WATER POLLUTION CONTROL RESEARCH SERIES 16130 04/70
GUIDELINES:
BIOLOGICAL SURVEYS
AT PROPOSED HEAT DISCHARGE SITES
ENVIRONMENTAL PROTECTION AGENCY WATi'R QUALITY OFFICE
-------
WATER POLLUTION CONTROL RESEARCH SERIES
The Water Pollution Control Research Series describes the
results and progress in the control and abatement of pollu-
tion of our Nation's waters. They provide a central source
of information on the research, development, and demon-
stration activities of the Water Quality Office, Environ-
mental Protection Agency, through inhouse research and grants
and contracts with Federal, State, and local agencies, re-
search institutions, and industrial organizations.
Inquiries pertaining to the Water Pollution Control Research
Reports should be directed to the Head, Project Reports
System, Office of Research and Development, Water Quality
Office, Environmental Protection Agency, Washington, B.C. 20242.
-------
GUIDELINES: BIOLOGICAL SURVEYS AT PROPOSED
HEAT DISCHARGE SITES
by
Ronald R. Garton, Ph.D.
Research Aquatic Biologist
National Thermal Pollution Research Program
Pacific Northwest Water Laboratory
Corvallis, Oregon 97330
and
Ralph D. Harkins, Ph.D.
Mathematical Statistician
Chief, Pollution Surveillance Branch
Robert S. Kerr Water Research Center
Ada, Oklahoma 74820
ENVIRONMENTAL PROTECTION AGENCY
Water Quality Office, Northwest Region
Pacific Northwest Water Laboratory
200 Southwest 35th Street
Corvallis, Oregon 97330
April 1970
For sale by the Superintendent of Documents, U.S. Government Printing Office
Washington, D.C., 20402 - Price $1
-------
EPA Beview Notice
This report has been reviewed by the Water Quality Office,
EPA, and approved for publication. Approval does not signi-
fy that the contents necessarily reflect the views and poli-
cies of the Environmental Protection Agency, nor does mention
of trade names or commercial products constitute endorsement
or recommendation for use.
-------
PREFACE
The key words in the title of these Guidelines are "proposed
--- sites." The modern goal of environmental quality control on
many fronts is pollution prevention and environmental protection.
The horizons are moving out past simple abatement of existing pol-
lution and recovery from damages. Heat pollution is not unique in
this respect.
In keeping pace with the changing times, the scientist and
engineer are faced with different and more complex problems. It
had been a comparatively simple matter to survey an existing source
of heat pollution and to document its effects by measuring tempera-
ture gradients of the receiving water and collecting biologic data
within and without the zones of temperature increase. In survey-
ing a proposed discharge site, the investigators are faced with
two new challenges. First, they must design a sampling program
without absolute knowledge of how the heat will behave, collect
data to assay the biologic resource in jeopardy, and assure that
the data are adequate for defensible pre- and post-operational
comparison in the event that some heat discharge does occur.
Second, their pre-operational data and their form of presentation
must be sufficiently conclusive to influence and support very
important, and probably costly, management decisions.
-------
The authors present here general survey procedures that have
a reasonably high probability of success anywhere. Deviations
from the suggested procedures should be made only on the basis of
special knowledge of the peculiarities of a specific site and for
the purpose o.f raising the probability of success.
ii
-------
ABSTRACT
A quantitative approach is presented for the biological
portion of thermal discharge siting surveys and discharge monitor-
ing. Three types of studies are covered: Type I is a very general
study of the aquatic system and the pertinent literature; Type II
is a comprehensive pre-operational study designed to supply data
for management decisions on power plant siting and data to' serve
as baseline for possible future comparison; and Type III is a post-
operational continuation of Type II to detect possible effects if
a thermal discharge to a natural water body is allowed.
Two methods are recommended for location of sampling stations
by use of a grid system based on planned outfall design. Sample
collection and handling methods and frequency of sampling are
suggested for fish, macroinvertebrates, plankton, periphyton and
aquatic macrophytes.
Methods of data handling recommended include diversity and
redundancy indices and a combination of the two into one value for
a test for dispersion. A scale of importance is suggested for
organisms of special value in either an economic or ecologic sense.
For statistical analysis of the data, four appropriate methods are
recommended and sample problems are provided to illustrate data
handling and conclusions to be drawn from the tests.
Key Words: Thermal pollution, siting surveys, biological
sampling, data analysis, thermal power plants, ecological studies.
iii
-------
CONTENTS
Chapter Page
INTRODUCTION ">
PLANNING THE SURVEY 5
Survey Types 5
Organisms to be Sampled ... ..... 6
Location of Sampling Stations ..... ''
Rivers 12
Lakes, Oceans or Estuaries ^
Sampling Methods 17
Fish 18
Macroinvertebrates 22
Plankton 24
Planktonic Animals 24
Phytoplankton . 25
Periphyton 26
Aquatic Macrophytes. .... 27
Chemical and Physical Parameters 28
DESIGNING THE STATISTICAL MODEL. .... 31
Fish. 32
Macroinvertebrates . 34
Plankton, Periphyton, and Aquatic Macrophytes .... 40
STATISTICAL ANALYSIS .......... 43
APPENDIX 47
The Factorial Arrangement: AOV (Parametric) 47
Sample Size Estimation and Confidence
Intervals 64
The Factorial Arrangement: AOV (Nonparametric). ... 66
The Factorial Arrangement: AOCV (Parametric) 77
The Factorial Arrangement: AOCV (Nonparametric) ... 92
IV
-------
CONTENTS (CONT.)
Page
LITERATURE CITED ..................... 95
BIBLIOGRAPHY ....................... 101
-------
LIST OF FIGURES
Figure Page
1 Transect Grid for River Sampling 13
2 Transect Grid for Lake or Ocean
Sampling 16
3 Sampling Grid and Heated Area for Example
Problem 52
VI
-------
LIST OF TABLES
Table Page
A-l Raw Data Sheet 49
A-2 Sample Problem Raw Data Sheet (Weights
per Unit Area Sampled) 50
A-3 Interaction Table 54
A-4 Sample Problem Interaction Table (Weights
per Unit Area Sampled) 55
A-5 Analysis of Variance . . 60
A-6 Sample Problem Analysis of Variance 61
A-7 Contingency Table for Nonparametric AOV 69
A-8 Sample Problem Raw Data Sheet (Temperature) ... 79
A-9 Sample Problem Interaction Table
(Temperature) 80
A-10 Sample Problem Analysis of Covariance 88
vii
-------
INTRODUCTION
This publication provides a plan for prediction and detection
of biological effects of thermal additions to the aquatic
environment. A quantitative approach is presented for pre-
installation sampling at sites of potential thermal discharges and
for monitoring of biological effects attributed to these discharges.
Waste heat is a significant pollutant in the United States.
Heat added to natural bodies of water in this country comes from
many sources including industrial processes, air-conditioners,
and irrigation, but the largest amount comes from the electric
power generating industry. In 1964, almost one-half of the water
used in the United States was for cooling and condensing by the
power and manufacturing industry (1). Electric power plants, which
use stream, reservoir and lake waters to cool condensers, used
about 80 percent of this cooling water. Therefore, the projected
growth of electric power production provides a good index of
thermal pollution potential for the near future. Power generation
has approximately doubled each ten years during this century and
estimates of future demands indicate a shortening of the time span
for similar increases (1, 2). Heat rejection from nuclear and fossil
power plants is expected to increase almost nine-fold by the year
2000 (3) and there will be an accompanying rise in waste heat output
from manufacturing industries. The net effect will be a great in-
crease in the amount of waste heat rejected to the environment.
-------
This additional waste heat must be managed in a manner that
will maintain or enhance the physical, chemical, and biological
nature of water resources. The environment must be protected by
judicious selection of industrial sites for development, by in-
corporation of waste heat treatment into plant design and operation,
or by regulation of discharge of waste heat to conform with water
quality standards. In all cases, the potential ecological effect
of thermal discharges must be considered.
To predict the effects of thermal discharges on the aquatic
environment and on our use of this public resource, we must first
be able to assess the nature of the environment as it is at present.
To do this, we must have adequate and reliable techniques for
sampling and evaluating the aquatic biota in the areas of potential
effect.
A biological survey conducted at a potential site of thermal
discharge should:
1. Establish the nature and value of the resource in
jeopardy.
2. Provide a basis for projection of consequences of
thermal discharges and for determination of controls
to prevent detrimental effects.
3. Provide a baseline which can be used for future
comparison.
-------
If a thermal discharge is allowed, a monitoring program
should provide data for detection of possible changes in the biota,
Minimum amount of time and expense is required to accomplish
the task of evaluation if the essential data are obtained without
wasted effort and the data are of maximum utility. These points
are considered basic to the sampling program described in this
paper.
The emphasis in this paper is upon the biota, hence, results
should be interpreted by an aquatic ecologist. However, the
complexity of the problem requires a multidisciplinary approach
and the advice of competent chemists, engineers, and statisticians
must be available if needed.
-------
PLANNING THE SURVEY
Careful planning of the survey is necessary to assure
(1) that adequate information is obtained for a true characterization
of the aquatic community and (2) that valuable time and energy are
not spent collecting useless data.
Survey Types
Site surveys fall into three broad categories based upon
the amount of detail required at the time. The three types of
surveys fit into a natural sequence.
The first type is a very general survey of the area and of the
pertinent literature. This survey could be conducted in a very
short time with a minimum amount of manpower and expense. It
provides a general idea of the kinds of organisms present and
a rough idea of their relative numbers. In some cases, this
information might be sufficient in itself to determine whether a
proposed thermal power plant should be installed and if once-through
cooling should be allowed. As an example, if the preliminary
survey shows that the proposed site is an important spawning area
for cold-water fish, a heated effluent should not be allowed in this
area. When the preliminary survey is not conclusive, it can be used
in planning the second type of survey.
The second type is a much more detailed survey requiring more
manpower and money. This survey should provide an appraisal of
-------
numbers and kinds of organisms present as well as data on
important chemical and physical characteristics of the environment.
This survey should furnish regulatory agencies with the information
needed for the final decision as to whether a thermal discharge
should be allowed at that proposed site, and if so, what controls
should be imposed.
The third survey type is a continuation of the second. If the
power plant or other potential source of thermal pollution is
installed, the receiving body of water must be monitored to detect
any changes and to plan remedial action if such changes are detrimental
to the resource. Because of the variability from year to year in
biota and meteorological conditions, the combination of the second
and third type surveys should include at least three years pre-
operational data and at least three years post-operational data.
This may vary somewhat, depending on the nature of the area and
the degree of annual variability encountered. The plans for the
second and third phases should be the same since both sets of data
must be complete enough to give a detailed and reliable picture of
the numbers and kinds of biota present. In these phases, adequate
statistical planning is essential to ensure that post-operational
data can be compared with pre-operational data to detect significant
changes.
-------
Organisms to be Sampled
Since the basic purpose of these surveys is to aid in
protection of the public resources, it is logical that most of the
organisms studied will be those which are of special importance to
the public. These include such commercial species as salmon,
shellfish, etc., and fish such as trout and bass which are important
for recreational purposes. Essential organisms in the food web
of these species are also important. Nuisance forms such as blue-
green algae may be very significant because their presence decreases
the recreational value of a body of water.
Expected changes in fish populations will vary with species
and with the chemical and physical characteristics of the receiving
water. In an extreme case, the fish may be killed by the hot water,
but the effects are nearly always more subtle. Fish may be driven
out of a heated area at high temperatures, but attracted at lower
temperatures. Not all fish will be affected in the same way or to
the same degree, so there may be changes in species composition.
Fish passageways or spawning areas may be blocked to eliminate a
species in some areas, but overall productivity may be increased in
others. The range of possibilities is great and the sampling
problems will be unique to each site.
Benthic macroinvertebrates are sampled for three principal
reasons:
6
-------
1. They are important members of the food web and their
well-being is reflected in the well-being of the
higher forms such as fish.
2. Many invertebrates, such as the marine shellfish, are
important for commercial and recreational use.
3. Because of their relative immobility, they are
especially good indicators of pollution at the bottom
of the water column.
A feature of their use as indicators of pollution is the
fact that they do not reflect average levels of pollution, but are
more likely to reflect the extremes. The invertebrates cannot
avoid pollutants and many are slow to repopulate an area, so
detrimental effects are not easily erased. Consequently, the
structure of the community may be determined during those months
in which temperatures are highest.
The use of benthic species for detection of thermal effects
is limited by the unique property of the heated effluent water. The
effluent may be dispersed to float on the underlying receiving water
without ever coming in contact with the benthos. Still, benthic
sampling cannot be eliminated because a floating heated layer may
promote oxygen problems and may kill emerging aquatic insects or
planktonic stages of marine invertebrates.
-------
Zooplankton are essential in most aquatic systems, but due to
the high variability in numbers and species composition, it is
very difficult to arrive at valid conclusions using zooplankton
data unless samples are taken in greater numbers and with more
frequency than will usually be practical in the type survey being
proposed here. Also, in a body of flowing water, the zooplankton
at any spot are largely the product of upstream conditions and not
of conditions at the point of sampling. Because of these problems,
pre-operational zooplankton data at potential sites of heat
discharges are of dubious worth and only limited and specialized
plankton sampling is recommended.
In some areas, it will be necessary to sample for the immature
planktonic forms of fish and shellfish. In estuaries or the ocean,
the planktonic eggs and larvae of many species may be killed by
passage through the condenser system. Thus, the invertebrate
community could be severely harmed without the hot water layer ever
touching the bottom where the adults are found. In fresh water,
the same thing may happen to young fish. The presence of such
planktonic forms must be determined as part of the Phase I and II
siting surveys.
