United States
Environmental Protection
Agency
Prevention, Pesticides
and Toxic Substances
(7101)
EPA 712-C-96-144
April 1996
&EPA Ecological Effects Test
Guidelines
OPPTS 850.2500
Field Testing for
Terrestrial Wildlife
I

Public Draft"

-------
Introduction
This guideline is one of a series of test guidelines that have been
developed by the Office of Prevention, Pesticides and Toxic Substances,
United States Environmental Protection Agency for use in the testing of
pesticides and toxic substances, and the development of test data that must
be submitted to the Agency for review under Federal regulations.
The Office of Prevention, Pesticides and Toxic Substances (OPPTS)
has developed this guideline through a process of harmonization that
blended the testing guidance and requirements that existed in the Office
of Pollution Prevention and Toxics (OPPT) and appeared in Title 40,
Chapter I, Subchapter R of the Code of Federal Regulations (CFR), the
Office of Pesticide Programs (OPP) which appeared in publications of the
National Technical Information Service (NTIS) and the guidelines pub-
lished by the Organization for Economic Cooperation and Development
(OECD).
The purpose of harmonizing these guidelines into a single set of
OPPTS guidelines is to minimize variations among the testing procedures
that must be performed to meet the data requirements of the U. S. Environ-
mental Protection Agency under the Toxic Substances Control Act (15
U.S.C. 2601) and the Federal Insecticide, Fungicide and Rodenticide Act
(7 U.S.C. 136, etseq.).
Public Draft Access Information: This draft guideline is part of a
series of related harmonized guidelines that need to be considered as a
unit. For copies: These guidelines are available electronically from the
EPA Public Access Gopher (gopher.epa.gov) under the heading "Environ-
mental Test Methods and Guidelines" or in paper by contacting the OPP
Public Docket at (703) 305-5805 or by e-mail:
guidelines@epamail.epa.gov.
To Submit Comments: Interested persons are invited to submit com-
ments. By mail: Public Docket and Freedom of Information Section, Office
of Pesticide Programs, Field Operations Division (7506C), Environmental
Protection Agency, 401 M St. SW., Washington, DC 20460. In person:
bring to: Rm. 1132, Crystal Mall #2, 1921 Jefferson Davis Highway, Ar-
lington, VA. Comments may also be submitted electronically by sending
electronic mail (e-mail) to: guidelines@epamail.epa.gov.
Final Guideline Release: This guideline is available from the U.S.
Government Printing Office, Washington, DC 20402 on The Federal Bul-
letin Board. By modem dial 202-512-1387, telnet and ftp:
fedbbs.access.gpo.gov (IP 162.140.64.19), or call 202-512-0135 for disks
or paper copies. This guideline is also available electronically in ASCII
and PDF (portable document format) from the EPA Public Access Gopher
(gopher.epa.gov) under the heading "Environmental Test Methods and
Guidelines."
i

-------
OPPTS 850.2500 Field testing for terrestrial wildlife.
(a)	Scope—(1) Applicability. This guideline is intended to meet test-
ing requirements of both the Federal Insecticide, Fungicide, and
Rodenticide Act (FIFRA) (7 U.S.C. 136, et seq.).
(2) Background. The source material used in developing this har-
monized OPPTS test guideline is OPP 71-5 Simulated and Actual Field
Testing for Mammals and Birds (Pesticide Assessment Guidelines, Sub-
division E—Hazard Evaluation; Wildlife and Aquatic Organisms) EPA re-
port 540/09-82-024, 1982. This guideline may be used with OPPTS
850.7100.
(b)	Test standards—(1) Test substance. Unless specified otherwise,
data is to be derived from testing conducted with an end-use product. An
end-use product may be the applicant's own product or a typical end-use
product.
(2)	Test conditions. The test conditions for conducting a field test
should resemble the conditions likely to be encountered under actual use
of the product. Specifically, the pesticide should be applied to the site
at the rate, frequency, and method specified on the label.
(3)	Endangered species. Studies should not be conducted in critical
habitats or areas containing, or suspected to contain, endangered or threat-
ened plants or animals which may be threatened by the tests to be con-
ducted.
(4)	Residue levels. When the test substance is applied under field
condition testing, and residues of the test substance can be detected in
warm-blooded animals, residues should be determined in selected tissues
of test organisms and in vegetation, soil, water, sediments, and other ap-
propriate environmental components. If residues of the test substance can-
not be detected in warm-blooded animals, the applicant should consult
with the Agency before beginning the test.
(5)	Other standards. Additional standards for conducting field tests
are not delineated because of the wide variety of mechanisms by which
a pesticide may enter the environment, and because of the great variety
of food sources and habitats that may be affected. Any additional standards
for conducting these tests will be provided by the Agency in writing fol-
lowing consultation between the applicant or registrant and the Agency,
and will take into account the variety of mechanisms, food sources, and
habitats mentioned above.
(c)	General—(1) Introduction, (i) Data from full scale terrestrial
field studies are required by 40 CFR 158.145 on a case-by-case basis to
support the registration of an end-use product intended for outdoor applica-
tion. Because these studies are complex and costly, the Agency requires
1

-------
these tests to evaluate only those products that appear to pose significant
risks to nontarget wildlife.
(ii) Laboratory tests generally are amenable to a high degree of stand-
ardization. In contrast, field study protocols must retain a high degree of
flexibility. Because standardization of field studies is made difficult by
variables such as chemical mode of action, use pattern, crop type, method
of application and species density and diversity, this series of guidelines
provides only a general outline for field studies. Specific protocols must
be developed as needed and submitted to the Agency for review. However,
despite the variability among field studies, several key elements common
to most field studies can be identified. This guideline was prepared to
identify and discuss these elements as they pertain to terrestrial vertebrates,
and to provide a better understanding of the purpose of field studies.
(2) Objective of field studies, (i) The purpose of the field study is
either to refute the assumption that risks to wildlife will occur under condi-
tions of actual use of the pesticide or to provide some quantification of
the risk that may occur. The purpose is twofold because the FIFRA re-
quirement to determine unreasonable adverse effects implies the need for
a risk-benefit analysis. Thus, if the assumption of risk cannot be refuted,
and in order to refine the risk-benefit analysis, field studies should quantify
the adverse effects that would occur from actual use of the pesticide.
(ii)	A study designed to refute hazard is unusual in biological re-
search. Typically, an investigator is more concerned about concluding with
a high degree of confidence that an effect occurred, not that it failed to
occur. FIFRA specifies that a pesticide is to be registered only if EPA
determines it will not cause unreasonable adverse effects. While the dif-
ference between an objective of "will cause" and "will not cause" may
seem trivial, it substantially influences study design and the evaluation
of data.
(iii)	The adverse effects to wildlife that can result from the use of
pesticides can be classified as those that affect populations of wildlife and
those that affect individuals but not the entire population. Either of these
effects may warrant regulatory action, including cancellation or suspen-
sion. An adverse effect that results in a reduction in local, regional, or
national populations of wildlife species is clearly of great concern. A pes-
ticide that can repeatedly or frequently kill wildlife is also of concern even
if these repeated kills may or may not affect long-term populations. The
terrestrial field study, accordingly, must be designed to assess both of these
types of effects.
(iv)	The field study must be designed to provide data that show
whether wildlife species will not be affected significantly by a pesticide
under normal pesticide use practices. To achieve this objective fully at
the population level, detailed knowledge of the population dynamics and
2

-------
varying environmental conditions for each species potentially at risk must
be available. The theoretical aspects of population dynamics are well docu-
mented in the literature. However, empirical data are available for only
a few species (see paragraph (h)(8) of this guideline). A study designed
to provide the needed data would include information on age structure,
age-specific survival and reproductive rates, and the nature and form of
intrinsic and extrinsic regulatory mechanisms. Such a study, when coupled
with the influence of pesticide application on these parameters, would re-
quire several, if not many, years in order to begin to give meaningful
results. The cost of obtaining such data could make these studies imprac-
tical, if not impossible.
(v)	The essential question is: How can these studies be performed
in a practical, economical manner and still provide data that can show
that the pesticide under study will not reduce or limit wildlife populations
or repeatedly kill wildlife?
(vi)	The question can be answered by examining the potential influ-
ence pesticides can have on wildlife. These effects include:
(A)	Direct poisoning and death by ingestion, dermal exposure, and/
or inhalation.
(B)	Sublethal toxic effects indirectly causing death by reducing resist-
ance to other environmental stresses such as diseases, weather, or preda-
tors.
(C)	Altered behavior such as abandonment of nests or young, change
in parental care, or reduction in food consumption.
(D)	Reduced food resources or alteration of habitat.
(E)	Lowered productivity through fewer eggs laid, reduced litter size,
or reduced fertility.
(vii)	These effects can manifest themselves in a population through
reduced survival and/or lower reproductive success. However, if a field
study shows that actual use of a pesticide does not affect survival and/
or reproductive success or that only minor changes occur, it would seem
reasonable to conclude that the use of the chemical will not significantly
impact wildlife. Further, if a field study provides estimates on the mag-
nitude of survival and reproductive effects, it is possible to make reason-
able projections on the meaning of the effects to nontarget populations
by using available information on the species of concern and basic theories
of population dynamics. While less than ideal, field studies that collect
information on survival and reproductive effects and use these data to ad-
dress population parameters should provide a reasonable basis for evaluat-
ing potential impacts. This is not to imply that effects on populations are
3

-------
the only concern, however, a study adequate to assess these effect will
also assess the degree of risk to individual wildlife.
(viii) This guideline emphasizes avian and mammalian wildlife. The
Agency is also concerned about other terrestrial organisms such as
nontarget plants, invertebrates, amphibians, and reptiles. Plants and inver-
tebrates are excluded here from direct study, except as sources of food
or pesticides to wildlife. Testing guidelines for nontarget insects and plants
are in Groups C and D, respectively, of the 850 guideline series. Estab-
lished protocols, especially for acute and chronic toxicity testing, are avail-
able for birds and mammals, but not for reptiles and amphibians. Further,
the Agency assumes that "protection" for reptiles and amphibians is pro-
vided through the risk assessment process for birds and mammals. Occa-
sionally, however, it may be necessary to adapt these field techniques to
apply specifically to reptiles and/or amphibians.
(3) General approach, (i) Field studies required to support registra-
tion have evolved into two types, screening and definitive. The types of
studies required depends on the available data on the chemical in question.
If the available information is limited to laboratory toxicity data on a lim-
ited number of species, coupled with effective environmental concentra-
tions (EECs), a screening field study may be appropriate to determine if
impacts are occurring and, if so, to what species. If a screening study
indicates impacts are occurring, or if other available data suggest that dele-
terious effects have occurred or are extremely likely, the study design
should be quantitative, evaluating the magnitude of the impacts in a defini-
tive study. For some chemicals it may be appropriate to proceed directly
to a definitive study without the screening phase. Careful consideration
needs to be given to the likelihood of impacts occurring in order to deter-
mine which approach to use. When there is insufficient information to
indicate which species are at risk in the field but available data strongly
suggest adverse effects will occur, it may be appropriate for a field study
to begin with the general approach of a screening study, followed by a
quantitative phase that focuses on the species affected in the screening
phase. In certain instances there may be sufficient data and information
for the Agency to decide additional testing including field testing is not
necessary prior to a regulatory action.
