Trophic state of lakes in north central Florida

MISSING IMAGE

Material Information

Title:
Trophic state of lakes in north central Florida
Series Title:
Florida Water Resources Research Center Publication Number 13
Physical Description:
Book
Creator:
Brezonik, Patrick L.
Shannon, Earl E.
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Notes

Abstract:
General limnological and trophic conditions of 55 lakes and ponds in north and central Florida were established over an extensive one year sampling period. Florida lakes are typically shallow and in a sandy terrain. Most of the lakes have soft water, and high organic color is a common but variable property. Trophic conditions range from ultraoligotrophy in the sand-hill lakes of the Trail Ridge region to hypereutrophy in some large drainage lakes in Alachua County and in the Oklawaha River Basin. Trophic data were analyzed by multivariate techniques, and logical trophic groups derived by cluster analysis. A quantitative index of trophic state (TSI) was derived using 7 trophic indicators, and the TSI values were used to establish quantitative relationships between lake trophic conditions and watershed characteristics. Nitrogen and phosphorus budgets were calculated for the lakes based on land use and population patterns in the watersheds, and critical loading rates were estimated from the budgets and the trophic conditions.

Record Information

Source Institution:
University of Florida Institutional Repository
Holding Location:
University of Florida
Rights Management:
All rights reserved by the source institution and holding location.
System ID:
AA00001463:00001


This item is only available as the following downloads:


Full Text













Publication No. 13

Trophic State of Lakes in North Central Florida

By

Patrick L. Brezonik and Earl E. Shannon

Department of Environmental Engineering Sciences
University of Florida
Gainesville















TROPHIC STATE OF LAKES IN NORTH CENTRAL FLORIDA


by



PATRICK L. BREZONIK

and

EARL E. SHANNON


PUBLICATION NO. 13

FLORIDA WATER RESOURCES RESEARCH CENTER






RESEARCH PROJECT TECHNICAL COMPLETION REPORT


OWRR Project Number B-004-FLA


Matching Grant Agreement Numbers

14-31-0001-3068 (1970)
14-31-0001-3068 (1971)

Report Submitted: August 3, 1971






The work upon which this report is based was supported in part
by funds provided by the United States Department of the
Interior, Office of Water Resources Research as
Authorized under the Water Resources
Research Act of 1964.














TABLE OF CONTENTS


ABSTRACT . . . . .

CHAPTER 1. EUTROPHICATION AND FLORIDA LAKES .

A. INTRODUCTION . . .

B. NATURE OF EUTROPHICATION. . .

C. QUANTIFYING EUTROPHICATION. . .

D. COMPOSITION OF THE LAKE STUDY GROUP .

CHAPTER 2. EXPERIMENTAL PROCEDURES. . .

A. SAMPLING METHODS . . .

B. PARAMETERS EVALUATED AND EXPERIMENTAL
TECHNIQUES . . .

C. MULTIVARIATE ANALYTICAL METHODS .

CHAPTER 3. LIMNOLOGICAL RESULTS . .

A. MORPHOMETRIC AND PHYSICAL FEATURES.

B. GENERAL CHEMICAL CHARACTERISTICS. .


Page




2

2

2

6




. 15

15
. 1

. 2










. 2













. 28
. 6
. 11

. 15

. 15


. 16

. 19

. 28

. 28

. 32


C. PHYTOPLANKTON AND MACROPHYTE CHARACTERISTICS.

D. SEDIMENTS . . . . .

CHAPTER 4. CLASSIFICATION AND QUANTIFICATION OF
TROPHIC CONDITIONS IN FLORIDA LAKES . .

A. DEVELOPNMET'I' OF A TROPHIC CLASSIFICATION
SYSTEM FOR FLORIDA LAKES . . .

B. DEVELOPMENT OF DISCRIMINANT FUNCTIONS TO
CLASSIFY LAKES OUTSIDE THE ORIGINAL SAMPLE
GROUP . . . . .

C. FORMULATION OF TROPHIC STATE INDICES. .

CHAPTER 5. RELATIONSHIPS BETWEEN TROPHIC STATE AND
WATERSHED ENRICHMENT FACTORS . . .

A. INTRODUCTION . . . .











B. NITROGEN AND PHOSPHORUS BUDGETS . .

C. RELATIVE IMPORTANCE OF VARIOUS NUTRIENT
SOURCES . . . . .

D. STATISTICAL ANALYSIS OF TSI vs. NITROGEN
AND PHOSPHORUS LOADING RATES . .

E. CRITICAL NUTRIENT LOADING RATES:
APPLICATION TO LAKE MANAGEMENT. . .

F. EFFECT OF DEPTH ON LAKE CAPACITY TO
ASSIMILATE NUTRIENTS. . . .

G. SOURCES OF UNCERTAINTY . . .

H. RELATIONSHIPS BETWEEN TROPHIC STATE AND
GENERAL WATERSHED CONDITIONS . .

I. RELATIONSHIP BETWEEN TSI AND TOTAL WATERSHED
AREA . . . . .

ER 6. CONCLUSIONS. . . . .


APPENDIX . .

ACKNOWLEDGEMENTS .

BIBLIOGRAPHY .

ADDENDUM . .


S. 92

S. 95

. 96


Page

65


72


74


CHAPT











ABSTRACT


TROPHIC STATES OF LAKES IN NORTH CENTRAL FLORIDA


General limnological and trophic conditions of 55 lakes
and ponds in north and central Florida were established over
an extensive one year sampling period. Florida lakes are
typically shallow and in a sandy terrain. Most of the lakes
have soft water, and high organic color is a common but vari-
able property. Trophic conditions range from ultraoligotrophy
in the sand-hill lakes of the Trail Ridge region to hyper-
eutrophy in some large drainage lakes in Alachua County and
in the Oklawaha River Basin.

Trophic data were analyzed by multivariate techniques,
and logical trophic groups derived by cluster analysis. A
quantitative index of trophic state (TSI) was derived using
7 trophic indicators, and the TSI values were used to estab-
lish quantitative relationships between lake trophic condi-
tions and watershed characteristics. Nitrogen and phosphorus
budgets were calculated for the lakes based on land use and
population patterns in the watersheds, and critical loading
rates were estimated from the budgets and the trophic condi-
tions.









Brezonik, P.L. and Shannon, E.E.
TROPHIC STATES OF LAKES IN NORTH AND CENTRAL FLORIDA
Completion Report to the Office of Water Resources Research,
Department of Interior, July, 1971, Washington, D.C. 20240
KEYWORDS: eutrophication/ nitrogen/ phosphorus/ multivariate
analysis/ water quality/ lakes/ nutrients/ Florida/ models.










CHAPTER 1. EUTROPHICATION AND FLORIDA LAKES


A. INTRODUCTION


Although Florida has more than 7500 lakes (Florida Board
of Conservation 1969), limnological investigations of these
lakes have been few and limited to special interests. Most
detailed studies have been centered on a few unusual or recre-
ationally important lakes; for example, Mud Lake (Marion
County) (Bradley and Beard, 1969; lovino and Bradley, 1969),
Lake Mize CAlachua County) (Brezonik and Harper, 1969; Keirn
and Brezonik, in press) and Lake Apopka (Orange and Lake Coun-
ties) (for a review, see Sheffield and Kuhrt, 1970). Yount
(1963) has reviewed most pre-1960 limnological studies in
discussing some general features of Florida lakes.

However, as a group Florida lakes are almost limnologi-
cally unknown. Threatening of the recreational assets of
Florida lakes by cultural encroachment and consequent nutrient
enrichment has stimulated studies on these lakes. In 1968
the University of Florida Department of Environmental Engi-
neering initiated an extensive survey of the physical, chemi-
cal and biological characteristics of 55 lakes in north and
central Florida. The investigation had five main objectives:
i) to determine the basic limnological features of lakes in
the region; ii) to assess the present water quality trophicc
state) characteristics of the lakes and provide baseline data
for future studies; iii) to evaluate the applicability of the
common trophic state indicators to sub-tropical lakes; iv)
to provide necessary data to develop an index of trophic state
for sub-tropical lakes; v) to study the relationships between
lake trophic state and lake watershed conditions influencing
trophic state.


B. NATURE OF EUTROPHICATION


Cultural lake eutrophication is an undesirable consequence
of the interaction between man and his environment. Many of
his agricultural, industrial, domestic and recreational activ-
ities are introducing excess nutrients into surface waters,
causing significant water quality deterioration. Since fresh
water is vital to the total well-being of the environment,
man has an obligation to protect his valuable lacustrine re-
sources. However, progress in solving the problem has been
retarded by the inherent complexity of the eutrophication pro-
cess, and considerable vagueness still exists concerning the
definition of cause and effect relationships in the overall
process (Brezonik, 1969; Putnam, 1969).









It is generally agreed that eutrophication involves nutri-
ent enrichment, and a lake in time responds to this enrichment.
This response is reflected in a lake's trophic state (eutrophic
condition). However, few efforts have been devoted to quan-
tifying the relationship of eutrophication to trophic state.

One of the problems in the study of lake eutrophication
is of a semantical nature; i.e. distinguishing between and
defining the causes, symptoms and effects. Considerable liter-
ature has been devoted to discussing these concepts. The mean-
ing of the term eutrophicationn" has been stated by Hasler
(1947) as being, simply, the enrichment of water, be it in-
tentional (cultural) or unintentional (natural). This nutrient
enrichment is generally considered as the causal mechanism in
the overall eutrophication process. As originally suggested
by Naumann (1919) perhaps primary consideration should be given
to nitrogen and phosphorus nutrients. The concept of trophic
state (degree of eutrophy) is difficult to define. Eutrophic
conditions are the consequences or effects of a lake's nutrient
enrichment, but there is no way to express this state in sim-
ple, quantitative terms. Much of the conceptual difficulty
with the idea of trophic state could have been avoided long
ago had limnologists defined trophic state in precise terms
as a measure either of a lake's productivity or of a lake's
nutrient status. Instead the term has been used to refer to
both characteristics. While correlated to a degree, produc-
tivity and nutrient status are both also functions of other
independent phenomena (e.g. hydrology and climate).

Adequate description of a lake's trophic state requires
consideration of several different physical, biological and
chemical characteristics. For this reason the concept of
trophic state is not only multi-dimensional but hybrid, as
suggested by Margalef (1958). The trophic state of a lake
cannot be measured directly because of its multi-dimensional
nature. However, it is evidenced by various symptoms called
trophic state indicators. A list of common indicators of
trophic state is in Table 1. Reviews of trophic state indi-
cators have been compiled by Fruh et al. (1966), Vollenweider
(1968), Hooper (1969) and Stewart and Rohlich (1967).

There has been no scarcity of lake classification schemes
and a review of such is beyond the scope of this report.
Birge and Juday (1927) made a fundamental distinction concern-
ing the origin of dissolved organic matter in lakes. Lakes
dependent on internal sources (_primary production) were auto-
trophic and lakes dependent on external sources were allo-
trophic. Later Aberg and Rohde (1942) related the classical
trophic types of lakes in a two-dimensional concept of auto-
trophy and allotrophy. This general approach was used for
the classification purposes in this study and the idealized
two-dimensional relationship is shown in Figure 1. Organic
color measurements were assumed to be indicative of external-












Table 1. Trophic Indicators and Their Response
to Increased Eutrophication1


Physical


Transparency Cd)
(Secchi disc
reading)
Morphometry (D)
(mean depth)


Chemical


Nutrient concentrations
CI) Ce.g. at spring
maximum)
Chlorophyll a (I)
Conductivity (I)
Dissolved solids (I)
Hypolimnetic oxygen
deficit CI)
Epilimnetic oxygen
supersaturation (I)
Sediment type


Biological2

Algal bloom fre-
quency (I)
Algal species di-
versity (D)
Littoral vegeta-
tion (I)
Zooplankton (I)
Fish (I)
Bottom fauna CI)
Bottom fauna di-
versity CD)
Primary production CI)


1(I) after parameter signifies value increases with eutrophi-
cation: (D) signifies value decreases with eutrophication.

2Biological parameters all have important qualitative changes,
i.e. species changes as well as quantitative Cbiomass) changes
as eutrophication proceeds.


From Brezonik (1969)




























O
00
o a
Z oCg a




z o0 0 0o
0 0odui 0 0 i- W
o w a.



0 W M.




J
Production) and Allotrophy (External Organic Input)
(n > < 0 00 a.
-_J 0 0 0

_j 0 _._ M




MEASURE OF TROPHIC STATE







Figure 1. Two-Dimensional Concept of Lake
Classification Based on Autotrophy (Internal Organic
Production) and Allotrophy (External Organic Input)









source dissolved organic matter and thus denote lake allo-
trophy. As originally suggested by Hansen (1962), colored
and relatively clear lakes were recognized as two fundamentally
different lake types. Within each of these types, oligo-,
meso-, and eutrophic state subdivisions could occur as deter-
mined by some measure of lake trophic state.


C. QUANTIFYING EUTROPHICATION


From a qualitative viewpoint the phenomenon of eutrophi-
cation is now fairly well understood. However, for lake
management eutrophication control qualitative facts are seldom
sufficient. For example, it is generally recognized that
increased nitrogen and phosphorus input to a lake will gener-
ate increased plant production. But information concerning
the precise nutrient loading rates that stimulate excessive
production and scum-forming algal blooms is sorely lacking.
Lakes are highly complex ecosystems, and the factors control-
ling nutrient cycling and primary and secondary production
in them are at best poorly understood. Furthermore, lakes
cannot be regarded as isolated entities, but the interactions
of the entire watershed with the lake itself must be taken
into account (Hutchinson, 1969). The general significanceof
various land use patterns and cultural activities as nutrient
sources are largely unknown, and in particular the total
nutrient loading rates for specific lakes of varying trophic
conditions are known with accuracy for only a few cases.

The complexities of the eutrophication problem suggest
the utility of systems analysis techniques and of mathematical
modeling in properly defining the problem and simplifying it
to the extent that solutions become feasible. The theory and
nature of mathematical ecosystem models have been discussed
in several recent papers and books (Moreau, 1969; Patten,
1969; Watt, 1968; and Thomann, 1971). In general mathematical
models can be divided into two types. Analytical or mechanis-
tic models consist of a series of equations (algebraic, or in
ecosystem models more commonly, differential) which attempt
to explain the fundamental (functional) relationships between
certain parameters. For example, differential equation models
of primary production have been developed (Patten, 1968) in
terms of the basic relationships between photosynthesis and
light intensity, nutrient levels, etc. Empirical or statis-
tical models are composed of approximate parameter relation-
ships which are derived by such techniques as regression,
multi-variate, or time series analyses. Such models are
attractive in management of complex systems where cause-
effect relationships are unknown. Empirical models can be use-
ful in predicting system response to changes in environmental
conditions, and they can give clues to the significance of
the relationships Ci.e. the dependency) between variables.
However their lack of foundation in causal relationships renders









empirically developed models susceptible to misuse and over-
extension (to conditions in which they may not be applicable).

The inherent complexities of nutrient enrichment and its
attendant effects on lakes imply that a purely deterministic
approach is beyond our present capabilities. While functional
relationships are known for various lacustrine phenomenon,
and relatively sophisticated analytical (i.e. differential
equations) models have recently been formulated for even as
complicated a process as planktonic production (Chen, 1970;
DiToro et al., 1970; Patten, 1968), the much larger scope of
the eutrophication problem precludes such approaches at the
present time, especially in the general case. For particularly
unique and valuable resources like Lake Tahoe or the St.
Lawrence Great Lakes, the manpower and time expenditures re-
quired for development of such models may be justified. This
seems not to be the case for the thousands of smaller and
locally important recreational lakes in the U. S. and else-
where. A simpler, less costly approach is required for these
lakes.

Where large numbers of lakes must be managed an attractive
possibility is the development of empirical models based on
data from a representative sample of the lakes in question.
Such management tools as critical nutrient loading rates can
be developed by empirical manipulation of basic limnological
and watershed information. While empirical models are perhaps
not the ultimate answer to eutrophication problems, they can
provide direction for further studies and models while simul-
taneously providing interim predictive capacities required
for proper water quality management.

Eutrophication is a multivariable problem and thus lends
itself to analysis by multivariate statistical techniques.
Beneficial applications of empirical multivariate models can
be anticipated in three major areas of eutrophication research,
and Table 2 summarizes potential applications in each area.
Because of the broad, multi-dimensional concept of trophic
state, multivariate techniques seem especially appropriate
for the long standing problem of rational lake classification.
Trophic classification systems can be useful in several ways:
a) for identification la certain class Cname) calls to mind
certain distinctive characteristics]; b) for organization of
our knowledge concerning the objects (lakes) being classified;
c) as the basis for development of theories regarding cause's
of phenomena associated with a particular class (e.g. what
do lakes in a class have in common that might induce their
similar behavior), and d) for management purposes Cdifferent
classes of lakes may have different "best uses" and require
different land use and water management controls).

The ill-defined concept of trophic state is in reference
to both a lake's general nutrient status and its productivity,













Table 2. Applications of Empirical Models
to Quantification of Eutrophication



1. Lake Classification

a. formation of logical lake groups according to multi-
dimensional concept of trophic state

b. delineation of the set of conditions (ranges for
indicator values) defining different trophic groups

c. determination of redundancy and uniqueness among
various trophic indicators


2. Quantification of Trophic State

a. development of uni-dimensional quantitative trophic
state index (TSI)

b. correlation of classical trophic indicator values with
water quality problems


3. Relationship between Lake Trophic State and Causative Factors

a. regression models of TSI vs. N and P loading

b. regression models of trophic state vs. population
and land use patterns (including basin hydrology and
morphometry).









which are not always correlated. The circumstances defining
a given state (e.g. eutrophy) are not at all agreed upon by
limnologists. No single measure of nutrient status or pro-
ductivity is satisfactory or sufficient, and the results one
obtains depend on which indicators are used. Thus the lim-
nologist is left with the difficult task of subjectively
deciding which indicators to use and which to disregard or
weigh less heavily.

Reviews on trophic state indicators have been published
elsewhere (Fruh et al., 1966; Vollenweider, 1968; Hooper,
1969). Selection of appropriate indicators is a difficult
task, but consideration of the following criteria should
facilitate the decision: a) an indicator should be quanti-
fiable in order to permit numerical differentiation between
lakes of varying trophic states, b) each indicator should be
unique (i.e. not measure the same lake characteristic as
another indicator), c) an indicator should have fundamental
significance in terms of the concept of trophic state (as
a general measure of a lake's nutrient and productivity status),
and d) an indicator should be sensitive to levels of enrich-
ment and relatively simple to measure. The uniqueness of
trophic indicators can be studied by several multivariate
statistical methods, including factor analysis (Shannon, 1969;
Lee, 1971), principal component analysis (Lee, 1971) and
cluster analysis (Goldman et al., 1968; Shannon, 1969). While
different geographical regions may require somewhat different
treatment, indicators should be widespread properties of
aquatic environments in order to insure general interpreta-
bility of the generated classes.

The subjectivity involved in forming logical trophic
classes from conflicting indicator data can be minimized with
certain multivariate techniques such as cluster analysis
(Sokal and Sneath, 1963). Another important classification
problem is the assignment of lakes outside the original sam-
ple group into appropriate pre-established classes. The
method of discriminant function analysis (Shannon, 1970;
Lee, 1971) is useful in this regard.

In order to predict and evaluate the consequences of
watershed management practices on trophic conditions in a
lake, trophic state must somehow be quantified. As discussed
above, this has heretofore been obviated by the multi-
dimensional nature of the trophic concept. Development of
a single numerical index of trophic state from a combination
of several important indicators avoids the misleading and
fragmentary situation arising when only one indicator is used
and the confusion which results when several indicators are
considered individually. An index also allows quantitative
interpretation of trophic state not otherwise feasible. At
least five applications and advantages derive from develop-
ment of a trophic state index: 1) a numerical index would be









valuable in conveying lake quality information to the non-
and semi-technical public; 2) an index would be useful in
comparing overall trophic conditions between lakes; 3) in
the dynamic process of lake succession and trpphic change,
an index would provide a means to evaluate the direction and
rate of changes; 4) an index would facilitate development of
empirical models of trophic conditions as a function of
watershed "enrichment" factors for predictive and manage-
ment purposes; 5) a properly developed index would be highly
relevant to (i.e. identified with) water quality from a human
(or user's) perspective. In contrast to the last point, many
indicators (especially qualitative species composition indi-
cators) are largely of academic or research interest.

On the other hand an index can be criticized as having
no real physical meaning and as improperly combining diverse
parameters (the "can't add apples and oranges" syndrome).
However, the first argument is irrelevant; a relative index
of trophic state, in so far as it reflects the trophic
concept, has value regardless of its interpretability in actual
physical terms. With proper selection of indicators and
rational development of an index, the second criticism can
be largely overcome, but it must be realized that no index
can or should be expected to supply the detailed information
available in the individual parameters.

Proper selection of indicators is a vital consideration
in developing an index of trophic state. Criteria discussed
previously with regard to trophic classification apply equally
here; that the individual indicators be quantifiable is of
course essential. The number of indicators desirable in an
index bears some discussion. Generally an index should include
sufficient indicators to account for the essential attributes
denoted by the broad trophic concept. As fewer variables
are used, the index becomes more unstable, i.e. a large de-
viation from "normal" for a given indicator will tend to af-
fect an index incorporating few variables more than one incor-
porating many. Use of only one variable could result in very
misleading rankings of lake trophicc states." For example
if plankton biomass (expressed as packed cell volume, numbers
per ml, or chlorophyll a) were the sole measure, lakes with
a dense and active macrophyte and periphyton population but
low phytoplankton levels would be misranked as oligotrophic.
Similar criticisms apply to any other single indicator, and
to a lesser extent when only a few indicators are used. How-
ever, redundant indicators (i.e. those that measure essentially
the same phenomenon as another indicator) should be avoided
to prevent biasing the index, i.e. weighing it too heavily
toward that aspect or phenomenon. For example, specific con-
ductance and dissolved solids should not both be used in an
index since they measure nearly the same thing.

The multivariate statistical method of principal component









analysis represents one means of deriving a single numerical
trophic state index from a number of indicators. Given such
an index, empirical models of trophic state as simple func-
tions of nutrient loading rates or other watershed enrich-
ment factors can then be developed by multiple regression
analysis or other appropriate means.


D. COMPOSITION OF THE LAKE STUDY GROUP


Fifty-five lakes from three different areas of north-
central Florida were selected for the study (Figure 2).
Table 3 lists the lakes by name and code number and gives the
surface area and mean depth of each. The study originated in
early 1968 with a survey of 33 lakes within Alachua County,
in which Gainesville and the University of Florida are located.
This group, comprising all accessible and potentially important
recreational lakes in the county, exhibits considerable di-
versity in trophic conditions. Most of the lakes are very
shallow, and moderate to high organic color is common, re-
flecting the large expanses of pine forest in the county.
The small lakes typically have outlets only during periods
of extended rain whereas the large lakes have permanent out-
lets. General physical features of the Alachua County lakes
and initial chemical and biological measurements were summar-
ized by Brezonik et al. (1969); Clark et al. (1962) have
described the geological formations and general land forms
which affect the lakes.

In early 1969 lakes from two important north-central
Florida lake regions outside of Alachua County were included
in the study. Sixteen lakes in the Trail Ridge region of
the Central Highlands (east of Alachua County) comprise one
of these groups. This scrub-oak, sand-hill region is richly
endowed with lakes, most of which are clear and lie within
small drainage basins. Lakes in the Trail Ridge area are
naturally low in nutrients and subject to only light cultural
influence. While still shallow and typically unstratified,
these lakes are generally deeper than lakes in the other
two groups. Anderson-Cue and McCloud Lakes are being used
as model lakes in a separate eutrophication study (Brezonik
and Putnam, 1968; Brezonik et al., 1969). Artificial nut-
rient enrichment of Anderson-Cue Lake has been proceeding
since 1967, and the relevant chemical, biological and physi-
cal characteristics of both lakes have been monitored since
1966.

The final group consists of six lakes in the upper
Oklawaha River Basin northwest of Orlando, Florida. Five of
the Oklawaha lakes are joined by watercourses with the general
pattern of flow being from Lake Apopka through Lake Dora to
Lake Eustis which drains into Lake Griffin. The effluent
















I-75


COUNTY
LAKES


TRAIL RIDGE
LAKES


k


LOCATION '
OF
STUDY AREAS


(' 5T. JOHN'S
RIVER


S OKI.AWAHIIA
SLAAKES


SOR LAA/2
I
.- ,. I- K..A A


0 10 20 350 KM.




Figure 2. Location of 55 Lakes in Study Areas
of North-Central Florida










Table 3. North-Central Florida Lakes
in this Study

Surface
Lake Mean Depth Area
Number Lake Name (meters) (hectares)

(1) ALACHUA COUNTY LAKES

1 Santa Fe* 5.5 1674
2 Little Santa Fe 4.8 467
3 Hickory Pond 3.4 27
4 Altho 3.6 222
5 Cooter Pond 2.2 86
6 Elizabeth 1.5 27
7 Clearwater 1.5 5
8 Hawthorne* 2.8 20
9 Little Orange 2.8 314
10 (Unnamed) Ten* 3.2 29
11 Moss Lee 3.6 52
12 Jeggord 3.0 64
13 Still Pond 1.1 5
14 Lochloosa 2.9 2235
15 Orange* 1.8 3324
16 Palatka Pond 0.8 4
17 Newnan's* 1.5 2433
18 Mize* 4.0 1
19 Calf Pond 1.6 11
20 (Unnamed) Twenty* 1.9 4
21 Meta 1.6 2
22 Alice* .9 29
23 BivinTs Arm* 1.5 58
24 Clear* 1.6 3
25 (Unnamed) Twenty-Five 1.0 6
26 Beville's Pond 3.1 2
27 (Unnamed) Twenty-Seven 3.8 4
28 Kanapaha 0.7 82
29 Watermelon Pond 1.5 213
30 Long Pond 1.2 5
31 Burnt Pond 2.2 22
32 Wauberg* 3.8 101
33 Tuscawilla 1.3 162






(cont'd).








Table 3 (cont'd).


Surface
Lake Mean Depth Area
Number Lake Name (meters) (hectares)

(2) OKLAWAHA RIVER BASIN LAKES

34 Apopka 1.3 12412
35 Dora* 3.0 2237
36 Harris 4.2 5580
37 Eustis 4.1 3015
38 Griffin 2.4 3533
39 Weir* 6.3 2301


(3) TRAIL RIDGE LAKES
40 Kingsley* 7.3 667
41 Sumter-Lowry 4.8 508
42 Magnolia 8.0 83
43 Brooklyn 5.7 253
44 Geneva 4.1 692
45 Swan* 4.8 227
46 Wall 2.1 31
47 Santa Rosa 8.1 42
48 Adaho 3.5 41
49 McCloud* 2.0 6
50 Anderson-Cue* 2.0 5
51 Suggs* 2.5 47
52 Long 3.4 104
53 Winnott 5.2 85
54 Cowpen 3.7 240
55 Gallilee 3.5 34




lakes in 19 lake sub-sample group (see text)









from Lake Griffin forms the Oklawaha River. Lake Harris also
flows into Lake Eustis. Lake Weir, although in the Oklawaha
River basin, does not discharge directly into the Oklawaha
River. All six lakes in this group are important recreational
lakes; in the past Lake Apopka was among the best known bass
fishing lakes in the country. However, considerable cultural
eutrophication (and consequently water quality impairment)
has occurred in the five connected lakes within recent years.
The watersheds of these lakes are utilized primarily for cit-
rus farming, but a large area on the north shore of Lake
Apopka is devoted to vegetable farming of muck soils (recovered
marshland).


CHAPTER 2. EXPERIMENTAL PROCEDURES


A. SAMPLING METHODS


The sampling schedule used in this study was designed to
provide information on the average chemical, biological and
physical characteristics of the 55 lakes over a one-year
period. Systematic sampling of all lakes began in June, 1969,
and all 55 lakes were sampled at four-month intervals up to
June, 1970. In order to obtain greater detail on seasonal
trends, a 19 lake sub-group from the 55 lakes was sampled at
two-month intervals during this same time period. The 19
lakes (denoted by asterisks in Table 3) were selected on the
basis of being representative of the different trophic types
present in the 55 lake group. It was felt that this sub-
group adequately reflected seasonal trends in lake character-
istics without sampling all 55 lakes on a closer time interval.

Water samples taken from the lakes for chemical and bio-
logical analysis were composites. The small lakes (surface
area less than 10 hectares and maximum depth less than 4
meters) were sampled at two stations over depth (surface,
middle, and bottom). These samples were combined into a
composite sample from which aliquots were taken for major
chemical characteristics, for nutrient analyses (preserved
with mercuric chloride), for primary production and chloro-
phyll analysis, and for plankton identification and counts
(preserved with formalin). For the larger lakes that were
relatively shallow (maximum depth <10 meters) the procedure
of sample collection was the same except that three stations
were sampled and composite. For the few deep lakes in which
stable stratification was evident, samples were composite
from the euphotic zone (estimated as twice the Secchi disc
reading) for biological analyses and from the entire water
column for major chemical analyses, and nutrient analyses were
done in profile on uncomposited samples taken at regular depth
intervals. Sediment samples were taken by Ekman dredge from









the deepest region of the lake.


B. PARAMETERS EVALUATED AND
EXPERIMENTAL TECHNIQUES


A total of 6 morphometric, 2 physical, 29 chemical and
6 biological parameters were evaluated for each lake during
the project. In addition 11 parameters were evaluated for
the lake sediments. Six land use and three population char-
acteristics were evaluated for each lake drainage basin.
Table 4 lists all the parameters measured at various times
during the project. The physical parameters were measured
in situ; biological and chemical parameters were determined
on the composite samples using standard limnological procedures
(see Brezonik et al. 1969 for details). Primary production
was measured in the laboratory with a "light box" procedure
rather than in situ in order to standardize light and temper-
ature conditions and offer a more uniform basis of comparison
among the lakes.

Bathymetric maps were available for about 20 of the
lakes (Kenner, 1964); the remainder were sounded and mapped
with a Heath Co. depth sounder as part of the project. Basic
morphometric parameters such as volume, mean depth, volume
development index and shoreline development index were
computed from the bathymetric maps by methods described in
Hutchinson (1957).

Land use patterns in the lake watersheds were determined
by aerial photograph and topographic map interpretation.
Lake watershed areas were outlined and planimetered from
United States Geological Survey (Scale: 1/24,000) topographic
maps. Recent (1965-1968) aerial photographs (Scale: 1" =
1667') were obtained for each watershed from the Florida Soil
Conservation Service Office. Using photogrammetric techniques,
areas of various types of land use patterns were delineated
and measured. Lake surface areas were also determined from
the aerial photographs.

The population in each watershed was characterized in
four categories. Residences on a shoreline were classified
as immediate cultural units (ICU). Other residences within
the lake watershed were categorized as remote cultural units
(RCU), The ICU's and RCU's were evaluated from aerial photo-
graphs. Residences served by sanitary sewer facilities were
not included in the two previous categories. Recent popula-
tion figures were obtained for all of the municipalities
served by sewage treatment plants within each of the lake
watersheds. These figures were converted to equivalent cul-
tural units by dividing by a factor of 2.5, which represents
the average population of a single rural family residence in










Table 4. Lake and Basin Parameters Evaluated
for this Study


Watershed


Land Use


Fertilized cropland
Pastured area
Forested area
Urban area
Unproductive cleared area
Total watershed area


Population Characteristics

Cultural units1 on lake
shore
Cultural units in rest
of basin
Sewage treatment plant
Cultural units


Morphometric


Bathymetric map
Mean depth
Shoreline development


Lake surface area
Maximum depth
Volume development


Physical


Temperature profile
Turbidity


Secchi disc transparency


Chemical


Acidity
Alkalinity
Ammonia
Calcium
Chloride
C.O.D.
Color
Copper
Dissolved oxygen
Fluoride
Iron
Magnesium
Manganese
Mercury
Nitrate
Nitrite


Organic nitrogen
Ortho phosphate
pH
Potassium
Silica
Sodium
Specific conductance
Strontium
Sulfate
Suspended solids
Total phosphate
Total solids
Zinc


Biological


Chlorophyll a
Total carotenoids
Algal identification and counts


Primary production
Algal species diversity


(cont'd).









Table 4 (cont'd).


Sediments


Ammonia
Organic nitrogen
Total phosphate
Sediment type (visual classi-
fication)
Benthic organisms


Volatile solids
C/N ratio
Iron
Manganese
Chlorophyll derivatives
Total carotenoids


ISee text for explanation of this term.









the State of Florida (U.S. Bureau of Census, 1961). Cultural
units of municipalities discharging sewage effluent directly
into a lake were classified as immediate sewage treatment
plant cultural units (ISPU). Cultural units of municipalities
discharging sewage effluent somewhere else in the watershed
were classified as remote sewage treatment plant cultural
units (RSPU). The total cultural units (TCU) in the water-
shed was obtained by summing the cultural units in each of
the four categories. Estimates of total watershed population
could in turn be obtained by multiplying the TCU by 2.5.


C. MULTIVARIATE ANALYTICAL METHODS


Relationships among the several trophic indicators and
watershed eutrophication factors were investigated by a vari-
ety of multivariate statistical techniques. This term is
used to describe statistical methods concerned with analyzing
data collected on several dimensions (variables) on a set of
objects or individuals. Some dependency is assumed among the
variables so that they are considered as a system. Because
of their multi-dimensional nature, these techniques are most
conveniently described using vector and matrix notation.
Theoretical aspects of these techniques are discussed by
Morrison (1967), Sokal and Sneath (1963), and Lee (1971).
The applications and computational aspects of the techniques
used in this study are described below; see Appendix for a
description of the terminology used for vectors, matrices,
and multivariate statistics.

1. Cluster analysis is concerned with the problem of
classifying N objects (e.g. lakes) into groups based on p
variables measured on each object, when the number of groups
that best fit the data is not predetermined. Expressed geo-
metrically, the method attempts to distinguish logical group-
ings of objects in the p-dimensional hyperspace described by
the p data attributes of the objects. Figure 3 illustrates
a simple bivariate cluster problem involving groups formed
by hypothetical data for color and productivity in lakes (cf.
Figure 1). Cluster analysis of objects is referred to as a
Q-type analysis; a second type which clusters the variables
measured on a set of objects is referred to as R-type cluster
analysis. Cluster analysis was used in this study to find
natural groupings of lakes, i.e. those with similar trophic
states or chemical characteristics, as measured by several
limnological parameters (indicators) considered simultaneously
and weighed equally. Cluster analysis progressively combines
a set of objects into a smaller and smaller number of groups
according to the degree of similarity among the objects; ob-
jects (lakes) with the greatest similarity are joined first.

The starting point for any cluster analysis is the N x



















































Primary Production


Hypothetical bivariate plot showing clusters formed
by data for organic color and primary production
in lakes. Solid circles represent clusters formed
around 4 groups with good in-group.similarity:
I. low color, low production; II. low color,
high production; III. high color, low production;
IV. high color, high production. Dashed lines
represent less similar clusters of (A) low color
and (B) high color lakes formed later (at higher
objective function values).


Figure 3.









p raw data matrix X. If it is desired to group objects, the
matrix X is normally transformed to the matrix of standardized
variates Z since the variables may have been measured in quite
different sized units. The standardized data are used to
calculate product-moment correlation coefficients for all
possible pairs of objects. The resultant N x N symmetric
matrix is called the similarity matrix Q, with general element
q1J being the correlation between objects i and j considering
the p variables measured on each object. The Q matrix repre-
sents the starting point of the cluster analysis.

The three basic elements of a cluster analysis are the
between-object distances, the clustering criterion and the
computational procedure (Padron, 1969). A multitude of
methods are available to evaluate the between object distance
(see Sokal and Sneath, 1963, for a review); popular distance
measures include the correlation coefficient between objects
and simple functions of the Euclidean distance. The distance
measure used in this project was proposed by Gower (1966):


d.. = [2(1-q..)]1/2, (1)


where d.. is the distance between the i-th and j-th objects
and q is the correlation coefficient or measure of simi-
larity between objects i and j.

Clustering criteria (a measure of the goodness of any
given allocation o'f objects into groups) usually include a
measure of within group similarity. In some cases, good
within group similarity implies good between group dissimilar-
ity. The clustering criterion used was minimization of an
objective function (OF):


OF = 500 (WBAR-BBAR), (2)


where WBAR is the average within group distance and BBAR the
average between group distance for any given allocation. The
constant is an arbitrary number used to scale the objective
function into a convenient range. Using the distance measure
in Eq.(1), the minimum value of the objective function in
Eq. (2) is -1000, implying complete similarity within groups
and complete dissimilarity between groups, A value of zero
represents a random grouping of the lakes (where the mean
within group and between group distances are equal). Consider-
ation of the OF value for any allocation and its change from
a previous allocation offers a means of determining the rela-
tive degree of similarity between the two groups or objects
joined. Computational procedures are usually heuristic in
the interest of solving large problems with an economy of









computer time. A clustering algorithm in Fortran IV developed
by Padron (1969) was used in the cluster analyses.

2. Discriminant function analysis is a multivariate
classification procedure which can be used to assign objects
into appropriate pre-established classes. Figure 4 illustrates
a simple example involving two groups formed by two variables.
Discriminant functions are linear combinations of variables
for which the separation between groups is a maximum. The
functions contain as many variables as there are dimensions
to the objects. When the population is divided into two
mutually exclusive groups, one discriminant function is suffi-
cient to determine the group to which an object belongs.

Fisher (1936) first formulated the method for the sepa-
ration of two groups of objects. This technique was later
generalized by Anderson (1958) so that linear discriminant
functions could be evaluated for distinguishing between multi-
ple groups.

Let i,T2 .... iT be the m populations under consideration.
In this study the populations, Ti, represented the different
trophic states to which a lake may belong. Associated with
each population are the multivariate probability density func-
tions p1(x), p2(x)... .p(x) (x is an observation vector of p
variablesT. It is desired to divide the space of observations
into m mutually exclusive and exhaustive regions P,,P2....Pm.
If an observation falls into Pi it is assumed to be a member
of population wi. Assume the distribution of fi to be normal
with mean vector pi and covariance matrix Z. The covariance
matrix Z is assumed to be common for all i populations. If
the costs of misclassification are equal and the a priori
probabilities qi of drawing an observation from 7i are known,
the region Pi is defined by those x satisfying

qk
Pi: Pik > loq i-, k=l,2,....,m;kJi, (3)


where y. is the linear discriminant function related to the
ith and kkth populations. The a priori probabilities of x
being in population i or k are given by qi and qk, respectively.
The discriminant function pik is given by


log Pi(x) -
Pik log Pk(x) 2i+ki+Pk


Usually j, jk and Zare not known and xi, xk and S are used
as their estimates (x. is the vector of sample means of the p
variables and S is the sample covariance matrix). The linear

























COLORED LAKES



co
0



H /
XC1

U CLEAR LAKES











.TURBIDITY (X)





DISCRIMINANT FUNCTION, V = aY bX
xy


Figure 4. Hypothetical Two-Dimensional Plot
Showing Relationship Between Discriminant Function
and Two Clusters of Inverse Secchi Disc Transparency
and Turbidity Data.
Clusters represent the envelopes of points (not shown) for
colored and clear (uncolored) lakes. Color decreases Secchi
disc visibility; hence colored-lakes tend toward higher (1/SD)
values for a given turbidity. Bell-shaped curves represent
idealized distribution on a given axis for data points within
each cluster.









discriminant function then becomes vik and is given by:


vik = [x ii+-'(X+ ) (5)
-k2 -- J -i-k

For sufficiently large samples vi. is considered to be a good
estimator of Pik- If the a prior probabilities qk and qi are
equal in Eq. 3, the region Pi is defined for pik>0.

The method used to calculate the linear discriminant
functions in this study was the stepwise procedure (BMDO7M)
described in Biomedical Computer Programs (Dixon, 1968). In
the stepwise procedure variables are brought into the dis-
criminant function one at a time based on an 'F' test for
significance. In essence, the most powerful discriminatory
variables are entered into the discriminant function first
and less important variables at later stages.

3. Principal component analysis is used to examine the
dependence structure of multivariate data and reduce the dimen-
sionality of the data by expressing the original observation
variables in terms of fewer component variables, which are
linear functions of the observation variables. A simple bi-
variate example of principal component analysis is shown in
Figure 5. Principal component analysis was used to derive
indices using the first principal components extracted from
trophic state correlation matrices of trophic indicators
measured on the lakes. When the variables are expressed in
different units, the matrix of sample correlations (R) between
all possible pairs of variables is used as the starting point
in the analysis. If p variables are involved, R is a p x p
symmetric matrix.

The first principal component yi of the correlation mat-
rix R is the linear combination


yi = az, (6)


where a' is the transpose of the first characteristic vector
--1
(eigenvector) of R associated with the largest characteristic
root X (eigenvalue) of R, and z is the vector of standardized
variables. The variance of y, is given by A1. The jt prin-
cipal component yj is given by


yj = a'z, (7)


where a. is the transposed eigenvector associated with the jth
largest~eigenvalue Aj of R.
j
























y! = aX + bX2
First [ i -i. ual
component


/ 2 = cX dX2

Second principal
component


Chlorophyll a (Xi)


'Hypothetical bivariate plot of pri .- production and
chlorophyll data showing relationship of first and
second principal comp-nents to original variables.
First component is defined to pass through long axis
of elliptical sample cluster configuration, giving
maximum variance of the cluster; second component
passes through short axis of sample cluster, giving
maximum variance in that direction.


