bibliography B-l
111
-------
LIST OF TABLES
Table 3-1: Subjects By Which the Aquatic Literature
Page
Table 3-2
Table 3-3
Table 3-4
Table 4-1
Table 4-2
Table 4-3
Table 4-4
Table 4-5
Table 4-5a
Table 4-5b
Table 4-6a
Table 4-6b
Table 4-6c
Table 4-6d
Table 4-6e
Table 4-6f
Table 4-6g
Table 4-6h
Table 4-6i
Fish Species for Which Literature Contains Information
on Effects of Acid Stress and /or Metal Toxic ity
Maturation and Exposure Times; Allotted for Different
Mean Concentrations of Selected Ions
Approximate Inorganic Aluminum Concentration for
Adirondack Ponded Waters
Fish Species Covered in the Elicitation
Elicitation Table Soecies: Salve jinus fontinalis
(Brook Trout)
Elicitation Table Species: SalveHnus fontinalis
(Brook Trout)
Condensed Elicitation Table Species: Salvelinus
fontinalis (Brook Trout)
Brook Trout - Stocked
Brook Trout - Stocked
Brook Trout - Stocked
Brook Trout - Self-sustaining
Brook Trout - Self-sustaining
Brook Trout - Self-sustaining
Brook Trout - Self-sustaining
Brook Trout - Self-sustaining
Lake Trout - Stocked ,
3-10
3-11
3-20
4-3
4-5
4-6
4-11
4-13
4-15
4-16
4-19
4-20
4-21
4-22
4-23
4-24
4-25
4-26
4-27
IV
-------
LIST OF TABLES
(Continued)
Page
Table 4-6j Lake Trout - Stocked 4-28
Table 4-6k Lake Trout - Self-sustaining 4-29
Table 4-61 Lake Trout - Self-sustaining 4-30
Table 4-6m Lake Trout - Self-sustaining 4-31
Table4-6n Small Mouth Bass 4-32
Table 4-6o Small Mouth Bass 4-33
Table 4-6p White Sucker 4-34
Table 4-6q Common Shiner 4-35
Table 4-6r Fat Heat Minnow 4-36
Table 5-1 Steps in the Estimation of Damages Based on Elicitation
Results 5-8
Table 5-2 Number of Elicitations Performed for Different Fish
Species in the Project 5-9
Table 5-3 Variables in Project Data Base 5-13
Table 5-4 Distribution of Fish Habitat Represented in Data
Files Sorted by pH Range 5-20
Table 5-5 Results of Sort of Adirondack Lakes by Alkalinity,
Elevation, and Fish Species 5-23
Table 5-6 Alkalinity Distribution of Adirondack Lakes by Elevation
and Major Drainage Basin 5-27
Table 5-7 Estimated Acidification Status Distribution of Adirondack
Lakes by Elevation and Major Drainage Basin 5-28
Table 5-8 Chemical Shift in Adirondack Lakes Given a 100 eq/1
Acid Addition 5-32
Table 5-9 Chemical Shift in Adirondack Lakes Given a 50 eq/1
Acid Addition 5-34
Table 5-10 Chemical Shift in Adirondack Lakes Given a 100 eq/1
Acid Addition 5-35
-------
LIST OF TABLES
(Continued)
Page
Table 5-11 Cumulative Probability Distributions Showing the
Expected Value of Damage to Stocked Brook Trout at
Acidification Increases of IQOyeq/l and 50 ueq/1 5-37
Table 5-12 Cumulative Probability Distributions Showing the
Expected Value of Damage io Self-Sustaining Brook
Trout at Acidification Increases of 100ueq/l
and50ueq/l 5-38
Table 5-13 Cumulative Probability Distributions Showing the
Expected Value of Damage to Stocked Lake Trout at
Acidification Increases of lOOyeq/l and 50ueq/1 5-39
Table 5-14 Cumulative Probability Distributions Showing the
Expected Value of Damage to Self-Sustaining Lake
Trout at Acidification Increases of 100 ueq/1
and 50 ueq/1 5-40
Table 5-15 Cumulative Probability Distributions Showing the
Expected Value of Damage to Small Mouth Bass at
Acidification Increases of 100ueq/1 and 50 ueq/1 5-41
Table 5-16 Cumulative Probability Distributions Showing the
Expected Value of Damage to White Suckers at
Acidification Increases of lOOueq/I and 50ueq/1 5-42
Table 5-17 Cumulative Probability Distributions Showing the
Expected Value of Damage to Common Shiner at
Acidification Increases of lOOueq/1 and 50ueq/I 5-43
Table 5-18 Cumulative Probability Distributions Showing the
Expected Value of Damage to Fat Head Minnow at
Acidification Increases of 100 ueq/1 and 50 ueq/1 5-44
Table 5-19 Comparison of the Expected Values for 100 and 50 ueq/1
Acid Additions to Stocked and Self-Sustaining Brook
Trout Populations 5-46
Table 5-20 Expected Value of Change Occurring to Fish Excluding
Brook Trout 5-48
Table 5-21 Results of Secondary Sort Representing Lakes Which
Contained Both Fish Records and Chemistry Data 5-50
Table 5-22 Results of Reducing Present Acidification by 50 ueq/1 5-53
Table 5-23 Damages Occurring to Lakes Capable of Sustaining
Self-Sustaining Brook Trout Populations Between
the Original pH Values of 5.0 and 7.2 5.56
VI
-------
LIST OF FIGURES
Page
Figure 3-1 Overview of the Initial Elicitation Structure 3-14
Figure 3-2 Egg Mortality of Brook Trout Reported in the
Literature 3-15
Figure 3-3 Relationship Between Aluminum Concentrations and pH
in Adirondack Lakes More than 610 Meters Elevation
3une 24-27, 1975 3-17
Figure 3-* The Relationship Between pH and pCa for Adirondack
Lakes More Than 610 Meters Elevation 3-18
Figure 3-5 Intermediate Elicitation Structure 3-25
Figure 3-6 Inorganic Monomeric Aluminum as a Function of
Sample pH from DriscoII, 1980 3-27
Figure 5-1 Cumulative Probability Distribution for Percent Reduction
in Habitat for Fish Species X Resulting from a pH
of 6.4 Using the Data from Example 1 5-5
Figure 5-2 Cumulative Probability Distribution for Percent Reduction
in Habitat for Fish Species X Resulting from a pH
of 5.9 Using the Data from Example 2 5-6
Figure 5-3 Comparison of Meter pH and Colorimetric (Hellige) pH
Readings for 100 Adirondack Lake Samples 5-16
Figure 5-4 Representation of the Spatial and Temporal Variability
Within Dart Lake in the Adirondack Mountain Region 5-17
Figure 5-5 Sort Algorithm for Adirondack Ponded Waters 5-19
Figure 5-6 The Relationship Between pH and Alkalinity for Adirondack
Lakes less than 610 Meters Elevation 5-22
Figure 5-7 Calculated pH Shift Assuming Different Levels of Acid
Inputs 5-31
Figure 6-1 Possible Techniques for the Estimation of Probability
Distributions 6-7
vu
-------
-------
1J) INTRODUCTION
This report presents the procedures and results of an exploratory study dimensioning the
uncertainty around estimates of regional biotic effects caused by fresh water acidifi-
cation. This information is meant to augment scientific results by presenting an inter-
pretation of current scientific information concerning acid deposition to policy makers.
As such, it must include a measure of the confidence which can be placed in uncertain
specific estimates, thereby providing policy analysts with an additional critical perspec-
tive.
The premise underlying the method depicted in this report is that an acid deposition
policy assessment will be made regardless of the uncertainty or incompleteness of the
scientific data. Considerable uncertainty regarding the effects of acid deposition will
probably remain even after the research sponsored by National Acid Deposition Task
Force is presented in 1989. As a result, policy assessment will be made with uncertain
information on the effects of acid deposition, the current rates of acidification, and the
reduction of acid deposition that would result from a reduction in precursor emissions.
This means that any decision to implement an acid deposition control strategy will con-
tain an implicit evaluation of the probability that a given policy option is the correct
choice. Assessment planners have two options:
o either policy makers can be left on their own in evaluating scientific
uncertainties, or
o scientists and experts can directly assist decision makers by providing
information on these uncertainties.
If the latter option is preferred, the question becomes one of determining how to provide
information on the extent of uncertainty in the scientific data so that it is both useful to
the decision maker and scientifically accurate.
There is no question that the techniques used in this project will be viewed as controver-
sial by some parties. Still, it is important that the uncertainty in the estimated
1-1
-------
scientific relationships be dimensioned and this report presents one procedure for
accomplishing this. There are, however, other possible techniques. If reviewers and
researchers view this approach as inadequate, the authors encourage these individuals to
suggest alternatives, particularly since it is important that this dimensioning of uncer-
tainty be performed. This is a "first try", and it is important that alternatives be con-
sidered. Still, the conclusions drawn from this project indicate that this approach, or
variations of this approach, are able to provide needed information.
It is important that the information provided by this type of analysis be placed in the
proper perspective. As with all techniques, there are potential pitfalls in its applica-
tion. Two particular concerns need to be stated:
1) Policy makers may have an incorrect or inaccurate perception of how
"good" experts can be in making these subjective assessments, and thus
can be misled by the results; and,
2) That the elicitation of expert subjective judgements does not become a
substitute for performing needed effects research.
The information provided by this type of project is meant to augment the more conven-
tional presentation of scientific results. These techniques help to put scientific esti-
mates and conclusions in perspective. The dimensioning of uncertainty is based on
current scientific research and knowledge, and is simply another way of presenting this
information. It should not be viewed as a substitute for new scientific research.
In this project, considerable care was taken to avoid forcing scientists to make judgments
or to portray results at a level of precision beyond what they felt was reasonable. Expert
judgments required by this type of study will, by necessity, be somewhat vague. As a
result, it may not be appropriate to attempt to obtain precise estimates of the proba-
bility distribution for each uncertain outcome. This limitation is explicitly recognized in
this study. This project focuses on obtaining estimates of the range of potential effects
with only limited expectations regarding the ability of scientists to accurately provide
probabilities associated with different outcomes within that range.
The report is organized in the following form. Chapter 2.0 presents some background on
the importance of dimensioning the scientific uncertainty in acid deposition research,
1-2
-------
and the reasoning which led to the selection o- . Chapter 3.0 details the
development of the estimation procedure. Chapter ' jas the performance of the
estimation procedure and presents the results of the .,. . ;. Chapter 5.0 shows how the
collected data can be interpreted and used in an assesi.vu-nt of regional damages to fish
populations. Chapter 6.0 presents conclusions drawn from the application of the
approach.
1-3
-------
-------
2.0 APPROACH BACKGROUND
This chapter presents background information useful for understanding the interpretation
of scientific uncertainty used in this project. The first section presents several reasons
for dimensioning the uncertainty around scientific estimates for use in policy assess-
ment. The second section discusses the philosophy of the approach, and compares it with
techniques used in other studies. The third section discusses how this information could
lit into an overall policy assessment model.
2.1 DIMENSIONING SCIENTIFIC UNCERTAINTY FOR POLICY ASSESSMENT
The conventional approach for incorporating scientific uncertainty in policy assessment
has been to discuss the weaknesses of the data, but then to select best estimates and
proceed with the calculations as if they were certain values. Although scientists care-
fully explain the limitations of their data and, therefore, the limitations of their results,
policy analysts still have to use these results in policy assessment. The policy analysts
may simply be unaware of how these factors affect the use of an estimate in a policy
assessment framework. By having the scientists dimension the uncertainty due to these
limitations, policy analysts will be provided with interpreted results which should, in
conjunction with other scientific results and testimony, provide a more comprehensive
perspective of the policy alternatives.
When uncertainty is directly incorporated into a policy assessment, two approaches are
generally used: statistical derivation of confidence intervals and estimation of a conser-
vative lower bound of effects. If estimates of the effects of acid deposition are derived
statistically from historical data, then confidence intervals around the estimated para-
meters have to be used as a measure of the uncertainty. Confidence intervals, while
useful measures of uncertainty, are based on a number of underlying statistical assump-
tions that may or may not be true. These include assumptions regarding the actual
distribution of the residual errors, the functional form of the model, and the inclusion of
all important causal variables correlated with variables included in the model. Further,
many effects estimates are based on laboratory experiments, or from samples taken in
limited geographic areas that then must be applied to larger regions. For these reasons,
2-1
-------
uncertainty bounds based en statistically derived confidence intervals typically express a
lower bound of the uncertainty.
The second approach for handling uncertainty in benefit-cost studies has been to try to
establish a lower bound for the estimate of benefits (reductions in damages). This is done
by selecting the parameters used in the benefits calculations conservatively, i.e., when-
ever uncertainty is present the researcher tries to err on the low side. By doing this, the
researcher hopes to guarantee that the actual benefit estimates are greater than or equal
to the estimated levels. This procedure can be useful in establishing the dominance of
selected policies. If a conservatively estimated lower bound of the benefits still exceeds
the costs of a control policy, then one can be quite sure that the actual, unknown bene-
fits will exceed costs. There is, however, no actual dimensioning of uncertainty in this
approach.
If the range of uncertain outcomes of strategies to control acid deposition is to be
dimensioned, it may be necessary to base the estimates of the uncertainty on subjective
evaluations by experts. Even if simulation models or statistical models are used, an
alternative dimensioning of the uncertainty can provide useful and, in some cases, neces-
sary information. It may also be necessary to dimension the uncertainty surrounding the
application of mechanistic models calibrated for narrow regions in order to obtain
regional estimates of the effects. Granting a rationale for the performance of subjective
assessments, two questions then arise. First, can subjective evaluations of uncertainty
be obtained, and second, how accurate will these assessments be?
Addressing the second question first, the use of subjective probability distributions will
be an improvement over a decision making framework that uses only a single "best
estimate" of uncertain damages. In choosing a single best estimate of uncertain effects,
the researcher or policy maker must be selecting this estimate from some underlying,
intuitive probability distribution. If there is not an underlying distribution, there is no
basis for claiming that any particular estimate is better than any other. By not express-
ing this intuitive distribution, the researcher is not utilizing all of the information at his
disposal. In fact, he may be obscuring information important to the correct interpreta-
tion of his results. As a researcher conducts experiments or performs studies, numerous
choices are made in the design and scope of the project. Often, these choices are based
on subjective probability assessments concerning the most likely structure of a causal
relationship. Scientific research, as well as environmental decision making, requires that
2-2
-------
numerous subjective probability assessment :& u\s made during the normal course of
events. Making these underlying, intuitive assessments explicit will only increase the
amount and quality of information available to decision makers.
Another advantage in the use of explicit procedures for dimensioning uncertainty is the
additional qualitative understanding that can be gained through the process of quantify-
^j.
ing important decision parameters. Examples of the use of subjective judgements in a
structured model have been evaluated in controlled laboratory studies. Many of these
studies have shown substantial benefits from the use of judgement based models when
compared with more common and untutored intuitive decision processes.** By having to
express the extent of the uncertainty in quantitative terms, researchers and policy-
makers are forced to give more thought to the information requirements important to
the decision problem. Explicit procedures for dimensioning uncertainty will also assure
consistency across different researchers. If procedures for dimensioning uncertainty are
not defined, many different and possibly non-comparable methods will likely be used by
the different researchers.
Many important policy and assessment questions can be addressed by dimensioning the
uncertainty around effects estimates:
1. It will help decision makers evaluate alternative policies with uncertain
implications.
2. By using explicit procedures for dimensioning uncertainty, information
critical to policy assessment can be subject to formaLpeer review with
respect to both its content and estimation procedures.
3. It will assist in the determination of research priorities by showing the
benefits of reducing uncertainty in different effects areas.
Keeney and Raiffa (1976, p. 364) discuss the advantages in qualitative understanding
that were gained through the construction of joint probability functions for potential
outcomes in an air pollution control decision problem.
**See S. H. Mclntyre (1982) for a discussion of the use of the models in making marketing
decisions.
2-3
-------
It will allow, in conjunction with economic analysis, important policy
questions to be addressed, particularly:
o the appropriate timing of control strategies
o the benefits and costs of delaying the implementation of control
strategies in order to collect additional information.
2.2 BACKGROUND TO THE APPROACH USED TO DIMENSION UNCERTAINTY
The approach for dimensioning uncertainty most often used in risk assessment has been
to have experts provide point estimates of the probability of occurrence of different
events. The precision implied in these point probability estimates simply may not be
realistic. In spite of the sophisticated probability encoding techniques that are available
(see Lichtenstein et al., in press, and Spetzler and Stael von Holstein, 1975), scientists
whose opinions regarding the likelihood of different events, which may be only vaguely
formulated, can feel manipulated by a precise probability encoding process. In response
to this concern, Suppes (197*) and most recently Feagans and Biller (1981) have used
estimates of probability intervals rather than less realistic point estimates of probabili-
ties for different effects estimates.
In developing a method for assessing health risks in the setting of National Ambient Air
Quality Standards (NAAQS), Feagans and Biller obtained upper and lower subjective
probabilities for each event under consideration. The vagueness or imprecision of the
experts' judgement in their approach is reflected in the difference between the upper and
lower probabilities. The method selected for use in this study is similar to the approach
used fay Feagans and Biller in that the imprecision in the experts' ability to express or
dimension uncertainty is directly addressed. The focus is on estimating the range of
potential effects (i.e., upper and lower bounds to effects) with only a simple weighting
scheme to express the likelihood of different outcomes within that range. The Feagans
and Biller approach is designed to provide considerably more detailed information. They
consider each outcome within the possible range of effects and have the expert estimate
upper and lower probabilities for each outcome. Although the elicitation depicted in
Chapter 4 of this report has less ambitious information goals, scientific experts were
able to provide the requested estimates more reliably and, therefore, view them as more
credible. The lack of information on causal mechanisms associated with acidification
2-4
-------
effects as
conservati
ferent effcx:
.rge variability in ecosystem response to acid deposition warrants
...'ons regarding the precision with which experts can associate dii-
.ras with probabilities.
The objective.' of this process is to obtain information concerning the uncertainty of
effects estimates, and not to implement a procedure that requires a level of precision
that can not be met. Wallsten et al. (1983) cite publications which show that U.S. Army
intelligence analysts do not use explicit quantitative probability estimates in their
analyses because they believe numeric estimates fail to convey the vagueness inherent in
their opinions. The method taken in this study falls in between the precise estimation of
a probability distribution and the refusal to incorporate any numeric probability esti-
mates.
23 A POLICY ASSESSMENT APPROACH
The framework used to dimension uncertainty in this project was developed from the
data requirements of several economic studies as well as other general policy assessment
approaches (see Violette and Peterson, 1982). There are a number of common character-
istics to the assessment approaches that were considered. One such characteristic is
that rather than focusing on optimizing, i.e., picking the best approach from among all
possible approaches, the policy approaches considered utilize a satisfying criteria.
Instead of selecting the best policy from all potential policy options, the approach
focuses on selecting the best policy from a limited set of options. For example, a pre-
selected set of policy options could encompass:
Policy 1 - a no change strategy, i.e., no new controls.
Policy 2 - a 20 percent decrease in precursor emissions.
Policy 3 - a W percent decrease in precursor emissions.
These three policy options are quite distinct. The information needed to pick the best
policy from among these three is very different than the information required to deter-
As was pointed out by Wallsten et a!. (1983), the use of non-numeric phrases to describe
uncertainty may cause problems because the phrases are understood differently by
different people.
2-5
-------
mine whether a 20 percent or 25 percent reduction in emissions is the better choice.
Choosing between policies this similar will be a difficult task, given current information
concerning the effects of acid deposition. Only limited resolution between different
policy options may be achievable due to uncertainty in effects estimates.
If the costs of the three policy options presented above are $0, $1 and $3 billion annually,
then:
o Policy 1 will be the best choice of the three if benefits from adopting
Strategy 2 are less than $1 billion and the benefits of Policy 3 are less
than $3 billion, i.e., the costs of each of the control strategies are
larger than the benefits.
o Policy 2 will be the best choice if the net benefits of a 20 percent
reduction in emission (benefits - costs) are positive and greater than the
net benefits of Policy 3.
o Policy 3 will be the best choice if the net benefits of a 40 percent
reduction in emissions are positive and greater than the net benefits of
Policy 2.
Violette and Peterson (1982) show how these selection criteria can be framed as the
probability that the benefits of reductions in emission fall into given intervals. Although
Vioiette and Peterson (1982) formulated the intervals in terms of average damages in
dollars per unit of acid deposition to facilitate comparisons across different emissions
reductions policies, the basic concept involves estimating the probability that the bene-
fits of a 20 percent reduction in emissions exceed its costs (i.e., the probability that
benefits are greater than or equal to $1 billion) and the probability that the benefits of a
40 percent reduction in emissions exceeds $3 billion. Estimating the probability that
benefits of a given control strategy are greater than or equal to a given cost of control is
a much simpler task than estimating the marginal damage curve (as economists conven-
tionally prescribe), or even estimating the probability that benefits equal a given number,
i.e., obtaining a point estimate. The assessment problem may also be construed as being
analagous to purchasing insurance against unwanted and uncertain future damages. In
this case, different acid deposition control strategies can be viewed as purchasing
insurance in order to prevent the occurrence of environmental damages. The decision
-------
maker would be asked to determine the extent of the premium which he would be wi',
to purchase in order to avoid different acid deposition-caused effects and their
associated probabilities.
The framework presented -above greatly reduces the information requirements of the
assessment. The dimensioning of the uncertainty may be somewhat vague and imprecise,
as the level of resolution is simply that necessary to select between distinctly different
policy options. It recognizes that imprecise or uncertain effects estimates will not be
sufficient to differentiate between slightly differing policy alternatives. One result of
this type of assessment is to eliminate clearly inappropriate policy options while specify-
ing a set of "satisfactory11 alternatives.
Even if the types of uncertainty bounds placed on effects estimates are not incorporated
within a policy assessment procedure of this type, they will still be valuable in accurately
portraying current scientific information to policy makers. For example, the procedure
employed in this project integrates many scientific complexities in a format which
presents regional damages to fish populations resulting from fresh water acidification. A
principal benefit of this approach is the clear display of scientific information not
previously available to the decision maker. The results of this project are meant to pro-
vide broadly defined and integrated perspectives to non-scientists dependent upon
scientific information for the performance of their responsibilities in policy assessment.
2-7
-------
-------
3.0 DEVELOPMENT AND STRUCTURE OF THE DAMAGE FUNCTION
ESTIMATION PROCEDURE
This chapter outlines and discusies the design of a process for presenting scientific infor-
mation on fresh water acidification in a manner useful for policy assessment. To be
useful in policy evaluation, the scientific estimates should be regional in scope and the
uncertainty in the estimates should be dimensioned. The elicitation process developed in
this project meets these two criteria. The development of the eiicitation is an important
part of this approach. If the elicitation were not carefully designed, it could mis-
represent current scientific information. In order for scientific and policy researchers to
have confidence in the results of the elicitation, this chapter details the logic used in the
design of the elicitation. The discussion is divided into four segments reflecting the
temporal and logical development of that process. Section 3.1 specifies the purpose of
the estimates and the scope of their application. The remaining sections outline the
work performed in order to accomplish these objectives. Section 3.2 reviews the
development of the initial elicitation framework. This initial elicitation structure served
as the starting point for developing a structure that would be credible to the scientific
community and still meet the needs of policy assessment. Section 3.3 shows how this
initial structure evolved as a result of presentations and pre-tests of the procedure. Sec-
tion 3.* presents the final form of the elicitation structure.
3.1 SCOPE AND PURPOSE OF THE ESTIMATES
As noted in Chapter 1, the purpose of this project was to assess a particular approach for
evaluating regional damages to aquatic biota caused by acid deposition.* Chapter 2
presented the background and justification for this approach in policy assessment. This
section addresses the use of this approach for quantifying the uncertainty in estimates of
regional damages to aquatic ecosystems. Broadly speaking, the purpose of the process
was to determine the response of selected aquatic populations, i.e., fish and more
* Because functional relationships between atmospheric acid loading and lake acidifica-
tion have not yet been adequately quantified in North American waters, the goal of the
project was to obtain predictions of the biotic consequences of regional changes in levels
of fresh water acidification.
3-1
-------
specifically, gamefish to increased hydrogen-ion concentrations in fresh water eco-
systems. The project could have, however, focused upon other aquatic populations (such
as macroinvertebrates), or on chemical and bijologic processes (such as oligotrophication),
or lastly, upon changes in the structure of aquatic communities.*
The number of fish species that could be analyzed was limited. Game fish species were
the focus of this application due to their more obvious social and economic value to
society.** Although some non-game species are known to be quite sensitive to the
effects of acidification, it is difficult to estimate economic damages resulting from their
decline or extinction. It is important, however, to realize that assessments of the
impacts of acid deposition should not be limited to only those effects that can be easily
quantified in economic terms.***
The aim of this project was to estimate the relationship between changes in fresh water
pH and loss of fish populations on a regional basis. The independent variable is the cur-
rent pH of ponded waters, not the precipitation pH. The broader question of how acid
deposition affects fish requires the development of additional information showing the
linkage between deposition and changes in lake water chemistry. The latter relationship
could be provided if two additional pieces of information were available: (1) the ionic
constituents of dry and wet deposition; and (2) the fate of those atmospheric ions
described in terms of their interaction with the edaphic environment and subsequent
* While the biologic effects of acidification caused by acid deposition have been widely
observed and reported at virtually all trophic levels in Scandinavia, Canada, and the
United States, the effects upon fish have been the most carefully investigated and
documented in the literature. In addition, the decline of fish populations in both
Scandinavia and North America has been a focal point of public and political concern.
