Office of Superfund Remediation and Technology Innovation
           and
           Office of Research and Development
Sediment Assessment and Monitoring Sheet (SAMS) #1
                  Using Fish Tissue Data to
                Monitor Remedy Effectiveness
                                              Receptors
                                              Trophic
                                              Transfer

                      OSWER Directive 9200.1-77D
                              July 2008

-------
                Using Fish  Tissue Data to Monitor Remedy
                Effectiveness
SEDIMENT ASSESSMENT AND MONITORING SHEET #
Background and
Purpose
This is the first fact sheet in the
Sediment Assessment and
Monitoring Sheet (SAMS) series
prepared by the Office of
Superfund Remediation and
Technology Innovation (OSRTI).
This sheet was prepared in
collaboration with the Office of
Research and  Development.
OSRTI anticipates that other
SAMS will be completed, some
in collaboration with other
federal agencies.

This document provides
technical guidance to the U.S.
Environmental Protection
Agency (EPA)  staff on
developing monitoring plans for
contaminated sediment sites.   It
also provides information to the
public and to the regulated
community on  how EPA intends
to exercise its discretion in
implementing monitoring plans.
This document does not impose
legally-binding requirements on
EPA, states, or the regulated
community, but suggests
monitoring approaches that may
be used at particular sites, as
appropriate, given site-specific
circumstances.
 Introduction
Chapter 8 of the Contaminated Sediment Remediation Guidance
for Hazardous Waste Sites (OSWER Directive 9355.0-85,
December 2005), presents an approach for developing an effective
monitoring plan (http://www.epa.gov/superfund/health/
conmedia/sediment/guidance.htm).  As stated in the Guidance, one
of the goals of monitoring is to "evaluate long-term remedy
effectiveness in achieving remedial action objectives (RAOs) and
in reducing human health and/or environmental risk." The
Guidance describes a successful remedy as one where "the
selected sediment chemical or biological cleanup levels have been
met and maintained over time, and where all relevant risks have
been reduced to acceptable levels based on the anticipated future
uses of the water body and the goals and objectives stated in the
ROD." The information in the following text box is Highlight 8-1
from the Guidance.

As stated in the last two measures, fish tissue contaminant
concentrations are often the key measures that need to be
monitored.  The Guidance, however, does not specify how, when,
or where to collect fish tissue samples. There are many factors
that can influence the measured concentrations of contaminants in
biota tissues. The site manager and technical team need to be
aware of these factors and consider them in developing the
sampling plan. This will help ensure that the data collected can be
used to evaluate remedy effectiveness and to evaluate the
protectiveness of the remedy during the five year review process.
  United States
  Environmental Protection
  Agency
 Office of Superfund Remediation and
 Technology Innovation and
 Office of Research and Development

           1
OSWER Directive 9200.1-77D
              July 2008

-------
Using Fish Tissue Data to Monitor Remedy Effectiveness
  Sample Measures of Sediment Remedy Effectiveness

  Interim Measures

     1.  Short-term remedy performance (e.g., Have the sediment cleanup levels been achieved?
        Was the cap placed as intended?)
     2.  Long-term remedy performance (e.g., Have the sediment cleanup levels been reached and
        maintained for at least five years, and thereafter as appropriate?  Has the cap withstood
        significant erosion?)
     3.  Short-term risk reduction (e.g., Do data demonstrate or at least suggest a reduction in fish
        tissue levels, a decrease in benthic toxicity, or an  increase in species diversity or other
        community indices after five years?)

  Key Measures

     4.  Long-term risk reduction (e.g., Have the remediation goals in fish tissue been reached or has
        ecological recovery been accomplished?)
This SAMS provides general information on the
collection and use of tissue residue data for
monitoring the effectiveness of sediment
remedies at Superfund sites.  This information
may also be useful in collecting baseline risk
assessment data, but that is not the focus of this
fact sheet. This fact sheet briefly discusses the
factors that may be important and provides
general recommendations for fish sampling at
typical Superfund sites. Although this fact sheet
focuses on finfish, much of the information is
applicable to shellfish, e.g., mussels and crabs.
These recommendations are  based on the
experience and expertise of EPA researchers
and program staff and are supported by peer-
reviewed literature. Nevertheless, these
recommendations may not apply at every site,
and project managers are encouraged to make
their own site-specific decisions concerning
effective monitoring plans. Dependent upon the
question(s) being asked, the  data requirements
may be relatively easy to meet, or could
necessitate large and costly efforts.