Fi gure 5,









In principal component analysis the main objective is
to explain as much of the variance in the original observa-
tions as possible with a minimum number of components. The
first principal component is that linear combination of vari-
ables which explains the maximum variance in the original
data; the second principal component is the linear combina-
tion of variables explaining as much of the remaining vari-
ance as possible, and so on. As many component variables
as original variables can be derived, at which point all the
variance is explained, but this subverts the purpose of the
procedure (i.e. reducing the dimensionality of the data).
The proportion of the total variation that any one component
yj explains is given by


j _.j (8)
tr(R) p

where X. is the jth eigenvalue of R and tr(R) is the trace
of R (sam of the diagonal elements). The trace of R is also
equal to p (the number of variables) since each diagonal ele-
ment of R has a value of unity.

Theoretical and computational aspects involved in cal-
culating principal components from covariance or correlation
matrices are presented by Morrison (1967). The BMDX 72 program
from the Biomedical Computer Programs Library (Dixon, 1968)
was used to perform the analyses in this project.

4. Canonical correlation is used to analyze the statisti-
cal relationships between two sets of variables considered in
vector form. In this project canonical analysis was used to
study the relationships between a trophic state vector con-
sisting of seven trophic indicator variables, and a eutrophi-
cation factor vector, consisting of several land use and popu-
lation characteristics of the lake watersheds. The advantage
of canonical correlation over conventional multi-regression
analysis is that the former allows one to study relationships
between two sets of variables without defining any one vari-
able as dependent and without assuming orthogonality (inde-
pendence among the variables). This method determines the
linear combination of the variables within each set which
produces the maximum correlation coefficient between the two
sets. Thus canonical analysis can be used to determine the
dependency structure, i.e. the nature and extent of covaria-
tion, between two sets of variables.

Consider a random vector x composed of observations on
p variables with a covariance matrix 2. This vector x may
be partitioned into two subvectors x_ and x2 with p1 and p
components, respectively. Usually the variables of each
subvector will have some common feature, e.g. let xi consist
of several trophic state indicators for a lake and the vari-








ables x2 be various eutrophication factors that influence
trophic state, For convenience, it is assumed that p, From the population, N independent observation vectors are
drawn and the p x p sample covariance matrix S calculated.
It is assumed that N > (p, + P2 + 1) and S is the unbiased
estimator of Z. The covariance matrix may be partitioned
into submatrices in a manner similar to x where



S = (9)
S12 S22


where the dimensions of S,,, S,2 and S22 are p, x p1, p1 x P2
and P2 x P2, respectively. Once a testing procedure (de-
scribed by Morrison, 1967) indicates a significant dependence
between x, and x2, the method of canonical correlation may
proceed.

In canonical correlation analysis the following question
is proposed. What are the linear compounds


'1 = b x1 .....'i- t = b' x
(10)
v1 = X2, ..... ,v = x
t -t

with the property that the sample correlation of 1, and v,
is greatest, the sample correlation of p, and v2 greatest
among all linear compounds uncorrelated with p1 and vi and
so on for t = min(pl,p2) possible pairs? These pairs of
linear compounds are called canonical variates. It should
be noted that the correlation matrix R could have been par-
titioned in a similar manner to S resulting in similar canon-
ical correlations. However, canonical variates based on the
correlation matrix are dimensionless and are expressed in
terms of the standardized variables. The BMDO6M program
(Dixon, 1968) was used to perform the canonical correlation
analyses,

5. Multiple regression analysis may be described as a
method to predict the value of one variable (Y) from the values
of other variables (X ). Variable Y is assumed to be depen-
dent on the values of the independent variables X.. Strictly
speaking multiple regression analysis is not a method of
multivariate analysis since variates are considered inter-
dependent in the latter, and no single variable can be con-
sidered as the "dependent variable." The general model of
(linear) multiple regression may be written as









Y = bo + bl,X,, + b2X2 + ... bpp X (11)


where Y is the dependent variable, Xi, X2, ... X are indepen-
dent variables, b0 is the intercept value, and bl, b2, ...b
are regression coefficients. The variables may be raw data
values or may be transformed values of the raw data. The
principle value of multiple regression analysis lies in its
predictive capacity (i.e. prediction of Y values from a mea-
sured set of Xi). The technique was used to evaluate statisti-
cal relationships between the trophic state index (TSI) and
eutrophication factor (land use and population) variables.
The BMDO2R program (Dixon, 1968) was used with the zero inter-
cept (i.e. bo O) option. This option was used since it is
desirable to have a situation where the TSI is equal to zero
when all the eutrophication factors are zero. The computer
program is a stepwise multiple regression procedure and adds
the variables to the equation in decreasing order of their
statistical significance (i.e. their partial correlation with
the dependent variable).


CHAPTER 3. LIMNOLOGICAL RESULTS


Detailed descriptions of the morphometry and physical
features of the lakes in the study group are the subject of
another report in this series. Similarly the chemical and
biological limnology of the lakes will be described in detail
in a third report (Brezonik, in preparation). This chapter
will describe the limnological results in general terms as
background information for analysis of eutrophication factors
and lake trophic conditions in the following chapters.


A. MORPHOMETRIC AND PHYSICAL FEATURES


The geology of Florida is dominated by a limestone sub-
stratum underlying the entire peninsula. In north-central
Florida the upper limestone deposits are of Eocene to Miocene
age and are covered by more recent deposits of sand and clay.
Thickness of the overlying formations ranges from a meter or
so (e.g. in southern Alachua County) to over 30 meters. The
limestone deposits give rise to a karstic topography through-
out the peninsula with artesian springs, sink holes and solu-
tion lakes as prominent features of the landscape.

The morphometry and physical features of Florida lakes
are to a large extent determined by the geological structure
and resulting topography. Table 5 summarizes these features
for the 19 lakes sampled bimonthly. In general the lakes are
shallow, and maximum depths of more than 10 m are uncommon.










Table 5. Morphometric Features
of Selected Florida Lakes


LAKE SURFACE MAXIMUM MEAN D a D b
AREA DEPTH (zm) DEPTH (E)
(hectares) (m) (m)

Santa Fe 1674 8.8 5.5 1.88 1.24
Hawthorne 20.4 4.3 2.8 1.95 1.10
#10 29.3 4.6 3.2 2.09 1.10
Orange 3324 3.0 1.8 1.80 1.63
Newnan's 2433 4.0 1.5 1.13 1.20
Mize 0.86 25.3 4.5 0.53 1.19
#20 3.7 3.4 1.9 1.68 1.28
Alice 28.6 1.5 0.9 1.80 1.66
Bivin's Arm 58.4 1.9 1.5 2.37 1.48
Clear 3.4 2.7 1.6 1.77 1.40
Wauberg 101 5.2 3.8 2.19 1.13
Dora 2237 4.9 3.0 1.84 2.24
Weir 2301 9.8 6.3 1.93 1.70
Kingsley 667 22.9 7.3 0.96 1.01
Geneva 692 8.8 4.1 1.40 1.58
Swan 227 9.4 4.8 1.53 1.37
McCloud 5.6 3.7 2.0 1.62 1.17
Anderson-Cue 4.5 4.6 2.0 1.30 2.12
Suggs 47.2 3.7 2.5 2.03 1.22

aDevelopment of volume index = 37/Zm.

bDevelopment of shoreline index = L/2/T-7.









Mean depths for all 55 lakes range from about 0.7 to 8.1 m,
and maximum depths range from about 1.0 to 25 m. Most of
the shallow lakes have U-shaped basins; i.e. the lake basin
walls are concave toward the water. The deeper lakes gener-
ally are more cone-shaped; in the deepest lake of the survey,
Lake Mize, the lake basin walls are considerably convex to-
ward the water. The trend can be seen by examining the volume
development indices (D ) in Table 5. Index values less than
1.0 indicate a convex Ytoward the water) lake basin while
values greater than 1.0 are indicative of U-shaped basins.
Lakes with a DV of 1.0 have a basin similar in form to that
of a cone (Hutchinson, 1957; Zafar, 1959).

Many small Florida lakes are hydraulically perched; i.e.
their connection to groundwater is with a perched water table
located above and not directly connected to the principal
aquifer in the peninsula, the Floridan aquifer. Most of the
small Alachua County and Trail Ridge region lakes are seepage
(Birge and Juday,1934) with no visible outlets or permanent
inlets, and water levels may vary as much as several meters
between dry and wet periods. Thus few lakes have a definite
land-lake interface, and the shorelines may be intermittently
submerged land. Water levels in the larger drainage lakes
(e.g. Newnan's, Orange and Lochloosa Lakes, Alachua County)
frequently are structurally controlled so that water level
variations are much smaller. Some of the Trail Ridge lakes
(e.g. Kingsley, Swan, Brooklyn), because of their occurrence
in a region of very sandy soil, do possess fine natural sandy
beaches in spite of the periodically wide fluctuations in
water levels.

Nearly all the natural lakes in Florida have been derived
or substantially modified by limestone solution processes.
Numerous lakes are situated in sink-hole depressions formed
by dissolution of underlying limestone (Stubbs, 1940;
Hutchinson, 1957). In some cases lake basins have originated
by other mechanisms (e.g. fluviatile action) but solution
activity has substantially modified the original basin (e.g.
Lake Tsala Apopka in Citrus County; Cooke, 1939). Many
small and some larger lakes are simple dolines which tend to
have simple circular basins. Perhaps the best example is
Kingsley Lake (Clay County), an almost perfectly circular
basin (shoreline development index, SD=1.01) about 3 km in
diameter. Lake Santa Rosa (SD=1.09), a lake 0.8 km in diameter
in Putnam County, is another example. S is defined as the
ratio of the actual length of a lake's shoreline to the min-
imum length (i.e. the circumference of a circle) which would
enclose an area equal to that of the lake surface. Other
lakes are complex dolines with more irregular shorelines.
For example, Lake Brooklyn (Clay County) consists of at least
9 separate solution basins and has an SD=2.37, and Cowpen
Lake (Putnam County) with an SD=1.80 consists of at least
5 basins.









The shallowness of Florida lakes suggests that thermal
stratification would be unimportant in these lakes, and in-
deed most lakes do not exhibit classical Birgean thermoclines
with stagnant hypolimnia as is common in terinc-'erate lakes.
Eight lakes are sufficiently deep to develop stable strati-
fication and oxygen deficient bottom waters; these are Lakes
Mize (Brezonik and Keirn, in press), Kingsley, Magnolia,
Moss Lee, Santa Rosa, unnamed lakes numbered 20 and 27, and
Beville's Pond. Climatic circumstances favor a long period
of stratification; for example Lake Mize is stratified from
February or early March till November. The surprising fea-
ture of some of the lakes is the shallowness at which stable
thermal stratification can occur. Lake No. 20 is only about
4 m deep but the temperature in the bottom meter is several
OC cooler than the minimum temperatures in the region during
summer. Lake No. 27 is only about 7 m deep, yet it has a
pronounced thermocline between 2.7 and 4.2 m (9-12 ft), and
the bottom water was 11.4C in June, 1969, which is only 1C
warmer than the mid-winter bottom temperature. Clearly morpho-
metric factors are important in producing the thermal stability
of these lakes. Both are fairly small (1.5-4.5 ha), are in
a rolling terrain and are surrounded by high pine forest.
Thermal stratification is not limited to the summer months;
temporary stratification can develop as a result of the highly
changeable weather that occurs during January and February.
While none of the lakes are meromictic, low to zero, dissolved
oxygen values in the bottom waters of Beville's Pond,
Lake No. 27, and Lake Mize throughout the year indicate the
bottom waters circulate rather incompletely even during winter.

At least 6 other lakes among the 55 exhibit incipient
thermal stratification. Typical of these are Lake Wauberg
and Hickory Pond. Stratification develops only near the
bottom in these shallow lakes, preventing the formation of
a distinct hypl.I'mlion, but the bottom water temperatures
during summer are at least as cool as the nocturnal minima
in the region so that fairly stable conditions can be assumed.
Low dissolved oxygen values in the bottom waters of these
lakes also imply stable stratification. These lakes are some-
what larger or less wind protected by forest than the small
lakes discussed previously. Size is obviously an important
factor in determining whether stratification will occur in
a lake. For example, neither Lake Santa Fe (surface area =
1650 ha, Z = 8.8 m) nor Lake Weir (surface area = 2300 ha,
Zm = 9.8 m ) have shown any evidence of stratification on
any sampling date.

Many other shallow lakes show signs of stratified con-
ditions even in the absence of a typical thermocline. Temper-
ature differences of 4-5C from top to bottom in lakes that
are only 2-4 m deep are common during summer, but the decline
is continuous with depth rather than confined to a narrow
layer (also see Yount, 1961). At surface temperatures of









25-300C, temperature differences of a few degrees are suffi-
cient to impart considerable stability to the water column
(Hutchinson, 1957). Bottom temperatures are greater than
regional nocturnal air temperatures during summer, and strati-
fication thus is not highly stable. However, oxygen deple-
tion in the bottom waters of several lakes implies a metas-
table circumstance (i.e. mixing is not a daily phenomenon).


B. GENERAL CHEMICAL CHARACTERISTICS


In order to determine general patterns in chemical compo-
sition among the lakes (i.e. classify the lakes into distinct
chemical types), a cluster analysis was performed on data
for six basic chemical parameters for the 55 lakes. The para-
meters considered were pH, alkalinity, acidity, conductivity,
color and calcium, and mean values for each lake over the
sampling period were used for the analysis. The resulting
cluster diagram is shown in Figure 6.

The 55 lakes fall into four easily interpreted groups:
(i) acid colored lakes, (ii) alkaline colored lakes, (iii)
alkaline (hardwater) clear lakes; and (iv) soft, clear lakes.
A comparison of the six chemical characteristics in these
4 lake types is shown in Table 6. Assuming the 55 lakes are
a reasonable cross-section of the lakes in north-central
Florida, several conclusions derive from the results in
Figure 6. Slightly less than 50 percent of the lakes are
classified as colored, and the bulk of these are also acidic.
Thus color would appear to be a common feature of Florida
lakes. However, all but three of the colored lakes lie in
Alachua County, which fact both implies a rather heterogene-
ous geography in the region and suggests that caution should
be observed in extrapolating the statistics of the sample
group to the population of Florida lakes.

Several other regional differences can be noted. The
alkaline-colored group is composed entirely of lakes from
Alachua County. Three (Newnan's, Orange, Lochloosa) are
large connected drainage lakes; the other two are seepage
or semi-drainage. All five lakes are moderately enriched.
The alkaline clear group includes the five culturally en-
riched lakes of the Oklawaha chain plus the small eutrophic
lakes of Alachua County. The soft water clear lakes are
located primarily in the Trail Ridge region and eastern
Alachua County, which geographically comprise one topographic
unit.

One conclusion that seems a valid extrapolation is that
Florida lakes generally have soft water; only the 12 alkaline
clear lakes can be considered to exhibit hardness, and even
here the degree is moderate. This may seem contradictory in





VALUE OF OBJECTIVE FUNCTION
-650 -600 -550 -500 -450

ADAHO
SUGGS -
WALL
LONG POND
JEGGORD
LIT SANTA FE
ELIZABETH
CALF POND
#10 LE____
MOSS LEE
BURNT POND ACIDIC
TUSCNAWILLA
ALTHO COLORED
LIT ORANGE LAKES
MIZE
BEVILLE'S
PALATKA
#27 ---
HICKORY
LOCHLOOSA-
ORANGE
WAUBERG ALKALINE
COOTER COLORED
NEWNAN'S LAKES
EUSTIS
GRIFFIN ----
DORA
HARRIS
# 20
APOPKA ALKALINE
HAWTHORNE CLEAR LAKES
BIVIN'S ARM
ALICE
KANAPAHA
# 25 ----
CLEAR
SANTA ROSA
COWPEN
SAND HILL
STILL POND
WINNOT
KINGSLEY
GENEVA
META
CLEARWATER
McCLOUD
MAGNOLIA
ANDERSON-CUE
LONG LAKE SOFT WATEI
WATERMELON CLEAR LAKE
BROOKLYN LEA LAE
GALLILEE
SWAN
WEIR
SANTA FE
Figure 6. Clustering Diagram of Fifty-Five Florida
Lakes Considering Six Chemical Characteristics












Table 6. Selected Chemical Characteristics
of Four General Lake Types


Characteristics


Colored-
Acidic


Colored-
Alkaline


Clear-
Alkaline


Clear-Soft
Water


pH 5.66a 7.63 8.38 5.83


Acidity 7.31b 1.18 1.00 2.00
(mg/l as CaCO,) 6.64c .47 1.27 .96


Alkalinity 2.36 11.69 92.14 2.80
(mg/l as CaCO3) 3.37 6.04 39.91 6.42


Conductivity 45.8 70.0 249 48.2
(pmho/cm) 10.1 11.9 123 25.6


Color 220 114 60 17
(mg/l as Pt) 121 45 30 19


Calcium 3.3 6.9 36.8 3.0
(mg/1) 1.6 1.5 16.2 2.2


adenotes median value

denotes mean value

Cdenotes the standard deviation









view of the solution origin of the lakes and the abundance
of hard water springs in Florida, but few Florida lakes are
spring fed. Rather, most of the lakes receive the bulk of
their water either directly from precipitation or by surface
and subsurface runoff from the sandy, low calcareous soils.
In fact several of the hard water lakes are not naturally
calcareous but have hard water because of cultural effects,
i.e. the influx of ground water as treated sewage or septic
tank drainage.

The mean and median values of the chemical parameters
in Table 6 indicate highly distinct and readily apparent
differences among the 4 lake types, perhaps much greater
than when the lakes are considered individually (as the
large standard deviations for some parameters would suggest).
The acidic-colored lake group has a much higher mean color
than the alkaline-colored group (220 to 114 mg/l as Pt), and
the high color probably contributes to the low pH values.
Color concentrations as high as 700 mg/l have been found in
some lakes (e.g. Lake Mize). Color certainly contributes
to acidity (cf. acidity values of the acidic-colored and
clear-soft water groups, both of which have acid pH values).
Color is the only parameter which has a significantly differ-
ent value in each of the 4 types and as such appears to be
an important chemical characteristic for distinguishing
between the lake types.


C. PHYTOPLANKTON AND MACROPHYTE CHARACTERISTICS


Algal identification and enumeration was done on all
55 lakes at each sampling. Because of year-round favorable
growth conditions (solar radiation and temperature), some
of the fertile lakes such as Apopka, Bivin's Arm and Dora
exhibit virtually continuous algal blooms. However maximum
bloom conditions usually obtain during summer. Lake Apopka
has exhibited phytoplankton blooms of 88,000 cells/ml or
higher, predominated by blue-green genera such as Lyngbya
and Microcystis and green genera such as Pediastrum and
Scenedesmus. Blooms of 32,000 cells/ml or more have been
found in Lake Dora. Newnan's Lake, a colored eutrophic lake,
has summer populations predominated by blue-green algae
(Microcystis, Anabaena, Spirulina). In winter this lake usually
produces an extremely dense bloom of Aphanizomenon, which
fixes nitrogen at high rates (Brezonik and Keirn, unpublished
data), However this alga is not present in the lake during
other seasons of the year and is not a common constituent
of the phytoplankton in other eutr.ophic lakes. Microcystis
and Anabaena are the summer bloom former in Bivin's Arm.
The latter organism is found in all lakes in which nitrogen
fixation has been detected, and seems to be the primary algal
agent for this process in all the lakes except Newnan's Lake.









Oligotrophic lakes have typically low algal populations.
For example, in Swan Lake (a clear soft water, oligotrophic
type) a summer 1969 population of about 36 organisms/ml was
dominated by the diatoms Synedra and Navicula and the green
alga Sphaerocystis. Dinobryon and Synura (class Chrysophyceae)
are common in the low pH, low ionic strength waters of the
soft water clear (oligotrophic) lakes as are a variety of
Desmidaceae (e.g. Staurastrum, Closterium, and Cosmarium).

Diatoms are comparatively rare in the plankton of Florida
lakes, especially in the oligotrophic soft water lakes. Low
silica concentrations in Florida lakes may in part account
for this distribution. An exception to this general trend
is Lake Apopka, which normally supports a high (although not
usually dominant) population of diatoms, including Melosira,
Tabellaria, and Navicula, and perhaps not coincidentally has
one of the highest silica concentrations (3.7 ppm) of the
55 lakes. Bivin's Arm with a mean silica content of 1.8 ppm
also supports a spring bloom of diatoms (Maslin, 1970).

The dominant primary producers in a number of the 55
lakes are floating macrophytes. For example, Lake Alice,
on the University of Florida campus was until recently covered
almost entirely by a dense crop of water hyacinth (Eichornia
crassipes). While a faculty-student effort succeeded in
mechanically clearing this lake (at least temporarily), the
plant is common in canals and other lakes (e.g. Lakes Tuscawilla
and Apopka). Chemical spraying is used to control the plant
in a number of lakes including Bivin's Arm and Lake Apopka.
Duckweed (Lemna minor) partially covers the surface of Lake
No. 27 throughout the year, while perhaps one-third of
Beville's Pond is covered by Salvinia during the summer months,
Such growths limit light penetration, drastically reducing
phytoplankton populations, and under severe conditions may
inhibit oxygen transfer from the atmosphere to the water.


D. SEDIMENTS


Florida lakes have a wide variety of sediment types,
including sand, peat, and sludge-like (ooze) deposits. In
some of the oligotrophic lakes a light nearly pure sand
bottom occupies most of the lake bottom, suggesting the geo-
logical newness of these lakes. Organic deposits in the lakes
range in color from light brown (peat) to nearly black (ooze)
and the sediment consistency similarly covers a wide range
with large fragments of plant remains evident in peat sedi-
ments and very fine, slowly settling particles in some of the
oozes. In many of the lakes there is no defined sediment-
water interface, Rather a gradation from thin suspensions
of sediment to more compact strata occurs often over depths
of a meter or more. This characteristic makes sampling of








bottom water (and surface sediments) rather difficult. In
shallow lakes the suspended sediments undoubtedly become mixed
with the overlying water during periods of wind stress, and
considerable nutrient exchange is thus effected. The carbon:
nitrogen ratios in nearly all the sediments are greater than
10 indicating a "dy!" type of sediment in Hansen's (1962) ter-
minology. A crude correlation also exists between C/N ratio
and trophic conditions. The most eutrophic lakes have C/N
ratios in the range 10-15 and oligotrophic lakes have gener-
ally higher ratios, but considerable scatter occurs when all
55 lake sediments are considered.


CHAPTER 4. CLASSIFICATION AND QUANTIFICATION OF
TROPHIC CONDITIONS IN FLORIDA LAKES


As discussed in Chapter 1, eutrophication and trophic
state are extremely complex, multivariable phenomena. At
present our understanding of them and their interrelationships
is primarily qualitative. A broad effort to quantify these
relationships was made using the statistical techniques de-
scribed in Chapter 2 and the collected limnological and water-
shed data. The analyses were applied to three major aspects
of eutrophication research listed in 1) the long standing
problem of rational classification of lakes according to trophic
state, 2) quantification of the presently nebulous term
trophicc state," and 3) delineating the relationships between
lake trophic conditions and watershed enrichment factors.
This chapter presents results for the first two aspects;
Chapter 5 discusses the third.


A. DEVELOPMENT OF A TROPHIC CLASSIFICATION
SYSTEM FOR FLORIDA LAKES


The multi-dimensionality of the trophic state concept
has heretofore obviated objective and consistent classifica-
tion of lakes according to their trophic states. In an attempt
to minimize subjectivity in delineating trophic classifica-
tions for Florida lakes, similarity (cluster) analyses were
performed on trophic indicator data from the 55 lakes. Seven
indicators; viz., primary production (PP), chlorophyll a
(CHA), total organic nitrogen (TON), total phosphorus (TP),
Secchi disc transparency (SD), conductivity (COND), and a
cation ratio (CR) due to Pearsall (1922) were chosen as the
dimensions describing the hybrid concept of trophic state
and were considered simultaneously in the cluster analysis
to derive logical lake groups according to their trophic
states (at least as defined by the 7 indicators).

The main considerations in selecting the first six








indicators are.that (i) they are quantitative, (ii) they are
fundamentally significant as measures of trophic state, (iii)
they satisfy Hooper's (1969) criteria for useful trophic
indicators reasonably well, and (iv) they applyto Florida
lakes. The first six indicators have all been used with some
degree of success in various lake classification schemes.

The selection of Pearsall's cation ratio (Na + K) was
a somewhat subjective attempt to incorporate Mg + Ca
information on the major cations into the concept of trophic
state without adding each cation as an individual indicator.
Pearsall (1922) reported that English lakes with high nitrate
and silica and a Na + K ratio less than 1.5 had periodic
algal blooms. Mg + Ca Thus this ratio should be inversely
related to increasing eutrophy. Parenthetically it might be
noted that many workers have suggested a general correlation
between high productivity and water hardness (Ca and Mg con-
centrations). This ratio has not been used to any extent in
other investigations, but it was suggested as a potentially
effective parameter for differentiation between lake trophic
types by Zafar (1959). For Florida lakes the cation ratio
appears to be a reasonably good indicator of trophic state
with high values of the inverse cation ratio being indicative
of eutrophic conditions.

Averages of the lake parameters over the one year sampling
period would seem the most appropriate values for the purposes
of statistical analysis. In some respects extreme values
(e.g. maximum nutrient concentrations, algal densities at
the height of bloom conditions, etc.) are more critical deter-
minants of a lake's water quality and may thus be better and
more sensitive indicators of trophic state. But extreme values
are less reproducible, and their magnitude depends greatly on
the vagaries of sampling frequency and climatic circumstances.
Since the breadth of this project precluded detailed (e.g.
weekly) sampling, it is felt that mean values are more appropri-
ate in the ensuing analysis. In order not to bias the means
toward summer conditions, the June, 1970, values were not
included in the computations. Means of the trophic indicators,
color, and turbidity for the 55 lakes are listed in Table 7.
So that each indicator would denote trophic state in a positive
sense (an increase in indicator value denotes an increase in
trophic state) the Secchi disc and cation ratio indicators
were inversely transformed. Obviously there are many more
possible indicators of trophic state that could be included.
Alternatively, it may be that fewer trophic indicators will
eventually prove sufficient to describe the concept of trophic
state. The selection of 7 indicators was a somewhat arbitrary
attempt to incorporate as much information into the concept
of trophic state as possible without getting into a prolifer-
ation of secondary or redundant indicators.

Because of the basic typological differences caused by












Trophic State Indicator, Color and Turbidity Data,


Color-
Lake Scaled
Number 1/SD 1/SD Cond TON TP PP CHA 1/CR COL TUR


1
2
3
4
5
6
7
8
9
10
YO 11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29


0.43
0.66
0.56
0.73
1.16
1.66
0.41
1.41
1.06
0.65
0.70
1.23
0.50
1.15
1.00
0.81
1.91
1.83
1.41
2.87
0.45
0.67
1.65
1.33
0.66
0.46
0.71
2.81
1.04


0.52
0.46
0.52
0.58
0.92
0.89
0.39
0.91
0.54
0.53
0.53
0.77
0.47
0.81
0.74
0.44
0.88
0.47
0.42
2.88
0.49
0.46
1.79
1.17
0.57
0.39
1.37
2.27
1.08


53.2
53.7
45.7
53.3
59.7
47.7
40.0
167.7
50.7
50.4
43.0
55.7
38.0
87.0
77.4
25.0
59.8
53.0
43.7
314.3
93.3
552.2
253.8
136.4
92.7
39.7
47.0
121.7
37.7


0.50
0.61
0.70
0.59
1.26
0.81
1.32
1.86
0.94
0.86
0.77
0.47
0.63
1.42
1.07
1.22
1.41
0.85
1.41
2.06
0.81
0.50
1.88
1.27
0.73
0.65
0.58
2.20
0.86


0.021
0.015
0.027
0.023
0.165
0.036
0.012
0.079
0.105
0.064
0.036
0.087
0.013
0.058
0.063
0.024


9.3
1.8
4.1
12.4
29.6
6.9
0.5
96.6
20.8
18.2
25.4
6.8
0.5
20.9
27.4
8.5


0.110 102.5
0.113 11.2
0.184 12.7
0.410 262.8
0.030 2.4
0.900 7.9
0.546 251.7
0.392 87.4
0.028 2.6
0.087 1.8
0.325 1.2
0.422 153.7
0.052 17.0


5.6
4.5
7.6
5.9
22.6
8.0
2.3
56.8
9.8
12.6
8.1
7.0
3.1
23.3
15.6
15.6
47.4
33.9
23.5
92.8
3.3
4.4
56.00
26.4
3.5
23.7
30.1
42.7
9,0


1.24
0.90
0.64
0.65
1.01
0.85
0.53
1.88
0.55
0.47
0.49
0.39
0.61
1.41
1.01
0.78
0.85
0.62
0.83
4.04
1.50
3.42
2.48
2.91
9.88
0.54
1.21
5.12
0.73


59.0
149.0
62.0
133.7
83.3
236.7
21.3
58.3
165.7
123.7
98.3
192.7
26.0
116.3
107.1
93.3
188.9
433.4
404.0
68.5
25.0
25.5
42.1
85.4
36.0
181.7
92.0
120.7
74.0


1.9
1.5
1.9
2.3
4.5
4.3
1.0
4.4
2.0
1.9
1.9
3.5
1.5
3.8
3.3
1.3
4.2
1.5
1.2
17.4
1.7
1.5
10.2
6.1
2.2
1.0
7.5
13.8
5.6


(cont'd.)


Table 7.











Table 7 (cont'd.)


Color
Lake Scaled
Number 1/SD 1/SD Cond TON TP PP CHA 1/CR COL TUR


30 1.01 0.61 31.0 1.00 0.052 1.4 7.6 0.77 253.5 2.4
31 1.64 0.89 63.0 1.69 0.478 86.8 29.0 1.63 351.7 4.3
32 1.13 0.88 66.2 1.67 0.169 103.8 37.3 1.23 74.8 4.2
33 1.04 0.62 51.8 1.19 0.292 20.2 8.5 1.50 434.5 2.5
34 4.54 4.39 314.7 4.45 0.380 337.7 60.4 3.85 78.0 27.3
35 2.66 3.13 313.0 3.33 0.384 310.7 50.4 3.45 96.4 19.0
36 0.94 0.68 210.0 1.18 0.037 30.1 14.5 3.68 38.7 2.9
37 1.31 1.65 251.7 2.22 0.167 92.8 23.8 3.34 47.0 9.3
38 1.50 1.91 255.3 2.63 0.183 218.3 47.3 3.12 36.0 11.0
39 0.51 0.53 135.8 0.82 0.019 12.3 6.5 0.51 8.8 1.9
o 40 0.21 0.39 52.8 0.35 0.011 2.9 1.8 1.15 10.6 1.0
41 0.30 0.45 28.0 0.19 0.011 0.8 1.3 0.63 7.7 1.4
42 0.27 0.40 26.0 0.18 0.012 1.0 1.5 0.64 10.7 1.1
43 0.23 0.45 30.3 0.28 0.011 0.8 1.9 0.81 9.3 1.4
44 0.32 0.50 49.2 0.35 0.016 2.6 1.5 0.61 9.8 1.7
45 0.31 0.45 44.3 0.27 0.011 3.5 1.6 0.51 5.7 1.4
46 0.73 0.37 42.0 0.67 0.025 6.8 5.1 0.47 151.0 0.9
47 0.16 0.37 37.0 0.19 0.011 0.6 1.7 0.48 2.0 0.9
48 1.16 0.37 37.7 0.72 0.027 7.1 5.3 0.60 336.0 0.9
49 0.21 0.41 34.8 0.30 0.017 0.9 2.4 0.59 4.1 1.1
50 0.27 0.39 38.2 0.29 0.018 1.1 3.2 0.63 3.0 1.0
51 1.09 0.48 46.2 0.69 0.036 6.7 3.4 0.58 280.7 1.6
52 0.24 0.40 52.0 0.09 0.012 0.3 1.4 0.61 10.0 1.1
53 0.30 0.54 41.3 0.56 0.023 1.3 2,6 0.61 23.3 2.0
54 0.21 0.46 45.7 0.25 0.010 0.4 1.6 0.59 5.0 1.5
55 0.29 0.43 38.0 0.36 0.014 1.9 1.9 0.65 12.3 1.3

11/SD (Secchi disc transparency)-' in m-l; Cond in mmho cm-"; TON and TP in mg N or
P/1; PP in g C/m3-hr; CHA in mg/m3; 1/CR (cation ratio)-' dimensionless; COL (color)
in mg/l as Pt; TUR (turbidity) in Jackson Turbidity Units. See text for explanation
of column 3 (color scaled 1/SD).









organic color (Figure 6 and Table 6), it seemed best to con-
sider clear and colored lakes as separate classes in each of
which a range of trophic subclasses could exist. In fact,
a cluster analysis of the 55 lakes considering the 7 trophic
indicators plus color divided the lakes into essentially the
same 31 clear and 24 colored lakes shown in Figure 6. (Lakes
Wauberg and Kanapaha, which are in the alkaline colored and
alkaline clear groups, respectively, in Figure 6, are the
only lakes which fall into the opposite groups in the color
plus trophic indicator cluster analysis.) The lakes in the
colored group had mean color levels greater than about 75 ppm,
whereas the clear lakes had color levels less than this value.
Thus the horizontal line separating clear and colored lakes
in Figure 1 would appear to have a value of about 75 ppm for
Florida lakes.

The lakes within each of the main classes were grouped
into subclasses of similar trophic state by performing cluster
analyses with the 7 trophic indicators. The clear lakes (Figure
7) formed three apparently natural groups which can be inter-
preted in the classical (oligotrophic-mesotrophic-eutrophic)
sense. Nearly all the lakes of the Trail Ridge region comprise
the oligotrophic Group A. These demonstrate a good within-
group similarity as denoted by the low objective function
values at which they are joined. The mesotrophic Group B
includes a few lakes from the Trail Ridge region (e.g. Kingsley
and Winnott) which have been subjected to some cultural in-
fluence. As a whole the lakes in Group B (especially those
from the Trail Ridge) are perhaps closer to being oligotrophic
than eutrophic, but nonetheless they are distinctly (if
slightly) more productive than the Group A lakes. The lakes
of the eutrophic Group C include the 5 Oklawaha chain lakes
plus the small eutrophic and hypereutrophic lakes in Alachua
County. The latter are located primarily in urbanized areas,
especially around Gainesville, and cultural sources seem to
exert a heavy influence on their trophic states. The meso-
trophic and eutrophic groups exhibit greater diversity in
values for the individual trophic indicators and consequently
are joined at higher objective function values.

The colored lakes exhibited considerable diversity, and
the results were not as interpretable in terms of classical
trophic groupings. Perhaps this reflects a basic difference
between colored and clear lakes; in this regard it should be
noted that the position of dystrophic lakes in the usual
trophic (i.e. nutritional state) classification has long been
a subject of contention (Hansen, 1962; see Brezonik et al.,
1969 for further discussion). As previously mentioned, we
have considered color (roughly equivalent to dystrophy) as
a major lake type parallel to a clear lake type (as proposed
by Hansen, 1962, and Stewart and Rohlich,1967), and a range
of nutrient states was considered possible for both. However
the results of the cluster analysis imply that a simple har-








VALUE OF THE 02.JECTIVE FUNCTION
-600 -500 -400 =300 -200
i -r- --i-----r--- t -~

(41) SA.D HILL

(42) MAGNOLIA
(55) GALLILEE

(43) BROOKLYN

(47) SANTA ROSA

(54) COWPEN

(52) LONG LAKE
(50) Mc CLOUD GROUP A
L_-

(49) ANDERSON-CUE

(44) GENEVA
(45) SWAN

(7) CLEARWATER
(53) WINNOTT
(13) STILL POND

(29) WATERMELON
GROUP B
(1) SANTA FE GROUP B
(40) KINGSLEY
(21) META

(39) WEIR

(25) No. 25
(36) HARRIS

(37) EUSTIS

(34) APOPKA
(35) DORA
GROUP C
(38) GRIFFIN
(8) HAWTHORNE

(20) No. 20

(32) WAUBERG
(23) BIVINS ARM

(24) CLEAR

(22) ALICE

Figure 7. Cluster Analysis of 31 Low-Color
(Clear) Lakes Considering 7 Trophic Indicators









monious oligo to eutrophic gradation may not occur in highly
colored lakes. Depending on where the vertical line is drawn
through the colored lake cluster diagram (Figure 8) one can
obtain classifications containing anywhere from 2 to 6 or
more groups. However none of these systems are completely
satisfactory with regard to interpretability of the groups.

Similarity cut-off line A in Figure 8 delineates a 5-
group classification for the purpose of the present discus-
sion; this system gives good within group similarity for groups
1, 2 and 3, and moderate within group similarity for group 4.
Group 5 would appear to be a residual group whose lakes are
similar only to the extent that they are different from the
other groups. Lake Kanapaha is the most dissimilar of the
colored lakes since it was the last lake to be incorporated
into a group, and a seven group classification could also be
drawn which would leave this lake in a group by itself. The
five groups can be interpreted and labeled as follows:
1. oligotrophic, 2. meso-eutrophic, 3. oligo-mesotrophic,
4. dystrophic, 5. residual. Group 4 is labeled dystrophic
because the lakes in this group are moderately to highly
acidic and have high organic color and low dissolved solids.
However pH was not one of the indicator variables, and dys-
trophy is not a lake type parallel to oligotrophy and eutrophy.
Several of the lakes in this group are very shallow (mean
depths of .1-1.5 meters) and are partially covered with
emergent and floating macrophytes (e.g. water hyacinths).
These lakes (Palatka Pond and Tuscawilla, for example) could
more accurately be described as senescent (bordering on ex-
tinction), but again this is not a recognized trophic state
comparable to oligo- and eutrophy. The remaining lakes
(Group 5) would appear to be a residual group whose members
(except for Beville's Pond and Lake No. 27, which are in fact
similar) are alike only in being different from the other
groups. Apparently there were not enough pairs or groups
of lakes of nearly adjoining trophic characteristics to form
groups with good within-group similarity.

At a higher objective function value (i.e. lower degree
of similarity) (line B in Figure 8), three groups can be
drawn: 1. oligotrophic, 2. mesotrophic, 3. eutrophic-dystrophic.
In this scheme Lakes Lochloosa and Orange would appear mis-
classified and some of the dystrophic (i.e. low pH, high
color) lakes like Palatka and Calf Ponds are classified with
obviously eutrophic lakes like Newnan's Lake in spite of the
low productivities and algal standing crop in the former.
The latter apparent misclassifications result from low Secchi
disc transparencies (caused partly by high color) and in
some cases from fairly high nitrogen and phosphorus levels,
which, because of high color and low pH, do not produce algal
blooms and high productivity. At a still lower level of
similarity (line C in Figure 8) two colored lake groups can
be formed; 1. oligo-mesotrophic and 2. eutrophic-dystrophic.










VALUE OF THE OBJECTIVE FUNCTION
-500 -400 -300 -200 -100


(2) LIT. SANTA FE

(4) ALTHO

(3) HICKORY POND

(46) WALL LAKE

(10) No. 10

(11) MOSS LEE .

(15) ORANGE -2

(14) LOCHLOOSA

(6) ELIZABETH

(48) ADAHO

(51) SUGGS _3 1.__

(9) LIT. ORANGE

(12) JEGGORD

(31) BURNT

(33) TUSCAWILLA

(5) COOTER ------

(19) CALF

(16) PALATKA
-- ----- 2
(30) LONG POND 3

(26) BEVILLE'S

(27) No. 27

(17) NEWNANS
A5
(18) MIZE

(28) KANAPAHA



A B C

Figure 8. Cluster Analysis of 24 Colored Lakes
Considering 7 Trophic Indicators
44









This is not a very useful classification since the groups
then contain highly dissimilar lakes and are not easily
interpreted in terms of classical trophic groups, It is
apparent that none of the colored lake classifications is
ideal. The five group classification is not readily interpret-
able in terms of classical trophic groups, and the two group
system has groups that are too broad to be of much use. The
three group classification has some advantages in terms of
interpreting classical trophic states, but some obvious mis-
classifications occur in this system.

Mean values and standard deviations of the trophic indi-
cators within the 3 clear lake subgroups and 5 colored lake
subgroups delineated by the cluster analyses (Figures 7 and
8) are presented in Table 8. All seven indicators appear
to reflect trophic levels reasonably well. Among the clear
lakes indicator mean values without exception increase in
each succeeding trophic group (from oligotrophy to eutrophy).
Among the colored lakes the same trend is noted although some
exceptions occur. There is little difference between mean
values for the colored oligotrophic (Group 1) and oligo -
mesotrophic (Group 3) groups; primary production and chloro-
phyll means are actually somewhat higher in the former than
in the latter. The high values for the residual group derive
from the hypereutrophic conditions in Newnan's and Kanapaha
Lakes; the other lakes in this group have varied indicator
values.