** Damage estimates were, however, performed on three non-game species during the
project. In two cases, because the scientist believed he knew substantially more about a
particular non-game species than candidate game species. In the other, a non-game
species was considered because of its status as an "indicator" of a system's response to
acidification. For additional detail, see Sections 3.2 and
*** Additional work is being undertaken to investigate the best way to incorporate
ecosystems effects that cannot be easily quantified in economic terms. A description of
the effects should allow decision makers to incorporate the non-economic effects within
the assessment. The authors of this report do not advocate an assessment based only on
effects that can be economically quantified.
3-2
-------
transport to the aquatic ecosystem.* Such an analysis would have the general structure
outlined below.
Step 1 - Estimate the probability that changes in emissions of precursors will result in
changes in acidic deposition.
Step 2 - Estimate the probability that changes in deposition will cause changes in
aquatic chemistry.
Step 3 - Estimate the probability that changes in given water quality parameters will
cause effects in given fish populations.
While the performance of the analysis outlined above would provide policy-makers with
considerably more information than provided by the performance of Step 3 alone, the
present analysis was considered appropriate given the scope and objective of this project.
There are a number of regions in the United States which contain fish populations inhab-
iting dilute lakes and streams exposed to acid deposition inputs (Haines, 1981; Pfeiffer
and Festa, 1980; Schofield, 1976, 1982; Baker, 1982; Arnold, 1980). These regions possess
generally low acid neutralizing capacity and have drainage systems situated in areas of
crystalline or metamorphic bedrock with shallow soils of low buffering capacity
(Omernik, 1982).** While numerous areas in the United States possess poorly buffered
surface waters, inventories of water quality and fish population status are incomplete for
all of these areas. The most extensive available inventory data and the only published
literature relating fish population decline to pH within the United States covers the
Adirondack Region of New York State. Consequently, this region was selected to
evaluate the use of this probabilistic approach as a tool for evaluating the regional
effects or impacts of acid deposition. The Adirondack Region is also useful in that a
* Atmospheric precipitation is water in equilibrium with atmospheric gases which con-
tains dissolved and suspended matter taken up by the water droplets during passage
through the atmosphere. As the water contacts the soils and bedrock of a region, it
interacts with the mineral constituents physically and chemically, leaching out the more
soluble fractions. This water flows into rivers and lakes, creating changes in the
concentrations of the ions in the surface water. These changes cause biotic responses
being examined in this report.
** Acid neutralizing capacity (ANC) is the equivalent sum of all bases which can be
titrated with a strong acid. The base neutralizing capacity (BNC), is the sum of all acids
which can be titrated with a strong base. Total ANC and BNC can be considered as the
composite of individual acid/base systems (Driscoll and Bisogni, 1982; Henriksen, 1981).
3-3
-------
number of previous damage assessments have been prepared ..'leufer and
Festa, 1980; Menz and Mullen, 1982). Comparison of the cui . .:•-i.vment with the
conclusions and procedures of its predecessors should allow for =1 nure complete review
of the current technique.
A large data set covering the Adirondack Region is maintained by the New York State
Bureau of Fisheries. The fish listing developed by the state contains information on
approximately 3,500 waters with over 35,000 records. However, the Department of
i
Environmental Conservation (DEC) has current I data for only 937 of the approximately
3,500 surface waters contained within the file. Information contained within the file is
derived from three main sources: a survey of water quality and fish populations in
Adirondack surface waters conducted from the '1920s to the present; a survey of 214 of
the 219 Adirondack lakes above 610 meters conducted in 1975 (Schofield, 1976); and the
results of a sampling program established in 1975 by the DEC to determine the "scope of
water quality impacts associated with acid ion deposition" (Pfeiffer and Festa, 1980).
Since 1975, the DEC has sampled 937 Adirondack ponds and lakes, or 34 percent of the
total number of lakes and ponds within the Adirondack Ecological Zone.* More compre-
hensive water chemistry investigations have been conducted upon single Adirondack
lakes, or in some cases, small watersheds. The Integrated Lake-Watershed Acidification
Study (1LWAS) studied lake acidification processes for three lake watersheds in the
Adirondack Park Region of New York (Chen, 1980). In addition, detailed chemical
analyses have been conducted by Dr. Charles Driscoll and co-workers for a small number
of lakes (Driscoll, 1980; 1982a; 1982b). The DEC has closely monitored six lakes in
Hamilton and Herkimer counties since 1978 (P|feiffer and Festa, 1980). While stream
water chemistry has not received the same emphasis as ponded water chemistry, recent
studies conducted by the New York DEC (Colquhoun, 1981, 1982) have indicated that
Adirondack streams are affected by pH and alkalinity depressions.
Data on fish populations within the Adirondacks are generally limited to presence-
absence data contained within the DEC fish listing. Standing crop data was, however,
collected for three brook trout ponds using interview techniques by DEC staff in 1973.
The DEC interviewed fishermen to estimate angling pressure per acre for each lake and
the number of fish caught. From this, the DEC estimated the standing crop of these
* Thirty-five additional Adirondack lakes and ponds were surveyed and test netted by the
DEC in 1982.
3-4
-------
three brook trout ponds, expressed as yield per acre in 1972 as: Black, 1.18 pounds/acre;
Shaver, LSI pounds/acre; and Whey 0.28 pounds/acre. However, considering that there
are some 2,759 lakes in the Adirondack Ecological Zone (3506 in the broader DEC fish
listing), such limited standing crop information is scarcely useful for the prediction of
fish populations in the region, or for the estimation of the effects of acidification upon
those fish communities.*
Areas other than the Adirondacks have also been surveyed for data on water chemistry
and fish populations, yet the majority of the data remains unpublished, unreleased, or
incompletely analyzed by the investigators. When these data become available, a
regional assessment similar to the one performed in this report could be conducted.
Investigators at the Brookhaven National Laboratory are currently conducting an assess-
ment of acidic precipitation effects on surface and groundwaters for the entire United
States. The major objectives of the research are: (1) quantification of the sensitivity of
freshwaters within the United States to acidification; (2) determination of the extent to
which they have been impacted; and (3) investigation of the impact of acidification on
groundwaters using current and historical data. The researchers are collating existing
data on water quality from STORET and 157 other sources. Data are entered into the
Acidification Chemistry Information Database (ACID). By September 1982, the ACID
database contained approximately 1.7x10° data values obtained on 5.7x10^ dates from
ft. 15x10* water quality stations in the eastern United States. Data are restricted to
stations which report that 60 percent of their alkalinity measurements are equal to or
less than 500 yeq/1, or those with pH measurements of 7.5 or less (Hendrey, 1983). In
addition to BNL's work, surveys have recently been conducted by the Fish and Wildlife
Service in New England (Haines, 1982, unpublished) and in the Mid-Atlantic states
(Arnold, 1980), The U.S. EPA Environmental Research Lab in Duluth has characterized
water chemistry in northern Minnesota, Wisconsin, and Michigan and is using what cur-
rent information exists on fish populations and a number of other parameters to evaluate
the characteristics and resource value of each of a number of different watershed lake
types.
* In 1980, Pfeifer estimated that there were approximately 2,877 individual lakes and
ponds within the Adirondack Ecological Zone. This number was revised in the 1981
Acidity Status Summary.
3-5
-------
Major scientific uncertainties concerning ths effects of hydrogen ion concentration upon
fish populations might usefully be divided in|o two domains. First, as noted above, there
is currently very little comprehensive inventory data covering water quality and fish
population status. Second, there is substantial uncertainty regarding fish population
responses to lake and stream water acidification.*
3.2 DEVELOPMENT OF THE PROCESS
The process can be divided into four segments: (1) the development of the initial elicita-
tion structure; (2) the pretest and evaluation of the initial structure, followed by revi-
sions to the elicitation; (3) the implementation of the final elicitation structure; and (<0
the analysis of the results. This section reviews the development of the elicitation
framework. The initial step in developing the elicitation was to review the available
scientific research literature covering the effects of pH depression upon aquatic
organisms; this is discussed in Section 3.2.1. The initial elicitation structure, discussed in
Section 3.2.2, was quite detailed in order to reflect the complexity of the direct and
indirect effects of pH shifts upon fish populations. Discussions with cooperating scien-
tists allowed important simplifications in the elicitation format to be made. Section 3.3
summarizes this procedural development, while emphasizing key issues relevant to the
assessment of fish population change as a function of change in pH. The final elicitation
structure is discussed in Section 3.*.
In order to develop and implement the analysis, the investigators used a highly interac-
tive approach. Section 3.2, and to a lessef extent, Section 3.3 attempt to convey the
evolution and development of the analysis ovjer time.
* Damage estimates relating pH values to fjish population declines are extremely uncer-
tain. Lake morphometry, primary productiqn, plus calcium and dissolved organic carbon
concentrations, to list but a few factors, appear to all have profound influences upon the
toxicity of a given lake or pond pH value to a fish population, and as these and other
parameters vary widely, plotting regional fish population response to pH values is very
problematic. In addition, recent work by Qriscoll and co-workers has shown significant
temporal and spatial variability of pH values, and other water chemistry parameters with-
in a small Adirondack lake (unpublished data). Given this degree of chemical variability
and, for example, apparent brook trout avoidance of low pH waters in favor of waters
with a higher pH, lakes with measured critical or lethal pH values could possibly sustain
fish populations.
3-6
-------
3.2.1 Literature Review
The first step in developing the elicitation was the compilation of an extensive file
covering the effects of low pH on aquatic organisms at different trophic levels. This
review of the broader aquatic effects literature had several purposes: (1) achieving an
appreciation of the primary scientific hypotheses concerning the effects of depressed pH
upon organisms in dilute acidified waters; (2) listing the critical scientific uncertainties
both from the perspective of hypothesis confirmation and policy assessment; and (3),
developing a working vocabulary, to enable the authors to participate in discussions with
members of the scientific community. The accomplishment of (3) enabled the interdis-
ciplinary translation of one analytic idiom, aquatic ecology and biology, into that of
another, policy assessment. Such interdisciplinary understanding was found to be helpful
in the development of the eiicitation logic and in the specification of its variables.
More specifically, the objectives of the literature review were to:
1. Identify the effects of pH depression and associated metal toxicity upon
aquatic organisms.
2. Identify the mechanisms by which pH depression and metal toxicity are
thought to affect aquatic organisms.
3. Identify the majority of aquatic organisms and all fish species on which
"acid deposition" research has been conducted.
*f. Identify the critical uncertainties of the effects of pH depression and
metal toxicity upon individual species and aquatic communities.
5. Determine if fish populations could be expected to decline as a result of
some greater sensitivity of essential or preferred prey organisms to
declining pH values; i.e., determine if indirect effects are important for
the calculation of fish population mortality.
A cross-referenced indexing system was utilized for the literature reviewed. Each paper
was categorized by species, subject, and metal where appropriate (See Table 3-1). Based
upon the literature review, a list was compiled of a number of important fish for which
3-7
-------
Table 3-1
Subjects By Which the Aquatic Literature Was Classified
Algae
Bioassay
Community
Development
Invertebrates
Lentic
Lotic
Macrophytes
Metal Toxicity
Physiological Damage
Phytoplankton
pH Stress - Acidification
Reproduction
Resistance
Respiration
Survival
Zoop lank ton
3-8
-------
data existed upon the effects of acidic conditions and associated i.-.. . . .> (Table 3-
2). Each fish species was characterized as being either a lentic, lo*':c. -.v lentic-lotic
inhabitant. Within each category, further subdivisions were based upon the predominant
feeding habits of adults. Three food habit categories were recognized: (1) planktivores;
(2) insectivores; and (3) piscivores. Although simplistic, this approach was useful in sort-
ing and evaluating the literature concerning possible food chain and trophic community
affects upon fish populations.
Five life stages were recognized for each species: (1) egg; (2) sac-fry; (3) fry; «0 juve-
nile; and (5) adult. This division allowed the investigators to record the effects of low pH
and toxic metal concentrations upon different reproductive life stages of the fish
species in question. Both field observations and laboratory experiments have suggested
to researchers that the reproductive life stages are particularly sensitive to low pH.
Furthermore, tolerance to low pH and elevated toxic metal concentrations varies signifi-
cantly among the early reproductive or development stages of different species (Baker,
1981; 1982).
Information derived from the literature was recorded as it pertained to the different life
stages of the species in question. Specifically, data were sought on chronic and acute
lethal levels of pH for each life stage, metal toxicity levels, and physiological and popu-
lation responses to each (see Table 3-3 for an example).
3.2.2 The Initial Elicitation Structure
The collection of the extensive set of data outlined in Section 3.2.1 was motivated by the
complexity of the biological effects and the specific elicitation procedure developed.
The framework of the initial elicitation structure resulted principally from interplay
between the investigator's interpretation of the scientific literature and the form of
analysis imposed by subjective assessment.
The structure of the elicitation was designed to incorporate a number of scientific com-
plexities resulting in a subjective damage function expressing fish population decline.
The decline was initially expressed as a percentage reduction in the fish population of a
given stream or lake habitat. Mortality was expressed as a function of the independent
3-9
-------
Table 3-2
Fish Species for Which Literature Contains
Information on Effects of Acid Stress and/or Metal Toxictty1
Salmo clarki
Salmo gairdneri
Salmo salar
Salmo trutta
Salveiinus fontinalis
Carassius auratus
Cyprinus carpio
Phoxinus phoxinus
Pimephales promelas
Catostomus commersoni
Esox lucius
Ictalurus nebulosus
Ictalurus punctatus
Cyprinodon nevadensis
Lebistes reticulatus
Lepomis macrochtrus
Perca fluviatilus
Rutilus rutiius
Brachydanio rerio
Jordanella floridae
Cutthroat trout
Rainbow trout
Atlantic salmon
Brown Trout
Brook trout
Goldfish
Common carp
Minnow
Fathead minnow
White sucker
Northern pike
Brown ballhead
Channel catfish
Desert pup fish
Common guppy
Bluegiil
Perch
Roach
Zebra fish
Flag fish
Species with only anecdotal information available were not included.
3-10
-------
Table 3-3
Example of Data Compilation
Salvelinus fonttnalis (Brook Trout)
lentic-lotic, insectivore.
EGG
pH STRESS
24.6% mortality at pH 4.65
(94) pH 4.5 mortality threshold
(153) pH less than 6.5 caused
significant decrease in egg
hatchability and growth,
pH 4.45 100% mortality
(74)1 Acclimation to sub-lethally
low pH increases survival
of fry in low pH, poor sur-
vival in pH 4.00-4.65.
(94) Survival of fry at pH 4.5.
(153) Alevin mortality 100% at
pH 5.00 in 3 month study.
METAL TOXICITY
(54) Cd: no affect 0.06-6.4pg/l
(69) Cu: 17.5 g/1 no effect
SAC-FRY
(69) 17.4-32.4pg/1 greatly affected
growth and survival.
(54) Cd: Noeffect0.06-6.4ug/l.
(82) Damage occurs at pH 5.2
and lower
FRY
(54) Cd: no effect on survival 0.06-6.4 yg/1
but discussed growth
(69) Cu: same as sac-fry
Access number for literature file.
3-11
-------
Table 3-3
Example of Data Compilation
(Continued)
Salvelinus fontinalls (Brook Trout)
lentic-lotic, insectivore.
pH STRESS
METAL TOXICITY
JUVENILE
(77) C>2 transport capacity of
blood decreases at low pH;
caused by Na loss because Na
balance requires energy,
therefore Na efflux at low pH
(87) low pH causes C^ uptake to
be inhibited, thus net Na loss
(153) 100% mortality at pH 4.45,
mortality 25% at pH 6.56
or lower
(54) Cd: no effect 0.06-6.4 yg/1
but reduced growth
(69) Cu: l7.4-32.5yg/l severely
affected growth and survival
ADULT
(153)1 Survive at pH 4.5, recom-
mended pH maintained at 6.5
or greater.
(168) Decreased survival at pH 4.2
over time. More tolerant to
low pH in winter than summer,
demand less in winter.
(69) Cu: 17.4-3.4yg/I no effect
(54) Cd: 3.4 mg/1 strongly affected
spawning males. 6.4pg/l killed.
1 Access number for literature file.
3-12
-------
variable, pH.* The initial structure proceeded from simple to more complex scenarios
(see Figure 3-1). Initially, the participating scientist would estimate damage based upon
a simple laboratory experiment. Because considerable research on the effects of pH
depression on fish has been performed in bioassay experiments, and because these
experiments allow for the variation of one variable while holding others strictly constant,
the process of estimating dose-response (pH-mortality) relationships was simplified
(Wood, C.M. and D.G. McDonald, 1981; Spry et al., 1981). From this simple scenario, the
scientist would participate in successive elicitations, with each new elicitation adding
more complexity. The last step in this process is an elicitation designed to estimate the
effects of pH decline on selected fish populations in natural habitats.
The gradient of complexity in the initial elicitation structure evolved from a previous
study conducted by ERC. In that work, scientists appeared to have greater success in
providing damage estimates for carefully specified situations in which one variable of an
experimental situation was changed, while the remainder were left constant. When faced
with more uncertain estimates of damages to receptors, e.g., a natural habitat or region,
participants expressed the belief that they knew too little to be able to provide any
damage estimate. Elicitation Steps 1, 2, and 3 in Figure 3-1 were included to serve as
preparatory steps so that participants would be able to complete Step 4 of the initial
elicitation structure.**
In order to apply the initial elicitation framework, the fisheries habitat of the Adiron-
dacks was divided into three parts: the lentic, the lotic, and the lentic-lotic. For each
of these habitats, three different trophic-niches were stipulated, to be filled by those
fish which could be considered either piscivores, insectivores, or planktivores. For each
trophic-niche, at least one species of fish was to be the subject of an elicitation. There-
fore, a minimum of three species of fish - each representative of a different ecological
* Adverse effects of acidification on fish may be a function of both pH levels and
elevated metal concentrations in surface waters. In the Adirondack mountain region,
aluminum, manganese, and zinc have exhibited increased concentrations at depressed pH
levels (Schofield, 1976). Aluminum has received the most attention as being potentially
toxic to fish populations, and research indicates that damage estimates based only on pH
levels may be in error (Baker 1982).
** Reluctance to express subjective judgements about uncertain environmental effects is
probably due to a number of factors, including: (1) unfamiliarity with the concept and its
application; (2) a belief that subjective judgements are not scientific and (3) the improper
specification of the variables in the relation being estimated.
3-13
-------
r
Figure 3-1
Overview of the Initial Elicitation Structure
First Step Elic|tation
Estimates of mortality for different reproductive life stages
of a fish species in a controlled flow-through apparatus.
Second Step Elicitation
Estimates of mortality for a species "population" comprised
of different reproductive life stages in a
controlled flow-through apparatus.
Third Step Elicitation
Estimates of mortality for a single species population
in three "representative" lentic-lotic habitats.
Fourth Step Elicitation
Estimates of mortality for a single species population,
incorporating interspecific interaction and possible changes
in trophic relationships occurring in three
"representative" natural habitats.
3-1*
-------
Figure 3-2
Egg Mortality of Brook Trout Reported in the Literature*,11
Lentic-Lotic
Tnsectivore
Egg
Salvelinas fontinalis (Brook Trout)
pH stress: (74) 24.6% mortality
at pH 9.65; (94) pH 4.5 mortality
threshold; (153) pH 6.5 signif.
egg hatchability and growth; pH
4.45 100% mortal.
I
Metal Toxicity
(69) Cu: 17.5 eg/1 no
effect eggs; (54) Cd:
no affect .06-6.4 eg/I
(183) Al:
°2 Consumpti
:ion
100-
JMortality
eggs
50-
T
3
l
4
i
5
PH
t
6
* Numbers in parenthesis refer to data file listings.
** Graph represents % mortality of eggs as a function of pH. Points are taken from the
literature.
3-15
-------
niche - were to be investigated for each habitat type. Because a number of game fish
would occupy similar positions within this framework, for example, brook trout and rain-
bow trout, the actual number of elicitations would have been greater than three. Inquiry
into non-game species was considered appropriate in order to better represent possible
interspecific effects and food chain disruptions.*
The investigators did not completely specify the fish species to be considered in the
!
analysis. It was decided that the final determination of the fish species to be analyzed
should be a product of the interview process between the investigators and the partici-
pating scientist. Because fisheries scientists or ecologists are generally more familiar
with certain fish species, it was assumed that results of the elicitation would be more
reliable if participating scientists were able to specify those particular fish species for
which they had the greatest confidence in their predictions. As a consequence of this
indeterminancy, the data obtained from the literature review was organized in parallel to
the logic of the elicitation procedure whenever it was available (see Figure 3-2). Every
reproductive stage of the fish species under consideration was evaluated in terms of
percent survival relative to a set of water quality parameters.
The literature review revealed that certain metals are present in increased concentra-
tions in acidic waters, namely zinc, manganese, and aluminum (Schofield, 1976). Pre-
sumably these elevated concentrations are a result of the increased solubility of these
metals at lower pH values. Of these metals, only aluminum has been found to be toxic to
fish at concentrations currently measured in acidified waters (Baker 1981; Baker and
Schofield 1982; Muniz and Leivestad, 1980). On the other hand, calcium has been shown
to decrease the membrane permeability of fish undergoing acid stress, and consequently
decrease the rate of sodium loss. A number of investigators have observed correlations
between decreased calcium content and fish mortality as well as correlation between
increased calcium concentration and fish survival (Spry, 1981). As a result, it was felt
necessary to specify the concentrations of aluminum and calcium in the water for each
pH level. The relationships between aluminum, calcium and pH were taken from regres-
sions computed by Schofield (1976) of Cornell University. These relationships were based
upon water chemistry samples taken from 214 lakes above 610 meters in elevation (see
Figures 3-3 and 3-*).
•Indirect effects and interspecific relations were explicitly structured into the logic of
the analysis in the fourth, and last step of the process (Figure 3-1).
3-16
-------
Figure 3-3
Relationship between aluminum concentrations and pH in Adirondack
lakes more than 610 meters elevation, June 24-27,1975
(log Y = 3.99-0 J4X, r = -0.675)
2000
1000
100
50
20-
PH
3-17
-------
Figure
The relationship between pH and pCa for Adirondack lakes
more than 610 meters elevation (Y = 22.23-3.91 X, r = O.S26)
•o
n
9*
»• • f • ••
• s • • •
3-18
-------
The second elicitation in the initial framework consisted in the aggregation of life stage
effects (see Figure 3-1). A single species population consisting of all reproductive life
stages was, by hypothesis, contained within a single flow through vessel. Assuming
values for the elapsed maturation time of each reproductive stage then enabled esti-
mates of class survivorship'at different pH levels. This "process was to be iterated
through a complete reproductive cycle for each fish population. The time allotted for
maturation of each reproductive stage is shown in Table 3-4.
The second step was designed to capture the effects of declining pH values and associ-
ated metal toxicities on the population structure of a given fish species. As previously
noted, one of the primary mechanisms proposed to explain the loss of fish populations is
the recruitment failure of a new year class. Field observations have been reported which
indicate that acid stressed fish populations often have missing year classes. Step 2 would
enable estimates of population structure given a range of pH values. Step 1, while con-
sidered analytically necessary for the performance of this judgement is obviously not
sufficient. If Step 1 were - in the opinion of the participating scientist - superfluous, he
could then directly perform the elicitation at Step 2.
The third elicitation was designed to obtain estimates of population reduction in three
lakes, and considered only direct effects on a single species. The lakes initially con-
sidered were Woods, Panther, and Sagamore. Each of these lakes is located within the
Adirondack Park Region of New York State and receives similar acidic inputs. Of the
three lakes, Woods is considered acidic (typical outlet pH between 4 and 5), Panther is
considered neutral (typical outlet pH near 7), and Sagamore possesses an outlet pH which
usually varies between that of Woods and Panther. The Woods, Sagamore and Panther
Lake Watersheds have been exhaustively characterized as the result of the Integrated
Lake-Watershed Acidification Study funded by the Electric Power Research Institute.
The characterization utilized for the initial eiicitation was based upon information
published by Chen et al. (1981).
Three lakes characterized by Dr. Charles Driscoll (1980) were also considered. These
were Big Moose Lake, North Lake and Little Moose Lake. Big Moose Lake and North
Lake are both acidic lakes with pH levels below 5, while Little Moose Lake is character-
ized as being neutral with pH values above 7. Dr. Joan Baker has also conducted recent
field studies examining fish populations at Big Moose, North, and Little Moose Lakes
(Baker, 1981). The conjunction of the chemical and biological surveys suggested that
3-19
-------
Table 3U
Maturation and Exposure) Times Allotted for
Different Reproductive Stages in the Initial Elicitation
Eggs
Sac Fry
Fry
Juvenile
2 Weeks
3-4 Weeks
8-10 Weeks
56-60 Weeks
3-20
-------
these lakes might provide better exa»r»c'.- xor the aquatic scientists during the elicita-
tion process. No final decision was made at this stage of the project, as the selection of
"representative" lakes was felt to require the assistance of the scientists selected to
conduct the pretest and evaluation of the elicitation framework.
