_. p. 0:36th

 

 

- .08 n:

2135935“ . _

  

:5? R35. . .

 

 

.u.~to.\v¢ I .u.

..u!~wl..o‘v.m

51... .

2!”.

.u. u.mmma...m.4iz

5|!
.2 .

 

12 Ill.

liai-

. I..I r

#5

2....

1.1‘

....

«awayaéﬁgd a.
_ ﬂ urging. #2.

22%. .
a»: . .
. 2.. .ﬁ.)

2mm;

.
it

Iv

2.

".2 ..
222.222..»

. .2, 22,

. 2...

J‘

.2. $1351.

a;
. It”.

. ....... .. .. .

2 .. .

, .. 1'3... .

a in.

ain’t...

né
x a,
.32.. ,

2.2.; at...
, r .

all‘l. 5‘

ﬁrm . ,. émﬁm

E -
bum. .. 21.55..
a .

.1 £152.24:

.93.:

r .2 f.
:2... .1

l .32.. .32 .
h.u..!...!u.... w.»
t - 2.... 1....

 

 

 

s'

2‘ " LIBRARY
2005 Michigan State
(AZ/5651012 University

This is to certify that the
dissertation entitled

LINKING GRADIENTS OF PREDATORS AND
PRODUCTIVITY TO THE COMPOSITION AND LIMITATION
OF BENTHIC ALGAL COMMUNITIES

presented by
Tara Lee Darcy-Hall

has been accepted towards fulﬁllment
of the requirements for the

Doctoral degree in Zoology

 

 

- fMajorfProfessor’s Signature

0 ”6712200?
Date

MSU is an Afﬁrmative Action/Equal Opportunity Institution

 

PLACE IN RETURN Box to remove this checkout from your record.
To AVOID FINES return on or before date due
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m
2/05 c:/ClRC/DatoDue.Indd-p.15

LINKING GRADIENTS OF PREDATORS AND PRODUCTIVITY TO THE
COMPOSITION AND LIMITATION OF BENTHIC ALGAL COMMUNITIES

By

Tara Lee Darcy-Hall

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Zoology
Program in Ecology, Evolutionary Biology & Behavior
W. K. Kellogg Biological Station

2004

ABSTRACT

LINKING GRADIENTS OF PREDATORS AND PRODUCTIVITY TO THE
COMPOSITION AND LIMITATION OF BENTHIC ALGAL COMMUNITIES

By

Tara Lee Darcy-Hall

Species often exhibit tradeoffs in their abilities to survive in different
environments. As a result, tradeoffs among species’ traits can lead to shifts in species
composition along ecological gradients. Furthermore, compositionally variable
communities usually respond differently to changes in the abundance of species within
adjacent trophic levels (1'. e. changes in their food or their predators). In particular,
communities along gradients in potential primary productivity can differ in the extent to
which they are limited by predators or resources. My work focuses speciﬁcally on one
tradeoff, between species’ abilities to compete for resources and resist predation, and how
this tradeoff mediates shiﬁs in benthic algal species composition and consequently alters
patterns of nutrient and predator limitation across gradients in productivity.

I investigate the competition-resistance tradeoff within the context of the keystone
predation hypothesis (KPH), which predicts that concomitant increases in resources and
predators will result in dominance by predator resistant species. Conversely, species that
are vulnerable, superior competitors should prevail in environments low in resources and
predator biomass. A mathematical model is used to generate predictions of how the
keystone predator tradeoff and shifts in species composition might affect patterns of algal

nutrient and predator limitation.

A survey of benthic algal communities in 14 southwest Michigan lakes indicated
that algal nutrient and predator limitation were similar to predictions of the mathematical
model and the KPH. Moreover, several species exhibiting strong correlations with
productivity and predator biomass possessed traits consistent with those expected given a
tradeoff among competition and resistance.

Similarly, in a large-scale mesocosm experiment focusing on the key role of
grazing predators in driving a competition-resistance tradeoff, I demonstrated that
ungrazed and grazed benthic algal communities differ substantially in their composition
and hence their patterns of nutrient and predator limitation across a productivity gradient.
The presence of more predator resistant algal species in grazed communities led to
reduced levels of predator limitation and stronger nutrient limitation. These results
provided a causative link between the competition-resistance tradeoff inherent in the
KPH and patterns of algal limitation.

Additionally, I used series of small and large-scale experiments as well as a
unique statistical technique, the fourth-corner analysis, to correlate the key traits of
competitive ability and predator resistance to environmental variables, providing a direct
test of the predictions of the KPH. The fourth-comer analysis revealed several trait-
environment correlations consistent with the KPH; however, results were context
dependent and demonstrated that other tradeoffs may also be important.

These results generally attest to the importance of a competition-resistance
tradeoff within the benthic algae. More importantly, I have elucidated a link between this

tradeoff, benthic algal composition, and patterns of algal nutrient and predator limitation.

Dedicated to my grandparents, William T. Lavey and the late Lorraine D. Lavey

iv

ACKNOWLEDGMENTS

The words that follow would never have been written without the unending
support, advice, and feedback from many individuals as well as the greater KBS
community. Rather than save the best for last, I begin by expressing my deepest
gratitude and love to my husband, Spencer Hall, whose loving, patient manner and
unfaltering support have been tantamount to the present accomplishment. I am especially
grateful to him for taking many hours to create the limitation model: he is the sole author
of Appendix A and the ﬁgures in Chapter 1 (and we thought we could never
collaborate!). Spencer has unofﬁcially been my sixth committee member, challenging
me to grow both intellectually and personally, and for that I am not only a better scientist,
but a better person as well.

I have been extraordinarily lucky to work with a brilliant and supportive
committee. My advisor, Gary Mittelbach, has consistently provided sage advice,
guidance, and reviewing over the past 6'/2 years. He was particularly supportive near the
end, when he patiently read (and re-read) my chapters, despite his hectic schedule. Kay
Gross has been an invaluable advocate during my time at KBS. Together with Spencer,
she and Gary have been the anchors without which I might never have persevered and
ﬁnished the task at hand. Alan Tessier was an indispensable reviewer and statistics guru,
whose manic and intellectual presence will be sorely missed at KBS. Jan Stevenson
introduced me to the beautiful and complicated world of the benthic algae and has kept
my work grounded within that context. Last, but not least, Mathew Leibold has been my

intellectual inspiration, helping me to ﬁnally gain conﬁdence in my credibility as a

scientist. My interactions with these ﬁve people, individually and as a group, have truly
been the highlights of my academic tenure at KBS.

While social outlets may be limited at KBS, the opportunities for deep and
enduring friendships have been anything but scarce. Natalie Dubois, Sarah Grandy,
Pamela Geddes, and Kristi Sherﬁnski are four intelligent, beautiful, and genuine women
who have graced me with their friendship. Their companionship has kept a smile on my
face and insanity at bay, and their friendship is my most valuable gift from KBS.

I count many, many others among my dearest friends at KBS, with whom I’ve
enjoyed countless conversations (scientiﬁc, political, and otherwise), game nights and
potlucks, interminable drives to campus, and of course, more than a few beers. In
particular, I would like to thank Chris Steiner, Eric Thobaben, Stuart Grandy, Meg Duffy, 2
Angie Roles, Kalina Manoylova, Amy Downing, Tom Getty (my fellow gossip), and
more recently Chad Brassil, Joyce Besch, Lena Sivars-Becker, and Claes Becker.

Much of my research would have been impossible without the help of many
people in the ﬁeld and the lab. S. Hall, Jeff Richardson, Jeremy Wojdak, Nate Dom, C.
Steiner, Erica Garcia, P. Geddes, Pam Woodruff, Drew Corbin, Madhav Machavaram, K.
Manoylova, G. Mittelbach, John Mittelbach, and my sister Christine Malumphy, have all
graciously helped me scrape clay pots, identify algae or measure snails over the years.

My warmest appreciation and thanks go to the incredible support staff at KBS.
These people are the pulse of this institution: Nina Consolatti, Alice Gillespie, Sally
Shaw, John Gorentz, Michelle Bosanoz, and J anell Lovan.

Finally, I would like to thank my parents, Butch Darcy and Jean Lavey-Fuhr, as

well as my siblings, Christine and Bryan, for a lifetime of love and support.

vi

TABLE OF CONTENTS

LIST OF TABLES .......................................................................................................... ix
LIST OF FIGURES .................................................. xi
CHAPTER 1
INTRODUCTION ............................................................................................................ 1
The keystone predation hypothesis ................................................................................... 1
A mathematical model for limitation ................................................................................ 3
Benthic algae as study organisms ..................................................................................... 6
Dissertation synopsis ........................................................................................................ 7
Literature Cited ............................................................................................................... 10
CHAPTER 2
PATTERNS OF NUTRIENT AND PREDATOR LIMITATION OF BENTHIC ALGAE
IN LAKES ALONG A PRODUCTIVITY GRADIENT ............................................... 12
Introduction ..................................................................................................................... 12
Methods ........................................................................................................................... 16
Study system — lakes ........................................................................................... l6
Limitation assays ................................................................................................ 18
Statistical analyses and algal enumeration methods ........................................... 20
Results ............................................................................................................................. 22
Limitation vs. lake productivity .......................................................................... 22
Species composition and environmental correlations ....... . .................................. 27
Discussion ....................................................................................................................... 33
Patterns of nutrient and predator limitation ........................................................ 33
Benthic algal compositional shiﬁs ...................................................................... 35
Linking species turnover to patterns of limitation .............................................. 38
Conclusions ......................................................................................................... 40
Literature Cited ............................................................................................................... 41
CHAPTER 3
LINKING HERBIVORE-MEDIATED SPECIES TURNOVER TO PATTERNS OF
BENTHIC ALGAL NUTRIENT AND PREDATOR LIMITATION ........................... 47
Introduction ..................................................................................................................... 47
Methods ........................................................................................................................... 50
Limitation metric ................................................................................................ 50
Study system ....................................................................................................... 51
Phase 1: Community establishment .................................................................... 51
Phase 2: Limitation assays .................................................................................. 53
Algal enumeration methods ................................................................................ 57
Statistical analyses .............................................................................................. 57
Results ............................................................................................................................. 59
Phase 1 — Algal community composition ........................................................... 59

vii

Phase 2: Limitation assays .................................................................................. 66

Discussion ....................................................................................................................... 69
Algal composition and the keystone predation hypothesis ................................. 69
Patterns of limitation of grazed and ungrazed communities ............................... 71
Conclusions ......................................................................................................... 75

Literature Cited ............................................................................................................... 76

CHAPTER 4

EVIDENCE FOR THE KEYSTONE PREDATION HYPOTHESIS IN THE BENTHIC
ALGAE: LINKING SPECIES’ TRAITS TO ENVIRONMENTAL VARIABLES IN A

FOURTH-CORNER ANALYSIS .................................................................................. 82
Introduction ..................................................................................................................... 82
Methods ........................................................................................................................... 85
Overview of the experimental approach ............................................................. 85
Outdoor mesocosm experiment .......................................................................... 85
Resource limitation bioassay .............................................................................. 88
Grazer resistance bioassay .................................................................................. 89
Statistical analyses .............................................................................................. 90
Results ............................................................................................................................. 93
Cattle-tank mesocosm experiment (the A- and C-matrices) .............................. 93
Resource competition and grazer resistance trait assays (the B-matrix) ........... 97
The fourth corner analysis ................................................................................ 101
Discussion ..................................................................................................................... 108
Literature Cited ............................................................................................................. 112
CHAPTER 5
SUMMARY .................................................................................................................. 116
APPENDICES .............................................................................................................. 120
APPENDIX A: THE LIMITATION MODEL ............................................................ 121
Model structure and description .................................................................................... 122
Resource and predator limitation in food chains and food webs .................................. 126
Literature Cited ............................................................................................................. 136
APPENDIX B: CHAPTER 2 APPENDICES ............................................................. 137
APPENDIX C: CHAPTER 3 APPENDIX .................................................................. 140
APPENDIX D: CHAPTER 4 APPENDICES ............................................................. 142

viii

LIST OF TABLES

CHAPTER 2

Table 2.1: Slopes, variation explained (R2), and signiﬁcance of regression statistics for
nutrient and predator limitation modeled as a ﬁmction of 4 explanatory variables: log
(total phosphorus) (TP), log (total nitrogen) (TN), log (chlorophyll a), and as a composite
of all three ....................................................................................................................... 26

CHAPTER 4

Table 4.1: Results of permutational MANOVAs on the relative biovolume of benthic
algal species composition for A) the nutrient supply-1i ght subset and B) the nutrient
supply-N:P ratio subset of the mesocosm experiment .................................................... 95

Table 4.2: D-matrix from a fourth-comer analysis, associating algal species traits from
trait bioassays to the environmental variables in the cattle tank mesocosm experiment.
All values are correlation statistics r(i,j) and probabilities (P) tested using 10,000
permutations. Values in bold are signiﬁcant at the P<0.05 level. None of the correlations
were statistically signiﬁcant at the table-wide level. A) F ourth-comer analysis without
light treatments. B) F ourth-corner analysis including light treatment .......................... 104

Table 4.3: Trait values (relative grth rates — RGR) and environmental correlations
(from the RDA) for algal species whose responses in the trait bioassays and the outdoor
mesocosm experiment drive the signiﬁcant correlations between algal traits and
environmental variables in the fourth-comer analysis .................................................. 107

APPENDIX B

Appendix 32.1: Abiotic and biotic variables measured in the 14 study lakes. Grazer
biomass is measured as ash-free dry mass (AF DM) .................................................... 138

Appendix 32.2: Correlations of the algal species in Figures 2.3 and 2.4 with the six
environmental variables included in the RDA. Correlations were calculated using the
species and environmental scores from the ﬁrst 4 ordination axes .............................. 139

APPENDIX C
Appendix C3.1: Slices of a signiﬁcant nutrient supply by grazer history interaction term

in the permutational MANOVA on the relative biovolume of benthic algal species
composition in Phase 1 (community establishment) mesocosms ................................. 141

ix

APPENDIX D

Appendix D4. 1: Two-way (A) and three-way (B) MANOVA’s and ANOVA’s from the
cattle tank mesocosm experiments. MAN OVA F- statistics and P-values are Pillai’ 3
Trace ............................................................................................................................. 143

Appendix D4.2: Species-environment correlations from RDA of cattle tank mesocosm
experiment. Species in bold are those included in the fourth-comer analysis. Species
with * were only present in the RDA including light treatments. Species are listed with a
taxonomic designation in parentheses: (D) = diatoms, (BG) = blue-green algae
(cyanophytes), (G) = green algae (chlorophytes) ........................................................ 144

Appendix D4.3: Post-hoe pairwise comparisons in the resource competition bioassay for
A) chlorophyll a (Tukey’s HSD), B) chlorophyll a (Tukey’s HSD) with low light
treatment, C) species composition (perMANOVA pairwise comparisons) with low light
treatment ....................................................................................................................... 146

LIST OF FIGURES

CHAPTER 1

Figure 1.1: Predator and resource limitation over gradients of enrichment in food chains
and food webs. Species 1 (solid line) is the more vulnerable, superior resource
competitor, while Species 2 (stippled line) is the less vulnerable, inferior resource
competitor. The dashed line is a region of coexistence of several species that are
increasingly predator resistant with enrichment ............................................................... 5

CHAPTER 2

Figure 2.1: Trophic level biomass as a function of primary productivity (total phosphorus
— TP) in 14 lakes in southwest Michigan. A) Benthic algal chlorophyll a accumulated on
control clay ﬂowerpots. B) Benthic grazer biomass (ash-free dry mass). All axes are on
log scales, but values are listed as non log-transformed values for ease of interpretation.
The statistical outlier, Douglas Lake, is labeled. Lines are linear regressions; the
regression for (A) does not include Douglas Lake ......................................................... 24

Figure 2.2: Nutrient limitation (closed circles) and predator limitation (open triangles) as
functions of 4 measures of lake potential primary productivity. X-axes of A-C are on log
scales, but values are given as non log-transformed values for ease of interpretation.
Lines are linear regressions of nutrient limitation (solid line) and predator limitation
(dashed line). Graphs are limitation versus: A) total phosphorus in the water column, B)
total nitrogen in the water column, C) chlorophyll a on the clay pots prior to the onset of
the limitation assays, and D) productivity composite .................................................... 25

Figure 2.3: Biplot of species scores (solid vectors) and environmental scores (dashed
vectors) from a redundancy analysis on algal species relative biovolume in 13 lakes.
Only species with a score on either axis of over 0.40 are included in the biplot. Species
are listed with a taxonomic designation in parentheses: (D) = diatoms, (BG) = blue-green
algae (cyanophytes), (G) = green algae (chlorophytes) ................................................. 28

Figure 2.4: Species-environment correlations from RDA for log-transformed total
phosphorus (TP; white bars) and log-transformed grazer biomass (black bars). Only
those species with correlations over 0.30 are included. Lettering of species is the same as
those listed in Figure 2.3, with two additional species: ao) Anomoneis vitrea (D) and ap)
Leptolyngbya sp. B (BG) ............................................................................................... 30

Figure 2.5: Nutrient limitation (closed circles) and predator limitation (open triangles)
vs. the lake site scores from the 2nd RDA axis (productivity-grazer axis). Lines are
linear regressions for nutrient (solid line) and predator limitation (dashed line) .......... 32

xi

CHAPTER 3

Figure 3.1: Experimental design for Phase 1 (community establishment) and Phase 2
(limitation assays). Open and closed circles represent terracota clay ﬂowerpot sampling
substrates. Arrows represent the movement of clay pots at the start of Phase 2 ........... 54

Figure 3.2: Means i 1 SE. of a) Total phosphorus, b) total nitrogen, c) chlorophyll a of
periphyton on terracota clay ﬂowerpots and (1) dissolved oxygen in cattle tank
mesocosms after 3 months of applied nutrient supply treatments. Closed circles are
systems containing herbivores and open circles are systems lacking herbivores ........... 60

Figure 3.3: Response of experimental benthic al gal communities to various nutrient
supply concentrations (CAP Axis 1) and the presence/absence of herbivores (CAP Axis
2). Ordination of site scores from CAP analysis. Open symbols are ungrazed mesocosms
and closed symbols are grazed mesocosms. Nutrient supply treatments are represented
by different shapes: LOW (circles), MED (triangles), HIGH (squares), and XHIGH
(stars) ............................................................................................................................... 62

Figure 3.4: a) Benthic algal composition divided into broad morphological and taxonomic
groups for ungrazed (O) and grazed (+) communities at each nutrient supply level. b)
Correlations of algal species with each CAP axis. Only species with correlations of over
0.35 are included. Species are listed with a taxonomic designation in parentheses: (D) =
diatoms, (BG) = blue-green algae (cyanophytes), (G) = green algae (chlorophytes) ..... 63

Figure 3.5: a) Nutrient and b) predator limitation of historically grazed (closed symbols)
andrungrazed (open symbols) benthic algal communities. Regression lines are LOWESS
ﬁts (tolerance = 0.6) for grazed (solid lines) and ungrazed (dashed lines) relationships
with net primary productivity (dissolved oxygen — DO) ............................................... 67

CHAPTER 4

Figure 4.1: Biplot of species scores (solid vectors) and environmental scores (dashed
vectors) from a redundancy analysis on algal species relative biovolume in the cattle tank
mesocosm experiment. Species are listed with a taxonomic designation in parentheses:
(D) = diatoms, (BG) = blue-green algae (cyanophytes), (G) = green algae (chlorophytes).
Only species with a correlation over 0.30 with either axis is included in the biplot.
Species in bold are those included in the fourth-comer analysis .................................... 96

Figure 4.2: Results of CAP analysis on algal species relative biovolume from resource
competition bioassay. A) Biplot of experimental treatments: low P (triangles), low Si
(stars), low nitrogen (circles), and controls (squares). B) Biplot of species correlations
with the CAP axes. Species are listed with a taxonomic designation in parentheses: (D) =
diatoms, (BG) = blue-green algae (cyanophytes), (G) = green algae (chlorophytes).
Species in bold are those included in the fourth-corner analysis .................................... 99

xii

Figure 4.3: Algal community composition and mean chlorophyll a (i 1 SE.) in the
grazer resistance bioassay ............................................................................................. 102

APPENDIX A

Figure A1.1: Biomass and nutrient partitioning along nutrient supply gradients in food
chains and food webs with plants and a grazer at equilibrium. In the food chain example,
two examples are shown: species 1 is a superior nutrient competitor (i. 2., lower R ) but rs
also more vulnerable to predation. As nutrient supply exceeds the plant’s R (level a for
species 1 and level b for species 2), biomass of the plant 1ncreases with enrichment until
the predator can invade (at level 0 for more vulnerable species, and level d for the less
vulnerable species). Once these thresholds are reached, plant biomass remains constant
(at the predators N ) w1th enrichment as predator biomass (P ) and free nutrient levels
(R ) 1ncrease. The rate at which P and R increase with enrichment 1s higher for species
1 (more vulnerable) than species 2 (less vulnerable). In the food web, coexistence is
possible at intermediate resource supply (between thresholds e and f) In this region, P
and R remain constant with enrichment, as increased nutrient supply rs reallocated
among species 1 (declining) and species 2 (increasing) .............................................. 127

Figure A1.2: Predator and resource limitation over gradients of enrichment in food
chains and food webs. In food chains, predator limitation is always higher for more
vulnerable species than less vulnerable species. Yet, resource limitation declines more
rapidly over the nutrient supply gradient for the more vulnerable species. At intermediate
resource supply, coexistence of the plants separates a low nutrient regime, dominated by
the superior nutrient competitor, to one dominated by the more resistant plant. In the
region of coexistence, predator limitation typically would decrease with enrichment while
resource limitation always does .................................................................................... 130

Figure A1.3: Predator and resource limitation in food webs with many plants displaying
the keystone-predation tradeoff of competitive ability versus predation resistance. Over
an enrichment gradient, plant assemblages become increasingly dominated by resistant
species. As a result, predation limitation remains low over a broad gradient of nutrient
supply (as compared to the food chain case). These increasingly resistant species
maintain a high level of resource limitation of the plant assemblage over this same
enrichment gradient ...................................................................................................... 134

xiii

CHAPTER 1
INTRODUCTION

Species exhibit tradeoffs in their ability to use various resources, their ability to
allocate biomass to various structures, and/or their ability to compete for resources or
colonize new habitats (reviewed in Kneitel and Chase 2004, Grover 1997, Tilman 1988).
Moreover, a species’ trait selected for in one environment may not prove favorable in a
different environment. Thus, species may exhibit tradeoffs between traits that confer
success in different environments, and these tradeoffs may drive shifts in species
composition along ecological gradients. Here, I focus on a tradeoff among benthic algal
species in their ability to procure nutrients and their resistance to predation. A
competition-resistance tradeoff has long been recognized as an important phenomenon
within ecological communities (Paine 1966, Lubchenco and Gaines 1981, Vance 1978,
Holt et al. 1994, Grover 1994, Leibold 1996, Chase et al. 2000). If prey species (e. g.
algae) exhibit a competition-resistance tradeoff, then prey community composition might
shift to dominance by more predator resistant species as both productivity and predator
biomass increase. This potentially critical role of predators in shifting prey community
structure has been investigated empirically (e.g. Steiner 2001 , Leibold 1999) and
developed theoretically as the keystone predation hypothesis (KPH; Leibold 1996, Holt et
al. 1994, Grover 1994).

The keystone predation hypothesis

The keystone predation hypothesis predicts that species compositional shiﬁs

across productivity gradients will involve a tradeoff between the ability of species to

escape predation and compete for resources. These broadly deﬁned terms actually

encompass a number of more speciﬁc, physiological or morphological species’ traits that
qualify as predator avoidance or competitive ability. For example, plants (or algae) can
be resistant to predators by possessing chemical or structural traits that augment defense.
Benthic algal species might be predator resistant by being tightly attached to their
substrate, by having a tangled, ﬁlamentous morphology that impedes grazing, or by
occupying a protected position low in the benthic mat. Algal species can also be predator
tolerant (sensu Chase et al. 2000), minimizing the effects of predation by rapid regrth
of affected tissues. Finally, algal species might overcome large predation effects by
maintaining a growth rate that is faster than the grazing rate.

Competitive ability in the KPH is deﬁned in the classic sense of resource
competition (Tilman 1982); that is, a species is competitive for a particular resource if it
can persist on low levels of that resource. A competitive species can be alternatively
viewed as one that acquires better access to a key resource (e. g. Grime 1979). Both
deﬁnitions of competition apply within the benthic algae and while both are addressed
within the following chapters, I focus primarily on the classic deﬁnition, as it is the one
used in the keystone predation hypothesis.

The competition—resistance tradeoff implicit in the KPH generates several
predictions concerning the nature of species compositional shifts along gradients in
productivity and predator biomass. First, the KPH predicts that edible, superior
competitors will dominate the prey community at low productivities, while more grazer
resistant, inferior competitors will dominate at high productivities. Second, the resultant
prey communities across this productivity and predator gradient can differ not only in

composition, but also in how they respond to changes within adjacent trophic levels.

It is well established that processes initiated from both the top-down and the
bottom-up are critical to community structure and biomass (Power 1992). A strong focus
in aquatic communities has been the relative importance of nutrients and herbivores on
primary producers within one system. However, surprisingly few studies have explored
how the relative effects of top-down (predator limitation) and bottom-up (nutrient
limitation) factors shift across communities along a productivity gradient, and no research
to date has attempted to attribute a mechanism to such patterns. Osenberg and Mittelbach
(1996) hypothesized that patterns of nutrient and predator limitation might relate to
herbivore-mediated shifts in prey species composition, thus linking the KPH to broader
food web interactions. In my work, I have adopted the deﬁnition and metrics of nutrient
and predator limitation as deﬁned by Osenberg and Mittelbach (1996). Limitation is
measured as the increase in the per capita growth rate of a population upon the removal
of a limiting factor (e. g. via the addition of nutrients or the removal of predators). The
objectives of my dissertation are to test the applicability of the KPH for benthic algal
communities, describe natural patterns of nutrient and predator limitation of lake benthic
algae, and to experimentally link the herbivore-mediated algal compositional shifts
predicted by the KPH to these patterns.

A mathematical model for limitation

I present a mathematical model in Appendix A that outlines the predicted patterns
of limitation given the parameters of two competing theories concerning trophic level
biomass and composition: the KPH and simple food chain models (Hairston et al. 1960,
F retwell 1977, Oksanen et al. 1981). Food chain models assume there is one species per

trophic level, while the KPH incorporates species heterogeneity within trophic levels.

Below, I refer to these theories as “food chain” and “food web” models, respectively.
The mathematics and assumptions resulting in the following predictions are explained in
thorough detail in Appendix A.

In food chains, resource (or nutrient) limitation is predicted to substantially
decrease along a gradient in nutrient enrichment (Figure 1.1), although this decline is
predicted to be slower for less vulnerable prey species (Species 2). Because less
vulnerable prey species persist at higher biomass, they deplete resources to a lower level,
causing higher resource limitation (Appendix A). Predator limitation of both prey types
should increase along a productivity gradient, but again at a slower rate for less
vulnerable prey, because mortality due to predation is lower.

In food webs, with compositional shifts to increasingly predator resistant species
with enrichment, resource limitation initially decreases, but then remains relatively
constant and high across much of the productivity gradient. Resource limitation remains
strong for less vulnerable species because these species maintain high biomass and thus
deplete their resources more per capita than their more vulnerable counterparts
(Appendix A). Therefore, resource limitation in the food web case is expected to be: a)
stronger than resource limitation in the food chain case (except at low levels of resource
supply), and b) stronger than predator limitation across gradients of enrichment (Figure
1.1). While predator limitation initially increases at low productivities due to the
presence of more vulnerable prey species, dominance by increasingly less vulnerable
species causes predator limitation to decrease and remain constant (or decrease slowly)

and low across a productivity gradient (Figure 1.1). These predictions are examined in

Food Chain Food Web

 

 

 

 

 

 

Species 1
.......... Species 2
3 — - - Coexistence
‘5 (several spP-I
'D
d)
a.
D.
C
o I~ l‘ l‘ w ‘
'41"- , — - - -
CB
3::
.§ ‘
_| a, \
u x
5 ‘~— ———————
3
0
(It
a:
a:

 

 

 

 

 

Nutrient Supply, 3

Figure 1.1: Predator and resource limitation over gradients of enrichment in food chains
and food webs. Species 1 (solid line) is the more vulnerable, superior resource
competitor, while Species 2 (stippled line) is the less vulnerable, inferior resource
competitors. The dashed line is a region of coexistence of several species that are
increasingly predator resistant with enrichment.

Chapters 2 and 3 using a survey of southwest Michigan lakes and a large-scale mesocosm
experiment.
Benthic algae as study organisms

Lake benthic algae are an excellent community for examining predicted changes
in species composition and limitation across productivity gradients. Freshwater benthic
algae are a diverse community of substrate-bound autotrophs. I have observed over 200
species in local lakes and they are broadly composed of 3 main taxonomic groups: the
glass cell-walled diatoms, green algae (chlorophytes), and the prokaryotic blue-green
algae (cyanophytes or cyanobacteria). Benthic algae can vastly range in size from small
unicellular diatoms (~ 30 um3) to relatively large ﬁlamentous species (~10,000 um3/cell),
and they embody a wide range of morphologies. For example, algal species can be
unicellular, colonial, gelatinous, mucilaginous, ﬁlamentous, or stalked. Furthermore,
benthic algae can dominate carbon ﬁxation in many lakes (Wetzel 1996), yet these
communities are often ignored when testing current ecological theory (Lowe 1996).