It is interesting to note that the colored lakes have
a much smaller range of values for most of the indicators
compared to the clear lakes. Thus the clear oligotrophic
lakes reflect much greater nutrient impoverishment than the
colored oligotrophic lakes, and similarly the apparent degree
of eutrophy is greater in the clear lakes. For example, the
range of mean primary production values in the clear lakes
is 1.3 to 150 mg C/m3-hr, while in the colored lakes the range
is 9.7 to 55.


B. DEVELOPiiEiIT OF DISCRIMINANT FU:i:'TIONS
TO CLASSIFY LAKES OUTSIDE THE ORIGINAL
SAMPLE GROUP


Discriminant functions were derived for the three trophic
classes delineated by cluster analysis of the 31 low-color
lakes and are presented in Table 9. In addition, the 55 lakes
were grouped into five trophic categories ranging from ultra-
oligotrophic to hypereutrophic based on the trophic state
index described in the next section. Discriminant functions
were derived for these classes and are shown in Table 10.
The colored lake group was too small and diverse to form
meaningful discriminant functions. Using the criterion de-










Table 8. Mean Values and Standard Deviations of Trophic State
Indicators Within Trophic State Groups

Trophic State Indicators


Primary Total Total Inverse
Group Production Chlorophyll a Phosphate Organic Secchi Conductivity Cation
(mg Cm3-hr) (mg/m3) (mg-P/1) Nitrogen Disc (pmho/cm) Ratio
(mg-N/1) (m-1)

(a) Clear Lakes
A. Oligotrophic 1.3a 1.8 .013 .25 .25 38.5 .61
1.0b .5 .003 .08 .05 8.5 .08
B. Mesotrophic 5.8 4.3 .023 .73 .47 61.5 .86
6.3 2.5 .014 .30 .24 35.1 .38
C. Eutrophic 150.2 39.5 .306 1.98 1.72 244.0 3.61
119.5 26.3 .251 1.10 1.12 129.0 2.13

(b) Colored Lakes
1. Oligotrophic 11.4 7.3 .032 .70 .67 48.0 .60
9.0 3.0 .017 .10 .06 5.1 .17
2. Meso-
Eutrophic 24.1 19.5 .060 1.24 1,07 82.2 1.21
4,6 5.5 .003 .25 .11 6.8 .28
3. Oligo-
Mesotrophic 9.7 6.7 .058 .72 1.24 47.6 .60
6.2 2.5 .035 ,17 .25 6.6 .16
4. Dystrophic 31.6 21.1 .213 1.36 1.59 46.0 1.36
230.0 24.5 .278 1,45 2.41 49.0 1.57
5. Residual 55.1 35.6 .211 1.13 1.54 64.2 1.67
71.9 9.5 .152 .68 .96 33.0 1.95


adenotes mean value


denotes standarddeviation











Table 9. Discriminant Functions
for Trophic Groups Delineated in Table 8a



V b 16(1/SD) + .27(PP) + .04 (CCfli' 62(TP) 34(TON)
AB
4.4(1/CR) .74(CHA) + 14.3


VAC 65(1/SD) + .77(PP) + .10(COND) 230(TP) 108(TON)

20(1/CR) 2.9(CHA) + 126


VBC = 49(1/SD) + .51(PP) + .06(COND) 167(TP) 75(TON)

16(1/CR) 2.2(CHA) + 112




alndicator abbreviations: 1/SD = inverse Secchi disc,
PP = primary production, COND = specific conductance,
TP = total phosphate, TON = total organic nitrogen,
1/CR = inverse of Pearsall's cation ratio, CHA =
chlorophyll a, all indicators in units given in
Table 8.
bSubscripts of discriminant functions indicate the
groups (from Table 8) being compared.








Table 10. Discriminant Functions for Five Groups
of Lakes Determined from Trophic State Indicator
Ranges (From Table 18)

Vb = H 23.99(1/SD)a + .47(COND) + 15.17(TON) + 148.76(TP4)
E + .38(PP) .04(CHA) + 8.19(1/CR) 289.10

VHM = -22.44(1/SD) + .53(COND) + 35.93(TON) + 289.93(TP4)
+ 91(PP) + ,77(CHA) + 11.11(1/CR) 369.94
VH = -34.88(1/SD) + .59(COND) + 74.56(TON) + 427.32(TP4)
+ .59(PP) + 2.45(CHA) + 23.39(1/CR) 451.87

VHU = -44.33(1/SD) + .64(COND) + 111.06(TON) + 490.95(TP4)
+ .33(PP) + 3.27(CHA) + 26.56(1/CR) 475.42
VE = 46,43(1/SD) + .o6(COND) + 20.77(TON) + 141.17(TP4)
+ .53(PP) + .81(CHA) + 2.92(1/CR) 80.85

VEO = -58.87(1/SD) + .12(COND) + 59.40(TON) + 278.56(TP4)
+ .21(PP) + 2.49(CHA) + 15.19(1/CR) 162.78

VEU = -68.32(1/SD) + .17(COND) + 95.90(TON) + 342.20(TP4)
-.05(PP) + 3.31(CHA) + 18.37(1/CR) 186.32
VM = -12.44(1/SD) + .07(COND) + 38.63(TON) + 137.39(TP4)
-.32(PP) + 1.68(CHA) + 12.28(1/CR) 81.93
VMU = -21.88(1/SD) + .12(COND) + 75.13(TON) + 201.03(TP4)
-.58(PP) + 2.50(CHA) + 15.45(1/CR) 105.48

V = -9.45(1/SD) + .05(COND) + 36.50(TON) + 63.64(TP4)
OU -.26(PP) + .82(CHA) + 3.18(1/CR) 23.55


aKey to indicator abbreviations identical to
Table 9
bSubscripts of discriminant functions refer to
groups labeled in Table 18: hypereutrophic (H),
eutrophic(E), mesotrophic (M), oligotrophic (0),
ultraoligotrophic (U).









scribed in Chapter 2, a lake belongs to Group A (oligotrophic)
if VAB and V, the respective discriminant functions between
the subscripted groups in Table 9, are both greater than or
equal to zero.

To demonstrate the application of this technique to lake
classification, trophic indicator data for three well known
North American lakes (Table 11) were assembled from various
sources, and the lakes were classified according to the discrim-
inant functions of Tables 9 and 10.

Data for Lake Tahoe was taken from Ludwig et al. (1964)
and Goldman and Armstrong (1968). Data for the two great
lakes was obtained from Saunders (1964), Putnam et al. (1966)
and Beeton (1969). As expected, Lakes Tahoe and Superior
were assigned to the oligotrophic class and Lake Erie to the
eutrophic class using the discriminant functions for the clear
lakes (Table 9). The discriminant functions of Table 10 de-
rived from all 55 lakes classified Lake Tahoe in the ultra-
oligotrophic group (U), Lake Superior with the oligotrophic
lakes (0), and Lake Erie in the mesotrophic group (M).

It should be emphasized that use of these three lakes
is for illustrative purposes only. The validity of assigning
large temperate lakes into classes delineated from a sample
of small sub-tropical lakes has not been tested. Certainly
the general effects of eutrophication are similar in all
"normal" lakes, and in this sense the examples are not in-
appropriate. However, if geographically broad (or universal)
trophic groups are to be delineated, the original sample
group should be similarly broadly based, which of course the
Florida lakes used to develop the discriminant functions are
not. A further word of caution regarding this method is the
deleterious effect of small sample size on the probability
of misclassification (Wallis, 1967). For good differentiating
power the functions should be based on sample groups of 50
or more.


C. FORMULATION OF TROPHIC STATE INDICES


The multivariate statistical method of principal component
analysis (Chapter 2) represents one means of deriving a single
numerical index from a number of indicators, and this technique
was used to derive indices from the trophic indicator data
for the 55 lakes.

As seen in Figure 6, organic color can be used to separate
lakes into the two fundamentally different classes of colored
and clear lakes. Nutrient enrichment may cause different ef-
fects in each class, i.e., various trophic indicators may
respond differently to nutrient enrichment in clear vs. colored












Table 11. Trophic Characteristics and Classification of
Three Well-Known North American Lakes by Discriminant Functions
in Tables 9 and 10a


1/SD PP COND TP TON CHA Trophic Trophic
Lake m-1 mg C/m3-hr pmho/cm mg P/1 mg N/1 1/CR mg/m3 Class Class
(Table 9) (Table 10)


Tahoe


Superior .10


\l
o Erie


.29


0.5


8.0


59


83


79


313


.007


.060


.09


.48


1.4 1.5 A (oligo-
trophic)

5.1 2.5 A (oligo-
trophic)

4.7 27.5 E (eu-
trophic)


U (ultraoligo-
trophic)

0 (oligotrophic)


M (mesotrophic)


adi abbreviations as in Table 9.
Indicator abbreviations as in Table 9.









lakes, and a single trophic index for all lakes could possibly
be inappropriate. Consequently separate trophic state indices
were developed from the correlative relationships of indicators
within each of the two basic classes defined previously by
cluster analysis. The annual mean values for each lake (Table
7) were used in the derivation of the indices. Table 12 lists
the means and standard deviations of each indicator for all
24 colored lakes and all 31 clear lakes, and Table 13 presents
the respective correlation matrices. The first principal
components, yco and ycl, extracted from the colored and clear
lake correlation matrices, respectively, are shown in Table
14. The first principal components extract a good portion
of the information from the R's since yco and ycl explain 72
and 71% of the total variances in their respective correlative
matrices.

The principal components are simple linear functions of
the 7 trophic state indicators with weighting factors for
each indicator. The indicator values are standardized values
(i.e. the actual raw value from Table 7 minus its mean value
and divided by its standard deviation from Table 12). The
trophic state index for each group of lakes, i.e. TSIco and
TSI,1, was derived by slightly modifying the respective first
principal components. The modification consisted in adding
a constant value to the principal component so that the TSI
would always be greater than zero. The constant was obtained
by evaluating the first principal component with raw data
values of zero for each indicator. A zero raw data value
results in a negative standardized value and hence a negative
value for y, which value was then added to y to produce the
TSI (see Table 14). Hence a hypothetical lake with zero pro-
ductivity and zero values for the other indicators would then
have a TSI of zero. (In actuality this would never occur
since even pure water has a finite Secchi disc transparency,
and all natural waters have a non-zero cation ratio.) In
general, lakes with increasingly positive indicator values
will exhibit correspondingly higher TSI's.

The TSI's of the lakes in each group were calculated by
substituting the standardized indicator values into the appropri-
ate TSI formula, and Table 15 presents the results for each
group ranked in descending order of TSI value. Thus the above
analysis indicates that Lake Kanapaha is the most eutrophic
colored lake and Wall Lake is the least eutrophic in this
group; similarly Lake Apopka (eutrophic) and Lake Santa Rosa
(oligotrophic) represent the extremes of trophic state within
the clear lake group.

The results of the cluster analyses (Figures 7 and 8)
are also included in Table 15 for comparative purposes. Rank-
ings of the clear lakes according to their TSI's are in ex-
cellent agreement with the clear lake groups formed by cluster
analysis. The first 12 lakes (in order of decreasing TSI)












Table 12.


Means and Standard Deviations of Trophic Indicators
Within the Colored and Clear Lake Groupsa


Group


1/SD


Colored Lakes


.53c


Clear Lakes


COND


53.2

20.2


TON


.99

.42


.88b 124.0 1.04

.97c 126.1 1.04


aSee Table 7 and text for explanation of symbols and units of expression.


bDenotes the mean.


CDenotes the standard deviation.


TP


.119

.130


.129

.209


PP


25.0

37.8


60.1

102.7


CHA


16.7

12.6


17.0

24.2


1/CR


.99

.94


1.83

1.94












Table 13. Correlation Matrices of Seven Trophic
Indicators for Colored and Clear Lake Groups



(a) Colored Lakes:


1/SD


COND


TON


1.000 .630 .720
1.000 .627
1.000


TP

.534
.484
.643
1.000


PP CHA


.782
.733
.818
.640
1,000


.646
.517
.658
.596
.705
1.000


(b) Clear Lakes:


1.000 .643 .931
1.000 .621
1.000


.559
.888
.481
1.000


.962
.638
.915
.586
1.000


.858
.603
.813
.543
.910
1.000


1/SD
COND
TON
TP
PP
CHA
1/CR


1/CR

.697
.792
.764
.685
.800
.529
1.000


1/SD
COND
TON
TP
PP
CHA
1/CR


.464
.522
.442
.396
.402
.392
1.000











Table 14. First Principal Components (y and y c)
and Trophic State Indices (TSIco and TSIcl)




(a) Colored Lakes:

yco = .848(1/SD) + .809(COND) + .887(TON) + .768(TP)
+ .930(PP) + .780(CHA) + .893(1/CR)


Cumulative Percent of Total Variance Explained
by ycol = 72%

TSIcol = Ycol + 933


(b) Clear Lakes:


Ycl


= .936(1/SD) + .827(COND) + .907(TON) + .748(TP)


+ .938(PP) + .892(CHA) + .579(1/CR)


Cumulative Percent of Total Variance Explained
by ycl = 71%

TSIcl cl + 4










Table 15. Lakes of Clear and Colored Groups
Ranked According to TSIcl and TSIco


a. Clear Lakes


Lake


TSIcl


Cluster
Group


Lake


TSIcl


Apopka
Twenty
Dora
Bivin's Arm
Griffin
Alice
Eustis
Hawthorne
Clear
Wauberg
Harris
Twenty-five
Watermelon
Weir
Meta
Clearwater

b. Colored


Kanapaha
Burnt
Newnan's
Lochloosa
Cooter
Calf Pond
Mize
Tuscawilla
Orange
Twenty-seven
Little Orange
Elizabeth


18.1
15.1
14.6
12.0
10.7
9.2
8.2
7.9
7.3
6.3
5.3
5.1
2.9
2.7
2.4
2.1


Lakes
TSIco


27.9
17.0
15.3
12.0
11.0
10.6
10.5
10.4
9.9
9.2
8.0
7.9


Santa Fe
Still Pond
Winnott
Kingsley
Geneva
Gallilee
Swan
Anderson-Cue
McCloud
Brooklyn
Cowpen
Long
Sumter-Lowry
Magnolia
Santa Rosa


1.9
1.6
1.4
1.3
1.2
1.2
1.1
1.1
1.0
1.0
1.0
0.9
0.9
0.9
0.8


TSIco

Ten 6.9
Palatka Pond 6.9
Jeggord 6.7
Moss Lee 6.3
Beville's Pond 6.2
Suggs 6.2
Adaho 6.1
Long Pond 6.1
Altho 6.0
Little Santa Fe 5.8
Hickory Pond 5.6
Wall 5.3


Cluster
Group









correspond to the lakes in eutrophic group C of Figure 7;
the next 8 lakes are in mesotrophic groupB, and the last 11
lakes comprise oligotrophic group A. Thus the TSI for clear
lakes can be used to separate classical trophic states
quantitatively. A TSIcl of about 5.0 would appear to be the
dividing line between mesotrophy and eutrophy, and a value
of about 1.2-1.3 separates mesotrophy and oligotrophy. Qual-
itative inspection of other trophic indicators for Lakes
Kingsley and Winnott suggests these lakes are more typically
oligotrophic than mesotrophic and the TSI dividing line should
perhaps be raised to 1.5. The colored lakes ranked according
to TSIco are in general agreement with the cluster analysis
(Figure 8), but some discrepancies are noted. For example,
Lakes Lochloosa and Orange have a high degree of similarity;
however, the lakes have high but somewhat dissimilar TSI
values, and four lakes have TSI rankings between the values
for the two lakes. Also Beville's, Palatka and Long Ponds
were clustered into eutrophic groups although their TSI
values indicate oligotrophy. The discrepancies in comparing
the two analyses probably arise within the cluster analyses
since the colored lakes exhibited considerable diversity and
did not form groups with good within-group similarity.

For management and identification purposes it would be
desirable to have a single trophic state index to rank all
lakes regardless of color. Large differences in the specific
conductance, primary production and cation ratio mean values
for the two groups (Table 8) and the cluster analysis of basic
chemical parameters (Figure 6) suggest a basic difference
which could possibly cause different trophic indicator responses
in the two types. On the other hand, that the two groups can
be viewed as two samples of one (larger and more diverse) popu-
lation, and a single TSI to rank all 55 lakes was developed
under this assumption.

Of the seven indicators used to assess trophic state in
this study, the one most directly affected by organic color
is Secchi disc transparency. This parameter is essentially
a function of color and turbidity, and a multiple regression
of inverse Secchi disc reading as dependent variable vs.color
and turbidity as independent variables produced the following
relationship;

1/SD = 0.003(Col) + 0.152(Tur) (12)

Data for the analysis were from Table 7, and the zero inter-
cept option was used in the regression analysis. The relation-
ship is significant at the 99% confidence level, and the per-
cent of variation in 1/SD explained by Eq. 12 is 96%. Using
Eq. 12, a color value of 75 mg/l, and turbidity values from
Table 7, new color-scaled inverse Secchi disc values were
calculated for each of the 55 lakes; the results are listed
in Table 7. A color value of 75 mg/l was chosen for the








scaling purposes because it represents the dividing line
between clear and colored lakes and is also in the middle
range (zone of best prediction) of the regression equation.

Once the Secchi disc values had been color scaled, the
correlative relationships between the seven trophic indicators
for all 55 lakes were subjected to a principal component
analysis, and the means, standard deviations and the correla-
tion matrix are given in Table 16. The first principal
component yt extracted from R is given by


yt = .919(1/SD) + .800(COND) + .896(TON) + .738(TP)

+ .942(PP) + .862(CHA) + .634(1/CR) (13)

yt extracts a good portion of the information from R and
explains 70% of the total variation in R. The TSI is given
by

TSI = yt + 5.19, (14)

where the value of 5.19 was determined as described previously
in the derivations of TSIco and TSIcl.

TSI's were calculated for each of the 55 lakes by sub-
stituting the standardized indicator values (computed from
Tables 7 and 16) into Eqs. 13 and 14. The lakes are ranked
in descending order of TSI in Table 17. Using the cluster
analyses of Figures 7 and 8 as a guide, the 55 lakes were
separated in terms of classical trophic state terminology
into five groups as follows: 1. Hyper-eutrophic (TSI>10),
2. Eutrophic (102TSI7), 3, Mesotrophic (7>TSI3), 4. Oligo-
trophic (3>TSI2), 5. Ultra-oligotrophic (TSI<2). These groups
are delineated and labeled in Table 17. The relative rank-
ings of the lakes in the TSI and TSI formulations of
Table 15 are also shown in Table 17. Comparison shows that
the clear lakes are ranked in almost identical order accord-
ing to the total (55 lake) TSI (excluding the interspersed
colored lakes) as they are by the TSIcl. Further it is ob-
vious that the clear lakes as a group are more extreme in
their trophic behavior than are the colored lakes; all but
one of the hypereutrophic lakes and all the ultraoligotrophic
lakes in Table 17 belong to the clear lake group. Nearly all
the colored lakes are included in the oligotrophic and meso-
trophic categories. Comparison of the colored lake rankings
according to the 55 lake TSI and the TSI also indicates a
general correspondence. The lake most out of order is Lake
Twenty-seven, which is the fourth listed colored lake in
Table 17 and the tenth ranked lake according to the TSIco.
Many of the other colored lakes are "misranked" by one or two
places, but there are no major discrepancies. Most of the
changes in relative rankings between Tables 15 and 17 probably













Table 16. Means, Standard Deviations, and Correlation
Matrix of Trophic State Indicators for 55 Lakes


Standard


Indicator


Mean

.84
93.1
1.02
.125
44.8
16.9
1.47


Deviation

.77
101.3
.82
.177
82.3
19.8
1.63


Correlation Matrix R;


1/SD COND TON

1.000 .617 .880
1.000 .582
1.000


1/SD
COND
TON
TP
PP
CHA
1/CR


1/SD
COND
TON
TP
PP
CHA
1/CR


TP

.542
.762
.500
1.000


PP

.927
.654
.890
.576
1.000


CHA

.784
.540
.788
.553
.859
1.000


1/CR

.502
.560
.474
.440
.478
.402
1.000












Table 17. Fifty-five Florida Lakes Ranked According
to Trophic State Index (TSI)


Rank in Rank in
Lake TSI Table 151 Lake TSI Table 151


1. Hypereutrophic group

Apopka
Twenty
Dora
Bivin's Arm
Griffin
Kanapaha
Alice
Eustis

2. Eutrophic group


Hawthorne
Clear
Burnt Pond
Wauberg
Newnan's


22.1
18.5
18.5
14.7
13.7
13.5
10.7
10.5


9.1
8.8
8.3
7.4
7.1


9A
2B
10A
3B


3. Mesotrophic group


Twenty-five
Harris
Twenty-seven


6.4
6.3
5.8


12A
11A
10B


Cooter Pond
Lochloosa
Tuscawilla
Calf Pond
Orange
Mize
Watermelon Pond
Little Orange
Weir
Elizabeth
Ten
Palatka Pond
Beville's Pond
Meta


4. Oligotrophic group

Jeggord
Moss Lee
Long Pond
Clearwater
Altho
Hickory Pond
Santa Fe


(cont'd.)


5.3
5.2
4.8
4.6
4.3
4.2
3.6
3.4
3.3
3.2
3.2
3.2
3.1
3.1



2.8
2.8
2.8
2.6
2.5
2.5
2.5


5B
4B
8B
6B
9B
7B
13A
11B
14A
12B
13B
14B
17B
15A


15B
16B
20B
16A
21B
23B
17A












Table 17 (cont'd.)


Rank in
Lake TSI Table 151


Suggs 2.3 18B
Little Santa Fe 2.3 22B
Adaho 2.2 21B
Wall 2.1 24B
Winnott 2.0 19A

5. Ultra-oligotrophic group

Still Pond 1.9 18A
Kingsley 1.9 20A
Geneva 1.8 21A
Gallilee 1.6 22A
Swan 1.5 23A
Anderson-Cue 1.5 24A
McCloud 1.5 25A
Brooklyn 1.5 26A
Cowpen 1.5 27A
Long 1.3 28A
Sumter-Lowry 1.3 29A
Magnolia 1.3 30A
Santa Rosa 1.3 31A


IRank from Table 15 according to TSIc1 (A values) and TSIo (B values).









result from the use of color-corrected Secchi disc trans-
parencies for the TSI values in Table 17, which presumably
should produce a more accurate relative ranking of the lakes
according to their trophic states.

The question concerning the soundness of one TSI for
both clear and colored lakes remains. A definitive answer
is perhaps impossible. However, the first principal component
on which the 55 lake TSI is based accounts for about as much
of the variance (70%) in the correlation matrix of trophic
indicators for all lakes as do the first principal components
for the clear and color groups, which accounted for 71 and 72%
of the variances in their respective correlation matrices.
Further, there appear to be no obvious misclassifications
or misrankings in the 55 lake TSI's. One of the major values
of the TSI concept is the possibility of ranking rather diverse
objects (lakes) in a logical and objective manner. Obviously
if the sample is too diverse, the rankings will have little
or no meaning. Thus extrapolation of the TSI concept to
development of a single, universal index for all lakes is
not suggested. To rank Arctic bogs, acid volcanic lakes,
tropical ponds and the Great Lakes on the same scale would
be pointless and meaningless. On the other hand, the more
"harmonious" the sample, the more meaningful and logical
(and easier) it will be to rank the objects. Th- relatively
harmonious series of clear lakes is easily and logically ranked
(Table 15). Inclusion of the colored lakes produces a more
diverse sample with an inevitable loss in clarity in inter-
pretation of the resulting TSI. Nevertheless, it is felt
that the 55 lake TSI is a useful, interpretable and logical
means of ranking Florida lakes.

Some interesting features of the TSI rankings deserve
mention. Lake Alice has been ranked in the hypereutrophic
group although it might be classified oligotrophic on the
basis of plankton productivity alone. Lake Alice has extremely
high nitrogen and phosphorus concentrations and supports a
profuse growth of water hyacinths, which along with a short
hydraulic detention time (in the order of 2-3 days), have
restricted plankton productivity. In this case, the other
trophic state indicators (nitrogen, phosphorus, and conduc-
tivity) have been sufficiently high to counteract the low
primary production and chlorophyll a values. Lake Twenty-
seven was also ranked higher than it would be on the basis of
plankton production alone. This lake is almost completely
covered with duckweed (Lemna minor), and as with Lake Alice
the other indicators have counteracted the low primary pro-
duction value.

The usefulness of the trophic state index can be best
determined by its application, e.g. in practical (e.g. manage-
ment and control) situations or in development of empirical
models relating trophic state to watershed enrichment factors









(see the next section). However, the validity ofthe approach
can be inferred from closer inspection of the TSI and its
component parameters. Table 18 presents the means and 95%
confidence intervals for the 7 trophic indicators in each
of the classes delineated in Table 17. In nearly every case
the mean parameter values increase in progressing toward more
eutrophic classes. However the large confidence intervals
for most parameters implies considerable overlap between the
classes delineated by any single indicator. These facts demon-
strate three important points. First, because of the over-
lap, any single parameter is inadequate to define trophic
state or trophic classes. Second, the wide and overlapping
ranges of indicator values preclude easy placement of lakes
into appropriate trophic classes since the values for a lake
could fit within the confidence intervals of the parameters
in two adjacent classes. Finally, the increasing mean values
in progressing toward eutrophic conditions imply that the
TSI provides at least an objective means of placing lakes
into appropriate trophic classes and suggests that the rela-
tive ranking of the lakes by their TSI values is reasonable.

The TSI described above reflects the general trophic
conditions of Florida lakes; whether it is the best index
that can be developed will have to be answered by further
work comparing its attributes with those of other indices
that might be developed. The seven indicators in the present
index reflect the major limnological consequences of eutro-
phication with the exception of macrophyte problems. Indices
with fewer variables would reflect a narrower concept of
trophic state and would be more likely to yield misleading
results.

Specific water quality problems resulting from eutrophi-
cation are not directly considered by the index, but some of
the indicators are indirectly related to such problems. For
example, chlorophyll a, a biomass parameter, might be correlated
with taste and odor problems arising from algal blooms; Secchi
disc transparency is associated with water turbidity, which
should be correlated with the length of sand filter runs in
water treatment plants. Perhaps other indices could be
developed which would be directly related to water quality
problems, but it is not always a simple matter to find ap-
propriate quantitative indicators for such purposes.

The index described above should be practical for rou-
tine assessment of general trophic conditions since the indi-
vidual parameters are commonly and rather simply measured.
The only exception possibly is primary production. This
parameter, while of fundamental significance to the trophic
state concept, also suffers from the fact that measured values
are highly variable in a given lake and are greatly dependent
on physical factors such as light and temperature. Perhaps
a simpler TSI not incorporating this parameter would prove












Table 18. Confidence Intervals for Trophic Indicators in
Five Lake Groups Delineated by Trophic State Index Valuesa


Parameter

TSI Range


Number of Lakes

Primary Produc-
tion
mg C/m3-hr

Chlorophyll a
mg/m3

o Total Phosphate
Smg P/1

Total Organic
Nitrogen
mg N/1

(Secchi Disc)-l
m-1

Specific Conduc-
tivity
-mho cm-1

[Ca] + [Mg]
[Na] + [K]


Ultraoligotrophic Oligotrophic


1.3-1.9


1.3 .7


1.9 .4


0.13 + .002



.29 + .08


.43 .02



39.6 5.3

.65 .10


2.0-2.9

12


8.6 3.3


7.7 2.3


.040 .012



.78 .10


.55 .08



50.6 + 11.5

.69 .13


Mesotrophic


3.0-6.9

17


17.3 8.5


19.5 7.5


.141 .085



1.08 .23


.73 .22



80.2 39.8

2.35 2.27


Eutrophic


7.0-9.9

5


95.4 10


39.4 15.8


.246 .221



1.58 .29


.94 .16



98.6 62.2

1.70 .97


Hypereutrophy

>10.0(10.0-22.1)

8


205 94


42.7 21.7


.424 .192



2.41 .96


2.31 .98



297 101

3.60 .64


values represent means 95% confidence interval.









more useful to governmental agencies faced with evaluating
the trophic characteristics of large numbers of lakes.


CHAPTER 5. RELATIONSHIPS BETWEEN TROPHIC STATE
AND WATERSHED ENRICHMENT FACTORS


A. INTRODUCTION


Empirical relationships between lacustrine trophic condi-
tions and watershed conditions can be developed by regression
analysis using the TSI as dependent variable and appropriate
conditions in the watershed as independent variables. A
general model for eutrophication can be written as:


TS = (N,M,H,S,t...), (15)


where TS is the trophic state resulting from nutrient (N)
loading (nitrogen, phosphorus and other essential nutrients),
M represents morphometric characteristics such as mean depth,
H represents h' ld--logical conditions (e.g. water detention
time), S is a sedimentation factor, and t is time. The re-
lationships among these parameters is presently too vague for
the development of functional relationships. However, simpli-
fied empirical approximations of Eq. 15 can be developed.

As a first approach models of the type


TSI = g(N,P) + C, (16)


were developed, where the TSI described in Eq. 14 represents
the trophic state parameter of Eq. 15, N and P represent annual
nitrogen and phosphorus loading rates, and C is an uncertainty
term. Although nitrogen and phosphorus are not the only nutri-
ents required for algal growth, it is generally agreed that
they are the two main nutrients involved in the lake eutrophi-
cation process. In spite of current controversy over the
role of carbon (Bowen, 1970; Legge and Dingeldein, 1970; Kerr
et al., 1970), researchers as a whole regard phosphorus as
the most frequent limiting nutrient in lakes. Vollenweider
(1968) and others have emphasized the importance of nutrient
(particularly nitrogen and phosphorus) supply in determining
a lake's trophic state. Although various lake factors, such
as mean depth, detention time, basin shape, and sedimentation
rate, affect the amounts of nutrients a lake can assimilate,
nutrient budget calculations represent a first step in quan-
tifying this dependence.









A few lacustrine nitrogen and phosphorus budgets have
been reported in the literature, e.g. Rohlich and Lea (1949)
for Lake Mendota, McGauhey et al. (1963) for Lake Tahoe and
Edmondson (1968) for Lake Washington. Vollenweider (1968,
1969) has summarized most of the budget calculations for
American European Lakes. Comprehensive evaluation of the
nutrient balance for a lake requires measurement of all poten-
tial nutrient sources and sinks (Table 19) over an extended
period in order to assess seasonal and other effects. Some
sources and sinks, e.g. groundwater, nitrogen fixation and
denitrification, require elaborate sampling and experimental
procedures to be adequately evaluated. Consequently, man-
power and time constraints have resulted in very few complete
nutrient balances being attempted. An alternative and simpler
method is to use literature estimates for nutrient exports
from various sources and information on the various land use
and population characteristics of the lake watershed. This
approach was used by Lee et al. (1966) for nitrogen and
phosphorus budget calculations for Lake Mendota. While per-
haps not as accurate as actual measurement, there is no other
realistic alternative when evaluating budgets for a large
number of small lakes.


B. NIrTI,,:I:Eli AND PHOSPHORUS BUE.DET.i


Partial nitrogen and phosphorus budgets for the 55 lakes
in the study group have been computed by this latter approach.
The budgets are referred to as partial since no attempt was
made to account for such sources as nitrogen fixation, leaves
and pollen and groundwater. Adequate data were not available
to evaluate most sinks, and consequently none were considered.
The partial budget calculations therefore estimate gross
supply or loading.

The morphometric, land use, and population figures for
each lake were determined according to the methods described
in Chapter 2. Watersheds were divided into forest, urban,
pasture, fertilized cropland, and cleared unproductive areas.
Table 20 lists the pertinent watershed and morphometric data
for each lake.

Literature figures for the expected contributions of
nitrogen and phosphorus from the various sources were compiled,
and the values used in this study are summarized in Table 21.
Where applicable, each value is accompanied by the literature
reference. Literature estimates were not available for two
sources. Muck (recovered marshland) and citrus farm contribu-
tions were calculated from average fertilizer composition and
application rates, assuming that 10 percent of the applied
nitrogen and one percent of the applied phosphorus was ex-
ported from the soil to the lake. The figures for percentage













Table 19, Potential Nitrogen and Phosphorus
Sources and Sinks for Lakes


(a) Sources


Natural

Precipitation on Lake Surface

Swamp Runoff

Virginal Meadowland Runoff

Forest Runoff

Soil Erosion

Aquatic Bird and Animal Wastes

Leaf and Pollen Deposition

Groundwater Influxes

Nitrogen Fixation

Sediment Recycling


(b) Sinks

Outlet Losses

Fish Catches

Aquatic Plant Removal


Cultural

Domestic and Industrial
Waste Waters

Agricultural Runoff

Managed Forest Runoff

Urban Runoff

Septic Tanks

Landfill Drainage











Denitrification

Volatilization

Ground Water Recharge

Sediment Losses


*Applies to nitrogen alone.









Table 20. Population and Land Use Data
for 55 Florida Lake Watershedsi


Unproductive Population
Mean Forested Urban Fertilized Pastured Cleared Immediate Remote Served
Depth Area Area Cropland Area Area Cultural Cultural by STPa
Name/No. (m.) (ha.) (ha.) (ha.) (ha.) (ha.) Units Units Facilities

Santa Fe 1 5.5 4424 191 60.6 206 137 209 91 0
Lit. Santa Fe 2 4.8 842 0 61.3 109 72.8 34 24 0
Hickory 3 3.4 95.5 0 29.3 119 0 0 6 0
Altho 4 3.6 666 73.8 17.1 21 21 13 58 0
Cooter 5 2.2 487 0 0 627 0 0 19 0
Elizabeth 6 1.5 156 0 0 5.2 0 6 3 0
Clearwater 7 1.5 18.1 0 0 0 0 0 0 0
Hawthorne 8 2.8 53.6 38.0 0 0 26.8 10 120 0
Lit. Orange 9 2.8 525 0 108 524 786 4 109 0
Unnamed 10 3.2 70.0 0 7.7 0 0 0 0 0
Moss Lee 11 3.6 148 0 0 7.7 5.2 0 0 0
Jeggord 12 3.0 207 0 0 23.2 15.5 4 0 0
Still 13 1.1 10.4 0 0 5.2 0 0 0 0
Lochloosa 14 2.9 17766 81.6 201 1232 1232 96 371 0
Orange 15 1.8 26405 182 488 1499 2298 54 381 0
Palatka 16 0.8 18.2 0 0 0 13.0 0 0 0
Newnan's 17 1.5 22136 876 71.3 1549 2324 79 792 0
Mize 18 4.0 15.5 0 0 0 0 0 3 0
Calf 19 1.6 100 8 0 0 8.1 0 29 0
Unnamed 20 1.9 38.4 16.5 0 0 2.1 3 11 0
Meta 21 1.6 8.2 4.9 0 0 8.4 0 0 0
Alice 22 0.9 56.8 288 129 0 0 0, 0 5100
Bivin's Arm 23 1.5 378 256 72.8 85.4 0 16 91 0
Clear 24 1.6 15.9 15.2 0 0 0 11 7 0
Unnamed 25 1.0 9.3 1.6 0 0 34.0 0 1 0
Beville's 26 3.1 12.5 5.2 0 0 9.4 3 27 0
Unnamed 27 3.8 26.3 0 0 0 20.2 0 7 0
Kanapaha 28 0.7 4043 1087 0 821 821 6 679 0
Watermelon 29 1.5 979 0 0 106 70.5 3 4 0


(cont'd.)










Table 20 (cont'd.)


Unproductive Population
Mean Forested Urban Fertilized Pastured Cleared Immediate Remote Served
Depth Area Area Cropland Area Area Cultural Cultural by STPa
Name/No. (m.) (ha.) (ha.) (ha.) (ha.) (ha.) Units Units Facilities

Long Pond 30 1.2 43.7 0 0 20.2 17.8 0 0 0
Burnt 31 2.2 129 0 0 55.4 38.8 2 18 0
Wauberg 32 3.8 258 16.1 0 123 0 6 4 0
Tuscawilla 33 1.3 963 66.3 0 154 103 1 59 0
Apopka 34 1.3 2384 467 17508 0 0 274 1157 6950
Dora 35 3.0 1233 931 6762 0 0 342 507 6500
Harris 36 4.2 5979 675 8672 0 3612 438 690 0
Eustis 37 4.1 1683 722 5271 0 900 355 554 7740
Griffin 38 2.4 5157 679 9605 0 1187 415 239 13850
Weir 39 6.3 320 139 1168 0 0 274 105 0
Kingsley 40 7.3 503 328 0 0 96.8 266 120 0
Sand Hill 41 4.8 1689 0 0 0 0 0 0 0
Magnolia 42 8.0 484 0 0 0 0 0 0 0
Brooklyn 43 5.7 667 32.3 0 0 0 167 24. 0
Geneva 44 4.1 741 205 0 0 270 165 388 0
Swan 45 4.8 460 0 0 0 97.5 107 7 0
Wall 46 2.1 401 12.1 0 73.2 114 0 25 0
Santa Rosa 47 8.1 122 0 0 0 0 26 15 0
Adaho 48 3,5 369 0 0 41.3 0 1 0 0
McCloud 49 2.0 40.5 0 0 0 11.3 Q 0 0
Anderson-Cue 50 2.0 48.9 0 0 0 8.1 0 0 50
Suggs 51 2.5 658 0 24.9 23.2 34.8 0 6 0
Long 52 3.4 547 0 0 0 0 10 5 0
Winnott 53 5.2 216 0 15.5 23.8 45.0 48 26 0
Cowpen 54 3.7 712 0 0 0 211 65 110 0
Gallilee 55 3.5 213 0 0 0 150 13 18 0


aSewage Treatment Plant











Table 21. Expected Quantities of Nitrogen
and Phosphorus from Various Sources


Quantity of Quantity of
Source Reference Nitrogen Phorphorus


Domestic Sewage

Fertilized Area

Citrus Farms

Muck Farms

Pastured Area

Unproductive
Cleared Area

Forested Area

Urban Area

Rainfall

Septic Tanks

Immediate

Remote

Domestic Ducks


Vollenweider (1968)







Miller (1955)


Brink (1964)

Sylvester (1961)

Weibel (1969)

Brezonik et al. (1969)







Paloumpis & Starret
(1960)


agrams/capita

bgrams/square

Cgrams/square

dgrams/septic

egrams/duck -


- year

meter of land use area year

meter of lake area year

tank year

year


3940a



2-.24b

.11b

.85b


.18b

.24b

.88b

.58


795a



.018b

.135b

.018b


.006b

.008b

.110b

.044c



138d

13.8d


2420d

970d


480e


90e









fertilizer losses were reported by Vollenweider (1968) and,
although approximate, probably represent lower limits. Septic
tank contributions were estimated using a similar procedure.
An average septic tank was assumed to have a daily effluent
volume of 475 liters with total nitrogen and phosphorus con-
centrations of 35 mg/l and 8 mg/l,respectively (Polta, 1969).
For septic tanks associated with immediate cultural units,
it was estimated that 25 percent of the nitrogen and 10 per-
cent of the phosphorus in the effluent were exported to the
lake. For remote cultural unit septic tanks it was estimated
that 10 percent of the nitrogen and 1 percent of the phosphorus
discharged eventually reached the lake.

Contributions from domestic sewage are expressed in Table
21 as the amount per capital per year. These sewage figures
were used only when effluent records for the individual plants
were not available. One lake (Mize) harbors a colony of 50
domestic ducks; estimated nitrogen and phosphorus contributions
from ducks are thus listed in Table 21. Several large lakes,
e.g. Griffin and Apopka, receive nitrogen and phosphorus via
citrus processing plant effluents. The magnitude of the con-
tributions were determined from average plant flow rates
and concentrations (Environmental Engineering, Inc., 1970).

The calculated nitrogen and phosphorus loading rates
for each of the 55 lakes are presented in Table 22 expressed
as grams per cubic meter of lake volume per year. Loadings
expressed per unit lake surface may be obtained by simply
multiplying the volumetric loading by lake mean depth (from
Table 20). In Florida lakes mean depths rarely exceed 5
meters and most lakes are completely mixed year round. Con-
sequently most of the analyses reported here pertain to the
volumetric loading rates.