This third elicitation included considerations of natural habitat sensitivity. The uncer-
tainty in estimates of the relationship between fish mortality and pH included, influences
of lake morphometry, such as uneven mixing, temperature stratification, different inlet,
groundwater upwelling, and outlet chemistries, pulse events, and lake turnover time.
The step 4 elicitation was designed to elicit fish population decline within each of the
three lakes incorporating both direct and indirect effects; e.g., interspecific competition,
replacement, and changes in the structure of the food chain.
The careful specification of watershed geology, soil type, and lake morphometry in Steps
3 and 4 was considered necessary in order to: (1) present standard sets of data to the
scientists participating within the elicitation; (2) partially reflect the range or variability
of chemical, biological, and geological parameters of Adirondack watersheds; (3) provide
some basis for a regional extrapolation of damages to fish populations; and (4) determine
the significance of the variability between "bioassy results" and "lake estimations" for
the assessment of regional effects. Because a number of the scientists were more
familiar with Adirondack water chemistry and fish populations than the remainder of the
participants, a careful specification of the particular lakes under consideration was
required. Identical data sets were required for the elicitation in order to prevent pos-
sible errors or bias in the aggregation of the cumulative damage functions prior to the
computation of regional effects. The lakes were to be selected so that they would be
considered representative of a class of Adirondack lakes sharing similar water quality
parameters. The investigators originally intended to extrapolate from these three
"representative" lakes to obtain estimates of the effects of changing pH records on fish
populations for the Adirondack Mountain Region. However, this initial elicitation struc-
ture was revised extensively prior to its application. The final elicitation structure made
this extrapolation unnecessary.
3-21
-------
33 EVOLUTION OF THE ELIOTATION STRUCTURE
The initial four step elicitation structure presented in the preceding section served as the
j
starting point for developing a structure which would be credible to the scientific com-
munity and still allow for the estimation of • regional effects. The initial elicitation
structure is complex relative to the final elicitation structure. The principal reason for
the complexity of the initial structure was the ^vestigators1 desire for the assessment to
directly capture as many differences in fish population response to different water
I
chemistry variables as possible. After scientific review, it became apparent that while
the initial structure, or something like it, would more explicitly address mechanisms of
fish mortality than would a more simplified approach, that the current inventory data
and time available for the performance of the project would not warrant the application
of these more complex formulations. In addition, the detailed specification of the varia-
bles in the initial elicitation made subjective estimates of the effects very difficult to
perform with any degree of confidence. Selecting the appropriate degree of resolution
for the framework proved, in itself, a difficult interdisciplinary problem. This is sig-
nificant for two reasons. First, it suggests that the selection of the variables for
"regional" assessments of acid deposition effects should be a function of two considera-
tions: the current state of scientific knowledge and the precision of other
complimentary analyses. Second, it suggests the continued importance of scientific con-
tributions to the development and application of assessment "models".
The initial elicitation was developed by the project investigators with only minimal dis-
cussion with outside scientists. The next steps consisted in: the presentation of the ini-
tial structure for comments; a pretest of the elicitation in order to identify problems;
and revisions prior to its actual implementation. Comments were .solicited from three
audiences — researchers studying the effects of depressed pH on fish populations, aquatic
chemists familiar with the Adirondack region, and decision theory researchers familiar
with techniques for eliciting expert judgements.
Before conducting an elicitation, a number of questions needed to be addressed. First,
did the structure of the elicitation reasonably represent the estimation problem?
3-22
-------
Seccrsd. w*:r the independent and dependent variables selected for the elicitation
appropriate?* Third, could scientists reasonably take into consideration the large
number of factors that influence the response of fish populations to low pH and acidifi-
cation? Lastly, what information could be compiled that would help the scientists esti-
mate the range of uncertainty around estimates of the effects of-pH on fish populations?
The first presentation of the initial elicitation structure was to the Duluth EPA Labora-
tory.** The purpose of the meeting was to present the initial elicitation framework to a
scientific audience for a preliminary review. Several recommendations were made for
improving the clarity of the presentation.*** A second presentation of the elicitation
structure was scheduled at Carnegie-Mellon University with a researcher**** familiar
with techniques for eliciting expert judgements. The purpose of this meeting was to
exchange ideas regarding the best method for eliciting the lower and upper bounds for
the effects of pH on fish populations. Potential biases in the manner in which individuals
estimate subjective parameters were considered, as were biases that could result from
the way in which questions were asked or from the presentation of information during the
elicitation.
* An important issue which needed to be considered during this pretest period was
whether the selected dependent and independent variables were appropriate. In the
initial elicitation framework the independent, or causal, variables were a given summer,
surface pH measurement, assumed to be taken in the epilimnion, and related
concentrations of total aluminum and calcium for typical Adirondack waters. The
dependent variable was to be a percent reduction in a species fish population. Scientists
were to consider differences in characteristics of Adirondack lakes in estimating the
range of possible effects for a given pH, including size of the lake, possible existence of
refuges within the lake; genetic variability in the susceptability of fish to pH depressions,
and the likelihood that a lake with a given summer surface pH record would experience
severe spring pH depressions.
** Dr. Orie Loucks, Dr. Gary Glass, and Dr. George Rapp of the EPA Duluth Laboratory
were generous with their time and made many helpful suggestions.
*** The Duluth scientists questioned the general scope of the approach and how the
scientific participants were selected. In order to address these and similar questions, the
project investigators prepared a memorandum for distribution to the selected elicitation
participants outlining the following points: the need for the extension of scientific
results to broader regions; the nature of the uncertainty implicit in such an extension;
and the mechanism proposed by ERC to dimension this uncertainty. It was decided to
include a summary volume of a previous report which suggested application of this
technique in order to dimension the uncertainty surrounding effects estimates, and a
progress report outlining work performed on the project.
**** Dr. Max Henrion of Carnegie-Mellon University was particularly helpful.
3-23
-------
The next set of meetings were with effects scientists for preliminary tests of the struc-
ture of the elicitation.* As a result of these discussions, substantial changes were made
in the elicitation. Many additional uncertainties were identified that affect regional
estimates of the effects of pH on fish populations. While the investigators initially
developed parameters -which "they felt necessary ito a regional assessment (3.2.2), scien-
tific participants often suggested important qualifications in the approach because of
either: (1) lack of inventory data; or (2) lack of scientific understanding of a mechanism
or effect. Figure 3-5 presents a revised intermediate elicitation structure. The revised
structure represents a clear simplification of this investigators' original schema. While
the output of both the initial and the intermediate eticitations remained the same, i.e.,
the percentage reduction of a fish population within a small set of exemplary lakes, the
intermediate elicitation structure aggregates thd elicitation of damages to the different
reproductive life stages (eggs, sac fry, and fry) into one elicitation. Damages to both
juveniles and adults were to be estimated in Step 2. Step 3 of the intermediate elicita-
tion structure provided the same output as Step fr of the initial elicitation structure (See
Figures 3-1 and 3-5). The elicitation of the damage functions was to be performed over
three different Adirondack lakes, Sagamore, Woods, and Panther. Dr. Baker advised
against the use of Big Moose, Little Moose, and North Lake, which were also being con-
sidered, because they were less likely to be representative of a broad range of Adiron-
dack lakes.**
The prediction of impacts upon fish populations in natural habitats as functions of pH
would thus encompass such considerations as: (1) the temporal and spatial variation in
* The assistance from Dr. Joan Baker of North Carolina State University and Dr. Car!
Schofield of Cornell University is gratefully acknowledged. Dr. Baker worked with the
project investigators for several days, asking helpful questions and offering suggestions
for improvement.
** The sampling sites at Big Moose, Little Moose, and North Lake are ail located at mod-
erately high elevations (approximately 550 m). All three lakes are larger than 150 ha.
While both Big Moose and North Lakes are "acjd" (pH is less than 5), Little Moose is a
neutral lake (pH is greater than 7). The geologic makeup surrounding all three lakes is
similar, with the area being underlain by siliceous bedrock (Driscoll, 1980).
3-24
-------
Figure 3-5
Intermediate Elicltation Structure1
1. EUcitation over damages to reproductive life stages (eggs, sac fry, fry)
2. EUcitation over damages to juveniles and adults
3. EUcitation over damages occurring in representative habitats.
1
One and Two are bioassay flow-through experiments. Three is based on set of
exemplary lakes.
3-25
-------
the lake water chemistry in relation to the occurrence of sensitive fish life history
stages; (2) correlations between occurrence of the particular fish species and water
quality within the Adirondacks; and (3) the results of laboratory bioassays and the deter-
mination of the survival of fish eggs and larvie as functions of pH. As with the initial
elicitation structure, the results of Steps 1 and 2 were not considered to be outputs for
the regional assessment, but rather preparatory guides for the performance of the final,
Step 3, elicitation.
In addition to the above changes, the expression of aluminum levels as a function of pH
for use in the elicitation was changed (from tjhat shown in Figure 3-3) to that shown in
Figure 3-6. Figure 3-6 is adapted from Oriscbll (1980) and represents the concentration
of inorganic monomeric aluminum as a function of pH. Work by Baker (1980) and Baker
and Schofield (1981) indicates that the species of aluminum present in dilute waters
influences the toxicity of those waters to fish species.
The final portion of the meetings consisted of a pretest of the elicitation interview
process. Initially, it was hoped that the interview and technique would be sufficiently
polished to allow for the performance of the intermediate elicitation. However, because
of a number of difficulties, the interview was not able to be completed.* The principal
obstacle was the data output from Step 3 of the elicitation. The output in the initial
eiicitation structure was expressed as the percentage reduction of a given fish population
within a lake. Because the data base maintained on fish in lakes by the New York DEC
contains only presence-absence data for fish populations within the Adirondacks, the use
of the elicited damage functions expressing change as percentage reductions of the fish
populations within representative lakes was problematic. To make the elicitation of the
damage function expressed in Step 3 of the intermediate elicitation structure approp-
riate, biomass or population statistics would heed to be developed for the Adirondack
Region. The Morpho-Edaphic Index was considered as a possible mechanism for the
prediction of fish biomass with the DEC data set. It was decided, however, that while
* The unexpected difficulties that were encountered illustrate the importance of pretest
applications of elicitations. Many important insights were obtained from an actual dry
run of the elicitation which were not apparent during the discussions of the procedure.
3-26
-------
Figure 3-6
Inorganic Monomerk Aluminum as a Function of Sample pH
From Driscoll, 1920
L-Inorg
(mg/Al)
.00-
3.6
3-27
-------
the MEI is useful as a predictor of fish biomass ovir a broad range of water chemistries,
the index would not be accurate for the dilute lakes within the Adirondack region. In
addition, even if the ME! could be used successfully to predict biomass, it is unable to
distinguish between the fish species that might or
While this might not be much of a problem for pri narily mono-culture fish habitats, for
example, brook trout ponds, the difficulties associated with its application to other lake
types and habitats precluded its application. j
might not be present within the lake.
During these meetings no decision was made concerning the treatment of these ques-
tions. However, because they were considered to ie potentially very important, a num-
ber of possible elicitation structures were discussed tha_t would satisfactorily meet the
* i
criteria of policy-relevance (i.e., dimension uncertainty and furnish regional effects
estimates) and scientific credibility, while also avoiding the problems posed by the use of
representative lakes and the need for the generation of biomass or population data.
.* THE FINAL ELICITATION STRUCTURE
The final form of the elicitation structure represents a considerable simplification of
both the intermediate and initial elicitation structures. It estimated the percentage of
Adirondack lakes which could possibly, but no longer can, sustain populations of a parti-
cular fish species at a given summer pH measurerr ent taken in the epilimnion. Thus, the
dependent variable was changed from a measure olf the reduction in fish biomass or popu-
lation to a measure of the reduction in a fish species habitat. This change responds to
considerations mentioned earlier, and reflects the general absence of inventory data
expressing biomass or number of fish per -unit of measurement.
Because a substantial gamefish stocking prograni exists in the Adirondacks, separate
elicitations were felt to be necessary for stocked fcnd non-stocked gamefish populations.
This adjustment was necessary because of different sensitivities of the reproductive and
adult life stages to pH depressions. Scientists were asked to estimate the percentage of
Adirondack lakes which could possibly, but no longer can sustain both: (1) self-sustaining
populations of a particular game fish; and (2) projperly stocked populations of juvenile
and adult game fish. For non-game species, for example, white sucker, the sensitivity of
only unstocked populations was evaluated.
3-28
-------
The obvious structural difference between the final elicitation structure and its prede-
cessors consists of the elimination of bioassay estimations of population mortality.
These were eliminated for two reasons. First, time required for the elicitation interview
would be significantly reduced. Second, the elicitation over the different life stages of
the fish species did not, in the opinion of several scientists, significantly contribute to
the third and most important elicitation result.*
An advantage inherent in the final elicitation structure is that the scientists directly
extended the estimation of regional effects over the Adirondack mountain region. The
method proposed in the initial and intermediate elicitation structures required the extr-
apolation of the elicitation results for three "representative" lakes to the entire Adiron-
dack region. While this procedure reduces uncertainty in the estimation of fish damages
for the "representative" lakes, the extrapolation to the Adirondack Region would be done
without the explicit dimensioning of the uncertainty by the scientists.
* The relative uncertainty of response in these hypothetical bioassay experiments was
quite small when compared to the uncertainty of other aspects of the regional estimation
of effects.
3-29
-------
-------
4.0 IMPLEMENTATION AND RESULTS
This chapter reviews the implementation of the final elicitation structure and the result-
ing damage relationships obtained through interviews with participating scientists.
Section 4.1 discusses the work done in preparation for the elicitations including the
selection of the participating scientists. Section 4.2 discusses the actual performance of
elicitation. Section 4.3 summarizes the results of the elicitations.
4.1 PREPARATION FOR THE ELICITATION
4.1.1 Selection of Participating Scientists
The selection of scientists for participation in the elicitation was an important part of
the project. The first step was to compile a list of authors from the peer review litera-
ture and government reports consulted by ERC during the literature review. Weighting
in selection was assigned on the basis of: (1) the number of recent publications; and
(2) the amount of work performed in the Adirondack mountain region. On the basis of
this initial review, a list of scientists was compiled. A number of Canadian scientists
were included within the list of possible participants in order to provide a comparison
between Canadian and American research communities.
The second step consisted in the contacting of possible participants and determining their
willingness to participate in the study. During these initial phone conversations, the
length of the proposed interview (approximately four hours), the research area, and the
information requirements of the project were outlined. On the basis of these discussions,
the number of potential candidates was narrowed down. On the other hand, a number of
additional candidates were proposed by those scientists initially contacted. Table 4.1
contains a list of the scientists who finally participated in the study and the dates of
their interviews with ERC.
4-1
-------
These elicitation interviews were complemented by discussions with Walter Kretser of
the New York State Bureau of Fisheries and Charles Driscoll of Syracuse University. The
purpose of the discussion with Mr. Kretser was to review the records maintained within
the Adirondack Waters Data Management System. As a result of conversations with Mr.
Kretser, the investigators obtained the complete data file for the 3506 ponded waters
within the Adirondack area. The discussions with Dr. Driscoll addressed; first, the
general review of the project's scope and methodology; and second, a review of the water
chemistry parameters being proposed for use in the study.
*.i-2 Preparatory Orientation and Review Material
Prior to the interview process, all the scientific participants received orientation packets
in the mail. The mailing took place approximately two weeks prior to the elicitation.
The purpose of this mailing was to introduce the approach and the project goals in more
detail to the elicitation participants. The package included three separate documents:
(1) A description of the meeting agenda and a brief explanation of the
rationale behind this type of approach.
(2) A project progress memo which reviewed the status of the work per-
formed on the project and outlined the remainder of work to be accomp-
lished.
(3) A summary of a previous report providing background information on
the approach.*
4.1.3 Compilation of Background Data for Use in the Elicitation
A notebook of data was compiled as a reference for the scientists to use during the elici-
tation. This data included a table adapted from Baker and Schofield (1981) listing the
mean concentrations of selected ions for Adirondack lakes and streams sampled during
* "Assessing the Benefits of Policies Designed to Reduce Acid Depositions A Decision-
Analytic Benefit-Cost Framework," (Violette and Peterson, 1982).
4-2
-------
Table 4.1
Interview Participants
Name
institution
Date
of Interview
Carl Schofield
Department of Natural Resources
Cornell University
Ithaca, New York
11/5/82
Martin Pfeiffer
Bureau of Fisheries, DEC
Route 86
Ray Brook, New York
11/8/82
Douglas Spry
Department of Biology
Mimaster University
Hamilton, Ontario
11/10/82
Gordon Craig
Ministry of Environment
Rexdale, Ontario
11/12/82
Gail Beges
Ministry of Natural Resources
Toronto, Ontario
11/15/82
3im MacLean
Ministry of Natural Resources
Toronto, Ontario
11/15/82
3oan Baker
North Carolina State University
Acid Deposition Program
11/17/82
4-3
-------
1977-1978 (^se IV '•- While this information was of only limited value for the
participants with e:.'tensi /c understanding and Knowledge of Adirondack waters, it pro-
vided a useful indicator of the constituents likely to be present in dilute Adirondack
waters for the Canadian participants. Relationships between pH and pCa and between
pH and inorganic monomeric aluminum were also included.
The relationship between pH and inorganic monameric aluminum was also represented in
a table (see Table ^-3). The only other data shown to all participants was the Environ-
mental Protection Agency's Critical Assessment Document Table *-24 (Baker, 1982
Draft).* The data was provided in order to supplement the scientists' judgment and to
provide a convenient and uniform reference soiree for all participants. In some cases,
however, participants supplemented this data by consulting their own sources during the
|
course of either the pre-elicitation interview orjthe elicitation itself.
*3. PERFORMANCE OF THE ELICITATION
The meeting with elicitation participants can b0 divided into four segments:
(1) an introductory briefing and interview
(2) discussion of the parameters to be estimated and variables considered
(3) the performance of the damage fundtion elicitation
a review and assessment of the elicijtation procedure and its results.
While each of the meetings followed the basici structure outlined above, there were dif-
ferences between the interviews. These differences are attributable more to the dif-
ferent information requirements of the scientific participants than any variability of
methodological design or implementation on th4 part of the investigators. This section
* A number of participants were shown data representing the temporal and spatial
variability of water chemistry within Dart Lfke in the Adirondacks (unpublished data
provided by Charles Driscoll).
-------
Table 4-2
Mean concentrations (mg/l)(±. standard error) of selected ions in softened, dechlorinated
water utilized in all laboratory experiments and in Adirondack lakes and streams sampled
during 1977-1978
Ion
Ca
Mg
Na
K
freeF
S°4
Adirondack
2.0
0.4
0.7
0.4
0.025
6.2
waters
±0.6
±0.1
±0.3
±0.1
±0.060
±0.9
Table adapted from Baker and Schofield, 1981.
4-5
-------
1
Table 4-3
Approximate Inorganic Aluminum Concentration for Adirondack Ponded Waters*
(mg/1)
pH
73.
6.9
6.6
6.3
6.0
5.7
5.*
5.1
4.8
4.7
4.5
4.4
4.2
3.9
3.6
Al
neg.**
neg.
neg.
neg.
.02
.04
.06
.11
.21
.24
.38
.44
.6
neg.**
neg.**
Note: At lower elevation, organic cornplexation could reduce the availability of inor-
ganic aluminum at or below pH 4.4.
* Used as background data for elicitation.
** Negligible levels of inorganic aluminum concentrations.
4-6
-------
will address the three segments of the interview listed above while indicating the range
in variability of the interview format and process.
4.2.1 Introductory Briefing
The first step in every interview consisted in the review of a briefing package prepared
by EEC staff in order to further acquaint participants with the project, its metho-
dological justification, and data requirements. The principal point made during this seg-
ment of the interview was that any decision to implement an environmental policy must
be based upon a probability assessment of the odds of benefits (monetary or non-
monetary) from the policy exceeding the costs of its implementation. The purpose of the
project was stated to be the construction and implementation of a framework which
would allow scientists to participate in the direct extension of the results of recent
scientific research into damage estimates useful for policy analysis. Secondly, it was
stressed that this was an evaluation of this approach for dimensioning and incorporating
scientific uncertainty in estimates of the regional effects of acid deposition.
Some time was spent explaining what the investigators believed were the principal
advantages to the approach being utilized in the analysis. The advantages expected are:
o First, that scientific information is less likely to be misinterpreted when
the uncertainty is directly and explicitly incorporated into the analysis;
o Second, it was stressed that the subjective probability estimates which
express the damage to environmental receptors in this approach are not
point estimates, but express the damages as a range.. For example,
rather than estimating that 20% of the lakes with a given summer pH
value would no longer support a self-sustaining brook trout population,
the approach estimates with a high confidence interval that, say,
between 5% and 60% of the lakes with a given summer pH value would
no longer support a self-sustaining brook trout population;
o Third, it was pointed out that because the scientist would be directly
responsible for the subjective extension of research results, that the
extension would better represent scientific opinion concerning the
effects than if it were performed by policy-makers or economists;
-------
Fourth, because the probability judgements are expressed quantitatively
and are explicitly represented the;y are amenable to scientific review
and criticism; and ;
Finally, the results of the elicitation can be readily and inexpensively
revised in the light of additional research results and conclusions.
At this point, the consensus judgement was: Q) an objective expression (i.e., statistically
based estimate) of regional fish damages caused by acid deposition is not currently
possible; and (2) regional estimates are required in order to assess policy alternatives
with respect to the mitigation of acid deposition.
The length of discussion concerning the theoretical and general aspects of the discussion
noted above varied significantly between the interviews. It was the investigators' inten-
tion to cover as fully as possible, given time Constraints, the queries and responses of the
participants. In one case, for example, the discussion centered - for some time - on the
need to perform any regional quantitative assessments of the damage occurring as a
result of acid deposition and its effects. It was stressed that while there have been a
number of attempts to incorporate scientific information and opinion within a policy
framework in the acid deposition debate, that none of those attempts have been viewed
as wholely satisfactory - either in terms of i^s explicit treatment of the scientific issues
or in its use of economic and decision analytic theory.*
The final step of the introductory briefing consisted of a very brief review of subjective
probability assessment and its previous applications within environmental policy deci-
sions.
* One important benefit of this project wa
individuals involved with the overall integration
policy analysis. The interdisciplinary discuss
the communication between scientists and
and assessment of the information for
ons that resulted were very informative.
-------
*-2-2 Variable Specification
This part of the interview was a review of the independent and dependent variables used
in the analysis. The discussion concerned the scope -of the eHcitartion and had several
purposes:
(I) to ensure that the participant understood what the variables repre-
sented;
(2) to outline the data compiled by the project investigators for use by
participants in the elicitation (discussed in Section 4.1.3);
(3) to determine if any important source of variability or uncertainty had
been omitted by the investigators;
(4) to agree upon the fish species for which estimates would be developed.
As noted in Section 4.3, the elicitation was to determine the percent of Adirondack
Lakes which could possibly, but no longer can, sustain a given fish population at a speci-
fied summer surface pH measurement. The independent variable was pH. However,
recognizing the potential importance of inorganic monomeric aluminum and calcium in
determining fish response to low pH values, regressions relating calcium and and inor-
ganic monomeric aluminum to pH in Adirondack waters were used to incorporate these
factors within the analysis. No other potentially toxic or mitigative metals were expli-
citly considered. No elicitations were performed without incorporating the effects of
monomeric inorganic aluminum and calcium upon the dependent variable.*
* Metals which consistently exhibit increased concentrations as functions of depressed
pH are aluminum, manganese, and zinc. However, neither manganese nor zinc have been
found to be toxic to fish at concentrations found in acidic surface waters. Aluminum,
however, has been found to be toxic to fish at concentrations as low as .1 to .2 mg/1, a
level within the range of concentrations measured in surface waters. Free aluminum ion
or aluminum hydroxide forms have been shown to be more toxic than organic aluminum
complexes. The solubility of inorganic aluminum in acidic Adirondack surface waters is
apparently regulated by some form of aluminum trihydroxide solid (Driscoli, 1980; Baker,
1981).
-------
The estimation of the percentage of Adirondack Lakes which would be incapable of sup-
porting fish populations at a given summer pH value is an, admittedly, complex and dif-
ficult task. Fish mortality — even at the sarmt pH — can be expected tc vary depending
upon a number of factors described by the monphometric and biogeologic characteristics
of the lakes. Bog, seepage, and eutrophic lakes can all be expected to have different
capabilities to support a fish population given the same pH measurement. In addition,
the presence or absence of ground water upwellings, episodic or pulse pH depressions,
uneven water mixing (the temporal and spatial variability of chemical species within the
lake), acid neutralizing or acidifying inlets, and the biologic production of alkalinity are
reported to be related to fish mortality within the literature. Genetic variability as well
as tiie possibility of avoidance behavior an
-------
Table 4-4
Fish Species Covered in the ElicitatLon
Game Species
Salmon idae
Brook Trout (Salve linus iontinalis)
Lake Trout (SalveUnus namavcush)
Centrarchidae
Small Mouth Bass (Micropterus dolomteui)
Non-Game Species
Catosmidae
White Sucker (Catostomus commersoni)
Cyprinidae
Fat Head Minnow (Pimephales promelas)
Common Shiner (Notropis cornutus)
-------
The Implemention of the Final Elicitation ructure
This section reviews the procedure followed during the actual estimation of damages by
the scientific participants.* The eliciting of the ranges of potential effects at each pH
level followed an established procedure. Compared with the variability of the other
components of the interview, its performance followed a standardized format.