(ii) Design differs between the screening study and the definitive
study. If the objective of the study is to determine if impacts are occurring,
"pass-fail" methods can evaluate whether or not animals are being
stressed by the application. These methods may include carcass searching,
residue analysis of species collected on study plots, residue analyses of
wildlife food sources found in and adjacent to the area of application, be-
havioral observations, and enzyme analysis. At the quantitative level (de-
finitive study), the objectives should include estimating the magnitude of
acute or secondary mortality caused by the application, the existence and
extent of reproductive effects, and the influence of pesticide use on the
4

-------
survival of species of concern. Methods that can be used to address these
objectives include mark-recapture, radio telemetry, line transect sampling,
nest monitoring, territory mapping, and measuring young to adult ratios.
(4) Sampling and experimental design. While examples of accept-
able experimental designs are given, it is beyond the purpose of this guide-
line to cover the fundamentals of this topic. Paragraph (f) of this guideline
provides a general outline for a field study protocol to be submitted to
the Agency for review. Paragraphs (d) and (e) of this guideline outline
points to be considered in designing screening and definitive field studies.
As stated under paragraph (c)(1) of this guideline, specific protocols must
be developed on a case-by-case basis and submitted to the Agency for
review.
(d) Screening study—(1) Objective and scope, (i) The screening
study is designed primarily to demonstrate that the hazard suggested by
lower tier laboratory or pen studies does not exist under actual use condi-
tions. The interpretations of screening study results, in most cases, are lim-
ited to "effect" versus "no-effect" determinations. If the study indicates
that the pesticide has caused little or no detectable adverse effect, it may
be reasonable to conclude that potential adverse effects are minor. When
effects are demonstrated, it may be necessary to determine the magnitude
of the effects, thus requiring additional testing if pesticide registration or
continued registration is still desired. Therefore, when information already
available shows that a product has caused adverse effects under normal
use conditions, the screening study may be of limited value. It may be
appropriate to proceed directly to a definitive field study in cases where
analysis of laboratory or other data strongly suggest that adverse effects
are likely to occur, and are unlikely to be attenuated by field use condi-
tions.
(ii)	Screening studies are limited to addressing the potential for acute
toxic effects, such as direct poisoning and death, and sublethal toxic effects
potentially affecting behavior and/or survival. In most instances, a screen-
ing study would not address chronic effects, such as reduced reproduction,
or effects such as changes in density or diversity of populations.
(iii)	Further laboratory and/or pen studies may be useful prior to pro-
ceeding to the field, or may be necessary to interpret results of the field
study. For example, additional toxicity data on species that are expected
to be exposed from the proposed use pattern may indicate which species
are more susceptible to the pesticide, allowing the study to be designed
to monitor those species in greater depth as well as to provide insight
into field results that show some species were affected more than others,
and additional laboratory studies may be unavoidable. If residue concentra-
tions in resident species are being used to indicate potential problems, the
relationship between tissue levels and the doses that causes adverse effects
must be estimated. If secondary poisoning is of concern, feeding secondary
5

-------
consumers (held in captivity) prey items collected in the field following
the application can be useful to evaluate this potential exposure route. Lab-
oratory toxicity tests for secondary consumers coupled with residue analy-
sis of prey items can indicate the potential for secondary poisoning of
nontarget species. In designing field studies, the utility of laboratory and/
or pen tests should not be neglected, and where appropriate their use is
encouraged.
(2)	Geographic area selection, (i) The selection of geographical
areas for evaluating pesticide impacts on wildlife can be difficult, particu-
larly for pesticides to be used on crops grown over large and diverse areas.
Studies should be performed in each biogeographic area where the pes-
ticide could be used. While this approach may be practical for uses re-
stricted to localized areas and conditions, many uses (e.g., corn, soybeans,
alfalfa) would require an inordinate number of studies in different geo-
graphic areas, due to the diversity and variability in wildlife species and
habitats involved. To keep the number of geographic areas at a manageable
level while still accomplishing the purpose of the field study, geographical
area selection should be biased toward situations likely to present the
greatest risk. If hazards appear to be low under these conditions, it can
be reasonably concluded that impacts under less severe conditions would
be minor.
(ii) A careful review of the species and habitats in the various geo-
graphical areas where the pesticide could be used is necessary to identify
the areas of highest concern. A sound understanding of the biology of
the species that are found in association with the potential use sites is
essential. Identifying these areas may require an extensive literature review
and consultation with experts familiar with the areas and species of con-
cern. The study area selected should be frequented by those species that
would have high exposure, based on their feeding or other behavioral as-
pects. If exposure and fate (e.g., degradation) parameters vary geographi-
cally, study area selection also should be biased towards maximizing resi-
dues available to wildlife. In some circumstances preliminary monitoring
of candidate areas may be necessary to determine which should be selected
for detailed study,
(3)	Study site selection, (i) Selection of study sites within each geo-
graphic area also is extremely important in designing field studies. Study
sites should be randomly selected throughout the study area. While this
approach may be practical for some areas such as rangeland or large con-
tiguous crops, random selection would require a large number of sites to
provide a representative sample due to the diversity and variability in wild-
life species and habitats in most areas. The cost and time requirements
of such studies would be unreasonable. To maximize the hazard, the sites
selected should have associated species that would be at highest risk from
the application, as well as a good diversity of species to serve as indicators
for other species not present at that specific location. The choice of study
6

-------
sites that are as similar as possible in terms of abundance, diversity, and
associated habitat will facilitate an analysis of the results.
(ii)	Under some circumstances, it may be difficult to decide before-
hand which species are likely to be at highest risk. In most cases, field
surveys of a number of sites may be needed to identify which sites should
be selected for detailed study. Even when high risk species can be identi-
fied, preliminary surveys may be needed to determine which sites have
adequate numbers of the high risk species as well as a good diversity
of other species.
(iii)	In general, study sites should be selected from what is considered
to be a typical application area, but at the same time, study sites should
contain the widest possible diversity and density of wildlife species. Identi-
fying potential study sites may require consultation with experts familiar
with the areas where studies are proposed, and preliminary sampling.
(iv)	In the initial evaluation of potential study sites, edge effect may
indicate which sites support the larger and more varied wildlife popu-
lations. As stated by Aldo Leopold (see paragraph (h)(23) of this guide-
line), "The potential density of game of low radius requiring two or more
types is, within ordinary limits, proportional to the sum of the type periph-
eries." (Type is defined as the various segments of an animal's environ-
ment used for food, cover, or other requirements.) If study sites are se-
lected to maximize edge effect the potential for high density and diversity
should be increased. One quantitative measure of edge and edge effect
(see paragraph (h)(15) of this guideline) is the distances around individual
plant communities in relation to the unit area of the community. Population
densities, in general, are positively related to the number of feet of edge
per unit area of community. Study sites chosen to maximize the ratio of
edge to core may serve to indicate sites with higher densities and diversi-
ties of wildlife species.
(v)	While this ratio can be helpful in selecting study sites, the other
characteristics of edge should not be neglected in screening potential study
sites. Density and diversity of wildlife species are also influenced by the
variety in the composition and arrangement of the edge component cover
types and its width. Also, interspersion (the plant types and their associa-
tion with one another) influences densities of wildlife species. The edge
effect is the sum of all the characteristics of edge and hence each compo-
nent needs to be considered. An agricultural field with a relative high edge
to core ratio may not have as high a density and diversity as one with
a lower ratio but greater variety, width, and interspersion. In general, edge
characteristics can be used to screen potential study sites; however, pre-
liminary sampling of prospective study sites will be needed to identify
study sites with adequate density and diversity of wildlife species.
7

-------
(4) Number of sites, (i) The number of sites needed can be estimated
using the binomial theorem. Briefly, the rationale is that for each study
site there are two possible outcomes, either effect or no-effect. Trials of
this type are known as binomial trials and, when repeated, the results will
approximate a binomial distribution. In this case, to use the binomial theo-
rem, it is necessay first to define the expected probabilities that birds or
mammals on a site are affected or not affected, after which the probability
of the discrete binomial random variable x for n replications can be used
to determine the minimum number of sites at a certain level of signifi-
cance.
(ii)	For purposes of illustration, a problem exists if some specific mor-
tality rate or level of some other variable occurs on more than 20 percent
of the potential application sites. Translated into binomial probabilities,
there is a 0.2 probability of a site showing an effect and a 0.8 probability
of a site not showing an effect. Therefore, if the results from the field
trial show that the number of sites affected is significantly lower than 0.2
n, it can be concluded that potential impacts will be below the stated level
of concern.
(iii)	To calculate the minimum number of sites necessary to show
a significant difference between the observed and expected, the following
formula for the probability of the binomial random variable x can be used
(see paragraph (h)(36) of this guideline):
where x = number of sites showing effects, n = number of sites, p = prob-
ability of a site showing an effect, and q = probability of a site not showing
an effect.
Then, solving for n, when x = 0, i.e.,
(iv) The minimum number of sites can be determined using this for-
mula. Continuing with the discussion example of 20 percent occurrence
of an effect as a level of concern (i.e., a 0.2 probability of an affected
Let P(x = 0) = a, then
a = qn
log a = n log q
n = log a/log q
8

-------
site, a 0.8 probability of a nonaffected site, and a 0.05 level of signifi-
cance), n would be:
9

-------
n = log 0.05/log 0.8
n = 13.43
Therefore, 14 is the minimum number of sites needed such that the prob-
ability is not greater than 0.05 that all sites surveyed would be unaffected.
In other words, if 20 percent of the application sites are actually affected,
there is only a 5 percent chance of finding all 14 sites unaffected when
n = 14. Moreover, if 20 percent of the application sites are actually af-
fected, we expect to find 1, 2, 3, and 4 sites affected with probabilities
of 0.15, 0.25, 0.25, and 0.17, respectively, when n = 14.
(v)	Under many circumstances, conducting this number of replications
may not be practical. However, the number of sites can be reduced if
site selection is biased toward hazard. While arguable, it seems logical
that if the worst cases are sampled, a less stringent level of significance
could be accepted. While this must be determined on a case-by-case basis,
the Agency believes a minimum acceptable level of significance under
worst-case conditions is 0.2 rather than 0.05 under average or normal use
conditions. At this level, eight sites showing no effect would be required
to conclude at the 0.2 level of significance that the effect occurred on
less than 20 percent of the application sites or—there is less than a 20
percent chance that all eight sites will be judged unaffected when n =
8 sites. Under some circumstances, this may not seem adequately protec-
tive. It should be noted, however, that based on this same design, it could
be concluded that, at the 0.1 level of significance, effects occur on less
than 30 percent of the application sites, and at the 0.05 level of signifi-
cance, effects occur on less than 40 percent of the application sites. Hence,
with eight sites, it could be concluded with a relatively high degree of
confidence that effects would occur on less than 40 percent of the applica-
tion sites. Because worst-case study sites were used, the Agency could
have additional confidence that adverse effects would occur on less than
20 percent of all normal application sites.
(vi)	Under some circumstances, particularly if endangered species
could be exposed from the proposed use, additional replication may be
desirable. Under these conditions, a high degree of confidence that an ef-
fect was a rare occurrence would be required. (Under no circumstances
should field studies on pesticides be conducted in areas where endangered
species could be exposed.)