Phytoplankton are considered primarily because of their
potential for nuisance. An increase in water temperature due to
waste heat may cause an increase in productivity of phytoplankton
and also cause a species shift to less desirable forms. Cairns (4)
-------
found that while there is significant species overlap, it is evident
that a rise in temperature is generally accompanied by a gradual
shift in algal species from the diatoms to the greens and finally
to the blue-greens at the highest temperatures. In terms of
desirable productivity for fish and other aquatic life, the shift
would be from more to less desirable forms. Eutrophication of
lakes and reservoirs is already becoming a very serious problem
in the United States and addition of waste heat could contribute to
this problem.
Periphyton samples are taken for an estimate of primary
productivity. Since periphyton includes the attached algae, an
estimation of algal growth is obtained with less problem of natural
variability than is encountered in sampling phytoplankton.
Aquatic macrophytes provide both food and shelter for fish and
fish-food organisms, and they may provide important feeding and
nesting sites for shore birds and waterfowl. Perhaps even more
importantly, aquatic macrophytes may become too plentiful and
create a nuisance by clogging water intakes and by choking canals
and bays to such an extent that boating is impossible and flow
for irrigation or other uses is severely restricted.
An increase in temperature may cause a shift in species
composition of aquatic plants or may cause accelerated growth of
the plants present. Anderson (5) stated that at Chalk Point power
plant on Chesapeake Bay, a species of Ruppia which grew profusely
-------
in the area and was used as food for migrating waterfowl was
replaced by a species of Potamogeton within two years after
the power plant began to discharge heated water into the bay.
On reservoirs administered by the Tennessee Valley Authority, there
is already a serious problem due to clogging of waterways by large
growths of Myriophyllum. Ceratophyllum and El odea have produced
similar problems in other places. These problems have not all
been caused specifically by rises in temperature. However, a
temperature rise might promote the problem in areas where additional
heat could increase the length of the growing season. It is
especially important to prevent the growth of such plants because
eradication becomes a long and costly procedure once larger areas
are infested.
Measurement of selected chemical and physical parameters should
be included in Type II and III surveys. Without such data, it will
be difficult to determine conclusively whether biologic effects
result from temperature alone, other pollutants associated with the
discharge, or other causes.
Location of Sampling Stations
To ensure that the sampling effort will result in data which
can be treated statistically, the sampling areas and times must
be determined in advance. Sample stations should be located on
a grid system designed to cover the heat-affected portion of the
water body and a portion of the surrounding area. The heat-affected
10
-------
areas will not be known precisely prior to installation of the
plant, so the location of the sample grid will be based upon
engineering predictions of movement and behavior of the heated
effluent.
The two sampling grids presented in this section are for
example and will not fit every situation. The grids should be
modified to fit the morphology and hydrology of the site, but the
basic idea of sampling within and outside of the proposed heated
area can still be followed.
Rivers
In a river, samples should be taken on at least four transects
(See Figure 1) located in relation to the plume area. References
(6, 7, 8, 9, 10) may be consulted for assistance in predicting the
extent of the thermal plume.
The first transect should be close to the effluent on the
upstream side, but out of the influence of the heated discharge.
The second transect should be close to the effluent on the down-
stream side and within the heated plume. The third transect
should be placed at approximately that point where the surface
temperature in the center of the plume has dropped to the point
halfway between ambient river temperature upstream from the
effluent and temperature of the effluent (0.5AT). The fourth
transect should be located at the point where a rise in river
11
-------
.9w .7w .3w .Iw
Close above Effluent
J EFFLUENT
Close below Effluent
(0.5^7 at surface
O.I A j at surface
FIGURE 1 TRANSECT GRID FOR RIVER SAMPLING
12
-------
temperature is barely detectable (0.1AT). Location of the lower
two transects will be determined for that time of year when the
river within the plume area is expected to be at maximum tempera-
ture and will be fixed thereafter.
For all organisms except fish, four sampling stations will
be established on each transect. These stations will be located
at 10, 30, 70, and 90 percent of the width at that time of year
for which the downstream reach of the plume is calculated. If the
nearshore area is appreciably different from the center sections
(more shallow, weedy, etc.), additional stations should be located
between the banks and the 10 and 90 percent stations. In a river
in which width fluctuates widely, it is important that the sampling
stations be located where they will not be dry at low flow. The
four or more stations on each transect should be permanently buoyed
so that samples for each period can be taken at the same spot.
Depths will be mentioned later for each type of sample.
For fish, the river should be divided along each transect
into three equal widths consisting of one-third of the width from
each bank and one-third in mid-channel. Samples will be taken in
these areas in the general vicinity of the transect line since
most fishing gear does not lend itself to precise placement on a line.
Additional stations may be necessary to include special migratory
routes or other areas of fish concentration.
For a post-operational survey, the same stations will be
sampled, but additional transect lines may be required to provide
13
-------
sufficient information to establish the boundaries of thermal
effect.
Lakes, Oceans or Estuaries
In lakes, reservoirs, or estuaries where the flow is of
sufficient velocity and persistence to resemble a river, the river
transect plan may be applied. But in most large lakes or at ocean
sites, a different sampling plan based upon distance from the out-
fall is warranted.
In these cases, a grid system should be established to include
the area of expected temperature change and enough of the surrounding
area to provide an unaffected zone for comparison.
A suggested method is to locate sampling stations on the inter-
sections of concentric circles about the outfall and transect lines
radiating from the outfall (See Figure 2).
The concentric circles should be placed to pass through the
most distant points where surface temperature increases above
ambient are predicted to be 0.9AT, 0.5AT and 0.1 AT at that time
of the year when the highest temperatures and greatest affected
area are predicted. References (11, 12, 13, 14) may be consulted
for assistance in predicting the extent of the thermal plume.
To set up the transect lines, determine the number of degrees
of angle of open water in the outermost concentric circle. Run the
transect lines from the point of effluent out through the points on
-------
8
p.l A T at surface
0.5 AT at surface
0.9AT at surface
EFFLUENT
20
FIGURE 2 TRANSECT GRID FOR LAKE
OR OCEAN SAMPLING
-------
the circle at 20, 40, 60, and 80 percent of the angular distance
from the shore on one side of the effluent to the shore on the other
side. For example: if the shoreline runs north and south, the open
water portion of the circle will be 180°. The points on the circle
will be at 20, 40, 60, and 80 percent of 180° or at the 036, 072,
108, and 144 degree points in relation to north.
This suggested method for setting up the sampling grid may
have to be changed to fit local conditions. More transects should
be added if the plant is on a point where the open-water portion
is so great as to put the transects too far apart and additional
samples may be needed along the shoreline if the 20 and 80 percent
radials are too far out to ensure sampling of shallow inshore areas.
Sampling Methods
Sampling methods are recommended herein for many kinds of
organisms in different environments, but it is not expected that
all types of sampling would be used on each survey. The preferable
sampling methods differ with each individual case. In each survey,
it will be desirable to sample benthos, fish, etc., but the methods
of taking the samples are to be chosen by the biologist-in-charge
to suit the situation.
Sampling methods are presented for organisms grouped by type
or by habitat preference. The statistical treatment of the data
discussed in sections to follow should be determined at the beginning
of the survey to ensure that sampling and analysis methods are
compatible.
16
-------
Fish
Duplicate samples should be taken at each station to quantify
the variability in sampling results. This may be done by sampling
during two successive periods or by sampling in two spots at a
station simultaneously. Fishing methods, once chosen, should
remain constant, and the sampling series should be conducted at the
same time each year.
Sampling should be conducted with sufficient frequency to
detect seasonal migrations and fluctuations in the population.
For example, anadromous species are in a river during two specific
periods. The young fish drift downstream after they hatch and,
later, the same fish move back upstream to spawn. Sampling times
and methods must be planned to sample the young fish moving down-
stream and the adults coming back up. In this case, there would
be just two general sampling periods per species. But there may
be several anadromous species in one river so several sampling
periods would be needed to catch every one. The same principle
may be applied to bass or any other warmwater fish which moves into
special areas of lakes or streams for spawning. Sampling for non-
anadromous species should be conducted at least four times each
year to coincide with the four seasons.
Each sampling series should provide the following: (1) identi-
fication of species present; (2) number of individuals of each
species; (3) biomass of each species; and (4) scale samples, if
17
-------
suitable, for later use in age and growth studies. Numbers and
weights may be converted to numbers or weights per catch effort
if effort varies significantly from one series to another.
The following is a list of suggested sampling methods. The
methods and gear must be tailored to each species at each site and
a combination will often be required to sample adequately every
important species of fish. The important thing is to obtain
adequate reliable data by whichever means are most suitable.
1. Drag seine - Relatively nonselective for both species
and size (very small fish may go through the mesh), but its use is
generally limited to small rivers or embayments where fish may be
trapped in small areas. Hoover (15) checked by mark and recapture
and dynamiting on small trout streams and determined an efficiency
of 70 to 100 percent. Gerking (16) found 88 percent efficiency
in a warmwater stream in Indiana. In thermal studies, it would be
most useful in backwaters or embayments or in small streams used
for industrial discharge.
2. Gill nets and trammel nets - These are used in both streams
and still water such as in lakes or estuaries. Their use is limited
by current which tends to clog the net with floating debris; drifted
nets are hampered by obstructions on the bottom. They are selective
for species and size of fish, but can be adapted by changing mesh
size, deary and Greenbank (17) found that they did not work well
with deep-bodied fish such as the Centrarchids, but that they are
quite efficient for many other fishes.
18
-------
3. Traps - There are several types of trap nets so the best
one must be chosen to fit the species and the morphology of the
area. Hoop nets and wing nets are limited by current and move-
ment of the fish; however, Cleary and Greenbank (17) state that
workers in eastern Iowa and field men of the Upper Mississippi
River Conservation Committee have found the trap net, despite its
limitations, to be the best all-around method of taking qualitative
and quantitative samples of fish in large streams. A mobile fish
trap suspended between two outboard motor boats was used
successfully to capture young salmon in Bellingham Bay, Washington
(18).
4. Electric shockers - Their use has been primarily in
small streams and lakes (19). Large streams and reservoirs would
be very hard to sample this way, but the shocker may be used in
special situations where fish are restricted to shallow areas.
5. Townets and trawls - These nets are useful for large
reaches of open water such as estuaries, bays, reservoirs and the
ocean where they have been used extensively for studies on Pacific
salmon (18, 20). The depth of tow must be adjusted to the fish
sought and morphology of the area.
6. Mark and recapture - This method entails release of
captured marked fish and subsequent recapture by any one of the
fishing methods already mentioned. A ratio is then set up for
each species:
19
-------
Marked fish recaptured = Total fish captured
Total marked fish Total fish in the water body '
to estimate total numbers of fish (20, 21). This method assumes
no significant loss or recruitment to the population between
capture series.
7. Electronic tracking - Tags which give off sonic or radio
waves are useful for tracing migration patterns of anadromous
fishes (22, 23) and may be very helpful in determining fish
passageways or other key sampling areas.
These are not the only available methods and no one will work
in all cases, but they are examples of successful methods being
used for fisheries surveys.
Where a substantial commercial or recreational fishery exists,
catch records by periods or by units of effort may furnish valuable
data for estimations of the fishery. They are difficult to interpret
as a measure of absolute numbers of fish present, but they do provide
information on fluctuations in abundance. If the commercial or
recreational fishery is especially important, the catch records may
be as pertinent as records of total abundance of the species.
However, consideration must be made of the fact that commercial
catch records often are influenced by factors other than abundance,
such as fluctuating price of fish, local opportunity for other more
lucrative jobs, or fishing techniques used.
20
-------
The ideal study of the fishery would include estimates of all
species of fish regardless of their commercial or forage value.
However, if budget or other considerations limit the intensity of
the study, it is more practical to work on a few species of fish
such as those receiving heaviest angling pressure. In this case,
specialized methods can be developed for the fish in question and
such population characteristics as ratio of adults to subadults,
rates of growth, etc., can be determined.
The information obtained must be used by regional ecologists
to predict (or, in Type III, detect) heat-induced changes in the
fish populations. This requires adequate temperature-tolerance
data for the fish and adequate data are not now available for all
species. In such cases, the workers will have to establish these
tolerances at the time of the survey by means of bioassay experi-
ments. The predictions of the ecologist can then be used to judge
the suitability of the site for a power plant and the possible
restrictions which must be imposed if the site is accepted.
Macroi nvertebrates
Invertebrates should be sampled at those permanently buoyed
locations mentioned in the section on location of sampling stations
and all series should be taken at the same locations. Duplicate
samples should be taken at each site, and the following information
recorded from each sample: (1) identification of species (or genus
21
-------
if appropriate); (2) numbers of individuals per species; (3) biomass
per species.
Since many of the invertebrates (aquatic insects, for example)
have relatively short life cycles, the sampling program must be
carried on throughout the year to detect seasonal changes.
Invertebrate bottom samples should be taken a minimum of once each
quarter. Artificial substrate samplers should be changed every
six weeks unless growth is so heavy that a shorter period is
indicated. In this case, they should still be retrieved at six-
week intervals, but set out so as to be exposed for a shorter time.
Sampling devices must be suited to the substrate and the
organisms to be sampled, but generally the choice will be made from
a few methods or a combination of methods. Before methods are
determined for the long-term Type II and III surveys, the preliminary,
Type I, survey should be conducted. This quick survey with a variety
of gear will enable the researchers to determine roughly the kinds
of organisms and type of substrate to be sampled so gear can be
chosen for the extended Type II and III surveys.