Although this fact sheet focuses on collecting
and using fish tissue contaminant data, surface
sediment samples should also be collected at the
same locations and same time as part of remedy
effectiveness monitoring.  It is important to try
to understand the relationship between the
contaminant levels in the surface sediment and
the resulting levels in the fish.  A biota sediment
accumulation factor (BSAF) approach is often
used to characterize this relationship and this
approach is most useful if both fish tissue and
sediment data are collected concurrently.
Depending upon site specific conditions, it may
also be important to collect surface water
samples at the same locations to further
understand the exposures resulting in
contaminant uptake in fish.
    Source: The Great Lakes National Program Office

-------
Using Fish Tissue Data to Monitor Remedy Effectiveness
Factors to Consider in Collecting Contaminant Residue Data in Fish and
Other Aquatic Organisms
Contaminant Types
Contaminants accumulate in biota to varying
degrees and at different rates. These variations
are a function of the contaminant, the organism,
and the environment.  Contaminants found most
frequently at Superfund sites may be divided
into three chemical classes: organic compounds,
metals, and organometallics like methylmercury
and tributyltin. PCBs and pesticides have been
the key organic contaminants of concern (COC)
at over half of the Superfund sediment sites.

The most important characteristic of organic
compounds that affects their ability to
bioaccumulate is their hydrophobicity; i.e., their
resistance to be dissolved in water. A
chemical's hydrophobicity is most often
expressed or measured using the n-
octanol/water partition coefficient, Kow, often
displayed as the logarithm, log Kow.
Fortunately, the grouping of compounds by their
Kow allows for some generalizations regarding
the expected accumulation of these organic
compounds in tissues.

The extent of bioaccumulation of a chemical is
also fundamentally related to the rates of
excretion and metabolism of the chemical in the
organism. Organic compounds that are very
slowly metabolized (if at all) are often highly
chlorinated, such as PCBs, dioxins and furans,
and DDTs.  In general, organic chemicals that
significantly bioaccumulate in fish are nonionic,
have a Kow greater than 105, and are not rapidly
excreted or easily metabolized.

The second major class of contaminants in
sediment at Superfund sites is metals.
Unfortunately, there are no generalizations
concerning bioaccumulation that can be made
for metals. The chemical properties that affect
the accumulation of metals can be different for
different forms or chemical species of the same
metal, e.g., their oxidation state.  The issues
surrounding the accumulation and effects of
exposures to metals are summarized in the
Framework for Inorganic Metals Risk
Assessment (U.S. EPA 2007).

Some metals can be transformed into
organometallic compounds that accumulate in
tissues to much greater levels than their
inorganic counterparts. The best example of
this is mercury.  While inorganic mercury is not
readily accumulated, the organic form,
methylmercury, accumulates substantially.

Organism Type and Lipid Content
Different taxonomic groups of organisms or
different life stages of the same organism can
accumulate contaminants differently. This is a
result of both the physiology and the life history
of the particular organism. Different classes of
organisms have different biochemical systems
that vary in their ability to degrade or
metabolize  contaminants, and some classes of
organisms have mechanisms to sequester and/or
excrete the  contaminant or detoxify it.

Because risks at Superfund sediment sites are
often driven by the ingestion offish  and
shellfish by humans or wildlife, these organisms
are the ones that usually should be sampled.
Since fish are mobile, they may be good
integrators of varying sediment conditions and
can be used to estimate typical exposures from
ingestion offish at a site. Some fish, however,
have very large foraging ranges and, depending
on site size, may not represent well the
exposures attributable to just the site.
Knowledge of the  life history of the  organism
may be needed in order to limit your selection
offish species to those that have exposures
reflective of site contaminants.  Small non-game
fish with high site  fidelity and organisms with
limited mobility, such as  clams or mussels,  can
be very useful in estimating contaminant
exposures from localized areas.