(vii)	The calculations under paragraph (d)(4)(iii) of this guideline are
for when x = 0, no effects are observed on any site. A similar approach
can be used to estimate the number of sites necessary to show a significant
result for a critical value of x > 0. Again the formula for the probability
of the binomial random variable can be used summing the probabilities
of x and all outcomes 
-------
ber of replications required to show statistical significance may be deter-
mined for a given level of significance for individual x values. That is:
P(X
-------
(ix)	Under any condition, it is extremely important with the binomial
approach to define the critical or threshold level for an effect, and to be
sure that the methods used are sensitive enough to detect an effect should
one occur. These assessments depend upon the species potentially at risk
as well as the parameter being sampled. It should be noted that the measure
of effect is not limited to mortality. Other parameters, such as residue
or enzyme levels, can be used. Whatever parameters are used, defining
the criteria level for an effect is extremely important, and when designing
studies this issue should be considered carefully.
(x)	Using this approach, control (reference) sites are not an absolute
necessity. While the Agency encourages their use, in some cases the addi-
tional information gained from the control sites for a screening study may
not justify the additional effort required. In most instances, control sites
would serve to protect from erroneously attributing effects due to other
causes to the pesticide. However, for most chemicals, this can be avoided
by employing methods, such as residue analysis and/or cholinesterase
(ChE) inhibition tests, that can be used to indicate if the pesticide contrib-
uted to the observed effect. Further, because studies have shown that it
is a relatively rare event to locate dead or sick animals in the wild except
under unusual conditions (see paragraph (h)(18) of this guideline), it is
unlikely to find dead animals that were killed by something other than
the pesticide being tested.
(xi)	Nevertheless, controls may be necessary when reliable methods
to confirm the cause of effect are not available. In these cases the binomial
design can be modified to a paired-plot binomial design, with a treatment
plot and a comparable control plot for each study site within an area. When
critical levels of effect and occurrence are defined, the binomial theorem
can be used for sample size determination, which gives 8 site pairs (16
paired plots) showing less than a defined difference between plots to con-
clude at the 0.2 level of significance that the effect occurred on less than
20 percent of the application sites. Alternatively, a quantitative difference
or, preferably, ratio of treated to control responses could be used to test
for a treatment effect on each of the measured response variables. (This
is discussed further under paragraph (e)(2) of this guideline.)
(5) Size of study sites. Study sites must be large enough to provide
adequate samples. The size is dependent on the methods used, the sensitiv-
ity required, and the density and diversity of species and their ranges. In
some cases, particularly with slow-acting poisons or where species at high
risk have relatively large home ranges, areas several times larger than the
treatment area may need to be examined. In some circumstances, several
fields in an area may be included in a single study site to account for
wide-ranging species of lower densities. Except in the unusual cir-
cumstance where fields are extremely large (e.g., forested and range areas),
the study site should never be less than an individual field and the sur-
rounding area. The nature of the surrounding area is discussed further
12
-------
under individual methods. Another consideration is the distance between
study sites—in general, sites should be separated adequately to ensure
independence, which is dependent mainly on the range of the species that
could be exposed.
(6)	Chemical application, (i) Consideration should be given to appli-
cation rates and methods. In general, the test conditions should resemble
the conditions likely to be encountered under actual use of the product.
In most instances the pesticide should be applied at maximum use rates
and frequencies specified on the label.
(ii)	If more than one application method is specified on the label,
the method that maximizes exposure of nontarget species should be used.
This evaluation should relate wildlife utilization of the area to exposure.
For example, if the crop is one that is used by avian species as preferred
nesting areas, feeding areas, or cover, ground application may be the meth-
od that maximizes exposure. However, if it is a crop with low utilization
by wildlife species, but with high utilization of its edges, aerial application
where drift could increase exposure may be more appropriate. In any case,
the method of application used must be consistent with the label.
(iii)	Equipment used may influence potential exposure of nontarget
species. There is a diversity of types of farming equipment that, depending
on the particular use pattern involved, could influence exposure. For exam-
ple, for pesticides applied in-furrow at planting there are several types of
covering devices employed on seeders, such as drag chains, drag bars,
scraper blades, steel presswheels, etc., in which the efficiency may vary
for covering the pesticide. In general, the various equipment normally used
for the particular pesticide application has to be evaluated to estimate the
potential influence of equipment choice on exposure. In some instances,
preliminary tests may be required to estimate which method and equipment
poses the highest exposure.
(7)	Methods, (i) This section provides a general outline of methods
appropriate for use in a screening field study and indicates some of their
limitations. While methods described have been found to be most useful,
a screening study is not limited to these methods. If other methods are
more appropriate, their use is encouraged. Because procedures should be
adapted to specific situations, the outlines presented should not be inter-
preted as strict protocols. Normally, different methods will be combined
to evaluate potential impacts. Due to the indefinite number of variables
and the unpredictability of wild animals, even normally reliable procedures
can sometimes prove inadequate.
(ii) The methods used in a screening study address exposure by mon-
itoring overt signs of toxicity such as mortality or behavioral modifica-
tions, or by evaluating parameters that indicate animals are under stress,
such as residue concentrations in tissues or degree of enzyme inhibition.
13
-------
Measurements of density and diversity of species are needed to aid in
evaluating the results. The following methods can be useful for screening
studies.
(A)	Carcass searches. (7) Searching for dead or moribund wildlife
is a basic method used in field studies to evaluate the impact of pesticides
on nontarget species. Carcass searches can roughly indicate the magnitude
of kills when adequate areas are searched and the reliability of the search
is documented. This latter point is extremely important. Rosene and Lay
(under paragraph (h)(30) of this guideline) indicated that finding even a
few dead animals suggests that there has been considerable mortality; fail-
ure to find carcasses is poor evidence that no mortality has occurred. The
reliability of the search is based upon the percentage of carcasses recov-
ered by searchers and the rate of disappearance. By knowing the reliability,
the significance of failure to find carcasses can be assessed and the extent
of the kill estimated.
(2)	Finding dead animals is seldom easy, even if every animal on
a site is killed. For example, three breeding pairs of small birds per acre
is considered a large population (see paragraph (h)(18) of this guideline),
and under average cover conditions, a small bird is difficult to detect.
Small mammals may be more abundant but, due to their typically secretive
habits, they are more likely to die under cover and can be even more
difficult to find than birds. Carcass searching specifically for mammals
should be attempted only when cover conditions permit reasonable search
efficiency. However, any vertebrate carcasses found should be collected,
even if the search is oriented primarily to one taxon.
(3)	Because results may be biased by scavenging and failure to find
carcasses, the sensitivity of this procedure should be determined. Under
conditions of heavy cover and/or high scavenger removal, other methods
may be more appropriate.
(4)	There are no standard procedures for carcass searches. Paragraph
(g) of this guideline outlines practices that have been used typically and
should be considered in designing searches.
(B)	Radio telemetry. (7) Radio telemetry has been found to be ex-
tremely useful for monitoring mortality and other impacts caused by pes-
ticide exposure of wildlife. Advances in miniaturizing electronic equip-
ment have made it feasible to track most vertebrate animals. Transmitters
weighing a few grams have been used to track species as small as mice.
Cochran's excellent summary of this technique provides additional details
(see paragraph (h)(5) of this guideline).
(2) Radio telemetry has the advantages of providing information on
the fate of individual animals following a pesticide application and of fa-
cilitating carcass recovery for determining the cause of death. Although
the initial cost of this technique may be more than for other methods,
14
-------
the increase in information obtained under some circumstances can more
than justify the cost. The method is particularly useful with less common
or wide-ranging species.
(3)	In addition to mortality, radio telemetry can be used to monitor
behavioral modification as well as physiological changes. Automatic radio-
tracking systems permit continual surveillance of the location of animals
(see paragraph (h)(5) of this guideline), which can be used to provide in-
sight into behavioral changes such as nest abandonment, desertion of
young, or decreases in activities such as flying or feeding. Radio telemetry
equipment is also available for the transmission of physiological data such
as heart rates or breathing rates (see paragraph (h)(25) of this guideline).
(4)	While this technique can provide very useful information on im-
pacts of pesticides to wildlife, other points need to be considered in addi-
tion to cost. Capturing animals alive and unharmed requires time, skill,
and motivation. For the method to be consistently successful, the investiga-
tor must be thoroughly familiar with the habits of the species under study
and with the various capture methods that can be used. Even for the most
experienced investigator, adequate sample sizes can be difficult to obtain.
(5)	Adequate sample size is very important. The binomial theorem
can be used to estimate minimum sample size per site, if the question
is limited to mortality. Briefly stated, to be sure that nontarget species
are not being affected by environmental concentrations greater than, for
example, an LC20, the expected binomial probabilities would be 0.2 for
mortality and 0.8 for nonmortality. Depending on the level of significance,
8 (a = 0.2) to 14 (a = 0.05) individuals would need to be monitored
per site (see paragraph (d)(4) of this guideline for further details on these
calculations). However, since the LC20 may differ between species, 8 to
14 individuals would be required for each species, unless laboratory tests
have documented relative species sensitivity. Further complications can
arise if radiotagged animals leave the area or if the movements of individ-
uals limit their exposure. If these complications occur at relatively low
rates, a few additional radio-tagged animals may be sufficient to overcome
these problems.
(C) Tests of ChE inhibition. (7) Measuring ChE concentrations in
animal tissues has been found to be a very useful field technique for evalu-
ating exposure of nontarget animals to ChE-inhibiting-chemicals (see para-
graphs (h)(18) and (h)(19) of this guideline). These chemicals, including
organophosphates and carbamates, affect the synaptic transmission in the
cholinergic parts of the nervous system by binding to the active site of
acetylcholinesterase (AChE), which normally hydrolyzes the neuro-trans-
mitter acetylcholine. Thus, ChE inhibitors permit excessive acetylcholine
accumulation at synapses, thereby inhibiting the normal cessation of nerve
impulses (see paragraphs (h)(6) and (h)(27) of this guideline).
15
-------
(2)	The depression of AChE activity, when measured and compared
to controls, can indicate the degree to which an animal is affected. Brain
ChE depression of >50 percent in birds has been found sufficient to as-
sume that death is pesticide-related (see paragraph (h)(24) of this guide-
line). Depressions of more than 70 percent are often found in dead birds
poisoned by these chemicals (see paragraphs (h)(1), (h)(2)), and (h)(33)
of this guideline), although some individual birds with less than 50 percent
inhibition may die. A 20 percent depression of brain ChE has been sug-
gested as an indication of exposure (see paragraph (h)(24) of this guide-
line). ChE concentrations in blood can also be used to indicate exposure,
avoiding the necessity of sacrificing the animal. However, blood ChE con-
centrations are influenced more by environmental and physiological factors
than are brain ChE concentrations. Because ChE activity varies among
species, the degree of depression must be based on an estimated normal
value for concurrently tested controls of the species potentially at risk.
Because of this difference between species, each case must be considered
unique (see paragraph (h)(19) of this guideline).
(3)	Although there are several calorimetric methods for determining
ChE activity, the general methods are similar. Brain tissues (or blood sam-
ples) are taken and analyzed for ChE concentrations. Comparisons are
made between pre- and posttreatment and between treated and untreated
areas. It is important to ensure that untreated controls have not been ex-
posed to any ChE inhibitors. It also should be noted that absolute enzyme
levels in the literature are derived from various different, although similar,
methods and are reported in different ways. For example, Ludke et al.