The following references will be useful for selection of
sampling methods and gear:
Bottom samplers - 24
SCUBA sampling methods - 25, 26
Artificial substrates - 24, 27, 28, 29, 30
22
-------
Sample handling is much the same no matter which type sampler
is used. In any case, the invertebrates will be sorted, counted
and identified, and weighed. See references (24, 29, 30) for
sample handling information.
As in the case of fish, prediction of effect requires reliable
temperature tolerance data for the forms present and this is not
always available. Predictions may sometimes be made on the basis
of data for related forms but for species of special importance
bioassay data may have to be acquired at the site.
Plankton
Planktonic animals. Occurrence of those organisms normally
known as zooplankton is so variable that the information gained
will generally not be worth the expense in siting surveys. But,
many of the commercially valuable invertebrates, such as clams and
oysters, have planktonic stages which must be included in the
survey. The presence of these forms can be inferred from the
benthic samples of adults which have planktonic eggs or larvae,
but actual distribution will have to be determined by sampling
with plankton tow nets or other devices which filter a measured
amount of water (18, 24).
Sampling should be conducted during the seasons when such
planktonic forms are present in the system. In some cases, this
may be only during a few months of the year. Samples should be
taken on the grid system so the same area can be sampled each time,
23
-------
but additional samples may be chosen to include areas of concen-
tration which the grid does not cover adequately.
Local currents and concentrations of planktonic forms should
be taken into special consideration for design of intake systems
to avoid harmful effects. If such devices are built, they must
be monitored in the Type III survey and the effluent water from
the condensers must be sampled to determine if the condenser
temperatures are lethal to those organisms going through the intakes.
If these temperatures are lethal, one can estimate the numbers of
planktonic organisms killed by determination of flow rate of
organisms through the condensers.
In sampling for the lethal effects of condenser passage, it
is extremely important that the organisms be held long enough after
passage to detect the effects. Organisms may appear well immed-
iately after passage, but be dead a day later as a result of the
temperature shock or other injuries. Exact timing will depend
upon the organisms and on preliminary tests but it should be fixed
after the study starts if results are to be comparable to detect
possible changes in mortality due to changes in season, plant
operation, etc.
Phytoplankton. In large open bodies of water, the phytoplankton
are so variable in number and distribution that sampling usually
does not produce data worth the time and effort expended. But, the
planktonic component of the algae is of interest in enclosed bodies
-------
of water such as small lakes or estuaries where the entire life
cycle of the organisms would be affected by the heated water. This
portion may be sampled by standard plankton techniques using a net
which strains a measured amount of water. Phytoplankton are then
concentrated and estimates are made of percent blue-green algae,
green algae and diatoms, ash-free dry weight and chlorophyll.
This sampling should be conducted at the buoyed stations determined
by the grid system at the same six-week intervals used for inverte-
brate substrates. See Reference 24 for details of sample collection
and handling.
Data may be presented in graphic form and an analysis should
be made as suggested in the statistics section to detect shifts
in algal types. Special note should be made of potentially trouble-
some forms.
Periphyton
Periphyton should be sampled with periphyton substrates,
such as the plexiglass plates used by Grzenda and Brehmer (31)
or the glass slides used by Patrick, et al. (32), and recommended
in FWPCA Methods for the Collection and Analysis of Freshwater
Plankton, Periphyton, and Macroinvertebrate Samples (24). These
artifical substrates are held in racks and submerged at a pre-
determined depth (24) at the buoyed sampling spots determined by
the grid system used. Slides should be exposed for two- to four-
week intervals (24) and samples should be taken at six-week intervals
to coincide with macroinvertebrate substrate sampling.
25
-------
Samples should be analyzed for percent diatoms, green and
blue-green algae, chlorophyll and ash-free dry weight (24).
Aquatic Macrophytes
In determining suitability of the site for heat discharge or
in monitoring effects, the main points to consider are: (1) dis-
tribution and abundance of present plant growth and (2) suitability
of bottom type for extension of these plant beds should growing
season be extended or conditions in some other way be made favorable
for plant growth.
A preliminary requirement is that existing plant beds be
mapped to show areas covered by each species. Such mapping is
greatly facilitated by use of regular and infrared aerial photo-
graphy accompanied by ground check to provide species identification
(33). In this manner, the entire area likely to be affected by
heated effluent can be mapped in a relatively short time without
expenditure of a great deal of time or money. Special consideration
should be given to water depth in the area. Distribution of most
of the attached aquatic plants is limited by depth so there will
be little danger of excessive growth of rooted aquatics in deep,
steep-sided reservoirs or rivers. The critical regions are in
lakes or rivers with extensive shallow littoral areas with bottoms
suitable for attachment of aquatic plants. Morphology of the
bottom should be determined as part of the general overall survey.
26
-------
The aquatic plant survey should be made at least three times
during the year if the potential for nuisance exists. It is of
primary importance that the main survey be made during the middle
part of the summer when aquatic plant growth is at its maximum.
Comparison of pre- and post-operational data collected during
spring and fall surveys will show whether or not plant growth or
distribution is affected by lengthening of the growing season as
a result of addition of the hot water.
Aquatic plant data can be presented as units of area covered
or as percent of total area covered by each species. If percent
is used, replication is not necessary and the data can be presented
graphically or analyzed as suggested in the section on statistics.
Chemical and Physical Parameters
The following chemical and physical parameters should be
included in a minimum sampling program:
1. Dissolved oxygen determinations should be made each
season during periods of maximum temperature and lowest river
flow, or at any other time of year when oxygen concentration may
get low enough to be a potential problem. This must be done during
the siting surveys and continued throughout the monitoring program
until it is demonstrated that no problems have been created.
2. Surface to bottom temperature profiles should be determined
at each grid station at six-week intervals when the benthic substrates
27
-------
are collected and temperatures at the intake and outfall should
be continuously monitored.
3. Residual Chlorine. If a discharge is allowed, this
should be monitored upstream from the discharge and in the discharge
area if chlorine is used as an algacide. Monitoring should be
planned to coincide with the plant chlorination schedule so slugs
of chlorine will be detected.
4. Dissolved copper, nickel and zinc. If a discharge is
allowed, these metals should be monitored upstream from and within
the discharge area until it is determined that there is no toxic
metal leaching from any part of the power plant system.
5. Nitrogen series (dissolved, nitrite, nitrate, ammonia,
Kjeldahl). These determinations must be made as part of the
initial survey at the buoyed sample stations at six-week intervals
to determine potential for increases in possible nuisance algae.
6. Phosphorus (total and ortho). Conduct the same as
nitrogen.
7. pH, hardness and alkalinity. Conduct the same as
nitrogen and phosphorus.
8. Pesticides should be monitored before and after installa-
tion if pesticide use is prevalent enough in the region to mask
the effect of temperature.
9. Salinity should be monitored in the marine environment
on the same schedule as nitrogen and phosphorus if the discharge
28
-------
and intake are arranged so as to cause drastic changes in currents
or in salinity of estuaries.
Samples should be taken one foot below the surface, at mid-
depth and one foot above the bottom at each of the sample sites
designated. (Temperature profile is taken at one-foot intervals
at the same site.) Analysis should be made in accordance to FWPCA
Methods for Chemical Analysis of Water and Hastes (34).
29
-------
DESIGNING THE STATISTICAL MODEL
This section outlines a few statistical techniques to be used
for evaluation of data collected by the recommended sampling
methods. More detail on specific tests is included in the Appendix.
Obviously, the materials and procedures presented herein do not
cover all possible situations. Before an investigation is begun,
the project officer and key staff should consult with a statistician
to design the survey and subsequent analysis to account for factors
peculiar to the environment being studied.
The statistical model should provide for the control of
factors in the system which affect the response factor. Affect-
ing factors are of two types: (1) those which can be fixed,
such as transect location or season; and (2) those which, although
not fixed, can be measured, such as velocity, dissolved oxygen,
nutrients, etc. This second set of factors is examined to
determine that they are not lethal or do not override the effect
of temperature. It should also be determined that levels for
synergistic toxicity are not reached. Such factors must be
investigated along with temperature to impart confidence to statements
of causality. The relative importance of affecting factors to
the organisms in a system should be determined by aquatic ecologists
who are familiar with the system and who are responsible for the
evaluation of that system. Response factors to be specifically
mentioned in this paper include those parameters measured for
-------
fish, macroinvertebrates, plankton, periphyton, and aquatic
macrophytes. Procedures are presented for reducing multiple
response factors to unique numerical values which can be used in
the appropriate statistical evaluations.
Fish
The count of fish per species is a multiple response
measurement. It is desirable to reduce these values to some
meaningful composite score, and since certain species are of
special value, the score should take this into account. It is
assumed that for a given geographic location it is possible to
rank the fish species according to their contribution to the
economic or esthetic well-being of the area. Those species having
the greatest value should be assigned a positive numerical value
and those with the least value, a lesser numerical value which
may be negative if the species is undesirable. Ties can occur.
For example, if a sample yields seven species of fish, three of
which are desirable and three not so desirable and one of which
is neutral, the simplest scale representing the relative magnitude
of importance of these seven species might be 3, 2, 1, 0, -1, -2,
-3. The final score used in the analysis would be the sum of the
products of the number of fish per species times the scalar value.
Under such a plan, locations having a high population of desirable
species will get a high score, an evenly mixed location should get
31
-------
a score centering around zero, and a location having a high
concentration of undesirable types will get a high negative score.
A change in population composition will be readily apparent in the
scores.
Other scale values can be chosen. For example, if extensive
data are available on the economic value of given fish species,
then an economic valued scale can be used. The weight per species
and average individual weight in each species are also multiple
responses which can be reduced to a composite score using the
same scale of importance as given above. Whatever scale is
chosen must be used for the duration of the investigation.
Count data are often non-normal; therefore, a nonparametric
statistical evaluation, such as the factorial arrangement, non-
parametric type given in the Appendix, should be used. A score
arrived at by the scale of importance is derived from a count so
analysis is by the same nonparametric methods. Two methods of
handling count data are given in the Appendix.
The statistical evaluation used is also dependent upon the
questions to be answered. If the primary question is whether there
has been a gross change over time, then an analysis is made of
the composite scores to detect changes in community composition.
If the interest is in determining shifts in a given species over
time, then an analysis is made for each species. In this
case, the average weight/species should be used, and a parametric
-------
analysis, as identified in the Appendix, is the proper evaluation
tool. In some studies, both methods may be used.
Macroinvertebrates
Standard procedures for presenting benthic data include
bar charts or graphs which present the data, but are not always
sufficient to show the combined effects of various chemical consti-
tuents and hydrographic conditions upon the structure of the aqua-
tic community. In an effort to alleviate this problem, various
diversity indices have been developed. Those showing the greatest
promise are given by Margalef (35) and are derived from information
theory.
Diversity indices are a measure of the species distribution
of the individuals of a community. The most diverse community
possible would contain a number of species with each having the
same number of individuals. The least diverse community would
have one species including all individuals.
The index for redundancy (R), often used with the diversity
index, is a measure of preponderance of individuals in a few
species. Redundancy is greatest with a large number of individuals
in one species and least with one individual per species, so R
decreases as diversity increases.
The use of diversity and redundancy indices is based upon
the theory in ecology that more stable communities are more diverse,
33
-------
i.e., they contain more species and the number of organisms is
relatively well distributed among the species. A stable
community would have a higher diversity and a lower redundancy
value than a less stable community. If the diversity decreases
and redundancy increases for the community below a thermal discharge,
one would assume that some species are being eliminated and that
the system is becoming less stable (36).
Diversity (IT) and redundancy (R) are computed by means of
the following equations derived from Wilhm (37).
S
H = 1 (log2N! -I Iog2n.!)
and _
1og2
log2(N-S+l)! - S Iog2 (^)!
where
n.j = Number of organisms in the ith species
S
N = I n.j = total number of organisms
S = Number of species
Base two logarithm is used.
This gives two values to describe a sample. These may be used
as they are or they can be reduced to one response value per
-------
sample which is comparable to other samples from the study area.
This is done by use of the following modified equation from
Kendall (38) for dispersion:
VR %
where
L = Location (sampling point) 1, 2,...n
KR = Rank of R at location L
Km3V = Highest rank of R
llldA
Kjq- = Rank of H" at location L
VR and VTT = Variance of R and R, respectively.
A dispersion (D. ) value is computed for each sample at
each time of sampling and is a relative measure of the difference
between the combination of diversity and redundancy in that sample
and in a theoretical sample where H" = 0 and R = 1, i.e., where
there are no organisms present at all. A sample with a dispersion
value of 2.0 would be much closer to the worst case than would a
station with a value of 23.0. In comparing data from several
locations, if all stations have high values except for those in
the plume, then something is decreasing the diversity and is
probably detrimental to the resource. In a before-and-after case,
we can look for changes in the dispersion values to point out
35
-------
changes in community. The reason(s) for change will be determined
from an examination of possible affecting factors.
As an example, in a body of water where physical conditions
(bottom type, current, etc.) are fairly uniform, the dispersion
values for five stations will fall within a narrow range and
"clump" at one point between 0 and infinity.
0
43125
0 - oo = Dispersion (D.) values (a positive number computed by
the equation for DL)
1 - 5 = Station numbers (L)
If stations are laid out downstream from 1-5 and some form
of toxicant (possibly heat) is introduced between stations 1 and
2, the distribution of DL values would change and could look
like this:
This would indicate a drastic change at station 2, but a gradual
recovery downstream to station 5, which has nearly the same D,
value as station 1.