Benthic algal functional form theory (Rosemond 1996, Steinman et al. 1992)
predicts species compositional shifts that are similar to those predicted by the KPH. For
instance, benthic algal species’ traits that confer effective nutrient uptake (e. g. erect or
ﬁlamentous forms) render these species vulnerable to herbivores. In general, defended
species are inferior competitors, due in part to costs of defense as well as their inherent
morphology (e. g. adnate). Therefore, there is a priori evidence for a competition-
resistance tradeoff in the benthic algae, which may drive species compositional shifts and

potentially inﬂuence species’ responses to the removal of limiting factors.

Dissertation synopsis

Chapter 2 of my dissertation describes natural patterns of nutrient and predator
limitation of benthic algae across a productivity gradient of ‘14 lakes in southwest
Michigan. In this study, I compare predictions of limitation patterns generated from the
KPH and simpler food chain models (e. g. Oksanen'et al. 1981). Because infra-trophic
level heterogeneity (i. e. the KPH) is a more realistic model assumption, I hypothesized
that patterns of benthic algal limitation across lakes would more closely resemble those
predicted by keystone predator models. I also surveyed benthic algal species
composition across lakes in an effort to relate species compositional shifts to changes in
environmental variables and patterns of nutrient and predator limitation. Speciﬁcally, I
examined whether the KPH predictions of prey species shifts qualitatively matched
benthic algal species turnover along a productivity gradient of 13 Michigan lakes.

In Chapter 3, I experimentally link the process of herbivore-mediated species
turnover (i. e. the KPH) to patterns of benthic algal limitation. In 1000-L, cattle tank
mesocosms, I established two sets of algal communities along a productivity gradient:
one community that developed in the presence of herbivorous grazers and one that
developed in their absence. After several weeks of community establishment, I assayed
each community for nutrient and predator limitation and then compared the patterns
between historically grazed and ungrazed communities. I hypothesized that algal
communities historically exposed to herbivores would exhibit herbivore-mediated shifts
in composition, and thus patterns of limitation would be similar to those predicted by the
KPH. In contrast, I expected that the composition of historically herbivore-naive

communities would shift solely on the basis of differences in productivity and that

predator resistant species would not dominate at higher productivities, generating patterns
of limitation resembling predictions from simple food chain models.

In Chapter 4, I evaluate the applicability of the focal'KPH tradeoff (i. e.
competitive ability vs. predator resistance) in benthic algal communities. In a series of
small mesocosm experiments, I measured several algal species’ traits, including
resistance to several herbivores and tolerance to limiting levels of several key resources
(i. e. nutrients and light). Concurrently, I established algal communities in large
mesocosms that varied in their light environments, nutrient concentration, and level of
herbivory. Finally, data from these experiments were assembled into matrices (a species
x species’ trait matrix, an environment x site matrix [abiotic and biotic variables from the
large mesocosms], and a species x site matrix [also from the large mesocosms]). These
matrices were analyzed using the fourth-corner method, which generates a fourth matrix
correlating species’ traits to environmental variables (Legendre et al. 1997). The KPH
predicts that grazer resistance will dominate at high productivities and superior
competitive ability will dominate at low productivities. Thus, I hypothesized that the
species’ traits of resistance to predators would be positively correlated with productivity
variables; whereas, traits of superior competitive ability would be negatively correlated
with productivity variables.

Findings from my dissertation research demonstrate the importance of a
competition-resistance tradeoff within benthic algal communities, not only for altering
species composition, but also for affecting patterns of nutrient and predator limitation.
Overall, this dissertation research addresses several novel and ecologically relevant

questions concerning patterns and mechanisms of algal limitation across environments,

while successfully combining rigorous theory, descriptive studies in natural systems, and

controlled experiments.

LITERATURE CITED

Chase, J. M., M. A. Leibold, A. L. Downing, and]. B. Shurin. 2000. The effects of
productivity, herbivory, and plant species turnover in grassland food webs.
Ecology 81: 2485- 2497.

Chase, J. M., M. A. Leibold, and E. Sims. 2000. Plant tolerance and resistance in food
webs: community-level predictions and evolutionary implications. Evolutionary
Ecology 14:289-314.

Grime, J. P. 1979. Plant Strategies and Vegetation Processes. Wiley, London, UK.

Grover, J. P. 1997. Resource Competition. Chapman and Hall, New York, New York,
USA.

Grover, J. P. 1994. Assembly rules for communities of nutrient-limited plants and
specialist herbivores. American Naturalist 143:158-282.

Holt, R. D., J. P. Grover, and D. Tihnan. 1994. Simple rules for interspeciﬁc dominance
in systems with exploitative and apparent competition. American Naturalist
144:741-771. .

Knietel, J. M. and J. M. Chase. 2004. Trade-offs in community ecology: linking spatial
scales and species coexistence. Ecology Letters 7:69-80.

Legendre, P., R. Galzin, M. L. Harmelin-Vivien. 1997. Relating behavior to habitat:
solutions to the fourth-comer problem. Ecology 78:547-562.

Leibold, M. A. 1996. A graphical model of keystone predators in food webs: trophic
regulation of abundance, incidence, and diversity patterns in communities.
American Naturalist 147:784—812.

Leibold, M. A. 1999. Biodiversity and nutrient enrichment in pond plankton
communities. Evolutionary Ecology Research 1:73-95.

Lowe, R. L. 1996. Periphyton patterns in lakes. In Algal Ecology: Freshwater Benthic
Ecosystems (R. J. Stevenson, M. L. Bothwell, and R. L. Lowe, eds.), pp. 57—76.
Academic Press, San Diego, California, USA.

Lubchenco, J. and S. D. Gaines. 1981. A uniﬁed approach to marine plant-herbivore
interactions. 1. Populations and communities. Annual Review of Ecology and

Systematics 12:405-437.

Oksanen, L., S. D. Fretwell, J. Arruda, and P. Niemela. 1981. Exploitation ecosystems
in gradients of primary productivity. American Naturalist 118:240-261.

10

Osenberg, C. W. and G. G. Mittelbach. 1996. The relative importance of resource
limitation and predator limitation in food chains. In Food webs: Integration of
patterns and dynamics. (G. A. Polis and K. O. Winemiller, eds), pp. 134-148.
Chapman and Hall, New York, New York USA.

Paine, R. T. 1966. Food web complexity and species diversity. American Naturalist
100:65-75.

Power, M. E. 1992. Top-down and bottom-up forces in food webs: do plants have
primacy? Ecology 73:733-746.

Rosemond, A. D. 1996. Indirect effects of herbivores modify predicted effects of
resources and consumption of plant biomass. In Food webs: Integration of
patterns and dynamics. (G. A. Polis and K. O. Winemiller, eds), pp. 149-159.
Chapman and Hall, New York, New York USA.

Steiner, C. F. 2001. The effects of prey heterogeneity and consumer identity on the
limitation of trophic-level biomass. Ecology 82:2495-2506.

Steinman, A. D., P. J. Mulholland, and W. R. Hill. 1992. Functional responses
associated with growth form in stream algae. Journal of the North American
Benthological Society 1 1:229-243.

Tilman, D. 1982. Resource Competition and Community Structure. Princeton
University Press, Princeton, New Jersey, USA.

Tilman, D. 1988. Plant Strategies and the Dynamics and Structure of Plant
Communities. Princeton University Press. Princeton, NJ, USA.

Vance, R. R. 1978. Predation and resource partitioning in one-predator-two prey model
communities. American Naturalist 112:797-813.

Wetzel, R. G. 1996. Benthic algae and nutrient cycling in lentic freshwater ecosystems.
In Algal Ecology: Freshwater Benthic Ecosystems (R. J. Stevenson, M. L.
Bothwell, and R. L. Lowe, eds.), pp. 641-667. Academic Press, San Diego,
California, USA.

11

CHAPTER 2
PATTERNS OF NUTRIENT AND PREDATOR LIMITATION OF BENTHIC
ALGAE IN LAKES ALONG A PRODUCTIVITY GRADIENT
INTRODUCTION
After decades of debate, ecologists now concur that bottom-up and top-down
forces simultaneously inﬂuence trophic level biomass and community structure (Oksanen
1981, McQueen et al. 1986, Power 1992, Osenberg and Mittelbach 1996). While many
studies have documented the relative effects of these forces within a community or
ecosystem (see reviews in Hillebrand 2002, Downing et a1. 1999, Carpenter and Kitchell
1993, Power 1992), much of this research has focused on the effects of nutrient supply
and/or top predators on whole trophic levels within one system, often ignoring
heterogeneity within trophic levels and between ecosystems (Power 1992, Hunter and
Price 1992). Few studies have attempted to quantify how the relative importance of
bottom-up (i. e. resource limitation) and top-down (i. e. predator limitation) factors
changes along environmental gradients (e. g. Hillebrand 2001), and none have done so
while linking the observed patterns to species heterogeneity (i. e. the presence of multiple
species) within a focal trophic level.
Simple food chain models (Hairston et al. 1960, F retwell 1977, Oksanen et al.
1981) provide a useful benchmark for generating predictions of how resource and
predator limitation change across gradients in potential primary productivity. The classic
ecosystem exploitation hypothesis (Oksanen et al. 1981) posits that top-down and
bottom-up limitation of a species or tropho-species will alternate in importance,

depending upon the number of trophic levels in a system. For ecosystems with an even

12

number of trophic levels, top-down limitation is predicted to dominate odd-numbered
trophic levels (e. g. primary producers and secondary consumers), regardless of the
overall productivity of the system. Therefore, across systems that vary in productivity,
but not in the number of trophic levels, resource (or nutrient) limitation of the primary
producer trophic level is expected to decrease in importance as nutrient supply increases,
while predator limitation is expected to increase in importance as consumer biomass
increases (Darcy-Hall, Chapter 1, Appendix A, Oksanen et al. 1981).

Alternative models predict how species heterogeneity will drive patterns of
trophic level biomass and community structure across ecosystems that vary in potential
primary productivity. For example, Osenberg and Mittelbach (1996) incorporate prey
heterogeneity in their predictions of patterns of resource and predator limitation across
potential primary productivity gradients. Speciﬁcally, they predict that resource
limitation should remain strong across systems that vary in productivity, due to a shift in
the relative abundances of edible versus inedible prey (Osenberg and Mittelbach 1996).
This expectation is in agreement with results of my limitation model, which predicts that
resource limitation in a food web context will be strong and positive, but will decrease
non-linearly across an enrichment gradient (Darcy-Hall, Chapter 1, Appendix A).
Several ﬁeld studies in aquatic systems have observed a positive correlation between the
biomass of adjacent trophic levels (Mills and Schiavone 1982, McNaughton et al. 1989,
Ginzburg and Akcakaya 1992, Leibold et al. 1997, Leibold 1999), indicating that
herbivores and primary producers increase concomitantly with productivity. This rise in
herbivore biomass can favor dominance by predator resistant forms among the primary

producers. These predator resistant forms incur very low levels of predator-induced per-

13

capita mortality rates and can therefore increase in abundance as nutrient levels increase
(Osenberg and Mittelbach 1996). Thus, as the nutrient supply of a system increases, one
would expect resource limitation to remain important for the producer trophic level and
for predator limitation to decrease as more grazer-resistant prey dominate (Darcy-Hall,
Chapter 1, Appendix A). However, Osenberg and Mittelbach’s (1996) verbal model
lacks a mechanism to drive the changes in prey heterogeneity that account for the
predicted patterns in resource and predator limitation.

The keystone predation hypothesis (KPH; Holt et al. 1994, Grover 1994, Leibold
1996) provides a mechanistic basis for the predictions laid out by Osenberg and
Mittelbach (1996). The KPH incorporates a tradeoff among prey species in their
competitive ability and resistance to predation and focus on the role of predators in
driving turnover among these species (Holt et al. 1994, Grover 1994, Leibold 1996). For
example, as the biomass of herbivores increases across a nutrient supply gradient, the
dominant algal species are expected to shift from edible, superior resource competitors at
low productivities to predator resistant, inferior resource competitors at higher
productivities. Thus, the KPH tradeoff facilitates predictions of how prey composition
should shift across productivity gradients and in turn provides a mechanistic
understanding of the consequent changes in resource and predator limitation.

I use the deﬁnition and metrics of resource (here, nutrient) limitation and predator
limitation as presented by Osenberg and Mittelbach (1996) to examine patterns of benthic
algal (i.e. periphyton) limitation in 14 southwest Michigan lakes that span a wide gradient
in potential primary productivity. Limitation is deﬁned as the extent to which a

population’s per capita growth rate is reduced by predators or insufﬁcient nutrient supply

14

(Osenberg and Mittelbach 1996). Limitation can be measured as the increase in the
population’s per capita growth rate upon the removal of the limiting factor, via the
addition of saturating nutrients or the removal of predators. The limitation metric aims to
directly measure the change in growth rate due solely to the limiting factor and therefore
must be measured in the short-term to avoid confounding the measurement with a
numerical (or functional) response of predators or a switch to limitation by another
resource (Osenberg and Mittelbach 1996, Downing et al. 1999). Downing et al. (1999)
found that in an experimental setting, limitation was a constant and reliable measure of
the change in population per capita growth of marine phytoplankton over a period of 2-7
days, post treatment initiation.

There are several reasons why assemblages of benthic algae are an excellent
experimental system with which to address questions of nutrient and predator limitation.
First, algal communities have the potential for a rapid response to the removal of limiting
factors, since algal division rates are on the order of approximately one generation per
day (Downing et al.1999). Second, the majority of studies addressing sources of
limitation of autotrophs have focused on phytoplankton communities. Finally, it is well
established in stream periphyton communities that both resource and predator limitation
are important factors affecting community physiognomy (physical structure) and biomass
(McCormick and Stevenson 1989, Mulholland et al. 1991, Hill et al. 1992, Rosemond et
al. 1993), but these patterns have rarely been investigated in lakes (Lowe 1996).

I compared the relationship between algal nutrient and predator limitation and
several productivity variables to assess how natural patterns of algal limitation compare

to predictions from classic ecosystem exploitation models and more recent keystone

15

predator models. I hypothesized that patterns of benthic algal limitation across lakes
would more closely resemble those predicted by the KPH due to heterogeneity within a
trophic level. That is, I expected nutrient limitation of benthic algae to initially decrease
at low productivities, but remain strong and positive at higher productivities.
Alternatively, I expected predator limitation values to be low and perhaps decline along
the productivity gradient of lakes (Darcy-Hall, Chapter 1, Appendix A). I also surveyed
benthic algal species composition across lakes in an effort to relate species compositional
shifts to changes in environmental variables and patterns of nutrient and predator
limitation. Speciﬁcally, I examined whether the KPH predictions of prey species shifts
qualitatively matched benthic algal species turnover along this productivity gradient of
lakes. Finally, I sought to link patterns of algal limitation to changes in algal composition
across lakes.
METHODS
Study system - lakes
My primary focus was to investigate patterns of algal limitation across lakes
(versus within lakes), therefore, I chose to sample more lakes rather to sample intensively
within a few lakes. The 14 study lakes are located within 50 km of the W. K. Kellogg
Biological Station (KBS; Hickory Comers, Michigan, USA) and encompass a wide
gradient in both total phosphorus (TP: 13.5 — 77.5 ug/L) and total nitrogen (TN: 208.9 -
1869 rig/L), both of which are potentially limiting nutrients for primary productivity in
these lakes. I used both of these variables, as well as algal standing crop (i. e. chlorophyll
a), as surrogates for system primary productivity. Total nutrients were sampled in June

2001 in the littoral zone of each lake and measured using spectrophotometry and standard

16

methods (Bachmann and Canﬁeld 1996, Crumpton et al. 1992, Prepas and Rigler 1982,
APHA 1980). All chlorophyll a measurements were conducted using cold ethanol
extractions followed by narrow-band ﬂuorometry (sensu W-elschmeyer 1994). These
productivity measures, as well as conductivity, pH, and grazer biomass, were included in
ordinations relating environmental variables to benthic algal species composition. All
variables except pH and conductivity required log-transformation to meet assumptions of
normality. Conductivity and pH measurements for all lakes were obtained from
independent data sets (Stephen J. Hamilton, KBS, and Robert Bachmann, University of
Iowa, unpubl. data), from which 2001 data were included when available, otherwise
summer values from other years were used. Prior to the limitation assays, macrograzers
were collected from sampling substrates (terracotta clay ﬂowerpots) and preserved in
70% ethanol for later identiﬁcation and measurement of ash-free dry mass (AF DM).
Herbivorous grazers in these lakes are primarily gastropods (Physa spp., Helisoma sp.,
Gyrauius sp., Amnicola sp, and Bithynia sp.), which ranged from 1% to 99% of grazer
biomass. Other dominant grazers included amphipods (Hyalla azteca), mayﬂy larvae
(Family Caenidae), caddisﬂy larvae (order T richoptera), and chironomid midge larvae
(family Chironomidae). All but one of the study lakes support four trophic levels:
benthic primary producers, their invertebrate grazers (insect larvae and gastropods),
invertebrate (insect larvae) and vertebrate (ﬁsh) predators, and piscivores (one lake, Duck
Lake, lacks piscivorous ﬁsh).
Limitation assays
Benthic algal communities were established on 10-cm (4-inch) terracotta clay

ﬂowerpots in all lakes for 30 days prior to the start of the assays. In each lake, I placed

17

16 clay pots each in separate 4.7 L(S-quart) plastic buckets which had 3, 15 cmz-windows
cut in them to allow water, algae, and grazers to move through the container. The
buckets were attached to ﬂoating wooden frames that allowed the clay pots to be
submerged in the lake water. Frames were anchored in the littoral zone in a linear array.
The clay pots were placed close to the surface in all lakes, thus eliminating differences in
light levels across lakes. I monitored algal biomass during a 30-day incubation period to
determine when a substantial algal mat had developed for the limitation assays. Each
week, one clay pot from each lake was sampled by scraping off of all algae with a
toothbrush, razor blade, and deionized water. The resulting algal slurry was
homogenized in a blender and then subsampled for chlorophyll a analysis.

Identical limitation assays were initiated in all 14 lakes over a 4-day period
beginning on July 25, 2001. Due to the distance and travel time between lakes, I was
only able to set up these assays in 3-4 lakes per day. Each assay included three,
randomly assigned treatments with 4 replicates each: nutrient additions, grazer removals,
and controls. In nutrient addition treatments, I added a biologically saturated nutrient
solution (NH4C1, NaN03, and KH2P04 in a 16:1 molar N:P ratio) to the interior of the
clay pots. The same nutrient solution was added to the clay pots in each lake. The porous
clay surface readily diffuses nutrients over time (Darcy-Hall unpubl. data, Fairchild et al.
1985, Marks and Lowe 1989, Wootton et al. 1996). Lake water was added to the interior
of control clay pots. Macrograzers (i. e. visible to the naked eye) were manually removed
from all clay pots and returned to all but the grazer removal treatment. Control
treatments mimicked natural conditions in that no nutrients were added and grazers were

present. After the treatments had been applied, the clay pots were placed in new plastic

18

buckets with lOS-um mesh covering the windows, with one window removed from all
plus-grazer buckets (controls and nutrient additions) to allow grazer migration.

Each limitation assay lasted for 6 days, after which benthic algal communities
from each pot were harvested (as described above) and transported on ice back to the
laboratory. Any benthic grazers present on the clay pots were collected for future
identiﬁcation and biomass measurements. Algae on clay pots from the same treatment
within a lake were combined into one composite sample per treatment; thus, there were
three algal slurries (1 per treatment) per lake. In general, composite sampling aims to
reduce the effects of spatial variability on composition or biomass among clay pots of the
same treatment and also reduces the number of samples that needed to be processed in
one day. While this approach did not allow me to measure intra-treatment variability
within a lake, my primary question in this study concerned patterns between lakes,
therefore I focused sampling intensity across versus within lakes.

Composite algal slurries were subsampled for chlorophyll a and community
composition. Chlorophyll a analyses were conducted on the same day that the algal
communities were harvested. The chlorophyll a data were used to calculate values of
nutrient and predator limitation. Limitation values were calculated using the following
equations (Osenberg and Mittelbach 1996, Downing et al. 1999):

LimN = ln(NA)t— ln(C) (la)

ln(GR) — ln(C)

 

(1b)

 

Limp =

where t is duration of the experiment in days, LimN is nutrient limitation and Limp is
predator limitation, and NA, GR, and C are the chlorophyll a measurements of the

nutrient addition, grazer removal, and control treatments, respectively. Values of

19

limitation are measures of the change in instantaneous per-capita growth rate of algal
communities after the removal of a limiting factor. Positive values of limitation indicate
that algal communities cumulatively increased in biomass, while negative values indicate
that algal biomass decreased.
Statistical analyses and algal enumeration methods

Average values of nutrient and predator limitation across lakes were calculated
for comparison with those observed in other across-system studies (e. g. Downing et al.
1999, Osenberg and Mittelbach 1996). Bias-corrected 95% conﬁdence intervals around
averages of limitation were bootstrapped (sensu Dixon 2001) using Matlab v. 5.3
(Mathworks, Inc. 1999). Values of nutrient and predator limitation were regressed
against each of the 3 productivity measures (TN, TP, and chlorophyll a) as well as a
composite variable represented by the ﬁrst principal components axis (64% explained
variance) from a principal components analysis (PCA) on the 3 original productivity
variables (PC loadings: TP: 0.79, TN: 0.84, chlorophyll: 0.76). I used the average of
chlorophyll a on clay pots measured immediately prior to the limitation assays and
chlorophyll a on control clay pots at the termination of the limitation assays as a
productivity measure in statistical analyses (regressions and ordinations). The average of
chlorophyll a over these two samplings (Appendix B2.1) provides a more accurate
representation of algal biomass over the 6-day limitation assay than either sampling
alone. The relationships between both sources of limitation and the productivity
variables were quantiﬁed using simple linear regressions (Systat v.8, SPSS, Inc.), while
the means of nutrient and predator limitation were compared with a paired t-test.

Algal communities from control clay pots were used to examine patterns of

20

species compositional change along the gradient of lakes. In one lake, Wintergreen Lake,
the control algal composition sample was lost, so algae were enumerated from the grazer
removal treatment. Because composition within lakes was more similar than composition
between lakes (data not shown), I did not expect the small differences between species
relative biovolume in within-lake treatments to change the overall interpretation of
results.

Benthic algal species were enumerated and identiﬁed using established methods
(Lowe and Pan 1996). Semi-permanent wet mounts of a sample were prepared and
examined at 125 0x magniﬁcation on a compound microscope. Three hundred natural
units (e. g. a cell, ﬁlament or colony) were enumerated and identiﬁed to the species level
(when possible) for each sample. In most cases, a second slide was prepared with acid-
cleaned diatoms mounted in Naphrax (Patrick and Reimer 1966) to identify diatoms to
the species level. A subsample of 10-15 organisms per species was measured for
biovolume calculations, using published geometric forms and volume equations for
various algal genera (Hillebrand et al. 1999). All algal community composition results
presented here are in units of relative biovolume.

Algal species composition was related to lake environmental variables using
redundancy analysis (RDA; Canoco v.4). Algal relative biovolume data were arcsine-
square root transformed prior to analysis. These data were additionally transformed
using a Hellinger distance transformation to avoid violating the strict linearity
assumptions of RDA (Legendre and Gallagher 2001).

RDA uses eigenanalysis to partition the variation in algal species composition

among several ordination axes (the ﬁrst 4 in Canoco v.4), which are constrained to

21

represent some combination of the environmental variables. RDA also uses a Monte
Carlo permutation method to test whether a signiﬁcant portion of the variation in algal
composition is explained by the ordination axes. Scores of sites (i.e. lakes) and species
are generated for each ordination axis and these values represent that site or species’
correlation with a given axis. Scores from any combination of two axes can be used to
graphically represent the results in a biplot. In addition, an RDA generates scores along
the ordination axes for each environmental variable and these can be transposed, as
vectors, on species’ biplots. In a species’ biplot, the angles between species’ vectors are
reﬂections of the correlations between species (i.e. how they often coexist) and the angle
between a species’ vector and an environmental vector reﬂects the correlation between
that species and a particular environmental variable. However, biplots only show
correlations in 2 dimensions (i. e. axes), so to better understand the relationship between a
species and an environmental variable, I calculated the correlations using the scores from
all 4 ordination axes. I use species-environment correlations to explore how algal species
composition across lakes relates to predictions of the KPH. Finally, I regress lake site
scores against both nutrient and predator limitation to examine how species composition
contributed to the observed patterns. Lake site scores are representative of the relative
positions of lakes in overall algal species composition and thus provide a univariate
measure of species composition.
RESULTS
Limitation vs. lake productivity
There was a generally positive relationship on a log scale between water column

TP and average algal biomass accumulation, measured as chlorophyll a (Figure 2.1A; r =

22

0.36, P > 0.19). Total phosphorus was also positively correlated with benthic grazer
biomass (Figure 2.18; r = 0.61, P = 0.020), indicating that the abundances of both
primary producers and herbivores were positively correlated with productivity. When a
clear statistical outlier (Douglas Lake; studentized residual = -3.7) was removed from the
TP-chlorophyll a analysis, chlorophyll was more strongly correlated with TP (r = 0.53, P
= 0.06). Further analyses do not include data from Douglas Lake and are restricted to the
other 13 lakes.

Values of nutrient and predator limitation as well as environmental variables are
listed for each lake in Appendix B2.l. In 13 out of the 14 lakes sampled, nutrient
limitation was positive, indicating that algal biomass increased upon the addition of
nutrients. The average value (i 95% conﬁdence interval) of nutrient limitation across all
lakes was 0.083 i 0.023/day, which is a value similar to those reported by Downing et al.
(1999) for marine phytoplankton (average ~ 0.1/day) and by Osenberg and Mittelbach
(1996) for freshwater benthic algae (0.098/day). Nutrient limitation did not exhibit a
consistent or signiﬁcant relationship with any of the productivity variables, but, similar to
expectations, nutrient limitation did remain strong in several lakes, despite high
productivity (Figure 2.2, Table 2.1). In fact, nutrient limitation was signiﬁcantly stronger
than predator limitation (average = —0.019 i 0.010/day) across nearly all lakes (paired t-
test: t1; = 4.38, P = 0.001), which is similar to the pattern observed in a survey of aquatic
systems by Osenberg and Mittelbach (1996). Interestingly, average predator limitation
was negative (i.e. the 95% conﬁdence interval around predator limitation average does

not overlap zero; Figure 2.2), indicating that algal biomass decreased after the removal of

23

w
P
on

I

15.8

 

Chlorophyll a (mg/m2)
S”
00
l

p Douglas Lake

2.5 1 1 l 1 l r 1 J
20 32 50 79

 

 

A

o

8
-l
ﬂw

100

_\
O

 

Grazer biomass (mg/m2)

 

 

0 l l l l l l l l
1 3 20 32 50 79

TP (pg/L)

Figure 2.1: Trophic level biomass as a function of primary productivity (total phosphorus
— TP) in 14 lakes in southwest Michigan. A) Benthic algal chlorophyll a accumulated on
control clay ﬂowerpots. B) Benthic grazer biomass (ash-free dry mass). All axes are on
log scales, but values are given as non log-transformed values for ease of interpretation.
The statistical outlier, Douglas Lake, is labeled. Lines are linear regressions; the
regression for (A) does not include Douglas Lake.

24

 

 

 

 

 

 

 

 

 

 

0.2 0.2
A B
s: . ,
“' .1
E 0.1— . o 0.1
G? 0
3- ° - .
C .13
33 J 6 o
:3 O-O'Zf‘z-{z ....... .. 0.0
:E 7 ...""'-‘, .............
.l ...
-0.1 l J 1 1 1 1 ' ”‘ -0_1 1 L 1 f‘ 1 a1 4
13 20 32 50 79 178 355 708 1413 2818
TP lug/L) TN (pg/L)

 

 

0.2

P
N

0.1 - ~ 4 '

P
.n
1
$
0

Limitation (per day)

A
I J
,
7 O ' '
C ,
o o I ’-
I
4‘ v

 

 

 

 

 

 

0.0 :‘m . 73...",“W o
.0.1 L‘A‘l 1 1 481 1 '01 1 191 1 1 1 A.
0.6 1.0 1.6 2.5 4.0 ' -1.0 0.0 1.0 2.0
Chlorophyll a (pg/cm?) Productivity Composite

Figure 2.2: Nutrient limitation (closed circles) and predator limitation (open triangles) as
functions of 4 measures of lake potential primary productivity. X-axes of A-C are on log
scales, but values are given as non log-transformed values for ease of interpretation.
Lines are linear regressions of nutrient limitation (solid line) and predator limitation
(dashed line). Graphs are limitation versus: A) total phosphorus in the water column, B)
total nitrogen in the water column, C) chlorophyll a on the clay pots, and D) productivity
composite.

25

Table 2.1: Slopes, variation explained (R2), and signiﬁcance of regression statistics for
nutrient and predator limitation modeled as a function of 4 explanatory variables: log
(total phosphorus) (TP), log (total nitrogen) (TN), log (chlorophyll a), and as a

composite.