(see paragraph (h)(24) of this guideline) used a modification of the method
reported in paragraph (h)(13) of this guideline and reported results of ChE
activity as nanomoles of acetylthiocholine iodide hydrolyzed per minute
per milligram of protein, whereas Bunyan et al. (see paragraph (h)(1) of
this guideline) used their own colorimetric method (in addition to a pH
change method) and reported the results as micromoles of acetylcholine
hydrolyzed per hour per milligram of protein. Therefore, without a tightly
standardized method, it is necessary to use concurrent controls of the same
species obtained from the general vicinity (but untreated) of the exposed
birds, rather than literature values. Because of the greater variation in plas-
ma ChE levels than for brain, more controls are necessary to evaluate
blood samples.
(4)	Tests for ChE activity can be used to help confirm cause of death
and monitor levels of exposure. In the latter case, 5 to 10 individuals of
each species are collected before treatment and at periodic intervals follow-
ing treatment. Mean inhibition of 20 percent or more is considered an
indication of exposure to a ChE inhibitor. Confirmation of cause of death
may be determined by analyzing brain tissue from wildlife found dead
following treatment and comparing the activity with controls. Inhibition
of 50 percent or more is considered strongly presumptive evidence that
16
-------
mortality was caused by a ChE-inhibiting compound. The cause-effect re-
lationship can be further supported by chemical analysis of the contents
of the digestive tract or other tissues for the chemical in question.
(5) For this technique to provide accurate information, prompt collec-
tion and proper preservation of specimens are essential. ChE concentra-
tions in tissues are influenced by time since death, ambient temperatures,
and whether or not reversible ChE inhibitors are being investigated. There-
fore, the response of postmortem brain ChE to ambient conditions can
seriously affect diagnosis of antiChE poisoning. Samples must be collected
shortly after death and frozen immediately to halt changes in tissue or
enzyme-inhibitor complexes. A technique for field monitoring and diag-
nosis of acute poisoning of avian species has been reviewed, discussing
sample collection, sample numbers, preservation procedures, and sources
of error (see paragraph (h)(19) of this guideline).
(D) Residue analysis. (7) Residue analyses of wildlife food sources
provide information about the level and duration of pesticide exposure.
Residue analysis of animal tissues also can indicate actual exposure levels.
If the relationship between tissue concentrations and toxic effects is known
for the species in question, residue analyses can provide a measure of the
degree to which animals are affected. For this application of residue analy-
ses, laboratory trials are necessary to establish the relationship between
residue levels and toxicity. In addition to death, these laboratory trials
should include such signs as anorexia, asthenia, asynergy, or ataxia. For
chemicals that are readily metabolized by vertebrates, residue analysis may
not be appropriate for diagnostic purposes. With many pesticides, it will
be necessary to analyze for residues of active metabolites also.
(2)	For determining residues on wildlife food sources, the investigator
should collect samples of insects, seeds, leafy parts of plants, etc., imme-
diately after pesticide application and at periods thereafter. Samples should
be analyzed for the chemical to determine potential exposure rate and dura-
tion. The application method needs to be considered in determining where
to take samples. If drift is likely, samples should be taken from habitats
surrounding the treatment sites as well as in the treated fields. Because
analysis can be costly, the investigator should consider carefully the num-
ber of samples necessary to provide adequate data. Where feasible, sam-
ples from different locations within a site should not be pooled. Separate
analysis of samples can provide data on the range and variability of expo-
sure as well as mean levels.
(3)	When residue analysis is used to evaluate exposure in nontarget
animals, the tissues selected for analysis differ depending on the purpose.
Heinz et al. (see paragraph (h)(18) of this guideline) indicated that for
many chemicals, residues in brains of birds and mammals can be used
to determine if death is pesticide-related. The authors believe that sublethal
exposure is judged better from residues in other tissues. Therefore, they
17
-------
proposed that analyses of whole body homogenates should be used to
quantify the body burden of a pesticide. If this is not feasible, analysis
of muscle tissue is suggested, because muscle residues reflect body burden
more nearly than those of any other tissue, and the amount of muscle
tissue is not unduly large. For persistent chemicals, it has been suggested
that residues in liver and fat tissues could be misleading for determining
acute body burdens. Liver is a processing organ and its residue level large-
ly represents current availability of the chemical. Residues in fat are great-
ly affected by changes in the amount of body fat, and are not dependable
indicators of body burden of the chemical; however, for some chemicals,
liver, fat, or other tissues may be good qualitative indicators that exposure
did occur. In general, laboratory trials or data gathered in metabolism or
other studies may be necessary to determine which tissues can provide
the most useful information. Residue analysis of eggs taken from nests
in treatment areas can indicate the degree of contamination that a treatment
has caused, as well as possible reproductive effects of the treatment.
(4)	Two approaches may be used to determine the number of samples
to be collected. Frequently, residue samples will be collected to establish
a mean value and confidence limits. To determine the number of samples
to collect, it is necessary to estimate the standard deviation and to set
an arbitrary limit from the mean value that is acceptable. Although the
mean value does not need to be estimated, it is also necessary to have
some idea of the mean so that the standard deviation can be estimated
and the limit can be set. The formula for the number of samples, presented
under paragraph (h)(34) of this guideline, is
n = 4g2/L2
for 95 percent probability, where a is the standard deviation and L is the
allowable limit around the mean. For example, to calculate the residue
concentrations on vegetation within ±10 ppm, with an estimate of the
standard deviation of 20 ppm, then n = 4(20)2/(10)2 means 16 samples
are required to have a 95 percent probability that the sample mean value
will be within ±10 ppm of the true mean.
(5)	In some situations, there may be little information useful for esti-
mating the standard deviation, or the standard deviation may be rather
large, thus requiring a very large sample size. For some types of samples,
such as residues in nontarget wildlife carcasses, the sample size cannot
be increased to permit more precision. The mean value of a parameter
certainly has utility, but it also is very important to establish confidence
limits around the mean. In general, the Agency will use the 95 percent
confidence limits (usually the upper boundary, as in the case of residues)
in the assessment of the data. This approach will substantially reduce the
impact of outliers but will still incorporate the range of reasonable values
into the assessment. In addition, the use of confidence limits reduces the
necessity for taking a large number of samples. Because the width of the
18
-------
confidence intervals decreases with increasing sample sizes, investigators
should take samples that are as large as feasible.
(6)	Since the sample size will nearly always be less than 30, the cal-
culation of confidence limits should be based on Student's t-distribution.
The t values are derived from tables available in most statistics books,
and the 95 percent confidence limits are:
X±(t05)(s-V^)
where s is the standard deviation estimated from the sample of size n.
(7)	Alternatively, the binomial approach may be used for determining
if residues, typically in collection of live nontarget animals, exceed a par-
ticular threshold value that indicates an effect. The required sample size
is the same as presented for the binomial approach in determining the
number of study sites. Specifically, in the preceding example a minimum
of 8 samples with none exceeding the threshold value or 14 samples with
one or none exceeding the threshold value indicates no effect at p = 0.2
in 20 percent of the samples. This approach requires the establishment
of threshold values which are determined on a case-by-case basis. In gen-
eral, residues reflecting an LC20 level of exposure would seem to be a
maximum acceptable effect concentration for a screening study. An LC50
should be determined in the laboratory for each species analyzed for resi-
dues, after which a group of animals would be exposed to an LC20 con-
centration to determine the mean threshold concentration of residues. Since
this approach is impractical for a screening study, it is suggested that the
mean residue concentration in bobwhite and/or mallards exposed to an
LC20 dietary concentration would provide an indication of threshold lev-
els.
(8)	The number and timing of collection periods must be considered
and should be based on the persistence of the specific chemical under
study. Where persistence in the field has not been adequately determined,
it may be necessary to sample at regular intervals (e.g., days 0, 1, 3, 7,
14, 28, 56) to provide data on degradation rates.
(E)	Behavioral observations. Observations of behavior sometimes
can be an extremely important indicator of treatment effects. Such observa-
tions might include characteristic signs of toxicity or behavioral changes
seen in test animals exposed to the pesticide in the laboratory. Other abnor-
mal behavior (e.g., territorial males abruptly ceasing singing, birds not
feeding, reduced avoidance of humans) also may be important.
(F)	Density and diversity estimates. (7) It is necessary to know the
number of individuals and variety of species on and around a study site
in order to indicate which species could have been exposed and to aid
in evaluating the significance of mortalities or other findings. In addition,
19
-------
preliminary information on density and diversity is necessary for site selec-
tion and to determine the size of study sites. Under some circumstances,
comparisons of density estimates between treatment and control sites, or
between before and after treatments, may be used to indicate pesticide
impacts. In general, the usefulness of these comparisons is limited in a
screening study due to the relatively small acreage involved. If mortality
occurs, replacement from outside is likely to be so rapid that losses are
replaced before censuses are completed. Seasonal changes, such as migra-
tion, molt, or incubation, that can affect real or apparent densities must
be considered.
(2)	Several techniques may be used to estimate the density and diver-
sity of wildlife species, including counts of animal signs, catch-per-unit
effort, mark-recapture, and line-transect sampling. Although the methods
selected depend on the species of concern, for the screening field test line
transect methods are likely to be the most useful for birds.
(3)	The major advantage of line-transect sampling is that it is rel-
atively easy to use in the field once a proper sample of lines has been
chosen. However, line-transect-sampling is not applicable to all species,
particularly those that are not easily observed. Individuals using line-
transects must be extremely competent in species identification.
(4)	In the line transect method, an observer walks a distance (L)
across an area in nonintersecting and nonoverlapping lines, counting the
number of animals sighted and/or heard (N), and recording one or more
of the following statistics at the time of first observation:
(z) Radial distance from observer to animal.
(zz) Right-angle distance from the animal sighted or heard to the path
of the observer or angle of sighting from the observer's path to the point
at which the animal was first sighted or heard.
(5)	Although the field procedures are simple, they must be understood
adequately and implemented well to obtain good estimates of density (see
paragraph (h)(3) of this guideline). The authors provide a thorough review
of the theory and design of line-transect-sampling and this monograph
should be reviewed for details.
(6)	For mammals, density and diversity estimates from capture data
may be the most practical for a screening study. There are several ways
of estimating the populations from capture data, some relatively simple,
that may provide adequate information for a screening study. Davis and
Winstead (see paragraph (h)(7) of this guideline), review the various meth-
ods available, explaining their advantages and disadvantages.
(8) Interpretation of results, (i) The numerous variables involved
in field studies makes a meaningful discussion of the interpretation of re-
20
-------
suits somewhat tenuous, particularly with the almost inexhaustible array
of results that could occur. Each study must be considered unique and
therefore will require a case-by-case analysis that incorporates not only
the actual study but other relevant information that is available.
(ii)	In general, the results of the screening field study should provide
information on acute poisoning and potential sublethal effects as suggested
by enzyme, residue, or other measurements. In addition, information has
to be developed on the density and diversity of species on the study sites
as well as the sensitivity of the methods used. If no effects are detected,
assuming that the methods used were adequate to detect levels of concern
and that the species on the study site represent a good cross-section of
the nontarget species expected to be at risk, the potential hazard indicated
by lower tier tests is refuted. Unless other hazards (e.g., reproduction) are
still of concern, additional tests would not normally be necessary. How-
ever, if an effect is detected on one or more study sites at rates equal
to or greater than concern levels, the hazard has not been refuted and addi-
tional tests may be necessary.