In general, the results indicate a positive correlation
between nitrogen and phosphorus supply and trophic state as
quantified by the TSI, but several discrepancies are evident.
Lakes Alice (22) and Kanapaha (28), although demonstrating
hypereutrophic characteristics, have nitrogen and phosphorus
loadings at least an order of magnitude higher than any of
the other hypereutrophic lakes. This can be attributed to
the fact that both lakes have had their natural watersheds
increased by cultural activities, which have resulted in very
short detention times for the lakes. Lake Alice receives 1
to 2 million gallons per day of sewage effluent and 10 to 12
million gallons per day of cooling water from University of
Florida facilities. Lake Kanapaha, which is connected with
a sinkhole draining an urbanized stream, has had its watershed
enlarged 2 to 3 fold by drainage diversion schemes. Thus the
hydraulic characteristics of these two lakes separate them
from the remainder of the study lakes, which receive runoff
from natural watersheds. In order to prevent severe bias
in the statistical analyses, these two lakes were excluded












Table 22. Calculated Nitrogen and Phosphorus
for Fifty-Five Florida Lakes


Lake


Typea


1
2
3
4
5
6
7
8
9
10
11
12
13
14
S15
16
17
18
19
20
21
22
23
24
25
26
27
28


aKey to Symbols: U
H
bTrophic State Index
CIn g/m3-yr


TSIb

2.5
2.3
2.5
2.5
5.3
3.2
2.6
9.1
3.4
3.2
2.8
2.8
1.9
5.2
4.3
3.2
7.1
4.2
4.6
18.5
3.1
10.7
14.7
8.8
6.4
3.1
5.8
13.5


NC PC


Santa Fe
Lit. Santa Fe
Hickory Pond
Altho
Cooter Pond
Elizabeth
Clearwater
Hawthorne
Lit. Orange
Unnamed
Moss Lee
Jeggord
Still Pond
Lochloosa
Orange
Palatka Pond
Newnan's
Mize
Calf Pond
Unnamed
Meta
Alice
Bivin's Arm
Clear
Unnamed
Beville's Pond
Unnamed
Kanapaha


.28
.32
2.25
.53
3.72
1.45
1.01
1.62
2.58
.54
.39
.57
1.73
1.15
1.85
2.70
2.61
2.05
2.42
3.99
3.00
106.00
6.86
4.31
2.07
2.89
.77
48.30


Lake.


Supplies


Typea


.015
.014
.051
.031
.101
.064
.051
.130
.082
.021
.020
.027
.072
.044
.071
.121
.118
.183
.132
.335
.250
18.000
.424
.405
.113
.187
.032
2.95.0


TSIb


NC PC


3.6 1.45


29 Watermelon
Pond
30 Long Pond
31 Burnt Pond
32 Wauberg
33 Tuscawilla
34 Apopka
35 Dora
36 Harris
37 Eustis
38 Griffin
39 Weir
40 Kingsley
41 Sandhill
42 Magnolia
43 Brooklyn
44 Geneva
45 Swan
46 Wall
47 Santa Rosa
48 Adaho
49 McCloud
50 Anderson-Cue
51 Suggs
52 Long
53 Winnott
54 Cowpen
55 Gallilee


5.94
2.27
.63
2.60
2.23
3.00
1.10
1.46
3.69
.29
.18
.29
.25
.26
.31
.26
3.27
.18
1.03
1.35
3.10
2,24
.55
.41
.42
.86


2.8
8.3
7.4
4.8
22.1
18.5
6.3
10.5
13.7
3.3
1.9
1.3
1.3
1.5
1.8
1.5
2.1
1.3
2.2
1.5
1.5
2.3
1.3
2.0
1.5
1.6


.062

.183
.092
.028
.124
.161
.127
.029
.077
.183
.010
.015
.015
.011
.016
.022
.015
.124
.009
.039
.058
.187
.071
.026
.016
.021
.036


- Ultraoligotrophic; 0 Oligotrophic; M Mesotrophic; E Eutrophic;
- Hypereutrophic.









from the sample group.


Anderson-Cue Lake (50) has a nitrogen and phosphorus
loading comparable to hypereutrophic Lake Dora (35), but a
TSI typical of ultraoligotrophic lakes (1.5, see Table 17).
Two reasons may be responsible for this discrepancy: (1) the
lake has not had sufficient time to equilibrate with its
nutrient supply and (ii) the TSI has not been sensitive to
the lake response. This lake has been artificially enriched
with nitrogen and phosphorus at approximately the present
loading rates since 1967 as part of a study of eutrophication
factors in Florida lakes (Brezonik and Putnam, 1968; Brezonik
et al., 1969). Prior to 1967, the lake was ultraoligotrophic
and similar in most aspects to the control, McCloud Lake (49).
Both lakes are still ultraoligotrophic according to their
TSI's although some increased growths of attached algae have
recently been noted in Anderson-Cue Lake. Since the TSI
accounts for phytoplankton production and biomass alone, this
response is not reflected in the TSI.


C, RELATIVE IMPORTANCE OF VARIOUS
NUTRIENT SOURCES


Budgets for six representative lakes are shown in Table
23 in order to compare the percentage contributions of the
various nutrient sources to the overall nitrogen and phos-
phorus budgets. In order to illustrate general trends occurring
in the transition from ultraoligotrophic to culturally hyper-
eutrophic conditions, one lake from each of the five trophic
groups is presented. In addition, Newnan's Lake is included
as an example of a naturally eutrophic lake. For the ultra-
oligotrophic and oligotrophic lakes, the natural nutrient
sources of rainfall and runoff from forested regions are domi-
nant, although Lake Santa Fe receives a small portion of its
nitrogen and phosphorus supply (21%) from cultural sources.
Orange Lake could perhaps be classified as naturally mesotrophic
since most of its nitrogen and phosphorus supply is derived
from natural sources.

Lakes Hawthorne and Dora have obviously been influenced
by the cultural activities in these watersheds. The former
receives the major portion of its nitrogen and phosphorus
supply from urban runoff and septic tanks while sewage effluent
and agricultural runoff have played a significant role in the
deterioration of Lake Dora. Newnan's Lake has a large, heavily
forested watershed, and the associated runoff appears to be
the predominant factor in the eutrophication of this shallow
lake. Eutrophication of this sort is virtually impossible
to control, whereas measures can be taken to control the cul-
tural sources degrading lakes like Hawthorne and Dora.












Table 23. Percentage Contributions From Various Cultural
and.Natural Sources for Selected Lakes


Nutrient Sewage


Urban
Runoff


Fertilized Pasture


Area


Area


Unproductive Rainfall
Cleared Forest Septic on Lake
Area Area Tanks Surface


Santa Rosa (U)


Santa Fe (0)


Orange (M)


SNewnan's (E)


Hawthorne (E)


Dora (H)


aSee Table 4 for key to symbols.


bNot significant (less than 1%).


Lake
and
Typea


Cul-
tural


0
0

5
1

10
2

2
N.S.b

N.S.
N.S.

74
14


0
0

7
3

11
6

14
7

N.S.
N.S.

N.S.
N.S.


0
0

1
N.S.

4
3


N.S.
N.S.


13
11

2
2

N.S.
N.S.

1
1

32
23

2
1









D. STATISTICAL ANALYSIS OF TSI vs. NITROGEN
AND PHOSPHORUS LOADING RATES


Results of the statistical analyses are summarized in
Table 24. Several regression relationships were tested
using both additive and multiplicative models. All the re-
gression results presented in Table 24 were significant at
the 99% confidence level. Using the magnitude of the multiple
correlation coefficient (R) as a criterion for choosing among
the regression equations, an additive equation (A) in Table
24 (b), including simple, interaction and quadratic terms,
explains the largest percentage of variation in TSI (R=.830).
However, equations B and C incorporating only the simple
loadings give comparable significance (R~.80), and inclusion
of the interaction terms thus provides only marginal increases
in R. The multiplicative model (Equations D and E) is the
least significant, and comparison of the additive and multi-
plicative equations suggests that the functional relationship
between TSI and nitrogen and phosphorus loadings may itself
be additive with one nutrient being more significant; i.e.
limiting. In Florida lakes it appears that the phosphorus
loading is the limiting factor since it is the first indepen-
dent variable incorporated by the stepwise procedure into
the regression equations, and it has the highest simple
correlation [.786, Table 24 (a)] with the TSI. However, too
much significance should not be placed on the above interpreta-
tions. Regression analysis is inherently empirical, and its
primary value lies in its predictive abilities rather than
in any analytical potential.

Canonical correlation analysis [Table 24 (c)] derived
a canonical variate of the seven trophic indicators that was
significantly correlated (.723) with the canonical variate
of nitrogen and phosphorus loadings. In general, the analysis
corroborates the regression results. For instance, phosphorus
loading is the more significant of the two loadings based
on the weighting factors in the canonical variate (1.19 for P
vs.-0.23 for N). The most heavily weighted trophic indicator
in the indicator canonical variate is total phosphorus concen-
tration (TP). Thus, the larger weightings associated with P
and TP illustrate the dependence of average total phosphorus
concentration upon the phosphorus loading. Vollenweider (1968)
observed a similar correlation between spring total phosphorus
concentrations and phosphorus supply for a group of European
lakes.

Although the regression and canonical correlation analyses
resulted in statistically significant relationships, there is
considerable disagreement between the predicted and observed
values of TSI, For example, Lake Griffin has an experimental
TSI of 13.7 and a predicted TSI, using equation A of Table 24
(b), of 9.6, a 30 percent error. Similar discrepancies exist











Table 24. Statistical Analyses of Relationships
Between TSI and N and P Loading Rates


(a) CORRELATION MATRIX:


TSI
N


TSI
1.000


N
.773
1.000


P
.786
.935
1.000


(b) STEPWISE REGRESSION ANALYSES:


Model


Loading Rate
Units


Additive
A


Equationb


F Ratioc


TSI = 26.1(PV)-242(Pv)2+1.12(Nv)2


48.1


Multiple
Correlation
Coefficient


.830


+28.7(Nv)(Pv)+2.37(Nv)


Multiplicative
D


TSI = 26.1(PV)+0.90(Ny)

TSI = 0.62(NS)+10.l(PS)

TSI = 1.08(PV).42(Nv).1

TSI = 0.84(PS)'48(NS)*20


43.2

46.4


15.6

14.1


.793

.804

.620

.600


Percent Variance
Explained by
Equation


68.9



62.9

64.5

38.5

36.0


(c) CANONICAL CORRELATION ANALYSIS:
Canonical Variate of Trophic
State Indicatorsd


Canonical Variate
of N and P Loadings


0.69(TP) + 0.64(1/SD) + 0.48(CL) 0.36(TN)
+ 0.34(PP) + 0.33(CD) + 0.17(1/CR) 1.19(P) .23(N)


Canonical
Correlation
Coefficient
.723c


aLoading rates per unit lake volume (V), per unit lake surface area (S).
bAbbreviations: TSI=trophic state index dimensionlesss); NS and PS =nitrogen and phosphorus
surface loading rates in g/m2-yr.; NV and Py =nitrogen and phosphorus volumetric loading
rates in g/m3-yr.
All significant at the 99% confidence level.
dKey to symbols: TP=total phosphorus (mg/l); 1/SD=inverse Secchi disc (m-l); Cl=chlorophyll a
(mg/m3); TN=total organic nitrogen (mg/l); PP=primary production (mg C/m3-hr.); CD=specific
conductance (pmho/cm); 1/CR=inverse of Pearsall's (1922) cation ratio=[(Ca)+(Mg)]/[(Na)+(K)].


~


~









for some of the other lakes with the average error being about
25 percent. Thus in spite of the strong trends demonstrated
by the significant regression relationships, there is substan-
tial scatter of the experimental data about the fitted regres-
sion surfaces. Several possible sources of uncertainty will
be discussed later.


E. CRITICAL NUTRIENT LOADING RATES:
APPLICATION TO LAKE MANAGEMENT


Of great interest in control of cultural eutrophication
is the development of critical loading rates, above which
eutrophic conditions might be expected to ensue.

Vollenweider (1968) has developed two types of critical
loading rates based on information from a number of European
and American lakes. Permissible loading rates are values
below which no eutrophication problems should occur, and dan-
gerous loading rates are values above which problems can be
expected. Loading rates in between these two figures may or
may not cause problems depending on other factors. Inspection
of various limnological data from the 55 Florida lakes indicates
that eutrophic conditions (and attendant water quality deter-
ioration) are associated with all lakes having TSI values
greater than 7.0 (i.e. the lakes in the eutrophic and hyper-
eutrophic classes of Table 17), and similarly lakes with TSI
values less than 4.0 have essentially no nutrient enrichment
problems. Using these as "dangerous" and "permissible" TSI
values, respectively, the nitrogen and phosphorus loading
rates associated with these values were computed, assuming
an N:P molar loading ratio of 15:1 as most appropriate. Criti-
cal rates were computed on both areal and volumetric loading
bases from appropriate regression equations, and Table 25
compares these results with those of Vollenweider (1968). It
appears that Florida lakes can assimilate nutrients at some-
what greater rates without becoming eutrophic than suggested
by Vollenweider's critical values, but the uncertainties in-
volved in both analyses prevent detailed interpretation.

Some interesting results were obtained through graphical
presentation of the relationships between the TSI and phos-
phorus and nitrogen supplies, respectively. In Figure 9, the
mean TSI for each trophic group is plotted against the cor-
responding mean phosphorus loading. Figure 10 represents a
similar treatment considering mean nitrogen loadings. The
horizontal bounded lines represent plus and minus one standard
error of the group loading mean. In both graphs the dependence
of TSI on nitrogen or phosphorus loading can be adequately
described by an exponential function similar to the classical
logarithmic growth curve. The least squares equation and
correlation coefficients are shown for each figure. That
both curves are similarly shaped is to be expected since the













Table 25. Critical Loading Rates
for Nitrogen and Phosphorus


Reference


Loading
Rate Units


Permissible Loading Dangerous Loading
(up to) -(in excess of)


N P N P

Shannon and Volumetric
Brezonik, (g/m3,yr) .86 .12 1.51 .22
1971c

Ibid. Areal(g/m2-yr) 2.0 .28 3.4 .49

Vollenweider Areal(g/m2-yr) 1.0 .07 2.0 .13
(1968)a

aFor lakes with mean depths of 5 m or less.










PHOSPHOJRUS
.20


SUPPLY (g/m2- yr)
.30 .40 .50
""f i |"" r~~"""


I--H ---
S,
12.00oo(P)
/ TS. I= .21e


/r r=.992











-VVOLUMETRIC LOADING CURVE

---SURFACE LOADING CURVE


.10
PHOSPHC:OUS


.15
SUPPLY


.20
g 3 -y
(g/m -yr)


.25


Figure 9. Mean TSI Values for Five Trophic Groups
vs. Annual Phosphorus Loading in (g/m3-yr and g/m2-yr.
Brackets indicate range for one standard error. Sym-
bols of trophic groups are: ultraoligotrophic (U),
oligotrophic (0), mesotrophic (M), eutrophic (E), hy-
pereutrophic (H).


20 r-


F-- C
F- P


.05


.30










NITROGEN
2.0


20


SUPPLY (g/m2-yr)
4.0 6.0


I-- E. y
/i


- 0 -'


I-- U


r
I----- Hl i---




-E S. I 8 1e".87 (N


r= .995


-1--







-' VOLUMETRIC LOADING CURVE

---SURFACE LOADING CURVE


1.0
NITROGEN


2.0
SUPPLY


3.0 4.0
( /m3- yr)


Figure 10. Mean TSI Values for Five Trophic Groups
vs. Annual Nitrogen Loading in g/m3-yr and g/m2-yr.
See Figure 1 for explanation of symbols.


16 1-


12 1


8 k


C~--~.~----~*-~r~-P--^li~-~9------~-Ii~-


i~i~--~-~s~i~i~m~~c~-~ --Fl^iZI--------I---~ITIIZ~Z-----rY-









nitrogen and phosphorus loadings are themselves highly corre-
lated [See Table 24 (a)]. The within-group deviation of load-
ings is considerably more pronounced for the phosphorus re-
lationships, particularly for the hypereutrophic and eutrophic
groups of lakes. Such deviations are to be expected when
representing the complex process of trophic state change in
terms of a single nutrient input. In addition, the hypereu-
trophic group is essentially unbounded at the upper end and
therefore not subjected to artificial boundary constraints
as are the other four groups. Quite likely change-sin the
limiting nutrient will occur over any extended range of trophic
state response. Thus, the relationships of Figuresl and 2
reflect only an average situation, and their major utility
probably lies in the area of lake management. For example,
given that either nitrogen or phosphorus is limiting and a
known or proposed nutrient loading, potential lake response
can be determined by consulting the appropriate relationship.
It should be emphasized that the graphical relationships in
Figures 9 and 10 are most applicable for shallow subtropical
lakes, and their use in other situations may be unwarranted.


F. EFFECT OF DEPTH ON LAKE CAPACITY
TO ASSIMILATE NUTRIENTS


As our data base expands it should be possible to incor-
porate other factors of Eq. 15 into empirical eutrophication
models. For example, mean depth is probably the most impor-
tant morphometric factor affecting eutrophication. Figure 11
indicates a slight mean depth-trophic state relationship ex-
ists for the 55 Florida lakes, with the most eutrophic lakes
having mean depthsof 4 m or less and the deepest lakes being
the most oligotrophic. As expected a large scatter occurs.
The proper relation of mean depth to eutrophication has been
confused by many. It is neither a trophic indicator nor a
causal factor per se. Rather mean depth affects the rate at
which a lake can assimilate nutrients and maintain desirable
trophic conditions. The graphical approach taken by Vollen-
weider (1968) might prove useful in quantifying the effects
of depth. The method plots nutrient (N or P) loading rates
vs. lake mean depth, and the lines delineating the regions in
which oligotrophic and eutrophic lakes occur are estimated
by inspection. Figures 12 and 13 illustrate this approach
for Florida lakes using phosphorus and nitrogen loading rates,
respectively. However it is obvious that insufficient shallow
oligotrophic and deep eutrophic lakes occur in the sample
group to permit accurate delineation of boundary lines. Per-
haps a better approach to evaluating the role of mean depth
in the trophic state calculus would be the method of response
surfaces (Box, 1954; Goldman, 1967).






























20 -


0 0


15 -


o o


0

o U
0


0
a


0 .
(-.
0c
rQ
o (I`C-


0O
0


0 L--__LoJ__ _____ ._ I ________ J
0 1 2 3 4 5 6 7 8 9

Mean Depth, M.




Figure 11. Trophic State Index (TSI) Values
vs. Mean Depth for the Florida Lakes






















HE HE
E HE


HE
U E
0 HE /


.2k


0
HE

M
f


- 10
7
7 U

Eutrophic Lake
(Vollenweider)


E
E/


E/
E U 0 00


U

E
0 0 ~ U
0 M
06 U' U U U
^ U 0 U


U
S

J


I


0 U
0 U


.05 Oligotrophic
Lakes
/ Florida)





Oligotrophic Lakes
(Vollemneider)
.02 I _
.5 1 2 5 10

MEAN DEPTH, M.

Figure 12. Annual Phosphorus Loading Rate
vs, Mean Depth for 55 Florida Lakes.
Each datum represents a lake in the trophic group denoted by
that symbol. U = ultraoligotrophic, 0 = oligotrophic, M =
mesotrophic, E = eutrophic, HE = hypereutrophic.


1.0


HE HE


.51-


Eutrophic
Lakes
(Florida)


M












HE HE


HE
0 E 0
UO


MM M
E t.


S 0


M
UO0


fUTROPHIC LAKES
t 1 (VOLLENWEIDER)









]k OLIGOTROPHIC LAKES
(VOLLENWEIDER)


HE
HK 0

0
0


M EUTROPHIC
LAKES
(FLORIDA)


10.0







5.0.


UO
M
UO OLIGOTROPHIC
E LAKES
0 (FLORIDA)
0 UO
0 0 ^


Oo o o
0 0
UO 0n
U071


u0


Key
UO Ultraoligotrophic
0 Oligotrophic
M Mesotrophic
E Eutrophic
HE Hypereutrophic


I I


1.0


2.0


10.0


Mean Depth (m.)


Figure 13. Annual Nitrogen Loading
vs.. Mean Depth for 55 Florida Lakes

Each datum represents a lake having the
trophic state denoted by that symbol.


HE


E M


2.0







1.0


0.2
0.


0 0


5


I I----- ~ ~ ~ I


-`"


11JL


0.51-









G. SOURCES OF UNCERTAINTY


The above analysis represents an attempt to approximate
the general trophic response function (Eq. 15) by the simple
relationship between TSI and nitrogen and phosphorus loadings
in Eq. 16. The uncertainty term in Eq. 16 represents the
discrepancy between values of TSI predicted by the function
g(N,P) and the actual (measured) TSI values, assuming the
TSI in fact represents-the true trophic status of the-lake.

Individual components of the uncertainty term may include
the following: (i) g is an approximation of f, (ii) the nitro-
gen and phosphorus supply calculations are in error, and (iii)
the TSI does not represent the concept of trophic state (TS)
completely. Approximations of f were obtained here by using
multiple regression techniques. These approximations included
only two of a number of potentially important variables, i.e.
nitrogen and phosphorus loadings. Trie loadings were estimated
using land use and population characteristics and literature
values of individual source contributions, a procedure that
contains some inherent uncertainty. The TSI may not completely
describe the concept of trophic state in spite of the fact
that it incorporates seven of the more significant trophic
state indicators. As previously discussed in reference to
Anderson-Cue Lake, it does not account for macrophyte and
periphyton biomass or primary production, which in some lakes
may constitute a significant proportion of total lake primary
production,


H. RELATIONSHIPS BETWEEN TROPHIC STATE
AND GENERAL WATERSHED CONDITIONS


Another approach to relating trophic state to watershed
factors is direct regression of lake conditions (expressed
by a TSI) to the extent of various land use practices and
population characteristics with the watershed (expressed on
a per unit lake volume or area basis), The trophic indicator
data (Table 7), the TSI values (Table 17), and the population
and land use data (Table 20) were used for these analyses.
In addition, the correlative relationship between the seven
trophic state indicators and the eutrophication factors was
investigated using canonical correlation analysis. The eutro-
phication factors were expressed on a per unit lake volume
basis in the ensuing analyses by dividing the values in Table
20 by the total lake volume, Thus, the units for land use
patterns were square meters for a particular land use per
cubic meter of lake water, and population characteristics
were expressed as number of cultural units per cubic meter
of lake water. The eutrophication factors could alternatively
have been expressed per unit lake surface area. However, it









seems more logical to express eutrophication factors for
shallow Florida lakes on a unit volume basis since the entire
volume is involved in assimilation and dilution of nutrient
influxes. Results considering land use and population factors
on a unit surface area basis were similar to the results ob-
tained on a volumetric basis (Shannon, 1970). For the reasons
discussed in the section on statistical analysis of the TSI
vs. nutrient loading rates, Lakes Alice and Kanapaha were ex-
cluded from the following analyses.

Results of regression analyses for TSI (as independent
variable) vs. various eutrophication factors are shown in
Table 26. Two regression equations are given; the first con-
siders TSI as a linear function of the land use patterns within
the watershed plus the immediate and remote cultural units.
The second considers TSI as a function of the land use patterns
plus total cultural units (TCU), i.e. the sum of remote,
immediate and sewage treatment cultural units. The independent
variables of the regression equations are written in the step-
wise order in which they were incorporated into the equation,
i.e. in decreasing order of their partial correlation with
TSI. Both equations in Table 26 are statistically significant
at the 99% confidence level and both explain about 80% of the
total variance in the TSI.

The first independent variable in both equations is the
fertilized cropland; other culturally influenced factors such
as urban area and immediate cultural units are important var-
iables in explaining the variance in TSI. A natural factor,
forested areas, is also important, but other factors like un-
productive cleared area, remote septic tanks and pastured
areas add little to the predictive abilities of the equations.
These results can be interpreted as suggesting that culturally
influenced factors (fertilized cropland, urban runoff, septic
tank drainage) are among the most important variables deter-
mining the trophic states of Florida lakes. However, it should
also be emphasized that regression analyses are inherently
empirical, and while they may suggest, they never prove cause-
effect relationships.

A canonical analysis of the seven trophic indicators
(Table 7) and six eutrophication factors (the land use areas
and total cultural units) (Table 20).for the 55 lakes is pre-
sented in Table 27. The correlation coefficient between the
two canonical variates f1 and fF is high (0,.94) and signifi-
cant at the 99% confidence level. In the trophic indicator
canonical variate (,I), primary production is weighted consider-
ably higher than the other indicators, suggesting it is of
fundamental importance in the trophic state-eutrophication
factor relationship. At the other extreme the cation ratio
has a low weight and appears to be of minor importance in
the relationship. Cultural factors (urban area and fertilized
cropland) carry the heaviest weightings in the eutrophication











Table 26. Stepwise Regression Analysis of TSI
vs. Eutrophication Factors Expressed Per Unit
Lake Volume2


(1) Regression Equation:

TSI = 14.95(HFA) + .64(FOR) + 2.72(ICU) + 1.59(URB)
59.6 73.9 80.0 81.2

.35(UCA) + .06(RCU) .02(PA)
81.5 81.5 81.5

F Ratio = 28.98***
Multiple Correlation Coefficient (r) = .903
Percent of total variation explained by the
regression equation = 81.5%

(2) Regression Equation:

TSI = 14-.49(HFA) + .61(FOR) + 2,23(URB) + .53(TCU)
59.6 73.9 79.4 80.0

+ .31(UCA) .01(PA)
80.3 80.3

F Ratio = 31.91***
Multiple correlation coefficient (r) = .896
Percent of total variation explained by the
regression equation = 80.3%

Key to. Eutrophication Factor i ; ibols

HFA = Heavily fertilized cropland (m2/m3)
FOR = Forested area (m2/m3)
ICU = Immediate cultural units (#/m5xl04)
URB = Urban area (m2/m3)
UCA = Unproductive waste cleared area (m2/m3)
RCU = Remote cultural units (#/m3xi04)
PA = Pastured area (m2/m3)
TCU = Total cultural units (#/m3xi04)

***Denotes significant F value at the 99% confidence level.



IValues below symbols in regression equation indicate cumulative
percent of total variance explained by independent variables
up to that point.












Table 27. Canonical Analysis
of the Relationship Between Seven Trophic Indicators
and Six Eutrophication Factors1



Canonical Variate 1: Linear function of trophic indicators

S= -0.36(1/SD) + 0.71(COND) 0.17(TON)
+ 0.25(TP) + 1.13(PP) 0.60(CHA) 0.09(1/CR)

Canonical Variate 2: Linear function of eutrophication factors

EF = -0.10(FOR) + 0,53(URB) + 0,79(HFA) 0.04(PA)
-0.06(UCA) 0.16(TCU)

Canonical correlation coefficient = 0.94***

***Significant at the 99% confidence level by a testing
procedure described in Morrison, 1967.


'See Table 26 for key to eutrophication factor abbreviations.









factors canonical variate (EF ); pasture and unproductive
cleared areas carry the lowest weightings, which corroborates
the regression results of Table 26. Thus in general it appears
that the major link between trophic state (fJ) and eutrophi-
cation factors ( fF) is one of primary production and the
cultural factors of urban and heavily fertilized areas.

Comparing the canonical correlation of trophic indicators
vs. nitrogen and phosphorus loadings (Table 24) with the
above analysis indicates a higher correlation coefficient was
obtained in the latter analysis. There is some inherent error
in using literature values of the expected nitrogen and phos-
phorus contributions from land use patterns in order to ob-
tain nutrient loadings, and this quite likely explains the
lower correlation for the analysis using the nutrient loading
rates. In other words, the land use and population character-
istics in their raw form contain more significant information
than the calculated nitrogen and phosphorus loadings.

Empirical relationships such as those in Tables 26 and
27 depend on the fact that runoff from various land use prac-
tices has different and to an extent defined nutrient enrich-
ment effects on receiving bodies of water. Similarly the
population within a watershed can be divided into a few main
gi. ,p3 (e.g, people on sewerage systems, people using septic
tanks in the immediate vicinity of a lake, etc.) which have
similar (within group) enrichment effects, Refinement of
this type of regression relationship could prove beneficial
to regional planners and land use (zoning) boards.

In evaluating the statistical relationships between
trophic state and eutrophication factors, the time element
has not been considered. It has been assumed that lake trophic
state as reflected by the TSI or certain trophic state indica-
tors was a result of the eutrophication factors at that time
or, in other words, trophic state and the eutrophication fac-
tors were in equilibrium at the time they were evaluated. In
reality, the trophic state of a lake is the result of the
eutrophication factors influence over a period of years.
For example, the hypereutrophic conditions of Lakes Apopka
and Dora in the Oklawaha group are due to the intense cultural
activities around the lake in the past two or three decades.
On the other hand, Anderson-Cue Lake has been subjected to
a high rate of nutrient enrichment over a period of three
years but remains in an oligotrophic condition, presumably
because it has not had sufficient time to demonstrate a re-
sponse. Very little is known about the response time of a
lake to nutrient enrichment, and as yet it is impossible to
quantify. However, it seems reasonable to assume that the
majority of the lakes are in a state of dynamic equilibrium
with their environments; the relatively high correlations
between causal factors and effects would seem to substantiate
this point.









I. RELATIONSHIP BETWEEN TSI AND TOTAL
WATERSHED AREA


A simpler relationship between watershed and lake trophic
state was recently proposed (Schindler, 1971) for lakes within
a similar geological region and in which cultural influences
are slight. In a nutrient poor terrain the atmosphere acts
as the major nutrient source. Assuming a steady-state exists
between nutrient-input (via -precipitation) and- nutr-ient export
to the lake, the rate of lake nutrient enrichment should be
directly proportional to the sum of lake area (Ao) plus water-
shed land area (Ad). Because nutrient influx will be diluted
in proportion to lake volume (V), Schindler (1971) hypothesized
that lake trophic conditions then should be proportional to
(Ad +A )/V. Many lakes in north-central Florida fit the above
conditions, and Figure 14 shows the crude correlation result-
ing when TSI is plotted vs. the watershed factor for these
lakes. Data-points in Figure 14 represent seepage and semi-
drainage lakes located in similar terrain in the Trail Ridge
and Alachua County regions of Figure 2. Drainage lakes and
those showing major cultural influences were excluded. The
hypothesis seems to have limited applicability under these
conditions but the scatter implies poor predictive abilities.
Thus the earlier statement that eutrophication is a complicated
phenomenon is again borne out, and simple relationships are
unlikely to explain more than general trends. From the point
of view of eutrophication control and lake management, a
compromise between highly complex mathematical models and
oversimplified empirical relationships, such as described in
the preceding analyses, would seem the most appropriate means
of effecting satisfactory results.


CHAPTER 6. CONCLUSIONS


The limnology of north and central Florida is dominated
by shallow solution type lakes in a sandy terrain. While
thermal stratification is not typical in these lakes, neither
is it rare, and stable stratification can occur in small ponds
as shallow as 3.5 meters deep. The waters of most lakes are
low in dissolved solids, soft and slightly to moderately acid.
Organic color is an important but geographically variable
feature of the lakes. Both acid and alkaline conditions occur
in colored waters, but the former are more prevalent. Appar-
ently few lakes are springfed, accounting for the paucity
of hard-water lakes.

Lake trophic state was envisioned as a multi-dimensional
or hybrid concept described by several biological, chemical
and physical indicators. Groups of lakes with similar trophic
state characteristics were formed using cluster analysis, and
























7.0 -


O
6.0

0
SO
5.0
H

0)
0 4.0


3.0- 0 0 0
0 00
0 0

2.0 O0

0 0 00 0
1.0 -



0 1 2 3 4 5 6 7 8 9

[Ao + Ad]/V, m-1





Figure 14. Trophic State Index vs. Total Watershed
Area/Lake Volume for Selected Florida Lakes
A = lake area, Ad = watershed land area

Only seepage or semidrainage lakes with minimum
cultural influences are plotted.









these groups could be interpreted in the classical (oligo-
trophic-mesotrophic-eutrophic) sense.

A trophic state index (TSI) was formulated using prin-
cipal component analysis incorporating seven trophic state
indicators. The TSI quantified the concept of trophic state
on a numerical scale, thus providing a method for ranking
and comparing lake trophic states.

The trophic states of Florida lakes are largely depen-
dent on gross nitrogen and phosphorus supplies (loading rates)
as evidenced by significant regression relationships between
TSI and N and P loading rates and significant canonical cor-
relation between seven trophic indicators and the N and P
loadings, Phosphorus loading was the most important vari-
able from a statistical viewpoint in the regression and canon-
ical relationships, and it might be inferred that on an average
basis phosphorus is the (most common) limiting nutrient for
Florida lakes.

Cultural nutrient sources are relatively unimportant in
oligotrophic lakes, but for many eutrophic lakes, cultural
sources are by far the most significant. Critical nutrient
loading rates were calculated for Florida lakes based on the
regression relationships. Florida lakes seem capable of
assimilating greater quantities of nutrients than suggested
by Vollenweider's critical loading figures, but the two studies
are in general agreement.

A positive correlation exists for Florida lakes, between
lake trophic state and lake watershed land use and population
characteristics. The relationship was verified by statisti-
cally significant multiple regression equations using the
TSI as the dependent variable and several watershed land use
and population characteristics as independent variables.
Canonical correlation analysis of several trophic state indi-
cators versus the population and land use characteristics
showed high correlation and corroborated the regression re-
sults. It appears that cultural influences have played a
major role in determining the trophic states of Florida lakes.
Regression and canonical analyses results indicate that the
most influential eutrophication factor from a statistical
viewpoint is fertilized cropland.

In spite of the statistically significant results ob-
tained in this study there are several sources of uncertainty
in the methodology. These sources have been discussed in the
text and should not be overlooked in studies of a similar
nature.











APPENDIX


MULTIVARIATE TERMINOLOGY


The term ''multivariate analysis" is used to describe
statistical- techniques- concerned with-analyzing data -collected-
for p different variables on N objects. For example, the
variables in this study are chemical, biological, and physi-
cal characteristics measured on several lakes representing
the objects. Some dependency is assumed among the variables
so that they are considered as a system, implying that no vari-
able can be separated from the group and considered individu-
ally. This feature distinguishes multivariate data and tech-
niques from their multi-dimensional nature multivariate tech-
niques are most conveniently-described using vector and matrix
notation.

Vector quantities in the text are underscored, for ex-
ample xi represents the vector of p variables for lake i.
Matrix quantities are denoted by capital letters, and scalar
quantities are denoted b :'ill letters. Vectors are column
vectors unless the transpose is indicated by priming the
vector (e.g. x.' is the transpose of x.). The inverse of a
matrix A is denoted by A-. The (ijT-th element of a matrix
A is denoted by aij.

Suppose that the assumptions of random and independent
sampling have been satisfied and observation vectors of p
variables are evaluated for N lakes. The resulting collec-
tion of data may be expressed in an Nxp (N rows and p columns)
raw data matrix:

x x ... x
11 12 ip
x x ... x (A-l)
21 22 2p
X =


XNl XN2 ... xNp


The X matrix is the starting point for most multivariate pro-
cedures. Analogous to the univariate situation where a ran-
dom variable x is considered to be normally distributed with
mean p and variance a2, multivariate data are considered to
be realizations of a p-dimensional random variable distributed
multivariate normal with mean vector i and covariance matrix
Z. As t and o2 are estimated by the sample mean x and the
sample variance s2 in the univariate case, V and Z are estimated









by the vector of sample means of the p variables' and the
sample covariance matrix S:

1 N
x- 1 x (A-2)
Nk= 1k

N
and S Z- (x-). (A-3)
N-lk=l

S is the p x p matrix of covariance between all possible pairs
of variables, i.e. si = the covariance between variables
xi and xj. S is a symmetric matrix, i.e, si = s.i, except
for i=j. The variance of variable xi is con ainea in the
element sii.

The matrix of sample correlations between all possible
pairs of variables is denoted by the matrix R where:


r = sij (A-4)
iJ ls J
ii jj

The matrix R can be computed from S by the expression:


R = D( ).S.D(- ), (A-5)
si Si


where D(-) denotes a matrix containing the reciprocals of
the si standard deviations in the diagonal elements and
zeros in all other elements of the matrix. The matrix R is
also p x p and symmetric.

When the variables under consideration are in different
units and ranges it is necessary to transform (or standardize)
them to a scale of common origin and units. The Z score method
is a commonly used standardization technique. The raw data
matrix X is transformed to the standardized matrix Z by

1
Z = (I-iE)XD (A-6)


where Z and X are the Nxp matrices of transformed and raw
variables, respectively. I is the NxN identity matrix with
l's on the diagonal and zeros elsewhere, E is an NxN matrix
with l's in every position and D is a pxp diagonal matrix with
reciprocals of the standard deviations on the diagonal ele-
ments and zeros elsewhere, The general element of Z is given
by










zij (A-7)
j











ACKNOWLEDGEMENTS


This research was supported in part by Office of Water
Resources Research Matching Grant DI 14-31-0001-3068 and a
grant from the State of Florida Game and Fresh Water Fish
Commission. A Federal Water Quality Administration Grant -
DON 16010 (H. D. Putnam, principal investigator), supported
a substantial portion of the project, especially in its early
phases.

A number of faculty colleagues have contributed advice
and encouragement, including Drs. Hugh D. Putnam, James P.
Heaney, William H. Morgan, and Jackson L. Fox. Dr. Fox was
especially helpful in providing needed assistance in sampling
and in the various biological aspects of the project. The
assistance of Dr. Morgan in administrative affairs and in the
completion of this report is truly appreciated.

Special thanks also go to Roger Yorton and Michael Keirn,
project assistants and graduate students in the Department of
Environmental Engineering, for their cooperation in arranging
and conducting the sampling trips and running the chemical
and biological analyses.









BIBLIOGRAPHY


Aberg, B. and Rohde, W., "Uber die Milieufaktoren in einigen
sudschwedischen Seen," Symb. Botan. Uppsala, Vol. 5,
1942, pp. 1-256.

Anderson, T. W., An Introduction to Multivariate Statistical
Analysis, John Wiley and Sons, Inc., New York (1958).

Beeton, A. M. 1965. Eutrophication of the St. Lawrence
Great Lakes. Limnol. Oceanogr., 10:240-254.

Birge, E. A. and Juday, C., "The Organic Content of the Waters
of Small Lakes," Proc. Amer. Phil. Soc., Vol. 66, 1927,
pp. 357-372.

Birge, E. A. and Juday, C. 1934. Particulate and Dissolved
Organic Matter in Inland Lakes. Ecol. Monographs, 4:440-474.

Box, G. E. P. 1954. The exploration and exploitation of
response surfaces: some general considerations and ex-
amples. Biometrics 10, 16-60.

Bowen, D. H. M. Environ. Sci. Technol. 4, 725-726 (1970).

Bradley, W. H. and M. E. Beard. 1969. Mud Lake, Florida; its
algae and alkaline brown water. Limnol. Oceanogr., 14:
1277-1279.

Brezonik, P. L., "Eutrophication: The Process and Its Modeling
Potential," Proc. Workshop Modeling the Eutrophication
Process, Univ. Florida, Gainesville, 1969, pp. 68-110.

Brezonik, P. L..and C. L. Harper. 1969. Nitrogen fixation in
some anoxic lacustrine environments. Science, 164:1277-1279.

Brezonik, P. L., Morgan, W. H., Shannon, E. E., and Putnam, H. D.
1969. Eutrophication factors in north central Florida
lakes. Univ. Florida Industr. Engrg. Exper. Station, Bull.
Ser. No. 134, Gainesville, 101 p.

Brezonik, P. L. and Putnam, H. D., "Eutrophication: Small
Florida Lakes as Models to Study the Process." Proceedings,
17th South. Water Resources and Poll. Contr. Conf., Univ.
North Carolina, 1968, pp. 315-333.

Brezonik, P. L. 1971. Nitrogen: sources and transformations
in natural waters. Presented at 161st Nat. Meeting,
Amer. Chem. Soc., Los Angeles, Calif., April, 1971.

Brink, N. in "Nordisk Killokium om Eutrofieringsproblemer,"
0. Skulberg, ed., Norsk Inst. Vannforskning,Blindern,
Norway, 1964.