 

 

 

 

Variable slope R2 F-statistic P-value
Nutrient Limitation

log TP 0.03 0.000 0.253 0.625
log TN 0078 0.137 2.897 0.117
log chlorophyll a -0.119 0.123 2.687 0.129
composite -0.019 0.092 2.213 0.165
Predator Limitation

log 113 -0.076 0.176 3.560 0.086
log TN 0007 0.000 0.036 0.854
log chlorophyll a -0.016 0.000 0.071 0.795
composite -0.012 0.028 1.350 0.270

 

26

grazers. Predator limitation did not exhibit signiﬁcant associations with any of the lake
productivity measures, but tended to decline along a gradient in TP (Figure 2.2, Table
2.1). The signiﬁcance of the regression analyses did not change with the inclusion of the
outlier, Douglas Lake.

Species composition and environmental correlations

The ﬁrst 4 RDA axes explained a signiﬁcant portion (total = 52.5%) of the
variation in algal species composition across lakes (F = 1.998, P = 0.015), accounting for
25%, 11.7%, 9.9%, and 5.9% of the total variation, respectively. The ﬁrst RDA axis was
highly correlated with conductivity (r = 0.86) and thus, algal species primarily sorted
across lakes based on levels of conductivity. The second RDA axis correlated positively
with the productivity variables (TP: r = 0.81, TN: r = 0.59, and CHL: r = 0.52) as well as
with grazer biomass (r = 0.41) and pH (r = 0.74). The third RDA axis correlated
negatively with TN (r = -0.55), while the fourth RDA axis correlated positively with
CHL (r = 0.67). The RDA biplot (Figure 2.3) uses axis scores from the conductivity and
overall productivity-grazer biomass axes (axes l and 2, respectively).

Several benthic algal species loaded strongly with the ﬁrst RDA (conductivity)
axis (Figure 2.3). Diatom species exhibited especially strong positive correlations with
the conductivity vector. For example, Achnanthes exilis, Denticula keutzingii,
Encyonopsis microcephala, Encyonema sp., and Navicula radiosa, all had correlations
over 0.55 with conductivity (see Appendix B2.2 for all species-environment
correlations). Alternatively, several chlorophytes demonstrated strong, negative
correlations with the conductivity vector. For instance, the ﬁlamentous Stigeoclom'um sp.

and the colonial Ankistrodesmusfalcatus, Oocystis spp., Pediastrum duplex, and

27

3V =5...qu mEocochoG Em
a. 2.5355 «2.2.0.3500 36
89 ...Eoec.sm. 5.585.... 2....
5. £30.63...» c.2552“. in

g 86333.36 «32636.88 an
AOL Eacoboa .553..on cm

5. < dm 83.88.35 Em

av dam Eacouonoo 093 8m
68 «32...... 2.880050 Ca
58 < dm 033%:qu Aom

Ag can? szwonocoom 6m

78 .832. 8.53..on 8m

6m. .382 53.85.... an

A9 .3 23.20.8095 Em

Ag .3... 9.8.800 AN

68 3.228 638833.. A»
AOL «535 3.52.9335. Ax
A0. dam Eacououoo =me ts
69478 8.82 9

5v chanson an.n.ch< A:

8. 3...... 28.32.: 2

8V dam E.....:ou0boo E28: 3
59 3.58» 8650233.. C
88 39:32.... 63889.5 3
8m. m .3 8.82 a

59 22.83 Eagoqmoﬁsio 3
EV 8.9... 252......m 2

6m. .95 5.525 E.

8. 28....6. 388...). c

E. ...£.Em 228.83. C.

a. $8.353 3.35.6 ..

8. .5558. 33.....8 :

89 n. d... 285.28.. E
83...: 8585.3. 3

A09 dm nascent“? c

5. $2300.98 93053.6 3
R: dm oEozoxucm 3

av mdm Eathmoo 8
5:928 qucocQEoe 3

R: 80.52 $3.32 An

oé

md

m_xm 65¢ a?
0.0

.AmotEQBOEoV

mam—m 58w 0 ADV .Amo§:mo§>ov new—m cavemen—n H AOmv .8536 H Rd “8855.8.“ 5 nosey—wave 08.95me a 5:5 Bum:
8d 86on .833 05 E vows—o5 2m ovd 5.5 mo 38 856 so 808 a £5 86on .25 .863 2 E 082303 9522 86on
Rm? so max—«Sc. 552360.. a Bob $.88.» 8:33 8.53 35:89.35 98 A898.» 23$ 8.88 360% me 335 ”mm Emma

md- o. T

 

 

 

_

an

I]

v3

,,,..;7 >.au.nu.ou
on

N10

5

.u .3

 

VClEl puZ

srxe

04

28

Scenedesmus bijuga, all exhibited correlations over —O.6O with the conductivity axis.

A suite of benthic algal species loaded strongly along the productivity/grazer
biomass axis (Figure 2.3). The species-environment associations along this axis are
particularly important for examining species turnover in a KPH context, so they are
presented in more detail. In most cases, the species-productivity (TN, TP, or CHL)
correlation and the species-grazer biomass correlation had the same sign and oﬁen had
approximately the same magnitude (Appendix B22). For instance, species that were
positively correlated with the productivity variables were also positively correlated with
grazer biomass. Figure 2.4 illustrates this trend for the species-TP and species-grazer
biomass correlations.

Several interesting compositional patterns emerge when examining the identity of
the species showing strong correlations with productivity and grazer biomass (Figure 2.4,
Appendix 82.2). First, all of the species strongly negatively correlated with TP and
grazer biomass were either small to medium-sized (i. e. edible) diatoms or cyanophytes,
including several nitrogen-ﬁxers (Epithemia turgida, Calothrix spp., and Nostoc sp. B)
(Figure 2.4). Nitrogen-ﬁxers are especially good nutrient competitors, as they are able to
persist in low nitrogen conditions (Wehr and Sheath 2003). Second, several
chlorophytes, including one grazer resistant ﬁlamentous species (Oedogom'um spp.) were

positively correlated with TP and grazer biomass, but none were negatively correlated
with these environmental variables. Third, half of the species with strong positive
correlations with TP and grazer biomass were plankton species (Scenedesmus bijuga,
F ragilaria crotonensis, Staurastrum sp. A, Scenedesmus quadricauda, and Pediastrum

boryanum), indicating that more plankton species tended to settle into the benthic

29

.65 m .% exam§~oamx~ Em 28 RC 8......” m..o=oEo=V Aom M202; 35:63 0.3 5.3 .m.~ 95me E 36:
805 mm 08% 2: mm 3.8% we mutated 6322: £5 02. .6>o maouﬂoboo 5:5 3.8% 085 35 A33 x83. mmeoE .oNEw
vogoﬁcmbmg EB A33 BE? EC mEonmmosn =38 wogoﬁcﬁumo— com «dd Eat 30:32.80 EoEdEEcoéoBomm Jam Saw?“

 

 

 

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

_,- ed.
d
a
- to. m
s
, O
. u
N6. 0
.5 Eu E F E as we as Us as m
J
. . . . m.
- t o e
h 6 A _. M 0 ca n— : E v— M"
«d u
0
O
a
ed w
$85055~99 mo. I . m.

30

correlations with TP and grazer biomass were plankton species (Scenedesmus bijuga,

F ragilaria crotonensis, Staurastrum sp. A, Scenedesmus quadricauda, and Pediastrum
boryanum), indicating that more plankton species tended to settle into the benthic
community in high productivity lakes. Finally, the species common at high productivity
were on average 7 times larger than those at low productivity. In general, larger species
are inferior competitors compared to small species (Steinman et al. 1992), due to their
higher nutrient requirements.

I hypothesized that changes in predator biomass would affect algal species
composition, subsequently inﬂuencing patterns of limitation across lakes along a
productivity gradient. I have shown that grazer biomass and productivity are positively
correlated in these lake systems and that algal species turnover occurred along this
gradient (Figures 2.1, 2.3, 2.4). In Figure 2.5, I used a univariate measure of species
composition (lake site scores from the RDA) to determine if there was, any association
between species compositional shifts and patterns of nutrient and predator limitation. If a
correlation exists, then species composition may be related to limitation. I chose lake site
scores from the productivity-grazer biomass RDA axis (axis 2), because this was the
primary axis of interest in the context of the KPH. Nutrient limitation was unrelated to
lake site scores (r = 0.03, P = 0.93), whereas predator limitation was signiﬁcantly
negatively correlated to the lake productivity-grazer biomass axis (r = -0.55, P = 0.051),
indicating that algal species composition resulting from gradients in productivity and
grazer biomass affected the extent of these communities’ responses to the removal of

predators.

31

 

 

 

 

 

0.2 r 1 I l
.
”>7
g 0.1_ O . . '—
(T) o
.9: o
C. ____Q_, . i
8
cu 4‘: '
4.: L
g OO:“~:A . ‘ —
j . A ~ JA .
N
{—4‘ A~~‘
1;.
~ ~ ‘ ~
~ ~ “
'01 l l l I .s

-2 -1 0 1 2 3

Lake site scores (2Ind RDA axis)

Figure 2.5: Nutrient limitation (closed circles) and predator limitation (open triangles)
versus the lake site scores from the 2nd RDA axis (productivity-grazer axis). Lines are
linear regressions for nutrient limitation (solid line) and predator limitation (dashed line).

32

DISCUSSION
Patterns of nutrient and predator limitation
The limitation model predicted that in food chains, nutrient and predator
limitation should strongly decrease and increase, respectively, as potential primary
productivity increases (Darcy-Hall, Chapter 1, Appendix A). Alternatively, the limitation
model predicted that for food webs (i.e. KPH context), nutrient limitation should decline
non-linearly, remaining strong at high productivities, while predator limitation should be
low but also a declining function of productivity (Darcy-Hall, Chapter 1, Appendix A).
The results of the benthic algal limitation assays conducted across these lakes more
closely matched predictions from the KPH than those of simple food chain models. First,
nutrient limitation remained stronger than predator limitation across all lakes. Second,
large values of nutrient limitation were observed along the entire gradient in TP and
chlorophyll. Lastly, while neither nutrient nor predator limitation varied signiﬁcantly
with productivity, predator limitation tended to decrease across a gradient in TP, contrary
to what would be expected from the ecosystem exploitation hypothesis (Oksanen et al.
1981).

While nutrient limitation did not signiﬁcantly decline with any of the productivity
variables, one could argue that I lacked power to detect a signiﬁcant decline of nutrient
limitation along a gradient of TN (Figure 2.2). Algal communities in especially low TN
lakes (e. g. Shaw and Bassett) responded more to the addition of nutrients (Whitford Lake
is an exception to this). Hillebrand and Kahlert (2001) similarly noted that lake
periphyton responded less to the addition of nutrients in higher TN systems; however,

they were unable to detect any signiﬁcant relationships between abiotic variables and the

33

strength of top-down or bottom-up effects in their 3 lake systems. Similarly, in a meta-
analysis of 85 experiments manipulating nutrient enrichment and herbivore presence in
the benthic algae, Hillebrand (2002) found no signiﬁcant relationship between
productivity and top-down or bottom-up effect size. A potential decline in nutrient
limitation with TN enrichment is not necessarily supportive of simple food chain model
predictions. In fact, a decline in nutrient limitation with any productivity variable cannot
be attributed to either food chain models or the KPH in the present study, because it is
impossible to determine where these particular lakes fall along the hypothetical nutrient
supply gradient in the limitation model (Darcy—Hall, Chapter 1, Appendix A). However,
as mentioned above, other evidence points to the KPH (or food web) expectations as
being more realistic. For example, one might expect from simple food chain models that
the importance of predator limitation would surpass that of nutrient limitation in more
productive systems (Darcy-Hall, Chapter 1, Appendix A, Figure Al .2). Regardless of the
productivity measure used, nutrient limitation was always as strong as or stronger than
predator limitation and predator limitation did not increase across lakes.

I had hypothesized that predator limitation would decline with grazer biomass and
productivity, because of the presence of more predator resistant species at higher
productivities. The strongly negative predator limitation values in high productivity lakes
indicated that algal community biomass decreased greatly upon the removal of grazers.

A decrease in algal biomass following the removal of grazers was unexpected, but not
unprecedented, as both theory (de Mazencourt et al. 1998) and short- (Sterner 1986,
Leibold unpubl. data) and long-term experiments (McCormick and Stevenson 1989,

McCormick and Stevenson 1991) have observed the stimulation of algal growth by

34

grazers via nutrient regeneration or removal of competitors. Both the KPH (Holt et al.
1994, Grover 1994, Leibold 1996) and models of defense theory (de Mazancourt and
Loreau 2000, de Mazancourt et al. 2001), assume that better defended species tradeoff
predator resistance with the ability to acquire nutrients. Thus, grazer resistant species
may often be nutrient limited. In the context of a benthic algal mat, grazer resistant,
nutrient-limited species can beneﬁt from herbivory, either by utilizing a limiting nutrient
excreted by herbivores as they consume more vulnerable species (de Mazancourt et al.
1998, de Mazancourt and Loreau 2000) or by gaining closer access to water column
nutrients and light via the removal of edible overstory species (McCormick and
Stevenson 1991, McCormick 1994, Steinman 1996). Given that I expected and observed
more grazer resistant species in lakes with higher grazer biomass (see below), it is not
surprising that it was within these lakes that predator limitation was often strong and
negative. Similarly, in a Canadian lake, Graham and Vinebrooke (1998) observed a shift
from a net negative to a net positive effect of grazers on benthic algal biomass, as algal
composition shiﬁed to more grazer resistant (i. e. prostrate, ﬁrmly attached species) forms
throughout the summer.
Benthic algal compositional shifts

The 13 lakes in this study varied widely in conductivity and pH (Appendix 82.1),
which appeared to be important abiotic variables affecting the distribution of algal
species amongst lakes. This is not surprising, as many algal species, especially diatoms,
are sensitive to levels of conductivity or alkalinity (Lowe 1996, Patrick and Reimer
1966). The strong relationship between conductivity and species composition was likely

driven by several lakes in this survey, particularly Lawrence, Tamarack, and Three Lakes

35

2, which are high conductivity, marl lakes with very similar algal compositions (data not
shown). As I had no a priori interest in exploring the relationship between conductivity
or pH and algal composition and as neither variable is important within a KPH context, I
have focused my discussion of algal compositional changes to those related to gradients

in productivity and grazer biomass.

The KPH predicts that in high productivity, high grazer biomass lakes, grazer
resistant forms of benthic algae should dominate; whereas, at low productivities, grazer
vulnerable, superior nutrient competitors should prevail. The patterns of benthic algal
composition across my study lakes lend some support to these hypotheses. Low
productivity lakes were dominated by predator-vulnerable, small to medium-sized
diatoms and cyanophytes. Several of the species at low productivity were undoubtedly
superior nutrient competitors (e. g. the nitrogen ﬁxers), and the generally small size of
these low productivity species may imply stronger competitive abilities when compared
to the larger species observed in high productivity lakes (Steinman et al.1992).

A variety of algal species, over half of which were planktonic, exhibited strong
positive correlations with TP and grazer biomass. Only one of the benthic species, the
tightly attached, ﬁlamentous chlorophyte, Oedogonium spp., displays a grazer resistant
morphology (Steinman et al. 1992, Graham and Vinebrooke 1998). The two other
benthic species, Gomphonema parvulum and G. truncatum, are edible, stalked diatoms
that are likely epiphytic on Oedogonium. The presence of edible species at high
productivities is not unexpected; in fact, edible species must be present in order to
support high populations of benthic grazers. The remaining species associated with high

productivity lakes are planktonic (or unknown in the case of the cyanophytes), several of

36

which (6. g. Fragilaria crotonensis, Pediastrum boryanum ) are commonly observed in
eutrophic systems (Wehr and Sheath 2003). In addition, most of these plankton species
are likely to be resistant to zooplankton grazing: F. crotonensis and P. boryanum on the
basis of size and Staurastrum sp. A and Scenedesmus quadricauda due to the presence of
arms or spines. The persistence of these plankton species in the benthic mat indicates
that they are somehow avoiding benthic predation. I propose two possible mechanisms
of predator avoidance. First, traits conferring grazer resistance in the plankton may also
apply within the benthos. Alternatively, nutrient-rich lakes may produce a constant re-
supply of phytoplankton to the benthos, resulting in phytoplankton immigration rates that
surpass the rate of benthic grazing. Given the descriptive nature of the present study, it is
not possible to distinguish between these two alternative explanations. Regardless of the
mechanism, these phytoplankton species ultimately escaped benthic predation and should
be considered grazer resistant.

I Tradeoffs in competitive ability and herbivore resistance have been documented
before in the periphyton (Graham and Vinebrooke 1998, Rosemond et al. 1993, Steinman
et al. 1992). However, these researchers deﬁne competitiveness as better access to water-
column nutrients and not as the ability to persist when nutrients are scarce (sensu Tilman
1982), the deﬁnition adopted by the KPH. The former view of competition may be more
appropriate for vertically structured benthic communities. Despite the semantic
differences in deﬁning competition, the KPH continues to be an appropriate context for
exploring benthic algal ecology. The fundamental tenet of the KPH is that predators

mediate shiﬁs in prey composition, due to a tradeoff among prey species in their ability to

37

resist predation and compete for resources. If the tradeoff exists, regardless of how the
key species’ traits of competition and resistance are deﬁned, the KPH remains applicable.

The overall pattern of species composition across the lake productivity gradient
provides support for the predictions of the KPH. There were more predator-vulnerable,
superior competitors strongly associated with low productivity lakes, while larger, grazer
resistant species dominated in high productivity systems. However, this study is
correlative in nature and thus I cannot conclusively state that the patterns in species
composition were caused by changes in herbivore biomass. I address this question
experimentally in Chapter 3. In addition, there is not enough known about individual
species’ traits to interpret these patterns as a broad demonstration of the KPH. I provide
a more thorough treatment of algal species’ traits in the context of the KPH in Chapter 4.

Linking species turnover to patterns of limitation

Different algal communities vary in the identity and the traits of the dominant
species. Variation in dominant algal species and species’ traits across communities
should lead to differences in the community-wide response to the removal of limiting
factors (i.e. limitation). As I have described above, species composition varied greatly
across the study lakes, and many species demonstrated strong associations with certain
environmental factors (e. g. TP or grazer biomass), indicating that species’ traits might
also vary across lakes. Given this variation in species’ traits, the composition of an algal
community in a particular lake should affect how that community responds to the
addition of nutrients or the removal of grazers. Among these lakes, algal species
composition along the productivity-grazer biomass gradient was clearly related to

predator limitation, but not to nutrient limitation.

38

One possible scenario to explain the predator limitation-species composition
correlation is that the presence of more grazer resistant species in lakes with high grazer
biomass translated into a less predator-limited community that in turn responded
negatively to the removal of grazers. This counterintuitive response was perhaps due to
the tendency of grazer resistant species to be nutrient limited, because of their size,
position in the algal mat, or the costs associated with defenses such as mucilaginous
exteriors (de Mazancourt and Loreau 2000). Nutrient limitation, on the other hand, did
not appear to be related to species composition along the productivity-grazer biomass
axis. This is not surprising as nutrient limitation was not signiﬁcantly related to any of
the productivity variables, and nutrient limitation did not signiﬁcantly correlate with lake
site scores from any of the other three axes (data not shown). It is possible that nutrient
limitation may indeed be related to species composition, but along some unmeasured
gradient. Alternatively, the strength of nutrient limitation could be more affected by
abiotic variables and less by species composition and individual species’ traits. Finally, it
is also possible that with only 13 lakes, I lacked the power necessary to observe a
relationship between nutrient limitation and species composition.

While the analysis of benthic algal composition provided support for predictions
of the KPH, it is obvious that it cannot fully account for the variability in algal species
composition across these lakes. The fundamental keystone predator model parameters,
herbivore biomass and potential primary productivity (and pH) accounted for only 27.9%
of the variation in algal species composition. Conductivity alone accounted for nearly as
much (25%). The low variation explained by the variables of interest is hardly

surprising, given that the survey of algae is conducted in natural systems, which are

39

inevitably highly variable in all components of the ecosystem. Therefore, it is crucial to
recognize that while the KPH tradeoff may be an important inﬂuence affecting limitation
patterns and species composition, it is only one of several simultaneously occurring
mechanisms. In Chapter 3, I use a more controlled, experimental approach to examine
the linkage between herbivore biomass, productivity, and patterns of algal limitation.
Conclusions

Inter-ecosystem patterns of nutrient and predator limitation of benthic algae in
this study mirrored predictions of both verbal (Osenberg and Mittelbach 1996) and
mechanistic keystone predator models (Holt et al. 1994, Grover 1994, Leibold 1996).
Nutrient limitation was stronger than predator limitation and remained so across a lake
productivity gradient. In these lakes, predator limitation did not increase across the
productivity and grazer biomass gradient, contradicting predictions of simple food chain
models (Darcy-Hall, Chapter 1, Appendix A, Oksanen et al. 1981). Additionally,
changes in benthic algal species composition across lakes and the traits of common
species qualitatively complied with the keystone predation hypothesis. This research
illustrates the importance of heterogeneity both within and among systems to the
elucidation of interactions between a community of organisms and the surrounding

abiotic and biotic environments.

40

LITERATURE CITED

APHA. 1980. Standard methods for the examination of water and wastewater, 151h
Edition. APHA, Washington, DC.

Bachmann, R.W., and DE. Canﬁeld, Jr. 1996. Use of an alternative method for
monitoring total nitrogen concentrations in Florida lakes. Hydrobiologia 32321-8.

Bjojk-Ramberg, S., and Anell, C. 1985. Production and chlorophyll concentration of
epipelic and epilithic algae in fertilized and nonfertilized subarctic lakes.
Hydrobiologia 126:213-219.

Borchardt, M. A. 1996. Nutrients. In Algal Ecology: Freshwater Benthic Ecosystems
(R. J. Stevenson, M. L. Bothwell, and R. L. Lowe, eds.), pp. 183-227. Academic
Press, San Diego, California, USA.

Brett, M. T. and C. R. Goldman. 1997. Consumer versus resource control in freshwater
pelagic food webs. Science 275:384-386.

Carpenter, S. R. and J. F. Kitchell. 1988. Consumer control of lake productivity.
Bioscience 38:764-769.

Carpenter, S. R. and J. F. Kitchell. 1993. The T rophic Cascade in Lakes. Cambridge,
UK: Cambridge University Press. 385 pp.

Chase, J. M., M. A. Leibold, and E. Sims. 2000. Plant tolerance and resistance in food
webs: community-level predictions and evolutionary implications. Evolutionary
Ecology 14:289-314.

Crumpton, W.G., T.M. Isenhart, and PD. Mitchell. 1992. Nitrate and organic N
analyses with second-derivative spectroscopy. Limnology and Oceanography
37:907-913.

Dillon Jr., R. T. 2000. The Ecology of Freshwater Molluscs. Cambridge, UK:
Cambridge University Press. 509 pp.

Dixon, P. M. 2001. The bootstrap and the jackknife. In Design and Analysis of

Ecological Experiments (S. M. Scheiner and J. Gurevitch, eds.) pp. 267-288.
Oxford University Press. New York, New York, USA.

Downing, J. A., C. W. Osenberg, and O. Sarnelle. 1999. Meta-analysis of marine
nutrient-enrichment experiments: variation in the magnitude of nutrient limitation.

Ecology 80:1157-1167.

F airchild, G. W., R. L. Lowe, and W. B. Richardson. 1985. Algal periphyton growth on

41

nutrient-diffusing substrates: an in situ bioassay. Ecology 66:465-472.

Feminella, J. W., and C. P. Hawkins. 1995. Interactions between stream herbivores and
periphyton: a quantitative analysis of past experiments. Journal of the North
American Benthological Society 14:465-509.

F retvvell, S. D. 1977. The regulation of plant communities by food chains exploiting
them. Perspectives in Biology and Medicine 20: 169-1 85.

Ginzburg, L. R., and H. R. Akcakaya. 1992. Consequences of ratio-dependent predation
for steady-state properties of ecosystems. Ecology 73:1536-1543.

Graham, M. D. and R. D. Vinebrooke. 1998. Trade-offs between herbivore resistance
and competitiveness in periphyton of acidiﬁed lakes. Canadian Journal of
Fisheries and Aquatic Sciences 552806-814.

Grover, J. P. 1994. Assembly rules for communities of nutrient-limited plants and
specialist herbivores. American Naturalist 143:158-282.

Hairston, N. G., F. E. Smith, and L. B. Slobodkin. 1960. Community structure,
population control, and competition. American Naturalist 44:421-425.

Hill, W. R., H. L. Boston, and A. D. Steinman. 1992. Grazers and nutrients
simultaneously limit lotic primary productivity. Canadian Journal of Fisheries
and Aquatic Sciences 49:504-512.

Hillebrand, H. 2002. Top-down versus bottom-up control of autotrophic biomass — a
meta-analysis on experiments with periphyton. Journal of the North American
Benthological Society 21:349-369.

Hillebrand, H., C-D. Dﬁrselen, D. Kirschtel, U. Pollingher, and T. Zohary. 1999.
Biovolume calculation for pelagic and benthic microalgae. Journal of Phycology
35:403-424.

Hillebrand, H. and M. Kahlert. 2001. Effect of grazing and nutrient supply on
periphyton biomass and nutrient stoichiometry in habitats of different
productivity. Limnology and Oceanography 46: 1881-1898.

Holt, R. D., J. P. Grover, and D. Tilman. 1994. Simple rules for interspeciﬁc dominance
in systems with exploitative and apparent competition. American Naturalist
144:741-771.

Hunter, M. D. and P. W. Price. 1992. Playing chutes and ladders: heterogeneity and the

relative roles of top-down and bottom-up forces in natural communities. Ecology
73:724-732.

42

Legendre, P. and E. D. Gallagher. 2001. Ecologically meaningful transformations for
ordination of species data. 0ecologia 129:271-280.

Legendre, P. and L. Legendre. 1998. Numerical Ecology, Developments in
Environmental Modelling 20, Elsevier. Amsterdam, The Netherlands.

Leibold, M. A. 1999. Biodiversity and nutrient enrichment in pond plankton
communities. Evolutionary Ecology Research 1:73-95.

Leibold, M. A., J. M. Chase, J. B. Shurin, and A. L. Downing. 1997. Species turnover
and the regulation of trophic structure. Annual Review of Ecology and
Systematics 28:467-494.

Leibold, M. A. 1996. A graphical model of keystone predators in food webs: trophic
regulation of abundance, incidence, and diversity patterns in communities.
American Naturalist 147:784-812.

Lowe, R. L. 1996. Periphyton patterns in lakes. In Algal Ecology: Freshwater Benthic
Ecosystems (R. J. Stevenson, M. L. Bothwell, and R. L. Lowe, eds.), pp. 57-76.
Academic Press, San Diego, California, USA.

Lowe, R. L. and Y. Pan. 1996. Benthic algal communities as biological monitors. In
Algal Ecology: Freshwater Benthic Ecosystems (R. J. Stevenson, M. L. Bothwell,
and R. L. Lowe, eds.), pp. 57-76. Academic Press, San Diego, California, USA.

Marks, J. C., and R. L. Lowe. 1989. The independent and interactive effects of snail
grazing and nutrient enrichment on structuring periphyton communities.
Hydrobiologia 185:9-17.

de Mazancourt, C. and M. Loreau. 2000. Grazing optimization, nutrient cycling, and
spatial heterogeneity of plant-herbivore interactions: should a palatable plant
evolve? Evolution 54:81-92.

de Mazancourt, C., M. Loreau, and L. Abbadie. 1998. Grazing optimization and nutrient
cycling: when do herbivores enhance plant production? Ecology 79:2242-2252.

de Mazancourt, C., M. Loreau, and U. Dieckmann. 2001. Can the evolution of plant
defense lead to plant-herbivore mutualism? American Naturalist 158: 109-123.

McCormick, P. V. 1996. Resource competition and species coexistence in freshwater
benthic algal assemblages. In Algal Ecology: Freshwater Benthic Ecosystems (R.
J. Stevenson, M. L. Bothwell, and R. L. Lowe, eds.), pp. 229-252. Academic
Press, San Diego, California, USA.

McCormick, P. V. 1994. Evaluating the multiple mechanisms herbivore-a1 gal
interactions in streams. Hydrobiologia 291247-59.

43

McCormick, P. V., and R. J. Stevenson. 1989. Effects of snail grazing on benthic algal
community structure in different nutrient environments. Journal of the North
American Benthological Society 8: 162-172.

McCormick, P. V., and R. J. Stevenson. 1991. Grazer control of nutrient availability in
the periphyton. 0ecologia 86:287-291.

McCormick, P. V. and R. J. Stevenson. 1989. Effects of snail grazing on benthic algal
community structure in different nutrient environments. Journal of the North
American Benthological Society 8: 162-172.

McNaughton, S. J ., M. Oesterheld, D. A. Frank, and K. J. Williams. 1989. Ecosystem-
level patterns of primary productivity and herbivory in terrestrial habitats. Nature
341:142-144.

McQueen, D. J ., J. R. Post, and E. L. Mills. 1986. Trophic relationships in freshwater
pelagic ecosystems. Canadian Journal of Fisheries and Aquatic Sciences
43:1571-1581.