(iii)	In interpreting if an effect has occurred in the context of the
binomial approach, care must be employed not to assume a level of preci-
sion in results that does not exist. Most detectable effects will exceed the
concern level for some methods used in these studies, due to the inherent
variability in the data collected to estimate the level of impact, particularly
when minimum sample sizes and areas are used. In some instances in in-
terpreting results it may be appropriate to use confidence limits of data
collected (or another measure of dispersion) to evaluate if concern levels
are exceeded. For example, when density estimates are used to estimate
percent mortality using the number of dead animals found during carcass
searching, the upper and lower confidence limits of the density estimate
may be more appropriate than the average, particularly when variability
of the density estimate is high. Whatever method is used, when effects
are detected that exceed concern levels they will be put into perspective
in the context of the entire study as well as other available information
to determine if or what additional data are needed. A "no-pass" result
does not necessarily mean that definitive field testing is automatically re-
quired.
(iv)	For example, a test may be run in an area where a species is
abundant, yet on a specific study site their numbers may be sufficiently
small that a single death exceeds the level of concern on that site. Statis-
tically, such a finding would indicate that the study did not pass according
to the binomial approach, and this would be the preliminary interpretation.
However, if the other sites had an adequate number of this species as
well as other species expected to be at risk and no other signs of impacts
are observed, the implications of the mortality would seem minor. On the
other hand, if diversity of species were extremely limited, it would have
greater significance. In other situations where the one dead bird is of a
21
-------
species with small numbers on most sites, but density and diversity of
other species is representative of nontargets expected to be at risk, another
screening study that looks at the species in which the effect was detected
may be appropriate. Conversely, a screening study showing that there is
appreciable mortality on most study sites may be sufficient for the Agency
to consider regulatory action.
(v)	In summary, the interpretation of results will go beyond the statis-
tical evaluation since the Agency must consider all the factors and cir-
cumstances peculiar to each test and site. The biological interpretation of
results is, and probably always will be, a matter of scientific judgment
based upon the best available data. In general, the judgmental aspects of
biological interpretation are more important for definitive studies than for
screening studies. Nevertheless, biological considerations often will be rel-
evant to screening studies. Study conclusions must integrate that which
is biologically significant with that which is statistically significant.
(vi)	Another consideration in the interpretation of results of a field
study is the attribution of effects to the pesticide being studied. A well-
designed study will include appropriate techniques to determine if an effect
is caused by a pesticide. In the absence of such techniques, the Agency
has no choice but to consider that any effects were as a result of the pes-
ticide use. As an example, measurement of ChE levels can provide infor-
mation, since it is generally accepted that inhibition of 20 percent indicates
exposure and inhibition of 50 percent or more indicates, in birds, that mor-
tality is due to an inhibitor (see paragraph (h)(4) of this guideline). If the
test chemical is the only ChE inhibitor used in the vicinity of the study
site, it can be reasonably assumed that a mortality associated with 60 per-
cent ChE inhibition is due to the test chemical. However, if other ChE
inhibitors are used near the site, additional information, such as residue
measurements, may be necessary to attribute death to the specific ChE
inhibitor being tested.
(e) Definitive study—(1) Objective and scope, (i) The definitive
study is a relatively detailed investigation designed to quantify the mag-
nitude of impacts identified in a screening study or from other information.
In contrast to the screening study, which monitors mainly the proportion
of the local population that is expected to be exposed, the definitive field
study examines a sample of the entire local population in the treated area.
Although a definitive study may be performed when laboratory studies
indicate a high potential for field mortality, it is more likely to be requested
when there is evidence that actual field mortality has occurred, as in a
screening study, or where reproductive effects are being investigated. The
objectives of the definitive study are:
(A) To quantify the magnitude of acute mortality caused by the appli-
cation.
22
-------
(B)	To determine the existence and extent of reproductive impairment
in nontarget species from the application.
(C)	To determine the extent to which survival is influenced.
(ii)	Due to the intense effort and time required to estimate these pa-
rameters, the definitive study should be limited to one or a few species
believed to be at the highest risk. If it can be shown that minimal (as
defined at the onset of the study) or no changes occur in study parameters
to high risk species, there is likely to be minimal potential for adversely
affecting other presumably low risk species from use of the pesticide in
question.
(iii)	The definitive study, in addition to estimating the magnitude of
effects of acute toxicants, also can be applied to estimating the magnitude
of chronic or reproductive effects. Although the discussion has emphasized
chemicals that are acutely toxic, with few exceptions it is applicable to
chemicals that cause chronic effects.
(iv)	In general, the definitive study will provide limited insight into
whether or not effects are within the limits of compensation for the species
of concern. However, using the data collected in these studies coupled
with available information on the species of concern and basic theories
of population dynamics, the meaning of the observed effects on the species
can be evaluated.
(2) Sampling and experimental design, (i) The principles of statis-
tical design of studies are well documented and it is beyond the scope
of this guideline to cover the fundamentals of this topic. However, there
are a few points on this topic that warrant discussion relative to the defini-
tive study.
(ii) In the design of field studies, it is necessary to consider carefully
what constitutes a sampling unit. Eberhardt, under paragraph (h)(9) of this
guideline, points out that special problems are faced in designing experi-
ments on wild animal populations. Study sites must be large in order to
limit the influence of boundary effects, such as movements into and out
of the area. Large study sites can be very expensive both in terms of actu-
ally applying the experimental treatment and in the assessment of results.
Eberhardt also states that numerous observations, even a full year of data,
on a single study site may result in very sound values for that site, but
do not provide a basis for inferences to other sites. Hurlbert (see paragraph
(h)(20) of this guideline) has discussed the problems associated with field
studies where there was no replication or replicates were not statistically
independent, which he terms pseudoreplication. Of the field studies he
evaluated, 48 percent of those applying inferential statistics had
pseudoreplication.
23
-------
(iii)	According to Eberhardt, under paragraph (h)(9) of this guideline,
lack of replication seems to be based on the mistaken assumption that
variances based on subsampling of sites (intrasite variability) are suitable
bases for comparing treatment effects (intersite variability). This, he be-
lieves, is not a valid basis for a statistical test, because it is the variance
of sites that are treated alike that is relevant to a test of treatment dif-
ferences. Although subsampling of sites may be necessary to collect the
data, it is the difference between sites that is important for analysis.
(iv)	An important point to consider in designing a definitive study
is to be sure that the study will detect a substantial impact when, in fact,
it occurs. In statistical terms this concept is referred to as the power of
the test. Experience with classical experimental designs with random as-
signment of experimental treatment and controls, has shown that the prob-
ability of a Type II error is generally high (unless very large numbers
of replicates are available). Eberhardt, under paragraph (h)(9) of this guide-
line, indicates that, all too often in field studies on impacts to wildlife,
either by default or lack of understanding, there is only a 50 percent chance
of detecting an effect, which he likens to settling the issue by flipping
a coin and doing no field study whatsoever. Since a definitive study is
carried out under the assumption that effects will occur, the Agency be-
lieves minimizing Type II errors is extremely important.
(v)	As suggested, the more generally used experimental designs re-
quire inordinately large sample sizes to obtain small Type II errors. For
example, based on a coefficient of variation of 50 percent (a relatively
homogeneous sample for the kinds of data collected in field studies;
Eberhardt (see paragraph (h)(8) of this guideline), a 20 percent minimum
detectable difference between means, a Type II error of 0.2 and a Type
I error of 0.05, the number of replications required can be estimated as:
n = (Zi_a + Zi_p)2 (CV)2 [ 1 + (1 - 8)2]/82
where
n = number of replications
Zi-a and Zi-p are critical values of the unit normal distribution
CV = coefficient of variation
8 = detectable mean difference expressed as the proportion of the
control group mean, i.e., 8 = (fii - (12) + |-ii
For the above example:
n = (1.65 + 0.84)2 (0.5)2 [1 + (1 - 2)2]/(0.2)2
n = 63.6 replicates
24
-------
Thus, for the above parameter, 64 replicates for both control and treatment
groups, or 128 total study plots, are required to detect a 20 percent dif-
ference between treatments and controls with an 80 percent chance of
being sure to detect a real difference at a 0.05 level of significance.
(vi)	With more sophisticated designs, the number of replicates can
be reduced under some circumstances and still meet the Agency's aim
to limit the probability of a Type II error to 0.2 with a detectable difference
of 20 to 25 percent. For example, a paired plot design can be used, sub-
stantially reducing the number of replicates required. Pairing serves to re-
duce the effective coefficient of variation by reducing the variation attrib-
utable to experimental error. The lower coefficient of variation reduces
the number of replicates. A quantitative difference or, preferably, a ratio
of treated to the total of treated and control responses, can be analyzed
statistically to test for a treatment effect on the measured response vari-
ables (see paragraph (h)(31) of this guideline). The logic of using paired
plots is that, while no two areas are ever exactly alike, two areas that
are not widely separated in space are ordinarily subjected to much the
same climatic factors, have populations with about the same genetic make-
up, and generally the two populations can be expected to follow much
the same trend over time, apart from a pesticide effect (see paragraph
(h)(8) of this guideline)). If all plots are approximately equal in area and
habitat and population densities between pairs are similar, it is postulated
that when no pesticide impacts occur, the mean ratio of treatment to treat-
ment plus control will equal one-half. Then a t-test or an exact
randomization test (see paragraph (h)(l 1) of this guideline) may be applied
to test whether the average number of survivors on the treated plots is
equal to the average number of survivors on controls.
(vii)	The number of pairs required can be estimated using the follow-
ing formula:
n = (Zi-a + Zi-p)2 x [4 qi/(pi)2 (1 + qi) c]
where
n = number of paired plots
Zi-a and Zi_p are critical Z scores
qi = survival ratio
pi = mortality ratio
c = mean number of survivors on control plots
(viii)	Therefore, at an 80 percent assurance of detecting a treatment-
induced impact of 20 percent or greater at a 0.05 level of significance
if c = 28,
25
-------
n = (1.65 + 0.84)2 x 4(0.8)/[(0.2)2(l + 0.8)(28)]
n = 9.84
Thus, 10 pairs of plots (20 total) with a mean of 28 individuals per plot
would be needed. Increasing the mean number of individuals per plot (c),
causes a reduction in n.
(ix) In some field situations, pairing may not be feasible. In these
situations, other designs would be more appropriate or a less rigorous de-
sign may have to be used. However, in planning field studies, the power
of the study design must be considered to determine the limitations of
the study. Studies with adequate replication are highly preferred to support
registration—the use of less replication will not necessarily render the
study inadequate. However, it is wrong to use a study with low power
to imply no biological damage, when the study is not capable of detecting
the damage if it occurred. In cases where large numbers of replicates are
impractical, subjective and biological knowledge should be used in a deci-
sion process to decide if there was a treatment effect. In most instances,
it is highly advisable to involve statisticians or biometricians who are fa-
miliar with this kind of field study in the planning and analysis phase
of the field work to avoid costly technical errors.