-------
To obtain the DL values, first compute diversity (H) and
redundancy (R) for each station or sample location. The values
of H are then ranked from lowest to highest. The corresponding R
values for each station are ranked from highest to lowest since R
decreases as H" increases. Since DL is a measure of difference
between each station and a control value, a control point is neces-
sary to ensure a common starting point for subsequent comparisons.
The control value for the H values is arbitrarily given a rank of
1, so H values for sample stations rank from 2 on. The control value
for R is one greater than the number of stations being compared.
The variances (VR and Vj^-) for each station, but not the control,
are computed from the ranked values by using the equation:
Var (R or H) = [(n3 - n) -
- t)]
where
n = Number of observations
t = Number of tied ranks.
For example, if the following values are obtained,
Computec
Values
Ranks
i R
H"
R
H"
Locations or
h 4
.61 .49
.98 1.34
5 4
2 3
L3
.37
2.45
3
4
station
L4
.22
3.25
2
5.5
numbers
L5
.20
3.25
1
5.5
Rank of
Control Value
6
1
37
-------
then for L-J :
VR = ygy [(53 - 5) - 0] = 2.00
" 5)
Comparing the control value to location 1 by means of the
equation for DL gives:
and comparing the control value to location 5 gives:
D _ (1 - 6)2 + (5.5 - I)2 = 25 20,25 =
U5 ~ 2 1.81 2 HT8T
The other values are computed the same way. These dispersion
values can then be plotted on a line from 0-24 for a graphic display
of the differences between stations.
It may be superfluous to use any statistical technique to
check for population shifts once the dispersion values are computed.
However, a nonparametric method of either type discussed in the
Appendix is recommended if a test is desired.
In the marine environment, or in any situation where it is
possible to establish a relative value for various benthic species,
38
-------
a scale of importance, such as the one recommended for fish
data, may be used for an indication of whether species shifts
are beneficial or detrimental to the system or to the economics
of a particular region.
Plankton, Periphyton and Aquatic Macrophytes
Data for these three groups will be presented, in part,
at least, as percentages. Plankton and periphyton may be weighed
for a measure of biomass, but there will also be an estimate of
percent total composition of algal types (green, blue-green, etc.).
Aquatic macrophyte data will generally be in terms of percent
plant coverage of an area.
For replicated data, the percentage values should be normalized
using the arc sin transformation so that standard statistical tests
can be used. This is accomplished by referring the percentage
value to the arc sin table in Steele and Torrie (39) or any other
statistics text. If done by computer, the equation: angle in
degrees = arc sin /percent can be used for the transformation. Arc
sin values may be expreseed as degrees or radians since one is a
multiple of the other.
For replicated data, the experimental error used in the tests
for significance should be calculated from the data. For unreplicated
data, the following method can be used to determine an experimental
error for the tests. This method gives the minimum experimental
39
-------
error which is usually lower than that calculated from replicated
data.
If the transformed value is reported in degrees, the
theoretical experimental error is 821/n and if in radians the
error is 0.25/n. The divisor, n, is the common denominator of
all the fractions used to compute the percentages. For example:
if 200 microscope grid squares are viewed on a periphyton-covered
glass plate and 100 squares are covered with green algae, the
percentage is calculated as 100/200 = 50 percent. The denominator
is 200 and the theoretical error term in degrees is 821/200. In
radians, it is .25/200. Although use of the error term is
explained in the Appendix in the section on Analysis qf Variance
(Parametric), it is included here to point out the importance of
using equal areas or volumes for each sample from which a percen-
tage is obtained. Unequal volumes or areas change the denominator
so the theoretical error term used in the statistical analyses
is not constant.
-------
STATISTICAL ANALYSIS
Experimental design includes consideration of sampling plan,
sampling methods and the statistical techniques to be used for
analysis of the resulting data. This section is devoted to
analysis techniques; the first two components have been discussed
in previous sections.
Investigators should design the ideal experiment, then
examine it in light of the deficiencies inherent in the actual
field situation. If the field survey plan violates the basic
assumptions of controlled experimental design, then revisions are
required. The following techniques do not violate the assumptions
of randomization of experimental units (i.e., fish, benthic
organisms, etc.) in the system nor the assumption of replication
necessary to obtain a valid estimate of experimental error.
Random sampling is not to be confused with haphazard sampling.
The idea is to obtain a random sample, representative of the
experimental unit(s) of interest at a preselected point. This
is commonly referred to as stratified random sampling (40).
While the sampling program is an essential aspect of the over-
all study, the final result is dependent upon the data analysis.
There are a number of statistical techniques which may be used in
handling biological data and proper use of one or more of the
suitable methods is needed before interpretation of field study
data can be accepted.
-------
This paper presents only two of a large number of basic
statistical techniques which could be used. However, the two
methods selected should enable the user to deal with the wide
variety of data associated with the biological responses to
thermal discharges. For other methods, consult a statistician
and any of the statistical references given.
The basic technique to be used is the factorially arranged
analysis of variance (or covariance). Seasonal effects exist,
but it is assumed that there is no real seasonal interaction with
treatment so any apparent interaction between seasons and treat-
ments is a measure of the experimental error. Factors which can
be considered as entering at random, but physically controlled,
are transect location, depth of sample collection, and stream
width location. These factors define, the statistical strata to be
sampled. Those considered to be entering at random, but not
physically controlled, include velocity, flow rate, temperature,
D. 0., and the other chemical parameters. These constitute the
set of factors upon which the response factor is regressed. The
response factor (number of fish, diversity, etc.) is measured to
detect yearly changes between pre- and post-construction periods
or between sampling locations.
If only the physically controlled factors are considered in
the statistical analysis, then a factorial experiment, randomized
complete block analysis of variance, is used. General methods
-------
for analyzing parametric as well as nonparametric data are given
and identified in the Appendix. For more thorough discussions of
the parametric analysis, consult the texts by Steele and Torrie (39),
Cochran and Cox (40), or Winer (41). The nonparametric factorial
method is discussed more thoroughly in Wilson (42).
If the physically uncontrollable but measurable effects are
to be used in the evaluation, then a factorial experiment:
randomized complete block analysis of covariance can be used. The
parametric analysis is discussed in Winer (41) and in Steele and
Torrie (39). The nonparametric analysis of covariance is discussed
in the paper by Quade (43). Summaries of these analyses are given
in the Appendix along with discussions of their use, extensions,
and interpretations.
A factorially arranged experiment allows the examination of
synergistic effects between the factors included in the model as
well as examination of each factor by itself. Failure to include
the interaction inflates the error term and thus masks significant
effects that exist.
Isolating and examining the effects of covariates (physically
uncontrollable but measurable factors) is aided by performing a
series of step-wise analyses of covariance and examining the net
effect upon the adjusted and unadjusted F values obtained. There
are computer programs available to do these calculations.
-------
The interpretation of results obtained from either of the
above-mentioned analyses is dependent upon the number of possible
explanations presented by data collected from the system being
studied and upon the knowledge and experience of the persons
conducting the study. Alone, statistical analysis of a limited
set of biologic data can lead to spurious "conclusions"; on the
other hand, a purported biological effect cannot be confidently
accepted unless it can stand the test of statistical significance.
In all cases, statistics and reason must be combined to arrive at
the best possible conclusions.
-------
APPENDIX
The appendix describes the factorial arrangement for analysis
of variance (AOV) and analysis of covariance (AOCV) techniques
for both parametric and nonparametric data. In some cases,
numerical examples are given to clarify the procedures. The aim
is to show the reader how to set up the analyses, arrange the
data, make the necessary computations, and understand the meaning
of the results. In an actual situation, the amount of data would
likely be so large that a computer would be necessary. Still,
the basic techniques illustrated by the following examples would
apply.
The Factorial Arrangement: Analysis of Variance (Parametric)
A factorial experiment allows the simultaneous examination
of the effect of a number of different factors. The analysis of
variance is based upon the fact that any system has a total
variability which can be partitioned or divided into portions
attributable to different factors recognized in the model. An
additional, unexplainable part of the variability is called random
variation, or the error term. The method for dividing the total
sum of squares used to obtain these variances in a system can best
be explained by use of examples so a brief general explanation is
given and then a simple numerical example is presented so the
statistical test can be carried to a conclusion.
-------
Consider the river situation. In this case, the following
factors and levels of factors are included:
Time at 2 levels: Pre- and post-construction.
Transects at 4 levels: Above outfall, immediately
below outfall, 0.5 AT below outfall, and 0.1 AT
below outfall.
Width at 3 levels for fish (at bank, midstream,
and left bank) and at 4 levels for benthos,
chemical data, etc.
Depth at 3 levels for chemical analyses (not
considered in biological analyses).
Seasons form four blocks.
A general diagram of the data arrangement for one block is
given in Table A-l. A portion of this table has been filled
in for the spring block of data collected. Any response, R,
can be uniquely identified by the proper use of subscripts to
denote season, transect, width, time, and replicate.
Blocks (seasons) = B, where h = 1, 2, 3, 4 (Spring,
Summer, Fall, and Winter).
Transect = T, . where i = 1, 2, 3, 4, (above, below,
1 0.5 AT, and 0.1 AT).
Width = W.. where j = 1, 2, 3, for fish (left, middle,
"J right) or 1, 2, 3, 4, for benthos (0.1W,
0.3W, 0.7W, 0.9W).
Construction Time = Chk where k = 1, 2, (before, after).
Response = Rhij-kl where 1=1,2 (first, second replicate)
For example: R2 3 2 2 1 is taken in tne summer at transect
0.5 AT in the center section of the stream (for fish) after
construction of the plant and is the first replicate-
-------
TABLE A-l
RAW DATA SHEET
"\ Width
\" (W)
Transect N.
dr\
i
9
0
0
4
CW Sur
W Sum
Construction
Time (C)
1
9
L
1
1
9
T
1
9
i.
-\
\
9
L.
n
<
1
Rl ,1,1 ,1,1
Rl ,1,1 ,1,2
Rep Sum
Rl, 1,1, 2,1
Rl, 1,1 ,2,2
Rep Sum
C11W11
C12W11
wn
Spring (Block
2
Rl, 1,2, 1,1
Rl ,1,2, 1,2
Rep Sum
Cnwi2
C12W12
w12
1)
3
Cnwi3
C12W13
W13
4
C11W14
C12W14
W14
sum
Tn
T,2
1 L.
T13
T11
cn
C12
Blk. tot
(BT)
^
CO
A
t/)
x/
C-)
0
CO
fy
UJ
j__
^
3
A
|
et
U_
n
o:
LU
s:
i
oo
rv
Q
u_
1
5
UJ
r\
UJ
o;
Grand Sum
A
£T - V T
GT1 ^ 'hi
h-1
II 1
rjr
b!2
RT
bT3
GT.
^
4 p
GCn=E Chl ^tot
1 h-l hi
GC2 IGT = yew =
GWr GW2, GW3, GW4 ^GC
-------
TABLE A-2
SAMPLE PROBLEM RAW DATA SHEET (WEIGHTS PER UNIT AREA SAMPLED)
^Nv Width
Transect \^
1
(above)
2
(below)
3
(0.5 AT)
4
(0.1 AT)
Block 1 - Spring
Const. Time
1 (before)_
2 (after)
1
2
1
2
1
2
CW Sum
W Sum
1
(0.1 W)
6,10
16
9,8
17
10,8
18
0,1
1
7,9
16
4,3
7
9,9
18
11 ,10
21
68
46
114
2
(0.3 W)
7,10
17
7,10
17
7,9
16
2,1
3
9,9
18
5,4
9
7,9
16
10,9
19
67
48
115
3
(0.7 W)
5,11
16
8,7
15
6,11
17
8,7
15
9,8
17
8,8
16
8,7
15
9,9
18
65
64
129
4
(0.9 W)
8,9
17
9,7
16
8,8
16
7,8
15
8,7
15
9,9
18
8,8
16
8,9
17
64
66
130
sum
131
101
116
140
264
224
488
Blgck 2 - Summer
Const. Time
1
2
1
2
1
2
1
2
1
12,13
25
14,12
26
12,12
24
0,1
1
10,9
19
4,5
9
11 ,13
24
9,8
17
92
53
145
2
11,13
24
11 ,11
22
11 ,14
25
3,2
5
10,11
21
7,6
13
13,10
23
10,9
19
93
59
152
3
10,13
23
14,9
23
10,13
23
12,12
24
11 ,14
25
10,13
23
9,12
21
12,12
24
92
94
186
4
12,12
24
13,12
25
11 ,14
25
13,10
23
11 ,11
22
12,12
24
13,13
26
12,13
25
97
97
194
sum
192
150
156
179
374
303
677
Grand Sums
323
251
272
319
638
527
259
267
315 1165
324
-------
Table A-2 is the same as A-l except that the squares are
filled with whole numbers which have been chosen to represent
average weights of benthic invertebrates of a species. In a
real situation, the weights could be in grams or any other
measure of weight. The numbers in the example were chosen to
avoid decimals and make the example simple. Note also that since
these are benthic data, there are four widths and not three as
would be the case if the table were set up for fish data. The
example is carried out for only two seasons, again for simplifica-
tion. The method of obtaining the sums should be evident from
the table. If it is not, any of the referenced statistics books
can be consulted.
The numbers in the example were chosen specifically to
illustrate possible biological changes after operation of a power
plant in a hypothetical situation illustrated by Figure 3.
In order to compute the sums of squares for the factors
under consideration, Interaction Table A-3 should be prepared.
Note that only single subscripts are used in Table A-3, since
the block (i.e., season) subscript, h, has been removed by summing
across blocks. The entries in this table are obtained by first
summing across replicates and blocks in Table A-l or A-2 to obtain
the T., Ck, and W. entries where T^ = £R for transect i; Ck = £R
for time; W. = £R for widths. For example, T^-jW^is the sum
of the spring, summer, fall and winter values for both replicates
-------
EFFLUENT
FIGURE 3. SAMPLING GRID AND HEATED AREA FOR
EXAMPLE PROBLEM.