-------
Using Fish Tissue Data to Monitor Remedy Effectiveness
For organic compounds, the lipid content of the
organism can greatly influence the degree of
accumulation. Generally, the higher the lipid
content, the greater the accumulation. Different
species offish (and other biota) have different
lipid contents. Additionally, the age and
physiological state, e.g.,  gravid females, of the
organism can also affect the lipid content.

Fish Species
The species selected for  monitoring can have a
substantial effect on the  degree of bio-
accumulation observed.  Species can vary greatly
in size, lipid content, feeding habits, and
movement patterns, and  these in turn influence
bioaccumulation.  Knowledge of the fish species
present and their feeding behaviors can assist in
the  selection of the most appropriate species to
meet the study objectives.  However, since many
sediment sites include a fish consumption
advisory as a component of the remedy, the state
public health agency  that is implementing the
advisory should be consulted on the selection of
species. This consultation also can be valuable in
determining where and how to best collect a
particular  species.

Sex of Organism
For some species, the tissue concentration can
be influenced by the sex of the  individual
(Rypel et al. 2007).  This can be because of
inherent differences in the type and amounts of
lipids between males and females, differences in
feeding habits, spawning, and other life history
parameters, or to differences in elimination rates
of the contaminant.  Spawning  can alter lipid
content and contaminant concentrations in
females resulting in either biased data and/or
increased variance in the data.
Sample Type and Size
Determining the most appropriate sample type
and size depends on how the data are going to
be used, the extent of the available baseline
data, and several other site-specific factors.
Within the Data Quality Objective process,
decisions regarding the contaminant detection
limit, fish size, number offish, and location and
number of sampling stations must be made. As
with most investigations, the effort needs to be
cost-effective, balancing the costs offish
collection and sample analyses against the need
for increased accuracy and certainty  in
determining levels of risk reduction.  Detection
limits may need to be lower than typical if the
remediation goal is low, especially if small
species are to be sampled.  For small species or
small individuals of a species (e.g., young of the
year fish), the mass of each individual is  often
below the mass needed to meet standard
analytical requirements; e.g., 10 gram wet
weight per organic scan and 0.5 gram for metal
scan (see EPA's Contract Laboratory Program
Web site
http://www.epa.gov/superfund/programs/clp/tar
get.htm).  Micro extraction and analysis
techniques exist but are not routinely available,
and typically require additional expertise in the
handling and storage of samples to avoid
significant artifactual contamination.

One must decide whether to use samples
composed of whole fish or fillet. If risks are
driven by human ingestion of contaminated fish,
then samples consisting of fillets are generally
analyzed.  If risks are driven solely by
ecological risks, then samples consisting of
whole fish are generally employed.  At many
Superfund sites, to reduce the number offish
collected and to obtain residue data compatible
with both human and ecological risk
evaluations, samples composed of the fillets and
samples of their offal (the reminder of the
carcass after filleting) are analyzed separately.
Using the analytical data from the fillet and
offal samples, whole body residues can be
      Source: The Great Lakes National Program Office

-------
Using Fish Tissue Data to Monitor Remedy Effectiveness
estimated if needed using a tissue-weighted
method.

Since most sediment remedies have been driven
by human risks from ingestion of contaminated
fish, monitoring plans for Superfund sites
should collect data that are compatible with the
data used by the state or other public health
agency to set or modify fish consumption
advisories. However, when evaluating
ecological risks, it may be more useful to collect
smaller, early indicator species as well.

The EPA Guidance for Assessing Chemical
Contaminant Data in Fish Advisories, Vol. 1 -
Fish Sampling and Analysis, Third Edition
(EPA-823-B-00-007, November 2000) states
that programs monitoring contaminant levels in
fish tissues for fishing advisories should pool
individual  fish  into composite samples.  This is
done to improve estimates of the mean chemical
residue in the fish population while reducing
analytical costs. However, there may be cases
where a more rigorous and more expensive
sampling plan that analyses individual fish is
warranted. At  such sites, it may be useful to
perform a power analysis21 in order to determine
the minimum number offish to be collected and
analyzed in order to detect a specified minimum
significant differences in chemical residues over
time and/or space; e.g., have the residues in fish
decreased by 50% in three yearsb?
If the sample size (i.e., number offish) is too
small, the measurements will not have the
precision needed to provide reliable detection of
the differences in chemical residues over time.
If the sample size is too large, resources will be
wasted because too many fish were collected
and analyzed. When fish are pooled into
composite samples, there are tradeoffs between
the number of composites (n) and number of
fish per composite (m) in terms of their impact
upon the estimate obtained for the population
variance.  EPA's guidance on fish advisories
provides detailed information on the interchange
between number of composites (n) and number
offish per composite (m) upon the measured
variance, and further, provides look-up tables
documenting statistical power of the hypothesis
tests with a variety of specified assumptions.
Power analysis can be performed using many
statistical packages, e.g., SAS, PASS, G*Power,
R, and S-Plus.