(3)	Study area and site selection. Selection of geographical areas
and study sites within the areas for the definitive test generally requires
the same considerations as for a screening study. For the definitive study,
however, the selected areas and study sites must have adequate populations
of the species of concern. Obviously, the crop of concern must be grown
on a representative portion of the area. In addition, consideration needs
to be given to whether the target pest species will be present. If it is not,
it is necessary to consider what influence its absence may have on potential
results. For example, if the pest is a major food source for nontarget spe-
cies, its absence could significantly influence results. Finally, the potential
variation in populations of concern over the geographical areas selected
should be considered. It may be difficult to find sites that are sufficiently
similar to provide paired plots, which limits the coefficient of variation
so that the desired sensitivity can be achieved.
(4)	Number and size of sites. As suggested under paragraph (e)(2)
of this guideline, the number of sites will depend upon the species density
on sites and the sensitivity required. Ideally, sample size should be large
enough so there will be an 80 percent probability of being sure to detect
a 20 percent difference when it exists. The size of the study site must
be large enough to provide adequate samples. The size depends on the
survey methods used, sensitivity required, and the density and range of
the species of concern. For a paired-plot design, the number of sites re-
quired is a function of the average density of the species. In general, the
breeding density of the species of concern can be used to provide a rough
26
-------
estimate of the size of area needed to provide adequate samples. However,
preliminary sampling most likely will be required to verify the estimates.
(5)	Methods, (i) Essentially, the methods used in a definitive study
are a means to quantitate reproductive and mortality rates of animals on
treatment and control areas. There are many texts and monographs avail-
able on methods of sampling to estimate these parameters. Anyone not
familiar with the theory and principles of the various techniques should
review these references in depth. The objective of this portion of the guide-
line is to provide a general guide to the various methods that could be
used in a definitive field study. In addition, these methods can be applica-
ble to some screening studies.
(ii)	The methods to be used in an individual field study will depend
on the nature of the identified concerns. Some methods are useful for in-
vestigating several types of concerns, and most types of concerns can be
studied by several methods. When the concern becomes more specific
(e.g., secondary hazards to raptors) or the use pattern and/or habitat type
is limited, the range of applicable methods tends to become more narrow.
(iii)	Methods described below are divided into three categories: Meth-
ods for assessing mortality and survival of adults and independent juve-
niles, methods for assessing reproduction and survival of dependent juve-
niles, and ancillary methods. The intent of this guideline is to present
methods that are likely to be useful in many situations, rather than an
exhaustive list of all available methods. The Agency encourages the use
of other methods when they are scientifically valid and have a high prob-
ability of detecting an effect.
(iv)	While it is absolutely essential to have a detailed plan that de-
scribes the selected actions (with contingencies) for achieving the study
objectives, investigators must remain flexible because unanticipated prob-
lems always come up in long term studies. Even with highly experienced
and resourceful field biologists, the most carefully planned studies can be
compromised due to the unpredictability of wild animals and natural
events. When a natural disaster occurs early in the study, it may be wise
to reinitiate the study. If the event occurs after substantial data already
have been collected (e.g., early in the second year of a multiyear study),
it may be more appropriate to extend the study an additional year or more
to help provide for the additional needs. If the study is to be terminated,
the report should describe thoroughly the nature of the events and the con-
sequences if they affect the study results.
(6)	Mortality and survival, (i) It is very important to understand
the autecology of the species being studied in order to select the most
appropriate methods for investigating them. In addition, the choice of par-
ticular methods must consider the applicability of the method based on
the pesticide use pattern and study site characteristics.
27
-------
(A) Mark-Recapture. (7) There are several mark-recapture methods
available, each based on the same basic premise. A sample of animals
is captured, marked, released, and another sample is collected where some
of the animals are captured again. The characteristics of this identifiable
sample are used to estimate population parameters. Mark-recapture studies
can provide information on:
(z) Size of the population.
(zz) Age-specific fecundity rates.
(z/z) Age-specific mortality rates.
(z'v) Combined rates of birth and immigration.
(v) Combined rates of death and emigration.
(2) Seber (see paragraph (h)(32) of this guideline) reviewed the var-
ious mark-recapture methods and subsequent statistical analyses. Less de-
tailed but still very useful reviews are provided in paragraphs (h)(4) and
(h)(16) of this guideline. Nichols and Pollock provide a valuable compari-
son of methods under paragraph (h)(26) of this guideline. The following
Table 2. provides a brief summary of some of the various mark-recapture
methods discussed in these references.
Table 2.—Mark-Recapture Techniques
Method
Applications/Requirements/Assumptions
Peterson Method (Lincoln Index)
Schumacher's Method	
Bailey's Triple Catch 	
Jolly-Seber Method 	
Estimation of population size. Usually only two sampling peri-
ods. Closed population.
Estimation of population size. More than two sampling periods.
Marking continues throughout sampling. Closed population.
Estimate of birth rate and death rate in addition to population
size. Requires data from two marking occasions and two re-
capturing occasions. Open population.
Estimates mortality and recruitment in addition to population
size. Requires more than two sampling periods and that each
animal's history of recapture be known. Open population.
(3) When considering the use of one of these mark-recapture models,
one must carefully evaluate the applicability of the method to the cir-
cumstances under consideration. While in theory mark-recapture tech-
niques should be an excellent method for evaluating effects of pesticides
on wildlife populations, some mark-recapture analyses are not particularly
robust; small deviations from their implicit assumptions can produce large
errors in the results (see paragraph (h)(4) of this guideline). However,
some of the more recent and sophisticated analytical methods are robust
and can deal with deviations from assumptions in closed populations (see
paragraph (h)(28) of this guideline).
28
-------
(4) Mark-recapture methods are particularly useful for small mammals
because these animals are seldom amenable to the visual and auditory ob-
servations necessary for using transect, territory mapping, or similar meth-
ods. However, mark-recapture also may be useful for birds provided a
sufficient number of birds can be captured and marked. In some situations,
birds may be ' 'recaptured'' with use of binoculars via visual observations
of marked individuals. Animals must remain marked for the duration of
the study. Typically, mammals are toe-clipped or ear-marked and birds
are banded. Marking should not make the animals more susceptible to the
effects of the pesticide (e.g., anticoagulants with toe clipping). Dyes may
be useful unless they are lost by wear or molting.
(B)	Territory mapping method. (7) A common spatial census meth-
od is territory mapping, wherein the territories of individuals are mapped
before and after treatment, on both treated and untreated plots. The method
is usually applicable when birds are defending territories. It involves a
series of census visits to the study sites during which birds located by
sight or song are recorded on a map. The information from all the visits
is plotted for each species. Birds exhibiting territorial behavior appear on
the map as clusters of individual contacts. The clusters are used to estimate
both the size and number of territories. The pre- and posttreatment cen-
suses for treated sites are compared with the pre- and posttreatment cen-
suses for control sites to determine changes in populations of territorial
individuals that may be attributed to the pesticide (see paragraph (h)(12)
of this guideline). Further details of this method are given by the Inter-
national Bird Census Committee under paragraph (h)(21) of this guideline,
and its application to evaluating impact caused by pesticides is reviewed
by Edwards et al. (see paragraph (h)(12) of this guideline).
(2) Problems with this method can occur. Under some circumstances,
replacement from outside the area can be so rapid that territories are re-
filled before the census is completed. There usually is a floating population
of silent, nonterritorial birds who may quickly reoccupy empty territories
(see paragraph (h)(35) of this guideline). The effects of replacement can
be overcome for some species by capturing and marking the territorial
individuals prior to treatment, so they can be distinguished from the float-
ers. Also, replacement may not be a problem when the study areas are
in the center of a relatively large treated area.
(C)	Radio telemetry. Radio telemetry can be an extremely useful
technique to provide information on the effects of a pesticide application
on nontarget species. As discussed for screening studies, radio telemetry
can be used to monitor for mortality as well as to provide useful informa-
tion on behavioral modification caused by the pesticide application. The
points discussed previously (for screening studies) generally are applicable
to definitive studies. However, for the definitive study, the number of
radio-tagged animals needed depends upon the variation between sites and
the sensitivity required. For example, with behavioral observation, intra-
29
-------
and intersite variation will influence the number of radio-tagged animals
required. In some instances, it might not be practical to radio-tag the num-
ber of animals required to provide a rigorously designed study. Under
these conditions, the limitations should be specified, and the maximum
number of animals that can be practically radio-tagged and monitored
should be used.
(D) Other methods for mortality and survival. Other techniques
for assessing density and diversity are discussed for screening studies;
most of these, especially line-transect methods, are useful for definitive
studies. Some methods, such as catch per unit effort or counts of animal
signs, do not provide actual measures of density but may still be used
to compare effects on treated and untreated plots.
(7) Reproduction and survival of dependent young. Some of the
techniques for assessing mortality and adult survival are also useful for
assessing reproduction and survival of young. Some, but not all, mark-
recapture methods can provide information on fecundity. Radio-tagging
nestlings or suckling young of moderate and large size animals may be
used to assess survival of dependent young. Radio telemetry and territory
mapping are useful for locating dens or nests for further study. The follow-
ing methods are more specific for assessing reproductive parameters.
(i)	Nest monitoring. (A) Nest monitoring is useful for evaluating the
effect of pesticides on breeding birds. The typical procedure is to search
the study site to find active nests and subsequently to check those nests
to determine their fate. Information collected on each nest should include
number of eggs laid, number hatched, number of young fledged, and if
and when the nest was abandoned or destroyed, both before and after pes-
ticide application. While all definitive studies should consider this tech-
nique, it also may be useful in screening studies.
(B) This technique is relatively straightforward. However, it may not
be practical if nests are scarce or otherwise hard to find. Because the
breeding success of birds can be highly variable and can be quite low,
it is sometimes difficult to obtain sufficient data on the success of the
same species in enough sites to yield satisfactory results for statistical com-
parison with controls (see paragraph (h)(18) of this guideline). In some
cases, artificial nest structures can be constructed to increase nest densities.
In a few situations where sufficient numbers are available, the technique
may be applicable to mammal den monitoring.
(ii)	Behavioral observations. Behavioral observations associated
with reproduction can be quite useful, especially for birds. Techniques are
simple, but labor intensive. When used, such observations most likely
would be combined with nest monitoring since both techniques require
locating reproductive sites. Typically, the frequency and duration of behav-
iors will be compared for treated and untreated plots. Incubation, parental
30
-------
care (especially feeding for altricial birds), and following behavior (for
precocial animals) are behaviors that are particularly amenable to such
study. Courtship, mating, and nest building are other behaviors that could
be studied in some situations, but locating sufficient numbers of animals
displaying these behaviors to permit quantitative analysis is difficult.
(iii) Age structure of populations. (A) Comparisons of young/adult
ratios of selected species between treated and untreated plots may indicate
reproductive effects. The timing of the application and of breeding of se-
lected species is critical. For assessing reproductive impairment or survival
of dependent young, per se, the duration of this technique should be lim-
ited to single breeding/rearing periods, which may be repeatedly assessed.
However, longer study periods that may even include several years can
be used to assess the combination of reproductive success and age-specific
mortality, even if the two cannot be separated.
(B) Obviously, use of this method requires that the age of individual
animals be determined. In some cases, it may be necessary only to distin-
guish among adults, subadults, and juveniles. In mammals, this may usu-
ally be accomplished by examining pelage, development of testes or
mammaries, or tooth eruption or wear characteristics. In birds, plumage
or characteristics of particular (species-dependent) feathers may be used.