-------
and all blocks for transect 1, time 1 (before construction), and
width 1. Adding across in Table A-3 gives the sums T.C, and
I K
down rows within transects gives intermediate sums T.W.. Adding
' J
down rows within time frames gives the sums C.W.. Adding across
K J
columns of intermediate row sums (T.W. values) gives the GT. values,
I J I
Adding C-|W, + C2W. gives the GW, values and adding C^ + CkW2
+ CkW3 + CkW4 gives the GCk values. The sums of the GT.., GW., and
GCk values should all be equal to their respective grand total
values for the corresponding factors in Table A-l. If they are
not, an error in addition has occurred.
Table A-4 illustrates numerically the development of an
interaction table (i.e., Table A-3). Data from Table A-2 were
used to construct Table A-4.
Sums of squares are computed in the following manner:
Total sum of squares = I R2h- .kl-GS=TTSS, where GS=(GTot)2/N
hijkl
(BT)h
Block sum of squares = £ -^ - - GS = BSS
(GT)?
Transect sum of squares = ) -TJ -- GS - TSS
i NT
(GW.)2
Width sum of squares = I -n-^- -- GS = WSS
M
(GCk)2
Construction sum of squares = T -TJ -- GS = CSS
k hC
51
-------
TABLE A-3
INTERACTION TABLE
Transect
Tl
TOTAL
T2
TOTAL
T3
TOTAL
T4
TOTAL
TOTAL
TOTAL
TOTAL
Time
Cl
C2
Cl
C2
Cl
C2
Cl
C2
Wl
T1C1W1
T1C2W1
T1W1
T2C1W1
T2C2W1
T2W1
T3C1W1
T3C2W1
T3W1
Wl
T4C2W1
T4W1
C1W1
C2W1
GW]
Width
W2
W2
T1C2W2
T1W2
T2C1W2
1 o'-'O"^
T2W2
T3C1W2
T3C2W2
T W
'3W2
T4C1W2
T4C2W2
T4W2
C1W2
C2W2
GW2
W3
T1C1W3
T1C2W3
T1W3
T2C1W3
T2C2W3
T2W3
T3C1W3
T3C2«3
T3W3
T4C1W3
T r u
'4U2W3
T4W3
C1W3
C2W3
GW3
W4
T1C1W4
T1C2W4
T1W4
T2C1W4
T2C2W4
T2W4
T3C1W4
T3C2W4
T3W4
T4C1W4
T r w
V2W4
T4W4
C1W4
c2w4
GW4
Total
T1C1
T1C2
GT1
T2C1
T2C2
GT2
T3C1
T3C2
GT3
T4C1
T4C2
GT4
GC1
GC2
G Tot.
IX)
-------
TABLE A-4
SAMPLE PROBLEM INTERACTION TABLE (WEIGHTS PER UNIT AREA SAMPLED)
Transect Time 0.1 W 0.3 W 0.7 W 0.9 W Total
Ti
TOTAL
T2
TOTAL
T3
TOTAL
T4
TOTAL
TOTAL
TOTAL
TOTAL
Cl
C2
Cl
C2
Cl
C2
Cl
C2
C^
c2w
GW
16+25=41
43
84
18+24=42
2
44
35
16
51
42
38
80
160
99
259
41
39
80
41
8
49
39
22
61
39
38
77
160
107
267
39
38
77
40
39
79
42
39
81
36
42
78
157
158
315
41
41
82
41
38
79
37
42
79
42
42
84
161
163
324
162
161
323
164
87
251
153
119
272
159
160
319
638
527
1165
oo
-------
The interaction sums of squares are computed from Table A-3
or A-4. (Note: The main effects sums of squares could also have
been calculated from the marginal totals in Table A-2.)
TW sum of squares = £ (T^W.)2
U-a TSS - WSS - GS = TWSS
TC sum of squares = I (T-C, )2
Ik
NTC
- TSS - CSS - GS = TCSS
WC sum of squares = T (C..W.)
K J
jk
2
Ncw
- CSS - WSS - GS = WCSS
TWC sum of squares = T (T.W.C. )2
.. * J k
TSS - CSS - WSS - TWSS
nTWC
- TCSS - WCSS - GS = TWCSS
Experimental error sum of squares = I (Rep Sum)2
hijk
NR
GS - BSS - TSS - WSS - CSS - TWSS - TCSS - WCSS - TWCSS = EESS
-------
(Rep Sum)
2
Sampling error of squares = I Rh''kl ~ - N - = SESS
hijkl R
Experimental error is the sampling error plus the interaction
between the replicate factor and blocking factor, whereas sampling
error is a measure of the lack of agreement among observations
from the same treatment cell.
The divisors in the above error equations (NR) can be obtained
by counting the number of observations which were added to get each
of the values in the numerator. This can become laborious; there-
fore, methods for calculating them have been developed. For the
numerical example given, let the number of blocks used be called
"b" (2), the number of replications be "r" (2), the number of
transects be "t" (4), the number of widths be "w" (4), and the
number of time periods be "c" (2). Then the total number of
observations, N, is equal to brtwc or 2x2x4x4x2=1 28;
NB (the number of observations per block) is rtwc or 2 x 4 x 4 x 2 = 64;
NT (number of observations per transect) is brwc or 2 x 2 x 4 x 2 = 32;
NW (number of observations per width) is brtc or 2x2x4x2= 32;
NC (number of observations per time period) is brwt = 64; NTW (number
of observations at each transect and width) is brc = 8; NTC = brw = 16;
NCW = brt = 16; and NTWC = br = 4; NR = 2.
Using the hypothetical data presented in Tables A-2 and A-4.
the sums of squares are computed by the equations just given:
55
-------
TTSS = 11,859 - - = 11,859 - 10,603 = 1,256
BSS
= 488L±J^2- 10,603 = 279
TSS - 323' + 251' * 272* + 319'. ,0;603 . ,,8
+ ?672 + 3T52
WSS = 32 - 10»603 = 102
CSS = 5272
TWSS = 87^521 - 118 - 102 - 10,603 = 117
TCSS =
WCSS = 17453 - 97 - 102 - 10,603 = 108
TWCSS =
f^ - 118 - 97 - 102 - 117 - 125 - 108 - 10,603 = 101
-------
EESS = 23^459 - 10,603 - 279 - 118 - 102 - 97 - 117 - 125
- 108 - 101 = 80
SESS = 11,859 - 11,730 = 129
As a check on the computations, the TTSS computed initially
can be compared to:
TTSS = BSS + TSS + WSS + CSS + TWSS + TCSS + WCSS + TWCSS + EESS
+ SESS = 1,256
The results from the above calculations are presented in an
analysis of variance table, Table A-6, which follows the form of
a general AOV table (A-5). The F values obtained are compared
to tabular F values from F tables with the given numerator and
denominator degrees of freedom at a selected level of significance.
It should be noted that if this analysis or a similar one containing
block effects is used, it is not strictly necessary to have
replication. If replication is not done, then EEMS is used in all
tests and no measure of true experimental error is obtained.
However, if comparisons between species at given stations or
transects are contemplated, replication is required to give a
valid estimate of the error rate in the system.
57
-------
TABLE A-5
ANALYSIS OF VARIANCE
Source
Total
Blocks
Transect
Width
Construction
Time
TW
TC
we
TWC
Experimental
Error
Sampling
Error
Degrees of
Freedom
rbtwc - 1
b - 1
t - 1
w - 1
c - 1
(t-D(w-l)
(t-D(c-l)
(w-D(c-l)
(c-1) W"
(b-D(twc-l)
btwc(r-l)
Sums of
Squares
TTSS
BSS
TSS
WSS
CSS
TWSS
TCSS
WCSS
TWCSS
EESS
SESS
Mean Squares
TMS=TSSvdf
WMS=WSSvdf
CMS=CSS,df
TWMS=TWSS-df
TCMS=TCSS-df
WCMS=WCSSvdf
TWCMS=TWCSSvdf
EEMS=EESSvdf
SEMS=SESS-df
F
TMS-EMS*
WMS^EMS*
CMSvEMS*
TWMSvEMS*
TCMS-EMS*
WCMSvEMS*
TWCMSvEMS*
EEMS/SEMS
*If EEMS/SEMS is significant, then EMS is not significantly different
from EEMS; if it is not significant, then EMS is not significantly
different from SEMS. For more detailed discussion, see Steele and
Torrie (39).
-------
TABLE A-6
SAMPLE PROBLEM ANALYSIS OF VARIANCE
Source
Total
Blocks
Transect
Width
Const. Time
TW
TC
we
TWC
Exp. Error
Sample Error
*Tabular F
**^n t->i-Q tha
DF
128
2 -
4 -
4 -
2 -
3 x
3 x
O x
3 x
(2-1
- i
i =
i =
i =
i =
3 =
1 =
1 =
3 x
= 127
1
3
3
1
9
3
3
1 = 9
) (4x4x2-1 )=31
(2x4x4x2) (2-1 )=64
ratios were
F = "i c
selected
not cinni
ss
1256
279
118
102
97
117
125
108
101
80
129
at the
f i rant
MS
--
39
34
97
13
42
36
11
2.
2.
5% 1
(i P
Computed
F
--
19
17
48
6
21
18
5
6 1
0
evel of
. . 1 .3 <
.5
.0
.5
.5
.0
.0
.5
.3
signi
- 1 fi1)
Tabular
F*
--
2.
2.
4.
2.
2.
2.
2.
1.
7
7
0
0
7
7
0
6**
ficance.
. thpn FMS
is not statistically significantly different from SEMS. That is,
there is no block by response interaction.
-------
For the analysis of variance, the F test is one-tailed on
the right. Therefore, when the computed F ratio exceeds the
tabular value, the variation due to the effect being tested is
statistically significant. If the computed F ratio is less than
the tabular F, the variation is not statistically significant.
Thus, it is seen from Table A-6 that the variations in
biomass of benthic invertebrates due to all main effects and
all interaction effects are statistically significant. In other
words, the variations are greater than would be expected from
experimental or sampling error.
In terms of the hypothetical heated water discharge, this
means that there was a statistically significant change in biota
after introduction of the heated discharge. The following
significant effects can be seen from Table A-6:
1. There was a difference in biota among transects. This
is reasonable since three transects are within the heated zone
and one is outside (See Figure 3).
2. There was a difference in biota at different stations
based on percent of river width. Only one side of the river is
covered by the heated water.
3. There was a difference attributable to construction
time. Invertebrate biomass decreased significantly after plant
operation began.
60
-------
4. There was an interaction between transect and location by
width. Each sample station is located at the intersection of a
transect line and a line of percent width. Samples at intersections
outside the heated area were different from those inside.
5. There was an interaction between transect and construction
time and between location by width and construction time. There
was also an interaction between all three. Transect and location
by width determine location of a sample and construction time
determines whether the sample point is covered by heated waters.
The statistical analysis is not the whole story, but this
information can be combined with graphic presentations of the
data to form a composite picture. The statistical tests are
important to show whether apparent changes are significant.
In the test on Table A-6, the F value was arbitrarily chosen
at the 5 percent level of significance to illustrate the point of
the sample problem. This is a commonly accepted level, but in an
actual field study where results are not significant at the 5 percent
level the F may be varied to determine that level at which signifi-
cance is found. Then, the F is chosen such that the tabular F value
is that value in the table which is under, but closest to the
computed F ratio. This will show at which level the variation due
to a specific factor is significant. Of course, the results may
be accepted with much more confidence at the 1-5 percent level than
at the 30-40 percent level.
6l
-------
Sample Size Estimation and Confidence Intervals
The first year's data can be used to determine if replication
is adequate for making decisions. Following are two methods for
arriving at required sample size (39, 40).
These calculations depend upon:
1. An estimate of a2, the true experimental error,
2. The size of the difference to be detected, 6 or d,
3. The assurance with which it is desired to detect
the difference (Type II error or 3),
4. The level of significance to be used in the actual
experiment (Type I error or a), and
5. Whether a one-tailed or two-tailed test is required.
The following equation is an approximation since it assumes
2(t + t,)2 S:
i2 =a2 r > ^-r,
where
r = Number of replicates
S2 = Best available estimate of a2
6 = Difference of practical importance to be detected
tQ = Student's t associated with Type I error (a,
rejection of a true hypothesis)
t-j = Student's t associated with Type II error (g,
acceptance of a false hypothesis) or the probability
of detecting 6 if it exists.
62.
-------
Use the table for tQ for error degrees of freedom at the a
level and t-j for error degrees of freedom at 2(1 - $) level.
Assume the statistical model to be used fits Table A-6. There
will be four seasons (blocks), four transects, four width measur-
ing points, six years (three pre- and three post-), and two replicates
Assume that a 30 percent average weight reduction is considered
a change of practical importance. The Type I error is 0.05 = a
and the Type II error is 0.10 = 3.
The error degrees of freedom in the completed experiment is
given by btwc(r - 1) = 4*4*4x6(2 - 1) = 384. The mean weight
from Table A-4 is 9.1, therefore <5 is 0.3 x 9.1 = 2.7 grams. Then,
> 2(1.96 + Q.842)2 (2.0/4) K
r
2.72
The value 2.0/4 is used since a test for reduction about the
mean is desired and four is the smallest number of means to be
compared in the ANOVA. Since a one-tailed test is indicated by
the term "reduction," the t values are taken at a/2 and 3/2. The
value computed is always rounded up.