The analysis of large numbers of individual fish
at several locations can be costly for some
contaminants.  A second potential issue with the
collection of large numbers offish is that,
depending on the site, it might not be possible to
collect the target number of organisms within
the specified size range.  The field collection
crew, no matter how good, can only collect what
organisms are present.
a The power of a statistical hypothesis test is the probability that the test will not reject the null hypothesis (H0: Residues in fish have not
  decreased by 20%) when the null hypothesis is false, i.e., the residues have actually decreased by 20%. In other words, power is the
  probability that the test will not make a Type II error, i.e., false negative (the apparent decrease is real, but is rejected as being not real).
  When one performs hypothesis tests, typically, one sets alpha to 5% (a=0.05), and this is the probability for the Type I error, i.e., false
  positive (the apparent decrease is not real, but is accepted as being real). Setting alpha does not set the probability for the Type II error
  (P). The power of a test is calculated as 1- p. Power analysis defines beta (P), with a given alpha (a), for the statistical hypothesis test in
  either a prospective or retrospective applications. In prospective applications (before data are collected), power analysis will enable the
  refinement of your sampling designs (number offish per composite and number of composites). See appendix for further information on
  power analysis.
b The 50% decrease in residues is illustrative. Other risk reduction goals can be used.

-------
Using Fish Tissue Data to Monitor Remedy Effectiveness
General  Fish Sampling Recommendations for Superfund Sediment Sites
[Note: These recommendations are meant for a typical sediment site. Larger or more complex sites may require more
sampling, while smaller or simpler sites may require less.]
Baseline Data
In order to implement a decision-oriented post-
remediation monitoring plan, adequate baseline
data (i.e., pre-remediation) should be available so a
statistical comparison of pre-and post-remediation
data can be made. Where the baseline data were
collected a long time ago, analytical methods
improvements and changes might require
resampling and analysis so that the pre- and post-
remediation data are comparable. If adequate
baseline data are not available from the RI, new
data will need to be collected before the remedy is
implemented.

The importance of having baseline data and post-
remedy monitoring data that are compatible can
not be over-emphasized. The fish collection
methods should be as similar as possible; e.g.,
same species, age classes/sizes, sexes, locations,
timing, etc. By making the collections similar
and minimizing variability, it will enhance your
ability to detect statistically significant smaller
reductions in tissue concentrations.

Sampling Frequency
When developing a fish or biota sampling plan, the
study objective must be known and be clearly
described; e.g., determine if at the time  of the first
or second five-year review, the level of post-
remedy risk reduction is acceptable and the remedy
is expected to reach the remediation goal in the
predicted time frame. If the observed rate of
decline is less than the predicted rate, however, one
needs to decide if any changes to the Record of
Decision (ROD) are warranted. Additional
remediation may be needed in order to achieve
protection in a reasonable time frame. If the
selected remedy includes dredging or capping, an
expectation for an effective remedy could be that,
five years after remedy completion, there has been
a 50% decrease in the fish tissue levels and thus
substantial progress towards meeting the RAOs
and cleanup levels.
The time frame needed to demonstrate
reductions can
vary greatly
depending on
the type and
scope of
remedy
implemented.
For some
capping
remedies          Source: USEPA, Region 9
reductions in fish tissue levels may begin
shortly after completion of the capping. For
some dredging remedies, however, due to
resuspension and release of contaminants
throughout the dredging project and the
formation of a residual sediment layer, there
may be a short-term increase in fish tissue levels
before reductions are observed.  For monitored
natural recovery (MNR) remedies, the rate of
reduction should be similar to the rate observed
before remedy selection. To improve the
confidence in evaluating reductions in risk, at
least two sampling events should be conducted
by the time of the first five-year review.