For carcasses or sacrificed animals, observation of the ossification of bones
or development of reproductive organs are useful. In other cases, particu-
larly where comparisons are made among populations in different years,
it may be appropriate to distinguish age classes of adults. In mammals,
tooth eruption, wear, or enamel layers, or eye lens weights are useful.
It is more difficult to separate age classes of many adult birds, although
overall plumage or feather characteristics can provide some indication. In
some slow-maturing birds (e.g., gulls), plumage may be used to distinguish
year classes of sub-adults. Additional details on aging birds and mammals
are presented by Larson and Taber (see paragraph (h)(22) of this guide-
line).
(8) Ancillary methods, (i) At least some ancillary methods are essen-
tial in every field study. As used here, ancillary methods are generally
of two types. Certain of these methods are important for determining the
nature or existence of effects or for establishing causal relationships. Oth-
ers of these methods do not address effects directly, but they provide im-
portant information for interpreting the results of the study.
(ii) Many of the methods for determining effects have been discussed
for screening studies. Enzyme analysis, such as for ChE inhibition, and
observations of signs of toxicity can show that animals were exposed to
or killed by a toxicant of a particular type. Where it is possible that animals
may be exposed to other pesticides of the same type (e.g., feeding in a
nearby area treated with other pesticides), residue analysis in nontarget
animals may be necessary to determine which specific pesticide caused
31
-------
the signs or alterations in enzymes. Even though carcass searches, per se,
are not recommended for definitive studies, it is still essential to recover
and analyze any carcasses found accidentally or obtained through radio-
tracking. Residue and/or enzyme analysis of live animals collected will
frequently be important.
(iii) Among the other ancillary methods, analysis of environmental
residues is crucial and will probably be necessary in nearly every definitive
field study. As discussed for screening studies, the most important environ-
mental residues are those that occur on or in wildlife food sources, which
may include insects, plant parts, or even other vertebrates, depending upon
the species that are the primary focus of the investigation. The investigator
should review the literature on food habits of the species being studied;
often it will be appropriate to assess food habits on the specific study
sites, particularly where the literature is not adequate to define food habits
in the agricultural ecosystem under study. Such an assessment should in-
clude the availability of food sources and the number of mobile animals
that spend only part of the time in and adjacent to treated sites. The habitat
should be thoroughly described to include both the morphology and spe-
cies that are relevant to wildlife. Frequently, it will be important to locate
and describe roosting, denning, or nesting sites for mobile wildlife that
use treated sites part of the time.
(9) Interpretation of results, (i) While each field study is unique,
some elements may be common among many field studies. When a defini-
tive field study is required, the requirement is based on one or more spe-
cific concerns that pertain to a specific chemical and one or several use
patterns. Because of the substantial diversity in the types of problems to
be assessed and the variety of available investigative methods, the key
to understanding and interpreting a field study lies in the development
of a sound protocol. All protocols will contain a description of the study
sites, or the characteristics to be used in selecting sites within a given
area, and the methods to be used in conducting the study. However, a
well designed protocol will go beyond this descriptive approach in three
ways.
(A) First, the well-designed protocol will contain a restatement of the
concerns to be addressed to ensure that there is an adequate understanding
of the Agency's position. The investigator should review the literature and
other available information that may bear upon the problem. It is possible
that the literature may contain a valid answer to the questions raised by
the Agency. Far more likely, the literature may orient the investigator to
address the concerns in a particular way. An example is provided by
Hegdal and Blaskiewicz (under paragraph (h)(17) of this guideline), who
conducted a study to address the Agency's concerns for secondary toxicity
to barn owls (specifically) from the use of an anticoagulant bait proposed
for use on commensal rodents in and around agricultural buildings. A re-
view of the literature by these investigators indicated to them that labora-
32
-------
tory studies suggested a legitimate potential for secondary poisoning to
exposed raptors, but that the food habits of barn owls consist primarily
of microtine rodents in most areas, suggesting a low potential for actual
exposure. Consequently, they designed their study to focus on barn owl
food habits and movements, and included an additive to the bait formula-
tion that would permit an identification of whether or not the barn owls
ate rodents that had fed on the bait. The study adequately demonstrated
that actual exposure of barn owls was quite limited, and the proposed reg-
istration for this use was subsequently approved. By using the available
literature on both the chemical and the particular species of concern, the
investigators were able to narrow the study while still providing sufficient
information for evaluation. However, it should be noted that this study
was not adequate for evaluating the potential for secondary toxicity in the
field to other predators that may have different food habits, or for other
use patterns that may result in exposure to different predators or scav-
engers.
(B)	Second, the well designed protocol will provide the reasons why
particular methods are being used, including, at least qualitatively, the
meaning that different results might have. For example, a protocol may
include collection of residues in nontarget animals, but it also should in-
clude a statement of purpose and meaning for such collection. Residues
may be used to indicate potential exposure to nontarget organisms through
analysis of their food, exposure in nontarget animals as a result of eating
contaminated food, or that a particular pesticide was likely to be the cause
of any observed effects. Interpretation of data is facilitated substantially
by a statement of what results were intended by using a particular tech-
nique. In the previously cited example (see paragraph (h)(16) of this guide-
line), it was clearly stated that collection of owl pellets was used to assess
general food habits and that use of a fluorescing dye in the bait was for
the purpose of ascertaining whether or not the owls fed on commensal
rodents that specifically had fed on the bait. The interpretation of the data
collected, once the purpose was stated, naturally led to the conclusion of
no-significant-exposure to the barn owls.
(C)	Third, the well designed protocol will contain an experimental
design that will indicate how the results can be assessed quantitatively.
The experimental design has been discussed in previous sections of this
guideline, but there are two facets that relate closely to the interpretation
of results: The difference that can be detected between treated and un-
treated plots and the power (ability) of the design to detect this difference.
An experimental design with number of replicates based on an estimated
coefficient of variation that closely approximates reality will allow the
study to detect a stated concern level some prescribed number of times
during the study time. The actual difference between treated and control
units is measured during the field study, but the design can form an initial
basis for interpretation when combined with the available information on
33
-------
the species of concern. As a result, the well-designed protocol should in-
clude a section on interpretation.
(ii)	Study methods for investigating acute mortality are more straight-
forward than for other kinds of effects. Nevertheless, there are sufficient
differences in the use of the data to preclude a constant interpretation.
The study may focus directly on the species of concern and may involve
little or no extrapolation, depending on such factors as the type and the
extent of use, the available toxicity data base, and home range of the spe-
cies. Extrapolation to other populations, regions, or uses might be nec-
essary. If the species of concern cannot be studied directly, it may be
necessary to extrapolate between species, involving interspecies dif-
ferences both in toxicological sensitivity and in ecological and population
parameters.
(iii)	The same kinds of considerations apply to reproductive impair-
ment and chronic toxicity, even though different, and often more laborious
and costly, investigative methods are involved. Where reproductive success
is impaired, information on species-specific variation in reproductive ecol-
ogy is necessary to understand how a particular degree of impairment may
relate to effects among various species. Such reproductive considerations
can include whether an avian species is a determinate or indeterminate
layer, the number of nestings per season for different geographic areas
in the use pattern, the length of the refractory period, as well as the specific
effect which can range from destruction of reproductive organs to behav-
ioral deficits such as nest abandonment. Considerations of reproductive
ecology among different species of mammals include delayed fertilization
or implantation, resorption of embryos or parental infanticide due to stress,
number of young per breeding cycle, etc. All of these factors, and many
others, are relevant to determining for different species the extent of effects
that could result in population reductions or lack of ability to recover.
(iv)	An analysis of whether or not a particular level of effect is going
to affect wildlife populations is species-specific. For any species (or sub-
species), the changes in population can be described very simplistically
by the equation: rate of population increase (r) = birth rate - death rate,
where values of r can be positive (population growth) or negative (popu-
lation reduction). Immigration and emigration are also important when the
concern is for specific populations of a species. These characteristics differ
among species, and data will not always be available. The application of
sound scientific judgment to the best available information will be the
basis for interpreting the results of a study. It may be necessary to compare
the results of the field study to laboratory data, especially where laboratory
data are available on a variety of species and/or effects and the field study
has focused on species other than those of direct concern. The use of ex-
trapolation techniques will he necessary where endangered species are of
concern or where other species cannot be studied directly.
34
-------
(v) The Agency would like to be able to obtain a standardized result
from a field study so that the result could be applied in a very consistent
manner. As discussed in previous sections of this guideline, the different
effects and species of concern will vary and will require the development
of specific protocols to address these factors. Although most of the various
techniques have some degree of standardization, the field study may com-
bine the individual techniques in a wide variety of ways to address specific
concerns. A standardized result might be attainable for the individual tech-
niques, although that result would still have to be applied differently for
various species, depending on their biology and ecological characteristics.
However, determining a result for the whole field study that would un-
equivocally lead to a statement of the degree of risk, while obviously desir-
able, is not currently practical.
(f) Suggested components of a field study protocol. The following
protocol was adapted from The Wildlife Managements Techniques Manual
(under paragraph (h)(29) of this guideline), and is recommended for stud-
ies submitted to the Agency for review.
(1)	Title.
(2)	Problem definition. The following information should be pro-
vided:
(i)	A review and summary of the available information on the pes-
ticide in relation to nontarget hazard, including use information.
(ii)	A precise statement of the goals and purpose of the study.
(iii)	A brief statement of the problem and the context in which it
exists, specifying the limits of the proposed work.
(iv)	Precise statements of the major hypotheses to be tested.
(3)	Methods and materials. This section should include the following:
(i)	A brief discussion of various methods and procedures that have
been or could be used to evaluate the problem. This discussion should
identify the strengths and weaknesses of each method or procedure dis-
cussed.
(ii)	Description of the procedure to be followed:
(A)	Identify the study areas selected and their general suitability for
achieving the objectives of the study or what criteria will be used to select
study areas.
(B)	Identify the species present or expected to be present on the study
areas, discussing characteristics pertinent to the problem being evaluated.
35
-------
(C)	State the research procedures, designs and sampling plans to be
used.
(7) Specify the kind and amount of data needed and to be sought.
(2) Describe in detail how all the data are to be obtained, including
details of application, instrumentation, equipment, sampling procedures,
and any other information.
(D)	Describe how the data are to be treated, including specifying what
statistics are to be calculated, what models will be used, what tests of
data will be used, etc.
(E)	Describe in detail the methods to be used to check the sensitivity
and accuracy of the procedures used.
(F)	Describe quality assurance procedures for application, instrumen-
tation, equipment, and records.
(G)	Briefly describe the resources (people, facilities, etc.) to be ap-
plied to the study.
(g) Carcass searches—(1) Design, (i) In designing carcass searches,
the following factors need to be known or determined:
(A)	Density of the species that are likely to be exposed. For example,
granular products are most likely to result in exposure to ground-feeding
animals; therefore, birds such as warblers or swallows should not be in-
cluded in density counts for such products.
(B)	Probability of finding dead animals if any are killed. This is de-
pendent on the probability of a carcass remaining on the study site (i.e.,
not being removed by scavengers) and the probability of detecting a car-
cass if it remains on the study site (search efficiency).
(C)	Size of the search area.