Mills, E. L., and A. Schiavone Jr. 1982. Evaluation of ﬁsh communities through
assessment of zooplankton populations and measures of lake productivity. North
American Journal of Fisheries Management 2: 14-27.

Mittelbach, G. G., C. W. Osenberg, and M. A. Leibold. 1988. Trophic relations and
. ontogenetic niche shifts in aquatic ecosystems. In Size-Structured Populations
(B. Ebenman and L. Persson, eds.), pp. 219-235. Berlin: Springer-Verlag.

Mulholland, P. J ., A. D. Steinman, A. V. Palumbo, J. W. Elwood, and D. B. Kirschtel.
Role of nutrient cycling and herbivory in regulating periphyton communities in
laboratory streams. Ecology 72:966-982.

Oksanen, L., S. D. Fretwell, J. Arruda, and P. Niemela. 1981. Exploitation ecosystems
in gradients of primary productivity. American Naturalist 118:240-261.

Osenberg, C. W. and G. G. Mittelbach. 1996. The relative importance of resource
limitation and predator limitation in food chains. In Food webs: Integration of
patterns and dynamics. (G. A. Polis and K. O. Winemiller, eds.), pp. 134-148.
Chapman and Hall, New York, New York USA.

Pan, Y. and R. L. Lowe. 1994. Independent and interactive effects of nutrients and
grazers on benthic algal community structure. Hydrobiologia 291:201-209.

Patrick, R. 1970. Benthic stream communities. American Scientist 58:546-549.

Patrick, R. and C. W. Reimer. 1966. The Diatoms of the United States Exclusive of

44

Alaska and Hawaii, Monogr. 13, Vol. 1. Academy of Natural Science,
Philadelphia, Pennsylvania, USA.

Planas, D. 1996. Acidiﬁcation effects. In Algal Ecology: Freshwater Benthic
Ecosystems (R. J. Stevenson, M. L. Bothwell, and R. L. Lowe, eds), pp. 497-530.
Academic Press, San Diego, California, USA.

Power, M. E. 1992. Top-down and bottom-up forces in food webs: do plants have
primacy? Ecology 73:733-746.

Prepas, BE, and RH. Rigler. 1982. Improvements in quantifying the phosphorus
concentration in lake water. Canadian Journal of Fisheries and Aquatic Sciences
39:822-829.

Reynolds, C. S. 1984. The ecology of freshwater phytoplankton. Cambridge University
Press. Cambridge, UK.

Riber, H. H. and R. G. Wetzel. 1987. Boundary-layer and internal diffusion effects on
phosphorus ﬂuxes in lake periphyton. Limnology and Oceanography 3211181-
1194.

Rosemond, A. D. 1996. Indirect effects of herbivores modify predicted effects of
resources and consumption of plant biomass. In Food webs: Integration of
patterns and dynamics. (G. A. Polis and K. O. Winemiller, eds.), pp. 149-159.
Chapman and Hall, New York, NewYork USA.

Rosemond, A. D., P. J. Mulholland, and J. W. Elwood. 1993. Top-down and bottom-up
control of stream periphyton: effects of nutrients and herbivores. Ecology
74:1264-1280.

Steinman, A. D. 1996. Effects of grazers on freshwater benthic algae. In Algal Ecology:
Freshwater Benthic Ecosystems (R. J. Stevenson, M. L. Bothwell, and R. L.
Lowe, eds.), pp. 341-373. Academic Press, San Diego, California, USA.

Steinman, A. D., P. J. Mulholland, and W. R. Hill. 1992. Functional responses
associated with growth form in stream algae. Journal of the North American

Benthological Society 11:229-243.

Sterner, R. W. 1986. Herbivores’ direct and indirect effects on algal populations.
Science 231 1605-607.

Tilman, D. 1982. Resource Competition and Community Structure. Princeton
University Press, Princeton, New Jersey, USA.

Wehr, J. D. and R. G. Sheath. 2003. Freshwater Algae of North America. Academic
Press, San Diego, California, USA.

45

Welschmeyer, N. A. 1994. F luorometric analysis of chlorophyll a in the presence of
chlorophyll b and phaeopigrnents. Limnology and Oceanography 39: 1 985-1992.

Wetzel, R. G. 1996. Benthic algae and nutrient cycling in lentic freshwater ecosystems.
In Algal Ecology: Freshwater Benthic Ecosystems (R. J. Stevenson, M. L.
Bothwell, and R. L. Lowe, eds.), pp. 641-667. Academic Press, San Diego,
California, USA.

Wootton, J. T., M. E. Power, R. T. Paine, and C. A. Pﬁster. 1996. Effects of
productivity, consumers, competitors, and El Nifro events on food chain patterns
in a rocky intertidal community. Proceedings of the National Academy of

Sciences 93: 13855-13858.

46

CHAPTER 3

LINKING HERBIVORE-MEDIATED SPECIES TURNOVER TO PATTERNS OF
BENTHIC ALGAL NUTRIENT AND PREDATOR LIMITATION

INTRODUCTION

Predators can play an important role in structuring communities by directly
affecting the species composition of prey communities. When predation pressure is
strong, we expect prey communities to be composed of more predator resistant, well-
defended species. In aquatic systems, there are numerous examples of herbivorous
predators causing shifts to more predator resistant algal prey communities. For instance,
herbivore-mediated effects have been observed in freshwater periphyton (or benthic
algae) (e.g. Rosemond et al. 1993, Lamberti et al. 1995), phytoplankton (e. g. Leibold
1999, Steiner 2001) and marine macroalgae (e.g. Lubchenco and Cubit 1980). However,
there are costs to defense, and it has been hypothesized that organisms, particularly
primary producers, will exhibit tradeoffs in their ability to compete for resources and
resistance to predation (Paine 1966, Lubchenco and Gaines 1981, Vance 1978, Holt et al.
1994, Grover 1994, Leibold 1996, Chase et a1. 2000). For instance, benthic algal species
with traits that confer superior competitive abilities (e. g. small size or erect forms) are
vulnerable to herbivores. Alternatively, defended species tend to be inferior competitors,
due in part to costs of defense as well as their inherent morphology (e. g. large or adnate)
(Steinman et al. 1992, Rosemond 1996). Despite an abundance of anecdotal
observations, direct evidence for this tradeoff is limited.

If prey species exhibit a tradeoff in their ability to acquire resources and their
resistance to predation, then theory predicts that prey community composition should

shift toward dominance by predator resistant species with changes in both predator

47

density and resource supply (Leibold 1996, Holt et al. 1994, Grover 1994). The ﬁrst
prediction is straightforward: higher intensities of predation favor more predator resistant
prey species. The second prediction involves an upward trophic cascade: higher
ecosystem resources result in higher abundances of herbivores (Leibold 1999, Ginzburg
and Akcakaya 1992, McNaughton et a1. 1989, Mills and Schiavone 1982), that in turn
exert more selection for resistance in their primary producer prey. This phenomenon has
been coined the keystone predation hypothesis (KPH; Leibold 1996, Holt et al. 1994,
Grover 1994) and it focuses on the role of predators in shifting prey community structure,
contributing to the positive correlations between the biomass of adjacent trophic levels.
The KPH predicts that vulnerable, superior competitors will dominate at low
productivities, while more predator resistant, less resource efﬁcient taxa will dominate at
high productivities (Leibold 1996, Holt et al. 1994, Grover 1994). As a result, resource
limitation (i. e. the extent to which prey increase upon augmentation of resources [sensu
Osenberg and Mittelbach 1996]) of prey is expected to remain strong across a gradient in
resource supply (Darcy-Hall, Chapter 1, Appendix A). For instance, competitively
inferior but defended species will be less limited by predators, yet still limited by
nutrients. Therefore, resource limitation of prey will remain strong along a productivity
gradient: at low productivities, resources are scarce but taxa are good resource
competitors, while at high productivities, resources are plentiful, but dominant taxa are
less efﬁcient at consuming those resources. Alternatively, predator limitation (the extent
to which producer biomass increases after the removal of predators) is expected to be low
and potentially decrease across a productivity gradient, as more predator resistant species

dominant at higher productivities (Darcy-Hall, Chapter 1, Appendix A). These general

48

predictions of the KPH are in sharp contrast with predictions from simple (i.e. one
species/trophic level) food chain models (Hairston et al. 1960, Oksanen et al. 1981). For
instance, these models posit that in a 2-trophic level system, herbivores will increase with
productivity, but primary producers will not. Consequently, the producer species is
expected to become less nutrient limited but more predator limited as potential primary
productivity and the biomass of herbivores increases (Darcy-Hall, Chapter 1, Appendix
A). Field studies of primary producers have observed consistent and strong nutrient
limitation across a productivity gradient of aquatic systems (Darcy-Hall, Chapter 1,
Downing et al. 1999, Osenberg and Mittelbach 1996). However, there has been no direct
effort to link herbivore-mediated shifts in prey composition to patterns of resource and
predator limitation.

One approach to linking herbivore-mediated species turnover to patterns of
limitation is to compare such patterns in communities that have developed with and
without herbivore pressure. I hypothesize that benthic algal communities historically
exposed to herbivores (“grazed” communities) will exhibit herbivore-mediated shifts in
composition, and thus patterns of limitation will be similar to those predicted by the
KPH. In contrast, I expect that the composition of historically herbivore-naive
(“ungrazed”) algal communities will shift solely on the basis of differences in nutrient
supply rate, and due to the absence of any grazing pressure, predator resistant species will
not dominate at higher productivities. Thus, I predict that at high productivities,
ungrazed communities will exhibit stronger predator limitation and less nutrient
limitation than historically grazed communities, a pattern resembling predictions from

simple food chains. Patterns at low productivity are less predictable, as the limitation

49

model indicates that limitation for both food chains and food webs may be similar at low
enrichment (i.e. high resource limitation and low predator limitation) (Darcy-Hall,
Chapter 1, Appendix A). I do not expect limitation in ungrazed communities to perfectly
mimic the expectations of simple food chain models, as these communities will actually
be food webs and will likely exhibit more complex patterns than those predicted for food
chains in the limitation model (Darcy-Hall, Chapter 1, Appendix A). Despite this, I
expect that limitation in ungrazed communities will be measurably different than in
grazed communities and in the directions predicted. Hence, this experiment aims to
directly examine the herbivore-mediated changes in algal composition and how these
shifts might drive patterns of algal limitation.
METHODS
Limitation metric

Limitation is deﬁned as the extent to which a population’s perqcapita growth rate
is reduced by predators or insufﬁcient nutrient supply and is measured as the increase in
the population’s per capita growth rate upon the removal of a limiting factor (e. g. via the
addition of saturating nutrients or the removal of predators) (Osenberg and Mittelbach
1996). Limitation is measured in the short-terrn to avoid confounding its direct effects
with the indirect effects caused by a numerical (or functional) response of predators or a
switch to limitation by another resource (Osenberg and Mittelbach 1996, Downing et al.
1999). Downing et al. (1999) found that in an experimental setting, limitation was a
constant and reliable measure of the change in algal population per capita growth over a

period of 2-7 days, post treatment initiation.

50

Study system

The experiment was conducted in 1000-L cattle tank mesocosms during the
summer of 2003 at Michigan State University’s W. K. Kellogg Biological Station (KBS)
experimental pond lab facility (Hickory Comers, MI, USA). Cattle tanks are known to
effectively represent pond environments (Leibold et al. 1997) and previous experiments
have demonstrated that a large subset of lake benthic algae persist in the mesocosm
environment (Darcy-Hall, Chapter 4). For example, a 2001 lake survey in southwest
Michigan, I identiﬁed 214 benthic algal species, 120 of which were observed in a cattle
tank experiment the following year (Darcy-Hall, Chapters 2, 4). The mesocosm
experiment was conducted in two phases: an initial, 3-month community establishment
period (Phase 1) followed by 1-week limitation assays (Phase 2).

Phase 1 .' Community establishment

Initial grazed and ungrazed algal communities were established in early May
2003 within a simple experimental context. Four levels of nutrient supply were crossed
with the presence or absence of a benthic grazer community, yielding 8 treatments that
were replicated 4 times. Cattle tank mesocosms were acid-washed, ﬁlled with well
water, and covered with a l-mm mesh lid, composed of ﬁberglass window screening.
Fourteen 3” (7.6 cm) terracotta clay ﬂowerpots in each mesocosm served as sampling and
experimental substrates for benthic algal communities. Nitrogen (as ammonium nitrate)
and phosphorus (as potassium phosphate) were added to the water column to establish
four levels of productivity. Target nutrient supply concentrations were 10 (LOW), 30
(MED), 75 (HIGH), and 200 (XHIGH) rig/L total phosphorus (TP), with nitrogen added

in a 50:1 molar ratio. These productivity levels encompassed natural variation in TP and

51

TN of both lakes and ponds local to KBS (Hall et al. in press, Darcy-Hall, unpubl. data).
Nutrients were added every 5 days throughout the experiment to maintain the
approximate target nutrient concentrations. Benthic grazers included three species of
gastropods (Physa sp., Helisoma sp., and Gyraulus sp.) as well as amphipods (Hyallela
azteca and Gammarus sp.). In nature, consumer densities increase concomitantly with
increasing producer density (reviewed Leibold et al. 1997), so I initially established an
increasing density of grazers among the four levels of nutrient supply. Grazer density
data from lake surveys (Darcy-Hall, Chapter 2) were used to determine the densities at
which to add grazers to mesocosms. Grazers were added at 0.09, 0.24, 0.63, 1.64 g dry
mass/m2, densities approximate to those expected at increasing levels of nutrient supply.

Arrays of clay pots were established in the littoral zone of seven lakes to grow
communities of algae for seeding the mesocosm experiment. The lakes were chosen to
span a wide gradient of TP (13.5 — 77.5 rig/L) and TN (208.9 - 1869 ug/L). To inoculate
mesocosms, benthic algae were scraped from several clay pots in all lakes and mixed
with deionized water. Resulting algal slurries were transported to the lab, sieved through
a 125 um sieve to remove micro- and macrograzers, combined and mixed in a carboy,
and then distributed in 100 mL aliquots to each cattle tank. Algal species composition
among lakes was expected to vary through time, so this inoculation process was repeated
once every 2 weeks for the duration of Phase 1 to allow continued colonization
opportunities for a diverse algal assemblage.

Benthic algal species were allowed to establish in these experimental
environments for 15 weeks (approximately 50 algal generations) prior to the initiation of

Phase 2. One week prior to the limitation assays, the algal communities from two clay

52

pots per cattle tank were combined and subsampled to measure pre-Phase 2 chlorophyll
a, grazer densities, and algal species composition. Algae were removed from clay pots as
described above. Grazers were preserved in 70% ethanol and algae for species
identiﬁcation were preserved in 10% formalin. All chlorophyll a measurements were
conducted on the same day as sampling, using cold ethanol extractions followed by
narrow-band ﬂuorometry (sensu Welshmeyer 1994). In addition, water from the
mesocosms was sampled to measure TP and TN content. Total nutrients were measured
using spectrophotometry and standard methods (Bachmann and Canﬁeld 1996, Crumpton
et al. 1992, Prepas and Rigler 1982, APHA 1980)

After sampling, the remaining 12 clay pots from each ungrazed cattle tank were
shifted into a grazed cattle tank (Figure 3.1) within the same nutrient supply treatment.
For example, clay pots from a no-grazer, LOW mesocosm were shifted into a randomly
chosen plus-grazer, LOW mescosm. Ultimately, l6 mesocosms containing both grazed
and ungrazed sets of algal communities were used in the limitation assays. Resident clay
pots in the grazed mesocosms were also lifted out of the water and moved to a new
location within the same tank. Predator limitation by deﬁnition measures the effect of the
removal of grazers on a communities’ per capita growth rate. Therefore, the ungrazed
algal communities required a period of grazing prior to the onset of the limitation assays.
From previous experiments, one week was determined to be long enough for grazers to
have an effect on biomass but not on community composition (Darcy-Hall, unpubl. data).

Phase 2: Limitation assays
Seven-day limitation assays were initiated on August 25, 2003. Due to the scale

of these experiments, I was only able to set up the assays in 8 cattle tanks per day;

53

.N 825 we 5% 05 «a ﬂea >20 «0 288308 05 E8052 958.2 .83533 mam—95m 89032.“ tam—o 8885“ 8323»
86:0 come—o can :25 .Amhmmmm count—E: N 893 can Ccoaamznﬁmo 3558803 ﬂ omens. com amine 3.885me ”mm 2&5

ex ex ex
name. :3 ooeomo SN
.3”. g...
o
@ ms ooooomo ms
ﬁr a...
0
camp. on oonwowo cm
“w... gm...
#0 A: oonwow cw
y 0$000 1}
o _w>o4
9.52.335 m3... gonna min. 33.6 02 3330360..."—

N 3.2: ~ 8.23

54

therefore, half of the assays began on August 26 while the remaining half began on
August 28. Within each cattle tank, 2 limitation assays took place: one within the
ungrazed set of clay pots and the other within the grazed set of clay pots. Each assay
included four, randomly assigned treatments, with 3 replicates each: nutrient additions
alone, grazer removals alone, nutrient additions and grazer removals, and controls.
Control treatments were those that resembled natural conditions, that is, no nutrients
added and grazers present. The nutrient addition treatment involved removing 25 ml
of water from the interior of the clay pot and either returning it (to controls and grazer
removal treatments) or replacing it with 25 mL of a saturated nutrient solution (NH4N03,
and KH2P04 in a 2:1 molar N:P ratio). The nutrient solution was approximately 1500
times more concentrated at a molar level than the highest nutrient supply rate in the
experiment (200 ugL TP). Nitrogen was added at such a low N:P ratio to avoid adding
toxic levels of nitrogen to the algal and grazer communities. The porous clay surface of
the ﬂowerpots readily diffuses nutrients over time (Darcy-Hall unpubl. data, Fairchild et
al. 1985, Marks and Lowe 1989, Wootton et al. 1996). Grazers were manually removed
from all clay pots and returned to all but the grazer removal treatment. Care was taken to
evenly redistribute the grazers among the nutrient additions and control treatments. Once
treatments had been applied, all clay pots were placed within an overturned 2-pint (~ l-L)
clear, plastic deli container with 2, 105-um mesh windows which retained herbivores but
allowed water movement.

Each limitation assay lasted for one week, after which benthic algal communities
from each pot were harvested as described above and subsampled for chlorophyll a

analysis and algal species identiﬁcation. Any grazers present on clay pots were also

55

harvested and preserved in 70% ethanol for biomass measurements. Chlorophyll a
analyses were conducted on the same day that the algal communities were harvested. The
chlorophyll a data were used to calculate values of nutrient and predator limitation.
Limitation values were calculated using the following equations:

LimN = ln(NA)-—ln(C) (1a)

1‘
Limp = ln(GR)t- 1n(c) (1b)

 

 

where t is duration of the experiment in days, LimN is nutrient limitation and Limp is
predator limitation, and NA, GR, and C are the chlorophyll a measurements of the
nutrient addition, grazer removal, and control treatments, respectively. Values of
limitation are measures of the change in instantaneous per-capita grth rate of algal
communities after the removal of a limiting factor. In general positive values of
limitation are expected, as algal communities should increase in biomass with the
remoVal of the limiting factor. ’

At the start of the limitation assays, diel dissolved oxygen patterns (a surrogate for
ecosystem productivity) were measured in the mesocosms to obtain a direct measure of
overall community metabolism. Dissolved oxygen (DO) measurements were measured
with a YSI® oxygen probe and were taken at dusk, dawn, and dusk, during a 24-hour
period at the start of the limitation assays. Net ecosystem productivity was calculated as
the difference in D0 between dawn and dusk (in units of mg/L/hour), and it was these
values that were used as a system productivity measure in regressions of limitation.

Algal species composition at the end of the limitation assays was also assessed in

communities in the LOW and XHIGH productivity mesocosms. Composition was

56

enumerated for the control, nutrient addition, and predator removal limitation assay
treatments, but only data for the control treatment are presented.
Algal enumeration methods .

Benthic algal communities were enumerated and identiﬁed using established
methods (Lowe and Pan 1996). Semi-permanent wet mounts of a sample were prepared
and examined at 1250x magniﬁcation on a compound microscope. Three hundred
natural units (e. g. a cell, ﬁlament or colony) were enumerated and identiﬁed to the
species level (when possible) for each sample. In most cases, a second slide was
prepared with acid-cleaned diatoms mounted in Naphrax (Patrick and Reimer 1966) to
identify diatoms to the species level. A subsample of 50 organisms per species was
measured for biovolume calculations, using published geometric forms and volume
equations for various algal genera (Hillebrand et al. 1999). All algal community
composition results presented here are in units of relative biovolume, which were arc-sine
square-root transformed prior to analyses.

Statistical analyses

The relationship between nutrient and predator limitation of grazed and ungrazed
algal communities and net primary productivity (DO) were explored qualitatively with
LOWESS regression (Trexler and Travis 1993; Systat, v.8, SPSS Inc.1998). In addition,
limitation values in historically grazed versus ungrazed communities were compared
quantitatively using a 2-way AN OVA, treating nutrient supply and grazing history
(ungrazed or grazed) as factors. Each type of limitation was ﬁrst tested together in a

MANOVA, and then separately as the global hypothesis of no difference was rejected in

57

the MANOVA. Slices of signiﬁcant interaction terms were conducted in SAS v.8 (SAS
Institute).

Shifts in algal community composition due to the Phase 1 treatments were
analyzed using several methods. First, I ran a permutational MANOVA (perMAN OVA;
McCardle and Anderson 2001, Anderson 2001, Anderson 2004) to test whether the
nutrient supply and grazer treatments signiﬁcantly affected algal community
composition. The perMAN OVA uses permutation tests on a distance matrix of raw data
to provide F-values and test statistics for each treatment factor, their interaction,
uncorrected post-hoc treatment comparisons, and slices of interaction terms (Anderson
2001). It relaxes the assumption of multivariate normality, which is usually an unrealistic
assumption with community data, due to the presence of many zeroes in the data set. The
perMAN OVA also allows the user to choose a distance measure other than the Euclidean
distance, which is an assumption of traditional MAN OVA. Here, I use the Bray-Curtis
distance matrix, as it is known to perform especially well with species abundance data
(Anderson and Willis 2003, McCune and Grace 2002, Legendre and Legendre 1998).

In addition, I used a canonical analysis of principal coordinates (CAP) analysis
(Anderson and Robinson 2003, Anderson and Willis 2003, Anderson 2002) to provide a
visual ordination of the community composition data and to explore which individual
species were most affected by the treatments. Unlike many other ordination methods, the
CAP analysis takes into account the correlation structure between variables (i.e. species)
(Anderson and Willis 2003). The CAP analysis also deals well with situations where the
number of species is greater than the number of samples or experimental units, which is

the case here (Anderson and Willis 2003). The CAP analysis ﬁrst performs an

58

unconstrained principal coordinates analysis (PCoA) on a Bray-Curtis distance matrix of
the transformed algal biovolume data and then uses the resulting principal coordinates
(PCO axes) in a constrained ordination. The type of constrained ordination performed
depends upon the hypothesis of interest. Here, a 2-way crossed design provided the
structure for the hypotheses concerning algal shifts in response to experimental.
treatments. Therefore, a matrix of dummy variables coding for the experimental
ANOVA design constrained the ordination of algal data in a canonical correlation
analysis (CCorA). The CAP analysis produces canonical axes that are representative of
the ANOVA treatments, the strength of which is measured as a squared canonical
correlation coefﬁcient (82) that is comparable to an R2 value. It also generates canonical
scores of each mesocosm for use in a biplot of the relative locations of each mesocosm’s
algal community along the ordination axes as well as individual species’ correlations with
the canonical axes.
RESULTS
Phase 1 — Algal community composition
The applied nutrient supply treatments created a gradient in overall productivity
(Figure 3.2), with higher nutrient supply resulting in increased TP (Figure 3.2a) and TN
(Figure 3.2b) in the water column, higher chlorophyll a on clay pots (Figure 3.2c) and
higher net ecosystem productivity measured by DO (Figure 3.2d). Differences in TP and
chlorophyll across nutrient supply treatments depended upon the presence or absence of
grazers (interaction terms: FTp;3,24=6.7, P=0.002; ch;3,24=3.3, P=0.03 8), while differences
in TN and DO were affected only by nutrient supply (nutrient supply: F'm;3,24=92.7,

P<0.0001, Fm;3,24=44.48, P<0.0001).

59

150— a 7 — ' b

 

 

 

 

 

   

 

 

 

 

U, U,

3 E

E 50~ E
O 1 1 1 1 1 0 1 1 1 1 1
0 50 100 150 200 250 0 50 100 150 200 250
37 1: 077

3? c E d

E 45, 0.6-

‘m s

3; c 0.5

W 0

—. m

'5 Q 0.4

‘ o

g- '0 03

E E '

.C O

0.2

0 £3.
O 1 1 1 1 1 D 0.1 I 1 1 1 J
0 50 100 150 200 250 0 50 100 150 200 250

Nutrient supply treatment (pg/L TP)

Figure 3.2: Means :1: 1 SE. of a) Total phosphorus, b) total nitrogen, c) chlorophyll a of
periphyton on terracota clay ﬂowerpots and (1) dissolved oxygen in cattle tank
mesocosms after 3 months of applied nutrient supply treatments. Closed circles are
systems containing herbivores and open circles are systems lacking herbivores.

60

Benthic algal species composition clearly differed across the productivity and
grazer treatments. A biplot of the CAP analysis’ canonical scores (Figure 3.3) shows
clear separation between both nutrient supply and grazer treatments. The ﬁrst and second
CAP axes represent variation due to the nutrient supply treatment (52. = 0.82) and the
grazer treatment (522 = 0.46), respectively. Thus, lower nutrient supply mesocosms plot
negatively along the 1St CAP axis, while mesocosms from higher nutrient supply plot
positively (Figure 3.3). In addition, mesocosms containing herbivores tended to be
negative along the 2nd CAP axis, while ungrazed communities are positive (Figure 3.3).

A perMAN OVA of algal species composition revealed a signiﬁcant interaction
between the nutrient supply and grazer presence/absence treatments (F 3,24 = 1.76, P =
0.027). Slices of this interaction showed a signiﬁcant difference between grazed and
ungrazed communities in the LOW and XHIGH treatments (Appendix C3.1). In
addition, among the ungrazed communities, composition in the XHIGH treatment was
signiﬁcantly different from all other nutrient supply treatments (Appendix C3.1). Among
the grazed communities, composition in the XHIGH treatments differed signiﬁcantly
from only the LOW and MED treatment while the HIGH treatment also differed from the
LOW treatment (Appendix C3.1).

The algal composition of major taxonomic groups also exhibited dramatic
changes in relative biovolume across treatments (Figure 3.4a). A correlation biplot of
benthic algal species (Figure 3.4b) provided species-speciﬁc correlations with nutrients
(lSt CAP axis) and herbivore presence/absence (2nd CAP axis). Many of the species
present at the ends of the productivity gradient were strongly correlated with the nutrient

supply and grazer CAP axes (Figure 3.4b). In general, ﬁlamentous chlorophytes

61

0.2

 

 

A
'A'
F 0013 1:1
0.1 - ‘ {3*
A
A
0 a
N 0.0 - CA I *
w
- D
i? o : *
% -0.1 - § Al] 5
o ‘k
c
-O.2 - *
-0.3 1 1 l 1 l n
-0.3 -0.2 -0.1 0.0 0.1 0.2 0.3

CAP Axis 1

Figure 3.3: Response of experimental benthic algal communities to various nutrient
supply concentrations (CAP Axis 1) and the presence/absence of herbivores (CAP Axis
2). Ordination of site scores from CAP analysis. Open symbols are ungrazed mesocosms
and closed symbols are grazed mesocosms. Nutrient supply treatments are represented
by different shapes: LOW (circles), MED (triangles), HIGH (squares), and XHIGH

(stars).

62

100 ‘

 

 

 

Percent composition

 

I Filamentous Cyanophytes
E Small Diatoms

Large Diatoms

I Small Colonial Chlorophytes
U Large Colonial Chlorophytes

Filamentous Chlorophytes

Figure 3.4: a) Benthic algal composition divided into broad morphological and taxonomic
groups for ungrazed (O) and grazed (+) communities at each nutrient supply level. b)
Correlations of al gal species with each CAP axis. Only species with correlations of over
0.35 are included. Species are listed with a taxonomic designation in parentheses: (D) =
diatoms, (BG) = blue-green algae (cyanophytes), (G) = green algae (chlorophytes).