Species
Tissue residue data should always be  analyzed
on a species-specific basis; i.e., tissues of
different species should not be combined. The
species should have a limited foraging range in
order to be representative of the exposures
caused by the contaminated sediment in the
study area.  This increases the likelihood that
the contaminant tissue level is representative of
the sediment and/or water exposure level at the
particular sampling location.

Generally, at least two species that are
commonly caught by local anglers or
subsistence fishers and represent different
trophic levels in the food chain should be
collected and analyzed for contaminant tissue

-------
Using Fish Tissue Data to Monitor Remedy Effectiveness
levels. For sites where the risk is based only on
ecological risk, important prey species for the
receptors of concern should be collected.  The
species should be reasonably easy to collect in
the future.  If it might be difficult to obtain a
sufficient number of a selected species of a
specified size range, an alternate species or a
different size range should be identified before
the sampling team is mobilized.

In inland waters, one bottom feeder and one
commonly sought after predator/game fish
species should be collected. To develop a
comparative national set of data for Superfund
remedies,  site managers are encouraged to use a
bottom-feeder, such as channel catfish, brown,
black or yellow bullhead, and a game fish such
as smallmouth bass, largemouth bass, or
walleye. Depending on the location, and the
needs of state and other trustee agencies, it may
also be important to sample a trout or salmon in
addition to, or instead  of, one of the other fishes.
In most saltwater bodies, two fish species or one
fish and one shellfish species that are commonly
sought after by recreational or subsistence
anglers should be collected.

Size/Age
To minimize variability, fish from the same age
class should be collected. As a surrogate for age
class, in order to save the expense of aging fish,
fish of similar length can be collected. The
relationship between length and age varies
greatly for each species and for each location
and depends greatly on water temperature,
population density, and food availability. A
good goal, however, is that the smallest fish
sampled is no less than 75% the length of the
largest fish (EPA-823-B-00-007, November
2000). State fish and game agencies will often
have data on this relationship, but for many of
the preferred species discussed above, a 4 year
old fish is often about  12 inches in length.  Fish
should also be within the legal size limit.  Fish
of this size/age are often more abundant, easier
to catch, and may respond sooner to reductions
to exposure concentrations after remedy
implementation than older, larger fish.
Although older fish often have higher
contaminant levels, the objective is to measure
changes in tissue levels, not to estimate
maximum concentrations. For sites where
ecological risk drove the remediation, sizes
should be consistent with the size of the prey
typically consumed by the receptor(s) of
concern.

Sex
It is recommended that collections offish be
done in a manner to avoid sampling unequal
numbers of males and females as much as
possible. Consult your local fisheries experts on
the life history of the species of interest. To
minimize variability, relatively similar numbers
of males and females should be analyzed.

Sample Locations
The optimum number of sample locations varies
depending on site size and on the extent of
variations in sediment conditions, habitat types,
and hydrology.  Because of the typical variation
observed historically in fish tissue residues from
different locations within the same water body,
fish should be collected from at least three site
locations. At large sites, or at riverine sites that
contain more than one impoundment, it may be
necessary to collect fish from more than three
sampling stations.  If there are areas that are
preferred fishing locations, these areas should
be given special consideration for sampling.

Sample Time
Fish should typically be collected at the same
time of year and under similar stream flow
conditions.  Since the lipids content can be low
in the spring, spring sampling should be avoided.
At sites where methylmercury is driving the need
for remediation, fish should be collected during
times  of active methylation, typically in late
summer or early fall. To help minimize
variability between individuals, sampling should
not be done 2-4 weeks before or after the
spawning season.