Full Text

PAGE 1

"I .... Publication No. 13 Trophic State of Lakes in North Central Florida By Patrick L. Brezonik and Earl E. Shannon Department of Environmental Engineering Sciences University of Florida Gainesville

PAGE 2

TROPHIC STATE OF LAKES IN NORTH CENTRAL FLORLDA_ by PATRICK L. BREZONIK and EARL E. SHANNON PUBLICATION NO. 13 FLORIDA WATER RESOURCES RESEARCH CENTER RESEARCH PROJECT TECHNICAL COMPLETION REPORT OWRR Project Number B-004-FLA Matching Grant Agreement Numbers 14-31-0001-3068 (1970) 14-31-0001-3068 (1971) Report Submitted: August 3, 1971 The work upon which this report is based was supported in part by funds provided by the United States Department of the Interior, Office of Water Resources Research as Authorized under the Water Resources Research Act of 1964. ',"

PAGE 3

TABLE OF CONTENTS ABSTRACT CHAPTER 1. EUTROPHICATION AND FLORIDA LAKES A. INTRODUCTION ..... B. NATURE OF EUTROPHICATION. C. QUANTIFYING EUTROPHICATION .. D. COMPOSITION OF THE LAKE STUDY GROUP CHAPTER 2. EXPERIMENTAL PROCEDURES. A. SAMPLING METHODS ..... B. PARAMETERS EVALUATED AND EXPERIMENTAL Page 1 2 2 2 6 11 15 15 TECHNIQUES. . . .. ..... 16 C. MULTIVARIATE ANALYTICAL METHODS CHAPTER 3. LIMNOLOGICAL RESULTS A. MORPHOMETRIC AND PHYSICAL FEATURES. 19 28 28 B. GENERAL CHEMICAL CHARACTERISTICS. 32 C. PHYTOPLANKTON AND MACROPHYTE CHARACTERISTICS. 35 D. SEDIMENTS .. CHAPTER 4. CLASSIFICATION AND QUANTIFICATION OF TROPHIC CONDITIONS IN FLORIDA LAKES . A. DEVELOPMENT OF A TROPHIC CLASSIFICATION SYSTEM FOR FLORIDA LAKES ....... B. DEVELOPMENT OF DISCRIMINANT FUNCTIONS TO CLASSIFY LAKES OUTSIDE THE ORIGINAL SAMPLE 36 37 37 GROUP . . . . 45 C. FORMULATION OF TROPHIC STATE INDICES. .. 49 CHAPTER 5. RELATIONSHIPS BETWEEN TROPHIC STATE AND WATERSHED ENRICHMENT FACTORS. . 64 A. INTRODUCTION. . . . .. 64

PAGE 4

Page B. NITROGEN AND PHOSPHORUS BUDGETS . 65 C. RELATIVE IMPORTANCE OF VARIOUS NUTRIENT SOURCES . . . . 72 D. STATISTICAL ANALYSIS OF TSI VB. NITROGEN AND PHOSPHORUS LOADING RATES. . 74 E. CRITICAL NUTRIENT LOADING RATES: APPLICATION TO LAKE MANAGEMENT .. 76 F. EFFECT OF DEPTH ON LAKE CAPACITY TO ASSIMILATE NUTRIENTS. 80 G. SOURCES OF UNCERTAINTY. 84 H. RELATIONSHIPS BETWEEN TROPHIC STATE AND GENERAL WATERSHED CONDITIONS. 84 I. RELATIONSHIP BETWEEN TSI AND TOTAL WATERSHED AREA. . 89 CHAPTER 6. CONCLUSIONS. APPENDIX ACKNOWLEDGEMENTS BIBLIOGRAPHY ADDENDUM 89 92 95 96 101

PAGE 5

ABSTRACT TROPHIC STATES OF LAKES IN NORTH CENTRAL FLORIDA General limnological and trophic conditions of 55 lakes and ponds in north and central Florida were established over an extensive one year sampling period. Florida lakes are typically shallow and in a sandy terrain. Most of the lakes have soft water, and high organic color is a common but variable property. Trophic conditions range from ultraoligotrophy in the sand-hill lakes of the Trail Ridge region to hypereutrophy in some large drainage lakes in Alachua County and in the Oklawaha River Basin. Trophic data were analyzed by multivariate techniques, and logical trophic groups derived by cluster analysis. A quantitative index of trophic state (TSI) was derived using 7 trophic indicators, and the TSI values were used to establish quantitative relationships between lake trophic conditions and watershed characteristics. Nitrogen and phosphorus budgets were calculated for the lakes based on land use and population patterns in the watersheds, and critical loading rates were estimated from the budgets and the trophic conditions. Brezonik, P.L. and Shannon, E.E. TROPHIC STATES OF LAKES IN NORTH AND CENTRAL FLORIDA Completion Report to the Office of Water Resources Research, Department of Interior, July, 1971, Washington, D.C. 20240 KEYWORDS: eutrophication/ nitrogen/ phosphorus! analysis/ water quality/ lakes/ nutrients/ models. I

PAGE 6

CHAPTER 1. EUTROPHICATION AND FLORIDA LAKES A. INTRODUCTION Although Florida has more than 7500 lakes (Florida Board of Conservation 1969), limnological investigations of these lakes have been few and limited to special interests. Most detailed studies have been centered on a few unusual or recreationally important lakes; for example, Mud Lake (Marion County) (Bradley and Beard, 1969; Iovino and Bradley, 1969), Lake Mize (Alachua County) (Brezonik and Harper, 1969; Keirn and Brezonik, in press) and Lake Apopka (Orange and Lake Coun ties) (for a review, see Sheffield and Kuhrt, 1970). Yount (1963) has reviewed most pre-1960 limnological studies in discussing some general features of Florida lakes. However, as a group Florida lakes are almost limnologically unknown. Threatening of the recreational assets of Florida lakes by cultural encroachment and consequent nutrient enrichment has stimulated studies on these lakes. In 1968 the University of Florida Department of Environmental Engineering initiated an extensive survey of the physical, chemical and biological characteristics of 55 lakes in north and central Florida. The investigation had five main objectives: i) to determine the basic limnological features of lakes in the region; ii) to assess the present water quality state) characteristics of the lakes and provide baseline data for future studies; iii) to evaluate the applicability of the common trophic state indicators to sub-tropical lakes; iv) to provide necessary data to develop an index of trophic state for sub-tropical lakes; v) to study the relationships between lake trophic state and lake watershed conditions influencing trophic state. B. NATURE OF EUTROPHICATION Cultural lake eutrophication is an undesirable consequence of the interaction between man and his environment. Many of his agricultural, industrial, domestic and recreational activities are introducing excess nutrients into surface waters, causing significant water quality deterioration. Since fresh water is vital to the total well-being of the environment, man has an obligation to protect valuable lacustrine resources. However, progress in solving the problem has been retarded by the inherent complexity of the process, and considerable vagueness still exists concerning the definition of cause and effect in the overall process (Brezonik, 1969; Putnam, 1969). 2

PAGE 7

It is generally agreed that eutrophication involves nutrient enrichment, and a lake in time responds to this enrichment. This response is reflected in a lake's trophic state (eutrophic condition). However, few efforts have been devoted to quantifying the relationship of eutrophication to trophic state. One of the problems in the study of lake eutrophication is of a semantical nature; i.e. distinguishing between and defining the causes, symptoms and effects. Considerable literature has been devoted to discussing these concepts. The meaning of the term "eutrophication" has been stated by Hasler (1947) as being, simply, the enrichment of water, be it intentional (cultural) or unintentional (natural). This nutrient enrichment is generally considered as the causal mechanism in the overall eutrophication process. As originally suggested by Naumann (1919) perhaps primary consideration should be given to nitrogen and phosphorus nutrients. The concept of trophic state (degree of eutrophy) is difficult to define. Eutrophic conditions are the consequences or effects of a lake's nutrient enrichment, but there is no way to express this state in simple, quantitative terms. Much of the conceptual difficulty with the idea of trophic state could have been avoided long ago had limnologis.ts defined trophic state in precise terms as a measure either of a lake's productivity or of a lake's nutrient status. Instead the term has been used to refer to both characteristics. While correlated to a degree, productivity and nutrient status are both also functions of other independent phenomena (e.g. hydrology and climate). Adequate description of a lake's trophic state requires consideration of several different physical, biological and chemical characteristics. For this reason the coricept of trophic state is not only mUlti-dimensional but hybrid, as suggested by Margalef (1958). The trophic state of a lake cannot be measured directly because of its mUlti-dimensional nature. However, it is evidenced by various symptoms called trophic state indicators. A list of common indicators of trophic state is in Table 1. Reviews of trophic state indicators have been compiled by Fruhetal. (1966), Vollenweider (19682, Hooper (1969) and Stewart andRohlich (1967). There has been no scarcity of lake classification schemes and a review of such is beyond the scope of this report. Birge and Juday (1927) made a fundamental distinction concerning the origin of dissolved organic matter in lakes. Lakes dependent on internal sources production) were autotrophic and lakes dependent on external sources were allotrophic. Later Aberg and Rohde (1942) related the classical trophic types of lakes in a two-dimensional concept of autotrophy and allotrophy. This general approach was used for the classification purposes in this study and the idealized two-dimensional relationship is shown in Figure 1. Organic color measurements were assumed to be indicative of external-3

PAGE 8

Table 1. Trophic Indicators and Their Response to Increased Eutrophicationl Physical Transparency (d) (Secchi disc reading) Morphometry CD} (mean depth) Chemical Nutrient concentrations (I) (e.g. at spring maximum) Chlorophyll a CI) Conductivity-CI) Dissolved solids Cll Hypolimnetic oxygen deficit CI 2 Epilimnetic oxygen supersaturation (I) Sediment type Biologica12 Algal bloom frequency (I) Algal species diversity (D) Littoral vegeta-tion (I) Zooplankton (I) Fish (I) Bottom fauna (12 Bottom fauna di-versity (D) Primary production (I) 1(12 after parameter signifies value increases with eutrophication: (D) signifies value decreases with eutrophication. 2Biological parameters all have important qualitative changes, i.e. species changes as well as quantitative (biomass) changes as eutrophication proceeds. From Brezonik (1969) 4

PAGE 9

0::: 9 o (.) t c U lLI 0::: Z 0 -I (!) 0 0::: (.) C/) lLI -I o T lLI 0::: ::) C/) lLI :::E La.! <:) (.) ::t: Q. 0 0::: t-0 (!) -I 0 (.) ::t: Q. 0 0::: t-0 (!) -I 0 2 (.) ::t: Q. ::t: (.) 0 Q. 0::: 0 l: t-0::: Q. ;:) to lLI 0 0::: 0::: C/) t-lLI lLI ;:) Q. 2 lLI >-::t: (.) (.) 1111111 -(.) ::t: ::t: Q. Q. ::t: 0 0 Q. 0::: 0::: 0 t-t-o::: ;:) L 0 ,t-lLI C/) ;:) 0::: lLI lLI 2 MEASURE OF TROPHIC STATE ,.. 0,' f" c:. "'f Figure 1. TWO-Dimensional Concept of Lake Classification Based on Autotrophy (Internal Organic Production) and Allotrophy (External Organic Input) 5

PAGE 10

source dissolved organic matter and thus denote lake allotrophy. As originally suggested by Hansen (1962), colored and relatively clear lakes were recognized as two fundamentally different lake types. Within each of these types, oligo-, meso-, and eutrophic state subdivisions could occur as determined by some measure of lake trophic state. C. QUANTIFYING EUTROPHICATION From a qualitative viewpoint the phenomenon of eutrophication is now fairly well understood. However, for lake management eutrophication control qualitative facts are seldom sufficient. For example, it is generally recognized that increased nitrogen and phosphorus input to a lake will generate increased plant production. But information concerning the precise nutrient loading rates that stimulate excessive production and scum-forming algal blooms is sorely lacking. Lakes are highly complex ecosystems, and the factors controlling nutrient cycling and primary and secondary production in them are at best poorly understood. Furthermore, lakes cannot be regarded as isolated entities, but the interactions of the entire watershed with the lake itself must be taken into account (Hutchinson, 1969). The general significancem various land use patterns and cultural activities as nutrient sources are largely unknown, and in particular the total nutrient loading rates for specific lakes of varying trophic conditions are .known with accuracy for only a few cases. The complexities of the eutrophication problem suggest the utility of systems analysis techniques and of mathematical modeling in properly defining the problem and simplifying it to the extent that solutions become feasible. The theory and nature of mathematical ecosystem models have been discussed in several recent papers and books (Moreau, 1969; Patten, 1969; Watt, 1968; and Thomann, 1971). In general mathematical models can be divided into two types. Analytical or mechanistic models consist of a series of equations (algebraic, or in ecosystem models more commonly, differential) which attempt to explain the fundamental (functional) relationships between certain parameters. For example, differential equation models of primary production have been developed (Patten, 1968) in terms of the basic relationships between photosynthesis and light intensity, nutrient levels, etc. Empirical or statistical models are composed of approximate parameter relationships which are derived by such techniques as regression, multi-variate, or time series analyses. Such models are attractive in management of complex systems where cause-: effect relationships are unknown. Empirical models can be useful in predicting system response to changes in environmental conditions, and they can give clues to the significance of the relationships (i. e. the dependency) between variables. However their lack of foundation in causal relationships renders 6

PAGE 11

empirically developed models susceptible to misuse and overextension (to conditions in which they may not be applicable). The inherent complexities of nutrient enrichment and its attendant effects on lakes imply that a purely deterministic approach is beyond our present capabilities. While functional relationships are known for various lacustrine phenomenon, and relatively sophisticated analytical (i.e. differential equations) models have recently been formulated for even as complicated a process as planktonic production (Chen, 1970; DiToro et al., 1970; Patten, 1968), the much larger scope of the eutrophication problem precludes such approaches at the present time, especially in the general case. For particularly unique and valuable resources like Lake Tahoe or the St. Lawrence Great Lakes, the manpower and time expenditures required for development of such models may be justified. This seems. not to be the case for the thousands of smaller and locally important recreational lakes in the U. S. and elsewhere. A simpler, less costly approach is required for these lakes. Where large numbers of lakes must be managed an attractive possibility is th.e development of empirical models based on data from a representative sample of the lakes in question. Such management tools as critical nutrient loading rates can be developed by empirical manipulation of basic limnological and watershed information. While empirical models are perhaps not the ultimate answer to eutrophication problems, they can provide direction for further studies and models while simultaneously providing interim predictive capacities required for proper water quality management. Eutrophication is a multivariable problem and thus lends itself to analysis by multivariate statistical techniques. Beneficial applications of empirical multivariate models can be anticipated in three major areas of eutrophication research, and Table 2 summarizes potential applications in each area. Because of the broad, multi-dimensional concept of trophic state, multivariate techniques seem especially appropriate for the long standing problem of rational lake classification. Trophic classification systems can be useful in several ways: a) for identification Ia certain class (name) calls to mind certaj..n disti.nctive characteristics]; b} for organization of our knowledge concerning the obj ects (lakes 1 being classified; cl as the basi.s for development of theories regarding causes of phenomena associated with a particular class (e. g. what do lakes. in a class have in common that might induce their similar behavior Land d 1 for management purposes (different classes. of lakes may have different IIbest uses" and require different land use and water management controls). ThB ill-defined concept of trophic state is in reference to both. a lake's general nutrient status and its productivity, 7

PAGE 12

Table 2. Applications of Empirical Models to Quantification of Eutrophication 1. Lake Classification a. formation of logical lake groups according to multi-dimensional concept of trophic state b. delineation of the set of conditions (ranges for indicator values} defining different trophic groups c. determination of redundancy and uniqueness among various trophic indicators 2. Quantification of Trophic State a. development of uni-dimensional quantitative trophic state index (TSI2 b. correlation of classical trophic indicator values with water quality problems 3. Relationship between Lake Trophic State and Causative Factors a. regression models of TSI vs. Nand P loading b. regression models of trophic state vs. population and land use patterns (including basin hydrology and morphometry) 8

PAGE 13

which are not always correlated. The circumstances defining a given state (e.g. eutrophy) are not at all agreed upon by limnologists. No single measure of nutrient status or productivity is satisfactory or sufficient, and the results one obtains depend on which indicators are used. Thus the limnologist is left with the difficult task of subjectively deciding which indicators to use and which to disregard or weigh less heavily. Reviews on trophic state indicators have been published elsewhere (Fruh et al., 1966; Vollenweider, 1968; Hooper, 1969). Selection-or-appropriate indicators is a difficult task, but consideration of the following criteria should facilitate the decision: a) an indicator should be quantifiable in order to permit numerical differentiation between lakes of varying trophic states, b) each indicator should be unique (i.e. not measure the same lake characteristic as another indicator), c) an indicator should have fundamental significance in terms of the concept of trophic state (as a general measure of a lake's nutrient and productivity status), and d) an indicator should be sensitive to levels of enrichment and relatively simple to measure. The uniqueness of trophic indicators can be studied by several multivariate statistical methods, including factor analysis (Shannon, 1969; Lee, 1971), principal component analysis (Lee, 1971) and cluster analysis (Goldman et al., 1968; Shannon, 1969). While different geographical regionS-may require somewhat different treatment, indicators should be widespread properties of aquatic environments in order to insure general interpretability of the generated classes. The subjectivity involved in forming logical trophic classes from conflicting indicator data can be minimized with certain multivariate techniques such as cluster analysis (Sokal and Sneath, 1963). Another important classification problem is the assignment of lakes outside the original sample group into appropriate pre-established classes. The method of discriminant function analysis (Shannon, 1970; Lee, 1971) is useful in this regard. In order to predict and evaluate the consequences of watershed management practices on trophic conditions in a lake, trophic state must somehow be quantified. As discussed above, this has heretofore been obviated by the multidimensional nature of the trophic concept. Development of a single numerical index of trophic state from a combination of several important indicators avoids the misleading and fragmentary situation arising when only one indicator is used and the confusion which results when several indicators are considered individually. An index also allows quantitative interpretation of trophic state not otherwise feasible. At least five applications and advantages derive from development of a trophic state index: 1) a numerical index would be 9

PAGE 14

valuable in conveying lake quality information to the nonand semi-technical public; 2) an index would be useful in comparing overall trophic conditions between lakes; 3) in the dynamic process of lake succession and trophic change, an index would provide a means to evaluate the direction and rate of changes; 4) an index would facilitate development of empirical models of trophic conditions as a function of watershed "enrichment" factors for predictive and management purposes; 5) a properly developed index would be highly relevant to (i.e. identified with) water quality from a human (or user's) perspective. In contrast to the last point, many indicators (especially qualitative species composition indicators) are largely of academic or research interest. On the other hand an index can be criticized as having no real physical meaning and as improperly combining diverse parameters (the "can't add apples and oranges" syndrome). However, the first argument is irrelevant; a relative index of trophic state, in so far as it reflects the trophic concept, has value regardless of its interpretability in actual physical terms. With proper selection of indicators and rational development of an index, the second criticism can be largely overcome, but it must be realized that no index can or should be expected to supply the detailed information available in the individual parameters. Proper selection of indicators is a vital consideration in developing an index of trophic state. Criteria discussed previously with regard to trophic classification apply equally here; that the individual indicators be quantifiable is of course essential. The number of indicators desirable in an index bears some discussion. Generally an index should include sufficient indicators to account for the essential attributes denoted by the broad trophic concept. As fewer variables are used, the index becomes more unstable, i.e. a large deviation from "normall1 for a given indicator will tend to affect an index incorporating few variables more than one incorporating many. Use of only one variable could result in very misleading rankings of lake I1trophic states." For example if plankton biomass (expressed as packed cell volume, numbers per ml, or chlorophyll a) were the sole measure, lakes with a dense and active macrophyte and periphyton population but low phytoplankton levels would be misranked as oligotrophic. Similar criticisms apply to any other single indicator, and to a lesser extent when only a few indicators are used. However, redundant indicators (i.e. those that measure essentially the same phenomenon as another indicator) should be avoided to prevent biasing the index, i.e. weighing it too heavily toward that aspect or phenomenon. For example, specific conductance and dissolved solids should not both be used in an index since they measure nearly the same thing. The multivariate statistical method of principal component 10

PAGE 15

analysis represents one means of deriving a single numerical trophic state index from a number of indicators. Given such an index, empirical models of trophic state as simple functions of nutrient loading rates or other watershed enrichment factors can then be developed by multiple regression analysis or other appropriate means. D. COMPOSITION OF THE LAKE STUDY GROUP Fifty-five lakes from three different areas of northcentral Florida were selected for the study (Figure 2). Table 3 lists the lakes by name and code number and gives the surface area and mean depth of each. The study originated in early 1968 with a survey of 33 lakes within Alachua County, in which Gainesville and the University of Florida are located. This group, comprising all accessible and potentially important recreational lakes in the county, exhibits considerable diversity in trophic conditions. Most of the lakes are very shallow, and moderate to high organic color is common, reflecting the large expanses of pine forest in the county. The small lakes typically have outlets only during periods of extended rain whereas the large lakes have permanent outlets. General physical features of the Alachua County lakes and initial chemical and biological measurements were summarized by Brezonik et al. (1969); Clark et al. (1962) have described the geological formations land forms which affect the lakes. In early 1969 lakes from two important north-central Florida lake regions outside of Alachua County were included in the study. Sixteen lakes in the Trail Ridge region of the Central Highlands (east of Alachua County) comprise one of these groups. This scrub-oak, sand-hill region is richly endowed with lakes, most of which are clear and lie within small drainage basins. Lakes in the Trail Ridge area are naturally low in nutrients and subject to only light cultural influence. While still shallow and typically unstratified, these lakes are generally deeper than lakes in the other two groups. Anderson-Cue and McCloud Lakes are being used as model lakes in a separate eutrophication study (Brezonik and Putnam, 1968; Brezonik et al., 1969). Artificial nutrient enrichment of Anderson-Cue Lake has been proceeding since 1967, and the relevant chemical, biological and physical characteristics of both lakes have been monitored since 1966. The final group consists of six lakes in the upper Oklawaha River Basin northwest of Orlando, Florida. Five of the Oklawaha lakes are joined by watercourses with the general pattern of flow being from Lake Apopka through Lake Dora to Lake Eustis which drains into Lake Griffin. The effluent 11

PAGE 16

LOCATION OF STUDY AREAS TRAIL RIDGE LAKES SANTA FE '---, RIVEH .42 F I o 10 2.0 -:;0 KM. Figure 2. Location of 55 Lakes in Study Areas of North-Central Florida 12 GEORGE

PAGE 17

Lake Number Table 3. North-Central Florida Lakes in this Study Lake Name Mean Depth (meters) (1) ALACHUA COUNTY LAKES 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23. 24 25 26 27 28 29 30 31 32 33 Santa Fe Little Santa Fe Hickory Pond Altho Cooter Pond Elizabeth Clearwater Hawthorne* Little Orange (Unnamed) Ten* Moss Lee Jeggord Still Pond Lochloosa Orange* Palatka Pond Newnan's* Mize* Calf Pond (Unnamed) Twenty* Meta Alice* Bivinls Arm* Clear* (Unnamed) Twenty-Five Beville's Pond (Unnamed) Twenty-Seven Kanapaha Watermelon Pond Long Pond Burnt Pond Wauberg* Tuscawilla 13 5.5 4.8 3.4 3.6 2.2 1.5 1.5 2.8 2.8 3.2 3.6 3.0 1.1 2.9 1.8 0.8 1.5 4.0 1.6 1.9 1.6 9 1.5 1.6 1.0 3.1 3.8 0.7 1.5 1.2 2.2 3.8 1.3 Surface Area (hectares) 1674 467 27 222 86 27 5 20 314 29 52 64 5 2235 3324 4 2433 1 11 4 2 29 58 3 6 2 4 82 213 5 22 101 162 (cont'd).

PAGE 18

Table 3 (cont'd). Surface Lake Mean Depth Area Number Lake Name (me't.'ers) (hectares) (2 ) OKLAWAHA RIVER BASIN LAKES 34 Apopka 1.3 12412 35 Dora* 3.0 2237 36 Harris 4.2 5580 37 Eustis 4.1 3015 38 Griffin 2.4 3533 39 Weir* 6.3 2301 (3) TRAIL RIDGE LAKES 40 Kingsley* 7.3 667 41 Sumter-Lowry 4.8 508 42 Magnolia 8.0 83 43 Brookln 5.7 253 44 Geneva 4.1 692 45 Swan* 4.8 227 46 Wall 2.1 31 47 Santa Rosa 8.1 42 48 Adaho 3.5 41 49 McCloud* 2.0 6 50 Anderson-Cue* 2.0 5 51 Suggs 2.5 47 52 Long 3.4 104 53 Winnott 5.2 85 54 Cowpen 3.7 240 55 Gallilee 3.5 34 lakes in 19 lake sub-sample group (see text) 14

PAGE 19

from Lake Griffin forms the Oklawaha River. Lake Harris also flows into Lake Eustis. Lake Weir, although in the Oklawaha River basin, does not discharge directly into the Oklawaha River. All six lakes in this group are important recreational lakes; in the past Lake Apopka was among the best known bass fishing lakes in the country. However, considerable cultural eutrophication (and consequently water quality impairment) has occurred in the five connected lakes within recent years. The watersheds of these lakes are utilized primarily for cit rus farming, but a large area on the north shore of Lake Apopka is devoted to vegetable farming of muck soils (recovered marshland) CHAPTER 2. EXPERIMENTAL PROCEDURES A. SAMPLING METHODS The sampling schedule used in this study was designed to provide information on the average chemical, biological and physical characteristics of the 55 lakes over a one-year period. Systematic sampling of all lakes began in June, 1969, and all 55 lakes were sampled at four-month intervals up to June, 1970. In order to obtain greater detail on seasonal trends, a 19 lake sub-group from the 55 lakes was sampled at two-month intervals during this same time period. The 19 lakes (denoted by asterisks in Table 3) were selected on the basis of being representative of the different trophic types present in the 55 lake group. It was felt that this subgroup adequately reflected seasonal trends in lake characteristics without sampling all 55 lakes on a closer time interval. Water samples taken from the lakes for chemical and biological analysis were composites. The small lakes (surface area less than 10 hectares and maximum depth less than 4 meters) were sampled at two stations over depth (surface, middle, and bottom). These samples were combined into a composite sample from which aliquots were taken for major chemical for nutrient analyses (preserved with mercuric chloride), for primary production and chlorophyll analysis, and for plankton identification and counts (preserved with formalin). For the larger lakes that were relatively shallow (maximum depth <10 meters) the procedure of sample collection was the same except that three stations were sampled and composited. For the few deep lakes in which stable stratification was evident, samples were composited from the euphotic zone (estimated as twice the Secchi disc reading) for biological analyses and from the entire water column for major chemical analyses, and nutrient analyses were done in profile on uncomposited samples taken at regular depth intervals. Sediment samples were taken by Ekman dredge from 15

PAGE 20

the deepest region of the lake. B. PARAMETERS EVALUATED AND EXPERIMENTAL TECHNIQUES A total of 6 morphometric, 2 physical, 29 chemical and 6 biological parameters were evaluated for each lake during the project. In addition 11 parameters were evaluated for the lake sediments. Six land use and three population characteristics were evaluated for each lake drainage basin. Table 4 lists all the parameters measured at various times during the project. The physical parameters were measured in situ; biological and chemical parameters were determined on the composite samples using standard limnological procedures (see Brezonik et ale 1969 for details). Primary production was measured inthe laboratory with a "light box" procedure rather than in situ in order to standardize light and temperature conditions and offer a more uniform basis of comparison among the lakes. Bathymetric maps were available for about 20 of the lakes (Kenner, 1964); the remainder were sounded and mapped with a Heath Co. depth sounder as part of the project. Basic morphometric parameters such as volume, mean depth, volume development index and shoreline development index were computed from the bathymetric maps by methods described in Hutchinson (1957). Land use patterns in the lake watersheds were determined by aerial photograph and topographic map interpretation. Lake watershed areas were outlined and planimetered from United States OeQlogical Survey (Scale: 1/24,000) topographic maps. Recent (1965-1968) aerial photographs (Scale: 1" = 1667') were obtained for (each watershed from the Florida Soil Conservation Service Office. Using photogrammetric techniques, areas of various types of land use patterns were delineated and measured. Lake surface areas were also from the aerial photographs. The population in each watershed was characterized in four categories. Residences on a shoreline were classified as immediate cultural units (ICU). Other residences within the lake watershed were categorized as remote cultural units (RCU). The lCU's and RCU's were evaluated from aerial photographs. Residences served by sanitary sewer facilities were not included in the two previous categories. Recent population figures were obtained for all of the municipalities served by sewage treatment plants within each of the lake watersheds. These figures were converted to equivalent cultural units by dividing by a factor of 2.5, which represents the average population of a single rural family residence in 16

PAGE 21

Table 4. Lake and Basin Parameters Evaluated for this Study Land Use Fertilized cropland Pastured area Forested area Urban area Watershed Unproductive cleared area Total watershed area Bathymetric map Mean depth Morphorrietric Shoreline development Temperature profile Turbidity Acidity Alkalinity Ammonia Calcium Chloride C.O,D. Color Copper Dissolved oxygen Fluoride Iron Magnesium Manganese Mercury Nitrate Nitrite Chlorophyll a Total carotenoids Physical Chemical Biological Algal identification and counts 17 Population Characteristics Cultural unitsl on lake shore Cultural units in rest of basin Sewage treatment plant Cultural units Lake surface area Maximum depth Volume development Secchi disc transparency Organic nitrogen Ortho phosphate pH Potassium Silica Sodium Specific conductance Strontium Sulfate Suspended solids Total phosphate Total solids Zinc Primary production Algal species diversity (cont 'd).

PAGE 22

Table 4 (cont'd). Ammonia Organic nitrogen Total phosphate Sediments Sediment type (visual classification) Benthic organisms ISee text for explanation of this term. 18 Volatile solids CIN ratio Iron Manganese Chlorophyll derivatives Total carotenoids

PAGE 23

the State of Florida (U.S. Bureau of Census, 1961). Cultural units of municipalities discharging sewage effluent directly into a lake were classified as immediate sewage treatment plant cultural units (ISPU). Cultural units of municipalities discharging sewage effluent somewhere else in the watershed were classified as remote sewage treatment plant cultural units (RSPU). The total cultural units (TeU) in the watershed was obtained by summing the cultural units in each of the four categories. Estimates of total watershed population couid in turn be obtained by multiplying the TCU by 2.5. C. MULTIVARIATE ANALYTICAL METHODS Relationships among the several trophic indicators and watershed eutrophication factors were investigated by a variety of multivariate statistical techniques .. This term is used to describe statistical methods concerned with analyzing data collected on several dimensions (variables) on a set of objects or individuals. Some dependency is assumed among the variables so that they are considered as a system. Because of their nature, these techniques are most conveniently described using vector and matrix notation. Theoretical aspects of these techniques are discussed by Morrison (1967), Sokal and Sneath (1963), and Lee (1971). The applications and computational aspects of the techniques used in this$udy are described below; see Appendix for a description of the terminology used for vectors, matrices, and multivariate statistics. 1.Cluster Analysis is concerned with the problem of classifying N objects (e.g. lakes) into groups based on p variables measured on each obj ect, when the number .of groups that best fit the data is not predetermined. geo metrically, the method attempts to distinguish logical groupings of obJects in the p-dimensional hyperspace described by the p data attributes of the objects. Figure 3 illustrates a simple bivariate cluster problem involving groups formed by hypothetical data for color and productivity in lakes (cf. Figure 1). Cluster analysis of objects is referred to as a analysis; a second type which clusters the variables measured on a set of objects is referred to as R-type cluster analysis. Cluster analysis was used in this study to find natural groupings of lakes, i.e. those with similar trophic or chemical characteristics, as measured by several li:mnological parameters (indicators) considered simultaneously and weighed equally. Cluster analysis progressively combines a set of objects into a smaller and smaller number of groups according to the degree of similarity among the objects; objects (lakes) greatest similarity are joined first. The starting point for any cluster analysis is the N x 19

PAGE 24

H CJ ,-I o o CJ CD bfl H o .,;-----------........ / / B \ I \ I \ I I \ I \ / /' / "'"---------------------/' A "-/' "/ \ / .. \ I. \ \ I II J \ / \ 'j / ./ ......... _-----_._-----Primary Production Figure.3. Hypothetical bivariate plot showing clusters formed by data for organic color and primary production in lakes. Solid circles represent clusters formed around 4 groups with good similari ty: 1. low color, low production; II. low color, high production; III. high color, low production; IV. high color, high production. Dashed lines represent less similar clusters of (A) low color and (B) high color lakes formed later (at higher objective function values). 20

PAGE 25

p raw data matrix X. If it is desired to group objects, the matrix X is normally transformed to the matrix of standardized variates Z since the variables may have been measured in quite different sized units. The standardized data are used to calculate product-moment correlation coefficients for all possible pairs of objects. The resultant N x N symmetric matrix is called the similarity matrix Q, with general element qij being the correlation between objects i and j considering the p variables measured on each object. The Q matrix represents the starting point of the cluster analysis. The three basic elements of a cluster analysis are the between-object distances, the clustering criterion and the computational procedure (Padron, 1969). A multitude of methods are available to evaluate the between object distance (see Sokal and Sneath, 1963, for a review); popular distance measures include the correlation coefficient between objects and simple functions of the Euclidean distance. The distance measure used in this project was proposed by Gower (1966): d .. = [2(1-q .. )Jl / 2 (1) lJ lJ where d .. is the distance between the i-th and j-th objects and qijlJiS the correlation coefficient or measure of simi larity between objects i and j. Clustering criteria (a measure of the goodness of any given allocation of objects into groups) usually include a measure of within group similarity. In some cases, good group similarity implies good between group dissimilarity. The clustering criterion used was minimization of an objective function (OF): OF =500 (WBAR-BBAR), (2) where WBAR is the average within group distance and BBAR the average between group distance for any given allocation. The constant an arbitrary number used to scale the objective function into a convenient range. Using the distance measure in Eq. (1), the minimum value of the objective function in Eq. (2) is -1000, implying complete similarity within groups and complete dissimilarity between groups, A value of zero represents a random grouping of the lakes (where the mean within group and between group distances are equal). Consideration of the OF value for any allocation and its change from a previous allocatiDn offers a means of determining the relative degree of similarity between the tWD groups or objects joined. Computational procedures are usually heuristic in the interest of solving large problems with an economy of 21

PAGE 26

computer time. A clustering algorithm in Fortran IV developed by Padron (1969) was used infue cluster analyses. 2. Discriminant function analysis is a multivariate classification procedure which can be used to assign objects into appropriate pre-established classes. Figure 4 illustrates a simple example involving two groups formed by two variables. Discriminant functions are linear combinations of variables for which the separation between groups is a maximum. The functions contain as many variables as there are dimensions to the objects. When the population is divided into two mutually exclusive groups, one discriminant function is sufficient to determine the group to which an object belongs. Fisher (1936) first formulated the method for the separation of two groups of objects. This technique was later generalized by Anderson (1958) so that linear discriminant functions could be evaluated for distinguishing between multiple groups. Let TIl,TI2'" TIm be the m populations under consideration. In this study the populations, TIi' represented the different trophic states to which a lake may belong. Associated with each population are the multivariate probability density functions Pl(x), an observation vector of p variables). It is desired to divide the space of observations into m mutually exclusive and exhaustive regions P1,P2 .Pm If an observation falls into Pi it is assumed to be a member of population TIi' Assume the distribution of TIi to be normal with mean and covariance The covariance matrix is assumed to be common for all i populations. If the costs of misclassification are equal and the a priori probabilities qi of drawing an observation from TIl are known, the region Fi is defined by those x satisfying where P'k is the linear discriminant function related to the ith andl kth populations. The a priori probabilities of x being in population i or kare given by qi and qk' respectively. The discriminant function Pik is given by = log Pi log Pk(x) (4 ) Usually and are not known and and S are used as their estimates (x. is the vector of sample means of the p variables and S is th sample covariance matrix). The linear 22

PAGE 27

COLORED LAKES CLEAR LAKES .TURBIDITY (X) 'DISCRIMINANT FUNCTION; Vxy .= aY -bX Figure 4. Two-Dimensional Plot Showing Relationship Between Discriminant Function and Two Clusters of Inverse Secchi Disc Transparency and Turbidity Data. Clusters the envelopes of points (not shown) for colored and .cle:ar (uncolored) iakes. Color decreas.es Secchi disc visibility;hencecolored lakes tend toward bigher (l!SD) values for a given turbidity .. Bell .... shaped curves represent distribution on a given axis for data points within each cluster. 23

PAGE 28

discriminant function then becomes vik and is given by: vl k == [x !( x. +x, ) ] IS (Xl' +xk ) -2 -l -J --(5 ) For sufficiently large samples vil;\: is considered to be a good estimator of 'J.lik" If the a priorl and qi are equal in Eq. 3, the region Pi is defined for 'J.lik>O. The method used to calculate the linear discriminant functions in this study was the stepwise procedure (BMD07M) described in Biomedical Computer Programs (Dixon, 1968), In the stepWise procedure variables are brought into the criminant function one at a time based on an IF' test for significance. In essence, the most powerful discriminatory variables are entered into the discriminant function first and less important variables at later stages. 3. Principal component analysis is used to examine the dependence structure of multivariate data and reduce the dimensionality of the data by expressing the original observation variables in terms of fewer component variables, which are linear functions of the observation variables. A simple bivariate example of principal component analysis is shown in Figure 5. Principal component analysis was used to derive indices using the first principal components extracted from trophic state correlation matrices of trophic indicators measured on the lakes. When the variables are expressed in different units, the matrix of sample correlations (R) between all possible pairs of variables is used as the starting point in the analysis. If p variables are involved, R is a p x p symmetric matrix. The first principal component Yl of the correlation matrix R is the linear combination y,== a' z "" -1-' ( 6 ) where a' is the transpose of the first characteristic vector (eigenvctor) of R associated with the largest characteristic root (eigenvalue) of R, and! is the vector of standardized variables. The variance of Yl is given by The jth principal component Yj is given by (7) where a, is the transposed eigenvector associated with the jth largestJeigenvalue/l., of R. J 24

PAGE 29

N ?<: r! -P C) :::s 'd 0 H jl., H jl., Chlorophyll (Xl) Yl = aXI + bX2 First principal component Second principal component Figure 5. tiypothetical bivariate plot of primary production and chlorophyll data showing relationship of first and second principal components to original variables. First component is defined to pass through long axis of elliptical &le cluster configuration, giving maximum variance of the cluster; second component passes through short axis of sample cluster, giving maximum variance in that direction. 25

PAGE 30

In principal component analysis the main objective is to explain as much of the variance in the original observations as possible with a minimum number of components. The first principal component is that linear combination of variables which explains the maximum variance in the original data; the second principal component is the linear combination of variables explaining as much of the remaining variance as possible, and so on. As many component variables as original variables can be derived, at which point all the variance is explained, but this subverts the purpose of the procedure (i.e. reducing the dimensionality of the data). The proportion of the total variation that anyone component Yj explains is given by tr(R) Aj p (8 ) where A. is the jth eigenvalue of Rand tr(R) is the trace of of the diagonal elements). The trace of R is also equal to p (the number of variables) since each diagonal element of R has a value of unity. Theoretical and computational aspects involved in calculating principal components from covariance or correlation matrices are presented by Morrison (1967). The BMDX 72 program from the Biomedical Computer Programs Library (Dixon, 1968) was used to perform the analyses in this project. 4. Canonical correlation is used to analyze the statistical relationships between two sets of variables considered in vector form. In this project canonical analysis was used to study the relationships between a trophic state vector consisting of seven trophic indicator variables, and a eutrophication factor vector, consisting of several land use and population characteristics of the lake watersheds. The advantage of canonical correlation over conventional multi-regression analysis is that the former allows one to study relationships between two sets of variables without defining anyone variable as dependent and without assuming orthogonality (independence among the variables). This method determines the linear combination of the variables within each set which produces the maximum correlation coefficient between the two sets. Thus canonical analysis can be used to determine the dependency structure, i.e. the nature and extent of covaria tion, between two sets of variables. Consider a random vector composed of observations on p variables with a covariance E. This vector x may be partitioned into two subvectors Xl and with PI and Pz components, respectively. Usually the variables of each subvector will have some common feature, e.g. let consist of several trophic state indicators for a lake and the vari-26

PAGE 31

abIes be various eutrophication factors that influence trophic state, For convenience, it is assumed that Pl (PI + pz + 1) and S is the unbiased estimator of t:. The covariance matrix may be partitioned into submatrices in a manner similar to x where where the dimensions of S11' SIZ and Szzare PI x PI' PI X Pz and pz x Pz, respectively. Once a tenting procedure (described by Morrison, 1967) indicates a significant dependence between Xl and the method of canonical correlation may proceed.-In canonical correlation analysis the following question is proposed. What are the linear compounds ]Jl = b' ,]Jt = b' -1 -t VI = c 1 ,vt = c' _.1 -t with the property that the sample correlation of ]Jl and VI is greatest, the sample correlation of ]Jz andvz greatest among all linear compounds uncorrelated with ]JI and VI and so on for t min(Pl'PZ) possible pairs? These pairs of linear compounds are called canonical variates. It should (10) be noted that the correlation matrix R could have been partitioned in a similar manner to S resulting in similar canonical correlations. However, canonical variates based on the correlation matrix are dimensionless and are expressed in terms of the standardized variables. The BMD06M program (Dixon, 1968) was used to perform the canonical correlation analyses. 5. Multiple regression analysis may be described as a method to predict the value of one variable (Y) from the values of other variables (X.), Variable Y is assumed to be dependent on the values of1the independent variables X.. Strictly 1 speaking multiple regression analysis is not a method of multivariate analysis since variates are considered interdependent the latter, and no single variable can be considered as the "dependent variable." The general model of (linear) multiple regression may be written as 27

PAGE 32

(11) where Y is the dependent variable, Xl' X2 X are independent variables, b o is the intercept value, and b2 b are regression coefficients. The variables may be raw dataP values or may be transformed values of the raw data. The principle value of multiple regression analysis lies in its predictive capacity (i.e. prediction of Y values from a measured set of Xi)' The technique was used to evaluate statistical relationships between the trophic state index (TSI) and eutrophication factor (land use and population) variables. The BMD02R program (Dixon, 1968) was used with the zero intercept (i.e. bo=O) option. This option was used since it is desirable to have a situation where the TSI is equal to zero when all the eutrophication factors are zero. The computer program is a stepwise multiple regression procedure and adds the variables to the equation in decreasing order of their statistical significance (i.e. their partial correlation with the dependent variable). CHAPTER 3. LIMNOLOGICAL RESULTS Detailed descriptions of the morphometry and physical features of the lakes in the study group are the subject of another report in this series. Similarly the chemical and biological limnology of the lakes will be described in detail in a third report (Brezonik, in preparation). This chapter will describe the limnological results in general terms as background information for analysis of eutrophication factors and lake trophic conditions in the following chapters. A. MORPHOMETRIC AND PHYSICAL FEATURES The geology Of Florida is dominated by a limestone substratum underlying the entire peninsula. In north-central Florida the upper limestone deposits are of Eocene to Miocene age and are covered by more recent deposits of sand and clay_ Thickness Of the overlying formations ranges from a meter or so (e.g. in southern Alachua County) to over 30 meters. The limestone rise to a karstic topography throughout the peninsula with artesian springs, sink holes and solution lakes as prominent features of the landscape. The morphometry and physical features of Florida lakes are to a large extent determined by the geological structure and resulting topography. Table S summarizes these features for the 19 lakes sampled bimonthly. In general the lakes are shallow, and maximum depths of more than 10 m are uncommon. 28