63

Figure 3.4 (cont’d).

b

Correlation with CAP Axis 2

1.0

0.5

 

0.0

 

 

 

l

 

 

 

 

  

 

-0.5

0.0

J
0.5 1.0

Correlation with CAP Axis 1

a) Large Oedogonium spp. (G)

b) Scenedesmus brasiliensis (G)

c) Cosmarium sp. B (G)

d) Heteroleibleinia sp. (BG)

e) Pediastrum boryanum (G)

f) Stigeoclonium sp. (basal cells) (G)
g) Rhizoclonium sp. (G)

h) Epithemia adnata (D)

i) Encyonopsis microcephala (D)

j) Achnanthidium minutissimum (D)
k) Epithemia turgida (D)

l) Leptolyngbya sp. F (BG)

m) Tetraedron minimum (G)

n) Cosmarium sp. A (G)

64

o) Leptolyngbya sp. D (BG)
p) Achnanthes exilis (D)

q) Chroococcus minor (BG)

r) Oocystis parva (G)

s) Geminella sp. (G)

t) Pseudanabeana sp. D (BG)
u) Calothrix spp. (BG)

v) Pseudanabaena sp. E (BG)
w) Scenedesmus bijuga (G)

x) Leptolyngbya sp. G (BG)
y) Mougeotia spp. (G)

2) Rhopalodia gibba (D)

a) Bulbochaete spp. (G)

ab) Cladophora sp. (G)

dominated the biovolume of ungrazed algal communities. The identity of these
ﬁlamentous chlorophytes shifted from species such as Mougeotia spp., Cladophora sp.,
and Rhizoclonium sp. at low and intermediate productivities (i. e. these species showed
strong negative correlations with the 1St CAP axis; Figure 3.4b) to primarily large
Oedogonium species at high productivities (i. e. this species had a strong positive
correlation with the 1St CAP axis; Figure 3.4b).

The most striking impact of grazers was the drastic reduction in the relative
biovolume of ﬁlamentous chlorophytes, regardless of system nutrient supply (Figure
3.4a). The understory species that beneﬁted from the removal of the ﬁlamentous
overstory differed along the productivity gradient. At low productivity, the relative
biovolume of diatoms (small-celled Achnanthes exilis, Achnanthidium minutissimum, and
Encyonopsis microcephala and large-celled Epithemia adnata, E. turgida, and Fragilaria
crotonensis) increased in the presence of grazers (Figure 3.4). Alternatively, at very high
productivity, the biovolume of the large, planktonic chlorophytes Cosmarium sp. and
Pediastrum boryanum, and the predator tolerant basal cells of Stigeoclonium sp.
increased with herbivory (Figure 3.4). Thus, at low productivity, herbivory favored an
understory of generally edible diatom species, although at least one predator tolerant
species was present in low numbers (Stigeoclonium sp. — the large colonial chlorophyte in
Figure 3.4a). In contrast, in high productivity systems, herbivory generally favored more
predator resistant (or tolerant) species (mainly Stigeoclonium sp.).

Overall, herbivores tended to shift algal composition from a community
dominated by a vulnerable, ﬁlamentous overstory to one primarily composed of less

vulnerable understory species. The herbivore treatment clearly resulted in drastically

65

different algal communities that were then introduced into a common environment and
tested for differences in the degrees of nutrient and predator limitation. I expected that
nutrient limitation of ungrazed algal communities would decrease across the productivity
gradient while predator limitation would increase, adherent with predictions from food
chain models. In contrast, I expected that grazed algal communities would exhibit strong,
but declining nutrient limitation and declines in predator limitation across the
productivity gradient, due to an herbivore-mediated shift toward more predator resistant,
less resource efﬁcient taxa at high productivities.
Phase 2 — Limitation assays

As predicted, the algal communities that developed under different herbivore and
nutrient environments exhibited different degrees of nutrient and predator limitation
(Figure 3.5); however, the extent of these differences varied with overall system
productivity (DO). An initial MAN OVA (using the Phase 1 experimental design) on
nutrient and predator limitation demonstrated a signiﬁcant global hypothesis test (Pillai’s
trace = 0.915, F2,23=124.2, P<0.0001). A separate ANOVA on nutrient limitation
revealed a signiﬁcant interaction (F334 = 3.49, P=0.031) between nutrient supply level
and grazing history (i. e. ungrazed communities versus grazed communities). Slices of
the interaction show that nutrient limitation of grazed and ungrazed communities differed
signiﬁcantly at LOW nutrient supply (F124 = 6.86, P = 0.015) and marginally at XHIGH
(F134 = 3.23, P = 0.085), the same two treatments with the largest differences in
composition (Figure 3.3, Appendix C3.1). Although the patterns of limitation were
complex, in general the ungrazed algal communities at high productivities showed a

stronger decline in nutrient limitation than the grazed communities (Figure 3.5a).

66

 

 

 

 

 

 

 

 

 

0.3 r ,
.0 a) Nutrient
0.2 - t.
It;
3 0.1—
C
.0
E 0.0
E
-' -o.1—
_02 l l L L J
'0.1 0.2 0.3 0.4 0.5 0.6
0.3 r
b) Predator
3 0.2 —
(U
E
1: ' A
99- e ......... A 4
E 0.0 A'"; ......... i
-0.1 —
_0_2 1 1 1 1 1
0.1 0.2 0.3 0.4 0.5 0.6

Productivity - DO (mg/L/hr)

Figure 3.5: a) Nutrient and b) predator limitation of historically grazed (closed symbols)
and ungrazed (open symbols) benthic algal communities. Regression lines are LOWESS
ﬁts (tolerance = 0.6) for grazed (solid lines) and ungrazed (dashed lines) relationships
with net primary productivity (dissolved oxygen — DO).

67

In the nutrient addition treatments of the limitation assays, I observed 25-50%
grazer mortality that may have confounded the nutrient limitation results. To assess the
effect of grazer mortality, I used an alternative method of calculating nutrient limitation
in the absence of grazers (i. e. by subtracting chlorophyll values in predator removal
treatments from chlorophyll in treatments where nutrients were added and predators
removed) and compared the results to those using the original method (see Methods).

The only change resulting from the alternative method (data not shown) was that nutrient
limitation no longer differed between LOW grazed and ungrazed communities. Because
this difference may be confounded by grazer mortality, I do not discuss this result further.

On average, predator limitation in ungrazed algal communities was higher than in
grazed communities, especially at low productivities (Figure 3.5b). An ANOVA on
grazed and ungrazed predator limitation demonstrated a signiﬁcant interaction between
nutrient supply rate and grazing history (F334 = 4.28, P = 0.015). Slices of this interaction
indicated that predator limitation differed signiﬁcantly between LOW ungrazed and
grazed communities (F 1,24 = 14.45, P = 0.0009).

To measure predator limitation in ungrazed algal communities, these algae had to
be exposed to grazers one week prior to the initiation of the one-week limitation assays.
One concern might be that this period of grazing rendered the ungrazed and grazed
communities so similar that no differences in limitation would be apparent. To address
this issue, I compared algal species composition at the end of Phase 1 to composition two
weeks later, in control communities at the end of the limitation assays. I tested the effects
of time and treatment (4 levels: LOW and XHIGH, grazed and ungrazed) on algal

composition in a two-way perMANOVA. The perMANOVA revealed a signiﬁcant time

68

by treatment interaction (F 3,24 = 2.25, P = 0.01), slices of which indicated that only the
XHIGH ungrazed communities exhibited a signiﬁcant effect of time (t1 = 2.80, P = 0.03).
Thus, by the end of the limitation assay, the species composition of ungrazed XHIGH
communities was signiﬁcantly different than when grazers were ﬁrst introduced.
DISCUSSION
The presence of herbivores clearly altered benthic algal community composition
and in turn affected how the community responded to the removal of limiting factors.
However, differences in the limitation of ungrazed and grazed communities depended
upon the underlying productivity of the ecosystem. At intermediate productivities,
grazed and ungrazed algal communities did not differ in the extent of nutrient or predator
limitation. In contrast, these factors differed most in the endpoint communities; that is, in
the communities where nutrient supply rate differed by more than an order of magnitude
(10 and 200 rig/L). Thus, I will focus on the algal composition and limitation differences
in these communities.
Algal composition and the keystone predation hypothesis
The algal compositional shifts observed along productivity and herbivore
gradients were consistent with the expectations of the keystone predation hypothesis.
Herbivore-free communities were dominated by vulnerable, ﬁlamentous chlorophytes,
and more predator resistant species beneﬁted from herbivory in high productivity
systems. In most cases, ﬁlamentous chlorophytes are considered predator resistant, but
they can be vulnerable if herbivory occurs before ﬁlaments outgrow the edible, early
stages of development (Dudley and D’Antonio 1991, DeNicola et al. 1990). Because

grazers were present in the mesocosms before substantial benthic algal mats had

69

developed, it is highly probable that herbivory prevented signiﬁcant ﬁlamentous
chlorophyte establishment.

At low productivity, herbivore grazing of the ﬁlamentous overstory beneﬁted the
understory community of diatoms. In this experiment, the indirect, positive effects of
grazing (i. e. the removal of the overstory) on diatoms likely outweighed the direct,
negative effects on these morphologically edible species. In contrast, at higher
productivities, herbivore consumption of the ﬁlamentous chlorophytes resulted in the
dominance of a predator tolerant species (sensu Chase et al. 2000), the basal cells of
Stigeoclonium, and potentially predator resistant planktonic chlorophytes, Cosmarium sp.
and P. boryanum (see Darcy-Hall, Chapter 2).

While the distribution of predator resistant algal species along the productivity
gradient ﬁts the predictions of the KPH, it is not clear whether the same is true of
competitive (sensu Tilman) algal species. In general, ﬁlamentous chlorophyte species
have higher nutrient requirements than most species, (due to their large size; Steinman et
al. 1992), so it may appear that grazing fell more heavily on the inferior nutrient
competitor (contrary to expectations). However, in vertically structured benthic algal
communities, superior competitors may be more appropriately viewed as those species
that more effectively acquire nutrients. Because the overstory of ﬁlamentous
chlorophytes would have direct contact with water-column nutrients, they were, in this
case, the competitively superior species (Rosemond 1996, Steinman et al. 1992).
Therefore, herbivory fell on the competitive dominant, which is generally consistent with
the predictions of the KPH. Others have observed a similar tradeoff between access to

nutrients and predator resistance among benthic algal species (e. g. Graham and

70

Vinebrooke 1998, Rosemond et al. 1993, Steinman et al. 1992). For example, Graham
and Vinebrooke (1998) found that competitive growth forms of periphyton (i. e. tall,
loosely attached, overstory species) were most abundant when grazers were absent, but
that species composition shifted to ﬁrmly attached, short and tall growth forms when
herbivores were present.
Patterns of limitation of grazed and ungrazed communities

Nutrient and predator limitation differed in ungrazed and grazed communities,
likely due to grazer-mediated compositional differences. These compositional
differences between grazed and ungrazed communities were most dramatic at low and
very high productivities (see Figure 2.3), resulting in concomitantly larger differences in
the extent of nutrient and predator limitation. In addition, the direction of these
differences in nutrient and predator limitation supported the predictions of the KPH.

Nutrient Limitation — Nutrient limitation in both grazed and ungrazed
communities declined with enrichment, and similar to the predictions of the limitation
model (Darcy-Hall, Chapter 1, Appendix A), ungrazed nutrient limitation declined more
strongly than grazed nutrient limitation. As described above, algal communities
coexisting with herbivores were comprised of more predator tolerant, competitively
inferior algal species that were likely nutrient limited. For example, dominant
Stigeoclonium basal cells increased in the nutrient addition treatments of the limitation
assays (data not shown), potentially contributing to higher values of nutrient limitation in
grazed communities.

The shape of the relationship between nutrient limitation and productivity was

surprisingly similar in ungrazed and grazed treatments. Interestingly, nutrient limitation

71

exhibited a non-linear response to variation in net primary productivity, peaking at
intermediate productivity and declining thereafter. While the non-linearity was
unexpected in ungrazed communities, the response is consistent with expectations for
grazed communities. For instance, the predicted relationship between resource limitation
and productivity in food webs (see Figure 1.1, Darcy-Hall, Chapter 1) corresponds to a
portion of the relationship between nutrient limitation and productivity in Figure 3.2 (the
segment of the graph aﬁer ~0.23 mg/L/hr DO). It is possible that the range of mesocosm
productivity in this experiment includes lower levels than those used for the limitation
model parameters.

The peak in nutrient limitation at intermediate productivity may be due to the low
algal biomass of communities at corresponding levels of productivity in this study.
Hillebrand (2002) found that the effect of nutrient additions decreased with initial algal
biomass. Similarly, nutrient limitation (the result of a nutrient addition) was particularly
high when pre-limitation assay chlorophyll a was low (compare Figure 3.2c and Figure
3.5a). Thus, initial algal biomass clearly affected the extent to which the community
responded to the addition of nutrients. However, initial algal biomass did not explain the
entire response, because nutrient limitation of grazed and ungrazed communities differed
at high productivities, but initial chlorophyll a values did not.

Predator Limitation — Predator limitation in communities with and without
herbivores differed strongly at low productivity, yet converged at higher productivities.
Similar to predictions of the limitation model, predator limitation in low productivity,
grazed communities was lower than in ungrazed communities, likely due to the presence

of more grazer resistant species. Indeed, there were higher percentages of Stigeoclonium

72

sp. basal cells, which on average, responded negatively to the removal of herbivores (data
not shown). However, there were more predator resistant species in grazed communities
along the entire productivity gradient, thus one might expect predator limitation in all
grazed mesocosms to surpass that of ungrazed communities.

The absence of a difference between predator limitation in grazed and ungrazed
communities at higher productivities may be potentially explained in two ways. First, the
large difference in predator limitation at low productivity may involve a positive effect of
nutrient recycling by herbivores. This hypothesis is supported by the fact that predator
limitation in grazed communities is negative at low productivity. Yet at higher
productivities, the net effect of predators in the historically grazed communities declined
and predator limitation was generally positive and low. Second, contrary to my
predictions for simple food chain models, predator limitation in historically ungrazed
communities did not increase across the productivity gradient. In fact, predator limitation
in ungrazed communities at intermediate to high productivities did not differ from that of
grazed communities. This may be due in part to the timing of herbivory and the
composition of these XHIGH communities. Grazed XHIGH communities were primarily
composed of the predator tolerant basal cells of Stigeoclonium sp.; thus, it was not
surprising that predator limitation values were low. Ungrazed XHIGH communities were
dominated by ﬁlamentous chlorophytes, which were vulnerable to herbivory in grazed
communities, because grazing occurred at early stages of the ﬁlaments’ development.
However, when the ungrazed clay pots, with their fully developed ﬁlamentous
communities, were exposed to herbivores prior to the limitation assays, these ﬁlaments of

Oedogonium spp. were almost certainly resistant (Darcy-Hall, Chapter 2, Steinman et al.

73

1992). Therefore, at the time of the limitation assays, both ungrazed and grazed algal
communities were effectively predator resistant, perhaps causing predator limitation
values to be low in both treatments.

Comparisons to other studies - This study illustrates that herbivores can
signiﬁcantly alter the species composition of benthic algal communities, but that the
magnitude of this effect varied with overall system productivity. The variation in the size
of the herbivore effect was nonlinear since the largest effects were observed at the lowest
and highest nutrient supply rates, but not at intermediate levels. Very few studies have
examined the relationship between productivity and herbivore effects on composition,
although Chase et al. (2000) found decreasing effects of large herbivores on grassland
plant species composition as the productivity (i. e. precipitation) of grasslands increased.

Only a handful of studies have experimentally examined the relative effects of
resources and predators in habitats that vary in productivity (e. g. Hillebrand and Kahlert
2001), and only one (Darcy-Hall, Chapter 2) examines patterns of short-term limitation.
Hillebrand and Kahlert (2001) did not ﬁnd a conclusive pattern of nutrient versus grazer
effects across a productivity gradient of lakes; however, their gradient included only 3
systems.

In some respects the pattern of limitation of grazed communities in this study and
the one observed in 13 southwest Michigan lakes (Darcy-Hall, Chapter 2) were similar.
First, in both cases nutrient limitation was stronger than predator limitation across the
entire productivity gradient studied. Additionally, negative values of predator limitation
were observed in both cases; however, the most negative values of predator limitation in

lakes were in high productivity systems, while the opposite was true in the mesocosm

74

experiment. It is no surprise though that strong relationships between limitation and
productivity were easier to detect in a simpler, experimental context than in the complex
and highly variable environment of lake ecosystems.
Conclusions

This research demonstrates a clear effect of herbivores in shifting benthic algal
species composition along gradients in system nutrient supply. Furthermore, these
herbivore-mediated shifts in composition lead to differences in the extent of algal nutrient
and predator limitation in historically grazed versus ungrazed communities. In general,
nutrient limitation in grazed algal communities was stronger than in ungrazed algal
communities, especially in systems with high nutrient supply. In low nutrient supply
systems, predator limitation in historically ungrazed algal communities appeared to be
stronger than in grazed communities. These differences in nutrient and predator
limitation were consistent with predictions of the limitation model and the keystone
predator hypothesis. Thus, my results provide evidence for a link between keystone
predator-mediated compositional shifts and patterns of benthic algal nutrient and predator

limitation in habitats along a productivity gradient.

75

LITERATURE CITED

APHA. 1980. Standard methods for the examination of water and wastewater, 15th
Edition. APHA, Washington, DC.

Anderson, M. J. 2004. PERMANOVA_2factor: a FORTRAN computer program for
permutational multivariate analysis of variance (for any two-factor ANOVA
design) using permutation tests. Department of Statistics, University of

Auckland, New Zealand.

Anderson, M. J. and Robinson, J. 2003. Generalized discriminant analysis based on
distances. Australian & New Zealand Journal of Statistics 45: 301-318.

Anderson, M. J. and T. J. Willis. 2003. Canonical analysis of principal coordinates: an
ecologically meaningful approach for constrained ordination. Ecology 84:51 1-
525.

Anderson, M. J. 2002. CAP: a FORTRAN computer program for canonical analysis of
principal coordinates. Department of Statistics, University of Auckland, New
Zealand.

Anderson, M. J. 2001. A new method for non-parametric multivariate analysis of
variance. Austral Ecology 26:32-46.

Bachmann, R.W., and DE. Canﬁeld, Jr. 1996. Use of an alternative method for
monitoring total nitrogen concentrations in Florida lakes. Hydrobiologia 323:1-8.

Bjojk—Ramberg, S., and Anell, C. 1985. Production and chlorophyll concentration of
epipelic and epilithic algae in fertilized and nonfertilized subarctic lakes.
Hydrobiologia 126:213-219.

Borchardt, M. A. 1996. Nutrients. In Algal Ecology: Freshwater Benthic Ecosystems
(R. J. Stevenson, M. L. Bothwell, and R. L. Lowe, eds.), pp. 183-227. Academic
Press, San Diego, California, USA.

Chase, J. M., M. A. Leibold, A. L. Downing, and J. B. Shurin. 2000. The effects of
productivity, herbivory, and plant species turnover in grassland food webs.
Ecology 81 :2485-2497.

Chase, J. M., M. A. Leibold, and E. Sims. 2000. Plant tolerance and resistance in food
webs: community-level predictions and evolutionary implications. Evolutionary

Ecology 142289-314.

Crumpton, W.G., T.M. Isenhart, and PD. Mitchell. 1992. Nitrate and organic N
analyses with second-derivative spectroscopy. Limnology and Oceanography

76

37:907-913.

DeNicola, D. M., C. D. McIntire, G. A. Lamberti, S. V. Gregory, L. R. Ashkenas. 1990.
Temporal patterns of grazer-periphyton interactions in laboratory streams.
Freshwater Biology 23:475-489.

Dixon, P. M. 2001. The bootstrap and the jackknife. In Design and Analysis of
Ecological Experiments (S. M. Scheiner and J. Gurevitch, eds.) pp. 267-288.
Oxford University Press. New York, New York, USA.

Downing, J. A., C. W. Osenberg, and O. Sarnelle. 1999. Meta-analysis of marine
nutrient-enrichment experiments: variation in the magnitude of nutrient limitation.
Ecology 80:1157-1167.

Dudley, T. L. and C. M. D’Antonio. 1991. The effects of substrate texture, grazing, and
disturbance on macroalgal establishment in streams. Ecology 72:297-309.

Fairchild, G. W., R. L. Lowe, and W. B. Richardson. 1985. Algal periphyton growth on
nutrient-diffusing substrates: an in situ bioassay. Ecology 66:465-472.

Feminella, J. W., and C. P. Hawkins. 1995. Interactions between stream herbivores and
periphyton: a quantitative analysis of past experiments. Journal of the North
American Benthological Society 14:465-509.

F retwell, S. D. 1977. The regulation of plant communities by food chains exploiting
them. Perspectives in Biology and Medicine 20: 169-185.

Ginzburg, L. R., and H. R. Akcakaya. 1992. Consequences of ratio-dependent predation
for steady-state properties of ecosystems. Ecology 73: 1536-1543.

Graham, M. D. and R. D. Vinebrooke. 1998. Trade-offs between herbivore resistance
and competitiveness in periphyton of acidiﬁed lakes. Canadian Journal of
Fisheries and Aquatic Science 55:806-814.

Grover, J. P. 1994. Assembly rules for communities of nutrient-limited plants and
specialist herbivores. American Naturalist 143:158-282.

Hairston, N. G., F. E. Smith, and L. B. Slobodkin. 1960. Community structure,
population control, and competition. American Naturalist 44:421-425.

Hall, S.R., V.H. Smith, D.A. Lytle, and MA. Leibold. In press. Constraints on primary
producer N:P stoichiometry along N:P supply ratio gradients. Ecology.

Hill, W. R., H. L. Boston, and A. D. Steinman. 1992. Grazers and nutrients

simultaneously limit lotic primary productivity. Canadian Journal of Fisheries
and Aquatic Science 49:504-512.

77

Hillebrand, H. 2002. Top-down versus bottom-up control of autotrophic biomass — a
meta-analysis on experiments with periphyton. Journal of the North American
Benthological Society 21:349-369.

Hillebrand, H., C-D. Dﬁrselen, D. Kirschtel, U. Pollingher, and T. Zohary. 1999.
Biovolume calculation for pelagic and benthic microalgae. Journal of Phycology
35:403-424.

Hillebrand, H. and M. Kahlert. 2001. Effect of grazing and nutrient supply on
periphyton biomass and nutrient stoichiometry in habitats of different
productivity. Limnology and Oceanography. 46:1881-1898.

Holt, R. D., J. P. Grover, and D. Tilman. 1994. Simple rules for interspeciﬁc dominance

in systems with exploitative and apparent competition. American Naturalist
144:741-771.

Lamberti, G. A., S. V. Gregory, L. R. Ashkenas, J. L. Li, A. D. Steinman, and C. D.
McIntire. 1995. Inﬂuence of grazer type and abundance on plant-herbivore
interactions in streams. Hydrobiologia. 306: 179-188.

Legendre, P. and E. D. Gallagher. 2001. Ecologically meaningful transformations for
ordination of species data. 0ecologia 129:271-280.

Legendre, P. and L. Legendre. 1998. Numerical Ecology, Developments in
Environmental Modelling 20, Elsevier. Amsterdam, The Netherlands.

Leibold, M. A. 1999. Biodiversity and nutrient enrichment in pond plankton
communities. Evolutionary Ecolog/ Research 1:73-95.

Leibold, M. A., J. M. Chase, J. B. Shurin, and A. L. Downing. 1997. Species turnover
and the regulation of trophic structure. Annual Review of Ecology and
Systematics 28:467-494.

Leibold, M. A. 1996. A graphical model of keystone predators in food webs: trophic
regulation of abundance, incidence, and diversity patterns in communities.
American Naturalist 147:784-812.

Lowe, R. L. 1996. Periphyton patterns in lakes. In Algal Ecology: Freshwater Benthic
Ecosystems (R. J. Stevenson, M. L. Bothwell, and R. L. Lowe, eds), pp. 57-76.
Academic Press, San Diego, California, USA.

Lowe, R. L. and Y. Pan. 1996. Benthic algal communities as biological monitors. In

Algal Ecology: Freshwater Benthic Ecosystems (R. J. Stevenson, M. L. Bothwell,
and R. L. Lowe, eds.), pp. 57-76. Academic Press, San Diego, California, USA.

78

Lubchenco, J. 1978. Plant species diversity in a marine intertidal community:
importance of herbivore food preference and algal competitive abilities.
American Naturalist 112123-39.

Lubchenco, J. and J. Cubit. 1980. Heteromorphic life histories of certain marine algae as
adaptations to variations in herbivory. Ecology 61:676-687.

Lubchenco, J. and S. D. Gaines. 1981. A uniﬁed approach to marine plant-herbivore
interactions. 1. Populations and communities. Annual Review of Ecology and
Systematics 12:405-437.

Marks, J. C., and R. L. Lowe. 1989. The independent and interactive effects of snail
grazing and nutrient enrichment on structuring periphyton communities.
Hydrobiologia 185 :9- 1 7.

McCardle, B. H. and M. J. Anderson. 2001. Fitting multivariate models to community
data: a comment on distance-based redundancy analysis. Ecology 82:290-297.

McCormick, P. V. 1994. Evaluating the multiple mechanisms herbivore-algal
interactions in streams. Hydrobiologia 291:47—59.

McCormick, P. V., and R. J. Stevenson. 1989. Effects of snail grazing on benthic algal
community structure in different nutrient environments. Journal of the North
American Benthological Society 8: 162-172.

McCormick, P. V., and R. J. Stevenson. 1991. Grazer control of nutrient availability in
the periphyton. 0ecologia 86:287-291.

McCune, B. and J. B. Grace. 2002. Analysis of Ecological Communities. MjM Software
Design. Gleneden Beach, Oregon, USA.

McNaughton, S. J ., M. Oesterheld, D. A. Frank, and K. J. Williams. 1989. Ecosystem-
level patterns of primary productivity and herbivory in terrestrial habitats. Nature
341:142-144.

Mills, E. L., and A. Schiavone Jr. 1982. Evaluation of ﬁsh communities through
assessment of zooplankton populations and measures of lake productivity. North
American Journal of Fisheries Management 2: 14-27.

Mulholland, P. J ., A. D. Steinman, A. V. Palumbo, J. W. Elwood, and D. B. Kirschtel.
1991. Role of nutrient cycling and herbivory in regulating periphyton
communities in laboratory streams. Ecology 72:966-982.

Oksanen, L., S. D. Fretwell, J. Arruda, and P. Niemela. 1981. Exploitation ecosystems
in gradients of primary productivity. American Naturalist 118:240-261.

79

Osenberg, C. W. and G. G. Mittelbach. 1996. The relative importance of resource
limitation and predator limitation in food chains. In Food webs: Integration of
patterns and dynamics. (G. A. Polis and K. O. Winemiller, eds.), pp. 134-148.
Chapman and Hall, New York, New York USA.

Paine, R. T. 1966. Food web complexity and species diversity. American Naturalist
100:65-75.

Patrick, R. and C. W. Reimer. 1966. The Diatoms of the United States Exclusive of
Alaska and Hawaii, Monogr. 13, Vol. 1. Academy of Natural Science,
Philadelphia, Pennsylvania, USA.

Power, M. E. 1992. Top-down and bottom-up forces in food webs: do plants have
primacy? Ecology 73:733-746.

Prepas, EB, and RH. Rigler. 1982. Improvements in quantifying the phosphorus
concentration in lake water. Canadian Journal of Fisheries and Aquatic Sciences
39:822-829.

Rosemond, A. D. 1996. Indirect effects of herbivores modify predicted effects of
resources and consumption of plant biomass. In Food webs: Integration of
patterns and dynamics. (G. A. Polis and K. O. Winemiller, eds.), pp. 149-159.
Chapman and Hall, New York, NewYork USA.

Rosemond, A. D., P. J. Mulholland, and J. W. Elwood. 1993. Top-down and bottom-up
control of stream periphyton: effects of nutrients and herbivores. Ecology
74:1264-1280.

Steiner, C. F. 2001. The effects of prey heterogeneity and consumer identity on the
limitation of trophic-level biomass. Ecology 82:2495-2506.

Steinman, A. D. 1996. Effects of grazers on freshwater benthic algae. In Algal Ecology:
Freshwater Benthic Ecosystems (R. J. Stevenson, M. L. Bothwell, and R. L.
Lowe, eds.), pp. 341-373. Academic Press, San Diego, California, USA.

Steinman, A. D., P. J. Mulholland, and W. R. Hill. 1992. Functional responses
associated with growth form in stream algae. Journal of the North American

Benthological Society 11:229-243.

Sterner, R. W. 1986. Herbivores’ direct and indirect effects on algal populations.
Science 231 :605-607.

Trexler, J. C. and J. Travis. 1993. Nontraditional regression analyses. Ecology 7421629-
1637.

Tuchman, N. C. and R. J. Stevenson. 1991. Effects of selective grazing by snails on

80

benthic algal succession. Journal of the North American Benthological Society
10:430-443.

Vance, R. R. 1978. Predation and resource partitioning in one-predator-two prey model
communities. American Naturalist 112:797-813.

Wehr, J. D. and R. G. Sheath. 2003. Freshwater Algae of North America. Academic
Press, San Diego, California, USA.

Welschmeyer, N. A. 1994. Fluorometric analysis of chlorophyll a in the presence of
chlorophyll b and phaeopigrnents. Limnology and Oceanography 39:1985-1992.

Wetzel, R. G. 1996. Benthic algae and nutrient cycling in lentic freshwater ecosystems.
In Algal Ecology: Freshwater Benthic Ecosystems (R. J. Stevenson, M. L.
Bothwell, and R. L. Lowe, eds.), pp. 641—667. Academic Press, San Diego,
California, USA.

Wootton, J. T., M. E. Power, R. T. Paine, and C. A. Pﬁster. 1996. Effects of
productivity, consumers, competitors, and El Niﬁo events on food chain patterns

in a rocky intertidal community. Proceedings of the National Academy of
Sciences 93:13855-13858.