The initial step was to introduce a table similar to that shown in Table *.5.** The ERC
investigator began by asking the participant if there were any possibility of a summer pH
measurement of 6.9 or above resulting in a reduction in the number of lakes which could
support a fish population, in this example a non-$tocked population of brook trout. After
some discussion of the question, the expert concluded that there was, in his opinion, no
chance of a set of Adirondack lakes with a sumrjier pH of 6.9 or greater losing fish habi-
tat. As a result, zeros were entered in the low,i midpoint, and high columns of the table
(See Table
-------
Table 4-5
Eiicitation Table
Species: Salvelinus fontinalis (brook trout)
Range of Impacts
Constituents
pH6.9
pH6.8
pH6.7
pH6.6
pH6.5
pH6.3
pH6.2
Low
Low
Low
Low
Low
Low
Low
Low
midpoint
Expert:
Date:
Notes: Self sustaining population
High
High
High
High
High
High
High
High
ft-13
-------
bound to the range. As a result, 0 percent was entered in the low column. On the other
hand, to establish the high end of the range, the 'investigator would ask the participant to
consider the greatest possible impacts to the set of Adirondack lakes which both possess
summer pH measurements of 6.8 and self-sustaining brook trout populations. After some
discussion, the participant indicated that th^ high end of the range should be 15
percent. As a result, a 15 percent was entered in the high column (see Table *-5a).
These lower and upper bounds were chosen to represent a 95 percent confidence
interval.* Thus, the lower and upper bounds entered in the table do not determine the
absolute limits to the range of possible damages but, instead, serve to define effective
limits such that there is only a small probability that damages fall outside the range. In
this case, the assumed 95 percent confidence interval implies that there is a 2.5 percent
chance that the damages will be below the lower bound, and an equivalent probability
that damages will be above the upper bound.
Once the range of impacts for a given pH had been specified in the above manner, the
participants were questioned regarding where within this range the actual outcome was
most likely to fall. It was quite difficult to elicit responses regarding the most likely
damage outcomes within the estimated range. A procedure was developed which was
acceptable to the scientists and still provided tome information on where the most likely
outcome would fall. The procedure followed was to enter the mid-point of the range in
the table; for example, if the range were 0 percent to 15 percent the mid-point would be
7.5 percent. Upon entering the value for the mid-point, the participant was asked to
indicate whether the percentage of lakes no longer able to sustain fish populations would
be likely to lie above or below the mid-point. In our example, at pH value of 6.8, the
participant indicated that the range between 0 percent and 7.5 percent was more likely
to contain the true value than the range between 7.5 percent and 15 percent. Once the
participant decided whether the impacts were1 likely to be above or below the midpoint of
the range, an "X" was placed either above or below the mid-point in order to express this
weighting of the probabilities (see Table 4-5b).
The elicitation continued until estimates for ail the pH levels were developed. In many
cases the participant felt that the incremental change of .1 pH was not sufficient to
i
change the estimates listed above. In the current example, the participant, didn't change
* See Section 5.1 for a discussion of the statistical interpretation of the range expressed
by the scientific experts.
-------
.able f-5a
Eiicitation Table
Species: Salvelinus_ fontinalis (Brook Trout)
Range of Impacts
Constituents
Low
pH6.9
Midpoint
High
(0) {) (0) () (0)
Low
pH6.8
Low
pH6.7
Low
pH6.6
Low
pH6.5
Low
Low
pH6.3
Low
pH6.2
High
( 0 ) ( X ) ( 7.5 ) ( ) ( 15 )
High
High
High
Hieh
High
High
Expert:
Date:
Notes: Self Sustaining Population
-------
Table 4-5b
Condensed Elicitation Table
Species: Salvelinus fontinalis (Brook Trout)
Range of Imbacts
Constituents
pH6.9
pH6.S
pH5.5
pH5.2
pH5.0
Low
( 0 )
Low
Low
Low
Low
Low
Low
Low
Low
Low
midpoint
") ( )
midpoint
midpoint
midpoint
midpoint
midpoint
midpoint
midpoint
midpoint
High
( 0 )
High
( 0 ) ( X ) ( 7.5 ) ( ) < 15 )
High
( 10 ) ( X ) ( 20 ) ( ) ( 30 )
High
( 10 ) ( ) ( 35 ) ( X ) ( 60 )
midpoint
High
( 30 ) ( X! ) ( 55 ) ( ) { 80 )
High
) ( X ) ( 65 ) { ) ( 90 )
High
) ( ) ( 65 ) ( X ) ( 90 )
High
( 60 ) ( ) ( 75 ) ( X ) ( 90 )
High
( 90 ) ( ) ( 95 ) ( X ) { 100 )
High
{ 100 ) { ) { X ) ( ) ( 100 )
-------
the range of the damage function until the pH had reached 5.5. At that pH, the concen-
tration of inorganic monomeric aluminum was estimated to be approximately .05 mg/1,
and the pCa was approximately *.3. At these values, the lower bound of effects was
estimated to be 10 percent and the upper bound 30 percent (see Table *-5b).
In some cases, the participants were so uncertain that they were unable to weight either
the "high" or "low" interval. In such a case, no weighting was placed on either side of the
midpoint.
Not all of the elicitations began at the highest pH values, 6.9 and above, then proceeding
to lower pH values. In a number of elicitations, participants elected to begin at pH
values around 5.0, working toward both the higher and lower extreme. This is not sur-
prising, because the majority of bioassay experiments, as well as anecdotal literature,
note effects at a pH of 5.0 or less in dilute surface waters or their laboratory equi-
valent. The investigators do not believe that this change in the procedure affected the
estimated ranges.
Review and. Assessment of the Elicited Damage Functions
The final step in the eiicitation interview consisted in a review of the range and weight
which resulted from the eiicitation. During the review, the participant and the investi-
gator examined the range of impacts listed at the different pH values. If the participant
believed that the estimates accurately represented his beliefs, no changes were made.
On the other hand, if there were distributions which seemed inappropriate, then these
were reconsidered and changed where appropriate. Often this amounted to the alteration
. " -**
of just a few values. In some cases, however, substantial revisions were required before
the participant believed the results to accurately reflect his/her perceptions of the
possible range of damages or impact.
Following these revisions to the estimates, if sufficient time was available, the interview
was closed with a brief discussion concerning the participant's evaluation of the approach
and his confidence in the use of its results.
-------
RESULTS OF THE ELICITATION
This section presents the data obtained from the participating scientists that was
directly incorporated within the calculation of regional changes due to the incremental
shift of regional fresh water chemistry. Table fc-6 presents the results of each elicita-
tion. The data will be interpreted in Chapters 5 and 6.
-------
Table 4-6a*
Brook Trout - Stocked
pH
7.0
6.9
6.8
6.7
6.6
6.5
6.*
6.3
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Low Weight
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5 x
5 x
8 x
8 x "
8 x
30
30
50 x
50 x
50 x
70
Mid 'Point 'Weight
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2.5
2.5
10 x
10 x
10 x
27.5
27.5
31.5
31.5
31.5
50 jc
50 x
72.5
72.5
72.5
85
High
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5
5
20
20
20
50
50
55
55
55
70
70
95
95
95
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-19
-------
Table 4|6b*
Brook Trout | Stocked
pH Low Weight
7.0 0
6.9 0
6.8 0
6.7 0
6.6 0
6.5 0
6.4 0
6.3 0
6.2 0
6.1 0
6.0 0
5.9 0
5.8 0
5.7 0
5.6 0
5.5 0
5.4 0
5.3 0
5.2 0
5.1 0
5.0 0 x
4.9 0 x
4.8 10
4.7 10
4.6 25
4.5 25
4.4 40
4.3 40
4.2 50
4.1 70
4.0 70
Mid Point Weight High
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 .0
0 0
0 0
12.5 25
12.5 * 25
12.5 25
12.5 25
12.5 25
37.5 75
37.5 x 75
42.5 x 75
42.5 x 75
62.5 x 100
62.5 x 100
70 -x 100
70 x 100
75 x 100
85 x 100
85 x 100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-20
1
-------
Table 4-6c*
Brook Trout - Stocked
pH
7.0
6.9
6.8
6.7
6.6
6.5
6.4
6.3
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Low Weight
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5
15
30 x
40
60
85
95
100
100
100
100
100
100
100
Mid Point Weight
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2.5
7.5 x
20
40
55
72.5 x
90 x
97
100
100
100
100
100
100
100
High
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5
10
25
50
70
85
95
99
100
100
100
100
100
100
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-21
-------
Table 4-6d*
Brook Trout - Self-sustaining
.. pH
7.0
6.9
6.8
6.7
6.6
6.5
6.*
6.3
63
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5.3
53
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Low
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
10
10
10
10
10
30
30
40
40
40
40
60
90
90
90
100
Weight
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Mid Point Weight
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
20
20
20
35 x
35 x
55
55
65
65
65 x
65 x
75 x
95 x
95 x
95 x
High
0
15
15
15
15
15
15
15
15
15
15
15
15-
15
15
30
30
30
60
60
80
80
90
90
90
90
90
100
100
100
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
-------
Table 4-6e*
Brook Trout - Self-sustaining
pH
7.0
6.9
6.8
6.7
6.6
6.5
6.4
6.3
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Low Weight
0
0
0
0
0
0
0
0
0
0
0
0
0 x
0 x
0 x
0 x
0 x
0 x
10 x
10 x
25
25
25
25
50
50
50
80
80
80
100
Mid Point Weight
0
0
0
0
0
0
0
0
0
0
5
5
15
15
25
25
37.5
37.5
42.5
42.5
50 x
50 x
50 x
62.5 x
75 x
75 x
75 x
90 x
90 x
90 x
100
High
0
0
0
0
0
0
0
0
0
0
10
10
30
30
50
50
75
75
75
75
75
75
75
100
100
100
100
100
100
100
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-23
-------
Table 4-6f*
Brook Trout - Self-sustaining
pH
7.0
6.9
6.8
6.7
6.6
6.5
6.4
6.3
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5J
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Low
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5
5
5
5
5
10
10
10
10
10
20
20
20
20
20
75
Weight
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Mid Point Weight
0
0
0
0
0
0
0
0
0
0
5
5
5
5
5
17.5
17.5
17.5
17.5
17.5
37.5
37.5
37.5
37.5
37.5
50
50
50
50
50
87.5 x
High
0
0
0
0
0
0
0
0
0
0
10
10
10
10
10
30
30
30
30
30
65
65
65
65
65
80
80
80
80
80
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
tiie midpoint.
4-24
-------
Table 4-6g*
Brook Trout - Self-sustaining
pH
7.0
6.9
6.8
6.7
6.6
6.5
6.4
6J
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.*
5J
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Low Weight
0
0
0 x
0 x
0 x
0 x
0
0
0
0
10
10
20 x
20 x
20 x
20 x
40
40
60
60
80
80
90
90
90
90
100
100
100
100
100
Mid Point Weight
0
0
7.5
7.5
7.5
7.5
7.5 x
7.5 x
7.5 x
7.5 x
50 x
50 x
55
55
55
55
65
65
80
80
90
90
95
95
95
95
100
100
100
100
100
High
0
0
15
15
15
15
15
15
15
15
90
90
90
90
90
90
90
90
100
100
100
100
100
100
100
100
100
100
100
100
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-25
-------
Table 4-6ht
Brook Trout - Self-sustaining
pH
7.0
6.9
6.8
6.7
6.6
6.5
6.4
6.3
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4,5
4.4
4.3
4.2
4.1
4.0
Low
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5
5
5
5
10
10
10
20
25
35
45
50
75
85
95
95
95
Weight Mid Point Weight
0
0
0
0
0
0
0
0
0
0
x 7.5
x 7.5
x 7.5
10 x
x 17.5
x 20
x 25
30 x
32.5 x
35 x
40 x
45 x
55 x
65 x
x 70
x 72.5
86.5 x
92.5 x
97.5 x
97.5 x
97.5 x
High
0
0
0
0
0
0
0
0
0
0
15
15
15
20
30
35
50
55
55
60
70
70
85
95
95
95
98
100
100
100
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-26
-------
Table 4-6i*
Lake Trout - Stocked
pH
7.0
6.9
6.8
6.7
6.6
6.5
6.4
6J
62
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Low Weight
0
0
0
0
0
0
0
0
0
0
0 x
0 x
10
10
10
10
25
25
40
40
60
60
60
60
60
100
100
100
100
100
100
Mid Point Weight
0
0
0
0
0
0
0
0
0
0
5
5
20
20
20
20
50 x
50 x
70 x
70 x
80 x
80 x
80 x
80 x
80 x
100
100
100
100
100
100
High
0
0
0
0
0
0
0
0
0
0
10
10
30.
30
30
30
75
75
100
100
100
100
100
100
100
100
100
100
100
100
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-27
-------
Table 4-6 j*
Lake Trout -'Stocked
pH
7.0
6.9
6.S
6.7
6.6
6.5
6.4
6.3
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.*
53
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Low Weight
0
0
0
0
0
0
0
0
0
0
0
0
0
0 x
0 x
0 x
0 x
0 x
10 x
10 x
30 x
30 x
50
80
80
100
100
100
100
100
100
Mid Point Weight
0
0
0
0
0
0
0
0
0
0
0
0
0
12.5
12.5
25
25
25
55
55
65
65
75
90
90
100
100
100
100
100
100
High
0
0
0
0
0
0
0
0
0
0
0
0
0
25
25
50
50
50
100
100
100
100
100
100
100
100
100
100
100
100
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-28
L
-------
Table 4-6k*
Lake Trout - Self-sustaining
PH
7.0
6.9
6.8
6.7
6.6
6.5
6.4
6.3
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
tow Weight
0
0
0
0
0
0 x
0 x
0 x
0 x
10 x
40
40
50
50
50
50
80
80
80
80
100
100
100
100
100
100
100
100
100
100
100
Mid Point
0
0
0
0
0
5
10
10
10
35
57.5
60
70
70
70
70
90
90
90
90
100
too
100
100
100
100
100
100
100
100
100
Weight High
0
0
0
0
0
10
20
20
20
60
75
80
90
90
90
90
x 100
x 100
x 100
x 100
100
100
100
100
100
100
100
100
100
100
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-29
-------
Table 4-61*
i
Lake Trout - Self-sustaining
pH
7.0
6.9
6.8
6.7
6.6
6.5
6.4
6.3
63
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Low
0
0
0
0
0
0
0
0
0
0
0
10
10
10
10
10
10
10
25
25
50
50
80
100
100
100
100
100
100
100
100
Weight
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
1 Mid Point Weight
0
0
0
0
0
5
5
5
5
5
5
30
30
30
45
45
55
55
62.5
62.5
75 x
75 x
90 x
100
100
100
100
100
100
100
100
High
0
0
0
0
0
10
10
10
10
10
10
50
50
50
90
90
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-30
-------
Table 4-6m*
Lake Trout - Self-sustaining
pH
7.0
6.9
6.8
6.7
6.6
6.5
6.4
6.3
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5J
53
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Low
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5
5
5
5
5
10
10
10
10
10
20
20
20
20
20
75
Weight
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Mid Point Weight
0
0
0
0
0
0
0
0
0
0
5
5
5
5
5
17.5
17.5
17.5
17.5
17.5
37.5
37.5
37.5
37.5
37.5
50
50
50
50
50
87.5 x
High
0
0
0
0
0
0
0
0
0
0
10
10
10
10
10
30
30
30
30
30
65
65
65
65
65
80
80
80
80
80
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-31
-------
Table 4-6n*
Small Mouth &ass
pH Low Weight ' Mid Point
7.0
6.9
6.8
6.7
6.6
6.5
6.4
6J
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
0
0
0
0
0
0
0
0
10
10
20
20
20
30
30
30
30
40
40
40
70
70
90
90
100
100
100
100
100
100
100
0
0
0
0
0
0
0
0
17.5
17.5
35
35
35
50
50
55
55
65
65
65
85
85
95
95
100
100
100
100
100
100
100
Weight High
0
0
0
0
0
0
0
0
25
25
50
50
50
x 70
x 70
x 80
x 80
x 90
x 90
x 90
100
100
100
100
100
100
100
100
100
100
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-32
-------
Table 4-6o*
Small Mouth Bass
pH
7.0
6.9
6.8
6.7
6.6
6.5
6.*
6.3
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4,3
4.2
4.1
4.0
Low Weight
0
0
0
0
0
0
0
0
0
0
0
0
10
30
40
50 x
60
100
100
100
100
100
100
100
100
100
100
100
100
100
100
Mid Point Weight
0
0
0
0
0
0
0
0
0
0
0
2.5
20
40
50
65
75
100
100
100
100
100
100
100
100
100
100
100
100
100
100
High
0
0
0
0
0
0
0
0
0
0
0
5
30
50
60
80
90
100
100
100
100
100
100
100
100
100
100
100
100
100
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-33
-------
Table 4-6p*
White Sucker
pH
7.0
6.9
6.8
6.7
6.6
6.5
6.4
6.3
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Low
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5
5
5
20
20
20
20
20
30
30
40
50
50
93
93
100
Weight
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Mid Point Weight
0
0
0
0
0
0
0
2.5
2.5
10
10
10
10
20
20
27.5
27.5
27.5
35
35
45
50
50
55 x
55 x
67.5 x
74.5 x
74.5 x
96.5 x
96.5 x
100
High
0
0
0
0
0
0
0
5
5
20
20
20
20
40
40
50
50
50
50
50
70
80
80
80
80
85
99
99
100
100
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-341
-------
Table 4-6q*
Common Shiner
PH
7.0
6.9
6.8
6.7
6.6
6J
6.*
6J
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5J
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Low Weight
0
0
10
10
19
10
10
40 x
40 x
40 x
50
50
. 50
50
50
50
50
70
90
90
100
100
100
100
100
100
100
100
100
100
100
Mid Point Weight
0
0
25 x
25 x
25 x
25 x
25 x
60
60
60
65
65
65
65
65
70
70
80
95
95
100
100
100
100
100
100
100
100
100
100
100
High
0
0
40
40
40
40
40
80
80
80
80
80
80
80
80
90
90
90
100
100
100
100
100
100
100
100
100
100
100
100
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-35
-------
Table 4-6r*
Fat Head Minnow
PH
7.0
6.9
6.8
6.7
6.6
6.5
6.4
6J
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
53
52
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Low Weight
0
0
0 x
0 x
10 x
10 x
10
10
10
10
20 x
20 x
30 x
30 x
50
50
50
60 x
60 x
60
80
80
100
100
100
100
100
100
100
100
100
Mid Point Weight
0
0
10
10
15
15
20
20
20
20
30
30
50
50
60
60
75
75
75
75
90
90
100
100
100
100
100
100
100
100
100
High
0
0
20
20
30
30
30
30
30
30
40
40
70
70
70
70
90
90
90
90
100
100
100
100
100
100
100
100
100
100
100
* Where no weighting is indicated by the placement of an x, the weighting is placed over
the midpoint.
4-36
-------
5.0 INTERPRETATION AND USE OF THE ELICITATION DATA
This chapter addresses the use and application of the estimates obtained from the elici-
tations. Section 5.1 discusses the conversion of the basic data provided by the eiicita-
tions into marginal and cumulative probability distributions. Section 5.2 discusses the
data bases used by the investigators in order to apply these results to the lakes in the
Adirondack region. Section 5.3 presents and discusses the results of applying the cumu-
lative probability distributions developed in Section 5.1 to the data base discussed in
Section 5.2.
There are a number of ways to interpret and use the estimates obtained from the elicita-
tion. One way is to simply show the range of estimated effects. Since each expert esti-
mated lower and upper bounds of effects, these can be used to dimension the range of
effects due to a change in pH. If two additional assumptions are made, probabilistic
estimates of damages from acid deposition can also be obtained.
5.1 STEPS NEEDED TO OBTAIN PROBABILISTIC EFFECTS ESTIMATES
In order to obtain probabilistic estimates of the extent of change occurring to fish popu-
lations, the results of the elicitations presented in Table
-------
5.1.1 Assumptions
The initial assumptions made were that the range estimated by the scientist represented
a 95 percent confidence interval and that, when a weight was placed on one side of the
range mid-point, the effects were twice as likely to occur in the weighted segment as in
the unweighted segment. It is important to recognize that both the use of a 95 percent
confidence interval and a 2:1 weighting are assumptions. There is no quantative or stat-
istical justification for either assumption. However, a 95 percent confidence interval
was felt to be reasonable since during the elicitation the scientists were asked to dimen-
sion a range of potential effects so that, in their opinion, the actual impact would almost
certainly lie within that range. Assuming that the range represents a high confidence
interval is consistent with the way the elicitation was conducted. The assumption of a
2:1 weighting is more problematic since the scientists were not asked during the elicita-
tion to determine what the relative probabilities were between the weighted and un-
weighted segments. Whenever the scientists were asked to actually assign different
probabilities to the different segments, they resisted making the estimate. It was felt
that the simple weighting used in the elicitation pushed the limits of the scientists ability
to dimension the uncertainty surrounding the effects estimates. The scientists simply did
not feel that they could provide reliable estimates of actual probabilities, but they felt
they could indicate on which side of the range mid-point the actual outcome would most
likely fall. Given that this was all the information that could be reliably obtained, the
task facing the project investigators was to design a method that credibly used this
sparse data.
Since the specification of a confidence interval and interpretation of the weighting are
both assumptions, it is important to perform a sensitivity analysis around any selected
values. Confidence intervals of 90 percent anld 80 percent were used along with a 95 per-
cent confidence interval to test the sensitivity of the confidence interval assumption. In
addition to the 2:1 weighting assumption, sensitivity analysis was performed using a 3:2
and 5A weighting.
These different confidence intervals and weighting assumptions were felt to bound the
reasonable range of assumptions. The 95 percent confidence interval is quite a high con-
fidence interval (ie., the probability that the actual value would fall outside the scien-
tists' estimated range is only 5 percent). Conversely, the 80 percent confidence interval
was felt to be a lower bound. Given that the stated purpose of the elicitation was to
5-2
-------
have the scientist estimate a range within which he was almost certain the actual effects
would fall, it seems reasonable to assume that the range represented at least an 80 per-
cent confidence interval. The weightings used (2:1, 3:2, and 5:*) range from the assump-
tion that, if a weight is placed on one side of the mid-point, the scientist feels that the
actual outcome would be twice as likely to fall on the weighted side, to a 5:4 weighting
where the scientist would be very uncertain regarding which side of the midpoint the
actual outcome will most likely fall, even though he is willing to assign a weight.
The results of the sensitivity analysis are shown in more detail in Appendix A; however,
it was found that the assumed confidence interval had a significant effect on the damage
estimates, but the assumed interpretation of the weighting had a minor impact. This is
an important finding since the interpretation of the weighting is the most uncertain
assumption. Further, this result justified the emphasis placed on obtaining accurate esti-
mates of the potential range of effects while settling for a simple weighting scheme to
indicate where within this range the most likely outcome might fall.*
Two examples will be presented to show in more detail how these confidence intervals
and weighting assumptions are translated into probability distributions.
Example 1
Assume that for fish species X and a pH of 6.
-------
This indicates that the scientist felt that at a pH of 6.4 the percent of Adirondack lakes
that could possibly, but no longer can, support fish populations of species X would fall in
the range bounded by 0 and 30 percent, and that the expert felt that the outcome was
more likely to fall between 0 to 15 percent than between 15 and 30 percent. Assuming a
90 percent confidence interval and a 2:1 weighting results in the cumulative distribution
shown in Figure 5-1. Given this cumulative distribution, the following holds true:*
Probability 0
Probability 15
Probability 30
damages
damages
damages
15 equals .60;
30 equals .30; and
100 equals .10.
Example 2
Assume that for fish species X and a pH of 5.9, the scientist estimated the range and
assigned a weight as shown below:
5.9
Lower
Bound
30
Weight
( )
Range
Mid Point
55
Weight
(X)
Upper
Bound
SO
Using the same definitions of these elicitatipn results as in Example 1, but assuming a
80 percent confidence interval with a 3:2 weighting results in the cumulative probability
distribution shown in Figure 5.2. Given this cumulative probability distribution, the fol-
lowing holds true:
30 equals .10
55 equals .32
80 equals .48
100 equals .10
o
0
o
0
Probability 0
Probability 30
Probability 55
Probability 80
damages
damages
damages
damages
* Changing the interpretation of the weights from 2:1 to 3:2 would change the probabil-
ities of damages falling into the three specified intervals from .60, .30, and .10 to .54,
.36, and .10. A 5:4 weighting results in probabilities of .50, .40, and .10 for each
interval. Recall that the estimate of the number of lakes damaged was found, in this
application to be relatively insensitive to the weighting interpretation.
5-4
-------
Figure 5-1
Cumulative Probability Distribution for Percent Reduction in Habitat*
for Fish Species X Resulting from a pH of 6.4 Using the
Data from Example 1
(assumes a 90 percent confidence interval and a 2:1 weighting)
Cumulative
Probability
.10
.30-
.60-
1.0-
-.90
.80
.7
-.60'
.50-
.40-
.30-
.20-
.10-
-0-
10
zo
JO
tfO
50
60
70
80
100
.60 prob-
ability of
damages fall-
ing within
this interval
.30 prob-
ability of
damages falling
within this
interval
Percent Damages
.10 probability
of damages falling
within this
interval
* Percent of lakes which could, but no longer can, support fish populations of species X.