To obtain sample size sufficient to compute a confidence
interval no larger than a specified size, use the equation
S2 q* (P,n2)
63
-------
where
S2 = Variance of the smallest number of means to
be compared (4)
q = Value from the studentized range table for
Ma
four means (P) and degrees of freedom equal to
the error degrees of freedom from the final
survey (n~)
F = Value of Snedecor's F at the y significance
level for n~ degrees of freedom and n-| (current
sample size) degrees of freedom.
d = Half-length confidence interval - again taken
as 30 percent of the overall mean, or 2.7 grams.
This gives
r = 0.5(3.02) (1.0) .. 1 0
2.72
This indicates that replication is not required; however,
one should refrain from obtaining fewer than two replicate samples,
Methods for setting confidence intervals are to be found in
any of the parametric statistical texts referenced.
The Factorial Arrangement: Analysis of Variance
(Nonparametric)
If the raw data is in the form of counts, or if it is badly
skewed, it is often desirable to analyze it by nonparametric
-------
techniques. The procedure by Wilson (42) has proved to be very
satisfactory for evaluating counting data in water quality studies
In this procedure, the chi-square statistic for a contingency
table is broken into components in much the same way as a total
sum-of-squares is decomposed in analysis of variance computations.
This provides a nonparametric test of the hypothesis concerning
main effects and interaction much like the parametric test using
two-way analysis of variance.
Wilson explains use of the test for two factors with inter-
action to include more than two factors and expands this test in
his paper (42). Siegel (44) gives methods for cases where inter-
actions are not of interest. A simple example of the factorial
test is given in the section to follow; however, the more basic
references will have to be consulted for application to a large-
scale sampling program.
In the sampling program proposed in this paper, there are
five factors: blocks (seasons), replications, transects, widths,
and time (pre- and post-construction). In the actual survey
comprised of six years data, time will be in years and a special
test will have to be made for a strict pre- and post-construction
test. In the sampling problem to be presented, only transect
and time are considered for simplicity. In an actual situation,
the method becomes so awkard that a computer is necessary to
handle the table construction and analyses.
65
-------
For the sample problem, we assume that two sets (series) of
fish samples are taken before plant operation and two after
operation on each of the four transects shown in Figure 3. Since
there are three stations per transect, the total number of samples
is: 2 times (before and after construction) x 2 series per time
x 4 transects x 3 stations per transect = 48. The data represent
total fish counts per transect.
The following step-wise process is required to complete the
analysis. First, compute the median value for the entire set of
data. This divides the data into two sets with nearly equal
numbers of observations where n= is the number of samples equal
a
to or greater than the median and n, is the number of samples
less than the median. For example, if the numbers of fish
caught per station vary from one to 100, the median value could
be 46. If so, half of the samples have 46 or more fish and the
other half have fewer than 46 fish.
This information is set up in a two-way contingency table
such as Table A-7 which considers only transect and construction
time. The cell entries in the table are the numbers of observations
taken before and after construction on each transect. They are
divided into two groups depending upon whether the number of fish
caught is equal to or greater than the median or less than the
median. For example, in transect one, 14 samples before construction
had a number of fish equal to or greater than the median number of
fish per sample. In this case, 14 samples had 46 or more fish.
66
-------
TABLE A-7
CONTINGENCY TABLE FOR NONPARAMETRIC AOV
Sample Size
> MpH i ^ n
a*-j
< Median
J-i
a'-j + b'-j
^vTransect
^\.
Time ^\^^
Before
After
Before
After
1
14
9
23
10
15
25
48
2
16
6
22
8
18
26
48
3
16
8
24
8
16
24
48
4
14
13
27
10
11
21
48
Total
.'1.
60
36
96
bfi.
36
60
96
192
67
-------
Ten of the before samples in transect one had fewer than the median
number of fish.
Certain checks can be made to ensure that counting errors
are not made. The total number of samples in each transect
should equal 48 with 24 samples taken before construction and 24
taken after.
The information in the contingency Table A-7 will be used
to perform three chi-square tests to determine whether:
1. There is significant overall variability
in the data,
2. There is a significant difference between
before and after data,
3. There is a significant difference between
transects.
The total chi-square value for test 1 is computed from the
following equation:
x2 = II
T i j
where
nijna/n
af .j = Represents the number of observations
in the cell in row i and column j which
are equal to or greater than the median
-------
-ji = Represents the number of observations in
the cell in row i and column j which are
less than the median
na = a
/ij
i J
i J
na = nb = n/2
For the numerical computation:
/. . = 14, 9, 16 ... 13
J,, = 10, 15, 8 ... 11
na =96 nb = 96
69
-------
n = 192
24XT9T=12
= 24XT9T=12
= (14-12)2+(16-12)2+(16-12)2+(14-12)2+(9-12)2+(6-12)2+
(8-12)2+(13-12)2
12
-f (10-12)2+(8-12)2+(8-12)2+(10-12)2+(15-12)2+(18-12)2+
12
42 + 42 + 22 + 32 + 62 + 42 + I2
12
42 + 42 + 22 + 32 + 62 + 42 + I2
12
4+16+16+4+9+36+16+1 + 4+16+16+4+9+36+16+1
12
12
_ 102 . 102 _ 204
TO
-------
X2 has (re - 1) degrees of freedom
T
where
r = 2
c = 4.
So, (re - 1) = 8-1 = 7.
At 0.05 level of significance x2 = 14.1
7
Since the calculated x2 of 17.0 is greater than the table x2 of
T T
14.1, we conclude that the total variability among the samples is
significant.
The row chi-square value for test (2) is computed from the
following equation:
x -
R
I
i
~2~
ni.na/n ni-Vn
where
For the numerical computation:
a/i. = 36, 60
b/i. = 60, 36
n. =96
-------
96_
192
T5? = 48
2 (J6-48)2 , (60-48)2 , (60-48)2 , (36-48):
X = 48 48 48 48
X2 has (r - 1) degrees of freedom
R
where
r = 2
r-1 = 1.
At a 0.5 level of significance,
X2 = 3.8.
1
Since the calculated x2 of 12.0 is greater than the table x2 of
R R
3.8, we conclude that the difference between samples taken
before construction and after construction is significant.
The column chi-square value for test (3) is computed from
the following equation:
-------
= I
n..,na/n
where
"u-
For the numerical computation:
. = 25, 26, 24, 21
. = 23, 22, 24, 27
= 48
48 1
n n 96 2
n .n n .n.
X2 _ (25-24)2+ (26-24)2+ (24-24)2+ (21-24)2
C 24
(23-24)2+ (22-24)2+ (24-24)2+ (27-24)2
24
_ 1+4 + 0 + 9^1+4 + 0 + 9
24
24
73
-------
- = 1 2
24 '^
X2 has (c - 1) degrees of freedom
C
where
c = 4
c-1 = 3.
At the 0.5 level of significance,
X2 = 7.8.
3
Since the calculated x2 of 1.2 is less than the table x2 of
C C
7.8, we conclude that the difference between samples on different
transects is not significant.
The interaction between transect location and sample time
(before and after) is denoted by x2 and is computed by
subtraction. The value is
X2 = X2 - X2 - X2 = 17 - 12 - 1.2 = X2 = 3.8.
I,(r-l)(c-l) T R C 1,3
The tabular chi-square value with three degrees of freedom is
X2 = 7.8. Since x2 < X2> it is concluded that there is no inter-
3 1,3 3
action between transect location and sampling period, i.e., there
was a uniform reduction in biological forms from transect to trans-
sect from the before to after implementation period.
-------
If this were a real situation, we might conclude, based upon
the statistical tests, that the number of fish in the area changed
after construction of the power plant. From the data, we can see
that the number declined so this would be a detrimental effect.
There was not a significant difference between transects so the
decline was apparently widespread and not just limited to the
plume area. The hot water could have reduced the food supply in
the general area so fewer fish were supported or there may be other
reasons for the decline in fish population. The information
resulting from the chemical, benthic, and other sampling programs
is essential in answering questions of this type.
The Factorial Arrangement: Analysis of Covariance (Parametric)
The analysis of covariance, combining regression analysis and
analysis of variance, is used in this paper to determine the
effect of an independent covariate (X), such as temperature, upon
a response factor (R), such as biomass of invertebrates per unit
of area sampled. This is done by adjusting the sums of squares
and mean sums (from analysis of variance) to a common value of
the covariate. If, for example, the variability in benthos weight
between transects is eliminated when the values are adjusted to a
common temperature, one would assume that temperature was a primary
factor in causing the variation between transects.
In order for this test to be valid, one must assume that the
covariates are not affected by the response factors. For measurements
75
-------
of fish and most benthos, this assumption is probably not violated
since it is doubtful that the fish or benthos will materially
affect D.O. concentration, stream velocity, temperature, pH, nutrients,
or pesticides. It is possible, though, that periphyton, phytoplankton,
and aquatic macrophytes do affect D.O. and nutrient concentrations.
In this latter case, extreme care must be used in interpreting
results because of this dependency. If the response measured (benthos)
is affected by the covariate (temperature) and not vice versa, then
that covariate can be used without unduly compromising the inter-
pretations.
Biological data may be collected at widely spaced time intervals
in comparison to the physical and chemical data. Therefore, some
decision must be made by the principal investigator concerning what
value to use as a covariate. For example, with temperature, it
could be the mean, median, maximum, minimum, or a weighted value
per time period. In some cases, more than one of these could be
used to determine whether maximum temperature had more effect than
mean temperature, or median temperature, etc.
In order to demonstrate analysis of covariance, an example
problem is presented which is an extension of the analysis of variance
(parametric) problem starting on page The response factor is
still the weight of invertebrates per area of sample. The covariate
is temperature and we assume the same situation illustrated in
Figure 3 (page 52). Good texts, such as Snedecor and Cochran (45),
Steele and Torrie (39) or Winer (44) describe this test in detail.
76
-------
TABLE A-8
SAMPLE PROBLEM RAW DATA SHEET (TEMPERATURE)
N\ Width
Transect \^
1
(above)
2
(below)
3
(0.5 AT)
4
(0.1 AT)
Block 1 - Spring
Const. Time
1 (before)
2 (after)
1
2
1
2
1
2
CW Sum
W Sum
1
(0.1 W)
50,51
101
51 ,51
102
51,50
101
71,72
143
50,51
101
61,62
123
49,49
98
52,53
105
401
473
874
2
(0.3 W)
50,50
100
50,49
99
51 ,51
102
68,67
135
51 ,51
102
60,61
121
50,bO
100
53,52
105
404
460
864
3
(0.7 W)
49,51
100
50,50
100
50,50
100
51,50
101
52,50
102
51 ,50
lol
50,51
101
50,50
100
403
402
805
4
(0.9 W)
49,50
99
51,50
101
49,50
99
50,50
100
49,50
99
50,49
99
51 ,!>1
102
49,49
98
399
398
797
sum
802
881
848
809
1607
1733
3340
Block 2 - Summer
Const. Time
1
2
1
2
1
2
1
2
1
65^64
129
64,64
128
64,64
128
84,84
168
65,65
130
75,74
149
65,64
129
67,66
133
516
578
1094
2
65,65
130
65,64
129
~64,65
129
83,82
165
64,65
129
73,74
147
65,65
130
67,67
134
518
575
1093
3
64,65
129
64,65
129
65,63
128
65,65
130
r56,65
131
65,65
130
65^,65
130
65.64
129
518
518
1036
4
65,65
130
65,65
130
65,65
130
64,65
129
65,64
129"
65,64
129
64,66
130
64,64
128
519
516
1035
sum
1034
1107
1074
1043
2071
2187
4258
Grand Sums
1836
1988
1922
1852
3678
3920
1968
1957
1841 7598
1832
-------
TABLE A-9
SAMPLE PROBLEM INTERACTION TABLE (TEMPERATURE)
Transect Time 0.1 U 0.3 W 0.7 W 0.9 W Total
Tl
TOTAL
T2
TOTAL
T3
TOTAL
T4
TOTAL
TOTAL
TOTAL
TOTAL
Cl
C2
Cl
C2
Cl
C2
Cl
C2
C.,W
c£w
GW
101+129=
230
230
460
101+128=
229
311
540
231
272
503
227
238
465
917
1051
1968
230
228
458
231
300
531
231
268
499
230
239
469
922
1035
1957
229
229
458
228
231
459
233
231
464
231
229
460
921
920
1841
229
231
460
229
229
458
228
228
456
232
226
458
918
914
1832
918
918
1836
917
1071
1988
923
999
1922
920
932
1852
3678
3920
7598
oo
-------
The data for the covariate are arranged in Tables A-8 and A-9
in exactly the same manner as were the invertebrate data in Tables
A-2 and A-4. From Tables A-2 and A-8, we see that in block 1,
width 1, transect 1, construction time 1, the numbers of inverte-
brates collected per sample area were 6 and 10. Temperatures at
the corresponding locations and times were 50 and 51 degrees,
respectively.
All sums of squares are computed for the covariate in the
same manner as they were computed for the response factors in
the analysis of variance. Equations for the calculations have
already been given in the analysis of variance section. In addition,
the cross-products must be computed for use in computing adjusted
factor sums of squares.
To separate the sums of squares, those for the response factor
(invertebrates) will be subscripted by "R," those for the covariate
by "X," and those for cross-products by "RX." Response sums of
squares are taken from the analysis of variance section and the others
are calculated as follows. Numbers are taken from Tables A-2, A-4,
A-8, and A-9.
(GT t)2/N = GSR for response (invertebrates)
= GSY for covariate (temperature)
A
= GSRX for cross-products.
79
-------
GSR = TOT = 10>603
6SY = -= 451,013
A I to
_ (1165)(7598) .