81

CHAPTER 4
EVIDENCE FOR THE KEYSTONE PREDATION HYPOTHESIS IN THE
BENTHIC ALGAE: LINKING SPECIES’ TRAITS TO ENVIRONMENTAL
VARIABLES IN A FOURTH-CORNER ANALYSIS
INTRODUCTION
Species are adapted to particular environments and as a result, traits that confer

high ﬁtness in one environment may not necessarily prove favorable in a different
environment. In many cases, traits exhibit functional tradeoffs, preventing species from
being well adapted to multiple environments. When this occurs, species composition is
predicted to change along ecological gradients. For example, Gleeson and Tilman (1990)
demonstrated a tradeoff among plant species in their relative allocation of biomass to
above and below- ground structures along a successional gradient, and Tilman (1988) also
argues for a general tradeoff between allocation to roots and shoots along gradients in soil
fertility. Species may also exhibit tradeoffs in their ability to compete for resources and
their resistance to predation along gradients in potential primary productivity (Paine
1966, Lubchenco 1978, Vance 1978, Holt et al. 1994, Grover 1994, Leibold 1996, Chase
et al. 2000). Increases in ecosystem resource supply (i. e. potential primary productivity)
often lead to concomitant increases in both primary producer and consumer biomasses
(Leibold 1999, Ginzburg and Akcakaya 1992, McNaughton et a1. 1989, Mills and
Schiavone 1982). Thus, if primary producer-prey species exhibit a competition-
resistance tradeoff, then prey community composition might shift to dominance by more
predator-resistant prey as both productivity and predator biomass increase. This

phenomenon has been termed the keystone predation hypothesis (KPH; Leibold 1996,

82

Holt et al. 1994, Grover 1994), which focuses on the role of predators in shifting prey
community structure.

The keystone predation hypothesis (Leibold 1996, Holt et al. 1994, Grover 1994)
predicts how species’ traits should vary across systems of different resource supply. At
low productivities, when predator levels are also low, vulnerable prey species with
superior resource efﬁciency are expected to dominate. Alternatively, as resources and
predators increase, prey communities will shift in dominance to more predator
invulnerable species. By the nature of the implicit tradeoff, these resistant species should
be relatively poor resource competitors.

A number of studies have documented the predicted predator-mediated shi its in
prey species composition along productivity gradients (Darcy-Hall, Chapters 2, 3, Steiner
2003, Chase et a1. 2000, Bohannan and Lenski 1999, 2000); however, only one of these
studies (Steiner 2003) has directly examined the traits of the resultant prey communities.
This is surprising, as a key component to testing the KPH model is that the prey species
favored under speciﬁc conditions (e. g. high productivity-high predator biomass) exhibit
the hypothesized traits (e. g. predator resistance). Instead, most studies provide anecdotal
evidence that species’ traits match those predicted by the competition-resistance tradeoff
(e. g. Darcy-Hall Chapter 2, Hillebrand 2000, Rosemond et al. 1993). We still lack direct
evidence for strong associations between the traits of competitive ability and resistance
and the expected environmental contexts.

Here, I examine this question within a predator-prey system composed of benthic
algal prey and benthic grazer predators. A competition-resistance tradeoff has been

suggested in the periphyton (e. g. Hillebrand 2000, Graham and Vinebrooke 1998,

83

Rosemond et al. 1993), but has never been directly addressed. To investigate benthic
algal trait-environment correlations, I combine short-term assays measuring algal traits of
competitive ability and grazing resistance with a long-term, outdoor mesocosm study of
algal species composition across an array of environmental conditions. The key algal
traits of competitive ability and resistance are measured as relative growth rates under
various conditions of resource limitation or grazing regime, respectively. Data from
these experiments are analyzed using the fourth-comer analysis, a statistical technique
that correlates species’ traits to environmental variables (Legendre et al. 1997). This
technique has been used successfully by other researchers examining trait-environment
correlations (e.g. Hooper et al. in press, Legendre et al. 1997). I hypothesize that benthic
algal species’ traits of superior resource competitive ability will be most common in low
nutrient supply environments and will thus be negatively correlated with higher
productivity environments. In contrast, I expect algal grazing resistance traits to be more
common at high nutrient supply (and thus high grazer environments) and to therefore
exhibit positive correlations with high productivity.

Empirical and theoretical studies of the KPH have generally focused on shifts in
prey composition along gradients in nutrient supply. However, primary producers may
also compete for light, yet I know of no studies examining how primary producers (i. e.
algae) might tradeoff in their ability to compete for light and to resist predators. Most
studies that have examined species’ abilities to compete for light have focused on
tradeoffs between light and nutrients (e.g. Huisman et al. 1999, Grover 1997). As part of
this experiment, I also included treatments measuring algal light competition to gain a

preliminary understanding of how light availability may ﬁrnction within a keystone-

84

predator context. This effort will motivate future research concerning the nature of
benthic algal tradeoffs involving light, nutrients, and resistance to predators.
METHODS
Overview of the experimental approach

The fourth-comer analysis uses three matrices to generate a species’ traits by
environment correlation matrix (the “D-matrix”): a species x site matrix (the “A-
matrix”), a species by species’ trait matrix (the “B-matrix”), and a site x environment
matrix (the “C-matrix”). In this study, the A- and C-matrices were generated from a ﬁeld
experiment using cattle tank mesocosms (i. e. the sites) in which I created a wide-range of
environmental conditions (based on variation in nutrient supply, herbivore biomass, and
light levels) across which a regional species pool of benthic algae were allowed to sort.
At the conclusion of the mesocosm experiment, algal species composition data and
various abiotic and biotic environmental variables from each mesocosm were compiled to
create the A- and C-matrices. Algal traits for the B-matrix were measured in two shorter-
terrn bioassays, which measured changes in algal species’ composition in several
resource-lirniting or grazing environments. The change in each species’ absolute
biovolume (also referred to as a species’ relative growth rate) in each treatment of the
two bioassays was used to represent these species’ traits for the B-matrix. Below, I
describe each of these experiments and bioassays in detail as well as the speciﬁcs of the
fourth-comer analysis.

Outdoor mesocosm experiment
The experiment used to develop the A- and C-matrices was conducted in 1000-L

cattle tank mesocosms during the summer of 2002 at the W. K. Kellogg Biological

85

Station’s experimental pond lab facility (Hickory Corners, MI, USA). Cattle tank
mesocosms were acid-washed and ﬁlled with well water prior to the initiation of the
experimental design. All cattle tanks were afﬁxed with a 1-mm mesh lid, composed of
ﬁberglass window screening. The main treatments were four nutrient supply
concentrations [LOWl (13:1 N:P ratio), LOW2 (50:1 N:P ratio), MED, and HIGH (both
50:1 N:P ratio)] and 2 light treatments (shaded/unshaded, at LOW2 and HIGH nutrient
concentrations). Each of these six treatments was crossed with the presence or absence
of a grazer community, yielding 12 treatments that were replicated 4 times. Six 4-inch
(~10-cm) terracotta clay ﬂowerpots served as sampling substrates for the benthic algal
communities in each mesocosm. Nitrogen (as ammonium nitrate) and phosphorus (as
potassium phosphate) were added to the water column to establish the four levels of
productivity. Target nutrient supply concentrations were 10 (LOWl and LOW2), 30
(MED), and 75 (HIGH) rig/L total phosphorus (TP), with nitrogen added in a 50:1 molar
ratio (or 13:1 in the LOWl treatment). The design included a low N:P ratio treatment to
provide species with an opportunity to invade a low nitrogen environment. These
productivity levels encompassed natural variation in TP and total nitrogen (TN) of local
lakes and ponds near KBS (Hall et al. in press, Darcy-Hall, unpubl. data). Nutrients
were added every 5 days throughout the experiment to maintain the approximate target
nutrient concentrations. Low light treatments were achieved by adding shadecloth to the
lids of the low light mesocosms. Shadecloth blocked approximately 95% of ambient
light entering the cattle tank. Benthic grazers were added at approximately 0.5 g dry
mass/m2 and included 3 species of gastropods (Physa sp., Helisoma sp., and Gyraulus

sp.) as well as amphipods (Hyallela azteca). Zooplankton (Dahpnia spp.) were also

86

added to each cattle tank to control populations of phytoplankton that could potentially
shade periphyton communities.

Arrays of clay pots were established in the littoral zone of six lakes for
accumulation of algal inoculum for the mesocosm experiment. The lakes were chosen to
span a wide gradient of TP (13.5 — 77.5 ug/L) and TN (208.9 - 1869 ug/L). To seed the
mesocosms, benthic algae were scraped from several clay pots from each lake and mixed
with deionized water. Resulting algal slurries were sieved through a 125-um sieve to
remove micro- and macrograzers, combined and mixed in a carboy, and then distributed
in 100-mL aliquots to each cattle tank. The inoculation process was repeated once every
3 weeks for the duration of the experiment to incorporate natural temporal variation in
algal species composition.

Algal species were allowed to sort among these experimental environments for 16
weeks, after which endpoint communities were sampled twice: on 3-4 September, 2002,
and 17-18 September, 2002. Only data from the ﬁrst sampling period are presented here.
During each sampling, the algal and grazer communities from three clay pots per cattle
tank were combined and subsampled to measure chlorophyll a, algal species
composition, and grazer densities. Algae were removed from clay pots as described
above, and samples for species identiﬁcation were preserved in 10% formalin. Algal
enumeration was conducted as in previous work (Darcy-Hall, Chapters 2, 3), and all
analyses of species compositional shifts were conducted using arc-sine, square-root
transformed species relative biovolume data. Chlorophyll a measurements were taken on
the same day as sampling, using cold ethanol extractions followed by narrow-band

ﬂuorometry (sensu Welshmeyer 1994). Grazers were preserved in 70% ethanol and

87

biomass was calculated using length-mass regressions (C. W. Osenberg, unpubl. data).
In addition, several abiotic variables were measured during the ﬁrst sampling period.
Total (TP and TN) and dissolved nutrients (ammonia, nitrate, soluble reactive
phosphorus, and silica) were measured using spectrophotometry and standard methods
(Bachmann and Canﬁeld 1996, Crumpton et al. 1992, Prepas and Rigler 1982, APHA
1980). Temperature, conductivity, and pH were measured using a Hariba® multi-sensor
probe. Dissolved oxygen (DO) measurements were measured with a YSI® oxygen probe
and were taken at dusk, dawn, and dusk, during a 24-hour period. Net ecosystem
productivity was calculated as the difference in D0 between dawn and dusk (in units of
mg/L/hour). Additionally, light measurements (umol/mz/s) at the bottom of each cattle
tank were sampled with a LI-COR® light meter (LI-COR Biosciences, Lincoln, NE,
USA).
Resource limitation bioassay

Resource limitation bioassays were conducted to measure the response of algal
species to limiting levels of several key resources. Algae were exposed to 'four limiting
resource environments (P, N, silica, and light) and a control with all resources at
saturating levels. Each experimental treatment included the target limiting resource at
very low levels, with the other resources maintained at saturating levels. The bioassay
was conducted in 19-L plastic buckets, and target limiting nutrient values were
ascertained from the literature (Borchardt 1996): 3 ug/L P, 100 ug/L N, and 100 ug/L Si.
Because the nutrient content of the laboratory well water often exceeded these values,
buckets were ﬁlled with l-L well water and 8-L deionized water, and nutrients were

added back at saturating levels where necessary. Saturating levels of key nutrients were:

88

50 ug/L P, 1350 ug/L N (50:1 N:P ratio), and 5000 ug/L Si. The light-limiting
environment was created using 95% shadecloth. Each of the ﬁve treatments was
replicated 4 times, for a total of 20 experimental units. Experimental arrays were
arranged on two tables in the pond lab in a randomized, complete block design, with each
table as a block. All buckets were placed under full-spectrum ﬂuorescent lights and
exposed to a 16-hour light, 8-hour dark cycle.

A hyperdiverse slurry of algae collected from local lakes and one clay pot-
substrate were added to each bucket. Algae quickly consumed the added nutrients
(Darcy-Hall, unpubl. data); so to maintain the treatment conditions, nutrients were added
to the buckets every 2 days to maintain saturating levels where necessary (i.e. there were
no additions of target limiting nutrients). Loading rates were determined from pilot
experiments that examined how quickly levels of soluble reactive phosphorus (SRP),
ammonia (NH4+), nitrate (N03) and silica (Si) decreased (Darcy-Hall, unpubl. data).
Algae from clay pots were sampled after two weeks (as described above) for chlorophyll
a and species composition, and the water column of each bucket was sampled for
dissolved nutrient analyses.

Grazer resistance bioassay

Algal communities that had established on clay pots in two treatments of the
cattle tank experiment were brought into the laboratory for the grazer resistance assay.
The two source mesocosm treatments were the grazed, 10 ug/ L nutrient supply (LOW +)
and the grazed, 75 ug/L nutrient supply (HIGH +) treatments. One clay pot from three
replicates of each source treatment was exposed to three grazing treatments (two grazer

species and a no-grazer control) in the laboratory. The two herbivores were species of

89

gastropod snails (Physa sp. and Helisoma sp.). Thus, the experimental design crossed
two source environments with three grazer treatments, each replicated three times for a
total of 18 experimental units (19-L plastic buckets). All grazers were added at 0.1 ug
dry mass/m2, a value based on natural densities in a high-grazer biomass lake (Darcy-
Hall, Chapter 2). Clay pots and grazers were placed in buckets ﬁlled with well water and
N and P were added to each bucket to achieve a medium level of productivity (~ 30 ug/L
TP). Clay pots were exposed to a l6-hour light, 8-hour dark cycle via full spectrum
ﬂuorescent lights in the laboratory. After 10 days, clay pots were sampled (as above) for
chlorophyll a, algal community composition, and grazer density.

Algal traits were calculated the same way in both bioassays. The absolute
biovolumes of each species in the control and experimental treatments (e. g. Low P or
Helisoma) were entered into the equation for limitation (see Darcy-Hall, Chapter 2 & 3).
The resultant value is identical to a relative growth rate (Harper 1977) for each species in
each limiting condition. Below, I will refer to these traits as, for example, “phosphorus
competition” or “resistance to Helisoma in HIGH+ environments”.

Statistical analyses

My main goal was to test for correlations between nutrient competition traits and
productivity variables, as predicted by the KPH. Therefore, I did not include light
treatments in initial analyses of species compositional shifts and the fourth-comer
analysis. Instead, I focused on nutrient and herbivore effects on species composition in
all experiments. In subsequent analyses, I examined species trait correlations associated

with all three treatments (nutrient supply, herbivore presence/absence, and light), and I

90

present an abbreviated analysis of these results. A more detailed examination of the
effects of light on algal species composition will be treated elsewhere.

Tests of treatment establishment — Treatment effects of log (x+l)-transformed
productivity variables and grazer biomass in all experiments were tested initially by
MANOVA (when necessary; SAS, v.8, SAS Institute) on several variables of interest,
followed by univariate ANOVA’S (Systat, v.8, SPSS, Inc.) of each variable when
appropriate. The long-term mesocosm experiment was not set-up in a fully crossed
design, thus two subsets of data, each containing a fully crossed design, were tested
separately. The ﬁrst subset was a 2-way analysis testing the effects of nutrient supply
(low N:P ratio and 10, 30, 75 ug/L TP) and herbivore presence/absence. The second
subset was a 3-way analysis testing the signiﬁcance of nutrient supply (10 ug/L and 75
ug/L TP), herbivore presence/absence, and light (shaded/unshaded) treatments.

Analyses of species composition — My general approach to examining the
responses of species to experimental treatments was ﬁrst to test for signiﬁcant treatment
effects with permutational MANOVA (perMANOVA; see Darcy-Hall Chapter 3,
McCardle and Anderson 2001, Anderson 2001, Anderson 2004, Jones 2002, for details)
and then conduct an ordination of species composition to visualize treatment separation
and/or individual species’ responses. The cattle tank mesocosm data set was divided into
two subsets, as described above. Each perMAN OVA used 10,000 permutations to test
for signiﬁcant effects and a-posteriori pairwise comparisons were tested when
appropriate.

Relationships between algal species composition and environments within the

mesocosm experiment were analyzed with a redundancy analysis ordination (Canoco v.4;

91

see Legendre and Legendre 1998, ter Braak and Smilauer 1998, Darcy-Hall, Chapter 2,
for details). Algal compositional data were Hellinger-distance transformed (see Darcy-
Hall Chapter 2 and Legendre and Gallagher 2001) and were constrained using ﬁve
environmental variables from the cattle tank experiment (DO, TP, TN, N:P ratios, and
grazer biomass), all of which were also included in the C-matrix of the fourth-comer
analysis. I used the same variables in both analyses to facilitate comparisons between
fourth-comer results and individual species’ responses in the cattle tanks.

Species compositional changes in the resource competition assay were analyzed
with a 1-way perMAN OVA (4 levels), followed by post-hoe treatment comparisons
(Jones 2002). Individual species’ responses and separation of treatments were analyzed
using a canonical analysis of principal coordinates (CAP; see Anderson and Robinson
2003, Anderson and Willis 2003, Anderson 2002, Darcy-Hall Chapter 3, for details).
Composition data were analyzed to maximize separation between the four treatments (i. e.
a discriminant analysis was run on principal coordinates) and 10,000 permutations were
used to test for the signiﬁcance of the ﬁrst canonical axis.

I used a 2-way perMAN OVA on algal relative biovolume data in the grazer
resistance bioassay to test for effects of grazer type (3 levels) and source of algal
community (2 levels). The only signiﬁcant factor affecting algal composition was the
community’s source mesocosm, thus I did not perform an ordination of species’ data.

The fourth-corner analysis — To avoid an undeﬁned solution to the matrix
equation, only those species that occurred in all treatments of both the grazer and
resource bioassays were included in the fourth-comer analysis. The data pertaining to

this suite of species were assembled into the A- (species x site) and B- (species x species’

92

trait) matrices. The A-matrix was composed of 1’s and 0’s only, representing either
presence or absence, respectively, of each species in each cattle tank. The B-matrix
contained the relative grth rates of each species under each treatment of the two
bioassays. My main hypothesis concerned trait correlations with environmental variables
that might encompass system primary productivity, so the C-matrix (environmental
variables x site) was composed of 5 abiotic and biotic productivity variables (DO, TP,
TN, N:P, and light levels [in a second analysis only]), and grazer biomass for each of the
32 experimental units of the mesocosm experiment. These 3 matrices were entered into a
FORTRAN program (Legendre 2002), which performed the matrix algebra necessary to
produce the D-matrix. In addition, 10,000 permutations of columns of the A-matrix were
performed (Model 2; Legendre et al. 1997) to determine the statistical signiﬁcance of the
D-matrix correlations between algal traits and environmental variables. Holm’s
procedure (Rice 1989) was employed to assess the table-wide signiﬁcance (or = 0.05) of
the D-matrix.
RESULTS

Cattle-tank mesocosm experiment (the A- and C—matrices)

The nutrient supply treatments created a gradient in overall productivity, with
higher nutrient supply resulting in increased TP, TN, and N:P ratios in the water column,
higher chlorophyll a (chl) on clay pots, and higher net ecosystem productivity measured
by DO (see Appendix D4.1 for MANOVA and ANOVA results). Grazer biomass did not
signiﬁcantly differ along the established productivity gradient (F117 = 1.011, P = 0.385).

Light levels at the bottom of shaded mesocosms (average = 49.6 umol/mZ/sec d: 0.49

93

S.E.) were reduced by 92% compared to light reaching the bottom of unshaded cattle
tanks (average = 650.0 umol/mZ/sec i 12.7 S.E.).

Benthic algal community composition differed due to variation in nutrient supply,
herbivore biomass, and (marginally) their interaction (perMANOVA results, Table 4.1A)
The ﬁrst 4 axes of a redundancy analysis constraining variation of benthic algal relative
biovolume to the environmental variables found in the mesocosms (Figure 4.1) explained
a signiﬁcant amount of variation (21 .9% total) in species composition (F = 1.516, P =
0.034). The ﬁrst RDA axis represents 10.6% of the variation in species composition and
correlates strongly with general productivity (r90 = 0.62, rm = 0.60, rm» = 0.42, er =
0.40). The 2nd RDA axis represents 5.7% of the variation in species composition and
correlates strongly with grazer biomass (rch = 0.54). Species scoring negatively along
the 1St RDA axis dominated in low productivity environments, while species scoring
positively along this axis dominated in high light environments. A suite of species,
composed primarily of diatoms and cyanobacteria, dominated low productivity
environments, while one species (large species of the ﬁlamentous chlorophyte
Oedogonium) correlated strongly with high DO (Figure 4.1). A complete list of species-
environment correlations is in Appendix D4.2. Two species (basal cells of Stigeoclonium
and Scenedesmus acutiformis) exhibited grazer resistance and correlated strongly with
high herbivore biomass, while several species (e.g. the ﬁlamentous chlorophytes
Bulbochaete spp. and Geminella spp.) prospered in environments with low grazer
biomass (Figure 4.1).

A perMAN OVA of the subset of mesocosm data including the light treatment

indicated that light explained a signiﬁcant portion of variation in species composition and

94

Table 4.1: Results of permutational MANOVAs on the relative biovolume of benthic
algal species composition for two subsets of the mesocosm experiment: A) the nutrient
supply-light subset and B) the nutrient supply-N:P ratio.

 

 

 

 

 

 

Factor df F-statistic P-value

Nutrient supply 3 1.685 0.050

Grazer 1 2.540 0.022

Nutrient supply * grazer 3 1.611 0.062

Error 24

ggnilicant post-hoc comparisons

Nutrient supply levels 1 and 4 1 1.765 0.001
Factor df F-statistic P-value
Nutrient supply 1 1.130 0.314
Light 1 7.200 0.001
Grazer 1 0.995 0.379
Nutrient supply * light 1 0.621 0.616
Nutrient supply * grazer 1 1.059 0.345
Light * grazer 1 3.634 0.021
Nutrient supply * light * grazer 1 1.205 0.280
Error 24

 

95

Ewing HoEoo-§o-« 05 E wows—0E 805 2a Eon E 86on .833 65 E coca—0E E ES 856 E3

36 55 cote—oboe a “E? 866% >30 .AmoSanHoEov 3w? :6on u av Amorﬁaocgov cam—m seamen—n N am: .mEoEc H Rd
”momoﬁeﬂmm E nonmewuov 038553 a 6:5 BB: 2m 86on .Eoﬁtonxo EmooomoE E3 2:8 05 E 682963 63:22 36on
Rm? :0 E9328 323568 a 80m @863 00:33 8.88 358:835 98 $.88? 33$ 88% 5.28% 00 53mm ”:5 SEE

 

m_x< (mm 1.
A9 Essie noeomsﬂ mm . . . . .
A0vn__oo.nuun d» 53.6.0325 «a O F m C O O m 01 O Fr
33 maze-.30 3888280 on 0. Fr
A00 $56336 «5:668:80 um _l _ _ _

EV 8am msoaqocm 8

,w .3 6:036:66:me on

0 dm meomnmcmgomm mm

A08 < .3 3636:2581

. . EC .3 £380

A00 dam 52530360 096..
Aomv m .8 £585:

69 < .3 $585:

.0 .8 6393253

.m .3 6395033

(mam gaucbouqoq

69 m...» gauges...
«=33 .0

35.0 .0

523.8333»: .0
Seaman...“ .0

av 6:095 «5020:5530
6v .8 $3558

«29:. .m

§ «352.. seasim

5v 26:589an mamqocoxucm
A00 m d» 553530

g .3 90:3me

A09 SEES 388850
69 a3 €528

A00 dam $563038

A09 do memoocch<

A09 mean? mmqmooemﬁ?‘
8V 538.6%..58 EEESﬁEuc.

c 8 x 3.82 L md-

.0
(U

\\

c»
‘\

I

n_._. 4...

on Announce. . 00.00000
x 000000

z.—- A-noun-ooIIOIIu-o-noon-noon... 00

. n

\
\.
‘4

anonm..or.:-...x_Ecoo.Uun..:>Bx>~

. c
O
O
O
O
O

0

O

O

O

9

O

O

>

1V «.3
r- 5-..? 3.0m

4! -3.:.E.om

I

ll

rt
In 0 -
1
SD09 V008 puZ

h
N
"a

U
(U

E
<

 

96

there was a signiﬁcant light by grazer interaction, but no signiﬁcant main effects of
nutrient supply and grazers (Table 4.1B). When light was included in the RDA, an
additional 6.6% of variation in algal species composition was explained and light
correlated strongly (r = -0.71) with the 1St RDA axis (data not shown). Appendix D4.2
includes species-light correlations from this RDA.

The overall goal of the mesocosm experiment was to provide a diversity of
environments across which a heterogeneous species pool of algae could sort. The
presence or absence of these species in the particular environments (the A-matrix)
provided the foundation for linking algal traits to environmental variables.

Resource competition and grazer resistance trait assays (the B-matrix)

Resource competition assay — Two experimental units were lost from the resource
competition assay due to a mix-up in experimental application. Thus, the low P and low
light treatments were only replicated 3 times. The low nutrient treatments of the
competition assay resulted in reduced levels of the key, limiting nutrient (compared to
controls), but not always to target limiting levels. Mean levels (i 1 S. E.) of NH; and
N03' in low nitrogen treatments were 180 ug/L i 6.4 and 75.7 ug/L :t 0.33, respectively,
compared to 1146 ug/L NH4+ i 512 and 87.0 ug/L NO3' i: 0.67 in controls (an 84%
reduction). Dissolved silica in low silica treatments averaged 2.3 mg/L d: 0.20, compared
to 4.6 mg/L i 1.7 in controls (a 50% reduction). Soluble reactive phosphorus levels in
low phosphorus treatments averaged 8.4 ug/L :t 0.79, compared to 10.1 ug/L :l: 1.8 in
controls (a 17% reduction). Control experimental units received nutrient additions every
2 days to maintain saturating levels, but these nutrients were often quickly consumed,

contributing to the apparent similarity between nutrient levels (particularly SRP) in

97

control and experimental treatments. Even though the low nutrient treatments never
received additional inputs of the target nutrient throughout the assay, levels did not
always decline as expected. The P and Si treatments in particular might not have reached
a truly limiting condition, and this corresponded to there being no observed difference in
chlorophyll a on clay pots in P or Si treatments relative to controls (Appendix D4.3).
Chlorophyll a on clay pots differed signiﬁcantly among treatments (Fm = 50.39,
P<0.0001), but this was primarily due to strong declines in the low nitrogen treatment
(Appendix D4.3).

Algal species composition did not differ signiﬁcantly among the nutrient
limitation treatments (perMANOVA results: F3,” = 1.22, P = 0.247) and a CAP analysis
of algal species relative biovolume did not show signiﬁcant separation among the 4
treatments (Figure 4.2a, canonical test statistic of 1St CAP axis, 812:0.64, P = 0.104),
corroborating the perMANOVA results. A correlation biplot of algal species (Figure
4.2b) indicates that many algal species thrived under the saturating conditions in the
controls, but only a few (e. g. large species of Oedogonium in low N treatments) beneﬁted
from limiting nutrient conditions (Figure 4.2b).

If the low light treatment is included in the analysis, the 1St CAP axis explains a
signiﬁcant portion of the variation between groups (812 = 0.84, P = 0.004), and a
perMANOVA also exhibits signiﬁcant differences among treatments (F114 = 1.569, P =
0.051). Post-hoe comparisons revealed that these differences were mainly due to
differences between communities at low light and the other experimental treatments and
controls (Appendix D4.3). Several species (primarily diatoms) dominated in low light

conditions in both the resource competition assay and in the mesocosm experiment;

98

 

 

0.3 _
0.2 — ‘
n
e
0.1 ~—
0) a ‘k
.9 A
(U
9,: oo —- ‘1' A n
0
(3“ 0 ﬁ %
A
-0.2 —-
-03 1,, 1 1 1 1 I 1 l
-O.3 -0.1 0.1 0.3 0.5
1st CAP axis

Figure 4.2: Results of CAP analysis on algal species relative biovolume from resource
competition bioassay: A) Biplot of experimental treatments: low P (triangles), low Si
(stars), low nitrogen (circles), and controls (squares). B) Biplot of species correlations
with the CAP axes. Species are listed with a taxonomic designation in parentheses: (D) =
diatoms, (BG) = blue-green algae (cyanophytes), (G) = green algae (chlorophytes).
Species in bold are those included in the fourth-comer analysis.

Figure 4.2 (cont’d).

99

Correlation with 2nd CAP axis

.o-uong—xh'mz'co-Mbaocrm

 

-1.0

q
a“"0 e
m 9Q 39
x] W a
“ hadfah f
r aa la
agk
v a

l l l I
-0.5 0.0 0.5 -1 .0

Correlation with 1st CAP axis

Ankistrodesmus falcatus (G) r Medium Oedogonium spp.
Chroococcus sp. (BG) 5 Small Oedogonium spp.
Small Cosmarium spp. (G) t Pediastrum boryanum (G)
Fragilaria brevistriata (D) u Pseudanabaena catentata (BG)
F. capucina v Pseudanabaena sp. D

F. construens w Pseudanabaena sp. E
Geitlen’nema tenue (BG) X Rhopalodia gibba (D)
Gloeocystis ampla (G) y Scenedesmus abundans (G)
Gloeocystis vesiculosa (G) 2 S. bijuga

Gomphonema angustatum (D) aa 8. incrassatulus

Leptolyngbya sp. C (BG) ab S. obliquus

Limnothrix sp. B (BG) ac S. serratus

Limnothn'x Iauterbomii ad Staurosira construens (D)
Mougeotia spp. (G) ae Staurosira construens var. venter
Navicula cryptotenellum (D) af Staurastrum spp. (G)
Nitzschia palea (D) ag Tetraedron spp. (G)

Large Oedogonium spp. (G) ah Tychonema sp. (BG)

100

however, the identities of the species were not necessarily the same in both experiments.
While Achnanthidium minutissimum exhibited similar responses in both experiments
Achnanthes exilis and Encyonopsis microcephala did not- These results highlight the
context dependency of individual species’ responses.