-------
Using Fish Tissue Data to Monitor Remedy Effectiveness
Sample Size and Type
The goal is to collect enough fish and analyze
enough fish samples to be able to determine
reliably whether adequate decreases in tissue
concentrations are occurring.  For example, the
hypothesis test may evaluate whether there has
been a 50% reduction in chemical residues in
fish five years after completion of the
remediation. For typical sites, in order to
control costs and help minimize variability, a
sampling plan should consider pooling fish into
composite samples rather then analyzing
individual fish.  A sampling plan that specifies a
minimum of 5 composites consisting of at least
5 fish per composite is often adequate to
determine with a confidence level of 90-95% if
the post-remediation concentrations have
decreased at least 50%. Although it depends on
the variance, if a smaller decrease needs to be
detected, more composites will typically be
needed. Increasing the number of composite
samples analyzed provides greater
improvements than increasing the number of
fish per composite in the ability to detect
smaller significant reductions in chemical
residues in fish (see EPA  Guidance for
Assessing Chemical Contaminant Data in Fish
Advisories, Vol. 1- Fish Sampling and Analysis,
Third Edition, EPA-823-B-00-007, November
2000). However, compared to the additional
cost for contaminant analyses of more
composites, the cost of collecting more fish per
composite is less.
  Source: The Great Lakes National Program
In such cases where little or no baseline data are
available, fish should be collected during
remedial design in order to have a basis for
future comparisons offish tissue residues.
Since there is no estimate of variance to use in a
power analysis, it may be beneficial to collect
more than five composites and/or more than five
fish per composite. Then, based upon the
measured population variance of the initial
samples,  the design can be changed to collect
the most  appropriate number of composite
samples that would allow adequate
determination of mean residue concentrations.
This will allow site managers to maximize
certainty  in their conclusions about any changes
in the mean residue levels as a result of the
remediation.

When contaminant concentrations in tissue are
measured, lipid contents of the tissue should be
determined.  Lipid  determination allows for
lipid normalization of the data and assists in
data interpretation and calculation of BSAFs.
When BSAFs are used, it is just as important
that the contaminant concentrations in sediment
are normalized on a percent organic carbon
basis. This is particularly important when
evaluating non-polar compounds and strongly
bioaccumulating compounds that have an
affinity for lipids. However, lipid  data can also
be used to evaluate the relative health or status
of an organism. Individuals with low lipid
contents may be unhealthy, starved, or may
have recently lost lipid-soluble contaminants
due to egg laying, thereby transferring
contaminants and biasing the bioaccumulation
data. Data from fish with unusually low lipids
relative to data from other sampling events
should be reported  but discounted in the
analyses.

When measuring lipid contents, the specific
analytical technique should be considered.
Total non-polar lipids is an acceptable
determination, and EPA (2000) recommends
using the dichloromethane extraction solvent
method.
  Contact Information
  For questions on this fact sheet, please contact
  Marc Greenberg (732.452.6413), or
  Stephen Ells (703.603.8822) of OSRTI, or
  Lawrence Burkhard (218.529.5164) of ORD.

-------
Using Fish Tissue Data to Monitor Remedy Effectiveness
References
Rypel, A. L., R. H. Findlay, J. B. Mitchell, D. R. Bayne.  2007.  Variations in PCB concentrations
      between genders of six warmwater fish species in Lake Logan Martin, Alabama, USA.
      Chemosphere. 68:1707-1715.

U. S. Environmental Protection Agency. 2000.  Guidance for Assessing Chemical Contaminant Data in
      Fish Advisories, Vol. 1, Fish Sampling and Analysis, Third Edition. EPA-823-B-00-007. Office
      of Water.  Washington, DC.

U.S. Environmental Protection Agency. 2003. Methodology for Deriving Ambient Water Quality
      Criteria for the Protection of Human Health (2000).  Technical Support Document Volume 2:
      Development of National Bioaccumulation Factors. EPA-822-R-03-030, Office of Water,
      Washington, DC.

U.S. Environmental Protection Agency. 2005.  Contaminated Sediment Remediation Guidance for
      Hazardous Waste Sites. EPA-540-R-05-012.  Office of Solid Waste and Emergency Response
      (OSWER) Directive 9355.0-85. Washington, DC.

U.S. Environmental Protection Agency. 2007.  Framework for Metals Risk Assessment. EPA

-------
Using Fish Tissue Data to Monitor Remedy Effectiveness
Appendix:
Significance and Power of Residue Comparisons; Refinement Techniques
To determine the most appropriate number of individual fish per composite, and the number of
composite samples to analyze, statistical precision and power analyses may be performed.  Many
statistical packages, e.g., SAS, PASS, G*Power, R, and S-Plus, provide tools for sample size estimation,
hypothesis testing, and power analysis.