PAGE 33

Table 5. Features of Selected Florida Lakes LAKE SURFACE MAXIMUM MEAN AREA DEPTH (zm) DEPTH (z) (hectares) (m) (m) Santa Fe 1674 8.8 5.5 Hawthorne 20.4 4.3 2.8 '#10 29.3 4.6 3.2 Orange 3324 3.0 1.8 Newnan's 2433 4.0' 1.5 Mize 0.86 25.3 4.5 #20 3.7 3.4 1.9 Alice 28.6 1.5 0.9 Bivin's Arm 58.4 1.9 1.5 Clear 3.4 2.7 1.6 Wauberg 101 5.2 3.8 Dora 2237 4.9 3.0 Weir 2301 9.8 6.3 Kingsley 667 22.9 7.3 Geneva 692 8.8 4.1 Swan 227 9.4 4.8 McCloud 5.6 3.7 2.0 Anderson-Cue 4.5 4.6 2.0 Suggs 47.2 3.7 2.5 aDevelopment of volume index = 3z /Zm DV 1. 88 1. 95 2.09 1. 80 1.13 0.53 1. 68 1. 80 2.37 1. 77 2.19 1. 84 1. 93 0.96 1. 40 1.53 1. 62 1. 30 2.03 bDevelopment of shoreline index = L/2/TIA. 29 a DL b 1. 24 1.10 1.10 1. 63 1. 20 1.19 1. 28 1. 66 1. 48 1.40 1.13 2.24. 1. 70 1. 01 1. 58 1. 37 1.17 2.12 1. 22

PAGE 34

Mean depths for all 55 lakes range from about 0.7 to 8.1 m, and maximum depths range from about 1.0 to 25 m. Most of the shallow lakes basins; i.e. the lake basin walls are concave toward the water. The rleeper lakes generally are more cone-shaped; in the deepest lake of the survey, Lake Mize, the lake basin walls are considerably convex to ward the water. The trend can be seen by examining the volume development indices (Dy) in Table 5. Index values less than 1.0 indicate a convex toward the water) lake basin while values greater than 1.0 are indicative of U-shaped basins. Lakes with a DV of 1.0 have a basin similar in form to that of a cone (Hutchinson, 1957; Zafar, 1959). Many small Florida lakes are hydraulically perched; i.e. their connection to groundwater is with a perched water table located above and not directly connected to the principal aquifer in the peninsula, the Floridan aquifer. Most of the small Alachua County and Trail Ridge region lakes are seepage (Birge and JudaY,1934) with no visible outlets or permanent inlets, and water levels may vary as much as several meters between dry and wet periods. Thus few lakes have a definite land-lake interface, and the shorelines may be intermittently submerged land. Water levels in the larger drainage lakes (e.g. Newnan's, Orange and Lochloosa Lakes, Alachua County) frequently are structurally controlled so that water level variations are much smaller. Some of the Trail Ridge lakes (e.g. Kingsley, Swan, Brooklyn), because of their occurrence in a region of very sandy soil, do possess fine natural sandy beaches in spite of the periodically wide fluctuations in water levels. Nearly all the natural lakes in Florida have been derived or substantially modified by limestone solution processes. Numerous lakes are situated in sink-hole depressions formed by dissolution of underlying limestone (Stubbs, 1940; Hutchinson, 1957). In some cases lake basins have originated by other mechanisms (e.g. fluviatile action) but solution activity has substantially modified the original basin (e.g. Lake Tsala Apopka in Citrus County; Cooke, 1939). Many small and some larger lakes are simple dolines which tend to have simple circular basins. Perhaps the best example is Kingsley Lake (Clay County), an almost perfectly circular basin (shoreline development index, SD=l.Ol) about 3 km in diameter. Lake Santa Rosa (SD=1.09), a lake 0.8 km in diameter in Putnam County, is another example. SD is defined as the ratio of the actual length of a lake's shoreline to the minimum length (i.e. the circumference of a circle) which would enclose an area equal to that of the lake surface. Other lakes are complex dolines with more irregular shorelines. For example, Lake Brooklyn (Clay County) consists of at least 9 separate solution basins and has an SD=2.37, and Cowpen Lake (Putnam County) with an SD=1.80 consists of at least 5 basins. 30

PAGE 35

The shallowness of Florida lakes suggests that thermal stratification would be unimportant in these lakes, and in deed most lakes do not exhibit classical Birgean thermoclines with stagnant hypolimnia as is common in temperate lakes. Eight lakes are sufficiently deep to develop stable stratification and oxygen deficient bottom waters; these are Lakes Mize (Brezonik and Keirn, in press), Magnolia, Moss Lee, Santa Rosa, unnamed lakes numbered 20 and 27, and Beville's Pond. Climatic circumstances favor a long period of stratification; for example Lake Mize is stratified from February or early March till November. The surprising feature of some of the lakes is the shallowness at which stable thermal stratification can occur. Lake No. 20 is only about 4 m deep but the temperature in the bottom meter is several C cooler than the minimum temperatures in the region during Summer. Lake No. 27 is only about 7 m deep, yet it has a pronounced thermocline between 2.7 and 4.2 m (9-12ft), and the bottom water was 11.4c in June, 1969, which is only 1C warmer than the mid-winter bottom temperature. Clearly morphometric factors are important in producing the thermal stability of these lakes. Both are fairly small (1.5-4.5 ha), are in a rolling terrain and are surrounded by high pine forest. Thermal stratification is not limited to the summer months; temporary stratification can develop as a result of the highly changeable weather that occurs during January and February. While none of the lakes are meromictic, low to zero, dissolved oxygen values in the bottom waters of Beville's Pond, Lake No. 27, and Lake Mize throughout the year indicate "the bottom waters circulate rather incompletely even during winter. At least 6 other lakes among the 55 exhibit incipient thermal stratification. Typical of these are Lake Wauberg and Hickory Pond. Stratification develops only near the bottom in these shallow lakes, preventing the formation of a distinct hypollmnion, but the bottom water temperatures during SUmmer are at least as cool as the nocturnal minima in the region so that fairly stable conditions can be assumed. Low dissolved oxygen values in the bottom waters of these lakes also imply stable stratification. These lakes are somewhat larger or less wind protected by forest than the small lakes discussed previously. Size is obviously an important factor in determining whether stratification will occur in a lake. For example, neither Lake Santa Fe (surface area = 1650 ha, Zm =8.8 m) nor Lake Weir (surface area = 2300 ha, Zm =9.8 m ) have shown any evidence of stratification on any sampling date. Many other shallow lakes show signs of stratified conditions even in the absence of a typical thermocline. Temperature differences of 4-5C from top to bottom in lakes that are only 2-4 m deep are common during summer, but the decline is continuous with depth rather than confined to a narrow layer (also see Yount, 1961). At surface temperatures of 31

PAGE 36

25,.....309C, temperature differences of a few degrees are sufficient to impart considerable stability. to the water column (Hutchinson, 1957). Bottom temperatures are greater than regional nocturnal air temperatures' during. summer, and stratification thus is not highly stable. HoweVer, oxygen depletion in the bottom waters of several lakes a metastable circumstance (i.e. mixing is not a daily phenomenon). B. GENERAL CHEMICAL CHARACTERISTICS In order to determine general patterns in chemical composition among the lakes (i.e. classify the lakes into distinct chemical types), a cluster analysis was performed on data for six basic chemical parameters for the 55 lakes. The parameters considered were pH, alkalinity, acidity, conductivity, color and calcium, and mean values for each lake over the sampling period were used for the analysis. The resulting cluster diagram is shown in Figure 6. The 55 lakes fall into four easily interpreted groups: (i) acid colored lakes, (ii) alkaline colored lakes, (iii) alkaline (hardwater) clear lakes; and (iv) soft, clear lakes. A comparison of the six chemical characteristics in these 4 lake types is shown in Table 6. Assuming the 55 lakes are a reasonable cross-section of the lakes iri Florida, several conclusions derive from the results in Figure 6. Slightly less than 50 percent of the lakes are classified as colored, and the bulk of these are also acidic. Thus color would appear to be a common feature of Florida lakes. However, all but three of the colored lakes lie in Alachua County, which fact both implies a rather heterogeneous geography in the region and suggests that caution should in extrapolating the statistics of the sample group to the population of Florida lakes. Several other regional differences can be noted. The alkaline-colored group is composed entirely of lakes from Alachua County. Three (Newnan's, Orange, Lochloosa) are large connected drainage lakes; the other two are seepage or semi-drainage. All five lakes are moderately enriched. The alkaline clear group includes the five culturally enriched lakes of the Oklawaha chain plus the small eutrophic lakes of Alachua County. The soft water clear lakes are located primarily in the Trail Ridge region and eastern Alachua County, which geographically comprise one topographic unit. One conclusion that seems a valid extrapolation is that Florida lakes generally have soft water; only the 12alkaline olear lakes can be considered to exhibit hardness, and even here the degree is moderate. This may seem contradictory in 32

PAGE 37

VALUE OF OBJECTIVE FUNCTION -6FF---600 i -55.-". _____ --::-5. '? ADAHO SUGGS WALL LONG JEGGO LIT SA ELIZA CALF RD NTA FE r BETH --POND LEE #10 MOSS BURNT ALTHO LIT 0 MiZE BEViL POND-.,WILLA==:=] RANGE LE'S KA PALAT 11:27 HICKO RY r l I 1 I ACIDIC COLOR ED LAf
PAGE 38

Table 6. Selected Chemical Characteristics of Four General Lake Types Characteristics ColoredAcidic pH 5.66a Acidity 7.31b (mg/l as CaC03 ) 6.64c Alkalinity 2.36 (mg/l as CaC03 ) 3.37 Conductivity 45.8 (jJmho/cm) 10.1 Color 220 (mg/1 as Pt) 121 Calcium 3.3 (mg/l) 1.6 adenotes median value bdenotes mean value Co1oredAlkaline 7.63 1.18 .47 11.69 6.04 70.0 11.9 114 45 6.9 1.5 cdenotes the standard deviation 34 Clear ... Alkaline 8.38 1.00 1. 27 92.14 39.91 249 123. 60 30 36.8 16.2 Clear-Soft Water 5.83 2.00 .96 2.80 6.42 48.2 25.6 17 19 3.0 2.2

PAGE 39

view of the solution origin of the lakes and the abundance of hard water springs in Florida, but few Florida lakes are spring fed. Rather, most of the lakes receive the bulk of their water either directly from precipitation or by surface and subsurface runoff from the sandy, low calcareous soils. In fact several of the hard water lakes are not naturally calcareous but have hard water because of cultural effects, i.e. the influx of ground water as treated sewage or septic tank drainage. The mean and median values of the chemical parameters in Table 6 indicate highly distinct and readily apparent differences among the 4 lake types, perhaps much greater than when the lakes are considered individually (as the large standard deviations for some parameters would suggest). The acidic-colored lake group has a much higher mean color than the alkaline-colored group (220 to 114 mg/l as Pt), and the high color probably contributes to the low pH values. Color concentrations as high as 700 mg/l have been found in some lakes (e.g. Lake Mize). Color certainly contributes to acidity (cf. acidity values of the acidic-colored and clear-soft water groups, both of which have acid pH values). Color is the only parameter which has a significantly different value in each of the 4 types and as such appears to be an important chemical characteristic for distinguishing between the lake types. C. PHYTOPLANKTON AND MACROPHYTE CHARACTERISTICS Algal identification and enumeration was done on all 55 lakes at each sampling. Because of year-round favorable growth conditions (solar radiation and temperature), some of the fertile lakes such as Apopka, Bivin's Arm and Dora exhibit virtually continuous algal blooms. However maximum bloom conditions usually obtain during summer. Lake Apopka has phytoplankton blooms of 88,000cells/ml or higher, predominated by blue-green genera such as Lyngbya and and green genera such as PediaatrUm and Scenedesmus. Blooms of 32,000 cells/ml or more have been found in Lake Dora. Newnan's Lake, a colored eutrophic lake, has summer populations predominated by blue-green algae (Microcystis, Anabaena, Spirulina). In winter this lake usually produces an extremely dense bloom of Aphanizomenon, which fixes nitrogen at high rates (Brezonik and Keirn, unpublished data). However this alga is not present in the lake during other seasons of the year and is nota common constituent of the phytoplankton in other eutrophic lakes Microcystis and Anabaena are the summer bloom formers in Bivin's Arm. The latter organism is found in all lakes in which nitrogen fixation has been detected, and seems to be the primary algal agent for this process in all the lakes except Newnan's Lake. 35

PAGE 40

Oligotrophic lakes have typically low algal populations. For in Swan Lake (a clear s.oft water, oligotrophic type) a summer 1969 population of about 36 organisms/ml was dominated by the diatoms Syriedra and Navi.cula and the green alga Sphaerocystis. Dinobryon and Synura Chrysophyceae) are common in the low pH, low icmic strength waters of the soft water clear (oligotrophic) lakes as are a variety of Desmidaceae (e,g. Staurastrum, Closterium, and Cosmarium). Diatoms are comparatively rare in the plankton of Florida lakes, especially in the oligotrophic soft water lakes. Low silica concentrations in Florida lakes may in part account for this distribution. An exception to this general trend is Lake Apopka, which normally supports a high (although not usually dominant) population of diatoms, including MelOSira, Tabellaria, and NaVicula, and perhaps not coincidentally has one of the highest silica concentrations (3.7 ppm) of the 55 lakes. Bivin's Arm with a mean silica content of 1.8 ppm also supports a spring bloom of diatoms (Maslin, 1970). The dominant primary producers in a number of the 55 lakes are floating macrophytes, For example, Lake Alice, on the University of Florida campus was until recently covered alm.ost entirely by a dense crop of water hyacinth (Eichorriia crassipes). While a faculty-student effort succeeded in mechanically clearing this lake (at least temporarily), the plant is common in canals and other lakes (e.g. Lakes Tuscawilla and Apopka). Chemical spraying is used to control the plant in a number of lakes including Bivin's Arm and Lake Apopka. Duckweed (I:iemna ruinor) partially covers the surface of Lake No. 27 throughout the year, while perhaps one-third of Beville's Pond is covered by SalVinia during the summer months, Such growths limit light penetration, drastically reducing phytoplankton populations, and under severe conditions may inhibit oxygen transfer from the atmosphere to the water. D. SEDIMENTS Florida lakes have a wide variety of sediment types, including sand, peat, and sludge-like (ooze) deposits. In som,e of the oligotrophic lakes a light nearly pure sand bottom occupies most of the lake bottom" suggesting the geo .... logical newness of these lakes. Organic deposits in the lakes range in color from light brown (peat) to nearly black (ooze) and the sediment consistency similarly covers a wide range with large fragments of plant remains evident in peat sedi ments and very fine, slowly settling particles in some of the oozes. In many of the lakes there is no defined sedimentwater interface. Rather a gradation from thin suspensions of sedim,entto m,ore compact strata occurs often over depths of a meter or more. This characteristic makes sampling of 36

PAGE 41

bottom water (and surface sediments) rather difficult. In shallow lakes the suspended sediments undoubtedly become mixed with the overlying water during periods of wind stress, and considerable nutrient exchange is thus effected. The carbon: nitrogen ratios in nearly all the sediments are greater than 10 indicating a Hdyl1 type of sediment in Hansen's (1962) minology. A crude correlation also exists between CIN ratio and trophic conditions. The most eutrophic lakes have CIN ratios in the range and oligotrophic lakes have gener ally higher ratios, but considerable scatter occurs when all 55 lake sediments are considered. CHAPTER 4. CLASSIFICATION AND QUANTIFICATION OF TROPHIC CONDITIONS IN FLORIDA LAKES As discussed in Chapter 1, eutrophication and trophic state are extremely complex, multivariable phenomena. At present our understanding of them and their interrelationships is primarily qualitative. A broad effort to quantify these relationships was made using the statistical techniques described in Chapter 2 and the collected limnological and watershed data. The analyses were applied to three major aspects of eutrophication research listed in 1) the long standing problem of rational classification of lakes according to trophic state, 2) quantification of the presently nebulous term "trophic state," and 3) delineating the relationships between lake trophic conditions and watershed enrichment factors. This chapter presents results for the first two aspects; Chapter 5 discusses the third. A. DEVELOPMENT OF A TROPHIC CLASSIFICATION SYSTEM FOR FLORIDA LAKES The of the trophic state concept has heretofore obviated objective and consistent classification of lakes according to their trophic states. In an attempt to minimize subjectivity in delineating trophic classifications for Florida lakes, similarity (cluster) analyses were performed on trophic indicator data from the 55 lakes. Seven indicators; viz., primary production (PP), chlorophyll a (CHA), total organic nitrogen (TON), total phosphorus (TP), Secchi disc transparency (SD),conductivity (COND), and a cation ratio CCR) due to Pearsall (1922) were chosen as the dimensions describing the hybrid concept of trophic state and were considered Simultaneously in the cluster analysis to derive logical lake groups according to their trophic stateB (at least as defined by the 7 indicators). The main considerations in selecting the first six 37

PAGE 42

indicators are. that (;iJ they are quantitative, (ii) they are fundamentally significant as measures of t.rophicstate, (iii) they satisfy HQoper's (1969) criteria for useful trophic indicators reasonably well, and (1v ) they apply to Florida lakes. The first Six indicators have all beeri used with some degree of success in various lake classification schemes. The selection of Pearsall's cation + K) was a somewhat subjective attempt to incorporate Mg + Ca information on the major cations into the concept of trophic state without adding each cation as an individual indicator. Pearsall (1922) reported that English lakes with high nitrate and silica and a Na + K ratio less than 1.5 had periodic algal blooms. Mg + Ca Thus this ratio should be inversely related to increasing eutrophy. Parenthetically it might be noted that many workers have suggested a general correlation between high productivity and water hardness (Ca and Mg concentrations). This ratio has not been used to any extent in other investigations, but it was suggested as a potentially effective parameter for differentiation between lake trophic types by Zafar (1959). For Florida lakes the cation ratio appears to be a reasonably good indicator of trophic state with high values of the inverse cation ratio being indicative of eutrophic conditions. Averages of the lake parameters over the one year sampling period would seem the most appropriate values for the purposes of statistical analysis. In some respects extreme values (e.g. maximum nutrient concentrations, algal densities at the height of bloom conditions, etc.) are more critical determinants of a lake's water quality and may thus be better and more sensitive indicators of trophic state. But extreme values are less reproducible, and their magnitude depends greatly on the vagaries of sampling frequency and climatic circumstances. Since the breadth of this project detailed (e.g. weekly) sampling, it is felt that mean values are more appropri ate in the ensuing analysis. In order not to bias the means toward summer conditions, the June, 1910, values were not included in the comput.ations. Means of the trophic indicators, color, and turbidity for the 55 lakes are listed in Table 7. So that each indicator would denote trophic state in a positive sense (an increase in indicator value denotes an increase in trophic state) the Secchi disc and cation ratio indicators were inversely transformed. Obviously there are many more possible indicators of trophic state that could be included. Alternatively, it may be that fewer trophic indicators will eventually prove sufficient to describe the concept of trophic state, The selection of 7 indicators was a somewhat arbitrary attempt to incorporate as much information into the concept of trophic state as possible without getting into a prolifer ation of secondary or redundant Because of the basic typological differences caused by 38

PAGE 43

Table 7. Trophic State Indicator, Color and Turbidity Color Lake Scaled Number l/SD l/SD Cond TON TP PP CHA l/CR COL TUR 1 0.43 0.52 53.2 0.50 0.021 9.3 5.6 1. 24 59.0 1.9 2 0.66 0.46 53.7 0.61 0.015 1.8 4.5 0.90 149.0 1.5 3 0.56 0.52 45.7 0.70 0.027 4.1 7.6 0.64 62.0 1.9 4 0 0.58 53.3 0.59 0.023 12.4 5.9 0.65 133.7 2.3 5 1.16 0.92 59.7 1. 26 0.165 29.6 22.6 1. 01 83.3 4.5 6 1.66 0.89 47.7 0.81 0.036 6.9 8.0 0.85 236.7 4.3 7 0.41 0.39 40.0 1. 32 0.012 0.5 2.3 0.53 21.3 1.0 8 1.41 0.91 167.7 1. 86 0.079 96.6 56.8 1. 88 58.3 4.4 9 1. 06 0.54 50.7 0.94 0.105 20.8 9.8 0.55 165.7 2.0 10 0.65 0.53 50.4 0.86 0.064 18.2 12.6 0.47 123.7 1.9 LV 11 0.70 0.53 43.0 0.77 0.036 25.4 8.1 0.49 98.3 1.9 \0 12 1. 23 0.77 55.7 0.47 0.087 6.8 7.0 0.39 192.7 3.5 13 0.50 0.47 38.0 0.63 0.013 0.5 3.1 0.61 26.0 1.5 14 1.15 0.81 87.0 1. 42 0.058 20.9 23.3 1. 41 116.3 3.8 15 1. 00 0.74 77.4 1. 07 0.063 27.4 15.6 1. 01 107.1 3.3 16 0.81 0.44 25.0 1. 22 0.024 8.5 15.6 0.78 93.3 1.3 17 1. 91 0.88 59.8 1. 41 0.110 102.5 47.4 0.85 188.9 4.2 18 1. 83 0.47 53.0 0.85 0.113 11.2 33.9 0.62 433.4 1.5 19 1.41 0.42 43.7 1. 41 0.184 12.7 23.5 0.83 404.0 1.2 20 2.87 2.88 314.3 2.06 0.410262.8 92.8 4.04 68.5 17.4 21 0.45 0.49 93.3 0.81 0.030 2.4 3.3 1. 50 25.0 1.7 22 0.67 0.46 552.2 0.50 0.900 7.9 4.4 3.42 25.5 1.5 23 1.65 1.79 253.8 1. 88 0.546 25L. 7 56.00 2.48 42.1 10.2 24 1.33 1.17 136.4 1. 27 0.392 87.4 26.4 2.91 85.4 6.1 25 0.66 0.57 92.7 0.73 0.028 2.6 3.5 9.88 36.0 2.2 26 0.46 0.39 39.7 0.65 0.087 1.8 23.7 0.54 181.7 1.0 27 0.71 1. 37 47.0 0.58 0.325 1.2 30.1 1.21 92.0 7.5 28 2.81 2.27 121. 7 2.20 o .422 7 42.7 5.12 120.7 13.8 29 1.04 1.08 37.7 0.86 0.05217.0 9".0 0.73 74.0 5.6 (cont lei)

PAGE 44

Table 7 (cont'd.) Color Lake Scaled Number l/SD l/SD Cond TON TP PP CHA l/CR COL TUR 30 1. 01 0.61 31. 0 1. 00 0.052 1.4 7.6 0.77 253.5 2.4 31 1. 64 0.89 63.0 1. 69 0.478 86.8 29.0 1. 63 351. 7 4.3 32 1.13 0.88 66.2 1. 67 0.169 103.8 37.3 1. 23 74.8 4.2 33 1. 04 0.62 51. 8 1.19 0.292 20.2 8.5 1. 50 434.5 2.5 34 4.54 4.39 314.7 4.45 0.380 337.7 60.4 3.85 78.0 27.3 35 2.66 3.13 313.0 3.33 0.384 310.7 50.4 3.45 96.4 19.0 36 0.94 0.68 210.0 1.18 0.037 30.1 14.5 3.68 38.7 2.9 37 1. 31 1. 65 251. 7 2.22 0.167 92.8 23.8 3.34 47.0 9.3 38 1. 50 1. 91 255.3 2.63 0.183 218.3 47.3 3.12 36.0 11.0 +:=39 0.51 0.53 135.8 0.82 0.019 12.3 6.5 0.51 8.8 1.9 0 40 0.21 0.39 52.8 0.35 0.011 2.9 1.8 1.15 10.6 1.0 41 0.30 0.45 28.0 0.19 0.011 0.8 1.3 0.63 7.7 1.4 42 0.27 0.40 26.0 0.18 0.012 1.0 1.5 0.64 10.7 1,1 43 0.23 0.45 30.3 0.28 0.011 0.8 1.9 0.81 9.3 1.4 44 0.32 0.50 49.2 0.35 0.016 2.6 1.5 0.61 9.8 1.7 45 0.31 0.45 44.3 0.27 0.011 3.5 1.6 0.51 5.7 1.4 46 0.73 0.37 42.0 0.67 0.025 6.8 5.1 0.47 151.0 0.9 47 0.16 0.37 37.0 0.19 0.011 0.6 1.7 0.48 2.0 0.9 48 1.16 0.37 37.7 0.72 0.027 7.1 5.3 0.60 336.0 0.9 49 0.21 0.41 34.8 0.30 0.017 0.9 2.4 0.59 4.1 1.1 50 0.27 0.39 38.2 0.29 0.018 1.1 3.2 0.63 3.0 1.0 51 1. 09 0.48 46.2 0.69 0.036 6.7 3.4 0.58 280.7 1.6 52 0.24 0.40 52.0 0.09 0.012 0.3 1.4 0.61 10.0 1.1 53 0.30 0.54 41. 3 0.56 0.023 1.3 2.6 0.61 23.3 2.0 54 0.21 0.46 45.7 0.25 0.010 0.4 1.6 0.59 5.0 1.5 55 0.29 0.43 38.0 o .J6 0.014 1.9 1.9 0.65 12.3 1.3 Il/SD (Secchi disc transparency)-l in m-l ; Cond in TON and TP in mg N or P/l; PP in g C/m3-hr; CHA in mg/m3 ; l/CR (cation ratio)-l dimensionless; COL (color) in mg/l as Pt; TUR (turbidity) in Jackson Turbidity Units. See text for explanation of column 3 (color scaled l/SD).

PAGE 45

organic color (Figure 6 and Table 6), it seemed best to sider clear and colored lakes as separate classes in each of which a range of trophic subclasses could exist. In fact, a cluster analysis of the 55 lakes considering the 7 trophic indicators plus color divided the lakes into essentially the same 31 clear and 24 colored lakes shown in Figure 6. (Lakes Wauberg and Kanapaha, which are in the alkaline colored and alkaline clear groups, respectively, in Figure 6, are the only lakes which fall into the opposite groups in the color plus trophic indicator cluster analysis.) The lakes in the colored group had mean color levels greater than about 75 ppm, whereas the clear lakes had color levels less than this value. Thus the horizontal line separating clear and colored lakes in Figure 1 would appear to have a value of about 75 ppm for Florida lakes. The lakes within each of the main classes were grouped into subclasses of similar trophic state by performing cluster analyses with the 7 trophic indicators. The clear lakes (Figure 7) formed three apparently natural groups which can be interpreted in the classical (oligotrophic-mesotrophic-eutrophic) sense. Nearly all the lakes of the Trail Ridge region comprise the oligotrophic Group A. These demonstrate a good withingroup similarity as denoted by the low objective function values at which they are joined. Themesotrophic Group B includes a few lakes from the Trail Ridge region (e.g. Kingsley and Winnott) which have been subjected to some cultural influence. As a whole the lakes in Group B (especially those from the Trail Ridge) are perhaps closer to being oligotrophic than eutrophic, but nonetheless they are distinctly (if slightly) more productive than the Group A lakes. The lakes of the eutrophic Group C include the 5 Oklawaha chain lakes plus the small eutrophic and hypereutrophic lakes in Alachua County. The latter are located primarily in urbanized areas, especially around Gainesville, and cultural sources seem to exert a heavy influence on their trophic states. The mesotrophic and eutrophic groups exhibit greater diversity in values for the individual trophic indicators and consequently are joined at higher objective function values. The colored lakes exhibited considerable diversity, and the results were not as interpretable in terms of classical trophic groupings, Perhaps this reflects a basic difference between colored and clear lakes; in this regard it should be noted that the position of dystrophic lakes in the usual trophic (i.e, nutritional state) classification has long been a subject of contention (Hansen, 1962; see Brezonik etal., 1969 for fUrther discussion). As previously mentioned.-We have considered color (roughly equivalent to dystrophy) as a major lake type parallel to a clear lake type (as proposed by Hansen, 1962, and Stewart and Rohlich,1967) and a range of nutrient states was considered possible for both. However the results of the cluster analysis imply that a simple har41

PAGE 46

VALUE OF THE OBJECTIVE FUNCTION -600 -500 -400 -300 -200 I i I I I (41) SAND HIll (42) MAGNOLIA (55) GAlUlEE (43) BROOKLYN I (47) SANTA ROSA -.....;.-----., 1----..., (54) COWPEN I (52) LONG LAKE (50) Me CLOUD (49) ANDERSONCUE (44) GENEVA (45) SWAN I -(7) CLEARWATER ----------------------, (53) WI NNOTT (13) STILL POND I (29) WATERMELON ---------'1 I GROUP A GROUP B (I) SANTA FE (40) "(NGSlEY (21) META (39) WEIR (25) No. 25 (36) HARRiS (37) EUSTIS (34) APOPKA DORA (38) GRIFFIN ( 8 ) HAWTHORNE' (20) No. 20 (32) WAUBERG (23) BIVINS ARM (24) CLEAR (22) ALICE I I I GROUP C L I 1 Figure 7. Cluster Analysis of 31 Low-Color (Clear) Lakes Considering 7 1rophic Indicators 42

PAGE 47

monious oligo ..., to eutrophic grada.tion may not :occur in highly colored lakes. Depending on where the. vertical line is drawn through the colored lake cluster diagram (FigureS) one can obtain clas.sificationscontaining anywhere from 2 to 6 or more groups. However none of these systems are completely satisfactory with regard to of the groups. Similarity line A in Figure 8 delineates a 5group classification for the purpose of the present sion; this system gives good within group similarity for groups 1, 2 and 3, and moderate within group similarity for group 4. Group 5 would appear to be a residual group whose lakes are only to the extent that they are different from the other groups. Lake Kanapaha is the most dissimilar of the colored lakes since it was the last lake to be incorporated into a group, and a seven group classification could also be drawn which would leave this lake in a group by itself. The five groups can be interpreted and labeled as follows: 1. oligotrophic, 2. meso-eutrophic, 3. oligo-mesotrophic, 4. dystrophic, 5. residual. Group 4 is labeled dystrophic because the lakes in this group are moderately to highly acidic and have high organic color and low dissolved solids. However pH was not one of the indicator variables, and dys trophy is not a lake type parallel to oligotrophy and eutrophy. Several of the lakes in this group are very shallow (mean depths of .1-1.5 meters) and. are partially covered with emergent and floating macrophytes (e.g. water hyacinths). These lakes (Palatka Pond and Tuscawilla, for example) could more accurately be described as senescent (bordering on tinction), but again this is not a recognized trophic state comparable to oligoand eutrophy. The remaining lakes (Group 5) would appear to be a residual group whose members (exceptfQr Beville's Pond and Lake No. 27, which are in fact similar) are alike only in being different from the other groups. Apparently there were not enough pairs or groups of lakes of nearly adjoining trophic characteristics to form groups with good within-group similarity. At a higher objective function value (i.e. lower degree of Similarity) (line B in Figure 8), three groups can be drawn: 1. oligotrophic, 2. mesotrophic, 3. In this scheme Lakes Lochloosa and Orange would appear mis classified and some of the dystrophic (i.e, low pH, high cQlor) lakes like Palatka and Calf Ponds are classified with obviously eutrophic lakes like NewnantsLake in spite of the low productivities and algal standing crop in the former. The. latter apparent misclassifications result from low Secchi disc transparencies (caused partly by high color) and in some cases from fairly high nitrogen and phosphorus levels, which, because of highc610r and low pH, 40 not produce algal blooms and high productivity. At a still lower level of sim,j.,larity (lineC in Figure 8) two colored lake groups can be formed; 1. and 2. eutrophic ...... dystrophic. 43

PAGE 48

(2) LIT. SANTA FE (4) ALTI-IO HICKORY P OND (46) WALL LAK E (10) No. 10 (II) MOSS LEE (15) ORANGE -(14) LOCHLOOSA (6) ELIZABETH (48) ADAHO (51) SUGGS (9) LIT. ORANGE (12) JEGGORD (31) BURNT (33) TUSCAWILL A (5) COOTER --, (19) CALF (16) PALATKA (30) LONG POND (26) BEVILLE'S (27) No. 27 (17) NEWNANS (18) MIZE (28) KANAPAHA VALUE OF THE FUNCTION -500 -400 -300 -200 -100 r I I i II I I I I r I -, r I J I I I I I }-I I I-..-.-! L 1-I I 3 2 1 I 14 y I-5 A 1 I I I I I B I I C 2 Figure 8. Cluster Analysis of 24 Colored Lakes Considering 7 Trophic Indicators 44

PAGE 49

This is not a Vel:'Y useful classification since the groups then conta,in highly dissimilar lakes and are nQteasily interpreted in terms of classical trophic gl:'OUps. ltis apparent that none of the colored lake classifications is ideal. The five group classification is not readily interpretable in terms of classical trophic groups, and the two group system has groups that are too broad to be of much use. The three group classification has some advantages in terms of interpreting classical trophic states, but some obvious misclassifications occur in this system. Mean values and standard deviations of the trophic indicators within the 3 clear lake subgroups and 5 colored lake subgroups delineated by the cluster analyses (Figures 7 and 8) are presented in Table 8. All seven indicators appear to reflect trophic levels reasonably well. Among the clear lakes indicator mean values without exception increase in each succeeding trophic group (from oligotrophy to eutrophy). Among the colored lakes the same trend is noted although some exceptions occur. There is little difference between mean values for the colored oligotrophic (Group 1) and oligo -mesotrophic (Group 3) groups; primary production and chlorophyll means are actually somewhat higher in the former than in the latter. The high values for the residual group derive from the bypereutrophic conditions in Newnan's and Kanapaha Lakes; the other lakes in this group have varied indicator values. It is interesting to note that the colored lakes have a much smaller range of values for most of the indicators compared to the clear lakes. Thus the clear oligotrophic lakes reflect much greater nutrient impoverishment than the colored oligotrophic lakes, and similarly the apparent degree of eutrophy is greater in the clear lakes. For example, the range of mean primary production values in the clear lakes is 1.3 to 150 mg C/m3_hr, while in the colored lakes the range is 9.7 to 55. B. DEVELOPMENT OF DISCRIMINANT FUNCTIONS TO CLASSIFY LAKES OUTSIDE THE ORIGINAL SAMPLE GROUP Discriminant functions were derived for the three trophic classes delineated by cluster analysis of the lakes and are presented in Table 9. In addition, the 55 lakes were grouped into five trophic categories ranging from ultraoligotrophic to hypereutrophic based on the trophic state index described in the next section. Discriminant functions were derived for these classes and are shown in Table 10. The colored lake group was too small and diverse to form meaningful discriminant functions. Using the criterion 45

PAGE 50

-l= 0\ Table 8. Mean Values and Standard Deviations of Trophic State Indicators Within Trophic State Groups Trophic State Indicators Primary Total Group Production Chlorophyll a Phosphate (mg Cm3-hr) (mg/m3 ) (mg-P/l) rCiTCiear-Lakes A. Oligotrophic 1. 3 a 1.8 .013 1. Ob .5 .003 B. Mesotrophic 5.8 4.3 .023 6.3 2.5 .014 C. Eutrophic 150.2 39.5 .306 119.5 26.3 .251 (b) Colored Lakes 1. Oligotrophic 11. 4 7.3 .032 9.0 3.0 .017 2. Meso-Eutrophic 24.1 19.5 .060 4.6 5.5 .003 3. OligoMesotrophic 9.7 6.7 .058 6.2 2.5 .035 4. Dystrophic 31. 6 21.1 .213 230.0 24.5 .278 5. Residual 55.1 35.6 .211 71. 9 9.5 .152 adenotes mean value bdenotes standard deviation Total Organic Nitrogen (mg-N/l) .25 ,08 .73 .30 1. 98 1.10 .70 .10 1. 24 .25 .72 .17 1. 36 1. 45 1.13 .68 Inverse Secchi Conductivity Cation Disc Ratio (m-l) .25 38.5 .61 .05 8.5 .08 .47 61. 5 .86 .24 35.1 .38 1. 72 244.0 3.61 1.12 129.0 2.13 .67 48.0 .60 .06 5.1 .17 1.07 82.2 1. 21 .11 6.8 .28 1. 24 47.6 .60 .25 6.6 .16 1. 59 46.0 1. 36 2.41 49.0 1. 57 1. 54 64.2 1. 67 .96 33.0 1. 95

PAGE 51

VAB b VAC VBC Table 9 Discriminant Eu'nctions for Trophic Grou,ps Delineated in Table8a 16 (l/SD) + .27(PP) + .04 (COND) ,.... 62 (TP) 34(TON) 4.4(l/CR) -.74(CHA) + 14.3 = 65(l/SD) + .77(PP) + .10 (COND) 230 (TP) ,... 108(TON) 20 (l/CR) -2.9 (CHA) + 126 = 4g(1/SD) + .51(PP) + .06(COND) -167(TP) ,.... 75(TON) -. 16 (l/CR) 2.2 (CHA) + 112 alndicator abbreviations: l/SD = inverse Sec chi disc, PP = primary production, COND = specific conductance, TP = .total phosphate, TON = total organic nitrogen, l/CR = inverse of Pearsall's cation ratio, CHA = chlorophyll a, all indicators in units given in Table 8. -bSubscripts of discriminant functions indicate the groups (from Table 8) being compared. 47

PAGE 52

Table 10 Discriminant FunctiQns.fQr Five Groups of Lakes D.etermined from TrQphicSt.ate. Indicator Ranges (From Table 18) VHE b = 23.99(1/SD)a + .47(COND) + 15.17(TON) + 148.76(TP4) + .38(pp) .o4(CHA) + 8.19(1/CR) 289.io VHM VHO VEM YEO VEU VMO VMU YOU = = -22.44(.1/Sb) + .53(COND) + 35.93(TON) + 289.93(TP4) + 91(PP) + .77(CHA) + 11.11(I/CR) -369.94 + .59(COND) + 74.56(TON) + 427.32(TP4) + .59(PP) + 2.45(CHA) + 23:39(1/CR) -451.87 = -44.33(1/SD) + .64(COND) + 111.06(TON) + 490.95(TP4) .33(PP) + 3.27(CHA) + 26.56(1/CR) -475.42 = = = = = = ,..,46,43(l/SD) + .06(COND) + 20.77(TON} + 141.17(TP4) + .53(PP) + ,81(CHA) + 2.92(1/CR) 80.85 -58.87(1/SD) + .12(COND) + 59.40(TON) + 278.56(TP4) + .21(PP) + 2.49(CHA) + 15.19(l/CR) 162.78 -68.32(1/SD) + .17(COND) + 95.90(TON) + 342.20(TP4) -.OS(PP) + 3.31(CHA)" + 18.37(1/CR) 186.32 -12.44(1/SD) + .07(COND) + 38.63(TON) + 137.39(TP4) -.32(PP) + 1.68(CHA) + 12.28(1/CR) 81.93 -21.88(1/SD) + .12(COND) + 75.13(TON) + 2bl.03(TP4) -.58(PP) + 2.50(CHA) + 15.4S(1/CR) -ioS.48 + + 36.50(TON) + 63.64(TP4) -.26(pp) + .82(CHA) + 3.18(1/CR) -21.55 a Key to indicator abbreviations identiDal to Table 9 bSubscripts of discriminantfunctlons refer to groups labeled in Table 18: hypereutr.ophic (H), eutrophic (E), mes6trophic (M), oligotrophic (0), ultraoligotrophic(U) 48

PAGE 53

scribed in Chapter 2, a lake belongs to Group A (oligotrophic) if VAB and VAC' the respective discriminant functions between the groups in Table 9, are both greater than or equal to zero. To demonstrate the application of this technique to lake classification, trophic indicator data for three well known North American lakes (Table 11) were assembled from various sources, and the lakes were classified according to the discriminant functions of Tables 9 and 10. Data for Lake Tahoe was taken from Ludwig et al. (1964) and Goldman and Armstrong (1968). Data for the-rwo-great lakes was obtained from Saunders (1964), Putnam et al. (1966) and Beeton (1969). As expected, Lakes Tahoe and-Superior were assigned to the oligotrophic class and Lake Erie to the eutrophic class using the discriminant functions for the clear lakes (Table 9). The discriminant functions of Table 10 derived from all 55 lakes classified Lake Tahoe in the ultraoligotrophic group (D), Lake Superior with the oligotrophic lakes (0), and Lake Erie in the mesotrophic group (M). It should be emphasized that use of these three lakes is for illustrative purposes only. The validity of assigning large temperate lakes into classes delineated from a sample of small sub-tropical lakes has not been tested. Certainly the general effects of eutrophication are similar in all "normal" lakes, and in this sense the examples are not inappropriate. However, if geographically broad (or universal) trophic groups are to be delineated, the original sample group be similarly broadly based, which of course the lakes to develop the discriminant functions are not. A further word of caution regarding this method is the deleterious effect of small sample size on the probability of misclassification (Wallis, 1967). For good differentiating power the functions should be based on sample groups of 50 or more. C. FORMULATION OF TROPHIC STATE INDICES The multivariate statistical method of principal component analysis (Chapter 2) represents one means of deriving a single numerical index from a number of indicators, and this technique was used to derive indices from the trophic indicator data for the 55 lakes. As seen in Figure 6, organic color can be used to separate lakes into the two fundamentally different classes of colored and clear lakes. Nutrient enrichment may cause different effects in each class, i.e., various trophic indicators may respond differently to nutrient enrichment in clear vs. colored 49

PAGE 54

Vl 0 Lake Tahoe Superior Erie Table 11. Trophic Characteristics and Classification of Three Well-Known North American Lakes by Discriminant Functions in Tables 9 and lOa l/SD PP m-I mg C/m3-hr .04 0.5 .10 8.0 .29 59 COND ymho/cm 83 79 313 TP mg P/l .007 .014 .060 TON mg N/l .09 .14 .48 l/CR 1.4 5.1 CHA Trophic mg/m3 Class (Table 9) 1.5 A (oligo-trophic) 2.5 A (oligo-trophic) 4.7 27.5 E (eu-trophic) alndicator abbreviations as in Table 9. Trophic Class (Table 10) U (ultraoligo-trophic) o (oligotrophic) M (mesotrophic)

PAGE 55

lakes, and a single trophic index for all lakes could possibly be inappropriate. Consequently separate trophic state indices were developed from the correlative relationships of indicators within each of the two basic classes defined previously by cluster analysis. The annual mean values for each lake (Table 7) were used in the derivation of the indices. Table 12 lists the means and standard deviations of each indicator for all 24 colored lakes and all 31 clear lakes, and Table 13 presents the respective correlation matrices. The first principal components, y co andy 01' extracted from the colored and clear lake correlatlon matrlces, respectively, are shown in Table 14. The first principal components extract a good portion of the information from the R's since Yco and Ycl explain 72 and 71% of the total variances in their respective correlative matrices. The principal components are simple linear functions of the 7 trophic state indicators with weighting factors for each indicator. The indicator values are standardized values (i.e. the actual raw value from Table 7 minus its mean value and divided by its standard deviation from Table 12). The trophic state index for each group of lakes, i.e. TSIco and TSIcl' was derived by slightly modifying the respective first principal components. The modification consisted in adding a constant value to the principal component so that the TSI would always be greater than zero. The constant was obtained by evaluating the first principal component with raw data values of zero for each indicator. A zero raw data value results in a negative standardized value and hence a negative value for y, which value was then added to y to produce the TSI (see Table 14). Hence a hypothetical lake with zero productivity and zero values for the other indicators would then have a TSI of zero. (In actuality this would never occur since even pure water has a finite Secchi disc transparency, and all natural waters have a non-zero cation ratio.) In general, lakes with increasingly positive indicator values will exhibit correspondingly higher TSI's. The TSI's of the lakes in each group were calculated by substituting the standardized indicator values into the appropriate TSI formula, and Table 15 presents the results for each group ranked in descending order of TSI value. Thus the above analysis indicates that Lake Kanapaha is the most eutrophic colored lake and Wall Lake is the least eutrophic in this group; similarly Lake Apopka (eutrophic) and Lake Santa Rosa (oligotrophic) represent the extremes of trophic state within the clear lake group. The results of the cluster analyses (Figures 7 and 8) are also included in Table 15 for comparative purposes. Rankings of the clear lakes according to their TSI's are in excellent agreement with the clear lake groups formed by cluster analysis. The first 12 lakes (in order of decreasing TSI) 51

PAGE 56

\.Jl I\.) Table 12. Means and Standard Deviations of Trophic Indicators Within the Colored and Clear Lake Groupsa Group l/SD COND TON TP PP CRA l/CR Colored Lakes 1.13b 53.2 .99 .119 25.0 16.7 .99 .53c 20.2 .42 .130 37.8 12.6 .94 Clear Lakes .88b 124.0 1. 04 .129 60.1 17.0 1.83 .97c 126.1 1. 04 .209 102.7 24.2 1. 94 aSee Table 7 and text for explanation of symbols and units of expression. bDenotes the mean. cDenotes the standard deviation.