5-5
-------
Figure 5-2
i
Cumulative Probability Distribution for Percent Reduction in Habitat*
lor Fish Species X Resulting from a pH of 5.9 Using the
Data from Example 2
(assumes a 80 percent confidence interval and a 3:2 weighting)
Cumulative
Probability
.10
.32-
.10-
1.0
.80-
.70-
.60
.50
.30-
.20'
.10-
0-
10 20 30 40 50 60
Percent Damages
70 80 90 100
* Percent of lakes which could possibly, but no longer can, support fish populations of
species X.
5-6
-------
5.1.2 Procedure for Estimating Damages from Acid Deposition
The preceding section presented the assumptions required to convert the range of esti-
mated damages and the weight assigned by the scientist at each pH into a cumulative
probability distribution for that pH. This section will outline the steps in the estimation
of damages from acid deposition based on the elicitations. A simple illustration of these
steps and a presentation of results from a test case sensitivity analysis of the confidence
interval and weighting assumptions can be found in Appendix A. Table 5.1 shows the
steps in the estimation of damages based on the elicitation conducted with the scientists.
Step One is the calculation of aggregate cumulative probability distributions of damages
for each species of fish at each pH level. For example, five elicitations were conducted
for self sustaining populations of brook trout. Table 5.2 shows the number of elicitations
conducted for each fish species. The five different self sustaining brook trout elicita-
tions were combined to form aggregate cumulative probability distributions of damages
at each pH by first calculating cumulative probability distributions for each scientist's
elicitation and then averaging the distributions. A computer program using simple
numeric techniques was constructed to perform these computations.
Step Two involves determining the present distribution of pH levels across Adirondack
lakes. The basic source of this data is the Adirondack Water Management System data
base maintained by the New York State Department of Environmental Conservation.
This data base and others used are described in Section 5.2.
Step Three is the presentation of a hypothetical case used to derive the change in the
distribution of pH values caused by increases or decreases in acid loading.
Step Four is the calculation of the incremental damages from the shift in the distribution
of pH values calculated in Step Three. This is a conditional probability calculation. It is
the probability of incremental damages given the current distribution of fish habitat in
the Adirondacks.
5-7
-------
Table 5pl
Steps in the Estimation of Damages
Based on EUcitatton Results
Step 1: Obtain aggregated cumulative probability distributions for each species of
i
fish at each pH by averaging the cumulative probability distributions for each
scientist.
Step 2: Obtain current distribution of pH values for Adirondack lakes.
Step 3: Derive the change in the distribution of pH values across Adirondack lakes
caused by increases in acid loading.
Step *: Calculate the incremental damages over present damages due to a downward
shift in the distribution of pH values across Adirondack lakes.
Output: Given a predicted change in the distribution of pH values across Adirondack
lakes, a cumulative probability function of the incremental number of lakes
expected to be damaged.
5-S
-------
Table 5-2
Number of Elicitations Performed for
Different Fish Species in the Project
Fish Species
Number of
Elicitations
Brook Trout
Self Sustaining Population
Stocked Population
Lake Trout
Self Sustaining Population
Stocked Population
Small Mouth Bass
Common Shiner
Fat Head Minnow
White Sucker
5
3
3
2
2
1
1
1
5-9
-------
The step four calculation of incremental damages requires that the correlation between
the different distributions of effects at different pH levels be specified. Since a regional
i
assessment spans lakes with different pH values^ estimates of the probability of different
levels of damages are derived from the combined probability distribution for all lakes.
To calculate this combined distribution, the distributions of damages to lakes segmented
by pH class must be convoluted (i.e., combined)!* To perform this step, it is necessary to
know whether these different probability distributions are independent or whether there
is some correlation across these distributions. Referring back to Figure 5-1, the figure
shows the probability distribution of damages if or a fish species in lakes with a pH of
6.4. A similar distribution could be graphed for lakes with a pH of 6.3. The assumption
of perfect correlation between these two distributions would mean that if actual
damages at a pH of 6.4 are high, then actual damages at a pH of 6.3 will also be high. In
terms of the cumulative probability distributions, this means that if the actual damage
outcome turns out to be that outcome associated with a .9 cumulative probability for the
6.4 pH distribution, the actual damage outcome for lakes with a pH of 6.3 will also be
that level of damages also associated with the .9 probability. The assumption of perfect
independence implies that finding high levels of actual damages in lakes with one pH does
not change the probability of high or low damages in lakes at neighboring pH values.
Instead of assuming perfect correlation or perfect independence, a compromise assump-
tion of partial correlation between distributions is possible. With partial correlation, a
high damage outcome at one pH level does not imply a high damage outcome at a neigh-
boring pH level, but it does imply that the probability of high damages at the neighboring
pH levels increases.
As a result, three assumptions are possible — perfect correlation across distributions,
some form of partial correlation, or perfect independence. The assumption of perfect
correlation results in the most uncertainy (i.e., the greatest dispersion in outcomes away
from the mean). This stems from the fact that, with perfect correlation, if damages at
one pH turn out to be very high, the damages at neighboring pH levels will also be very
high. Thus, extreme values are always coincident. This tends to increase the probability
that actual outcomes will fail in the tails of the joint distribution. However, it does not
change the upper and lower bounds of the distribution. The assumption of perfect in-
* Convolution is the term for multiplying prbbability distributions.
5-10
-------
dependence gives the tightest distribution, in terms of measures of central tendency
around the mean; and, therefore, results in the lowest amount of uncertainty. Thus, the
assumptions of perfect correlation and perfect independence can be used to bound the
uncertainty.
The appropriate dependence assumption was discussed with scientists during the pretest
phase of the project. The scientists indicated it reasonable to assume that if actual
damages at one pH level were in the low part of the distribution, then actual damages at
neighboring pH levels will also be in the low portion of their distributions. This implies
that underlying factors which make estimates of damages uncertain are correlated across
pH levels, and that if scientists learned that actual damages caused by pH 5.4 were in the
low part of the range, then they would revise the estimate at a pH of 4.8 downward to
account for this finding.* As a result, an assumption of perfect or partial correlation is
appropriate, and an assumption of perfect independence is inappropriate. For this study,
perfect correlation was assumed. This assumption provides an upper bound to the
uncertainty in the estimates and it is mathematically much easier to work with than
partial correlation. By assuming perfect correlation, it is possible to sum calculated
distributions of damages across different "clusters" of lakes to obtain total distributions
of damage to lakes. A "cluster" of lakes is a set of lakes that have the same current pH
level and the same ending pH level after the shift in pH levels.
Appendix A presents a simple example of these calculations and the results of a "Test
Case" sensitivity analysis of the weighting and confidence interval assumptions.
* In review drafts of this report, some scientists interpreted correlation as meaning that
during the elicitation, if the weight was placed on the high side of the midpoint at one
pH, then the weight would have to be placed on the high side of the midpoint at other pH
levels. This is incorrect. The assumption of perfect correlation implies nothing about
the placing of weights during the elicitation. The scientist place the weights in the
manner they feel most appropriate. Assumptions of correlation or independence
influence how the resulting distributions are combined.
5-11
-------
3.2 DESCRIPTION OF ADIRONDACK DATA USED IN THE ANALYSIS
In order to apply the probabilistic damage estimates discussed in Section 5.1, it was
necessary to obtain a comprehensive and current water chemistry and fish population
i
data file for the Adirondack Mountain Region. The primary data set used was provided
by the New York State Department of Environmental Conservation (DEC). Published and
unpublished data provided by a number of othelr investigators were also used. These in-
cluded data collected in 1975 by Dr. Carl Schofield (unpublished data); a characterization
of 623 Adirondack lakes provided by Dr. Schofield (unpublished data); and the Acidity
Status of Lakes in the Adirondack Region of 'New York in Relation to Fish Resources
(Pfeiffer and Festa, 1980) and its 1981 Update. This section will describe salient charac-
teristics of the DEC data as it was represented on the Adirondack Waters Management
System Tape, as well as the manner in which the data were interpreted and used in a
regional assessment of fish damages.
Data in the Adirondack Waters Data Management System covers 3506 ponded waters in
the Adirondack area. To sort the data into meaningful units for analysis, the variables in
Table 5-3 were reorganized into a working data base. The data base is not entirely com-
prehensive, as not every ponded water has a complete record. For example, while there
are a total of 3820 pH records within the data base, only 2409 of these are contained
within the Chem-Current File (the file containing most current chemistry survey data for
those waters which possess chemistry surveys). Similarly, while there are a total of 3820
alkalinity records for Adirondack ponds, only 1516 are contained within the Chem-
Current File.* The DEC is, however, augmenting the records with acidification field
monitoring programs and increased efforts to record field station data on the Adirondack
Waters Management System Tape. Since 1975, the DEC has taken pH and alkalinity
measurements in over 937 ponded waters (Pfeiffer and Festa, 198&, unpublished, 1982).
The pH measurements were obtained with a'pH meter under air-CO2 equilibrium condi-
tions. In the initially published results, 25 percent of the surveyed waters registered pH
readings below 5.0 (Pfeiffer and Festa, 1980),
* The total number of alkalinity records ate the sum of the Chem-Historic and Chem-
Current Files. There were a total of 509 'records for calcium and 560 for magnesium
within the Data Base.
5-12
-------
Table 5-3
Variables in Project Data Base
DATA BASE
Location and Status Records
1. Record type
2. Watershed
3. P#
-------
Table 5-3
Variables in Project, Data Base
(Continued)
Liming Records
32. Date Recorded
33. Month Limed
3*. Year Limed
\
Chemistry Parameter Records
(lor pH, alkalinity, NO2, NO3, SO^, Al, Ca, Mg)
35. Date of Samples
36. Depth (ft.)
37. Measurement
38. Method
39. Remarks
W. Source
41. Sample Type
<>2. Year
*3. Acidification Classification
-------
There is some controversy concerning the accuracy of many of the pH records contained
within the data base. This concern resulted from a 1979 survey conducted by the DEC's
Lake Acidification Studies Unit. The DEC possessed historical pH records for 138 of
these waters collected between 1930 and 193*. In order to compare current levels of
acidification with these historical records, the investigators recorded the current pH
values using both a pH meter and a Hellige comparitor. In reporting the results, Pfeiffer
and Festa (1980) showed that the 1979 colorimetric (Heilige comparitor) readings were
consistently higher than the pH meter determinations (see Figure 5-3). However, when
Schofield (1981) compared colorimetric measurements to meter pH measurements, he
concluded that the agreement between the two methods was much closer than stated by
Pfeiffer and Festa. Schofield concluded that the pH meter determinations reported in
Pfeiffer and Festa were between .5 and 1.0 pH units too low. Hence, the 396 pH deter-
minations for the ponded waters sampled in 1979 are probably between .5 and 1.0 pH
units in error. The investigators have been unable to determine what percentage of pH
listings recorded by the DEC and listed in the data base are in error.
In order to assess the potential bias in data base pH records, a regression developed by
Schofield (1976) relating pH to pCa was used by the investigators to determine the pH of
the 509 Adirondack waters for which both pH and calcium data were available. The
mean listed pH value for the sample was 6.1, while the mean pH value calculated using
calcium concentration data and the regression relating pH to pCa was 6.3. The standard
deviation of the pH values was .8*f, while the standard deviation of the pH values calcu-
lated from calcium concentrations was 1.12. While the difference between the two
values is important, the investigators didn't believe it to be large enough to justify not
using the pH records contained within the data base. The seasonal variability of the pH
within individual lakes caused by biologic production, snowmelt, and in some cases, fall
rainstorms, is probably equal to or greater than the variation expressed in the calculation
of the lake pH values. Figure 5-4 demonstrates the considerable variation in the pH of
one Adirondack lake's summer-ice free-surface pH values. In the case of the example,
the values range from a low of ^.9 to a high of 5.50 in late summer. Readers, however,
should be aware of this systematic bias in the characterization of pH values in the data
base used in this analysis.
5-15
-------
Figure 5-3
Comparison of Meter pH and Colorimetric (Hellige) pH Readings
for 100 Adirondack Lake Samples*
8
I
Q.
Ld
O
UJ
X
OBSERVED
REGRESSION
/^-THEORETICAL LINE
OF EQUAL READINGS
* 0.6639 X +2.534
* Taken from Pfeiffer and Festa, 1980.
5-16
-------
Figure
Representation of the Spatial and Temporal Variability Within
Dart Lake in the Adirondack Mountain Region*
Ice Cover
pH Xsopleth
A M
TIME (months)
* (Driscoll, 1982, unpublished data). Dart Lake is an acidic drainage lake located in the
Adirondack Region of New York State. It possesses a well-defined inlet and outlet.
5-17
-------
Records of fish populations within the Data Base reflect the presence or absence of fish
i
species within the particular ponded water surveyed. All of the fish species for which
damage elicitations were performed are included the Data Base. A data sort was per-
formed to establish an inventory of Adirondack lakes posessing both chemistry records
and presence-absence data for fish populations. Figure 5-5 illustrates the sort pro-
cedure. Since 1975, the DEC has obtained pH and alkalinity measurements for approxi-
mately 972 ponded waters in the Adirondack Region of New York.* In searching the
Data Base for chemistry records from 1960 to the present, a total of 1217 lakes were
i
recorded. Of these, 550 possessed fish records,. Records dated from 1960 to the present
were considered current or indicative of the present pH and fish population status. Of
the 1217 examined, 280 were surveyed prior to 1975. Assuming some change in the pH of
those 2SO lakes between 1960 and the present indicates that our data probably over-
estimates current pH and fish populations status within the region. There has been no
published estimate of the rate of acidification in the Adirondack Region between 1960
and the present, but W.D. Watt et al. (1979) estimated that for lakes above pH 5 in Nova
Scotia, Canada, pH decreased approximately p.5 pH from Gorham's measurements taken
in 1955 to their own made in 1977. Below pH 5, typical Adirondack waters possess in-
creasing buffer intensities as a result of the mobilization of aluminum, and the
» '
shift could be expected to be smaller. In searching the files for a pH record, the
algorithm also looked for calcium and alkalinity records. Of the 1217 lakes with pH
records, there were 151 which did not possess alkalinity records.
Step 2 calculated the pH range of the ponded water's pH value. Lakes were assigned one
out of 42 possible pH ranges. Increments of ,1 pH between 8.0. to below 4.0 were used
(see Table 5-4). After the assignment of a pH range, the program assigned the lake to
an elevation class — either below or equal to 1500 feet, between 1501 and 2000 feet, and
above 2000 feet. If there was no record, or if the elevation was blank, the elevation
class was set equal to "unknown". High elevation lakes in the Adirondack Mountain
Region have been noted to possess both depressed pH values and reduced fish populations
(Pfeiffer and Festa, 1980; Schofield, 1976b, 1981, 1982). In a 1975 survey of 214 lakes
above 2000 feet, Dr. Schofield found that approximately 50 percent of the surveyed lakes
* 937 measurements were recorded between 1975 and 1981. In 1982, an additional 35
lakes were surveyed.
5-18
-------
Figure 5-5
Sort Algorithm for Adirondack Ponded Waters
For each lake in the data base:
Read index record (1-7)
Is there a pH vaJue or a pH indicator (Ca, alkalinity) record later than 1960?
YES
NO
NO
Read pH record, choosing lab pH if available, and calculate pH range; or, if
no pH record is available, read Ca record and calculate; 0.1, if no Ca record
is available, read alkalinity and calculate pH.
Is there an elevation record later than 1960? If so, calculate elevation
class. If not, or if elevation is blank, set evaluation equal to "unknown".
Is there an alkalinity record later than 1960? If so, calcualte alkalinity
class. If not, or if alkalinity is blank ,set alkalinity class to "unknown".
Is there at least one fish species record later than 1960?
YES
Read all job species records with the same data as the most recent one and
set flags for each of the six species being looked for.
Add 1 to total for each fish species found (broken down by alkalinity, pH
range, and elevation class).
Add 1 to total for all fish species (broken down by alkalinity, pH range, and
elevation class).
Add 1 to total for all takes (broken down by alkalinity, pH range, and eleva-
tion class).
GO TO NEXT LAKE.
5-19
-------
Table
Distribution of Fish Habitat Represented1 in Data Files Sorted by pH Range*
pH Range
8.0
7.90-7.99
7.80-7.89
7.70-7.79
7.60-7.69
7.50-7.59
7.40-7.49
7.30-7J9
7.20-7.29
7.10-7.19
7.00-7.09
6.90-6.99
6.80-6.89
6.70-6.79
6.60-6.69
6.50-6.59
6.40-6.49
6.30-6.39
6.20-6.29
6.10-6.19
6.00-6.09
5.90-5.99
5.80-5.89
5.70-5.79
5.60-5.69
5.50-5.59
5.40-5.49
5.30-5.39
5.20-5.29
5.10-5.19
5.00-5.09
4.90-4.99
4.80-4.89
4.70-4.79
4.60-4.49
4.50-4.59
4.40-4.49
4.30-4.39
4.20-4.29
4.10-4.19
4.00-1.09
4.00
Brook
Trout
2
1
1
1
4
3
6
5
4
10
12
13
12
8
11
17
13
20
19
11
27
16
16
16
8
17
8
5
7
9
11
3
10
6
5
4
1
3
2
2
1
0
take
Trout
0
0
3
1
0
2
1
2
2
4
4
2
1
1
2
2
1
5
7
3
5
4
5
3
2
0
2
1
2
0
0
0
0
0
0
1
0
0
0
0
0
0
Small Mouth
Bass
7
0
5
5
3
7
1
2
5
5
3
3
6
4
2
8
5
5
7
3
8
0
2
2
1
1
2
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
White
Sucker
8
1
4
3
6
6
5
5
9
9
18
12
9
11
12
16
10
14
17
9
18
13
7
7
5
9
6
3
6
7
3
1
1
2
1
2
0
0
0
1
0
0
Fat Head
Minnow
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0-
0
0
0
0
0
0
0
0
0
0
Common
Shiner
0
0
0
0
0
0
1
0
0
0
0
1
3
1
1
3
2
1
1
1
1
4
0
1
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
All Fish
Species
15
2
6
8
9
13
8
8
12
16
26
22
22
IS
18
27
21
27
30
17
38
19
21
20
11
20
13
8
9
12
13
4
10
7
6
4
1
3
2
3
1
0
* Numbers equal number of lakes within each pH category which have fish records.
5-20
-------
had pH values below 5.0. Fish surveys conducted at the same time indicated that 82
percent of these acidic, high elevation lakes are devoid of fish. Of the 1217 lakes
examined in the subset of Adirondack ponded waters which had pH, alkalinity, or calcium
records, Table 5-5 shows that 1*1 were above 2000 feet. Of these, 99 or approximately
70 percent, possessed alkalinity readings less than 50 eq/L The same table indicates
that of the 550 lakes which had records of fish populations and chemical records, only 52
were known to be above 2000 feet.*
After assigning the lake to an elevation class, the program determined if there was an
alkalinity record entered in 1960 or later. If so, the lake was assigned to one of the
following alkalinity classes: above 200 yeq/1 alkalinity, between 100-200, 50-100, or
below 50 eq/1 alkalinity. If there was no alkalinity record, the alkalinity record for that
lake was set equal to unknown. 200 yeq/1 was established as the lower bound of the high-
est category because it has commonly been used as the boundary between "sensitive" and
"insensitive" waters prior to the initiation of anthropogenic acidification (Hendrey et al.,
1980b) According to Glass and Brydges (1981) biological effects due to acidification
begin when aquatic systems reach alkalinities of approximately 100 ueq/1. The range
below 100 yeq/1 was divided into two segments: one between 50 and 100 ueq/1 and one
below 50 ueq/1. Referring to Figure 5-6, it can be seen that there is a dramatic reduction
in lake pH as lakes develop alkalinity values less than 50yeq/l in the Adirondacks.
After assigning the lake to a pH, elevation, and alkalinity category, the program
recorded which, if any, of the 6 species of fish for which elicitations were performed
were present within the lake either in 1960 or later. A summary of this data, not includ-
ing elevation and alkalinity records, is shown in Table 5-4. Table 5-5 contains a summary
of the data excluding pH categories for the Adirondack Mountain Region.
Table 5-6 indicates that of the 350 lakes in the data base for which there were records of
brook trout samplings and chemistry records, 1* percent, or 49 lakes, possessed pH
values above 7; 43 percent, or 151 lakes, had pH values between 6.00 and 6.99; 32 per-
cent, or 113 lakes, had pH values between 5.0 and 5.99; and 37 lakes had pH values
* In both cases, however, there is a substantial fraction of lakes for which elevation
records do not exist; 493 and 201 respectively for the total population of lakes and for
the subset containing fish, respectively.
5-21
-------
Figure 5-i6
the Relationship Between pH and Alkalinity for Adirondack Lakes less than
610 Meters Elevation. 3gne 24-27,1975.*
Aieq/l
100
Alkalinity
. • \
-v.^,_i
*.-
"' pH "
A.5
7.0
* Taken from bchofield, 1976.
5-22
-------
Table 5-5
Results of Sort of Adirondack Lakes by ABcalinity, Elevation, and Fish Species
Alkalinity*
Elevation:
= 1500 ft.
1501 - 2000 ft.
2000 ft.
Unknown
Total
200
16
10
1
27
5*
Brook
100-200
4
21
7
24
56
Trout
50-100
7
24
8
21
60
50
12
43
30
37
122
Unknown
9
24
4
21
58
Total
48
122
50
130
350
Small Mouth Bass
Alkalinity*
Elevation
= 1500 ft.
1501 -2000 ft.
2000 ft.
Unknown
Total
200
IS
2
0
19
39
100-200
8
7
0
8
23
50-100
2
9
0
7
18
50
4
6
0
1
11
Unknown
2
2
0
8
12
Total
34
26
0
43
103
Fat Head Minnow
Alkalinity*
Elevation
= 1500 ft.
1501 - 2000 ft.
2000 ft.
Unknown
Total
Alkalinity*
Elevation
= 1500 ft.
1501 - 2000 ft.
2000 ft.
Unknown
Total
200
1
0
0
0
1
200
2
2
0
8
12
100-200
0
0
0
0
0
Lake
100-200
1
6
0
7
1*
50-100
0
0
0
0
0
Trout
50-100
3
5
2
5
15
50
0
0
0
0
0
50
0
15
3
3
21
Unknown
0
0
0
0
0
-
Unknown
0
3
0
3
6
Total
1
0
0
0
1
Total
6
31
5
26
68
Alkalinity values in ueq/1
5-23
-------
"iable
Results of Sort of Adirondack Lakes by AJcalinity, Elevation, and Fish Speck
Continued
1
Common Shiner
Alkalinity*
Elevation
c 1500 ft.
1501 -2000 ft.
2000 ft.
Unknown
Total
Alkalinity*
Elevation
= 1500 ft.
1501 - 2000 it.
2000 ft.
Unknown
Total
Alkalinity*
Elevation
= 1500 ft.
1501 - 2000 ft.
2000 ft.
Unknown
Total
Alkalinity*
Elevation
= 1500 ft.
1501 - 2000 ft.
2000 ft.
Unknown
Total
200
it
1
1
1
7
200
31
12
1
35
79
200
53
16
1
58
128
200
87
37
1
94
219
100-200
0
3
0
0
3
White
100-200
10
25
2
27
64
Total: All
100-200
17
35
7
42
101
Total:
100-200
44
83
13
85
225
50-100
0
3
2
1
6
Sucker
50-100
5
19
4
12
40
Fish Species
50-100
14
32
9
30
85
All Lakes
50-100
25
69
21
60
175
50
0
4
0
1
5
50
9
30
11
17
67
50
18
63
31
45
157
50
26
143
99
179
447
Unknown
0
0
0
1
1
Unknown
6
10
0
10
26
Unknown
13
28
4
34
79
Unknown
23
46
7
75
151
Total
4
11
3
4
22
Total
61
96
18
101
276
Total
115
174
52
209
550
Total
205
378
141
493
1217
* Alkalinity values inyeq/1
5-24
-------
between 4,0 and 4.99. There were no records of brook trout lakes below pH 4.O.* The
total number of lakes included in our data sort of the Data Base is 1217, or approxi-
mately 35 percent of the total 3506 ponded waters in the Adirondack region. Of the
2,759 lakes which comprise the Adirondack Ecological Zone, only 937, or 34 percent,
have been entered into the data base. According to the DEC (1982 unpublished data), the
average size of the ponded waters in the Adirondack Ecological Zone is 89.3 acres, while
the average size of the waters in the 937 lakes sub-sample is 221.1 acres. The DEC cal-
culated the remaining 1822 unsurveyed ponds to have an average size of approximately
21.5 acres. The majority of these smaller ponds are found at high elevation sites on
generally more sensitive soils, and, consequently, can be expected to have, on average,
lower pH values than those ponded waters contained within the sample constructed for
this analysis.