For readers desiring detailed information on statistical power analysis, consult Cohen (1998).  For a less
detailed, but well-written and informative description of power analysis as it pertains to sampling
design, see EPA (1987) (EPA/430/09-87-003, Bioaccumulation Monitoring Guidance:  Strategies for
Sample Replication and Compositing - Volume 5). Note, this guidance document was written to detect
significant differences in chemical residues between sampling stations. Further note, the analysis for
detecting differences between sampling stations is same as that for detecting differences between
sampling dates.

Other Federal Agencies have documents and information on statistical power analysis and the
determination of sample sizes.  The U.S. Fish and Wildlife Service (USFWS) has a Web site,
http://www.umesc.usgs.gov/ltrmp/stats/statistics.html, that provides detailed information on Sampling
Design and Statistics for their Long Term Resource Monitoring Program (LTRMP).  The site discusses
these techniques by using primarily population endpoint examples. However, these techniques are
appropriate for examining changes in chemical residues in fish. The U.S. Army Corps of Engineers
(US-ACE) has a document that discusses these techniques in the context of evaluating dredged
sediments (Clarke and Brandon). Additionally, EPA and US-ACE have jointly published a document on
sediment evaluation that includes an extensive  section on statistical methods (EPA 1998) covering
power analysis and determination of sample size.

Sampling plans should use a Type I error rate (false positives; i.e., the apparent decrease is not real) of
a=0.05 or 0.10, while minimizing the Type II error rate (false negatives; i.e., the decrease although not
detected is real), and maximizing the statistical power, (i.e., 1-P).  In order to perform statistical precision
and power analyses, a good/accurate estimate of the population variance from baseline data is required.

In Figure A-l, sampling precision curves are provided for three sampling designs for four different
population variances (expressed in terms of coefficients of variation (CVs)). Using an a=0.05, five fish
per composite will provide low relative errors (i.e., adequate sampling precision to obtain the mean)
with the collection of five or more replicate composites when the coefficient of variation (CV) between
fish is below 75% (Figure A-l). In other words, 95% of the time, the  error in estimation of mean tissue
concentration will be less than 30%, 20%, and  10% when the CV is less than or equal to 75%, 50%, and
25%, respectively.  Figure A-l was determined using the equation (Snedecor & Cochran, 1989): s2 =
(CV- zi_a/2)2/(mn) where e is the relative error in estimation, CFis the  coefficient of variation, z(i_a/2) is z-
value from a normal distribution, m is the number offish per composite, and n is the number of
composites.  Based upon EPA's analyses on the interchange between number offish per composite and
total number of composites taken, increasing the number of replicate composite samples will have a
greater impact than increases to the number of individual fish per composite upon precision of the
estimate (EPA-823-B-00-007, November 2000).  Increasing either the number offish per composite or
the number of composite samples taken will further improve precision.
                                             A-1

-------
Using Fish Tissue Data to Monitor Remedy Effectiveness
In Figure A-2, the Power to detect a minimum detectable decrease (MDD) in tissue concentrations is
shown to range widely depending on the desired MDD and the CV between sample means. Higher
levels of Power can be achieved for MDDs between 35-50% for sample sizes (i.e., number of replicate
composites) between 5 and 10 when CV=25%. Similarly, for CVs as high as 50%, reasonable levels of
Power can be expected when sample sizes approach 10 replicate composites. The precision and Power
analyses are closely interrelated in that when one evaluates Power for a composite sampling program,
the total number of individual fish to be collected for the composite samples shouldn't become
exceeding large and not manageable from a collection and/or ecosystem capacity standpoints.