PAGE 57

(a) l/SD COND TON TP PP CRA l/CR (b) l/SD COND TON TP PP CRA Table 13. Correlation Matrices of Seven Trophic Indicators for Colored and Clear Lake Groups Colored Lakes: l/SD COND TON TP PP CHA 1. 000 .630 .720 .534 .782 .646 1. 000 .627 .484 .733 .517 1. 000 .643 .818 .658 1. 000 .640 .596 1. 000 .705 1. 000 Clear Lakes: 1.000 .643 .931 .559 .962 .858 1. 000 .621 .888 .638 .603 1. 000 .481 .915 .813 1. 000 .586 .543 1.000 .910 1. 000 l/CR 53 l/CR .697 .792 .764 .685 .800 .529 1. 000 .464 .522 .442 .396 .402 .392 1. 000

PAGE 58

Table 14. First Principal Components (Yeo and Ycl) and Trophic State Indices (TSIco ana TSIcl) (a) Colored Lakes; Yco = .848(1/SD) + .809(COND) + .887(TON) + .768(TP) + .930(PP) + .780(CHA) + .893(1/CR) Cumulative Percent of Total Variance Explained by Ycol = 72% TSIcol = Ycol + 9.33 (b) Clear Lakes: Ycl = .936(1/SD) + .827(COND) + + .748(TP) + .938(Pp) + .892(CHA) + .579(1/CR) Cumulative Percent of Total Variance Explained by Ycl =71% TSIcl =Ycl + 4.76 54

PAGE 59

Table 15. Lakes of Clear and Colored Groups Ranked According to TSIcl and TSIco a. Clear Lakes Lake Apopka Twenty Dora Bivin's Arm Griffin Alice Eustis Hawthorne Clear Wauberg Harris Twenty-five Watermelon Weir Meta Clearwater 18.1 15.1 14.6 12.0 10.7 9.2 8.2 7.9 7.3 6.3 5.3 5.1 2.9 2.7 2.4 2.1 b. Colored Lakes TSIco Kanapaha Burnt Newnan's Lochloosa Cooter Calf Pond Mize Tuscawilla Orange Twenty-seven Little Orange Elizabeth 27.9 17.0 15.3 12.0 11. 0 10.6 10.5 10.4 9.9 9.2 8.0 7.9 Cluster Group C C C C C C C C C C C C B B B B 3 3 3 1 3 3 3 3 1 3 2 2 55 Lake Santa Fe Still Pond Winnott Kingsley Geneva Gallilee Swan Anderson-Cue McCloud Brooklyn Cowpen Long Sumter-Lowry Magnolia Santa Rosa TSIcl 1.9 1.6 1.4 1.3 1.2 1.2 1.1 1.1 1.0 1.0 1.0 0.9 0.9 0.9 0.8 TSTco Ten 6.9 Palatka Pond 6.9 Jeggord 6.7 Moss Lee 6.3 Beville's Pond 6.2 Suggs 6.2 Adaho 6.1 Long Pond 6.1 Altho 6.0 Little Santa Fe 5.8 Hickory Pond 5.6 Wall 5.3 Cluster Group B B B B A A A A A A A A A A A 1 3 2 1 3 2 2 3 1 1 1 1

PAGE 60

correspond to the lakes in eutrophic group C of Figure 7; the next 8 lakes are in mesotrophic groupB, and the last 11 lakes comprise oligotrophic group A. Thus the TSI for clear lakes can be used to separate classical trophic states quantitatively. A TSlcl of about 5.0 would appear to be the dividing line between mesotrophy and eutrophy, and a value of about 1.2-1.3 separates mesotrophy and oligotrophy. Qualitative inspection of other trophic indicators for Lakes Kingsley and Winnott suggests these lakes are more typically oligotrophic thanmesotrophic and the TSI dividing line should perhaps be raised to 1.5. The colored lakes ranked according to TSlco are in general agreement with the cluster analysis (Figure 8), but some discrepancies are noted. For example, Lakes Lochloosa and Orange have a high degree of similarity; however, the lakes have high but somewhat dissimilar TSI values, and four lakes have TSI rankings between the values for the two lakes. Also Beville's, Palatka and Long Ponds were clustered into eutrophic groups although their TSI values indicate oligotrophy. The discrepancies in comparing the two analyses probably arise within the cluster analyses since the colored lakes exhibited considerable diversity and did not form groups with good within-group similarity. For management and identification purposes it would be desirable to have a single trophic state index to rank all lakes regardless of color. Large differences in the specific conductance, primary production and cation ratio mean. values for the two groups (Table 8) and the cluster analysis of basic chemical parameters (Figure 6) suggest a basic difference which could possibly cause different trophic indicator responses in the two types. On the other hand, that the two groups can be viewed as two samples of one (larger and more diverse) population, and a single TSI to rank all 55 lakes was developed under this assumption. Of the seven indicators used to assess trophic state in this study, the one most directly affected by organic color is Secchi disc transparency. This parameter is essentially a function of and turbidity, and a multiple regression of inverse Secchi disc reading as dependent variable vs.color and turbidity as independent variables produced the following relationship; l/SD =0.003(Col) + 0.152(Tur) (12) Data for the analysis were from Table 7, and the zero intercept option was used in the regression analysis. The relationship is significant at the 99% confidence level, and the percent Df variation in l/SD explained by Eq.12 is 96%. Using Eq. 12, a color value of 75 mg/l, and turbidity values from Table 7, new color-scaled inverse Secchi disc values were calculated for each of the 55 lakes; the results are listed in Table 7. A color value of 75 mg/l was chosen for the 56

PAGE 61

scaling purposes because it represents the dividing line between clear and colored lakes and is also in the middle range (zone of best prediction) of the regression equation. Once the Secchi disc values had been color scaled, the correlative relationships between the seven trophic indicators for all 55 lakes were subjected to a principal component analysis, and the means, standard deviations and the correlation matrix are given in Table 16. The first principal component Yt extracted from R is given by Yt =.919(1/SD) + .800(COND) + .896(TON) + .738(TP) + .942(PP) + .862(CHA) + .634(1/CR) (13) Yt extracts a good portion of the information from Rand explains 70% of the total variation in R. The TSI is given by TSI =Yt + 5.19, (14) where the value of 5.19 was determined as described previously in the derivations of TSIco and TSlcl' TSI1s were calculated for each of the 55 lakes by substituting the standardized indicator values (computed from Tables 7 and 16) into Eqs. 13 and 14. The lakes are ranked in descending order of TSI in Table 17. Using the cluster analyses of Figures 7 and 8 as a guide, the 55 lakes were separated in terms of classical trophic state terminology into five groups as follows: 1. Hyper-eutrophic (TSI>iO), 2. Eutrophic 3. Mesotrophic (7)TSIL3), 4. Oligo,...., trophic C3 >TSI.L2), 5 Ultra-oligotrophic (TSI<2). The se groups are delineated and labeled in Table 17. The relative rankings of the lakes in the TSlcl and TSI 0 of Table 15 are also shown in Table 17. 8omparison shows that the clear lakes are ranked in almost identical order ing to the total (55 lake) TSI (excluding the interspersed colored lakes) as they are by the TSlcl' Further it is vious that the clear lakes as a group are more extreme in their trophic behavior than are the colored lakes; all but one of the hypereutrophic lakes and all the ultraoligotrophic lakes in Table 17 belong to the clear lake group. Nearly all the colored lakes are included in the oligotrophic and mesotrophic categories. Comparison of the colored lake rankings according to the 55 lake TSI and the TSlc also indicates a general correspondence. The lake most ouB of order is Lake Twenty-seven, which is the fourth listed colored lake in Table 17 and the tenth ranked lake according to the TSlco' Many of the other colored lakes are "misranked" by one or two places, but there are no major discrepancies. Most of the changes in relative rankings between Tables 15 and 17 probably 57

PAGE 62

Table 16. Means, Standard Deviations, and Correlation Matrix of Trophic State Indicators for 55 Lakes Standard Indicator Mean DeViation l/SD .84 .77 COND 93.1 101. 3 TON 1. 02 .82 TP .125 .177 PP 44.8 82.3 CRA 16.9 19.8 l/CR 1. 47 1.63 Correlation Matrix R: l/SD GOND TON TP PP CRA l/CR l/SD 1.000 .617 .880 .542 .927 .784 .502 COND 1. 000 ,582 .762 .654 .540 .560 TON 1. 000 .500 .890 .788 .474 TP 1.000 .576 .553 .440 PP 1.000 .859 .478 CRA 1. 000 .402 l/CR 1.000 58

PAGE 63

\Jl \D Lake 1. 2. 3 Table 17. Fifty-five Florida Lakes Ranked According to Trophic State Index (TSI) Hypereutrophic group Apopka Twenty Dora Bivin's Arm Griffin Kanapaha Alice Eustis Eutrophic group Hawthorne Clear Bur.nt Fond Wauberg Newnan's Mesotrophic group Twenty-five Harris Twenty-seven Rank in TSI Table 151 Lake 22.1 lS.5 lS.5 14.7 13.7 13.5 10.7 10.5 9.1 S.S S.3 7.4 7.1 6.4 6.3 5.S lA 2A 3A 4A 5A IB 6A 7A SA 9A 2B lOA 3B 12A llA lOB 4. Cooter Pond Lochloosa Tuscawilla Calf Pond Orange Mize Watermelon Pond Little Orange Weir Elizabeth fJ;1en Palatka Pond Beville's Pond Meta Oligotrophic group Jeggord Moss Lee Long Pond Clearwater Altho Hickory Pond Santa Fe Rank in TSI Table 151 5.3 5.2 4.S 4.6 4.3 4.2 3.6 3.4 3.3 3.2 3.2 3.2 3.1 3.IL 2.8 2.8 2.8 2.6 2.5 2.5 2.5 5B 4B SB 6B 9B 7B 13A lIB 14A 12B 13B 14B 17B 15A 15B 16B 20B 16A 21B 23B 17A (cont'd.)

PAGE 64

Table 17 (cont'd.) Rank in Lake TSI TablelS1 Suggs 2.3 18B Little Santa Fe 2.3 22B Adaho 2.2 21B Wall 2.1 24B Winnott 2.0 19A 5. Ultra-oligotrophic group Still Pond 1.9 18A Kingsley 1.9 20A 0\ Geneva 1.8 21A 0 Gallilee 1.6 22A Swan 1.5 23A Anderson-Cue 1.5 24A McCloud 1.5 25A Brooklyn 1.5 26A Cowpen 1.5 27A Long 1.3 28A Sumter-Lowry 1.3 29A Magnolia 1.3 30A Santa Rosa 1.3 31A lRank from Table 15 according to TSIcl (A values) and TSIco (B values).

PAGE 65

result from the use of color-corrected Sec chi disc transparencies for the TSI values in Table 17, which presumably should produce a more accurate relative ranking of the lakes according to their trophic states. The question concerning the soundness of one TSI for both clear and colored lakes remains. A definitive answer is perhaps impossible. However, the first principal component on which the 55 lake TSI is based accounts for about as much the variance (70%) in the correlation matrix of trophic indicators for all lakes as do the first principal components for the clear and color groups, which accounted for 71 and 72% of the variances in their respective correlation matrices. Further, there appear to be no obvious misclassifications or misrankings in the 55 lake TSI's. One of the major values of the TSI concept is the possibility of ranking rather diverse objects (lakes) in a logical and objective manner. Obviously if the sample is too diverse, the rankings will have little or no meaning. Thus extrapolation of the TSI concept to development of a single, universal index for all lakes is not suggested. To rank Arctic bogs, acid volcanic lakes, tropical ponds and the Great Lakes on the same scale would be pointless and meaningless. On the other hand, the more "harmonious" the sample, the more meaningful and logical (and easier) it will be to rank the objects. The harmonious series of clear lakes is easily and logically ranked (Table 15). Inclusion of the colored lakes produc.esamore diverse sample with an inevitable loss in clarity in interpretation of the resulting TSI. Nevertheless, it is felt that the 55 lake TSI is a useful, interpretable and logical means of ranking Florida lakes. Some interesting features of the TSI rankings deserve mention. Lake Alice has been ranked in the hypereutrophic group although it might be classified oligotrophic on the basis of plankton productivity alone. Lake Alice has extremely high nitrogen and phosphorus concentrations and supports a profuse growth of water hyacinths, which along with a short hydraulic detention time (in the order of 2-3 days), have restricted plankton productivity. In this case, the other trophic state indicators (nitrogen, phosphorus, and conductivity) have been sufficiently high to counteract the low primary production and chlorophyll a values. Lake Twentyseven was also ranked higher than it would be on the basis of plankton production alone. This lake is almost completely covered with duckweed and as with Lake Alice the other indicators have counteracted the low primary pro duction value, The usefulness of the trophic state index can be best determined by its application, e.g. in practical (e.g. management and control) situations or in development of empirical m,odels relating trophic state to watershed enrichment factors 61

PAGE 66

(see the next section). However, the validity of the approach can be inferred from closer inspection of the TSI and its component parameters. Table 18 presents the means and 95% confidence intervals for the 7 trophic indicators in each of the classes delineated in Table 17. In nearly every case the mean parameter values increase in progressing toward more eutrophic classes. However the large confidence intervals for most parameters implies considerable overlap between the classes delineated by any single indicator. These facts demonstrate three important points. First, because of -the overlap, any single parameter is inadequate to define trophic state or trophic classes. Second, the wide and overlapping ranges of indicator values preclude easy placement of lakes into appropriate trophic classes since the values for a lake could fit within the confidence intervals of the parameters in two adjacent classes. Finally, the increasing mean values in progressing toward eutrophic conditions imply that the TSI provides at least an objective means of placing lakes into appropriate trophic classes and suggests that the relative ranking of the lakes by their TSI values is reasonable. The TSI described above reflects the general trophic conditions of Florida lakes; whether it is the best index that can be developed will have to be answered by further work comparing its attributes with those of other indices that might be developed. The seven indicators in the present index reflect the major limnological consequences of eutrophication with the exception of macrophyte problems. Indices with fewer variables would reflect a narrower concept of trophic state and would be more likely to yield misleading results. Specific water quality problems resulting from eutrophication are not directly considered by the index, but some of the indicators are indirectly related to such problems. For example, chlorophyll a, a biomass parameter, might be correlated with taste and odor arising from algal blooms; Secchi disc transparency is associated with water turbidity, which should be correlated with the length of sand filter runs in water treatment plants. Perhaps other indices could be developed which would be directly related to water quality problems, but it is not always a simple matter to find appropriate quantitative indicators for such purposes. The index described above should be practical for routine assessment of general trophic conditions since the individual parameters are commonly and rather simply measured. The only exception possibly is primary production. This parameter, while of fundamental significance to the trophic state concept, also suffers from the fact that measured values are highly variable in a given lake and are greatly dependent on physical factors such as light and temperature. Perhaps a simpler TSI not incorporating this parameter would prove 62

PAGE 67

Table 18. Confidence Intervals for Trophic lndicators in Five Lake Groups Delineated by Trophic State Index Valuesa Parameter Ultraoligotrophic Oligotrophic Mesotrophic Eutrophic Hypereutrophy TSI Range 1.3-1.9 2.0-2.9 3.0-6.9 7.0-9.9 >10.0(10.0-22-;1) Number of Lakes 13 12 17 5 8 Primary Production mg C/m3-hr 1. 3 .7 8.6 3.3 17.3 8.5 95.4 10 205 94 Chlorophyll a mg/m3 -1.9 .4 7.7 2.3 19.5 7.5 39.4 15.8 42.7 21.7 0\ Total Phosphate LA) mg P/l 0.13 .002 .040 .012 .141 .085 .246 .221 .424 .192 Total Organic Nitrogen mg N/l .29 .08 .78 .10 1. 08 .23 1. 58 .29 2.41 .96 (Secchi Disc)-l m-1 .43 .02 .55 .08 .73 .22 .94 .16 2.31 .98 Specific Conductivity llmho cm-1 39.6 5.3 50.6 11.5 80.2 39.8 98.6 62.2 297 101 [Ca] + [Mg] .65 .10 .69 .13 2.35 2.27 1. 70 .9'J 3.60 .64 [Na] + [K] aValues represent means % confidence interval.

PAGE 68

more useful to governmental agencies faced with evaluating the trophic characteristics of large numbers of lakes. CHAPTER 5. RELATIONSHIPS BETWEEN TROPHIC STATE AND WATERSHED ENRICHMENT FACTORS A. INTRODUCTION Empirical relationships between lacustrine trophic conditions and watershed conditions can be developed by regression analysis using the TSI as dependent variable and appropriate conditions in the watershed as independent variables. A general model for eutrophication can be written as: TS =j"(N,M,H,S,t ... ), where TS is the trophic state resulting from nutrient (N) loading (nitrogen, phosphorus and other essential nutrients), M represents morphometric characteristics such as mean depth, H represents hydrological conditions (e.g. water detention time), S is a sedimentation factor, and t is time. The relationships among these parameters is presently too vague for the development of functional relationships. However, simplified empirical approximations of Eq. 15 can be developed. As a first approach models of the type TSI = g(N,P) + C, (16) were developed, where the TSI described in Eq. 14 represents the trophic state parameter of Eq. 15, Nand P represent annual nitrogen and phosphorus loading rates, and C is an uncertainty term. Although nitrogen and phosphorus are not the only nutrients required for algal growth, it is generally agreed that they are the two main nutrients involved in the lake eutrophication process. In spite of current controversy over the role of carbon (Bowen, 1970; Legge and Dingeldein, 1970; Kerr et al., 1970), researchers as a whole regard phosphorus as thelTIost frequent limiting nutrient in lakes. Vollenweider (1968) and others have emphasized the importance of nutrient (particularly nitrogen and phosphorus) supply in determining a lake's trophic state. Although various lake factors, such as mean depth, detention time, basin shape, and sedimentation rate, affect the amounts of nutrients a lake can assimilate, nutrient budget calculations represent a first step in quantifying this dependence. 64

PAGE 69

A few lacustrine nitrogen and phosphorus budgets have been reported in the literature, e.g. Rohlich and Lea (1949) for Lake Mendota, OOcGauheyetal. (1963) for Lake Tahoe and Edmondson (1968) for Lake Washington. Vollenweider (1968, 1969) has summarized most of the budget calculations for American European Lakes. Comprehensive evaluation of the nutrient balance for a lake requires measurement of all potential nutrient sources and sinks (Table 19) over an extended period in order to assess seasonal and other effects. Some sourceB and e.g. groundwater, nitrogen fixation denitrification, require elaborate sampling and experimental procedures to be adequately evaluated. Consequently, manpower and time constraints have resulted in very few complete nutrient balances being attempted. An alternative and simpler method is to use literature estimates for nutrient exports from various sources and information on the various land use and population characteristics of the lake watershed. This approach was used by Leeet al. (1966) for nitrogen and phosphorus budget calculations for Lake Mendota. While perhaps not as accurate as actual measurement, there is no other realistic alternative when evaluating budgets for a large number of small lakes. B. NITROGEN AND PHOSPHORUS BUDGETS Partial nitrogen and phosphorus budgets for the 55 lakes in the study group have been computed by this latter approach. The budgets are referred to as partial since no attempt was made to account for such sources as nitrogen fixation, leaves and pollen and groundwater. Adequate data were not available to evaluate most sinks, and consequently none were considered. The partial budget calculations therefore estimate gross supply or loading. The morphometric, land use, and population figures for each lake were determined according to the methods described in Chapter 2. Watersheds were divided into forest, urban, pasture, fertilized cropland, and cleared unproductive areas. Table 20 lists the pertinent watershed and morphometric data for each lake. Literature figures for the expected contributions of nitrogen and phosphorus from the various sources were compiled, and the values used in this study are summarized in Table 21. Where applicable, each value is accompanied by the literature reference. Literature estimates were not available for two sources. Muck (recovered marshland) and citrus farm contribu,.... tions were calculated. from average fertilizer composition and application rates, assuming that 10 percent of the applied nitrogen and one percent of the applied phosphorus was exported from the soil to the lake. The figures for percentage 65

PAGE 70

Table 19. Potential Nitrogen and Phosphorus Sources and Sinks for Lakes -Sources Natural Precipitation on Lake Surface Swamp Runoff Virginal Meadowland Runoff Forest Runoff Soil Erosion Aquatic Bird and Animal Wastes Leaf and Pollen Deposition Groundwater Influxes Not FO to 1 r.ogen lxa lon Sediment Recycling (b) Sinks Outlet Losses Fish Catches Aquatic Plant Removal *Applies to nitrogen alone. 66 Gultural Domestic and Industrial Waste Waters Agricultural Runoff Managed Forest Runoff Urban Runoff Septic Tanks Landfill Drainage Denitrification* Volatilization* Ground Water Recharge Sediment Losses

PAGE 71

Mean Forested Depth Area Name/No. (m. ) (ha. ) Santa Fe 1 5.5 4424 Lit. Santa Fe 2 4.8 842 Hickory' 3 3.4 95.5 Altho 4 3.6 666 Cooter 5 2.2 487 Elizabeth 6 1.5 156 Clearwater 7 1.5 18.1 Hawthorne 8 2.8 53.6 Lit. Orange 9 2.8 525 Unnamed 10 3.2 70.0 Moss Lee 11 3.6 148 Jeggord 12 3.0 207 Still 13 1.1 10.4 Lochloosa 14 2.9 17766 Orange 15 1.8 26405 Palatka 16 0.8 18.2 Newnan's 17 1.5 22136 Mize 18 4.0 15.5 Calf 19 1.6 100 Unnamed 20 1.9 38.4 Meta 21 1.6 8.2 Alice 22 0.9 56.8 Bivin's Arm 23 1.5 378 Clear 24 1.6 15.9 Unnamed 25 1.0 9.3 Beville's 26 3.1 12.5 Unnamed 27 3.8 26.3 Kanapaha 28 0.7 4043 Watermelon 29 1.5 979 0\ -..::J Table 20. Population and Land Use Data for 55 Florida Lake Unproductive Urban Fertilized Pastured Cleared Area Cropland Area Area (ha.) (ha.) Cha.) (ha. ) 191 60.6 206 137 0 61.3 109 72.8 0 29.3 119 0 13.8 17.1 21 21 0 0 627 0 0 0 5.2 0 0 0 0 0 38.0 0 0 26.8 0 108 524 786 0 7.7 0 0 0 0 7.7 5.2 0 0 23.2 15.5 0 0 5.2 0 81.6 201 1232 1232 182 488 1499 2298 0 0 0 13.0 876 71.3 1549 2324 0 0 0 0 8 0 0 8.1 16.5 0 0 2.1 4.9 0 0 8.4 288 129 0 0 256 72.8 85.4 0 15.2 0 0 0 1.6 0 0 34.0 5.2 0 Q 9.4 0 0 b 20.2 1087 0 821 821 0 0 10q 70.5 Population Remote Served Cultural Cultural by STpa Unit:s Units Facilities 2091 91 0 34i 24 0 0 6 0 131 58 0 01 19 0 61 3 0 I O! 0 0 10i 120 0 4: 109 0 01 0 0 01 0 0 4' 0 0 0 1 0 0 961 371 0 54i 381 0 i oi 0 0 I 79 792 0 01 3 0 d I 29 0 31 11 0 0 I 0 0 a 0 5100 16 1 91 0 1Ji 7 0 d I 1 0 3! 27 0 d 7 0 e 679 0 3 4 0 I (cont'd.)

PAGE 72

Table 20 (cont'd.) Unproductive I Population Mean. Forested Urban Fertilized Pastured Cleared Innn.ed t e Remote Served Depth Area Area Cropland Area Area Cu1tur:a1 Cultural by STpa Name/No. (m. ) (ha. ) (ha. ) (ha.) (ha.) (ha.) unitls Units Facilities Long Pond 30 1.2 43.7 0 0 20.2 17.8 01 0 0 Burnt 31 2.2 129 0 0 55.4 38.8 21 18 0 Wauberg 32 3.8 258 16.1 0 123 0 6) 4 0 I Tuscawi11a 33 1.3 963 66.3 0 154 103 11 59 0 Apopka 34 1.3 2384 467 17508 0 0 274j 1157 6950 Dora 35 3.0 1233 931 6762 0 10 3421 507 6500 Harris 36 4.2 5979 675 8672 0 3612 438 690 0 Eustis 37 4.1 1683 722 5271 0 900 3551 554 7740 Griffin 38 2.4 5157 679 9605 0 1187 4151 239 13850 tveir 39 6.3 320 139 1168 0 0 105 0 Kingsley 40 7.3 503 328 0 0 96.8 266 120 0 Sand Hill 41 4.8 16 0 0 0 0 d 0 0 Magnolia 42 8.0 484 0 0 0' 0 d 0 0 I Brooklyn 43 5.7 667 32.3 0 0 0 1671 24 0 Geneva 44 4.1 741 205 0 0 2'70 388 0 I Swan 45 4.8 460 0 0 0 97.5 1071 7 0 Wall 46 2.1 401 12.1 0 73.2 114 d 25' 0 I Santa Rosa 47 8.1 122 0 0 0 0 26 15 0 Adaho 48 3.5 369 0 0 41..3 0 JJ 0 0 I McCloud 49 2.0 40.5 0 0 0 11.3 d 0 0 Anderson-Cue 50 2.0 48.9 0 0 0 8.1 d 0 50 Suggs 51 2.5 658 0 24.9 23.2 34.8 d 6 0 I Long 52 3.4 547 0 0 0 0 1q 5 0 Winnott 53 5.2 216 0 15.5 23.8 45.0 26 0 Cowpen 54 3.7 712 0 0 0 211 110 0 Galli1ee 55 3.5 213 0 0 0 150 II 18 0 I a Sewage Treatment plant 0\ (X)

PAGE 73

Table 21. Expected Quantities of Nitrogen and Phosphorus from Various Sources Quantity Source .:Reference Nit.roge.n Domestic Sewage Vollenweider (196S) 3940a Fertilized Area Citrus Farms 2'.'24 b Muck Farms .11b Pastured Area Miller (1955) .S5b Unproductive .18b Cleared Area Brink (1964) FClrested Area. Sylvester (1)61) .24b Urban Area Weibel (1969) .SSb Rainfall Brezonik etal. (1969) .58c -Septic Tanks Immediate 24 20 d Remote 970 d Domestic Ducks Paloumpis & Starret (1960) 480 e agrams/capita year bgrams/square meter of land use area year Cgrams/square meter of lake area '"" year dgrams/septic tank ,..., year egrams/duck year 69 of Quantity of PhorphoruB 795 a .OlSb .135b .018b .OO6b .OO8b .110b .o44c 13Sd 13.Sd 90 e

PAGE 74

fertilizer losses were reported by Vollenweider (1968) and, although approximate, probably represent lower limits. Septic tank contributions were estimated using a similar procedure. An average septic tank was assumed to have a daily effluent volume of 475 liters with total nitrogen and phosphorus con centrations of 35 mg/l and 8 mg/l,respectively (Folta, 1969). For septic tanks associated with immediate cultural units, it was estimated that 25 percent of the nitrogen and 10 percent of the phosphorus in the effluent were exported to the lakE:;--,_For_rernote cultural_unit_septic tanks it w_a_s_estirnat_ed that 10 percent of the nitrogen and I percent of the phosphorus discharged eventually reached the lake. from domestic sewage are expressed in Table 21 as the amount per capita per year. These sewage figures were used only when effluent records for the individual plants were not available. One lake (Mize) harbors a colony of 50 domestic ducks; estimated nitrogen and phosphorus contributions from ducks are thus listed in Table 21. Several large lakes, e.g. Griffin and Apopka, receive nitrogen and phosphorus via citrus processing plant effluents. The magnitude of the contributions were determined from average plant flow rates and concentrations (Environmental Engineering, Inc., 1970). Th caleulateEl nitrogen and phesphepus leading pates for each of the 55 lakes are presented in Table 22 expressed as grams per cubic meter of lake volume per year. Loadings expressed per unit lake surface may be obtained by simply multiplying the volumetric loading by lake mean depth (from Table 20). In Florida lakes mean depths rarely exceed 5 meters and most lakes are completely mixed year round. Consequently most of the analyses reported here pertain to the volumetric loading rates. In general, the results indicate a positive correlation between nitrogen and phosphorus supply and trophic state as quantified by the TSI, but several discrepancies are evident. Lakes Alice (22) and Kanapaha (28), although demonstrating hypereutrophic characteristics, have nitrogen and phosphorus loadings at least an order of magnitude higher than any of the other hypereutrophic lakes. This can be attributed to the fact that both lakes have had their natural watersheds increased by cultural activities, which have resulted in very short detention times for the lakes. Lake Alice receives 1 to 2 million gallons per day of sewage effluent and 10 to 12 million gallons per day of cooling water from University of Florida facilities. Lake Kanapaha, which is connected with a sinkhole draining an urbanized stream, has had its watershed enlarged 2 to 3 fold by drainage diversion schemes. Thus the hydraulic characteristics of these two lakes separate them from the remainder of the study lakes, which receive runoff from natural watersheds. In order to prevent severe bias in the statistical analyses, these two lakes were excluded 70

PAGE 75

Table 22. Calculated Nitrogen and Phosphorus Supplies for Lakes Lake a Type b TST N c c .p ....... Lake .. Type a i b ['31; NC pc 1 Santa Fe 0 2.5 .28 .015 29 Watermelon M 3.6 1. 45 .062 2 Lit. Santa Fe 0 2.3 .32 .014 Pond 3 Hickory Pond 0 2.5 2.25 .051 30 Long Pond 0 2.S 5.94 .183 4 Altho 0 2.5 .53 .031 31 Burnt Pond E 8.3 2.27 .092 5 Cooter Pond M 5.3 3.72 .101 32 Wauberg E i 7 4 .63 .028 6 Elizabeth M 3.2 1.45 .064 33 Tuscawilla M i 4.8 2.60 .124 7 Clearwater 0 2.6 1.01 .051 34 Apopka H 22.1 2.23 .161 8 Hawthorne E 9.1 1. 62 .130 35 Dora H R8.5 3.00 .127 9 Lit. Orange M 3.4 2.58 .082 36 Harris M 16.3 1.10 .029 10 Unnamed M 3.2 .54 .021 37 Eustis H 10.5 1. 46 .077 11 Moss Lee 0 2.8 .39 .020 38 Griffin H 13.7 3.69 .183 12 Jeggord 0 2.8 .57 .027 39 Weir M 3.3 .29 .010 13 Still Pond U 1.9 1. 73 .072 40 Kingsley U 1.9 .18 .015 --.:] 14 Lochloosa M 5.2 1.15 .044 41 Sandhill U 1.3 .29 .015 I-' 15 Orange M 4.3 1.85 .071 42 Magnolia U 1.3 .25 .011 16 Palatka Pond M 3.2 2.70 .121 43 Brooklyn U 1.5 .26 .016 17 Newnan's E 7.1 2.61 .11S 44 Geneva U i 1. 8 .31 .022 18 Mize M 4.2 2.05 .183 45 Swan U 11. 5 .26 .015 19 Calf Pond M 4.6 2.42 .132 46 Wall 0 2.1 3.27 .124 20 Unnamed H 18.5 3 .335 4/: Santa Rosa U i 1. 3 .18 .009 21 Meta M 3.1 3.00 .250 4$ Adaho U 2.2 1. 03 .039 22 Alice H 10.7 106.00 18.000 49 McCloud U 1.5 1.35 .058 23 Bivin's Arm H 14.7 6.86 .424 50 U 1.5 3.10 .187 24 Clear E 8.8 4.31 .405 51 Suggs 0 12.3 2.24 .071 25 Unnamed M 6.4 2.07 .113 Long U i 1. 3 .55 .026 26 Beville's Pond M 3.1 2.89 .187 53 Winnott M '2.0 .41 .016 27 Unnamed M 5.8 .77 .032 Cowpen U 1.5 .42 .021 28 Kanapaha H 13.5 48.30 2.950 55 Gall.ilee. U i 1 ... 6 .. ..86 .036 aKey to Symbols: U -Ultraoligotrophic; 0 -Oligotrophic; M -Mesotrophic; E -Eutrophic; H Hypereutrophic. bTrophic State Index cIn g/m3_yr

PAGE 76

from the sample group, Lake (50) has a nitrogen and phosphorus loading comparable to hypereutrophic Lake Dora (35), but a TSI typical of ultraoligotrophic lakes (1.5, see Table 17). Two reasons may be responsible for this discrepancy: (1) the lake has not had sufficient time to equilibrate with its nutrient supply and (ii) the TSI has not been sensitive to the lake response. This lake has been artifictally enriched ____ ................ _wit.h __ ni tra.g.enand_pho.sphorus_ at_approximat.elY'_thepr.esent loading rates since 1967 as part of a study of eutrophication factors in Florida lakes (Brezonik and Putnam" 1968; Brezonik et al., 1969). Prior to 1967, the lake was ultraoligotrophic andsimilar in most aspects to the control, McCloud Lake (49). Both lakes are still ultraoligotrophic according to their TSI's although some increased growths of attached algae have recently been noted in Lake. Since the TSI accounts for phytoplankton production and biomass alone, this response is not reflected in the TSI. C, RELATIVE IMPORTANCE OF VARIOUS NUTRIENT SOURCES Budgets for six representative lakes are shown in Table 23 in order to compare the percentage contributions of the various nutrient sources to the overall nitrogen and phos phorus budgets. In order to illustrate general trends occurring in the transition from ultraoligotrophic to culturally hypereutrophic conditions, one lake from each of the five trophic groups is presented. In addition, Newnan's Lake is included as an example of a naturally eutrophic lake. For the ultraoligotrophic and oligotrophic lakes, the natural nutrient sources of rainfall and runoff from forested regions are domi nant, although Lake Santa Fe receives a small portion of its nitrogen and phosphorus supply (21%) from cultural sources. Orange Lake could perhaps be classified as naturally mesotrophic since most of its nitrogen and phosphorus supply is derived from natural sources. Lakes Hawthorne and Dora have obviously been influenced by the cultural activities in these watersheds. The former receives the major portion of its nitrogen and phosphorus supply from urban runoff and septic tanks while sewage effluent and agricultural runoff have played a significant role in the deterioration of Lake Dora. Newnan's Lake has a large, heavily forested watershed, and the associated runoff appears to be the predominant factor in the eutrophication of this shallow lake. Eutrophication of this sort is virtually impossible to control, whereas measures can be taken to control the cul tural sources degrading lakes like Hawthorne and Dora. 72

PAGE 77

Table 23. Percentage Contributions From Various Cultural and. Natural Sources for S"elected Lakes Lake Unproductive Rainfall % and Urban Fertilized Pasture Cleared Forest Septic on Lake CulTypea Nutrient Sewage Runoff Area Area Area Area Tanks Surface tural Santa Rosa (U) N 0 0 0 0 0 47 13 40 13 P 0 0 0 0 0 30 11 59 11 Santa Fe (0) N 0 7 5 7 1 41 2 37 22 P 0 15 1 3 N.S. 25 2 54 21 Orange (M) N 0 1 10 11 4 57 N.S. 17 21 P 0 5 2 6 3 49 N.S. 35 13 ---1 (E) N 0 8 14 4 56 w Newnan I s 2 1 15 25 P 0 22 N.S.b 7 3 41 1 26 30 Hawthorne (E) N 0 36 N.S. N.S. 5 14 32 13 68 P 0 57 N.S. N.S. 2 6 23 12 80 Dora (H) N 13 4 74 N.S. N.S. 1 2 6 93 P 60 12 14 N.S. N.S. 1 1 12 87 aSee Table 4 for key to symbols. bNot significant (less than 1%).