Grazer resistance assay — Chlorophyll a levels was signiﬁcantly higher on clay
pots from HIGH+ mesocosms (source: Fm = 7.96, P = 0.015), but was not signiﬁcantly
affected by herbivory (grazer type: Fm = 0.429, P = 0.661; Figure 4.3). A perMANOVA
on algal species relative biovolume indicated that only the identity of the source
mesocosm community (and not grazer type) signiﬁcantly affected algal composition
(F1312 = 4.647, P = 0.003). Despite this, the effect of Helisoma and Physa snails on
certain algal groups depended on the source mesocosm community (Figure 4.3). For
instance, diatom relative biomass increased with Helisoma and Physa predation in algal
communities from LOW+ cattle tanks, but declined due to predation in communities
from HIGH+ cattle tanks (Figure 4.3). For this reason, I calculated separate values for
the B-matrix traits from each of the source communities in the grazer resistance bioassay.
More speciﬁcally, for each of the species included in the B-matrix, I calculated 4 average
trait values, one from each of the source communities (LOW+ and HIGH+) for both
Helisoma and Physa grazing.

T he fourth corner analysis

A total of 14 species were present in all treatments of the trait assays and thus data
for these species were assembled into matrices for the fourth-comer analysis. The fourth-
comer analysis produced a matrix of correlations between the algal species’ traits

(resource competitive ability/resistance to predation) and environmental conditions

101

Chlorophyll a (pg/clay pot)

$3305 85328.2 85% on“ E 5mm ﬂ av e =>EmoeoEo :35 can 52:89:80 515558 Em? ”me 2&5

2.252050 :35. I...» «5229 m magsnouoEoﬁEoBoﬁEw D
moth—50:30 I maﬁzneozoiioﬁo cubs; w 35.39.25 2.8552.”— N

DOHSOm USN “Gogmourﬁ
New + Ema Now + .304

mmEn. mEom=oI 6:50 82?. mEom=wI 6.500

- so

xix/755$? 575/5, 55;
5; 5517/11
5515/5/55

.\eo —.

o\oON

{eon

'v

 

$6?

$3

$8

{can

uomsodmoa %

- $8

$8

 
 

- $2:

102

within the cattle tank mesocosms. The KPH predicts that algal traits of superior
competitive ability should be associated with low productivity environments while
resistance to grazing should be associated with high productivity environments.
Therefore, I expected negative correlations between competition traits and productivity
and positive correlations between resistance traits and productivity. In general, when
traits differ in the sign of the correlation for the same environmental variable, this may
indicate a tradeoff between those traits.

The D-matn'x from the fourth comer analysis indicated that several correlations
between algal traits and environmental variables matched these a priori expectations
(Table 4.2A). For instance, several measures of system productivity (DO, TN, N:P ratio)
in the mesocosms were signiﬁcantly negatively correlated with phosphorus competition,
but signiﬁcantly positively correlated with resistance to Helisoma grazing (although this
was dependant on the source of the algal community). In addition, grazer biomass was
signiﬁcantly positively correlated with both Helisoma and Physa grazing (in HIGH+
treatments), indicating that the ability to resist grazing by these two snail species became
a more common trait with high grazer biomass. However, several environment-trait
correlations did not support the predictions of the KPH. For example, nitrogen
competition was positively correlated with DO, and resistance to Helisoma and Physa by
algal species in LOW+ source mesocosms was negatively correlated with several
productivity variables (6. g. DO and TN). It is not surprising that LOW+ algal
communities did not adhere to the competition-resistance tradeoff because these algal
communities likely experienced less grazing pressure in mesocosms and as a result were

composed of less grazer resistant species (e. g. large colonial chlorophytes [basal cells of

103

 

 

mood mood omod Ed «So mood Sod a
oﬁd m5? cord mood. «8.9 oood. SQ? 2d.
, coo >m_o\oEv 39:05 .390
mood mood add ohod Food mSo wood a
mood mood. mood o5? mFod. mood. mod. 2.?
8:8 Q2 5.9:
mono oFoo wood mood omvd mmvd mood a
god R3. 83 8rd. mood moo? «Ed. 3:
:33 E
mod Bod Ford voto mad mood mood a
mood EQQ mmod moo. mFod ooo.o so? 3:
:53 &
Food Bod omod mood mmod Bod Rod a
mood word. «to 33. mod «mod omod. 3:
325.5 8
+ 5oz .25 + 25. .Ea + :9; .6: + 26. ._m: _m 33 z 33 a 26.. saute,
.8582235
:9:

 

<

104

.mocogoob Em: @622: mam—o5 bEBﬁwaom Am moan—Boob Em: 3923 mmmboco HoEooﬁtsom A< ._o>o_ 023

-038 2: mo “:35ch 3305:on 0.53 28:22.80 2: mo 0:02 .35— modvm 2: do “:35:me 2o Eon E m02o> .3038:on
ooodﬁ memo: 868 AB mouzﬁooba 98 ASE moumcﬁm scum—choc 8d 829, :< .EoEtoaxo 88888 x53 233 05 E

833.8.» 35:80:28 05 8 mxommooﬁ “mob 88m $35 .36on Em? wccdmoOmmo .mmmboqo 858-538 o Bob xEoEAH ”N6 033.

 

 

oood oood oood oood Food oood otd oood a
oood oood. oood ooo¢ oood. oood ooo¢ oood 2....
. con >929: mmmEoB ENEO
oood oood oood ooFd oood oood wood Sod a
oood oood. Food omod. oood. oood mood oood. .3:
8:8 QHZ ._m_oE
oood mood oood oood oood oood oood oood n.
oood. oood oood. oood oood oood oood oood 2....
Aomm\oE\_oE& Eaton 8808?: um Em:
Food oood omod oood oood Food Food mood n.
oood Eod. oood Sod. oood. oood oood. Food. 2....
36.... E
oood oEd oood Food oood mono oood Food n.
Sod Food. oood oood. oood. oood. oood. too. ad.
:5... E
o3d oood FoFd Sod oood Sod Food oood n.
oood. mood. oood oood. oood. oood oood oood. 2d.
.5295 Co
+ no... son. + 26. .22... + no... .6: + 26. .6... £9. 26.. .m 26.. z 26.. 6 26.. 26.2.2,
_BcoE:o.:>:m
on...

 

3.6.3 o... 2.5

105

Stigeoclonium] in Figure 4.3). Although there was evidence for a competition-resistance
tradeoff in the D-matrix, this tradeoff did not operate in all contexts or in isolation of
other potential tradeoffs.

Most of the D-matrix environment-trait correlations were driven by several
abundant species that exhibited strong, corresponding responses in both the trait
experiments and the cattle tank mesocosms. For example, the positive correlation
between productivity (DO and TN) and resistance to Helisoma (source HIGH+) was
largely driven by species that responded positively to Helisoma grazing in the grazer
resistance assay and were found in high productivity environments in the mesocosms
(e. g. large Oedogonium species and Cosmarium sp. B). Other species, which exhibited
similar, corresponding responses in the trait assays and the mesocosm experiment
included Cosmarium sp. A, Leptolyngbya sp. A, Rhopalodia gibba, and the basal cells of
Stigeoclonium sp. Table 4.3 lists the traits and environmental correlations of these key
species. Several of these key species had competition and resistance traits that matched
the predictions of the keystone predation tradeoff, but only in some environments. For
example, Cosmarium sp. B was a superior nutrient competitor and vulnerable to
predation, but only in treatments from LOW+ source mesocosms (Table 4.3). Similarly,
Leptolyngbya sp. A was an inferior nutrient competitor and grazer resistant, but only not
in all source environments (Table 4.3).

If the trait of light competition is included in the fourth-comer analysis, less trait-
environment correlations are signiﬁcant (Table 4.2B), but those that are (e. g. trait-DO
correlations), exhibit the same approximate magnitudes and signs as the analysis without

light. Light reaching the bottom of the mesocosms was signiﬁcantly positively correlated

106

 

 

 

Sod ood oood ood 8.9 8.9 o._8 .oooo Egooooooom
oood. Food. did. oood. ooNo. ood. oood. o..oo.odo..m
ood 2.? oood ood ood ood ddo 56.8868 $5..
oood oood oot? oood. Sod. oood. < do 3.85934
oood ood oood ood oood oood m .do Eoooeooo
oood Ed wood. omod Food omod. < do Eatosooo
ceaﬂotoo Em: cenﬂwtoo ENE—0 cenﬂmtoo QHZ £539.50 2... 5.59950 a... team—9:8 OD
oood. Foo? oood oood or to. o Foo oood oood o_.8 .88 E=.:o.8ooom
ooo. 7 83 do. F- oooe oooé oood oood o Foo oood o.6o.odo§
oood. to. F- oood to. F- ooo¢ oood. oood oové ddo esooooooo moo...
oood oood oood. «mo; oood. oood. SN? omod. < do £35633
oood oooo- oood oood oood. oood Food oood m do Eoooeooo
oood omod oomd 33- oood 83 23 oooo < do Eotosooo
$0.: Ed +2.9 En. +10... ._ox $6.. .9. to: 26.. .w 2,3 z 2,3 d 26..
mom mom mom mom mom mom mom mom oooodo

 

107

.omoboco 588-558 2: 5 83¢?» .oocoaaohéo
28 $85 cow—o 5223 ocosﬂohoo oooomeB 05 03.6 E08598 EooooooE 88:8 05 com @3385 EB 05 E ooocomooo
80:3 860% Em? ooh Agog 2: 895 macaw—oboe .ooooﬁcohéo woo 95% I ooooo 53on 952.5 oo2o> 38,—. ”me 032.

with the traits of light, N, and Si-competition, as well as the ability to resist Physa in
LOW+ treatments (Table 4.2B). Both Cosmarium species and Stigeoclonium basal cells
and Leptolyngbya sp. A affected these correlations (Table 4.3).
DISCUSSION

Linking species’ traits to environmental contexts is key to elucidating the
mechanisms and generality of ecological tradeoffs. Through the design and
implementation of independent experiments as well as the use of a novel statistical
approach, I was able to directly correlate species’ traits and environmental variables.
Results of the fourth-comer analysis provided some evidence in support of the keystone
predation tradeoff in the benthic algae. Signiﬁcant negative correlations between
competition traits and mesocosm productivity and concomitant positive correlations
between resistance traits and productivity indicate that several algal species tradeoff in
their ability to procure resources and their resistance to predation; thus contributing to
compositional shifts along gradients in resource supply and herbivore presence.

Earlier studies of species compositional shifts along productivity (and often
predator) gradients in a variety of habitats have revealed evidence for the keystone
predation tradeoff (Darcy-Hall, Chapter 3, Steiner 2003, Bohannan and Lenski 2000,
Chase et al. 2000, Hillebrand 2000, Leibold 1999, Leibold et al. 1997). Most of these
studies either directly or indirectly documented predator-mediated prey shifts from
presumably vulnerable to less vulnerable species as system primary productivity
increases. However, only Steiner (2003) and the present study go the step further and
attempt to independently demonstrate that resultant prey communities exhibit the

expected traits. Steiner (2003) demonstrated that smaller (1'. e. edible) phytoplankton size

108

fractions exhibited higher growth rates than larger (1'. e. inedible) size fractions and that
the larger fractions were more resistant to zooplanktivorous predators. Despite his coarse
examination of algal traits, Steiner’s (2003) results support those of the present study’s
individual species-based test of keystone predator tradeoffs.

While several of the signiﬁcant correlations in the D-matrix supported the
predictions of the KPH, it is clear that algal traits (as deﬁned here) were context
dependent. In many instances, the same species exhibited differential responses to
herbivory in LOW+ and HIGH+ communities. For example, Cosmarium sp. B and
Rhopalodia gibba had opposite responses to herbivory in LOW+ versus HIGH+
communities (Table 4.3). Differential species’ responses in LOW+ and HIGH+
communities were potentially the result of the large differences in community
composition in the two nutrient environments (Figure 4.3). The response to herbivory by
a focal species would undoubtedly be affected by the identity of and responses to
herbivory of neighboring algal species, which are different in each environment. In
addition, several species, such as Cosmarium sp. A and Stigeoclonium responded
positively to one species of grazer, but negatively to the other species (Table 4.3). These
results highlight that the trait of grazer resistance is context dependent, and that different
species will beneﬁt from herbivory, depending on the species of grazing predator and the
composition and abundance of their competitors.

The context dependence of species’ responses to herbivory has precedence in both
plant and algal communities. For example, in a meta-analysis of plant species and their
traits’ responses to herbivory along a productivity gradient, Pakeman (2004) concluded

that only about 40% of species’ and species’ traits responses were consistent. The other

109

60% of species and trait responses were either context dependent, complex, or random
(Pakeman 2004). Similarly, Darcy-Hall (Chapter 3), Steiner (2003), and Hillebrand
(2000) found that the algal response to herbivory depended upon the productivity of the
habitat, which is consistent with the present results. Differential responses to herbivory
appear to be common in benthic algae. In a literature review, Steinman (1996) concluded
that factors such as grazer type, timing of the grazer-a1 gal interaction, and habitat can
affect not only algal compositional responses but also biomass and diversity responses.
Similarly, I found clear evidence that various grazer types can have opposing effects on
algal composition. These examples of the context dependency of algal responses to
herbivory can be at least partially explained by the effects of different grazer species as
well as by shifts in relative species composition in habitats of different productivity.
There was also evidence of other tradeoffs in the D-matrix. For instance, P-
competition correlated negatively with productivity (DO), but N-competition correlated
positively with DO. These opposing relationships were driven primarily by large species
of Oedogonium, which appeared to be poor competitors for phosphorus, but good
competitors for nitrogen. The mechanism of this response is unclear, but as has been
well documented in the phytoplankton literature, tradeoffs are common within and among
species in their ability to efﬁciently use various resources (reviewed in Grover 1997).
The inclusion of light treatments in both the mesocosm experiment and the
resource competition assay highlighted the importance of light in shifting algal species
composition. In both experiments, the light treatment was particularly strong, resulting in
over 90% light attenuation in low light treatments. The large difference between high

and low light treatments likely contributed to the drastic compositional effects, relative to

110

nutrient supply and herbivore treatments. Therefore, it is evident that subsequent models
of competition-resistance tradeoffs should incorporate a more thorough treatment of light
competition.

Future work concerning algal traits in the context of a competition-resistance
tradeoff may also beneﬁt from a broader deﬁnition of competition than the one used here.
The resource competition assays were designed to measure competition for a limiting
nutrient, hence I only measured classic, Tilman-type competition for resources. As
evidenced by previous results (Darcy-Hall, Chapters 2, 3), competition in benthic algal
communities can also be viewed as access to nutrients rather than survival in a limiting
condition. While I found evidence for a competition-resistance tradeoff among the
benthic algae using the classic measure of resource competition, it is possible that the
incorporation of a broader view of competition would strengthen the results.

In summary, the fourth-comer analysis provided support for the trait-environment
associations predicted by the KPH. The algal trait of P-competition correlated negatively
with several productivity variables, while the trait of resistance to Helisoma and Physa
grazing (in HIGH+ source communities) correlated positively with productivity.
Therefore, the competition-resistance tradeoff inherent to keystone predator models was

at least one of several potential mechanisms inﬂuencing benthic algal species
composition and coexistence. In addition, comparisons both within and among these
experiments highlight the context dependency of species’ responses — a phenomenon that

deserves more direct examination and review.

111

LITERATURE CITED

APHA. 1980. Standard methods for the examination of water and wastewater, 15th
Edition. APHA, Washington, DC.

Agrawal, A. 1998. Algal defense, grazers, and their interactions in aquatic trophic
cascades. Acta 0ecologica 19:331-337.

Anderson, M. J. 2004. PERMANOVA_2factor: a FORTRAN computer program for
permutational multivariate analysis of variance (for any two-factor ANOVA
design) using permutation tests. Department of Statistics, University of
Auckland, New Zealand.

Anderson, M. J. and Robinson, J. 2003. Generalized discriminant analysis based on
distances. Australian and New Zealand Journal of Statistics 45: 301 -3 1 8.

Anderson, M. J. and T. J. Willis. 2003. Canonical analysis of principal coordinates: an
ecologically meaningful approach for constrained ordination. Ecology 84:511-
525.

Anderson, M. J. 2002. CAP: a FORTRAN computer program for canonical analysis of
principal coordinates. Department of Statistics, University of Auckland, New
Zealand.

Anderson, M. J. 2001. A new method for non-parametric multivariate analysis of
variance. Austral Ecology 26232-46. ‘

Bachmann, R.W., and DE. Canﬁeld, Jr. 1996. Use of an alternative method for
monitoring total nitrogen concentrations in Florida lakes. Hydrobiologia 32321-8.

Bohannan, B. J. M. and R. E. Lenski. 2000. The relative importance of competition and
predation varies with productivity in a model community. American Naturalist
156:329-340.

Bohannan, B. J., and R. E. Lenski. 1999. Effect of prey heterogeneity on the response of
a model food chain to resource-enrichment. American Naturalist 153273-82.

Borchardt, M. A. 1996. Nutrients. In Algal Ecology: Freshwater Benthic Ecosystems
(R. J. Stevenson, M. L. Bothwell, and R. L. Lowe, eds.), pp. 183-227. Academic
Press, San Diego, California, USA.

ter Braak, C. J. F. and P. Smilauer. 1998. CANOCO reference manual. Centre for
Biometry. Wageningen, The Netherlands.

Chase, J. M., M. A. Leibold, A. L. Downing, and J. B. Shurin. 2000. The effects of

112

productivity, herbivory, and plant species turnover in grassland food webs.
Ecology 81:2485-2497.

Crumpton, W.G., T.M. Isenhart, and PD. Mitchell. 1992. Nitrate and organic N
analyses with second-derivative spectroscopy. Limnology and Oceanography
37:907-913.

Ginzburg, L. R., and H. R. Akcakaya. 1992. Consequences of ratio-dependent predation
for steady-state properties of ecosystems. Ecology 7 3:1536-1 543.

Gleeson, S. K. and D. Tilman. 1990. Allocation and the transient dynamics of
succession on poor soils. Ecology 71:1144-1155.

Graham, M. D. and R. D. Vinebrooke. 1998. Tradeoffs between herbivore resistance
and competitiveness in periphyton of acidiﬁed lakes. Canadian Journal of
Fisheries and Aquatic Sciences 551806-814.

Grover, J. P. 1997. Resource Competition. Chapman and Hall, New York, New York,
USA.

Grover, J. P. 1994. Assembly rules for communities of nutrient-limited plants and
specialist herbivores. American Naturalist 143:158-282.

Hall, S.R., V.H. Smith, D.A. Lytle, and M.A. Leibold. In press. Constraints on primary
producer N:P stoichiometry along N:P supply ratio gradients. Ecology.

Harper, J. L. 1977. Population Biology of Plants. Academic Press, London, UK.

Huisman, J ., R.R. Jonker, C. Zonneveld, and F .J . Weissing. 1999. Competition for light
between phytoplankton species: experimental tests of mechanistic theory.
Ecology 80:211-222.

Hillebrand, H., B. Worm, and H. K. Lotze. 2000. Marine microbenthic community
structure regulated by nitrogen loading and grazing pressure. Marine Ecology
Progress Series 204:27-38.

Holt, R. D., J. P. Grover, and D. Tilman. 1994. Simple rules for interspeciﬁc dominance
in systems with exploitative and apparent competition. American Naturalist
144:741-771.

Hooper, E. R., P. Legendre, and R. Condit. In press. Factors affecting community
composition of forest regeneration in abandoned, deforested land in Panama.

Ecology.

Jones, D. L. 2002. Users manual for FATHOM: a MATLAB toolbox for multivariate

113

ecological and oceanographic data analysis. Available from:
http://www.ramas.miami. edu/personal/djones/

Knietel, J. M. and J. M. Chase. 2004. Tradeoffs in community ecology: linking spatial
scales and species coexistence. Ecology Letters 7:69-80.

Legendre, P. 2002. http://www.fas.umontreal.ca/biol/casgrain/en/telecharger/index.html#
The4thComer

Legendre, P. and E. D. Gallagher. 2001. Ecologically meaningful transformations for
ordination of species data. 0ecologia 129:271-280.

Legendre, P. and L. Legendre. 1998. Numerical Ecology, Developments in
Environmental Modeling 20, Elsevier. Amsterdam, The Netherlands.

Legendre, P., R. Galzin, M. L. Harmelin-Vivien. 1997. Relating behavior to habitat:
solutions to the fourth-comer problem. Ecology 78:547-562.

Leibold, M. A. 1999. Biodiversity and nutrient enrichment in pond plankton
communities. Evolutionary Ecology Research 1273-95.

Leibold, M. A., J. M. Chase, J. B. Shurin, and A. L. Downing. 1997. Species turnover
and the regulation of trophic structure. Annual Review of Ecology and
Systematics 28 2467-494.

Leibold, M. A. 1996. A graphical model of keystone predators in food webs: trophic
regulation of abundance, incidence, and diversity patterns in communities.
American Naturalist 147:784-812.

Lubchenco, J. 1978. Plant species diversity in a marine intertidal community:
importance of herbivore food preference and algal competitive abilities.
American Naturalist 112223-39.

McCardle, B. H. and M. J. Anderson. 2001. Fitting multivariate models to community
data: a comment on distance-based redundancy analysis. Ecology 822290-297.

McNaughton, S. J ., M. Oesterheld, D. A. Frank, and K. J. Williams. 1989. Ecosystem-
level patterns of primary productivity and herbivory in terrestrial habitats. Nature
341:142-144.

Mills, E. L., and A. Schiavone Jr. 1982. Evaluation of ﬁsh communities through
assessment of zooplankton populations and measures of lake productivity. North

American Journal of Fisheries Management.2: 14-27.

Paine, R. T. 1966. Food web complexity and species diversity. American Naturalist
100:65-75.

114

Pakeman, R. J. 2004. Consistency of plant species and trait responses to grazing along a
productivity gradient: a multi-site analysis. J. Ecol. 92:893-905.

Prepas, EB, and RH. Rigler. 1982. Improvements in quantifying the phosphorus
concentration in lake water. Canadian Journal of Fisheries and Aquatic Sciences
39:822-829.

Rice, W. R. 1989. Analyzing tables of statistical tests. Evolution 43:223-225.
Rosemond, A. D., P. J. Mulholland, and J. W. Elwood. 1993. Top-down and bottom-up
control of stream periphyton: effects of nutrients and herbivores. Ecology

74: 1264-1280.

Steiner, C. F. 2003. Keystone predator effects and grazer control of planktonic primary
production. Oikos 101:569-577.

Steinman, A. D. 1996. Effects of grazers on freshwater benthic algae. In Algal Ecology:
Freshwater Benthic Ecosystems (R. J. Stevenson, M. L. Bothwell, and R. L.
Lowe, eds.), pp. 341-373. Academic Press, San Diego, California, USA.

Tilman, D. 1988. Plant Strategies and the Dynamics and Structure of Plant
Communities. Princeton University Press. Princeton, NJ, USA.

Vance, R. R. 1978. Predation and resource partitioning in one-predator-two prey model
communities. American Naturalist 112:797-813.

Welschmeyer, N. A. 1994. Fluorometric analysis of chlorophyll a in the presence of
chlorophyll b and phaeopigments. Limnology and Oceanography 39:1985-1992.

115

CHAPTER 5
SUMMARY

The research reported here provides a thorough investigation into the nature and
mechanisms driving patterns of benthic algal composition and limitation along gradients
in potential primary productivity. The results combined a mathematical model, data ﬁ'om
a ﬁeld survey of natural lakes, controlled mesocosm experiments and rigorous
multivariate analyses to provide convincing evidence for the importance of a
competition-predation resistance tradeoff among benthic algae. From a conceptual
standpoint, I was able to theoretically and empirically link questions from two important
ecological areas, one pertaining to keystone predator models of competition-resistance
tradeoffs and another concerning the relative effects of top-down and bottom-up factors
on trophic level biomass and community structure. In general, the benthic algae are an
under-studied group in lentic systems, despite their signiﬁcant contribution to primary
production. This fact alone provides an important justiﬁcation to the present research.

Generally, my results demonstrate that the competition-resistance tradeoff
implicit in keystone predator models can impact compositional shifts of benthic algae and
hence patterns of nutrient (bottom-up) and predator (top-down) limitation of algal growth
across habitats (i. e. lakes and mesocosms in southwest Michigan) that vary in potential
primary productivity. More speciﬁcally, my results provide evidence for several key
predictions of the keystone predator model. First, I demonstrated that both natural and
experimental patterns of nutrient and predator limitation of benthic algae are consistent
with expectations of the keystone predation hypothesis (KPH). Second, I show that

predators mediate benthic algal compositional shiﬁs in the manner predicted by the KPH.

116

Finally, I provide support for the expected species’ trait-environment associations under
the KPH.

In Chapter 2, I investigated how benthic algal composition and limitation varied
across lakes along a wide productivity gradient. I found that inter-ecosystem patterns of
nutrient and predator limitation of benthic algae were similar to predictions of the “food
web” case (i. e. KPH) in the limitation model, which incorporates species heterogeneity
within trophic levels. Nutrient limitation remained strong across a productivity gradient
of lakes, while predator limitation remained low and declined across the productivity and
grazer biomass gradient. These results contradicted predictions of classic, trophically
homogeneous food chain models. Additionally, benthic algal species composition across
lakes and traits of key species qualitatively complied with expectations of the KPH. This
research illustrated the importance of heterogeneity both within and among systems to
interactions between a community of organisms and the surrounding abiotic and biotic
environment.

In Chapter 3, I showed that herbivore grazing favored al gal communities
dominated by grazer resistant species. Similar to predictions from the KPH, systems
without herbivores were dominated by vulnerable, superior competitors, while systems
with herbivores were dominated by less vulnerable, inferior nutrient competitors. In
addition, these herbivore-mediated shifts in algal composition led to differences in the
extent of nutrient and predator limitation in historically grazed versus ungrazed
communities. Nutrient limitation of historically ungrazed communities was lower than
that of historically grazed communities, but primarily in high productivity mesocosms.

Predator limitation, on the other hand, was higher in ungrazed communities but only in

117

low productivity mesocosms. These results qualitatively matched predictions from the
KPH. Thus, this experiment provided evidence for a link between herbivore-mediated
compositional shiﬁs and patterns of benthic algal nutrient and predator limitation in
habitats along a productivity gradient.

In my ﬁnal chapter, I used a recently developed statistical technique (the fourth-
comer analysis) to test whether the algal traits of resource competitive ability and
predation resistance were correlated with the environments as predicted by the keystone
predation hypothesis. Keystone predator models predict that algal competitive ability for
resources should be associated with low productivity environments while resistance to
grazing should be associated with high productivity environments. Therefore, I expected
negative correlations between competition traits and productivity and positive
correlations between resistance traits and productivity. Indeed, algal phosphorus and
light-competition correlated negatively with several measures of productivity, while
resistance to Helisoma and Physa grazing correlated positively with several productivity
variables (although this effect depended on the source of algal community exposed to
grazers). However, several species’ trait-environment correlations did not match the
predictions of the KPH, indicating that other trade-offs may be important within benthic
algal communities and deserve further attention. In general, the competition-resistance
tradeoff inherent to keystone predator models was at least one of several potential
mechanisms inﬂuencing benthic algal species composition and coexistence.

Future research concerning tradeoffs within benthic algal communities should
focus in several key areas. First, the context dependency of algal species’ responses to

grazing should be investigated more speciﬁcally and more thoroughly. Research should

118

focus on how and why individual species’ responses vary with productivity of the habitat,
herbivore type, identities of dominant, coexisting species, and interactions between all of
these factors. One potential analytical method to address such questions might involve
partial multivariate analyses. Analyses such as partial redundancy analysis or partial
canonical correspondence analysis partition out variation due to certain variables while
controlling for other variables with known effect.

Second, the idea of competition within benthic algal communities should be more
rigorously investigated. Ideally, experiments involving individual or small groups of
species would measure relative competitive abilities in low and high nutrient
environments to assess both R* competition and nutrient acquisition competition. These
experiments could be paired with predator resistance assays to explore the competition-
resistance tradeoff in a population versus a community context.

Third, the role of light in affecting algal species composition remains
understudied, especially in the context of ecological tradeoffs such as the KPH.
Signiﬁcant priorities should involve studies focusing on the relative importance of light,
nutrients, and herbivores on species compositional shifts in natural and experimental
settings.

Finally, I would like to pursue questions relating to the relative edibility of algal
communities resulting from various levels and/or types of herbivory. It remains to be
demonstrated that presumably grazer resistant algal communities are poor sources of food
for herbivorous grazers. I would like to investigate the grth rates of various grazer
types on algal communities from sources varying in productivity and/or the presence or

absence of herbivores, perhaps from both natural and experimental systems.