ERC implemented a separate sort algorithm in order to differentiate between those
waters containing stocked and naturally reproducing populations of brook trout and lake
trout. The results of that sort indicated that there were a total of 77 lakes with stocking
records and recorded brook trout populations after 1960. The sort indicated that there
were 273 lakes which possessed apparently naturally reproducing brook trout popula-
*•
tions.* For lake trout, the sort indicated only one lake which was both stocked and
* For all lakes with chemistry values reported later than 1960, only 6 had pH values less
than 4.0.
* Pfeiffer (1979) states that there are 407 ponds within the Adirondack Ecological Zone
open to public fishing. Of these, he records that approximately 368 ponds, representing
some 12,694 acres, receive annual stocking with fall fingerling wild x domestic hybrids.
39 of the 407 ponds depend upon natural reproduction. 100 additional public brook trout
ponds are said to have lost their capacity to support brook trout. Pfeiffer predicts that
"about 25 percent of the existing viable acreage will succumb to acidifiction by 1992.
However, in a recent personal communication, Pfeiffer indicated that there had been a
slight improvement in the condition of brook trout ponds since the beginning of
monitoring in 1965. He indicated that, while the pH of ponded waters in central and
western New York had declined over the period, that the pH had increased slightly in
eastern and southern New York.
5-25
-------
possessed sampling records of lake trout. There were 67 lakes which apparently con-
i
tained self •sustaining populations of lake trout.*,
Consulting Region 5 stocking records for the period between January 1 and December 31,
1981, 287 ponds were found to be stocked with brook trout. Pfeiffer (1979) lists 368
ponds within the Adirondack Ecological Zone as being stocked with brook trout. The
DEC confirmed that at present, many stocking records have not been entered on the data
!
base, and that the data base does not accurately represent the stocking history of
Adirondack Jakes. As a consequence of the incompleteness of the files' current stock
records, the investigators did not, as initially planned, apply the damage estimations for
the stocked and unstocked populations of lak± trout to stocked or non-stocked, lakes.
i
This analysis, however, will be possible when ,the DEC enters the stocking data in the
tape. Damages to stocked and unstocked habitats were estimated.
The Data Base was further analyzed into the six drainage bases which jointly constitute
the Adirondack Region. The results of this sort are represented in Table 5-6. Elevations
apparently strongly influences alkalinity. 70 percent of the lakes above 2000 feet in the
I
data base have alkalinity values less than 50 yeq/1. 86, 85, and 68 percent of the lakes
above 2000 feet in the Raquette, Mohawk and Hudson, and Oswesatchie/Black River
Basins have recorded alkalinity values of less than 50 yeq/1. Of the lakes between 1501
feet and 2000 feet, 38 percent had alkalinity values less than 50 yeq/1. There was a sub-
stantial increase in lakes which possessed alkalinity values greater than 100 yeq/1. 42
percent of the lakes below 1500 feet have alkalinity records greater than 200 yeq/1. This
compares with 1 percent of the lakes above 2000 feet.
Table 5-7 represents the same terrain, according to the calcium level of Adirondack
lakes analyzed by elevation and major drainage basin. While direct comparison between
i
Tables 5-6 and 5-7 is difficult, the general trend of increasing alkalinity values and cal-
cium values with decreasing elevation is evident in both.
* Pfeiffer (1979) indicates that within the Adirondack Ecological Zone there are 61
public lakes with Lake Trout populations. 'Of these, 37 are maintained through natural
reproduction, while 2* receive stocking. There are an additional 33 private lake trout
waters, which Pfeiffer assumes have self-!sustaining populations. 38 lakes which once
contained lake trout are now devoid of this species.
3-26
-------
£
"o .2,
11
^3 C
| I
£ 1
4Q ><
U
oo — < o
**> ON — «
CM
(N
o o o o o O
— « — -1 — « _M \0
OOOOOO
— CMVD
<^\ *f\ r^ (N c*^ rs CN ~* ON CNI fsi
-^ CM r^
-------
c
ft
1
"3
3
1
1$
v £
£ V)
«
1 •
C ao
O 5i
-1 j
i ^
*" * 15
u Q •-
-a * 3
i2 §
H g -o
2 5
1/5 s
So
jj
M *j
4* 10
s Ji
«3 UJ
1 £
£ •
H
**
£
*•?
17
•M
V
t_-
V
CL
3
£
j£ C
> Q
If
°5
S1
JJ
•6*
«1
|S
£
o
5
4->
0)
3
o-
IT9
a:
c
•O
X
a;
4s S
£ *
s
(Q
C
3
"a.
5
u
£_>
•3 <_i
.«-«-.
3*
I
+<
«3
4)
UJ
*n»«n -^ oo — < /vi00—1
m^-CM — CM vo CM^^,
O 00 _. O CM IS. m Si! _ 0 CM JQ
^0 — 25
-H -M W <«> «««N.002 CMON^
- -ONgS 0«^2K
- « m oo CM^OSO °rM-£
iv. j*. ^ •* _. NO K
-00-. J^^;-»o ZTocM^
'
•^ M5 - ^ ^^S^S OOS§^
JT** ^ ^ .^ pg ^,^
«^ "J3 "> *TJ "> *o
• 4n4 • 4nA * ^*
•0, CM «A o tricvirio *} «r\ rs. tr\ ««\ CM
OO .» «• IS, CM CM
CM IV 00 00 \O
— « CM
C^«^'«A
CM e^ <»•
^p
er «• r^ CM o\ o CM
'*' ITS 00 — < «• 00 CA
z -<-<
4- OO OO
— • & f\
&• « oo O ON fs. oo
*— ' «*> r^ ITS -^ so -^
2 —•»«*> ON —
«— « <~»
*-vfM h*
— i CM r««
C^ CM CM OO CM ON OO
"^ rv. 10 <*\ «A -H
Z 04 f«>
,-^'R'R
NO — r^.
5^
CM \X> f^ — * C^\ <»N
^^ «»N ON u-i oo ON m
"7 -3- i/"v —4
rfLj -a- -»
^o*ot
-< —4 tN.
^. ON «• ON CM CM CM
*-*] <«N »^> \O -» .* —
Zj CM r^»
^^&
«r\ «r\ ON
^J — < — tv ON CM 00
^-1 CM CM O -SJ- 00 -^
Z| •» *
ffi
*2
o
« - H
^.7^5 -^ o
— ,A x
v5 xi o. g
_ o. a ~ .2
S - ~ v *-*
O »T-^-«- rt
? IS- 8 5
S E.|£ -s
•« j= ±; * ^
=5 u « w £
2 £ffl £
< H 3-
JU
CO
u
'o
t/>
JJ
«J| ?§S
^ « S-2
S*W Vrf ^^
«A 0 JJ-"
£ ^CM^ t^?
.2 --ACM* "Jig
X - £ g- o
rt ocx:
w i^iX
*
*
* + -*
, 5-28
-------
Furthermore, there are similarities between the calcium and alkalinity values within the
different drainage basins; for example, compare the records for the Mohawk Hudson
drainage above 2,000 feet in both tables.
5.3 CALCULATION OF INCREMENTAL DAMAGES CAUSED BY CHANGES IN FRESH
WATER CHEMISTRY
The first step in the determination of incremental damages or benefits to fish popula-
tions resulting from changes in fresh water chemistry was to propose a simple model of
system response for Adirondack lake waters assuming acid or base additions. The base
additions can be taken to represent either reductions in the input of acidic deposition; or,
presuming no change in the level of acidification, a mitigation strategy of base addi-
tion. The change in lake pH was calculated relative to a fixed pH value, when 50 and
100 yeq/1 of base and 50 and 100 yeq/I of acid were added to a fresh water system with
the following characteristics;*
1. open to the atmosphere
2. in equilibrium with gibbsite solid phase
3. ionic strength is approximately 0.001M
4. DOC (dissolved organic carbon) equal to 5 mg/1
5. sulfate concentration equal to 6 mg/1 as SO^~^
6. F~ concentration equal to .1 mg/1.
7. temperature fixed at 15 degrees centigrade
While the model is useful for the purposes of this project, it is a very unrefined estima-
tion of lake water chemistry and pH change within the Adirondack mountains. It asumes
that all Adirondack lakes have the chemistry outlined above, provides no time rate of
change for the aquatic chemistry, and makes no attempt to relate the change in aquatic
chemistry to changes in the rates or concentrations of acidic deposition. In order to
calculate the change in pH, two computer runs were executed. The first fixed or esta-
blished the pH at values from 4 to 8 in 0.5 pH increments. The model then calculated the
amount of acid or base required to shift and fix the pH values at a new steady state. The
* See Figure 5-7. The model calculated the extent of change in the pH of lakes relative
to the fixed pH values along the bottom axis of the graph.
5-29
-------
second run calculated a theoretical titration based upon the proton demand calculated in
the first run. The protons were added with the1 intention of adding ±50 and ±100 yeq/1
more than was required to fix the pH at the predetermined values. The results of the
runs are shown in Figure 5-7. 100 yeq/1 acid additions were considered the limiting upper
bound, and are approximately equal to the concentration of anthropogenic acid in the
Adirondacks if Henriksen's predictions are true| (Henriksen, 1982). The 100 yeq/l base
addition, would shift the distribution of lakes to approximately the original pH distribu-
tion found within ponded Adirondack waters prior to anthropogenic acidification. The
50 yeq/l increment acid and base additions were calculated in order to determine the
i
effects of smaller shifts in fresh water acidification.
Changes in lake pH represented in Table 5-8 were calculated using Figure 5-7. Notice
that the buffer intensity of the system seems weakest in the pH values ranging from
approximately pH 7 to pH 6.* As expected, there was a decreasing shift at either end of
the pH range. Lakes below 4.5 are shown not shifting under any scenario of base or acid
addition. At pH 4,5 and below, the model results were highly variable due to the in-
creased aluminum buffering in the system. However, lakes with pH records equal to or
below 4.5 are very unlikely to support fish populations in the dilute, oligotrophic water
characteristic of the Adirondack region.**
* Schofield (1976b) noted that in his survey there were relatively few lakes with pH
values between 5.5 and 6.0. He attributed this to the hypothesis that, for Adirondack
waters, this was the region of minimal buffering intensity. The model runs depicted in
Figure 5-7 indicate, however, that there is less buffer intensity between pH 7 and pH 6.
The model runs indicate that the Adirondack lakes are strongly buffered by the aluminum
system rather than the carbonate system. In a carbonate buffered system, the maximum
buffering intensity would occur at pH 6.3 as opposed to below 4.9. As discussed pre-
viously, while the aluminum buffering systeijn prevents or slows down further lake water
acidification, it tends to stabilize the systjem at very toxic conditions due to the re-
sultant combinations of pH and aluminum levels.
** DEC files indicate that 550 lakes contain at least one of the six species investigated
during this project. There are 10 records of fish populations in water with pH less than
4.5. Ninety percent of those low pH lakes contained brook trout, one contained white
suckers.
5^30
-------
Figure 5-7
Calculated pH Shift Assuming Different Levels of Acid Inputs*
1-0 _
.0 -
* To calculate the shift in pH of a lake given that lakes pH, select acid or base addition
line of interest and read change in pH axis. For example, for a lake with a pH of 6.5 and
100yeq/l acid addition, the change in pH would be approximately -2 pH units. The lake
would shift from a pH of 6.5 to a pH of 4.5.
5-31
-------
Table 5-8
Chemical Shift in Adirondack Lakes Given a 100 ueq/I Acid Addition
Current pH
8.0
7.9
7.8
7.7
7.6
7.5
7.*
7.3
7.2
7.1
7.0
6.9
6.8
6.7
6.6
6.5
6.*
6.3
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Change^ in pH
'0
.062
.083
.125
.187
.250
J75
.500
.730
.950
K30
1.58
1.78
1.87
1.95
1.87
1.81
1.75
1.62
1.50
1.42
1.31
1.19
1.06
.950
.875
.800
.700
.625
.500
.437
.375
.312
.220
.125
0
0
0
0
0
0
NewpH
8.0
7.8
7.7
7.6
7.4
7.2
7.0
7.2
6.5
6.1
5.7
5.3
5.0
4.8
4.6
4.6
4.6
4.5
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.5
4.5
4.5
4.5
4.5
4.4
4.3
4.2
4.1
4.0
5-32
-------
5.4 ESTIMATES OF DAMAGES/EFFECTS DUE TO CHANGES IN ACIDIFICATION
This section presents the results of the calculations described in Section 5.1. Both in-
cremental and decrementai f>H changes ior the six species of fish considered during the
eiicitation will be reported. Results for brook trout and lake trout are expressed both for
stocked and self-sustaining populations. The computations of incremental damage were
made using the following assumed parameters:
1. the weighting assignment was fixed at 3:2;
2. confidence intervals were set at both 80 and 90 percent.
Three of the four lake pH shifts calculated using the model described in Section 5.3.1.
were used in the final analysis:
1. 100 yeq/1 acid addition
2. 50 yeq/1 acid addition
3. 50 yeq/1 base addition
The first section presents the results for assumed increases in acid input of 50 yeq/1 and
100 yeq/1 for each cluster of lakes within a given pH range. The second section presents
the results of a base addition scenario which would result in the increase in pH of the
cluster of lakes. A slightly different procedure was used to examine the effects of base
additions to the lakes. The third section presents a comparison of the response of the
different fish species habitats and an evaluation of the results using two different cal-
culation procedures.
5.*.I Increase in Acidification - Predicted Effects Upon Adirondack Fish Habitats
Two scenarios postulating increases in lake water acidification were examined: (1) a 50
yeq/1 acid addition, and (2) a 100 ueq/1 addition. Both of these scenarios represent
rather large increases in acid deposition. Although estimates vary, Henriksen (1982)
estimates that approximately 100 yeq/1 of acidification can presently be attributed to
anthropogenic sources in the Adirondack region. Based upon this estimate, the two
scenarios represent roughly a 50 percent and a 100 percent increase over present levels
of acidification from antropogenic sources.
5-33
-------
Table 5-9
Chemical Shift in Adirondack Lakes Given a 50 peq/1 Acid Addition
Z Current pH
8.0
7.9
7.8
7.7
7.6
7.5
7.*
7.3
7.2
7.1
7.0
6.9
6.8
6.7
6.6
6.5
6.4
6J
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Change in pH
0
0
0
0
.062
.125
.187
.333
.437
.625
.75
.875
.975
1.125
1.187
1.187
1.083
KO
.937
.916
.875
.833
.275
.75
.666
.6
.5
•f75
.437
.375
.333
.25
.225
.125
.083
0
0
0
0
0
0
New pH
8.0
7.9
7.8
7.7
7.5
7.4
7.2
7.0
6.8
6.2
6.2
6.0
5.8
5.6
5.4
5.3
5.3
5.3
5.3
5.2
5.1
5.1
5.0
4.9
4.9
4.9
4.9
4.8
4.8
4.7
4.7
4.6
4.6
4.6
4.5
4.5
4.4
4.3
4.2
4.1
4.0
5-34
-------
Table 5-10
Chemical Shift in Adirondack Lakes Given a 100 ueq/1 Acid Addition
Current pH
8.0
7.9
7.8
7.7
7.6
7.5
7.4
7.3
7.2
7.1
7.0
6.9
6.8
6.7
6.6
6.5
6 A
6.3
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
5.1
5.0
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
Change in pH
0
.062
.083
.125
.187
.250
.375
.500
.730
.950
1.30
1.58
1.78
1.87
1.95
1.87
1.81
1.75
i.62
1.50
1.42
1.3 1
1.19
1.06
.950
.875
.800
.700
.625
.500
.437
.375
.312
.220
.125
0
0
0
0
0
0
New pH
8.0
7.8
7.7
7.6
7.4
7.2
7.0
7.2
6.5
6.1
5.7
5.3
5.0
4.8
4.6
4.6
4.6
4.5
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.6
4.5
4.5
4.5
4.5
4.5
4.4
4.3
4.2
4.1
4.0
5-35
-------
Tables 5-11 and 5-12 list the results for brook trout populations given the aggregate
cumulative probability distributions which resulted from the elicitations, the distribution
of Adirondack ponded waters containing brook trout records since 1960, and the dec-
remental shifts in lake pH values resulting fronfl the different acid addition runs.* The
results expressed in Table 5-11 assume that all 350 lakes examined contained populations
of stocked brook trout. Column one represents the cumulative probability of damages
i
occurring to the set of 350 lakes currently supporting brook trout populations for the
shift in pH values resulting from a 100 ueq/1 addition of acid, assuming a 3:2 weighting
and a SO percent confidence interval. The damages range from a lower bound of 4,7
lakes, or approximately 1.34 percent of the total sample, to an upper bound of 273.4 or
73.11 percent of the total sample. The median value is 166.5 lakes. Because the results
represent a cumulative probability distribution, the median represents the number of
lakes which divides the range of probabilities of damages into two equally likely seg-
ments. The expected value of damages is 1^9.5 lakes, or 45.57 percent of the total
sample. The expected value of damage is the mean of the probability distribution. If the
mean, as in this case, is less than the median value, the marginal probability distribution
representing the shift would be skewed towards the lower end of the probability range.
Conversely, if the mean is greater than the median, the expected changes in the marginal
distribution would be skewed towards the upper half of the range. The results expressed
in column one indicate that there is a 10 percent chance that damages occurring to the
sample of brook trout lakes would be equal td or less than 34.0 lakes. Similarly, .95 in-
dicates that there is a 95 percent probability that damages will be equal to or Jess than
26S.S Jakes.
Another way to use the information contained in the cumulative probability distribution
is to calculate the probability that damages,from the shift in distribution of pHs across
Adirondack lakes exceed a given level. Since the cumulative probability indicates the
likelihood that damages are equal to or less than a given level, one minus the cumulative
probability will equal the likelihood that damages exceed a given level. Column 1
(Table 5-11 — 80% confidence interval — indicates that with an increase of 100 yeq/1 of
acid, the probability that 25 percent of the lakes or 87.5 lakes, which currently support
stocked brook trout populations, would no longer be capable of sustaining stocked popuia
* Tables 5-13 through 5-18 present the results of the acid additions to the remaining fish
populations in the Adirondack mountains.
5-36
-------
1 s
«
"3
v
u
ai
2
S-l
O
U
uJ &
* j
lUI
2 « ^ -
a
T3 ?^
"O £ UJ
i * <*
- s I
1*2
5?
g— ^ ^.
•o C
S "* Si
V)
V
a.
x
UJ
to
4)
I
V
2
1
U
U
a-
V
o
o
«9
L.
4)
O U
OS C
V
§
u
(0
w - ^f
Sit
C.
O
U
|o
3 L.
fs, 00 (^ 00- « IN. CM 00
_^ _ tsj CM r^ ro
• • • •
3.90 SO O
'* S
«-* rv
CS
0 NO
- * CO 00 -
- - VD
oo
SO
"ffl
T3
4>
O.
X
LJ
5-37
-------
1
>
"8
«s
^L
UJ
i
¥
c
"5
7
J§
CM *"
fable 5-1
ributions
^ 5
£
•a
1
1
i-
1
a
w
•*•
0
M
Si
a
y
«—
g
J**
IM
•Q
5 •
< —
*i P"
fl V
4«* ^
Eo
^ °
i i
*Q re
J "er
fi 1
M »
I8
J>
±a
J^
o
4-»
V
»
1
w
*M
o
0
* "A
1"
IA
4>
0.
X
UJ
£
ra
V3
4-*
1
41
Di
^
f
^M
I
o
ft
f
3-
§
13
£
^ 4>
ON E
* §
XJ
HI
(J
"3
>
L.
4)
+••
JC
# V
O U
00 C
4i
2
«M
c
o
(J
fc
<-»
#7
O V
ON C
4J
S
<4H
g
U
~a
V
|HH4
# 4)
§ ^
U
T3
CM(ni»«n>oNOr^r^.oooNo— <-a--*r^\f>r-N. CO ^J- — «r»tvoOONOCMfnf^*«'%«r\Is»
O--ONOi-C^<*NOOvOU^ONr^OOCM\OOO
rvrx.t«^ON«riocMfMc*\j3'«Noo — "Noo— «sOc^ONONO
vts.oooOONON^
C3
!>!
ON
f4
•
•a-
ON
m
f^
vO
0^
0
VO
0>
13
>
•o
V
4-f
y
Q.
X
U
5-38
-------
I
-* o
a
(U
u
cr
4>
«J
i_
U
4)
TJ
•««*
•M
O
U
CJ
O U
CT\ C
f
g
U
9< v
32
4J
^
•«
g
U
«* >
> **
§2
oo
3 —. 00
• • •
» o rv
0)
«
Q.
£
5-39
-------
id
"o
— $
> s
11
ring the Expcx
Acidification
1 *
ributtons 5
Lake Troul
J2 oo
0 S
li
1 *
2 a
Q. $
> &
*5 u
- a?
i I
a ^
Mri
0
1
8
•o
s
1
««
«
I_
.c
0)
u
-cent Confide
)
£ |
"S *
C (jj
s 1
^H V
*•> •— I
* S
l_
CL
X
UJ
f
L«
ra
<2
"i
<±
u
j£
-MI
"a-
=1
o
10
cr
o
o
13
£
o° 4>
S c
4->
o 0
o\ c
ft)
o
u
13
V
J:
O > « o rv ^ ir\ oo vo «h 4- vO N« oo ^ ts oo oo
vo ^ — is. * — <«J>A oo (NjQON ^ rv oo o .-*« o m
•
^ ^ ^ Tt «n <*} •*. °? *. -. °* ^ «n -r0'. ^ *. ^ ^ ^. ^ ^
CM "^
* i^ C) "^ CZ^ *O CJ *O O *f"\ C3 "^ C3 *O C^ *^ CD *^ G> *^ * O
— ' _ _• <*> <* .!*• tfM<-> V0 V0 FN. r». 00 00 ON ON °£ _g
A3
>
•o
0)
u
Q.
X
UJ
-------
J
<0
3**
i
a-
4)
I
. 1
able 5-1:
ributions
*• «
*•*
a
1
2
p
cu
>
"£
3
1
0
reases
£
|
«
1
£
8
1/5
g
W
I5
a
a
N-t
0
_J
1
•o
A
o
o
in
*<3
4)
C
4)
•c
C
o
U
c.
a .c
o £
n3
TD X
C UJ
* <«
* O
u "-'
v>
in
V
a.
X
4>
L«
cd
(A
"i
V
f±
4)
H
««i
cr
o
>f\
cr
o
0.
o
o
13
JS
# OJ
o .0
ON C
V
a
"c
o
U
"rt
V
£
# «j
O U
00 C
41
2
<*i
§
1
1
90
Confidenc
1
2
c
S
*c
o
U
«ik.
•*. ^^
Cumulat
Probabil
«^ONONONr-«.rNMoo«rkCNc«>ivrx«ooooooNc'N«o.»v\
<\O ON CM WN/OO — » <••* £»• \C K O'-" (NC*N.-*>AVOOOON«OO
_ fx. ON — < «*\.fMis.vo.a-rvf>ONO— • CM <*> I** fM ON r^ rM
-SSSSSSSSSSSSSSSSS-SS
f^ oO ^ Q \o V-N sO "^ V^ OO <7^ *& ff^ "^ **^ ^* ON f^S vD O** f*1*
„„„„„„,„„„„
-NDON^rNOO^ONa-OOONNO^ _-^. ts. _ d- 00 ^ ON
(N *^
vc
trs
ON
9J
a.
x
UJ
-------
NC
1
v\
JU
(0
2
£
?
1
w
fr
i2
J
**
c
••«
1
*«
1
Q*—
^
3
*9
jg
8
§2
1
3
E
«M
O
M
SI
(Q
J
s
'43
U
0)
1
U
c
— : "
^it ^- "O
p^ a; aj
ft CL _C
O P
^ ^ a
T3 X
™ C t«J
et
"5- *° 1o
| f "J
^^ ^ ^^
O T3 ^1
^
CL
X
LU
^
4-1
n
a^
H
MMH
*&•
3.
O
o-
V
o
0
"3
I-
ftj
4->
^7
o u
c^> c
2
"c
o
U
_
>
aJ
?
# 4)
o o
."2
*M
g
U
>
&>
*->
jr
c^ O
° £
TJ
«*H
g
U
_^
to
>
+*
^ 43
~ £
c
o
U
li
If
OOSO^^^VOrvON-(N*^ l^>- 00 r%> v«l i^\ ^IHI i^w ^^ ^k *>M •cfr* f^ i^s oo oo oo vo cr\ H^
^^ fl^ flT\ f^ ^£) C^ 'C^ ^D f^ if^ 1^^ O ^^^ ^^J ^5 *LJ* ""^ t^ O^ f\ ^3 ^^
»s.^.ooo^oo<,c,oo^c,--^e,^ONON^ so
-S^S>§!CSooSSoo22SS^^~2vo §
or^.«^\-a-ON<^o^-
-------
1
<0
>
1
w
it
*""
1
^
bfl
.£
o
, *
•* M
laoie :>-
rtribution
«
S
IT
s ProbabiU
*•*
*J
J2
£
a
t!
O
1
s
»mt
+J
a
u
**~<
**4
•o
S
«••
10
»iM
6
1
2
9)
%
«f-t
S
0
-i
t
o
(A
1
5.
§
V)
~<3
>
a>
*•»
tc
Ui
£
a.
X
UJ
4>
(0
•
u
*->
£
^ V
§ -^
4>
2
*•«
"a
L.
a;
•M
•*•!