PAGE 78

D. STATISTICAL ANALYSIS OF TSI vs. NITROGEN AND PHOSPHORUS LOADING RATES Results of the statistical analyses are summarized in Table 24. Several regreSSion relatLonshLps were tested using both additive and multiplicative models. All the regression results presented in Table 24 were significant at the 99% confidence level. Using the magnitude of the multiple correlation coeffj.cient (R) as a cl'iterion_ for choQsing among the equations, an additive equation (A) in Table 24 (b), including Simple, interaction and quadratic terms, the largest percentage of variation in TSI (R=.830). However, equations Band C incorporating only the simple loadings give comparable Significance (R-.BO), and inclusion of the interaction terms thus provides only marginal increases in R. The multiplicative model (Equations D and E) is the least significant, and comparison of the additive and multiplicative equations suggests that the functional relationship between TSI and nitrogen and phosphorus loadings may itself be additive with one nutrient being more significant; i.e. limiting. In Florida lakes it appears that the phosphorus loading is the limiting factor since it is the first independent variable incorporated by the stepwise procedure into the regression equations, and it has the simple correlation [.786, Table 24 (a)] with the TSI. However, too much significance should not be placed on the above interpretations. RegreSSion analysis is inherently and its primary value lies in its predictive abilities rather than in any analytical potential. Canonical correlation analysis [Table 24 (c)] derived a canonical variate of the seven trophic indicators that was significantly correlated (.723) with the canonical variate of nitrogen and phosphorus loadings. In general, the analysis corroborates the regression results. For instance, phosphorus loading is the more significant of the two loadings based on the weighting factors in the canonical variate (1.19 for P vs. -0.23 for N). The most heavily weighted trophic indicator in the indicator canonical variate is total phosphorus tration (TP). Thus, the larger weightings associated with P and TP the dependence of average total phosphorus concentration upon the phosphorus loading. Vollenweider (1968) observed a similar correlation between spring total phosphorus concentrations and phosphorus supply for a group of European lakes. Although the regreSSion and canonical correlation analyses resulted in statistically significant relationships, there is considerable disagreement between the predicted and observed values of TSI. For example, Lake Griffin has an experimental TSI of 13.7 and a predicted TSI, using equation A of Table 24 (b), of 9.6, a 30 percent error. Similar discrepancies exist 74

PAGE 79

--...;J \J1 (a) Table 24. Statistical of Between TSI and Nand PLoading CORRELATION MATRIX: TSI N P TSI 1.000 N .773 1. 000 p. f786 .935 1.000 (b) STEPWISE REGRESSION ANALYSES: Multiple! Loading Rate Model Unitsa Equationb Correlation F RatioC Coefficieht Additive A V B V C S TSI = +28.7(NV)(PV)+2.37(Ny) TSI = 26.1(PV)+0.90(NV ) TSI = 0.62(NS)+10.1(PS ) 48.1 43.2 46.4 Multiplicative D V E S TSI = 1.08(PV)-lt2.(NV)Olt TSI = 0 84(p S S 15.6 14.1 (c) CANONICAL CORRELATION ANALYSIS: Canonical Variate of Trophic State Indicatorsd Canonical Variate of Nand P 0.69(TP) + 0.64(1/SD) -0.36(TN) + 0.34(pp) + 0.33(CD) + 0.17(I/GR) .... li19(P) .""' .. 23(N) .830 .793 .804 .620 .60.0 Percent Variance Explained by Equation 68.9 62.9 64.5 38.5 36.0 Canonical Correlation Coefficient .J23C aLoading rates per unit lake volume (V), per unit lake surface area (S). bAbbreviations: TSI=trophic state index NS and Ps and phosphorus surface loading rates in g/m2._yr.; NV and Pv =nitrogen and phosphorus volumetric loading rates in g/m3-yr. cAll significant at the 99% confidence level. dKey to symbols: TP=total phosphorus (mg/l); l/SD=inverse Secchi disc Cl=chlorophyll a (mg/m 3); TN=total organic nitrogen (mg/l); PP=primary production (mg C1m3-hr.); CD=specific conductance l/CR=inverse of Pearsall's (1922) cation ratio=8(Ca)+(Mg)]/[(Na)+(K)].

PAGE 80

for some of the other lakes with the average error being about percent. Thus in spite. of the strong trends demonstrated by the significant regression relationships, there is substan,... tial scatter of the experimental data about the fitted regression surfaces. Several possible sources of uncertainty will be discussed later. E. CRITICAL NUTRIENT LOADING RATES: Of great interest in control of cultural eutrophication is the development of critical loading rates, above which eutrophic conditions might be to ensue. Vollenweider (1968) has developed two types of critical loading rates based on information from a number of European and American lakes. Permissible loading rates are values beloW whichhoeutrophication problems should occur, and dan,"" gerous loading rates are values above which problems can be expected, Loading rates in between these two figures mayor may not cause problems depending on other factors, Inspection of various limnological data from the 55 Florida lakes indicates that eutrophic conditions (and attendant wat@r quality deterioration) are associated with all lakes having TSI values greater than 7.0 (i.e. the lakes in the eutrophic and hyper. eutrophic classes of Table 17), and similarly lakes with TSI values less than 4.0 have essentially no nutrient enrichment problems. Using these as "dangerous" and "permissible" TSI values, respectively, the nitrogen and phosphorus loading rates associated with these values were computed, assuming an N:P molar loading ratio of 15:1 as most appropriate. Criti,"" cal rates were computed on both areal and volumetric loading bases from appropriate regression equations, and Table 25 compares these results with those of Vollenweider (1968). It appears that Florida lakes can assimilate nutrients at some what greater rates without becoming eutrophic than suggested by Vollenweider's critical values, but the uncertainties involved in both analyses prevent detailed interpretation. Some interesting results were obtained through graphical presentation of the relationships the TSI and phos phorus and nitrogen supplies, respectively. In Figure 9, the TSI for ea6h trophic group is plotted against the cor responding mean loading. Figure 10 represents. a similar treatment considering mean nitrogen loadings. The horizontal bounded lines represent plus and minus one standard error of the group loading mean. In both graphs the dependence of TSI on nitrogen or phosphorus loading can be adequately described by an exponential function similar to. the classical logarithmic growth curve, The least squares equation and correlation coefficients are shown for each figure. That both curves are similarly shaped is to be expected since the 76

PAGE 81

Shannon and Brezonik, 1971c Ibid. Table 25. Critical Loading Rates for Nitrogen and Phosphorus Loading Permissible Loading -R-at@-gnits-N P Volumetric (g!m3 .... yr) .86 .12 Areal(g!m2 .... yr) 2.0 .28. Vollenweider Areal (g!m2. .... yr) 1.0 .07 (1968)a aFor lakes with mean depths of 5 ill or less. 77 Dangerous Loading N P 1.51 .22 3.4 .49 2.0 .13

PAGE 82

(f) ..... X W 0 Z w l-
PAGE 83

en I20 16 x 12 w ___ _____ ______ __c ___ __ z lIJ .....
PAGE 84

nitr.ogen and phosphorus loadings are themselves highly correlated [See Table 24 (a)J. The within-group deviation of loadings is considerably more pronounced for the phosphorus re lationships, particularly for the hypereutrophic and eutrophic groups of lakes. Such deviations are to be expected when representing the complex process of trophic state change in terms of a single nutrient input. In addition, the hypereutrophic group is essentially unbounded at the upper end and therefore not subjected to artificial boundary constraints .... -as.ar.e.-the.-o theI'-.four .. g.rou.p.s.-.... Qu:it e-l-i-ke I-JZ'-G-hange-s--l.I'l-t.R limiting nutrient will occur over any extended range of trophic state response, Thus, the relationships of Figures 1 and 2 reflect only an average situation, and their major utility probably lies in the area of lake management. For example, given that either nitrogen or phosphorus is limiting and a known or proposed nutrient loading, potential lake response can be determined by consulting the appropriate relationship. It should be emphasized that the graphical relationships in .. Figures 9 and 10 are. most applicable for shal-lowsubtrop-i.cal lakes, and their use in other situations may be unwarranted. F. EFFECT OF DEPTH ON LAKE CAPACITY TO ASSIMILATE NUTRIENTS As our data base expands it should be possible to incorporate other factors of Eq. 15 into empirical eutrophication models. For example, mean depth is probably the most important morphometric factor affecting eutrophication. Figure 11 indicates a slight mean depth-trophic state relationship exists for the 55 Florida lakes, with the most eutrophic lakes having mean depthsof 4 m or less and the deepest lakes being the most oligotrophic. As expected a large scatter occurs. The proper relation of mean depth to eutrophication has been confused by:many. It is neither a trophic indicator nor a causal factor per se. Rather mean depth affects the rate at which a lake can assimilate nutrients and maintain desirable trophic conditions. The graphical approach taken by Vollenweider (1968) might prove useful in quantifying the effects of depth. The method plots nutrient (N or P) loading rates vs. lake mean depth, and the lines delineating the regions in which oligotrophic and eutrophic lakes occur are estimated by inspection. Figures 12 and 13 illustrate this approach for Florida lakes using phosphorus and nitrogen loading rates, respectively. However it is obvious that insufficient shallow oligotrophic and deep eutrophic lakes occur in the sample group to permit accurate delineation of boundary lines. Per haps a better approach to evaluating the role of mean depth in the trophic state calculus would be the method of response surfaces (Box, 1954; Goldman, 1967). 80

PAGE 85

r-. H U) e 15 @ "{:! IJ) u cd 10 () -rl .r:: p.. o H 5 o o o o o o 0 o (> () 0 o o o o 0 o o o o o Mean Depth, M. Figure 11. Trophic State Index (TSI) Values vs. Mean Depth for the Florida Lakes 81

PAGE 86

1.0 HE HE .5 HE E HE HE o M Eutrophic Lakes (Florida) ./ ./ HE / ------------------HE./ t!> .2 .1 / ./ o ,,/ E o E,,/ HE M __ / ./ o M/ /' / /0 U E ./ M,,/ /' 0 0 /0 M u Eutrophic Lakes /' (Vollenweider) ,/ o M o o M U E o /' M U,/ u /6 u /' 0 U o B / / / ,/ Oligotrophic / ,/ Lakes Florida) Lakes (Vollemveider.) / / / o / / ,,/ .02 ___ ______ 1 2 5 10 MEAN DEPTH, M. Figure 12. Annual Phosphorus Loading Rate Vel. Mean Depth for 55 Florida Lakes. Each datum represents a lake in the trophic group denoted by that symbol. U = u1traoligotrophic, 0 = oligotrophic, M = mesotrophic, E = eutrophic, HE = l:).ypereutrophic. 82

PAGE 87

10.0 HE HE HE HE 0 E UO 5.0. 0 HK M 0 E 0 HE 0 o 0 E M HE EUTROPHIC LAKES (FLORIDA) --------_M tto 'd cO / 2.0 0 M UO E M M HE 0 0 M ..M -----UO UO ,r ..LAKES.(FLORIDA) o 0 0 UO o --------------H OUO UO UQ..-...... LAKES ____ 1-.0 -----0.5 OLIGOTROPHIC LAKES (VOLLENWEIDER) Key UO Ultraoligotrophic 0-Oligotrophic M -Mesotrophic E -Eutrophic HE -Hypereutrophic ________ __________ ____________ __________ 0.5 1.0 2.0 5.0 Mean Depth (m.) Figure 13. Annual Nitrogen Loading ys JVieanPe.p.tl1. tor 55 J;l'lorida Lakes Each datum represents a lake having the trophic state denoted by that symbol. 10.0

PAGE 88

G. SOURCES OF UNCERTAINTY The above analysis represents an attempt to approximate the general trophic response function (Eq. 15) by the simple relationship between TSI and nitrogen and phosphorus loadings in Eq. 16. The uncertainty term in Eq. 16 represents the discrepancy between values of TSI predicted by the function g(N,P) and the actual (measured) TSI values, assuming the in fact represents tbe-true trophic status of' -the lake. -Individual components of the term may include the following: (i) g is an approximation of f' (ii) the gen and phosphorus supply calculations are in error, and (iii) the TSI does not represent the concept of trophic state (TS) completely. Approximations of f were obtained here by using multiple regression techniques. These approximations included only two of a number of potentially important variables3 i.e. nitrogen and phosphorus loadings. The loadings were estimated using land use and population characteristics and literature values of individual source contributions, a procedure that contains some inherent uncertainty. The TSI may not completely describe the concept of trophic state in spite of the fact that it incorporates seven of the more significant trophic state indicators. As previouBly discussed in reference tn Anderson-Cue Lake, it does not account for macrophyte and periphyton biomass or primary production, which in some lakes may constitute a significant proportion of total lake primary production, H. RELATIONSHIPS BETWEEN TROPHIC STATE AND GENERAL WATERSHED CONDITIONS Another approach to relating trophic state to watershed factors is direct regression of lake conditions (expressed by a TSI) to the extent of various land use practices and population characteristics with the watershed (expressed on a per unit lake volume or area basis), The trophic indicator data (Table 7), the TSI values (Table 17), and the population and land use data (Table 20) were used for these analyses. In addition, the correlative relationship between the seven trophic state indicators and the eutrophication factors was investigated using canonical correlation analysis. The phication factors were expressed on a per unit lake volume basis in the ensuing analyses by dividing the values in Table 2Q by the total lake volume, Thus, the units for land use patterns were square meters for a particular land use per cubic meter of lake water, and population characteristics were expressed as number of cultural units per cubic meter of lake water. The eutrophication factors could alternatively have been expressed per unit lake surface area. However, it 84

PAGE 89

seems more logical to express eutrophication factors for shallow Florida lakes on a unit volume basis since the entire volume is involved in assimilation and dilution of nutrient influxes. Results considering land use and population factors on a unit surface area basis were similar to the results obtained on a volumetric basis (Shannon, 1970). For the reasons discussed in the section on statistical analysis of the TSI vs. nutrient loading rates, Lakes Alice and Kanapaha were ex cluded from the following analyses. Results of regression analyses for TSI (as independent variable) vs. various eutrophication factors are shown in Table 26. Two regression equations are given; the first con siders TSI as a linear function of the land use patterns within the watershed plus the immediate and remote cultural units. The second considers TSI as a function of the land use patterns plus total cultural units (TCU), i.e. the sum of remote, immediate and sewage treatment cultural units. The independent v-ariables of the regression equations are written in the step""" wise order in which they were incorporated into the equation, i.e. in decreasing order of their partial correlation with TSI. Both equations in Table 26 are statistically Significant at the 99% confidence level and both explain about 80% of the total variance in the TSI. The first independent variable in both equations is the fertilized cropland; other culturally influenced factors such as urban area and immediate cultural units are important var iables in explaining the variance in TSI. A natural factor, forested areas, is also important, but other factors like un...., productive cleared area, remote septic tanks and pastured areas add little to the predictive abilities of the equations. These results can be interpreted as suggesting that culturally influenced factors (fertilized cropland, runoff, septic tank drainage) are among the most .important variables deter mining the trophic states of Florida lakes. However, it should also be emphasized that regression analyses are inherently empirical, and while they may suggest, they never prove cause effect relationships. A canonical analysis of the seven trophic indicators (Table 7) and six eutrophication factors (the land use areas and total cultural units) (Table 20) for the 55 lakes is pre sented in Table 27. The correlation coefficient between the two canonical variates II and JEF is high (0.94) and signifi cant at the 99% confidence leveI. In the trophic indicator canonical variate (II)' primary production is weighted consider,.... ably higher than the other indicators, suggesting it is of fundamental importance in the trophic state_eutrophication factor relationship. At the other extreme the cation ratio has a low weight and appears to be of minor importance in the relationship. Cultural factors (urban area and fertilized cropland) carry the heaviest weightings in the eutrophication 85

PAGE 90

Table 26. Stepwise Regression Analysis of TSI vs, Eutrophication Factors Expressed Per Unit Lake Volume1 (1) Regression Equation: TSI = 14.95(HFA) + .64(FOR) + 2.72(ICU) + 1.59(URB) 59.6 73.9 80.0 81.2 .35(UCA) + .06(RCU) .02(PA) 81.5 81.5 81.5 F Ratio = 28.9B*** Multiple Correlation Coefficient Cr) = .903 Percent of total variation explained ldythe regression equation =81.5% (2) Regression Equation: TSI = + .61(FOR) + + .53(TCU) 59.6 73.9 79.4 80.0 + .31(UCA) .01(PA) 80.3 80.3 F Ratio =31.91*** Multiple correlation coefficient (r) = .896 Percent of total variation explained by the regression equation = 80.3% Key to Eu,trophicatiQn ';F'actor Sy'mbols: BFA = Heavily fertilized cropland CmZ/m3) FOR = Forested area (mZ/m3) ICU = Immediate cultural unitsC#/m3x lOIt) URB = Urban area CmZ/m3) UCA = Unproductive waste cleared area Cmz/m') RCU = Remote cultural units C#/m'xlOIt) PA = Pastured area CmZ /m 3 ) TCU = Total cultural units C#/m3x lOIt) ***Denotes significant F valu,e at the 99% confidence level. lValues below symbols in regression equati.on indicate curnu,latiye percent of total variance explained by independent variables up to that point. 86

PAGE 91

Table 27. Canonical Analysis of the Relationship Between Seven Trophic Indicators and Six Eutrophication Factorsl Canonical Variate 1: Linear function of trophic indicators fr + 0.71(COND) 0.17(TON) + 0.25(TP) + 1.13(PP) 0.60(CHA) -0.09(1/CR) Canonical Linear function of eutrophication factors hF = -0 10 (F OR) + o. 53 (URB) + O. 79 (HF A) O. 04 ( P A) -0.06(UCA) -0.16(TCU) Canonical correlation coefficient = 0.94*** ***Significantat the 99% confidence level by a testing procedure described in Morrison, 1967. lSee Table 26 for key to eutrophication factor abbreviations. 87

PAGE 92

factors canonical variate ((E ); pasture and unproductive cleared areas carry the weightings, which corroborates the regression results of Table 26. Thus in general it appears that the major link between trophic state ((I) and eutrophi cation factors (fEE) is one of primary prodfiction and the cultural factors of urban and heavily fertilized areas. Comparing the canonical correlation of trophic indicators vs, nitrogen and phosphorus loadings (Table 24) with the _C3.J::>9ve_ anC3.JYsj. higher __ _was_ obtained in the latter analysis. There is some inherent error in using literature values of the expected nitrogen and phos phorus contributions from land use patterns in order to obtain nutrient loadings, and this quite likely explains the lower correlation for the analysis using the nutrient loading rates. In other words, the land use and population character ...... istiDS in their raw form contain more significant information than the calculated nitrogen and phosphorus loadings. relationships such as those in Tables 26 and 27 depend on the fact that runoff from various land use practiDes has different and to an extent defined nutrient enrichment effects on receiving bodies of water. Similarly the population within a watershed can be dIvided into a few main -g-rQups(e. ,:}.-people on sewerag.esy-stems,peeple usingseptc tanks in the immediate vicinity of a lake, etc.) which have similar (within group) enrichment effects. Refinement of this type ot regression relationship could prove beneficial to regional planners and land use (zoning) boards. In evaluating the statistical relationships between trophic state and eutrophication factors, the time element has not been considered. ;It has been assumed that lake trophic state as reflected by. the TSI or certain tr.ophicstate indica ..... tors waS a r,esult of the eutrophication factOrs at that time or, in other trophic statea.nd the eutrophication fac,... tors were in equilibrium at the time they were evaluated. In reality, the trophic state of a lake is the result of the eutrophication factors influence over a period of years. For example, the hypereutrophic conditions of Lakes Apopka and Dora in the Oklawaha group are due to the intense cultural activities around the lake in the past two or three decades. On the other hand, Anderson-Cue Lake has been subjected to a high rate of nutrient enrichment over a period of three years but remains in an oligotrophic condition, presumably because it has not had sufficient time to demonstrate a sponse. Very little is known about the response time of a lake to nutrient enrichment, and as yet it is impossible to quantify. However, it seems reasonable to assume that the majority of the lakes are in a state of dynamic equilibrium with their environments; the relatively high correlations between causal factors and effects would seem to substantiate this point. 88

PAGE 93

I. RELATIONSHIP BETWEEN TSI AND TOTAL WATERSHED AREA A simpler relationship between watershed and lake trophic state was recently proposed (Schindler, 1971) for lakes within a similar geological region and in which cultural influences are slight. In a nutrient poor terrain the atmosphere acts as the major nutrient source. Assuming a steady-state exists -----------t-1-gn -J-anQ--nu-t-F-1-e-nt-e-*E>F-t -----to the lake, the rate of lake nutrierit enrichment should be directly proportional to the sum of lake area (Ao) plus water-shed land area (Ad)' Because nutrient influx will be diluted in proportion to lake volume (V), Schindler (1971) hypothesized that lake trophic conditions then should be proportional to (Ad +AQ)/V. Many lakes in north-central Florida fit the above conditlons, and Figure 14 shows the crude correlation result-ing when TSI is plotted vs. the wa-tershed factor for these lake-s-. Data-poin-ts in F-igu,& 1-4--I'&pr&s@nt s-&epag@-ar:J..Q-s-l-mi .... drainage lakes located in similar terrain in the Trail Ridge and Alachua County regions of Figure 2. Drainage lakes and those showing major cultural influences were excluded. The hypothesis seems to have limited applicability under these conditions but the scatter implies poor predictive abilities. Thu-s-thee-arlierstatement thateutrophicat"1:on is a c;omplicated phenomenon is again borne out, and simple relationships are unlikely to explain more than general trends. From the point of view of eutrophication and lake management, a compromise between highly complex mathematical models and oversimplified empirical relationships, such as described in the preceding analyses, would seem the most appropriate means of effecting satisfactory results. CHAPTER 6. CONCLUSIONS The limnology of north and central Florida is dominated by shallow solution type lakes in a sandy terrain. While thermal stratification is not typical in these lakes, neither is it rare, and stable stratification can occur in small ponds as shallow as 3.5 meters deep. The waters of most lakes are low in dissolved solids, soft and slightly to moderately acid. Organic color is an important but geographically variable feature of the lakes. Both acid and alkaline conditions occur in colored waters, but the former are more prevalent. Appar ently few lakes are springfed, accounting for the paucity of hard-water lakes. Lake trophic state was envisioned as a multi-dimensional or hybrid concept described by several biological, chemical and physical indicators. Groups of lakes w1th similar trophic state characteristics were formed using cluster analysis, and

PAGE 94

7.0 ,..... H 6.0 @ ] 5.0 H OJ +J Cll 4.0 __ ____ til t.I ..-l ..c:: 3.0 p. o E-t 2.0 1.0 0 o 0 0 0 8 o o 0 0 0 0 0 0 0 o o _____ __ __ I o 1 2 3 4 5 6 7 8 -1 m Jrigure14. Trophic State Index. v's. Total Watershed Area/Lake Volume Florida Lakes Ao = lake area, Ad = watershed land area Only seepage or semidrainage lakes with minimum cultural influences are plot.ted. 90 9

PAGE 95

these groups could be interpreted in the classical (oligosense. A trophic state index (TSI) was formulated using principal component analysis incorporating seven trophic state indicators. The TSI quantified the concept of trophic state on a numerical scale, thus providing a method for ranking and comparing lake trophic states. The t_rophic_states_ of -Florida lakes are largelydependent on gross nitrogen and phosphorus supplies (loading rates) as evidenced by significant regression relationships between TSI and Nand P loading rates and significant canonical correlation between seven trophic indicators and the Nand P loadings. Phosphorus loading was the most important able from a statistical viewpoint in the regression and canon ical relationships, and it might be inferred that on an average basis phosphorus is the (most common) limiting nutrient for Florida lakes. Cultural nutrient sources are relatively unimportant in oligotrophic lakes, but for many eutrophic lakes, cultural sources are by far the most significant. Critical nutrient loading rates were calculated for Florida lakes based on the regression relationships. Florida lakes seem capable of assimilating greater quantities of nutrients than suggested by Vollenweider's critical loading but the two studies are in general agreement. A positive correlation exists for Florida lakes, between lake trophic state and lake watershed land use and population characteristics. The relationship was verified by statisti cally significant multiple regression equations using the TSI as the dependent variable and several watershed land use and population characteristics as independent variables. Canonical correlation analysis of several trophic state indicators versus the population and land use characteristics showed high correlation and corroborated the regression results. It appears that cultural influences have played a major role in determining the trophic states of Florida lakes. Regression and canonical analyses results indicate that the most influential eutrophication factor from a statistical viewpoint is fertilized cropland. In spite of the statistically significant results ob in this study there are several sources of uncertainty in the methodology. These sources have been discussed in the text and should not be overlooked in studies of a similar nature. 91

PAGE 96

APPENDIX MULTIVARIATE TERMINOLOGY The term "multivariate analysis" is used to describe statics t-i-&al-t-e-e n&e-p-B-B-a-w-i-t-:8-ana-l y-z4-B-g-aa-ta--e e 0 tea for p different variables on N objects.. For example, the variables in this study are chemical, biological, and cal characteristics measured on several lakes representing the objects, Some dependency is assumed among the variables so that they are considered as a system, implying that no able can be separated from the group and considered ally. This feature distinguishes multivariate data and niques from their multi-dimensional nature multivariate tech niqu@s are-.-most--egnvenie-ntly--dese-pilded matl?ix notation. Vector quantities in the text are underscored, for ex ample represents the vector of p variables for lake i. Matrix quantities are denoted by capital letters, and scalar quantities are denoted by small le-t-ters. Vectors are c-olumn vectors unless the transpose is indicated by priming the vector (e.g. x.' is the transpose of xi)' The inverse of a rna trix A lS deAoted by A -1 The (ij T -th element of a matrix A is denoted by aij' Suppose that the assumptions of random and independent sam,pling have been satisfied and observation vectors of p variables are evaluated for N lakes. The resulting tion of data may be expressed in an Nxp (N rows and p columns) raw data matrix: x x X 1 1 1 2 1P I X X22 x (A-I) 21 2p X = XN1 XN2 xNp The X matrix is the starting point .for most multivariate procedures. Analogous to the univariate situation where a random variable x is considered to be normally distributed with mean 11 and variance .cr2,. multi va.riate .ctata.areconsidered to be realizations of a p-dimerisional random variable distributed multivariate normal with mean vector 11 and covariance matrix L:. AS)1 andcr2 are estimated by the Sample .mean x and the sample variance 82 in the univariate case, )1 and E are estimated 92

PAGE 97

by the vector of sample means of the p variables X and the sample covariance matrix S: and CA-3) S is the p x p matrix of covariance between all possible pairs of variables, e. si,i = cova::ian?e hetween variables x and x. S 1S a matr1x, 1.e. 8i' =s .. except i=j.J The variance of variable xi is the element sii' The matrix of sample correlations between all possible pairs of variables is denoted by the matrix R where: Slj Is .. s 11 JJ The matrix R can be computed from S by the expression: R = C 1 ) 'c 1 ) D-.S.D-, 'si si CA-4) (A-5) where DCJ:-.) denotes a matrix containing the reciprocals of the si standard deviations in the diagonal elements and zeros in all other elements of the matrix. The matrix R is also p x p and symmetric. When the variables under consideration are in different units and ranges it is necessary to transform (or standardize) them to a scale of common origin and units. The Z score method is a commonly used standardization technique. The raw data matrix X is transformed to the standardized matrix Z by CA-6) where Z and X are the Nxp matrices of transformed and raw variables, respectively. I is theNxN identity with l's on the diagonal and zeros elsewhere, E is an NxN matrix with lIs in every position and D is a pxp diagonal with reciprocals of the standard deviations on the diagonal elements and zeros elsewhere. The general element of Z is given by 93

PAGE 98

z lJ = X'ij,"",Xj Sj 94

PAGE 99

ACKNOWLEDGEMENTS This research was supported in part by Office of Water Resources Research Matching Grant DI 14-31-0001-3068 and a grant from the State of Florida Game and Fresh Water Fish Commission. A Federal Water Quality Administration Grant DON 16010 (H. D. Putnam, principal investigatgr) ,_ supported a sub-stantial porTion -of the proj ect, especially in its early phases. A number of faculty colleagues have contributed advice and encouragement, including Drs. Hugh D. Putnam, James P. Heaney, William H. Morgan, and Jackson L. Fox. Dr. Fox was especially helpful in providing needed assistance in sampling and in the various biological aspects of the project. The assistance of Dr. Morgan in administrative affairs and in the compretion of this report is truly appreciated. Special thanks also go to Roger Yorton and Michael Keirn, project assistants and graduate students in the Department of Environmental Engineering, for their cooperation in arranging and conducting the sampling trips and running the chemical and biological analyses. 95

PAGE 100

BIBLIOGRAPHY Aberg, B. and Ro hde, W., HUber die Milieufaktoren in einigen su.dschwedischen Seen, "Symb. Botan'. Up'psala, Vol. 5, 1942, pp. 1-256. Anders on, T. W., An Tntr'od\ictiontoMuTt'ivariate Statistical Analysis, John Wiley and Sons, Inc., New York (1958). -Beet troph-i c a '[1 on 0 r-'fhe --------Great Lakes. Limnol. Oceanogr., 10:240-254. Birge, E. A. and Juday, C., "The Organic Content of the Waters of Small Lakes," Proc. Amer. Phil. Soc., Vol. 66, 1927, pp. 357-372. Birge, E. A. and Juday, C. 1934. Particulate and Dissolved Organic Matter in Inland Lakes. Ecol. Monographs, 1:440-474. Box, G. E. P. 1954. The exploration and exploitation of response surfaces: some general considerations and examples. Biometrics 10, 16-60. Bowen, D.H. M. EmTiron 1., 725--.726 (197Q1. Bradley, W. H. and M. E. Beard. 1969. Mud Lake, Florida; its algae and alkaline brown water. Limnol. Oceanogr., 14: 1277-1279. -Brezonik, P. L., "Eutrophication: The Process and Its Modeling Potential," Proc. Workshop Modeling the Eutrophication Process, Univ. Florida, Gainesville, 1969, pp. 68-110. Brezonik, P. L.and C. L.HarpSr. 1969. Nitrogen fixation in some anoxic lacustrine environments. Science, 164,:1277-1279. Brezonik, P. L., Morgan, W. H., Shannon, E. E., and Putnam, H. D. 1969. Eutrophication factors in north central Florida lakes. Univ. Florida Industr. Engrg. Exper. Station, Bull. Ser. No. 134, Gainesville, 101 p. Brezonik, P. L. and Putnam, H. D., "Eutrophication: Small Florida Lakes as Models to Study the Process." Proceedings, 17th South. Water Resources and Poll. Contr. Conf., Univ. North Carolina, 1968, pp. 315-333. Brezonik, P. L. 1971. Nitrogen: sources and transformations in natural waters. Presented at 161st Nat. Meeting, Amer. Chern. Soc., Los Angeles, Calif., April, 1971. Brink, N. in "Nordisk Killokium om Eutrofieringsproblemer," O. Skulberg, ed., Norsk Inst. Vannforskning, Blindern, Norway, 1964. 96

PAGE 101

Chen, C. 1970. Concepts and utilities of an ecologic model. J. Sanit. Engrg. Div., Amer. Soc. Civil Engr. 96, 1085-1097. Clark, W. E., Musgrove, R. H., Menke, C. D. and Cagle, J. W., Jr. 1962. Interim report on the water resources of Alachua, Bradford, Clay and Union Counties, Florida. Florida Geol. Survey, Information Circular 92 p c:;ooke ,_C .. _W. 19 3g Scenery of Florida. F.lor ida_G.eo 1. Sur. vey, Geol. Bull. No U, 118 p. DiToro, D. M., O'Connor, D. J. and Thomann, R. V. 1970. A dynamic model of phyto-plankton populations in natural waters. Environ. Engrg. and Sci. Program, Manhattan College, Bronx, New York (mimeo). Dixon, W. J (Ed.), Biomedical Computer Programs Uni v. California Publ. in Automatic Computation No. 2, Calif, Press, Berkeley, 1968. Edmondson, W. T.,in "Water Quality Control," Univ. Washington Press, Wash., 1968. Engineering Ine., Gainesville, Florida, personal communication, 1970. Fisher, R. A., "The Use of Multivariate Measurements in Taxonomic problems,!! Annals of Eugenics, 7:179 (1936). Board of Conservation. 1969. Florida lakes. Part III. Gazetteer. Div. Water Resources, Tallahassee. 145 p. Fruh, E. G., Stewart, K. 00., Lee, G. F., and Rohlich, G. A. 1966. Measurement of eutrophication and trends. J.WaterPolT.Contr. Fed. 38, Goldman, C. R. 1967. Integration of field and laboratory experiments in productivity studies. In Estuaries, G. Lauff (ed.), Amer. Assoc. Adv. Sci., Washington, D.C. pp. 346,..,352. Goldman, C. R. and Armstrong, R. 1968. studies in Lake Tahoe, California. Limnol, 17, 49. Primary productivity Verh. Tnternat. Goldman, C, Gerletti, 00., Javornicky,P., Santolini, U., and DeAmezaga, E. 1968. Primary produc,.., tivity, bacteria, phytoand zooplankton in Lake Maggiore; Correlations and relationships with ecological factors. Mem. 1st. TtaT.ldrohio1. 23, 49':""127. 97

PAGE 102

Gower, J. C., "A Comparison of Some Methods of Cluster Analysis," Biometrics, Vol. 22, 1966, p. 623. Hansen, K. 1962. The dystrophic lake type. Hydrobiologia, 19: Hasler, A. D., "Eutrophication of Lakes by Domestic Drainage," Eco16gy, Vol. 28, 1947, pp. 383-395. F. F. "Eutro:Qhication InciJces and Their Relation tQ _____ Other Indices of Ecosystem Change," In: Eutrophica'tion: Causes,Consequences,Corre'ctiVes, National Acad. Sci., Wash., D.C., 1969, p. 225-235. Hutchinson, G.E. 1957. Wiley, New York. A treatise on limnology. Vol. 1. 1015 p. Hutchinson, G.E. 1969. Eutrophication past and present. In Eutrophication : Causes ,Conseq\ien'ces', CorrectiVe s, 'Rat. Acad. Sci" U. S., Washington, D.C. pp. Iovino, A. J. and W. H. 1969. The role of larval Chironomidae in the production of lacustrine copropel in Mud Lake, Marion County, Florida. Limnol. Oceanogr., 14 :.898=-905. Keirn, M. A. and Brezonik, P. L. 1971. Nitrogen fixation by in Lake Mize, Florida and in some lacustrine sediments. Limnol. Oceanogr.16 (in press). Kenner, W. E., "Maps Depths of Selected Lakes in J;i'lorida, II lnformation Circ. No. 40, .Pla. Geol. Surv., Tallahassee, Kerr, P. C., PariS, D. F., BDockway" D. L., "The Interrela .... tion of Carbon and Phosphorus in Regulating Heterotrophic and Autotrophic Populations in Aquatic Ecosystems, I' Water Poll. Contr. Res. Serf 16050, Fed. Water Qual. Admin., 1970. Lee, G. F.,etal., "Report on the Nutrient Sources of Lake Mendota-;rr Nutrient Sources Subcommittee of the Lake Mendota Problems Committee, Madison, Wis., 1966 (mimeo). Lee ", P. J. 1971.JVIultiVariate 'an'alys'isf'c:ir' 't'h:e'fis'h'e'r,i'eq biology. Fish. Res. Bd. Canada, Tech. Rept. 244, )1reshwater Institute, Winnipeg, Manitoba, 182p. Legge, R. F., Dingeldein, D., Canad. Res. Develop" 1970" p. 19,...,42. Ludwig, H. F., Kazmierczak, E., and Carter,R. C., "'Waste Disposal and the Future of Lake Tahoe ,'I J. ASCE,90 (3) : 27(1964),

PAGE 103

Margalef, R. 1958. Trophic typology versus biotic typology, as exemplified in the regional limnology of northern Spain. Verh. Internat. Verein. Limnol.13:339-349. Maslin, K. E. Ph.D. thesis, University of Florida, Gainesville, 1970. McGauhey, P. H., et al., "Comprehensive Study on Protection of Water Resources of the Lake Tahoe Basin through Controlled Waste ____ Tahoe, California, 1963. Miller" C. E., "Soil Fertility," Wiley, New York, 1955. Moreau, D. H. 1969. Concepts of mathematical models. In Modeling the eutrophication Proc. workshop, St. Petersburg, Florida, Univ. Florida, Dept. Envir. Eng., Gainesville, pp. 1-21. ------MorrIson, D. F. 1961: Multivariate statistical methods. McGraw-Hill, New York. Naumann, E., "Nagra Synpunkter Angaende Limnoplanktons Okologic Med Sarskild Hansyn Till Fytoplankton," Svensk. Bot.an. Tid-skr'-.,. .. VG.l.l-3 ,1919TP.Padron, M. 1969. An axiomatic basis and computational method for optional clustering. Dept. Industr. Systems Engineering Tech. Rept. No. 18, Univ. Florida, Gainesville. Paloumpis, A. A., Starret, W. C., Amer. Midl. Natur. 64, 406-421 (1960). -Patten, B. C. duction. 1968. Mathematical models of plankton proInt. Revue ges. Hydrobiol. 53, 357-408. Patten, B. C. 1969. Models of aquatic systems. In Modeling the eutrophication process Proc. workshop, St. Petersburg, Florida, Dept. Envir. Eng., Univ. Florida, Gainesville, pp. 32-53. Pearsall, W. H. 1922. A suggestion as to factors influencing the distribution of free-floating vegetation. J. Ecology 9, 241. Polta, R. P., in "Water Pollution by Nutrients-Sources, Effects and Control," Water Resources Res. Center, Bull. 13, Univ. Minnesota, Minneapolis, 1969. Putnam, H. D. (Ed.), "Modeling the Eutrophication Process," Proceedings, Workshop at St. Petersburg, Florida, Nov., 1969, Univ. Florida and Fed. Wat. Qual. Admin., 1970, 290 p. 99

PAGE 104

Putnam, H. D., Olson, T. A. and Odlaug, T. O. 1966. Primary production at a fixed station in Lake Superior. Proc. 9th Conf. Great Lakes Res., University Michigan, Great Lakes Res. Div. Rohlich, G. A., Lea, W. L., "The Origin of Plant Nutrients in Lake Mendota," Rept. to Univ. Wisconsin Lake Investigations Committee, Madison, Wis., 1949 (mimeo). _________ $aunders, G. W. __ of a Modified 14C Technique for Shipb6ard Estimation of Photosynthesis in Large Lakes, Great Lakes Research Division, Publication No. 8, Univ. of Michigan, Ann Arbor, Mich. (1962). Schindler, D. W. 1971. A hypothesis to explain differences and similarities among lakes in the Experimental Lakes Area, northwestern Ontario. J. Fish. Res.Bd.Canada 2 8, __ ?95 ... 3 01._ Sheffield, C. W. and Kuhrt, W. H. 1970. Lake Apopka -its decline and proposed restoration, p. 130-146. In Proc. conf. water poll. contr. Univ. Florida Exper. Sta. BUll. Ser. No. 135, Gainesville, pp. 130-146. Shannon, E. E. 1969. Multivariate techniques for the classification of lakes and the study of eutrophication. In Modeling the eutrophication process, Proc. workshop, St. Petersburg, Florida, Dept. Envir. Eng., Univ. Fla., Gainesville, pp. 175-212. Shannon, E. E. 1970. Eutrophication-trophic state relationships in north and central Florida lakes. Ph.D. thesis, Univ. Florida, Gainesville, 257 p. Sokal, R. R. and Sneath, P. H. 1963. Principles of numerical taxonomy, W. H. Freeman, San Francisco, Calif. Stewart, K. M. and Rohlich, G. A. 1967. Eutrophication -a review. Calif. State Water Qual. Control Bd. Publ. No. 34, 188 p. Stubbs, S. A. 1940. Solution a dominant factor in the geomorphology of peninsular Florida. Quart. J. Florida Acad. Sci., 5:148-167. Sylvester, R. 0., in "Algae and Metropolitan Wastes," U. S. Publ. Health Servo Rept. W61-3, p. 80-87, 1961. Thomann, R. V. 1971. Systems analysis and water quality management. Envir. Science Services Div., Envir. Res. Application, Inc., Wilton, Connecticut. 100

PAGE 105

U. S. Bureau of the Census, "United States Census of Population, 1960," U. S. Govt. Printing Office, Washington, D. C., 1961. Vollenweider, R. A., Arch. Hydrobiol. 66, 1-36 (1969). Vollenweider, R. V. 1968. The scientific basis of lake and stream eutrophication, with particular reference to phosphorus and nitrogen as eutrophication factors. Technical to 0 C._D., Paris, DAS/CSI/68-,---27 :_1-182 (mimeo LWallis, J. R. 1967. When is it safe to extend a prediction equation? -An answer based upon factor and discriminant function analysis. Water Resources Research. 3, 375-Watt, K. E. F. 1968. Ecology and resource management. McGrawHill, New York. Weibel, S H., ig"Eutroph:Lca t ion : Causes Consequences, Nat. Acad. Sci., Washington, D. C., 1969. Yount, J. L. 1961. A note on stability in central Florida lakes, with discussion of the effect of hurricanes. Limnol. Oceanogr., i:322-325. Yount, J. L. 1963. South Atlantic states, p. 269-286. In D. G. Frey (ed.), Limnology in North America. Univ. Wisconsin Press, Madison. Zafar, A. R. 1959. Taxonomy of lakes. Hydrobiologia,13: 287-299. ADDENDUM Publications arising from this project thus far are: Brezonik, P. L. 1971. Nitrogen: Sources and Transformations in Natural Waters. Presented at 161st Nat. Meeting American Chemical Society, Los Angeles, Calif., April, 1971. Proceedings of symposium to be published by J. Wiley. Brezonik, P. L. 1971. Morphometry and physical characteristics of Florida lakes. Florida Water Resources Research Center Publ. (in preparation). Brezonik, P. L. and Shannon, E. E. 1971. Eutrophication in Florida Lakes: Criteria for management. Presented at XVIIIth Congress of International Assoc. Limnol., Leningrad, U.S,S.R., August, 1971. Verh. internat. Verein Limnol. 18 (in press). 101

PAGE 106

Shannon, E. E. 1970. Eutrophication -trophic state relationships in north and central Florida lakes. Ph.D. thesis, Univ. Florida, Gainesville, 258 p. Shannon, E. E. and Brezonik, P. L. 1971a. Limnological characteristics of north and central Florida lakes. Limnol. Oceanogr. 16 (in press). Shannon, E. E. and Brezonik, P. L. 1971b. Eutrophication analysis: arnultiva.riate approach. J. Sanit. Eng. Div'J AIDer: Soc. Civil Eng.97 (inpre s s) Shannon, E. E. and Brezonik, P. L. 1971c. Relationships between trophic state and nitrogen and phosphorus loading rates. Submitted to Envir. Sci. Technol. 102