- 128
TTSSD = 1,256 (from AOV calculations)
K
TTSSV = 460,918 - 451,013 = 9905
TTSPRX = 68,974 - 69,154 = -180
BSSR = 279
BSSV
33402 + 42582
64
- 451,013 = 6,583
BSPRX = (488 x 3340) ^(677 x 4258) . 69fl54 = , ,355
TSSR = 118
TSS
_ (323 x 1836)+(251 x 1988) + (272 x 1922)+(319 x 1852)
~ 32
- 69,154 = -229
80
-------
WSSR = 102
WSS
1857' +1841' +
. 451>013 . 499
WSPn = (259 x 1968) + (267 x 1957) + (315 x 1841) + (324 x 1832)
RX 32
-69,154 = -225
CSSR = 97
CSSX = 3678^3920
TSP - (638 x 3678) + (527 x 3920)
LbKR)( - t-j-j
= _2]Q
TWSSR = 117
TWSSV
4602 + 4582 + ...2262
8
- 457 - 499 - 451,013
= 3>61^>256 - 457 - 499 - 451,013 = 438
o
TWSP = (84 x 460) + (80 x 458) + ...(84 x 226) _ (_22g) _ (_225)
RX 8
- 69,154 = 547^853 + 229 + 225 - 69,154 = -218
81
-------
TCSSR = 125
TTQC 9182 + 9182 + ...9322 , , m
TCSSy = - rg -- 4b/ - 457 - 451 ,013
= 7,238,332 _ 45y _ 457 _ 451 013 = 469
ID
TCSPDY = (162 x 918) + (161 x 918) + ...(16Qx932) _ (_22g) _ (_21Q)
KA I o
-69,154 = »» + 229 + 21Q _ 69jl54 = _241
WCSSR = 108
Q172 + Q??2 + Q14.2
WCSSX = is ^'^ - 457 - 499 - 451,013
= 7,239.560 _ 457 _ 499 _ 451 013 = 504
I b
WCSP = x 914) - (~
RX
- (-210) - 69,154 = ». + 225 + 21Q _ 69>154 = _233
TWCSSR = 101
= 230* + 230' +...226'
TWCSSX
- 451,013
= 1>847'162 - 457 - ... 451 ,013 = 454
82
-------
TWCSPRy = (41 x 230) + (43 x 230) + ... (42 x 226) _ (_22g) _ (_225)
RX 16
- (-210) - (-218) - (-241) - (-233) - 69,154 = 270j415 + 229
+ 225 + 210 + 218 + 241 + 233 - 69,154 = -194
EESSR = 80
EESSX = + 1QQ2 + ...1282 _ 451 j013 _ 6s583 _ 457 _ 499 _ 457
- 438 - 469 - 504 - 454 = 921^786 - 451,013 - 6,583 - 457 - 499
- 457 - 438 - 469 - 504 - 454 = 19
= (16 x 1Q1) + (17x1QQ) + ...(25x128)
KA
- (-229) - (-225) - (-210) - (-218) - (-241) - (-233) - (-194)
= 137'94Q - 69,154 - 1,355 + 229 + 225 + 210 + 218 + 241 + 233
+ 194 = 11
SESSR = 129
(502 + 512 + ...642) (1012 + 1Q02 + ...1282)
- "2
fc
= 460,918 - 9212'786 = 25
83
-------
SESPRX = [(6 x 50) + (10 x 51) + ...(13 x 64)]
_ [(41 x 1Q1) + (41 x 1QQ) + ...(42 x 128)]
= 68,974-^°-=4
As a check on the computations, the TTSS computed initially
can be compared to:
TTSS = BSS + TSS + WSS + CSS + TWSS + TCSS + WCSS + TWCSS
+ EESS + SESS
TTSSR= 279 + 118 + 102 + 97 + 117 + 125 + 108 + 101 + 80
+ 129 = 1,256
Calculated TTSSR = 1,256
TTSSX= 6583 + 457 + 499 + 457 + 439 + 469 + 504 + 454
+ 19 + 25 = 9,906
Calculated TTSS = 9,905. Difference is due to round-off
in component SSx's.
TTSPRX= 1355 - 229 - 225 - 210 - 218 - 241 - 233 - 194 + 11
+ 4 = -180.
Calculated TTSPRX = -180.
-------
The results of the above calculations are presented in an
analysis of covariance table, Table A-10. The sample error term
is combined with each of the terms in the top half of the table to
obtain the adjusted values in the bottom half.
The sums of the df and of products values are obtained by
addition of the sample error term and the factor being considered
(transect, width, etc). For example: transect df = 3 and sample
error df = 64; 64 + 3 = 67 df for the summed value. Transect X2 =
457 and sample error X2 = 25; 457 + 25 = 482. Transect RX = -229
and sample error RX = 4; -229 + 4 = -225, the summed value. This
operation is repeated for each term on top to fill in the bottom
left side of the table.
The adjusted values on the right side of the table are obtained
as follows.
One degree of freedom is lost in the error term for each covariate
used so adjusted df always equals df - 1. For example: in transect +
sample error, df = 67; 67 - 1 = 66 which is the adjusted df value.
The adjusted sample error sum of squares is obtained by use
of the equation ASESS = SESSR2 - [(SESSRX)2/SESSx2]
In this example, the calculation is
ASESS = 129 - = 128.4.
85
-------
TABLE A-10
SAMPLE PROBLEM ANALYSIS OF COVARIANCE
Source
Total
Block
Transect
Width
Construction Time
TW
TC
we
TWC
Experimental Error
Sample Error
"ransect + Sample
Irror
DIFFERENCE
df
127
1
3
3
1
9
3
3
9
31
64
67
Sums of Products
X2 RX R2
9905
6583
457
499
457
438
469
504
454
19
25
482
-180
1355
-229
-225
-210
-218
-241
-233
-194
11
4
-225
1256
279
118
102
97
117
125
108
101
80
129
247
Adjusted Values
df SS MS
63
66
3
128.4
142.0
13.6
2.04*1
4.53
F'
t-
2.2
Tabular F*
3.3
CO
cr\
Continued
-------
TABLE A-10 (CONT.)
SAMPLE PROBLEM ANALYSIS OF COVARIANCE
Source
Width + Sample
Error
DIFFERENCE
Construction Time
*- Sample Error
DIFFERENCE
TW + Sample Error
DIFFERENCE
TC + Sample Error
DIFFERENCE
WC + Sample Error
DIFFERENCE
TWC + Sample Errot
DIFFERENCE
df
67
65
73
67
67
» 73
Sums of Products
X2 RX R2
524
482
463
494
529
479
-221
-206
-214
-238
-229
-190
231
226
246
254
237
230
Adjusted Values
df SS MS
66
3
64
1
72
9
66
3
66
3
72
9
137.8
9.4
138.0
9.6
147.1
18.7
139.3
10.9
137.9
9.5
154.6
26.2
3.13
9.60
2.08
3.63
3.17
2.91
F1
1.5
4.7
1.0
1.8
1.6
1.4
Tabular F*
3.3
5.3
2.3
3.3
3.3
2.3
* Tabular F ratios were selected at the 2.5% level of significance to better illustrate the problem.
** EMS = SEMS from Table A-6.
CD
-------
The adjusted transect + sample error sum of squares is
obtained by use of the following equation:
ATSS + (SESSR2 + TSSR2) -
(SESSRX + TSSRX)2
SESSY2 + TSSV2
A A
In this example, it is:
ATSS = (129 + 118)
(4 - 229)2
25 + 457
= 141.97 rounded to 142.0.
Adjusted width + sample error sum of squares is:
AWSS = (SESSR2 + WSSR2) -
(SESS
WSSRX)*
SESSV2 + WSSV2
A A
AWSS = 129 + 102 -
(4 - 225)2
25 + 499
= 137.8.
The adjusted sums of squares for sample error and construction time,
TW, TC, WC, and TWC are calculated in a like manner.
The adjusted df and adjusted sums of squares for the treatment
(i.e., transect, width, etc.) + sample error are then used to
calculate the adjusted values of degrees of freedom, sums of squares
and mean squares for the treatment. Treatment adjusted values for
df are obtained by subtracting sample error df from the treatment
+ sample error df value being considered. For example: adjusted df
for transect = 66 - 63 = 3; for width it is 66 - 63 = 3; for
construction time, it is 64 - 63 = 1, etc.
-------
Treatment adjusted values for SS are obtained in the same
manner. For example: treatment adjusted SS for transect = 142.0 -
128.4 = 13.6. The rest are done in the same way.
Adjusted means squares are obtained by the following equation:
Adjusted MS = ^justed SS
Adjusted df
For example: adjusted MS for transect = ~^- = 4.53.
O
The calculated F value (F1) is obtained by dividing each
adjusted MS by the sample error MS. For example: F1 for transect
_ 4.53 _
' ~
The tabular F value is taken from a standard table of F ratios
using the adjusted sampling error and adjusted difference degrees
of freedom. For example, the F value for transect is obtained for
63 and 3 degrees of freedom and the value is 3.3 at the 2.5 percent
level of significance. The 2.5 percent level was chosen arbitrarily
because it best illustrated the point of the example problem.
If the calculated F1 value in Table A-10 exceeds the tabular
value we conclude that the variation in weight of invertebrates
per unit area of sample is significant for that factor (transect,
width, etc.) even after the data have been adjusted to a common
value for temperature. In this case, this does not happen since
the tabular F is always greater than the calculated F1. So, we
conclude that the differences in benthos are not significant after
the effect of temperature has been eliminated.
89
-------
In the analysis of variance problem, page 61, the effects of
transect, width, time, and all the combinations were significant.
Now, when the covariate (temperature) effect is eliminated by
adjustment to a common value, the effects of transect, width, time,
and the combinations are not significant. From this, we conclude
that the variation in invertebrate weight per unit area sampled is
due at least in part to the variation in temperature. Since the
data indicate that invertebrates were reduced in the hottest areas,
we conclude that the increase in temperature due to the power plant
is detrimental to the system.
This same analysis can be used for all the other possible
covariates such as D.O., stream velocity, etc. This will aid in
separating out effects of these other factors. Temperature is a
strong influence in this case, but the other factors could also
be important. For example, velocity changes due to intake or
outfall design may affect benthic organisms. This will show up
if velocity is used as a covariate. Since a computer is needed for
this analysis, all the probable covariates should be run at the
same time to give a broader picture.
Factorial Arrangement: Analysis of Covariance (Nonparametric)
Whenever the data exhibit extreme non-normality or are basically
a count, a nonparametric analysis is used. In this analysis, the
numerical values for response and the covariate are ranked and an
-------
ordinary linear regression is performed on the ranked values. The
output consists of adjusted responses. These adjusted responses
are then analyzed by analysis of variance. This method of analysis
is presented in detail in the paper by Quade (43).
-------
LITERATURE CITED
1. U. S. Department of the Interior, FWPCA. 1968. Cost of clean
water -- Vol. II. U. S. Government Printing Office,
Washington, D. C. 244 p.
2. Remirez, R. 1968. Thermal pollution -- Hot issue for industry.
Chemical Engineering 75(7):43-52.
3. Kolflat, Tor. 1968. Thermal discharges. Industrial Water
Engineering 5(3):26-31.
4. Cairns, J. Jr. 1956. Effects of increased temperature on
aquatic organisms. Industrial Wastes 1(4):150-152.
5. Anderson, R. 1968. Temperature and rooted aquatics. Presented
at Second Annual Thermal Workshop of the U. S. International
Biological Program. Solomons, Maryland. November 3-7, 1968.
6. U. S. Department of the Interior, FWPCA. 1968. Industrial
waste guide on thermal pollution. Pacific Northwest Water
Laboratory, Corvallis, Oregon 97330. 112 p.
7. Edinger, John Eric and John C. Geyer. 1968. Analyzing steam
electric power plant discharges. Journal of the Engineering
Division, Proceedings of the American Society of Civil
Engineers. SA4, August 1968:611-623.
8. Edinger, John Eric and Edward M. Polk, Jr. 1969. Initial
mixing of thermal discharges into a uniform current. Report
No. 1. Department of Environmental and Water Resources
Engineering, Vanderbilt University, School of Engineering,
Nashville, Tennessee.
9. Jen, Yuan, Robert L. Weigel, and Ismail Mobarek. 1966. Surface
discharge of horizontal warm-water jet. Jour, of the Power
Division, Proceedings of the American Soc. of Civil Engineers
P02, April 1966:1-30.
10. Edinger, John E. and John C. Geyer. 1965. Heat exchange in
the environment. Cooling water studies for Edison Electric
Institute, Research Project No. 49. The Johns Hopkins
University, Baltimore, Maryland.
-------
11. Harleman, R. F. and W. C. Huber. 1968. Laboratory studies on
thermal stratification in reservoirs. Proceedings of the
Specialty Conference on Current Research into the Effects
of Reservoirs on Water Quality, January 1968, Sponsored by
American Society of Civil Engineers. Tech. Rept. 17.
Department of Environmental and Water Resources Engineering,
Vanderbilt University, Nashville, Tennessee.
12. Baumgartner, D. J. and D. S. Trent. 1969. Ocean outfall design,
Part I, Literature review and theoretical development. In
Press. U. S. Department of the Interior, FWPCA, Pacific
Northwest Water Laboratory, Corvallis, Oregon 97330.
13. Cederwall,K. 1968. Hydraulics of marine waste disposal.
Hyd. Div. Chalmers Institutes of Technology, Groteberg,
Sweden, Publication No. 42.