To illustrate, the following example is provided.
  Assume:
      Baseline data:       5 composite samples with 5 fish per composite
                          Average  of 5 composites                  =1.5 mg/kg (ww)
                          Standard deviation of average of composites = 0.3

  Compute Population standard deviation (a)
             var ( z ) = G2/(nm)
             var(z) = (0.3)2
             n = 5         5 composite samples
             m = 5        5 fish per composite
             Population standard deviation       = ((0.3)2 x (nm))1/2 = (0.32 x 5 x 5)'/2
             Population standard deviation       =1.5
             Population coefficient of variation   = 100%

  Determine Minimum Detectable Decrease (MDD) with 80% Power (1-P)

      a = 0.10
      P = 0.20
      Power = 1-P = 0.80
      a' = 0.10 (one-sided test) where a' is a for a one-sided test and a/2 for a two-sided test.
      n-a'= 1.282
      zi.p = 0.8416
      n = 5
      m = 5
       c  1.52(1.282 + 0.8416)2   1.2822
       5>	^	'— +	
                 MDD2           2
       MDD-
               'l.52(1.282 + 0.8416)2
                                             A-2

-------
Using Fish Tissue Data to Monitor Remedy Effectiveness
Relative precision of estimated mean concentration

       s2=(CV*Zl_al2)2l(mn)

       s2 = (1. 00 *1.645)2 7(5*5) = 0.11
References

Clarke, J.U. and D.L. Brandon.  1996. Applications Guide for Statistical Analyses in Dredged Sediment
      Evaluations. U.S. Army Corps of Engineers, Washington, DC. Miscellaneous Paper D-96-2.

Cohen, J.  1988. Statistical Power Analysis for the Behavioral Sciences (2nd Edition).  Lawrence
      Erlbaum, Hillsdale, New Jersey.  567 pp.

Snedecor, G.W. and W. G. Cochran. 1989. Statistical Methods, 8th ed. Blackwell Publishing, Ames,
      Iowa.  503 pp.

U.S. Environmental Protection Agency.  1987. Bioaccumulation Monitoring Guidance: Strategies for
      Sample Replication and Compositing  Volume 5. EPA/430/09-87-003.  Office of Water.
      Washington, DC.

USEPA.  1998. Evaluation of Dredged Material Proposed for Discharge in Waters of the U.S. - Testing
      Manual. Inland Testing Manual. U.S. EPA Office of Water. EPA 823-B-98-004.  February
       1998.
                                            A-3

-------
Using Fish Tissue Data to Monitor Remedy Effectiveness

1.0-

2 0.8 -
LU
Ť 0.6-
3 0.4-
tŁ
0.2
o.o -
m=3
\
\
* \
\ X^
\ '•--... ~--


08 -

0.6 -
0.2 -
0.0 -
m=5
i

.V
•. \
\\
\ * ^
\ '-. *V
N X., "^^__


0.6 -

0.4 -
03 -
0.2 -
0.1 -
0.0 -
m-10
i

.V
•. \
\\
X '. X
\ \ X
X '-. "~^~









p\/— n oc;
	 CV=0.50
	 CV=0.75
	 CV=1.00







            5    10    15   20    25
               N (number of Composites)
                                      30
       5    10   15    20    25
          N (number of Composites)
                                                                             30
        5    10    15    20    25
          N (number of Composites)
                                                                                                                   30
Figure A-l.  Precision curves of relative error versus number of composites (N) for various values of the coefficient of variation (CV)
when a= 0.05. Calculations based m= 3, 5, or 10 individual fish per composite.
                   CV = 0.25
     1.0 -
     0.9 -
     0.8 -
     0.7 -
     0.6 -
     0.5 -
     0.4 -
     0.3 -
     0.2 -
     0.1 -
     0.0 -
             5    10   15   20   25
            N (number of composite samples)
                                     30
1.0 -
0.9 -
0.8 -
0.7 -
0.6 -
0.5 -
0.4 -
0.3 -
0.2 -
0.1 -
0.0 -
                                                          CV = 0.50
   0    5    10    15    20    25    30
       N (number of composite samples)
1.0 -
0.9 -
0.8 -
0.7 -
0.6 -
0.5 -
0.4 -
0.3 -
0.2 -
0.1 -
0.0 -
                                                                                                 CV = 0.75
        5    10    15    20    25
       N (number of composite samples)
                                                                                                                   30
Figure A-2.  Power (1-P) curves for a 25%, 35%, or 50% minimum detectable decrease (MDD) in contaminant concentrations between
means at a= 0.05, two-tailed, when the coefficient of variation (CV) for between-sample variability is 0.25, 0.50, and 0.75.
                                                                  A-4

-------