(D)	Number of carcasses found.
(ii)	These factors can be combined in the following formula:
n = d r e a p
where n = number of carcasses found, d = density in animals per acre,
r = proportion of carcasses remaining (nonremoval), e = search efficiency,
a = acres searched, and p = proportion of population killed.
(iii)	Carcass searches should be used only when there is a reasonable
potential to detect mortality. If such mortality does occur, the carcass
search should be able to detect it and therefore, carcasses should be found.
It is recommended that carcass searches be designed so that at least two
carcasses (n = 2) will be found if there is appreciable mortality. In general,
36
-------
preliminary sampling would be required to determine these factors. How-
ever, information from other field studies can be used in the planning
stages to determine if carcass searching would be appropriate for use under
anticipated conditions and to assist in developing the study design.
(iv)	The sensitivity of the carcass search approach is equivalent to
the percent detectable kill of the population. To determine the sensitivity,
the formula is adjusted:
p=n/drea
Since p is a proportion:
percent detectable kill = px 100 = n/ dreax 100
(v)	If any of the values of d, r, e, or a are zero, the equation cannot
be solved and the carcass search is not applicable (i.e., no density of birds,
no acres searched, no carcasses remaining, no remaining carcasses found).
However, other combinations of d, r, e and a, such as low density and
small acreage or low efficiency and high scavenger removal, can result
in a small denominator meaning that mortality can be detected only when
a high percentage of the population is killed. For example, in 5 acre fields
with only two birds per acre and r and e estimated at a moderate 0.5,
only an 80 percent or greater kill could be detected. In such situations,
it is necessary to increase one of the parameters to achieve a stated level
of detectability or else to use methods other than carcass searching. The
same equation can be used to estimate the minimum search area to detect
a given mortality level (p) by solving for a.
(2)	Search procedure. In general, depending on the sensitivity of
the search method relative to the habitat involved, corridors or plots should
be selected. These areas should be searched systematically by walking pre-
determined routes until the area has all been covered. Due to the con-
centration required to find dead animals, other activities that could distract
the attention of the searchers should be avoided during carcass searching.
In homogeneous situations, investigators should randomly select search
areas. However, in most studies it is advisable to stratify the sampling,
concentrating efforts in areas frequented by wildlife species such as woods
edges, ditch banks, field borders, fencerows, and other habitats where wild-
life concentrate.
(3)	Duration. Searches should begin on the day of application and
continue on a daily basis for as long as mortalities or other evidence of
intoxication occur. In general, a week or two following application should
be adequate. However, the length of time searches are continued should
be related to how long lethal concentrations are expected to be present.
Normally, the same areas should be searched each day.
37
-------
(4)	Estimating efficiency of carcass search, (i) Efficiency trials
should be conducted periodically (minimum 3 times per study site) during
the study to determine the proportion of carcasses that are detected. Just
prior to the initiation of a scheduled search, carcasses of animals represent-
ative of species found in the area should be variously placed within the
search area. If the study site includes edge habitat, carcasses should be
placed in the edges as well as in the fields. In general, carcasses should
be placed where animals would be most likely to die, depending on the
nature of the chemical. Searchers should not be aware that simulated mor-
talities have been placed; however, they should be aware that these trials
will occur during any scheduled search.
(ii) The number of carcasses placed should be approximately equal
to 20 percent of the estimated density of species on the search area. All
placed carcasses should be marked to distinguish them from actual kills.
The location of placed carcasses should be mapped so those not found
can be easily recovered following completion of that day's search activi-
ties, since unrecovered carcasses could bias study results. For example,
if a scavenger were to carry off a simulated kill and consume it at another
location on the study site, the remains could be erroneously classified as
pesticide-related if found. One way to avoid this problem would be to
dip carcasses in a nontoxic substance that fluoresces under ultraviolet light
so that the remains could be identified.
(5)	Estimating carcass removal rate, (i) Carcass removal should be
monitored to determine local variability in scavenger activity. The density
of both carcasses and scavengers can influence the rate of removal. Under
some conditions, large numbers of carcasses may attract scavengers. In
other situations a large number of kills may dilute removal rate due to
limited number of scavengers. Where it can be adequately documented
that removal of carcasses occurs almost exclusively either at night or dur-
ing the day, the timing of carcass searches may be adjusted to minimize
the effects of removal.
(ii)	Carcasses planted in monitoring trials should simulate mortalities
actually occurring from the pesticide. In most cases, small to moderate
sized species such as starlings or blackbirds, or laboratory bobwhite or
Japanese quail chicks may be used. Carcasses should be variously placed
within the general study areas and monitored daily for at least 5 days
or until 90 percent have been removed. The number used should approxi-
mate densities resulting from effects of the pesticide under study; however,
in most instances, this will not be known. Therefore, a density of approxi-
mately 20 percent of the population of nontarget species on the area is
recommended.
(iii)	Timing of carcass removal trials should be such that they do
not affect scavenger removal of pesticide-killed birds or the feather-spots
of the removed carcass could be erroneously classified as a pesticide kill.
38
-------
Location of placed birds should be recorded on maps and may be marked
in the field with small stakes or by other inconspicuous means, preferably
at a fixed distance and direction from the carcass.
(h) References. The following references should be consulted for ad-
ditional background material on this test guideline.
(1)	Bunyan, P.J. et al. Organophosphorus poisoning, some properties
of avian esterases. Journal of Agricultural and Food Chemistry 16: 326-
331 (1968).
(2)	Bunyan, P.J. et al. Organophosphorus poisoning, diagnosis of poi-
soning in pheasants owing to a number of common pesticides. Journal
of Agricultural and Food Chemistry 16: 332-339 (1968).
(3)	Burnham, K.P. et al. Estimation of density from line transect sam-
pling of biological populations. Wildlife Monographs, No. 72 (1980).
(4)	Caughley, G. Analysis of Vertebrate Populations. Wiley, NY
(1977).
(5)	Cochran, W.W., Wildlife telemetry, pp. 509-520, in Wildlife Man-
agement Techniques Manual. S.D. Schemnitz, Ed. The Wildlife Society,
Washington, DC (1980).
(6)	Corbett, J.R. The Biochemical Mode of Action of Pesticides. Aca-
demic, NY (1974).
(7)	Davis, D.E. and R.L. Winstead. 1980. Estimating the numbers
of wildlife populations, pp. 221-246, in Wildlife Management Techniques
Manual. S.D. Schemnitz, Ed., The Wildlife Society, Washington, DC
(1980).
(8)	Eberhardt, L.L. Quantitative ecology and impact assessment. Jour-
nal of Wildlife Management 4: 27-70 (1976).
(9)	Eberhardt, L.L. Appraising variability in populations studies. Jour-
nal of Wildlife Management 42: 207-237 (1978).
(10)	Eberhardt, L.L. Assessing the dynamics of wild populations.
Journal of Wildlife Management 49: 997-1012 (1985).
(11)	Edgington, E. Randomization Test. Dekker, NY (1980).
(12)	Edwards, P.J. et al. The use of a bird territory mapping method
for detecting mortality following pesticide application. Agro-Ecosystems
5: 271-282 (1979).
(13)	Ellman, G.L. et al. A new and rapid calorimetric determination
of acetylcholinesterase activity. Biochemical Pharmacology 7:88-95
(1961).
39
-------
(14)	Environmental Protection Agency. Pesticide Assessment Guide-
lines, Subdivision E - Hazard Evaluation: Wildlife and Aquatic Orga-
nisms. Office of Pesticides and Toxic Substances, Washington, DC (1982).
(15)	Hanson, W.R. Estimating the density of an animal population.
Journal of Research Lepidoptera 6: 203-247 (1967).
(16)	Hegdal, P.L. and R.W. Blaskiewicz. Evaluation of the potential
hazards to barn owls of Talon (brodifacoum bait) used to control rats and
house mice. Journal of Environmental and Toxicological Chemistry 3:
167-179 (1984).
(17)	Heinz, G.H. et al. Environmental contaminant studies by the Pa-
tuxent Wildlife Research Center, pp. 8-35. in Avian and Mammalian Wild-
life Toxicology. ASTM STP 693, E.E. Kenaga (Ed), American Society for
Testing and Materials, Philadelphia, PA (1979).
(18)	Hill, E.F. and W.J. Fleming. Anticholinesterase poisoning of
birds: field monitoring and diagnosis of acute poisoning. Journal of Envi-
ronmental and Toxicological Chemistry 1: 27-38 (1982).
(19)	Hurlbert, S.H. Pseudoreplication and the design of ecological
field experiments. Ecological Monographs 54:187-211 (1984).
(20)	Giles, R.H. Jr. Wildlife Management. Freeman, San Francisco,
CA. (1978).
(21)	International Bird Census Committee. Recommendations for an
international standard for a mapping method in bird census work. Bulletin
of the Ecolological Research Committee 9:49-52 (1970).
(22)	Larson, J. S. and R. D. Taber. Criteria of sex and age. pp. 143—
202, in Wildlife Management Techniques Manual. S. D. Schemnitz; (Ed.),
The Wildlife Society, Washington, DC (1980).
(23)	Leopold, A. Game Management. Scribner's, NY (1933).
(24)	Ludke, J.L. et al. Cholinesterase (ChE) response and related mor-
tality among birds fed ChE inhibitors. Archives of Environmental Contami-
nant Toxicology 3:1-21 (1975).
(25)	Moen, A.N. Wildlife Ecology: An Analytical Approach. Freeman,
San Francisco, CA (1973).
(26)	Nichols, J.D. and K.H. Pollock. Estimation methodology in con-
temporary small mammal capture-recapture studies. Journal of Mammal
64:253-260 (1983).
(27)	O'Brien, R.D. Insecticides: Action and Metabolism. Academic
Press, NY (1967).
40
-------
(28)	Otis, D.L. et al. Statistical interference from capture data on
closed animal populations. Wildlife Monographs 62:1-135 (1978).
(29)	Ripley, T.H. Planning wildlife management investigations and
projects, pp. 1-6. in Wildlife Management Techniques Manual. S.D.
Schemnitz (Ed.), The Wildlife Society, Washington, DC (1980).
(30)	Rosene, W. Jr. and D.W. Lay. Disappearance and visibility of
quail remains. Journal of Wildlife Management 27:139-142 (1963).
(31)	Scientific Advisory Panel. Final Scientific Advisory Panel
subpanel's report on the January 7-8, 1987 meeting concerning terrestrial
field studies. Environmental Protection Agency, Washington, DC (1987).
(32)	Seber, G.A.F. The Estimation of Animal Abundance and Related
Parameters. Macmillan, NY (1982).
(33)	Shellenberger, T.E. et al. The comparative toxicity of
organophosphate pesticides in wildlife, pp. 205- 210, in W.B. Diechmann,
(Ed.), Pesticide Symposium, Halos, Miami, FL (1970).
(34)	Snedecor, G.W. and W.G. Cochran. Statistical Methods. Sixth
Edition. The Iowa State University Press, Ames, IA (1967).
(35)	Stewart, R.E. and J.W. Alrich. Removal and repopulation of
breeding birds in a spruce-fir forest community. The Auk 68:471-482
(1951).
(36)	Walpole, R.E. and R.H. Myers.Probability and Statistics for En-
gineers and Scientists. Macmillan, NY (1972).
41
-------