# 0)
g ^
4>
^
*»H
§
U
W >s
> 4->
"*•• *EE
^2j5
3 «
£•§
at!
D.
««Nr->\oeOOCMm<'NtvOr\ir\ir\\orx.rv. <»\
\O\O<^\ONOOOCN-*\OOO-<~t-rvrMv.D— 'VOC-JOOrMi^N VO
•^ <»NNCPO---^CN|C-ICN4-lf-^«^f^*..S-ir\ir\\ONCh-.rt-.. tfNu-\tr\\os£vovov-osorN.rvrvis.ooooooooaNaNO h*
ON ON ON ON ON ON ON ON ON ON ON ON ON ON ON ON ON ON ON ON- O ON
— rO«AOu>O«'NO>'NO«^O>rvO>'NOM ^\i fsi m ff\ *+ ^K VN v\ iA vft i^. (s^. ^rt ^A nv /TV 0v ^
— ^ ^^ ^^» \^i ^^ i f^ ^y ^r "• i *• i vw >w i^* r^ ^w ww w^ w> j^ j
CTv —
-<3
V
o.
X
at
-------
J
^
1
1
1
£i
•S
eo *"
re «
t;
£
£
1
1
o.
v
.2
5
J
**
"8
3
s
1
1
to
,35
J*J
1
IS
2
•a
a.
2
1
i
*o
t
^1
1
o
i
_^
§
~
13
t
C
^*
U
VH
C
O
U
C
(U
U
1 f
o .£
"O (0
C X
to nj
O a,
PO _2
•£ j
T3
O
a.
X
LU
L.
W5
*-•
1
0)
a:
a>
H-
••^
o-
V
o
o-
U
o
o
"53
i-
£
Hi
C
o
O
_
1
£
<£ v
sl
I
"53
V*
§ C
2
•M
g
O
"53
^
*i
jC
Sy
C
U
^o
g
U
$£
"="2
E^J
3 £
cn^r^crM^ * a- * *»- ** ON <^
O
O— <(M(NCNIfM(M4*><«>rr><*>C<^(<^mr^4--i3>-d--4-IV^ <^l
0 0
-^*^»oo«,ooos^oxos^avas«7.avoooo oo
CM "^
_Jif\O«OOV>O>^\O"^O"'>O''NO«^O"JO^0Q W
m
T3
a.
X
UJ
-------
tions is equal to one minus .30 or 70 percent. Using this interpretation, the following
hold:
o probability that the number of lakes damaged will exceed one quarter of
the lakes (i.e., 87 lakes or more) is approximately .70.
o probability that the number of lakes damaged will exceed one half of
the lakes (i.e., 175 lakes or more) is approximately .45.
o probability that the number of lakes damaged will exceed three quarters
of the lakes (i.e., 262 lakes or more) is .10.
Column two represents the same calculation with a 90 percent rather than an 80 percent
confidence interval. As expected, the range has narrowed somewhat, especially at the
lower probability levels. The change in the expected value shifted from 159.5 to 162.5
lakes. This represents a change of .71 percent in the sample of lakes capable of sustain-
ing fish populations. The change in the median value is equal to .7* percent.
Columns three and four represent acid additions of 50 ueq/1. Column three is calculated
with an 80 percent confidence interval and column four with a 90 percent interval. As
can be seen the expected value of damage in both columns three and four is greater than
half the expected value of damages in columns one and two. This result was expected
based upon the aquatic chemistry model output shown in Figure 5-7. The change in the
pH level of the "typical" Adirondack lakes is not linear with respect to changes in the
amount of acid or base added to the system.
Table 5-19 compares the means, or expected values, for shifts occurring to stocked and
unstocked populations of brook trout given different acid additions. Comparing the
results expressed at the 90 percent confidence interval reveals that if the current distri-
bution of Adirondack lakes were to be further acidified with 100 yeq/1 of acid, the
expected value of damages would be 46 percent of the stocked lakes and 48 percent of
5-45
-------
r
Table 5-19
Comparison of the Expected Values lor }00 and 50peq/l Acid Additions
to Stocked and Self Sustaining Brook Trout Populations
+100peq/l Acid
Stocked Population
Confidence Intervals
80% 90%
Self Sustaining Population
Confidence Intervals
80% 90%
-159.5
-162.5
-160.9
167.3
+50 yeq/i Acid
Stocked Population
Confidence Intervals
80% 90%
Self Sustaining Population
Confidence Intervals
80% 90%
-76.8
-70.6
-9* .2
-97.0
5-46
-------
the self-sustaining lakes.* The stocked iai-.ss are considerably less sensitive in the
50 vjeq/1 acid addition, with expected losses of approximately 20 percent of 70 lakes as
compared with the approximate 28 percent reduction in self-supporting populations.
Presumably, the 8 percent difference can be attributed to the smaller pH shift in the 50
peq/1 addition.
The 100 ueq/l addition the water chemistry model predicted that lakes with pH values
equal to or less than 6.6 would shift to pH 4.6 or less (see Table 5-8). On the other hand,
the 50 jjeq/1 shift allowed lakes with initial pH values of 5.0 or greater to shift to pH
levels equal to or greater than pH 4.7 (Table 5-9). An examination of the lower bounds
and upper bounds from the elicitation for self-sustaining brook trout populations at pH
4.7 showed the average elicited lower bound to be a 40 percent loss of lakes and the
average upper bound to be 90 percent loss. Average lower and upper bounds for the
stocked populations were 37.7 percent to 76.3 percent at pH 4.7. The stocked population
at pH 4.5, however, was quite sensitive. The average lower bound was 51.7 percent,
while the upper bound was 90 percent. The average upper and lower bounds for the self-
sustaining population at 4.5 was almost identical, ranging from a lower bound of 50
percent to an upper bound of 93 percent. These results are similar to those found in the
laboratory analyses conducted by Baker and Schofield (Baker, 1981). In that study, Dr.
Baker determined that in order to prevent measurable reductions in the survival of early
life history stages of brook trout, the pH of the typical Adirondack surface water should
be greater than or equal to 4.8. This implies that as the pH declines, the difference in
sensitivity between stocked and unstocked lakes declined. At some pH levels, all fish are
affected.
Table 5-20 presents the expected values for changes occurring to the remaining five
species of fish considered during the project. Comparing the sensitivities of different
fish habitats considered in the analysis (50 yeq/1 acidification and 90 percent confidence
intervals) a ranking of fish population habitat sensitivities in Adirondack lakes can be
generated. From least sensitive to most sensitive, they are:
* Because of incompleteness in the stocking records entered in the DEC data base,
stocked populations and unstocked populations were differentiated by assuming that all
of the lakes are either stocked or self-sustaining. Damage calculations were then
performed on the entire set of lakes for which the investigators possessed both chemistry
and any population data concerning brook trout and lake trout.
5-47
-------
Table 5-20
Expected Value of Change Occurring to Fish Excluding Brook Trout1
0 = percent of baseline number of lakes
+100peq/l
+50ueq/l
Confidence Confidence
Interval Interval
80% 90%
Confidence Confidence
Interval Interval
80% 90%
Lake Trout Stocked
68 Lakes Total
Lake Trout Self Sustaining
68 Lakes Total
Small Mouth Bass
103 Lakes Total
White Sucker
276 Lakes Total
Common Shiner
122 Lakes Total
-35.6
(52)
-37.8
(56)
-59.5
(58)
-99.2
(36)
-19.7
(90)
-37.8
(56)
-37.6
(55)
-59.7
(58)
-100.6
(36)
-19.7
(90)
-21.4
(31)
-23.8
(35)
-37.6
(37)
Jf9.5
(18)
-12.6
(57)
-22.3
(33)
-24.8
(41)
-37.9
(37)
-49.6
(18)
13.2
(60)
* For brook trout, see Table 5-19. Fat head minnow results were not included because of
the small size of the sample contained within1 the DEC files.
5-48
-------
1. white sucker
2, stocked brook trout
3. self-sustaining brook trout
4. stocked lake trout
5. small rnouth bass
6. self-sustaining lake trout
7. common shiner
Laboratory results indicate that the white sucker is more sensitive to acidification in
Adirondack waters than brook trout (Baker, 1981). Our results do not necessarily contra-
dict those studies, but can be explained by the distribution of white sucker lake habitat
compared with brook trout habitat recorded in the DEC data base. Twenty-seven per-
cent of the lakes possessing records of white sucker have a pH greater than 7.00, while
only 1* percent of the brook trout lakes have a pH equal to or greater than 7.00. 27
percent of the white sucker habitat was predicted to shift to pH values no lower than
6.24 with an acid addition of 50 yeq/l. Similarly, 46 percent of the lakes with white
sucker have pH records where 6.0° is the base case. Assuming a 50 ueq/1 acid input,
these lakes would decrease to pH 5.12 or above. Only 43 percent of the brook trout lakes
have intial pH values equal to or greater than pH 6.0. While 73 percent of the white
sucker lakes won't shift below 5.12, only 57 percent of the brook trout lakes — under-this
scenario — remain above 5.12.
Caution must be used in interpreting the results listed above. They are not simply rela-
tions of sensitivity of different fish species to acidification. Rather, the results must_be
interpreted as being the relative sensitivities of regional fish habitats. The results listed
in Table 5-21 and Tables 5-12 through 5-19 should not be interpreted as representing the
sensitivity of different fish species to pH stress as expressed by the scientific partici-
pants within the study. These results are a sample application of the data developed in
the elicitation interviews, and are based upon additional inventory data and water
chemistry modeling information.
Decrease in Acidification — Predicted Effects Upon Brook Trout Habitat
One scenario was examined in which acidification was reduced by 50 yeq/1. This reduc-
tion could be achieved either by reductions in acidic inut into the lakes or by direct base
5-49
-------
Table 5-21
Results of Secondary Sort Representing Lakes Which Contained
Both Fish Records and Chemistry Data
1
Brook Trout
Alkalinity*
Elevation:
= 1500 ft.
1501 - 2000 ft.
2000 ft.
Unknown
Total
Alkalinity*
Elevation
= 1500 ft.
1501 - 2000 ft.
2000 ft.
Unknown
Total
200
16
10
1
27
54
200
18
2
0
19
39
100-200
4
21
7
2*
56
Small
100-200
8
7
0
8
23
50r100
7
2*
g
21
60
Mouth Bass
50-100
2
9
0
7
18
50
12
43
30
37
122
50
4
6
0
1
11
Unknown
9 48
24122
4 50
21130
58350
Unknown
2 34
2 26
0 0
8 43
12103
Total
Total
i
Fat Head i Minnow
Alkalinity*
Elevation
= 1500 ft.
1501 - 2000 ft.
2000 ft.
Unknown
Total
200
1
0
0
0
1
100-200
0
0
0
0
0
50-100
0
0
0
0
0
50
0
0
0
0
0
Unknown
0 1
0 0
0 0
0 0
0 1
Total
Lake i Trout
Alkalinity*
Elevation
= 1500 ft.
1501 - 2000 ft.
2000 ft.
Unknown
Total
200
2
2
0
12
100-200
1
6
0
8
1*
50-100
3
5
2
8
15
50
0
15
3
5
21
Unknown
0 6
3 31
0 5
3 3
6 68
Total
26
* Alkalinity expressed in yeq/1.
5-50
-------
Table 5-21
Remits of Secondary Sort Representing Lakes Which Contained
Both Fish Records and Chemistry Data
(Continued)
Common Shiner
Alkalinity*
Elevation
= 1500 ft.
1501 - 2000 ft.
2000 ft.
Unknown
Total
Alkalinity*
Elevation
= 1500 ft.
1501 - 2000 ft.
2000 ft.
Unknown
Total
Alkalinity*
Elevation
= 1500 ft.
1501 -2000 ft.
2000 ft.
Unknown
Total
Alkalinity*
Elevation
= 1500 ft.
1501 - 2000 ft.
2000 ft.
Unknown
Total
200
4
1
1
1
7
200
31
12
1
35
79
200
53
16
1
58
128
200
87
37
1
94
219
100-200
0
3
0
0
3
White
100-200
10
25
2
27
64
Total: All
100-200
17
35
7
42
101
Total:
100-200
44
83
13
85
- 225
50-100
0
3
2
1
6
Sucker
50-100
5
19
4
12
40
Fish Species
50-100
14
32
9
30
85
All Lakes
50-100
25
69
21
60
175
50
0
4
0
1
5
50
9
30
11
17
67
50
18
63
31
45
157
50
26
143
99
179
447
Unknown
0 4
0 11
0 3
1 4
1 22
Unknown
6 61
10 96
0 18
10101
26276
Unknown
13115
28174
4 52
34209
79550
Unknown
23205
46378
7141
75493
1SU217
Total
Total
Total
Total
* Alkalinity express inyeq/1.
5-51
-------
addition, i.e., liming. According to Henriksen's estimate of the present level of Adiron-
dack acidification, a 50 peq/1 reduction in acidity would equal approximately a 50 per-
cent reduction in anthropogenic acidification. A new algorithm was required to predict
the effects of a base addition to Adirondack waters. The algorithm developed for
predicting incremental damages, which has been previously described, could not be used
to calculate decremental changes in acidification.*
The calculation of the increase in fish habitat resulting from a decrease in acidification
involved a four step process, see column * of Table 5-22. First, the number of lakes
which could support fish, in our case, self-sustaining brook trout populations in a natural,
pre-acidified condition, was estimated. Only populations of self-sustaining brook trout
were evaluated in this scenario. The baseline distribution was calculated by assuming the
addition of 100 yeq/1 base to each lake cluster. This base addition was interpreted as if
it were equivalent to the removal of 100 yeq/1 of acidity. To evaluate the effects of
varying levels of acidification, it was necessary to make an assumption concerning the
ability of lakes in this baseline chemical condition to support fish populations. It was
assumed that 80 percent of the lakes with pH values between 7.2 and 5.0 in the baseline
distribution were originally capable of supporting self-sustaining brook trout popula-
tions. Examination of the actual current pH data indicates that only 1* percent of the
lakes containing brook trout populations have pH values above 7 and it is highly unlikely
that self-sustaining populations could have proliferated in the dilute waters charac-
teristic of the Adirondack region at pHs beljow 5.0.** A second sort of the Adirondack
Ponded Waters Data Management System tape was performed in order to classify those
lakes for which both fish survey records and pH records were available. Table 5-21
reports the results of this work. A total of 7^2 such lakes were identified. 350 of these
* This was because the original algorithm back calculated the original number of lakes
able to support fish from the ration of the number of lakes presently supporting fish and
the estimated damage levels. Where the estimated damages were very high, this caused
division by a very small number and this, in turn, caused unacceptable imprecision in the
results. For example, if damages at a giverj pH were estimated at 95 percent and 5 lakes
were found in that cluster, then the original number of lakes capable of sustaining fish
would be calculated as 5/(1.0 - .95) = 100. If, alternatively, the damages were estimated
as 97.5 percent, the original number of iakep would be 5/(1.0 - 97.5) = 200.
** This underestimates the number of lakes capable of supporting brook trout by not
considering organic acid waters with pH 5.0. It is known that these waters can sustain
brook trout populations at or below pH 5.0.
5-52
-------
Table 5-22
Results of Reducing Present Acidification by 50 ueq/1*
Cumulative
Probability
1.25
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
98.5
100peq/l
Acid Addition
5.3
16.0
25.2
33.1
39.9
45.3
50.7
56.2
61.8
67.7
73.6
79.8
86.3
96.2
107.6
123.1
152.0
180.0
209.2
241.0
359.6
Estimated Decrease in
50 ueq/1 Present Acidification
Acid Addition to 50yeq/l**
1.8
5.6
8.4
10.8
12.6
13.9
15.2
16.5
17.8
19.1
20.4
21.7
23.1
26.3
29.8
33.8
39.1
45.9
53.5
62.2
138,9
3.5
10.4
16.8
22.3
27.3
31.4
35.5
39.7
17.8
48.6
53.2
58.1
63.2
69.9
77.8
89.3
112.9
134.1
155.7
178.8
220.7
Expected Value
96.3
27.3
69.0
* The large difference between the extreme numbers (at the 98.5 percent probability
level) for the two distributions is due to the greater probability assigned to values lying
outside the range given in the elicitation. For the 90 percent confidence level, there is
assumed to be a 10 percent probability that the true value lies outside the range given.
For the 80 percent confidence level, this probability is assumed to b* 20 percent, effec-
tively doubling the assumed probability of-the occurrence of very high or low damages.
The result is that when base additions are considered, the upper end of the distribution
will be very sensitive to the elicitation results and the assumed confidence interval. To
adequately handle scenarios where there are base additions, changes should be made in
tine numeric algorithm that calculates incremental change. For acid additions, the cur-
rent algorithm for the calculations of damaged lakes is appropriate. The results were
calculated assuming a baseline distribution of 424 Adirondack lakes between pH values of
7.2 and 5.0 which were capable of sustaining populations of brook trout prior to acidifi-
cation
** Calculated by subtracting values of the 50yeq/l acid addition from the 100 ueq/1 addi-
tion.
5-53
-------
lakes were reported as containing brook trout. Using the pH shift presented in Figure
5-7, the number of lakes which were contained within the range between pH 5.0 and 7.2
i
were calculated after 100 yeq/1 of base was added to each lake cluster.
5. ,
The second step calculated the reduction in the number of lakes capable of supporting
self-sustaining brook trout populations given a ,50 yeq/1 addition of acid to the baseline
distribution of lakes.
The third step consisted of the calculation of the shift in brook trout habitat that would
result from a 100 yeq/1 acid addition to the ba'seline distribution of lakes. Such an addi-
tion is equivalent to a return to the present level of acidification from the predicted
baseline.
The fourth step consisted in the calculation of the effect of a 50 yeq/1 reduction in
acidity from the present distribution. This was accomplished by calculating the dif-
ference between the 100 yeq/I acid additifcn and the 50 yeq/1 acid addition, The
difference represents the improvement which could be expected by moving from a
current state of 100 ueq/1 of anthropogenic acidification to a future state of 50 yeq/1 of
anthropogenic acidification. The results shown on Table 5-22 indicate that brook trout
habitat could be expected to increase from a lower bound of 3.5 to an upper bound of
220.7 out of 424 lakes. The expected value of the change is equal to 16 percent, or 69
out of 424 lakes.
5.4.3. Alternative Damage Calculation
In order to further evaluate the validity of the damage estimates generated using the
elicitation results, an alternative method of calculating damages was used for "self-
sustaining" brook trout lakes in the Adirondacks, This method involved utilization of the
"baseline" pH distribution assumed to represent the chemical condition of the lakes prior
to anthropogenic acidification discussed in 5.4.2. Damage estimates were calculated by
considering changes from this calculated Baseline" pH distribution to new pH distribu-
tions reflecting varying levels of acidification. In order to evaluate the effects of vary-
ing levels of acidification, it was necessary to make the same assumption concerning the
ability of lakes in this baseline condition tp support fish populations; see Section 5.4.2.*
The calculation used the additional sort of the Adirondack Ponded Waters Data Manage-
5-54
-------
ment System tape depicted in T 1- '•'•). "Hie pH shift program, which was derived from
Figure 5-7, was then used to estirna-:^ damages due to a shift in pH from the baseline.
The following shifts were considered:
(1) the addition of 100 yeq/1 of acid to the baseline distribution of lakes.
This amounted to a return to the present state, as the baseline was cal-
culated by removing 100 yeq/1 of acidity from the present distribution
of lakes possessing brook trout records.
(2) the addition of 50 yeq/1 of acid to the baseline.
(3) the addition of 150 yeq/1 of acid; equivalent to the current distribution
plus 50 yeq/1 of acidification.
' (*) the addition of 200 yeq/1 of acid; equivalent to the current distribution
plus 100 yeq/1 of acidification.
A summary of these additions is shown in Table 5-23.
The results of the third shift, from the assumed baseline to the present state, are par-
ticularly interesting, since it can be used to provide some information on the accuracy of
the eiicitations and algorithms. Of course, the elicitations were designed to estimate the
effects of pH on fish populations. If other factors are causing damages to fish popula-
tions, then these figures would not necessarily correspond.
The present number of Adirondack lakes listed on the data base which are capable of
maintaining self-sustaining brook trout populations is 275. Using the incremental
damages generated by the elicitation results and the assumption that 80 percent of the
lakes between 5.0 and 7.2 could sustain brook trout populations, our approach predicted
that 327.7 lakes would be capable of sustaining brook trout populations. This discrepancy
can be explained by either:
* The investigators assumed that 80 percent of the lakes with pH values between 7.2 and
5.0 in the baseline distribution were originally capable of supporting self-sustaining brook
trout populations.
5-55
-------
Table 5-23
Damages Occurring to Lakes Capable of
Sustaining Self-Sustaining Brook Trout Populations
Between the Original pH Values of 5.0 and 7.2
Shift
Predicted
Number of
Damaged
Lakes
Predicted
Number of
Surviving
Lakes
Actual
Number of
Surviving
Lakes
(1) Assumed baseline reduction
of (100 ueq/Iacid)
(2) Addition of
+ 50 ueq/acid to baseline
27.3
396.7
(3) Present State estimated by
adding 100 ueq/1 acid
to baseline M
96.3
327.7
2751
150 yeq/I acid
addition to baseline
192.0
232.0
(5) 200 ueq/1 acid
addition to baseline
277,1
146.9
lakes equal the total number of lake's between pH values 5.0 and 7.2 after the
chemical shift caused by the addition of 100 peq/1 base (530) multiplied by .8. It is
assumed that 80 percent of the lakes between 5.0 and 7.2 could support brook trout.
JLJL
275 lakes are the number of observed lakes which contain brook trout populations on
the data base between a pH of 5.0 and 7.2.
5-56
-------
(1) the estimate that 80 percent of the lakes could sustain brook trout
populations is too high, or
(2) that the scientific elicitations were conservative estimates of the
damage, which is actually larger than they would predict, or
(3) there are factors other than pH that contribute to a reduction in brook
trout fish populations.
5-57
-------
-------
6J3 CONCLUSIONS
6.1 PROJECT OVERVIEW
This project demonstrated a framework for organizing scientific information in a manner
useful for policy assessment. In accomplishing this, two different sets of results were
developed. The first set of results, which is shown in Table 4-6, represents the results of
the elicitation interviews with participating scientists. The second set of results, which
are based on the elicitation results, are contained and discussed in Chapter 5. The
results in Chapter 5 provide an example of how the data generated in the elicitation
interviews can be applied. Because they are based upon inventory data that is expected
to be significantly improved prior to 1985, and because the uncertainty surrounding this
data wasn't, specifically dimensioned, the conclusions to be drawn from these results
should pertain to the adequacy of the procedure, instead of the specific numeric dimen-
sioning of the effects of acidification on Adirondack fish populations.
The framework was designed to interpret the scientific uncertainty in estimates of the
relationship between acidification and aquatic effects and to further assist in extrapo-
lating spatially limited scientific data to regional areas. If successful, it will augment
scientific research by reframing effects information so that non-scientists can better
evaluate the probability of different levels of regional environmental damage. The
framework does not, however, replace effects research, as it does not generate new
hypotheses or data. It simply organizes existing data and interpretation into a format
complementary to conventionally presented scientific results. To evaluate whether the
framework accomplishes its primary goal, two questions need to be answered:
1. Are the analytic procedures and the results of the framework credible
to the scientific community?
2. Are the results of the framework helpful to economists, policy analysts,
and decision makers?
A framework designed to interpret scientific data and knowledge must have scientific
credibility and endorsement if it is to be a useful too! for the assessment of acid deposi-
6-1
-------
tion policy alternatives. The framework developed and applied within this project has
not received general or widespread scientific comment. As a result, the credibility of
the approach is generally untested. The project has, however, been reviewed by a
number of the participating scientists and a review panel at the National Acid Precipi-
tation Program Effects Research Review Meeting.* Preliminary scientific review
suggests that scientists generally recognize tljie need to present the results of effects
research in a manner that enables decision makers to evaluate the regional importance
and uncertainty of acid deposition effects. A recognized advantage of the framework is
that members of the scientific community directly extrapolate from and interpret the
results of their investigations, rather than having less scientifically knowledgeable policy
analysts performing this extrapolation. Given the considerable uncertainty surrounding
the performance of regional damage estimates, which depend upon uncertain damage-
relations and inventory data, it seems unlikely that any framework addressing this
problem will receive unequivocal scientific endorsement.
The second question concerns whether the outputs of the procedure will be useful as
inputs to policy analysis and economic benefits studies. Here a more definitive answer is
available. The type of information produced by this project is required by all policy
assessment frameworks, and is also necessary for economic benefits studies. Policy
analysis cannot be performed without a regional characterization of the resources at risk
and the likelihood of different levels of damage on a regional level. To know that effects
have occured at dispersed sites, mainly in the northeastern U.S., is significant, yet
incomplete knowledge for the assessment. By dimensioning the uncertainty for the
aquatic and the other acid deposition caused effects many important policy questions can
be addressed. In particular, any policy assessment based on uncertain information must
have the uncertainty in the parameters dimensioned if responsible decisions are to be
made. In the past, ad hoc procedures for incorporating uncertainty into assessments have
been used by policy analysts. Procedures directly utilizing scientific expertise can signi-
ficantly improve this aspect of policy evaluation. In addition, it would allow analysts to
determine the value of additional scientific research and whether reductions in acid
deposition should be sought while additional effects research is being performed.