14. Fan, Loh-Nien. 1957. Turbulent buoyant jets into stratified
or flowing ambient fluids. W. M. Keck Lab. of Hydraulics
and Water Resources, Pubs. No. KH-R-15 and KH-R-18,
California Institute of Technology.
15. Hoover, E. E. 1938. Fish population of primitive brook trout
streams of northern New Hampshire. Trans. N. Amer. Wild!.
Conf. 3:486-496.
16. Gerking, S. D. 1949. Characteristics of stream fish populations
Investigations of Indiana Lakes and Streams 3:283-309.
17. Cleary, Robert E. and John Greenbank. 1954. An analysis of
techniques used in estimating fish populations in streams,
with particular reference to large non-trout streams.
Journal Wildlife Management 18(4):461-477.
18. U. S. Department of the Interior, FWPCA. 1967. Pollutional
effects of pulp and paper mill wastes in Puget Sound --
A report on studies conducted by the Washington State
enforcement project. Northwest Regional Office, Portland,
Oregon. 474 p.
19. Neil!, William H. 1969. Ecological responses of Lake Monona
(Dane County, Wisconsin) fishes to heated influent water.
Annual Progress Report to "Wisconsin Utilities Assoc."
Madison Gas & Electric Co. Coordinating co-author at
Department of Zoology, University of Wisconsin, Madison,
Wisconsin 53706.
93
-------
20. Sjolseth, D. E., E. 0. Salo, R. A. Bishop, and W. G. Williams.
1968. Final report - Studies of juvenile salmon in the
Nooksack River system and Bellingham Bay. Fisheries
Research Institute, College of Fisheries, University of
Washington, Seattle, Washington 98105. 53 p.
21. Rounsefell, George A. and W. Harry Everhart. 1953. Fishery
science - its methods and applications. John Wiley and
Sons, Inc., New York. 444 p.
22. Salo, E. 0. 1968. Final report for the period June 1, 1965-
September 30, 1968, estuarine ecology research project.
Fisheries Research Institute, College of Fisheries,
University of Washington, Seattle, Washington 98105. 80 p.
23. Coutant, C. C. 1969. Behavior of sonic-tagged chinook salmon
and steel head trout migrating past Hanford thermal discharges.
Biological effects of thermal discharges: annual progress
report for 1968. Battelle Northwest Laboratories Report
1050. Battelle Memorial Institute, Richland, Washington
99352. p. 21-27.
24. U. S. Department of the Interior, FWPCA. 1970. FWPCA methods
for the collection and analysis of freshwater plankton,
periphyton, and macroinvertebrate samples. FWPCA Division
of Water Quality Research, Analytical Quality Control
Laboratory, 1014 Broadway, Cincinnati, Ohio. 106 p.
25. Fager, E., A. Flechsig, F. Ford, R. Clutter, and R. Ghelardi.
1966. Equipment for use in ecological studies using
SCUBA. Limnol. and Oceanog. 11:503-509.
26. Somers, Lee H. 1967. Diving techniques as applied to
geological investigations of the Great Lakes. Proceedings,
Tenth Conference on Great Lakes Research held at the
University of Toronto, Toronto, Ontario, April 10-12, 1967.
p. 149-156.
27. Anderson, J. B. and William T. Mason, Jr. 1966. The use of
limestone-filled samplers for collecting macroinvertebrates
from large streams. U. S. Department of the Interior, FWPCA,
Water Pollution Surveillance System Applications and
Development Report No. 17. 19 p.
-------
28. Hester, F. Eugene and J. S. Dendy. 1962. A multiple-plate
sampler for aquatic macroinvertebrates. Iran. Amer. Fish
Soc. 91(4):420-421.
29. Mackenthun, Kenneth M. and William Marcus Ingram. 1967.
Biological associated problems in freshwater environments -
their identification, investigation and control. U. S.
Department of the Interior, FWPCA. U. S. Government
Printing Office, Washington, D. C. 287 p.
30. Mackenthun, Kenneth M. 1969. The practice of water pollution
biology. U. S. Department of the Interior, FWPCA. U. S.
Government Printing Office, Washington, D. C. 281 p.
31. Grzenda, A. R. and M. L. Brehmer. 1960. A quantitative method
for the collection and measurement of stream periphyton.
Limnol. and Oceanog. 5:190-194.
32. Patrick, R., M. H. Hohn and J. H. Wallace. 1965. A new
method for determining the pattern of the diatom flora.
Notulae Naturae 259:1-12.
33. Smith, John T. Jr., Editor in Chief. 1968. Manual of color
aerial photography. American Soc. of Photogrammetry,
105 N. Virginia Ave., Falls Church, Virginia. 550 p.
34. U. S. Department of the Interior, FWPCA. 1969. FWPCA methods
for chemical analysis of water and wastes, November 1969.
FWPCA Division of Water Quality Research, Analytical Quality
Control Laboratory, 1014 Broadway, Cincinnati, Ohio. 280 p.
35. Margalef, D. Ramon. 1957. Information theory in ecology.
Memorias de la Real Academia de Ciencias y Artes de
Barcelona 23:373-449.
36. Margalef, Ramon. 1968. Perspectives in ecological theory.
University of Chicago Press, Chicago, Illinois. Ill p.
37. Wilhm, J. L. 1967. Comparison of some diversity indices
applied to population of benthic macroinvertebrates in a
stream receiving organic wastes. Journal Water Pollution
Control Federation 39(10):1673-1683.
38. Kendall, M. C. 1966. Discrimination and classification.
C. E. I. R., ltd., London. 20 p.
95
-------
39. Steele, R. G. D. and J. H. Torrie. 1964. Principles and
procedures of statistics. McGraw-Hill Book Company, Inc.,
New York.
40. Cochran, W. G. and G. M. Cox. 1964. Experimental Designs.
Second Edition, John Wiley and Sons, Inc., New York.
41. Winer, B. J. 1962. Statistical principles in experimental
design. McGraw-Hill Book Company, Inc., New York.
42. Wilson, Kellogg V. 1956. A distribution-free test of analysis
of variance hypotheses. Psychological Bulletin 53(1):96-101.
43. Quade, Dana. 1966. Rank analysis of covariance. University
of North Carolina Institute of Statistics, Mimeo Series #483.
22 p.
44. Seigel, Sidney. 1956. Nonparametric statistics for the
behavioral sciences. McGraw-Hill Book Company, New York.
312 p.
45. Snedecor, George W. and William G. Cochran. 1967. Statistical
methods. The Iowa State University Press, Ames, Iowa.
593 p.
-------
BIBLIOGRAPHY
This section is included to present references, in addition
to those cited, which may prove useful to personnel planning field
studies. No attempt has been made to include every reference
available. Instead, especially pertinent papers and review papers
are listed. These may be used as a starting point and their
bibliographies consulted for additional papers of interest.
1. Adams, J. R. 1968. Thermal effects and other considerations
at steam electric plants - a survey of studies in the
marine environment. Pacific Gas and Electric Company
Report No. 6934:4-68. Pacific Gas and Electric Company,
4245 Hoi 1 is Street, Emeryville, California 94608.
2. American Public Health Association. 1965. Standard methods
for the examination of water and wastewater, including
bottom sediments and sludges. 12th edition. American
Public Health Association, Inc., 1790 Broadway, New York,
N. Y. 10019. 769 p.
3. Cushing, D. H. 1968. Fisheries biology; a study in population
dynamics. University of Wisconsin Press, Madison, Milwaukee,
and London. 200 p.
4. Edison Electric Institute. 1969. A summary of environmental
studies on water problems by investor-owned electric
utility companies recently completed, under way, and proposed.
Edison Electric Institute, 7530 Third Street, New York, N. Y.
10017.
5. Funk, J. L. 1958. Relative efficiency and selectivity of
gear used in the study of stream fish populations. Trans.
23rd No. Am. Wildlife Conf. 1958:236-248.
6. Holme, N. A. 1964. Methods of sampling the benthos. Adv.
Mar. Biol. 1964(2):171-260.
97
-------
7. Ingram, William M., Kenneth M. Mackenthun, and Alfred F-
Bartsch. 1966. Biological field investigative data for
water pollution surveys. U. S. Government Printing
Office, Washington, D. C. 139 p.
8. Lagler, Karl F. 1956. Freshwater fishery biology. Wm. C.
Brown and Company, Dubuque, Iowa. 421 p.
9. Lagler, Karl F., J. E. Bardach, and R. R. Miller. 1962.
Ichthyology. John Wiley and Sons, Inc., New York and
London. 545 p.
10. Littleford, Robert A., Curtis L. Newcombe, and Boland B.
Shepard. 1940. An experimental study of certain
quantitative plankton methods. Ecology 21(3):309-322.
11. Livingston, Robert Jr. 1965. A preliminary bibliography
with KWIC index on the ecology of estuaries and coastal
areas of the eastern United States. U. S. Department of
the Interior, Fish and Wildlife Service Special Scientific
Report - Fisheries No. 507. 352 p.
12. Mackenthun, Kenneth M. 1966. Biological evaluation of polluted
streams. Journal Water Pollution Control Federation
38(2):241-247.
13. Mason, William T. Jr., J. B. Anderson, R. Douglas Kreis, and
William C. Johnson. 1967. Collecting macroinvertebrates
in a polluted stream using rock-filled samplers. U. S.
Department of the Interior, FWPCA. Applications and
Development Report Number 28. 23 p.
14. Parker, Frank L. and Peter A. Krenkel. 1969. Thermal pollution:
status of the art. Report No. 3. Department of Environmental
and Water Resources Engineering, Vanderbilt University,
Nashville, Tennessee.
15. Pennak, Robert W. 1953. Fresh-water invertebrates of the
United States. Ronald Press Company, New York. 769 p.
16. Proffitt, Max A. 1969. Effects of heated discharge upon
aquatic resources of White River at Petersburg, Indiana.
Water Resources Research Center, Bloomington, Indiana.
Investigations No. 3. 101 p.
-------
17. Regner, Henry A. 1968. Fish size parameters useful in
estimating gill-net selectivity. The Progressive Fish
Culturist 31(l):57-59.
18. Starrett, W. C. and P. G. Barnickol. 1955. Efficiency and
selectivity of commercial fishing devices used on the
Mississippi River. 111. Nat. Hist. Surv. Bull. 26(4):
323-366.
19. Trembley, F. J. 1961. Research project on effects of
condenser discharge water on aquatic life. Institute
of Research, Lehigh University, Progress Report, 1956-
1959.
20. U. S. Department of the Interior. 1955. Fishery publication
index, 1920-54. Fish and Wildlife Service Circular 36.
254 p.
21. U. S. Department of the Interior, FWPCA. 1967- Temperature
and aquatic life. Laboratory Investigations Series No. 6.
Technical Advisory and Investigations Branch, 5555 Ridge
Avenue, Cincinnati, Ohio 45213.
22. Weber, Cornelius I. 1966. Methods of collection and analysis
of plankton and periphyton samples in the water pollution
surveillance system. U. S. Department of the Interior,
FWPCA. Water Pollution Surveillance System Applications
and Development Report No. 19. 25 p.
23. Weber, Cornelius I. and Ronald L. Raschke. 1966. Use of a
floating periphyton sampler for water pollution surveillance.
U. S. Department of the Interior, FWPCA. Water Pollution
Surveillance System Applications and Development Report
No. 20. 22 p.
24. Welch, Paul S. 1948. Limnological methods. McGraw-Hill
Book Company, Inc., New York. 381 p.
99
-------
Access/on Number
Subject Field & Group
010
SELECTED WATER RESOURCES ABSTRACTS
INPUT TRANSACTION FORM
Organization
Department of Interior, Federal Water Quality Administration, Northwest Region
Pacific Northwest Water Laboratory, Corvallis, Oregon
Title
GUIDELINES: BIOLOGICAL SURVEYS AT PROPOSED HEAT DISCHARGE SITES,
1 Q Author(s)
Garton, Ronald
HavUnc Dalnh
R.
n
16
21
Project Designation
Note
22
Citation
Water Pollution Control Research Series, FWQA, April 1970, 103 p., 3 fig, 10 tab,
69 ref.
Descriptors (Starred First)
23
*Thermal pollution, *Siting surveys, *Biological sampling, Data analysis
Thermal power plants, Ecological studies
25
Identifiers (Starred First)
27
Abstract
A quantitative approach is presented for the biological portion of thermal discharge
siting surveys and discharge monitoring. Three types of studies are covered: Type I
is a very general study of the aquatic system and the pertinent literature; Type II
is a comprehensive pre-operational study designed to supply data for management deci-
sions on power plant siting and data to serve as baseline for possible future compari-
son; and Type III is a post-operational continuation of Type II to detect possible
effects if a thermal discharge to a natural water body is allowed.
Two methods are recommended for location of sampling stations by use of a grid system
based on planned outfall design. Sample collection and handling methods and frequency
of sampling are suggested for fish, macroinvertebrates, plankton, periphyton and aqua-
tic macrophytes.
Methods of data handling recommended include diversity and redundancy indices and a
combination of the two into one value for a test for dispersion. A scale of importance
is suggested for organisms of special value in either an economic or ecologic sense.
For statistical analysis of the data, four appropriate methods are recommended and
sample problems are provided to illustrate data handling and conclusions to be drawn
from the tests.
(Garton-FWQA)
Abstractor
R. R. Gartnn
Insti tut ion
-S
'.I. S Department of Interior, FWQA
~ StSlD TO: WATER RESOUR
WRiIOZ (REV. JULY 1969)
WRSIC
WATER RESOURCES SCIENTIFIC INFORMATION CENTER
U.S. DEPARTMENT OF THE INTERIOR
WASHINGTON. D. C. 20240
* CPO: IQ69-359-339
------- |