Effects Research Review Meeting of the National Acid Precipitation Assessment
Program, Raleigh, North Carolina. February 21-25, 1983.
6-2
-------
T.-.e elicitations used in this project accomplished several important objectives. Esti-
mates of damage to fish populations were made directly by scientists participating in the
research. These effects estimates are not directly available from published scientific
experiments since most of these experiments are performed in the laboratory and cannot
easily be extrapolated to natural habitats or, where research results are available for fish
damages in natural habitats, few sites have been studied. The implications of the current
scientific research for estimating the effects of acid deposition on fish populations are
not straightforward and require the considered judgment of the scientists performing the
research. Since point estimates would be unreasonably precise given the current uncer-
tainties, the elicitation focused on estimating the expected range of potential effects for
a given pH value.
The estimates developed in this project have the advantage of being expressed quanti-
tatively. As such, they can be easily reviewed and, if need be, revised in the light of new
scientific information. For example, the damage relationships specified in this project
estimate effects as a function of pH, aluminum, and calcium concentrations. If new
research determines that some other variable is important to the prediction of effects,
then that variable can be incorporated. Similarly, scientists can easily revise the esti-
mates of effects either upwards or downwards as new information becomes available
from completed projects.
Another accomplishment of the elicitations used in this project is that the damage rela-
tions which resulted from the application of the framework integrate many different
mechanisms and possible causal agents of fish population decline, e.g., variations in lake
water chemistry due to groundwater upwellings, episodic pH shocks, and lake morpho-
metry. This characteristic of the damage relation allows for a relatively simple and
understandable dimensioning of the predicted effect and its uncertainty which will be
easily understood by policy analysts. This allows for the explicit representation of
current scientific opinions and the development of a regional perspective for acid depo-
sition effects.
The probabilistic structure of the project and its reliance upon explicitly represented
expert judgement in the development of regional effects estimates does, however,
generate considerable scientific skepticism. Scientists have tended to assume that this
framework is designed to find out something new about the world, rather than to provide
a structure for the incorporation of scientific information and advice within a highly un-
6-3
-------
certain policy debate. In this project, scientists were required to perform an assessment
role in addition to their responsibilites as effect? researchers. Consequently, some con-
fusion exists concerning the status of the project, i.e., is it really new scientific
research? The obvious answer is no. Recall that it is an assessment device useful for
i
organizing technical information. Its outputs represent the considered judgement of the
scientific community. As such, It is part o^ an integrated assessment intended to
organize and present scientific descriptions of the causal mechanisms and regional extent
of acid deposition effects. The rationale for the framework assumes that:
o Data sets containing statistically valid samples of observations on
effects will not be available prior tb 1985, and in some cases the 1989
assessments.
o Policy assessment requires the b^st information on scientific un-
certainty available. This is most likely to come from the scientists
performing the actual research.
6.2 CONCLUSIONS REGARDING THE SPECIFIC PROCEDURES USED IN THIS
PROJECT
A number of conclusions can be drawn that arp specific to the implementation of this, or
similar, assessment techniques. First, projects of this type require an iterative process
with extensive interaction between scientists and the project investigators. Both the
variables to be explicitly considered in the elicitation and the structure of the elicitation
should be determined in conjunction with participating scientists. This helps to assure
that the judgements being elicited can, in fact, be reasonably evaluated by the scientists.
One example of the importance of interactive communication between the project inves-
tigators and the participating scientists is the specification of variables for use in the
elicitation. The determination of the appropriate variables for use in policy assessment
can become as complex as it is important. For example, the independent and dependent
variables of the damage relations were changed several times during the project as a
result of communication between the investigators and participating scientists. The
process was iterative, with each succeeding set of variables being more appropriate to
the assessment than their predecessors. Primary factors influencing their selection were
6-4
-------
the scientists' understanding of effect mechanisms} the extent of avti-^-'-i inventory
information; and the size of the region over which effects were to be estimated. While
the project investigators suggested the original variables for the analysis, subsequent
specifications were the products of discussions with scientific participants.The linkage
between the information needs of the policy analyst and the scientific investigator is not
straightforward. Policy analysts are not able to adequately specify information needs to
the effects scientists without considerable and extended dialogue. As scientific informa-
tion and estimation improves, policy analysts will attempt to incorporate that informa-
tion within their assessments.
A second conclusion is that, when dimensioning the uncertainty in effects estimates, it is
important to try to achieve only that degree of resolution which is scientifically credible
or responsible. Scientific judgements regarding the uncertainty of different acid deposi-
tion effects estimates are imprecise. This makes it difficult for scientists to provide
judgemental estimates of a precise probability distribution for effects outcomes. If the
approach for dimensioning uncertainty requires a level of precision that the scientist
cannot achieve, then the scientist will not view the resulting estimates as being
credible. The elicitation used in this project explicitly recognizes the difficulty in mak-
ing meaningful point estimates for the probability of different levels of effects. Instead
of point estimates, the elicitation estimated the range of potential effects (i.e.,. upper
and lower bounds) with only a simple weighting scheme to express the likelihood of
different outcomes within the range. The lack of information on causal mechanisms
associated with acid deposition effects and the extreme variability in different eco-
system's response to acid deposition warrants conservative expectations regarding the
precision with which scientists can estimate the probability of different effects out-
comes.
A third conclusion is that inventory data concerning resources at risk may be the limiting
factor in the performance of assessments concerning acidification effects upon fish
populations. While there is considerable uncertainty concerning the effects of acidifica-
tion upon fish populations, there may be even more uncertainty in the current status of
regional fish populations and aquatic chemistry inventories. This project focused on
dimensioning the uncertainty in estimates of damages to lakes presently containing fish.
However, no attempt was made to dimension the uncertainty in estimates of the number
of lakes that currently support fish populations. If explored, it seems possible that the
uncertainty in the dose-response relationship for certain species of fish. Because of this
6-5
-------
relative absence of complete inventory data, statistical generalization over the outputs
of mechanistic or empirical effects models will be very uncertain. Techniques similar to
the approach employed in this project could be useful in dimensioning the uncertainty in
estimates of resources at risk.
Interpretation of the project results should be guided by the following remarks. First,
the project was designed to evaluate the utility of a particular device for incorporating
scientific knowledge and advice within policy debates. As such, the damage estimates
generated should be viewed as preliminary. Furthermore, the proper emphasis of the
report should be the results listed in Chapter 4, which are the outcomes of the different
elicitation interviews with participating scientists. Chapter 5 consists of a sample appli-
cation of those results utilizing a very simple water chemistry model and a data base
provided by the New York State DEC. The data from the elicitation interviews,
however, can be easily coupled with other data bases and water chemistry models in
order to yield different predictions of fish population damage due to acidification. This
report does not connect fish population response to acid deposition. The independent
variable throughout was the pH of Adirondack ponded waters. No attempt was made to
estimate the effect of acidification upon stream populations. Because stream fishing is
important in the Adirondacks, future regional effects estimates should address this area
as well. Finally, the damage estimates generated in Chapter 5 should not be interpreted
as damages to fish per se^ but rather to fish populations within their recorded lake habi-
tats. The sensitivity of these fish habitats to acidification is a function of many factors,
and is different from sensitivity recorded for the same species in bioassay experiments.
For example, while brook trout are known to be resistant to acidifcation effects, much
brook trout habitat is vulnerable to acidification because of its morphometric and bio-
geochemical characteristics.
63 AREAS FOR FUTURE RESEARCH
There are a number of other techniques tha} could be used to dimension uncertainty. All
use some variant of subjective probability eiicitations. The only exception is the use of
statistically derived confidence intervals. However, the use of statisical confidence
intervals for dimensioning uncertainty around acid rain estimates suffers from two
severe shortcomings. First, they are based on a set of statistical assumptions and model
specifications (e.g., functional form) that may not be correct; and second, statistically
valid examples are unlikely to exist for most effects areas. Still, there are techniques
6-6
-------
for combining sample information with the type of information generated by this
project. These techniques, based on Bayesian methods, should be explored whenever
sample data is available.
The elicitation technique used in this project estimated the range of possible effects with
very little information on the shape of the probability distribution within this range. This
process was selected because of the extreme uncertainty characteristic of acid deposi-
tion caused fish population effects. It might be possible, however, to estimate more
accurately the form of the probability distribution within the lower and upper boundaries
of the predicted effects.
Recall that the project segmented the range between the lower and upper bounds of the
effects by calculating the midpoint and then asking the scientist to indicate which seg-
ment of this range was more likely. For example, see line A in Figure 6-1:
Figure 6-1
Possible Techniques for the Estimation of
Probability Distributions
A.
B.
C.
pH. 4.6
pH. 4.6
pH. 4.6
10
10
10
20
4 20 3
30
30
30
X
X 40
1 40
50
50
2 50
A more precise estimate of the cumulative probability distribution can be obtained by
segmenting the range into fourths rather than halves as in line B of Figure 6-1. More
information could be obtained by having the scientist rank each segment within the range
from most Ukely to least likely, as shown on line C of Figure 6-1.
Another alternative would be to have the scientist directly estimate a probability distri-
bution for effects at each pH level. This would require the scientist to estimate, for
each pH level, the precise probability associated with each effects outcome or range of
outcomes. In this project, the effects outcome was a reduction in fish habitat. An elici-
6-7
-------
tation that would directly estimate the cumulative probability distribution was briefly
considered in the pretest and one trial application of this technique was conducted. This
elicitation technique was not used in the project because the estimation of a precise
cumulative probability distribution proved to,be extremely difficult for the scientist
given the current state of knowledge. Further^ the estimation of a cumulative probabil-
ity distribution for each of the pH levels and1 the different species of fish would have
proven to be even more lengthy than the four to five hour interview required by this
process. Still, a more complete exploration of this approach is probably warranted.
In recent work, Feagens and Biller (1981) hav^ designed elicitation techniques that esti-
mate upper and lower bounds to the probability of different effects outcomes. This
result is a direct estimate of the cumulative probability distribution accompanied with
bounding upper and lower distributions. This variation could also be explored. These and
other details of the application of subjectivei probability analysis deserve further con-
sideration with any application to deposition effects areas. The structure of the device
used to develop the probability distribution can be expected to vary across different
effects areas depending upon the level of uncertainty present in estimating regional
effects.
As suggested in the previous section, another useful research project would involve a
dimensioning of the uncertainty in estimates of resources at risk (e.g., fish population
data). To estimate effects, both dose-response and inventory data are needed. At
present, the fish inventory data is very limited and, therefore, the fish populations at risk
are uncertain.
The framework used in this project presents damage relationships based upon the judge-
ments of selected scientific participants. The opinions and decision making characteris-
tics of the participating scientists clearly influence the results presented in this report.
Selection criteria for participants and theijr justification are, therefore, important for
future studies. These could be developed ih conjunction with scientists already partici-
pating within the National Acid Precipitation Program and a number of external
reviewers.
6-8
-------
6.* CONCLUDING REMARKS
Regional damage assessments of acid deposition effects will continue to be characterized
by scientific uncertainty. As new scientific information and interpretation becomes
available, earlier estimates will be revised. If regional assessments are going to be
performed, direct scientific participation within the planning and implementation of
regional assessments is necessary. This project established that scientists are willing to
perform such a role. This role involves both the specification of the effects' model
parameters and the estimation of the parameter values. The importance of having scien-
tists help develop the parameters for effects assessment models has often been over-
looked. Conventionally, members of the scientific community have requested policy
analysts to provide a "wish list" describing the parameters necessary for the performance
of an analysis. This approach has led to considerable confusion. We recommend a joint
specification of assessment parameters where the scientist suggests appropriate para-
meters on the basis of current and predicted scientific information. Secondly, the parti-
cipating scientist should directly estimate the extent of effects resulting from acid
deposition. This project has determined that this can be credibly performed. In many
cases, the damage estimates were characterized by wide ranges, yet this doesn't indicate
that the estimates are inaccurate. It simply represents the degree to which scientists
are uncertain concerning the relationship between the parameters selected for analysis.
The credibility of the results and the process itself has been difficult to determine. In
large part, the readers of this report will be responsible for either adopting or rejecting
the technique and results. It must be remembered, however, that it is the first attempt
to explicitly dimension the uncertainty surrounding acid deposition effects. Regardless
of the technique used to develop regional damage estimates, those estimates will be
uncertain. This, or a similar technique, provides important additional information
concerning the confidence of those regional estimates. Finally, since the uncertainty in
the scientific information is so important in the assessment process, the uncertainty in
the scientific estimates must be dimensioned. If this approach is felt to be unacceptable,
then another must be suggested. The project investigators are aware of the shortcomings
and problems in this approach. The question is how to best perform a difficult task.
6-9
-------
-------
APPENDIX A
TEST CASE SENSITIVITY ANALYSIS OF THE
WEIGHTING AND CONFIDENCE INTERVAL ASSUMPTIONS
A-l
-------
APPENDIX A
TEST CASE SENSITIVITY ANALYSIS OF THE
WEIGHTING AND CONFIDENCE fNTERVAL ASSUMPTIONS
This appendix contains a brief illustration ojE the procedure developed to estimate
damages to fish populations caused by acidification. Table A-l presents the input pH
distribution data required by the analysis. It shows the "current" distribution of pH
values across Adirondack lakes containing brook trout as well as the shift in the
distribution of pH values assumed for a "sensitivity test case." This current distribution
of pH values was obtained from the Adirondack Water Management System data base
i
maintained by the New York DEC. In calculating the final results presented in Section
5.3, refinements in this pH distribution were made (see Section 5.2). However, this
distribution of pH values was used to test the sensitivity of the confidence interval and
weighting assumptions. A simple .2 pH shift across all the lakes was assumed. For
example, the 55 lakes with a current pH of 7.0 were assumed to shift to a pH of 6.8.*
Given this pH shift, a cumulative probability distribution of the number of additional
damaged lakes, i.e., lakes which can no longer support brook trout, is calculated.
This simple assumed shift of .2 pH was used to test the sensitivity of the results to the
confidence interval and weighting assumptions. Nine cases were considered:
Confidence Intervals of 95 percent, 90 percent and 80 percent
Weightings of 2:1, 3:2, 5:4
Table A-2 presents the results of four of the,se cases: 1) 95 percent confidence interval
with a 2:1 weighting; 2) 80 percent confidence interval with a 2:1 weighting; 3) 95
percent confidence interval with a 5A weighting; and 4) 80 percent confidence interval
with a 5:4 weighting.
* Since these 55 lakes have the same starting pH and same ending pH, they would repre-
sent a "cluster" of lakes as defined previously. The assumption of perfect correlation
allows the simple summing of the cumulative probability distributions across all
"clusters" of lakes to obtain an aggregate cumulative probability distribution of damages.
A-2
-------
Table A-l
Data Input: Assumed Change in the Distribution of pH Values
Across Adirondack Lakes Currently Containing Brook Trout
Sensitivity Test Case; A .2 pH Reduction
Current Distribution of
pH Values Across
Adirondack Lakes
Example Decremental
Change in the Distribution of
pH Values Across
Adirondack Lakes
PH
7.0
6.9
6.8
6.7
6.6
6.5
6.4
6.3
6.2
6.1
6.0
5.9
5.8
5.7
5.6
5.5
5.*
5.3
5.2
5.1
5.0
" 4.9
4.8
4.7
4.6
1.5
4.4
4.3
4.2
4.1
4.0
3.9
Number of Lakes
55
13
14
9
11
19
14
22
22
13.
34
16
17
17
8
23
11
6
9
11
16
10
20
17
16
11
4
8
4
3
4
0
pH
7.0
6.9
6.8
6.7
6.6
6.5
6.4
6.3
6.2
6.1
-6.0
5.9
5.8
5.7
5.6
5.5
5.4
5.3
5.2
5.1
3.0.
. 4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4.0
3.9
3.8
Number of Lakes
—
_
55
13
14
9
11
19
14
22
22
13
34
16
17
17
8
23
11
6
9
11
16
10
20
17
16
11
4
8
4
3
4
A-3
-------
r
Table A-2
Sensitivity Test Case Results:
Cumulative Probability Distribution Showing
the Probability That a| Given Number
of Lakes Will Be, Damaged
Case 1
95% Confidence Interval
2:1 Weighting
Case 2
80% Confidence Interval
2:1 Weighting
Cumulative
Probability
.05
.10
.20
.30
.40
.50
.60
.70
.80
.90
.95
Number of
Damaged Lakes
20.5
37.2
48.9
56,7
59.6
62.2
65.5
71.9
79.8
82.5
93.0
Cumulative
Probability
.05
.10
.20
.30
.40
.50
.60
.70
.80
. .90
.95
Number of
Damaged Lakes
23.1
36.0
43.7
54.8
60.6
63.8
68.1
76.4
80.2
87.3
126.5
Case 3
95% Confidence Interval
5:* Weighting
Case 4
80% Confidence Interval
5:* Weighting
Cumulative
Probability
.05
.10
.20
.30
.40
.50
.60
.70
.80
.90
.95
Number of
Damaged Lakes
23.5
41.4
52.5
58.0
61.0
63.0
66.1
72.7
81.0
87.5
94.9
Cumulative
Probability
.05
.10
.20
.30
.40
.50
.60
.70
.80
.90
.95
Number of
Damaged Lakes
26.4
39.5
48.5
57.3
62.3
64.9
69.5
77.7
82.8
88.7
126.1
A-4
-------
A comparison of Case 1 with Case 2 and Case 3 with Case * shows the sensitivity of the
calculated distribution of damages to differences in the assumed confidence interval.
This comparison shows a reasonably significant difference in the endpoints of the range
of damages when different confidence intervals are assumed. The same comparison also
show's the'Sensitivity of the distribution of damages due to different weighting assump-
tions. The differences between both the 50th fractile value and endpoints of the distri-
butions using different weightings are quite small. Based on this test case sensitivity
analysis, it was decided that future sensitivity analyses would focus on the confidence
interval assumption rather than on the assumed weighting. The standard set of runs for
each species was decided to be comprised of two scenarios: a 90 percent confidence
interval with a 3:2 weighting assumption and a 80 percent confidence interval also with a
3:2 weighting. However, a full sensitivity analysis was performed for brook trout.
A-5
-------
-------
SELECTED BIBLIOGRAPHY
Arnold, D., R. Light, and V. Dyrnond. 1980. Probable Effects of Acid Precipitation on
Pennsylvania Waters. U.S. EPA Corvallis Environmental Laboratory, EPA-600/3-
80-012.
Baker, J. 1981. Aluminum Toxicity to Fish as Related to Acid Precipitation and
Adirondack Surface Water Quality. Ph.D. dissertation, Cornell University, Ithaca,
New York.
Baker, 3.P. 1982. Effects on Fish of Metals Associated with Acidification. In Proceed-
ings of an International Symposium on Acidic Precipitation and Fishery Impacts in
Northeastern North America. Amer. Fish Soc., Bethesda, Maryland.
Baker, 3. and C. Schofield. 1982. Aluminum Toxicity to Fish in Acidic Waters. Water,
Air and Soil Pollution. 18:289-309.
Colquhoun, ]., 3. Symuia, R. W. Karcher, Jr. 1982. Report of Adirondack sampling for
stream acidification studies-1981 supplement. New York State Department of
Environmental Conservation, Technical Report 82-3.
Colquhoun, J., 3. Symuia, M. Pfeiffer, and 3. Feurer. 1980. Preliminary report of stream
sampling for acidification studies 1980. New York State Department of Environ-
mental Conservation, Technical Report 81-2.
Chen, C. W. 1981. Characteristics of Woods, Sagamore, and Panther Watersheds. In The
Integrated Lake Watershed Acidification Study. Contributions to the International
Conference on the Ecological Impact of Acid Precipitation. EPRI EA-1825. Project
1109-5. Interim Report. Lafayette, California. May 1981.
Chen, C. W. 1981. Introduction to the Integrated Lake-Watershed Acidification Study.
In The Integrated Lake Watershed Acidification Study. Contributions to the Inter-
national Conference on the Ecological Impact of Acid Precipitation. EPRI EA-
1825. Project 1109-5 Interim Report. Lafayette, California. May 1981.
B-l
-------
Driscoli, C.T. 1980. Chemical Characterization of Some Dilute Acidified Lakes and
Streams in the Adirondack Region of New York State. Ph.D. dissertation, Cornel!
University, Ithaca, New York.
Driscoli, C.T., 3.P. Baker, 3.3. Bisogni, and Q.L. Schofield. 1980 Effect of Aluminum
Speciation on Fish in Dilute Acidified Waters. Nature. 284:161-164.
Driscoli, C.T.. and 3.3. Bisogni, 1982a. Weak Acid/Base Systems in Dilute Acidified
Lakes and Streams in the Adirondack Regions of New York State, in Modeling of
Total Acid Precipitation Impacts. 3. L. Schnoor, ed. Ann Arbor Science,
Michigan. In press.
Driscoli, C.T., 3.P. Baker, 3.3. Bisogni, and C,L. Schofield. 1982b Aluminum Speciation
and Equilibrium in Dilute Acidic Surfacte Waters of the Adirondack Region of New
York State. In Geologic Aspects of Acid Rain. O. P. Bucher, ed. Ann Arbor
Science, Ann Arbor, Michigan. In Press.
Feagans, T.B., and W.F. Biller. 1981. A general method for assessing health risks
associated with primary national ambient air quality standards. Research Triangle
Park, NC: EPA Office of Air Quality Planning and Standards, May 1981.
Haines, T.A. 1981. Acidic Precipitation and its Consequences for Aquatic Ecosystems:
A Review. Trans. Amer. Fish. So. 110;669-707.
Henriksen, H. 1979. A Simple Approach for Identifying and Measuring Acidification in
Freshwater. Nature.
Henriksen, A. 1980. Acidification of Freshwaters— A Large Scale Titration, pp. 68-74.
In Ecological Impact of Acid Precipitation. Proceedings of an International
Conference, Sandefjord, Norway, March 11-14, 1980. Drablos, D., and A Tollan,
eds. SNSF-project, Oslo-As, Norway.
Henriksen, A., 1982a. Susceptibility of Surface Waters to Acidification. In Proceedings
of an International Symposium on Acidic Precipitation and Fishery Impacts in
Northeastern North America. Amer. Fish Soc., Bethesda, Maryland.
B-2
-------
Henriksen, A. 1982b. Changes in the Base Cation Concentrations Due to Freshwater
Acidification. NTNF Report OF-81623.
Keeney, R.L. and H. Raiffa. 1976. Decisions with Multiple Objectives. Wiley Series in
Probability and "Mathematical Statistics.
Lichtenstein, 5., B. Fischhoff, and L.D. Phillips. Forthcoming. Calibration of
probabilities: The state of the art to 1980. In D. Kahneman, P. Slovic, and A.
Tversky (eds.), Judgment under uncertainty; Heuristics and biases. New York:
Cambridge University Press, in press.
Menz, F.C., 3.K. Mullen. 1982. Acidification Impact on Fisheries: Substitution and the
Valuation of Recreation Resources. Presented before the Division of Environ-
mental Chemistry, American Chemical Society Meetings, Las Vegas, Nevada. April
1982.
Muniz, I.P. and H. Leivestad. 1980. Toxic Effects of Aluminum on the Brown Trout,
Salmo trutta L, pp. 320-321. D. Drablos and A. Tollan, eds. Proceedings of an
International Conference on the Ecological Impact of Acid Precipitation. Norway,
1980. SNSF Project.
Mclntyre, S.H. 1982. "An Experimental Study of the Impact of Judgement Based
Marketing Models. Management Science, Vol. 28, No. 1, January 1981.
Omernik, 3. 1982. Total Alkalinity of Surface Waters.
Research Laboratory. U.S. EPA.
Corvallis Environmental
Pfeiffer, M. and P. Festa. 1980. Acidity Status of Lakes in the Adirondack Region of
New York in Relation to Fish Resources. FW-P168(10/80), New York State
Department of Environmental Conservation, Albany, New York,
Schofield, C.L. 1976. Dynamics and Management of Adirondack Fish Populations.
Project Report April 1, 1975-March 31, 1976, No. F-28-R-4, Dept. of Environ-
mental Conservation, Albany, New York.
B-3
-------
Schofield, C.L. 1982. Historical Fisheries Changes as Related to Cscreases in Surface
Water pH. In Proceedings of an International Symposium on Acidic Precipitation
and Fishery Impacts in Northeastern North America. Amer. Fish. Soc., Bethesda,
Maryland.
Spetzier, C.5., and Stael von Holstein. 1975. "Probability encoding in decision analysis.
Management Science. 1975, 22, 3*0-358.
Spry, D.3., C.M. Wood, and P.V. Hodson. 198,1. The Effects of Environmental Acid on
Freshwater Fish with Particular Reference to the Softwater Lakes in Ontario and
the Modifying Effects of Heavy Metals. A literature review. Can Tech. Rep. Fish
Aquat. Sci. 999, 50 pp.
Suppes, P. 197*. "The measurement of belief". Journal of the Royal Statistical Society.
Series B, 197*, 36, 160-175.
Violette, D.M., and D.C. Peterson. 1982 Assessing the Benefits of Policies Designed to
Reduce Acid Deposition; A Decision Analytic Benefit-Cost Framework, prepared
for the Economic Analysis Divison, U.S. EPA under Contract #68-01-65*3, March
1982.
-
. DC
B-4
------- |