119

APPENDICES

120

APPENDIX A

THE LIMITATION MODEL

121

THE LIMITATION MODEL

with Spencer Hall

Model structure and description

Consider the familiar diamond food web model. Two plants compete for a
limiting resource and can potentially coexist with a generalist herbivore (Holt et al. 1994,
Leibold 1996, Grover 1997). This model mixes competition for a limiting resource, R,
among the two plants, N,- (Tilman 1982) with apparent competition among plants which
must withstand herbivory from a grazer, P (Holt 1977). This model can be described
with a simple system of equations representing the growth rates of the herbivore, plants,
and freely available resource as the net result of trophic ﬂuxes and nutrient recycling

(Grover 1997):

%=P[ZeifiN/' ‘0’] , (1'3)

dz]. = N,(ug,(R)—mj _fip) (Lb)

 

dR = “(5 ‘Rl‘Q./N‘Z(“1(R)_mj)+ PLZXQJ ‘equijj l"‘qu (1°C)

5 ’.~
In this model, grth rate of the herbivore (equ. la) is the balance between gains from
feeding on prey item Nj (following a linear functional response at grazing rate j}- and
conversion efﬁciency e,-), and losses at rate d. Because the herbivore’s ﬁmctional
responses for both prey items are linear, the prey items are linearly substitutable (Tilman
1982), and the model produces stable dynamics at equilibrium (Grover 1997). The

growth rate of plant j is the balance of gains from resource acquisition, at rate u]. (R), and

122

losses from predatory ( ij ) and non-predator (m1) sources. Finally, the change in free

nutrient concentration is the balance between gains and losses from dilution (at rate a),
where S is the total concentration of the incoming nutrient resource, the net effect of
nutrient uptake and instantaneous recycling of dead plants (where Q is the nutrient
content of the producer per unit carbon), recycling of producers eaten but not converted
into new grazer tissue (where q is the nutrient content of the grazer per unit carbon), and
recycling of dead producers. Although this free nutrient pool is open to gains and losses
from the outside environment, one can use the mass balance constraint (MBC) technique

to simplify the analysis. For this model, the MBC becomes:
R=S—ZQ_iNj—qP. (2)
,-

Now free resource dynamics follow an algebraic representation as the difference between

S, which becomes the total concentration of nutrients in the system, and those nutrients

locked up in plant (ZQJNJ. ) and grazer (qP) tissues (Grover 1997).
1

Initial progress is best made by making some preliminary assumptions about the

plants and grazer. First, we consider the case of linear uptake of nutrients by the plant,

such that u j (R) = u jR . This keeps the system simpler algebraically (and will be relaxed

elsewhere). Next, we assume that plant N1 is a superior nutrient competitor in the sense
that it can persist at the lowest concentration of freely available nutrients in the absence
of grazing (i. e., it has the lowest R; requirement, or R]. < Rz', following Tilman 1982),

where:

R. =—~". (3)

123

(Thus, mzul > m,u2 ). Yet, the herbivore prefers N 1 to N2 as it has a higher feeding rate

on the former (i.e.,f. > f2). As long as defense against herbivory is costly, we can then
assume that plant N1 trades off ability to compete for nutrients (low Rf) with its ability to
resist predation (high ﬁ). In mathematical terms, this tradeoff ensures that the nullclines

(with slope u j / f I. ) of the plant can cross in R-P space. The cross eventually occurs

because the nullcline for N2 in R-P space is steeper than that for N1 (i.e., ul /fl < u2 /f2 ,
or ul f2 < uzfl ) while the R-axis intercept (Rf) is lower for N1 than N2 (Grover 1997).

This cross of nullclines is critical for species turnover and potential coexistence along
enrichment gradients.

In this well-known case, the herbivore can mediate shifts in plant composition
along gradients of resource supply, S, and facilitate coexistence of the plants at
intermediate S. Assuming sufﬁcient resource supply to support the herbivore, at low and
high nutrient supply, the food chain comprised of the herbivore, plant, and resource

comes to an equilibrium:

 

 

 

NH,‘ = d (4.a)
' ejfi
. .S— .N.. —m. I
P :u.l( Q] .I ) I (4.1))
fj+quj
, .S— .N.’ + .
R-I-‘p =fl( Q] I ) mlq (4C)
fj+quj

This equilibrium requires that nutrient supply exceeds a minimum level,

S > Qij,P. +R..

I . If the herbivore is adaptive, it prefers to consume the prey for which

it is the best competitor (lowest M'p.) Thus, the herbivore can drive biomass of its

124

preferred prey (N1) to lower levels than its other prey (N2), i.e., since N1t < N2‘,
el fl > e2 f2 . It follows then, that unless there are large differences in plant stoichiometry

(Qj) among species, this minimal nutrient level required for invasion of the grazer is
lower for the superior nutrient competitor N1 than for the predation-resistant species N2
(since R]. < R2. and N 1 ,p‘ < N223). Once this nutrient level is surpassed, plant biomass
(iji) does not change with further enrichment but is instead regulated by the grazer’s
traits. However, the herbivore (Pi) and resource (R1; 13‘) both increase with nutrient supply
(Figure Al .1). One can readily show that both P8 and R1,]: increase in a food chain with
the more vulnerable species (N 1) than with the less vulnerable species (N 2), as the slopes
of the relationships between P‘ and S and between R“; and S are positive functions of j}
(Figure Al.l).

At intermediate nutrient supply in the food web, the two plants can potentially

coexist with the predator at an additional equilibrium:

 

 

 

 

Ru : mzf} "mlfg (5.3)
“2f: —ulf2
Pu : m2“: —m1u2 (5.b)
“2f: —ule
N." = sz —e2f2(S—R" ’qP“) (5-0)
ererz "ezszr
NS' = (Q. +e.f.(S-R" -qP") (5.d)
ererz ’ezszr

where now both free nutrient (Rn) and predator biomass (Pﬂ) do not change with

increasing nutrient supply. Instead, increasing nutrient becomes reallocated among the

125

plants as N; decreases and N2 increases (Figure A1 .1). Coexistence requires that nutrient

supply falls within the key range:

qP“ +Q2N,‘ +R" <S <qP“ +Q,N,‘ +R" . (6)
(thresholds e and f of Figure A1 .1, respectively) but also that the plants individually
impact the factor that most limits their own growth. In this model, the plant impacts its

predator at rate e I. f I. and its resource at ratio Qj. Therefore, coexistence requires that

el fl / e2 f 2 > Q1 /Q2 . Once these various requirements are met (intermediate resource

supply, tradeoff in competitive ability versus grazing resistance, and each species has a
greater relative impact on the factor most limiting its growth), coexistence is possible
(Leibold 1996, Grover 1997).
Resource and predator limitation in food chains and food webs

Whether the plant coexists with the predator alone or with its competitor and
predator, it is simultaneously limited by both resources and predation. Following
Osenberg and Mittelbach (1996), we deﬁne limitation as the effect on per capita growth
rate of the plant (or plant assemblage) as one of the limiting factors is removed.
Mathematically, this means that predator limitation (Limp) and resource limitation (LimR)

can be written generally as:

 

 

 

L' — aw, d 7
'm’” _ N dt an ( a)
-" N=NJ.‘,R=R’,P=O
L' aw, (7b)
zm = ‘
R Nidt

 

126

Figure Al.l: Biomass and nutrient partitioning along nutrient supply gradients in food
chains and food webs with plants and a grazer at equilibrium. In the food chain example,
two examples are shown. species 1 is a superior nutrient competitor (i. e., lower R ) but Is
also more vulnerable to predation. As nutrient supply exceeds the plant’s Rj (level a for
species 1 and level b for species 2), biomass of the plant 1ncreases with enrichment until
the predator can invade (at level c for more vulnerable species, and level (I for the less
vulnerable species). Once these thresholds are reached, plant biomass remains constant
(at the predators N ) w1th enrichment as predator biomass (P ) and free nutrient levels
(R ) 1ncrease. The rate at which P and R increase with enrichment 15 higher for species
1 (more vulnerable) than species 2 (less vulnerable). In the food web, coexistence is
possible at intermediate resource supply (between thresholds e and f) In this region, P
and R remain constant with enrichment, as increased nutrient supply 18 reallocated
among species 1 (declining) and species 2 (increasing).

127

Predator Plant

Free

Food Chain

 

Food Web

 

 

 

 

 

IIIIIIIIIIIII

 

 

 

 

 

 

 

— Species 1

"2" ------ Species 2

ti

(0

m , .............................
E .

.2

m

9.

>7.

3:

(D

C

0

D

m

J

C

.9.’
b .........
3 .......

Z

a d
be

 

 

Nutrient Supply, S

128

Figure Al.1

Predator limitation is evaluated as predators are instantly removed from the plant-

herbivore system at equilibrium (i. e., predators are reduced to P=O). Meanwhile,

resource limitation is measured with instant elevation of free nutrient levels to R which
greatly exceeds equilibrial free nutrient concentration, Rip...

In the food chain case, predator limitation becomes:

.

Limp = ujRj‘P — my]. (8)
which is a positive function of nutrient supply (S) since:

a(Limp) _ juj

5S fj+quj. (9)

 

This response of the predator limitation index to nutrient enrichment is identical to the
slope at which predator biomass increases with enrichment (following equation (4.b)).
Additionally, this slope is a positive ﬁmction of feeding rate 1?. Thus, at a given nutrient
concentration, predator limitation is larger and increases more quickly for the more
vulnerable prey (N 1) than for the less vulnerable prey (Figure A1 .2). This result makes
biological sense since the more vulnerable prey is depleted to a lower level, supports a
higher predator population, and enjoys higher available nutrient concentrations than the
more resistant prey (Figure A1 .1). Therefore, removing predators removes a large source
of mortality for the more vulnerable species but a smaller source of mortality for the

more resistant species. Furthermore, this predator limitation index’s S-axis intercept is
equal to the minimal S required to support the food chain (i.e., Q]. N j‘ P‘ + R]. ). Since this

minimal S is higher for the more resistant species, predator limitation is always higher for
the more vulnerable species along gradients of nutrient supply.

Similarly, nutrient limitation of the food chain becomes:

129

Food Chain Food Web

 

 

 

 

 

 

 

_ Species 1
......... Species 2
3 - - - Coexistence
H
m
'U
a: """"
h .....
n. 2 - _ .......
c .
O
I;
(B
2':
.E
_.| a, \
0
h ~ - .-
3 ...........
o "'1.
0)
a:
I:

 

 

 

 

 

Nutrient Supply, S

Figure A1.2: Predator and resource limitation over gradients of enrichment in food
chains and food webs. In food chains, predator limitation is always higher for more
vulnerable species than less vulnerable species. Yet, resource limitation declines more
rapidly over the nutrient supply gradient for the more vulnerable species. At intermediate
resource supply, coexistence of the plants separates a low nutrient regime, dominated by
the superior nutrient competitor, to one dominated by the more resistant plant. In the
region of coexistence, predator limitation typically would decrease with enrichment while
resource limitation always does.

130

LimR =uiR-ml. —f,P‘ (10)
where, again, R is the level to which nutrients are increased to relieve nutrient limitation

(R >> R}; p*). Resource limitation is a negative function of S, since:

6(LimR) =_ .iuj (11)
6S fj +quj

Thus, resource limitation decreases more quickly for the more vulnerable species (N 1)
than for the less vulnerable species (N2). Mathematically, this result arises because the
slope of the relationship between LimR and S is a negative function of increasing 1?
(Figure A1.2). Biologically, this result stems from the fact that the less vulnerable
species persists at higher biomass, supports few predators, but depletes available
resources more than the more vulnerable species. Thus, the less vulnerable species is
simultaneously more resource limited and less predator limited than its competitor
(Figure A1 .2).

In the food web region of resource supply in which both producers coexist, one
must examine limitation of the entire trophic level, NW. If NM, = N1 + N2, then the

aggregate plant trophic level increases with nutrient supply because:

aNW‘ = elfl -ezf2 > O (12)
as ererr "ezszz

 

 

which is true given the assumptions that species have higher impacts on the factor most
limiting them (ensuring a positive denominator) and that the predator is adaptive
(ensuring a positive numerator). The growth rate of the plant assemblage, then, is:

de, _ awl + dN,

— 13
dt dt dz ( )

131

following the chain rule from calculus. A bit of analysis reveals that predator limitation
of the plant assemblage can increase or decrease with enrichment. First, per capita
response of the growth rate of NW to the removal of thepredator is positive, since each
individual plant population’s growth rate increases with predator removal. Change in this

response to increases in S, following the chain rule, is:

af/as _ Harv... /65)
N... N 2

(0!

 

 

(14)

where f a dN /dt . The right-hand term is positive, but the left-hand term can be

to!

positive or negative since:

1_ frf2(er "ezxu1m2 ‘uzmrl

6S (ererz Tezszrxuzfl Turfz)

 

(15)

which is positive if e1 > e2 and negative otherwise (given the stipulations already placed
on the other terms in parentheses of equation (15) above). Thus, when the conversion
efﬁciency of the less preferred prey (e2) is less than the conversion efﬁciency of the more
preferred prey (e1), predator limitation of the plant trophic level always decreases with

enrichment. This case will likely occur when the predator is adaptive (since el fl > e2 f2 ).

When e2 > e1, predator limitation can remain as a negative function of plant biomass if
the right-hand portion of equation (14) exceeds the left-hand portion. This result makes
sense biologically because total plant biomass increases while predator biomass remains
constant in the coexistence region (Figure Al .2). Thus, the plant trophic level
experiences less predation per capita as S increases. Interestingly, nutrient limitation of
the plant assemblage is always a negative function of nutrient enrichment. Following
similar logic as above (equation (14)), it is straightforward to show that resource

limitation of the plant assemblage, while positive, is a non-linear, decreasing function of

132

enrichment as long as R > R” (which ensures that both left and right hand terms of
equation (14) become negative). This result echoes the change in species composition of
the plant assemblage. As the superior nutrient competitor decreases with increasing
nutrient supply, it becomes less nutrient-limited. Meanwhile, the increasing population
of inferior nutrient competitor along the enrichment gradient becomes more nutrient-
limited. Yet, apparently, the net balance of these factors ensures that nutrient limitation
decreases with enrichment in the coexistence zone.

Over a resource supply gradient, then, changes in species composition yield
higher resource limitation and lower predator limitation than expected from a simple food
chain. This fact becomes apparent after comparing a food chain with a superior
competitor-highly vulnerable species (Figure A1.2, “Food Chain” case) and a food web
with many species displaying the keystone-predation tradeoff between competitive ability
and predation resistance (“Food Web” case of Figure Al .2; Figure Al .3). In the food
chain, predator limitation would steadily increase while resource limitation decreased
along an enrichment gradient (Figure A1.2). Species turnover, however, greatly
attenuates this increase in predator limitation. Along a gradient of enrichment, species
turnover switches plant composition from a highly vulnerable species alone to two
species coexisting to a less vulnerable species alone to another species pair coexisting,
and so forth (Leibold 1996). As composition of the plant assemblage changes, the slope
of predation limitation in each single-species section becomes increasingly shallow (since
more resistant species eventually dominate). Meanwhile, predator limitation declines in
regions of plant coexistence (Figures A12, A13). Therefore, over a broad gradient of

enrichment, a zigzag pattern emerges (Figure A1.3). Furthermore, nutrient limitation of

133

 

 

 

 

 

 

 

h
0 N1
3 _,~.
2 ' N
c 0- ,,....N.
.o
a "L
:2."
.§
" ‘” K
3 a ........ N5
8 .N4
a)
m “N3
N1 N2

 

 

 

Nutrient Supply, S

Figure Al .3: Predator and resource limitation in food webs with many plants displaying
the keystone-predation tradeoff of competitive ability versus predation resistance. Over
an enrichment gradient, plant assemblages become increasingly dominated by resistant
species. As a result, predation limitation remains low over a broad gradient of nutrient
supply (as compared to the food chain case). These increasingly resistant species
maintain a high level of resource limitation of the plant assemblage over this same
enrichment gradient.

134

plant assemblages remains much higher with species turnover than expected from the
simple food chain. Here, species of decreasing vulnerability support less predator
biomass per capita but also deplete their resources more per capita than the highly
vulnerable species (Figures A12, A13). Resource limitation always remains high for a

plant assemblage displaying the keystone-predation tradeoff.

135

LITERATURE CITED

Grover, J. P. 1997. Resource Competition. Chapman and Hall, New York, New York,
USA.

Grover, J. P. 1994. Assembly rules for communities of nutrient-limited plants and
specialist herbivores. American Naturalist 143:158-282.

Holt, R. D., J. P. Grover, and D. Tihnan. 1994. Simple rules for interspeciﬁc dominance
in systems with exploitative and apparent competition. American Naturalist
144:741-771.

Holt, RD. 1977. Predation, apparent competition, and the structure of prey
communities. Theoretical Population Biology 12:197-229.

Leibold, M. A. 1996. A graphical model of keystone predators in food webs: trophic
regulation of abundance, incidence, and diversity patterns in communities.
American Naturalist 147:784-812.

Osenberg, C. W. and G. G. Mittelbach. 1996. The relative importance of resource
limitation and predator limitation in food chains. In Food webs: Integration of
patterns and dynamics. (G. A. Polis and K. O. Winemiller, eds.), pp. 134-148.
Chapman and Hall, New York, New York USA.

Tilman, D. 1982. Resource Competition and Community Structure. Princeton
University Press, Princeton, New Jersey, USA.

136

APPENDIX B

CHAPTER 2 APPENDICES

137

 

o8? mono «out. Now com . No Nova? Qt. Eozo 899953

oooo oooo goo mam moo no oooo. mom and 833
Foo. mooo 92 o.: own no o.~o~ mamv ammo :ommmm
moo Soo moo no? o2 oc. Boo Now x25
89? «moo 3.: nor mt o.o ooﬁ oom 8.: 8x85 .25..
oooo owoo v. 5 oo ooo og. ooom mow cog 8.9.8
ouoo. Boo QR o9 now «o o. am now one .28
oooo. oooo 34 t: EN 3 ooﬁ. new 85: .0385 222.2
Boo. oooo odor oo oov E. 3% o? Eat xosoeﬂ
omoo. oooo to 92 oz. oo 5.3? o5 E); 8:933
Soo Eoo. 3 mi no... no no; 0: goo N 3%.. 8?
omoo. oooo 3 oo ta oo o. av 32 ES new;
wood. onto You 3 on” no moon of 395
ovoo. oooo mom 0: moo o.» oﬁm 02 ES 2822,

 

33 .58 $8 .63
Booze: cosmos: $55.5 $53.5 €053 :55 :33
5585 22.52 soufﬂeo o __Eoeo_5 328328 Io E E 9.3

 

€93
mmoE bu ooﬁémo mo @2388 a $383 $530 $.83 bosom 3 05 E @2388 83353 385 38 38:? ”_.Nm x6593

138

Appendix B22: Correlations of the algal species in Figures 2.3 and 2.4 with the six
enviromnental variables included in the RDA. Correlations were calculated using the
species and environmental scores from the ﬁrst 4 ordination axes.

 

 

Species TP TN dilaophyll a gazer plums Willy pH
(HQ/L) (ML) (mg/"12) (mg/m?) (uS/CTTI)

Acrnanthes exilis 0.292 0.102 0.053 0.346 0.%8 0.019
Anphtpleua pal/with 0.043 0.044 0.42 0.0% 0.485 0.213
Arkistrodesms falcons 0.132 0.0% 0.357 0.162 0.706 0.12
Arumieis w’trea 0.308 0.317 0.25 0.209 0.077 0.219
Admaosa sp 0.209 0.178 0.027 0.358 0.6% 0.117
We oonferta 0.053 0.267 0.091 0.173 0.394 0.1%
Aphanothece stamina 0.065 0.137 0.20 0.102 0.552 0.5%
Calotfvix spp. 0.374 0.701 0.273 0.142 0.- 0.014
Groococcus Iimreticus 0%2 0.147 0.237 0.024 0.371 0.720
Chroococcus m'nuus 0.129 0.111 0.332 0.087 0.373 0.549
Cosmariun sp. 8 0.059 0.109 0.289 0.113 0.5% 0.067
CMwaeImtm stagtale 0.192 0.0% 0.038 0.041 0.136 0.691
(Lynne/Ia cwbifam‘s 0.284 0.148 0.2% 0.310 0.311 0.054
Denticda kedzing’i 0.232 0.053 0.132 0.248 0573 0.155
Encyonopsis m'crocedEVa 0.298 0.019 0.051 0.3% 0. 724 0.138
Encyonerm sp. 0.20 0.095 0.240 0.326 0.855 0.0%
Etithem'a tug'tb 0.348 0.%7 0.146 0.092 0%5 0.2%
Frag/aria crotonensis 0.405 0.046 0.527 0.404 0.090 0. 616
Gormhonema clevei 0.047 0.371 0.232 0.126 0.490 0.008
Gonphonema paMILm 0.698 0.6% 0.828 0.521 0.036 0.427
Gonphonema tnncatun 0.630 0.%3 0.754 0.487 0.0% 0.35
Lendyngna sp. A 0.201 0.089 0.214 0.293 0.308 0.172
Leptolynan sp. F 0.252 0.033 0.002 0.332 0.609 0.145
Leno/Mm sp. 8 0.28 0.045 0.449 0.305 0.078 0.24
Lirmothixlauetbm'ii 0.514 0.480 0.4% 0.393 0.193 0.308
Limwothix redekei 0.084 0. 189 0.352 0. 147 0.591 0.230
Abstogloia snith'i 0.509 0.823 0.287 0.268 0.202 0%2
Nancda radosa 0.133 0.130 0.233 0.144 0.551 0.1%
Ntzschia fortioda 0.292 0.163 0.711 0.293 0.217 0.216
Abstoc sp. A 0.087 0.411 0.333 0.097 0.547 0.027
Abstoc sp. 8 0.327 0.444 0.427 0.089 0.414 0.554
Large Oedogonitm spp. 0.449 0.333 0.234 0.356 0.346 0.379
Medan Oedogmiun spp. 0.096 0.205 0.233 0.027 0.353 0229
Small Oecbgomm spp. 0.046 0%3 0.185 0.1% 0.475 0.338
Oocystis spp. 0.238 0.255 0.207 0.229 0.694 0.191
Pedastnm boryantm 0.485 0.233 0.043 0.389 0.543 0.631
Pedastnm dip/ex 0.1% 0.075 0.308 0.162 0.617 0.026
Rhopaloda g'bba 0. 216 0.509 0.425 0.005 0.472 0.135
Scenedesmus bjjuga 0.349 0.150 0.054 0.419 0.7% 0.064
Sceneoamus cpadicauda 0.497 0.417 0.346 0.385 0.289 0.361
Stauastrun sp. A 0.481 0.29 0.037 0.386 0.544 0.628
Stigeoclon‘un sp. 0.165 0.074 0.309 0.165 0.621 0038

 

139

APPENDIX C

CHAPTER 3 APPENDIX

140

Appendix C3.1: Slices of a signiﬁcant nutrient supply by grazer history interaction term
in the permutational MAN OVA on the relative biovolume of benthic algal species
composition in Phase 1 (community establishment) mesocosms.

 

Factor df t-statistic P-value

 

Slices of nutrient supply (ungrazetﬂ

 

 

LOW vs. MED 1 1.288 0.202
LOW vs. HIGH 1 1.306 0.111
LOW vs. XHIGH 1 2.296 0.026
MED vs. HIGH 1 1.139 0.286
MED vs. XHIGH 1 1.840 0.030
HIGH vs. XHIGH 1 1.959 0.029
Slices of nutrient supply (grazed)

LOW vs. MED 1 1.332 0.144
LOW vs. HIGH 1 1.682 0.057
LOW vs. XHIGH 1 2.139 0.024
MED vs. HIGH 1 1.109 0.200
MED vs. XHIGH 1 1.628 0.030
HIGH vs. XHIGH 1 0.901 0.513
Slices of grazer histroyat:

LOW 1 1.551 0.058
MED 1 1.186 0.260
HIGH 1 1.406 0.108
XHIGH 1 2.364 0.031

 

141

APPENDIX D

CHAPTER 4 APPENDICES

142

 

 

 

 

 

on bow
..o.o nos 53 oto snoo n~.o onto 8o . omoo ntn .~ v 885 . .8: . >88 85.2
nooo oo.n nno.o «to onto nnn Rto 3.. . nnoo mod .n v 808 . .8:
onto moo moo o «oto nn.~ nnoo own . n...o otu .n v 88.8 . 89.. .832
.8o.ov noun .o8.ov «Sm Boo non onoo .o.o . «So 3.... .N v .8: . >88 85:2
<2 nuno on. «8o «no Koo En ..oo.o non . n8.o no... .n v 88.5
ooooo .n.v. noo.o oo... .o8.o «on o8.o no... . .8.o «no .n v .8:
.ooo.ov noon. .o8.ov «own 38o own. .Bo onn.o . .o8.ov Bun .N .. >88 85:2
__ 1593852 a
a bum
nnno nn.. n...o on ~.n.o .oo nnno no. 83 n... n 52 no. on n. 8881.88 .852
«noo otn ~no.o .o.o onto on.~ 38o nos. onno Bo . 88o ntn on n 850
goo ntn .8o.ov BR .oo.o nos .o8.ov 8.8 88o 3. n .o8.o non on n. 88852
88389... 85:22
8.9.“. 0888.“. 8.98 088.3 8.9.8 0883.“. 89d 0888.“. 8.9.; 088.3 8 39d 0888.“. .o 88o 6 so: coon“.
auz E a. 8 m .3888
<>oz< 8.35 <>oz<2

 

 

 

 

 

 

058,—. org—E 8o nos—gi an 85.88-... <>OZ<2
.nEoEtomxo 880858 5:8. 238 on. So... n.<>OZ< wan n.<>OZ<E av 83-85. can A<v 83-039 ”:5 55:35.4.

143

Appendix D4.2: Species-environment correlations from RDA of cattle tank mesocosm
experiment. Species in bold are those included in the fourth-comer analysis. Species
with * were only present in the RDA including light treatments. Species are listed with a
taxonomic designation in parentheses: (D) = diatoms, (BG) = blue-green algae
(cyanophytes), (G) = green algae (chlorophytes).

 

 

DO TP TN
Species (mg/L/hr) (HQ/L) (HQ/L)
Achnanthidium exigua‘ (D) 0.469 0.235 0.297
Achnanthes exilis (D) 0.168 0.028 0.158
Achnanthidium Ianceolata' (D) 0.187 0.303 0.363
Achnanthidium minutissimum (D) 0.485 0.316 0.431
Amphora pediculus* (D) 0.170 0.230 0.176
Aphanocapsa sp. (BG) 0.201 0.343 0.486
Aphanocapsa hyalina (BG) 0.254 0.326 0.341
Bulbochaete spp. (G) 0.224 0.337 0.183
Calothrix spp. (BG) 0.415 0.358 0.477
Chroococcus aphanocapsoides (BG) 0.071 0.027 0.123
Chroococcus minutus (BG) 0.334 0.198 0.364
Chroococcus prescotti (BG) 0.425 0.479 0.343
Chroococcus sp. (BG) 0.204 0.291 0.112
Cladophora sp. (G) 0.028 0.122 0.052
Closterium sp.‘ (G) 0.337 0.118 0.171
Cosmarium sp. A (G) 0.052 0.151 0.058
Cosmarium sp. 8 (6) 0.230 0.004 0.192
Encyonopsis microcephala (D) 0.309 0.169 0.325
Epithemia spp. (D) 0.040 0.083 0.066
Geminella spp. (G) 0.028 0.005 0.211
Gomphonema spp. (D) 0.023 0.061 0.102
Leptolyngbya sp. A (BG) 0.573 0.244 0.330
Leptolyngbya sp. 8 (BG) 0.430 0.389 0.381
Leptolyngbya sp. D (BG) 0.097 0.043 0.208
Leptolyngbya sp. E (BG) 0.054 0.078 0.029
Leptolyngbya sp. F (BG) 0.415 0.280 0.317
Limnothrix Iauterbomii (BG) 0.059 0.049 0.125
Limnothrix sp. A (BG) 0.287 0.164 0.386
Limnothn'x sp. B (BG) 0.213 0.288 0.232
Microcystis sp.‘ (BG) 0.007 0.154 0.142
Mougeotia spp. (G) 0.142 0.133 0.078
Nostoc sp. (BG) 0.176 0.168 0.200
Large Oedogonium spp. (G) 0.456 0.240 0.402
Medium Oedogonium spp. (G) 0.104 0.290 0.341
Small Oedogonium spp. (G) 0.121 0.059 0.075
Oocystis spp. (G) 0.163 0.212 0.350
Pseudanabaena sp. A (BG) 0.003 0.006 0.222
Pseudanabaena sp. C (BG) 0.041 0.014 0.047
Pseudanabaena sp. D (BG) 0.035 0.036 0.062
Pseudanabaena sp. E (BG) 0.018 0.098 0.057
Rhizoclonium sp.* (G) 0.027 0.265 0.277
Rhopalodia gibba (D) 0.106 0.200 0.268
Scenedesmus acutiformis (G) 0.136 0.149 0.079
Scenedesmus obliquus (G) 0.289 0.202 0.231
Stigeoclonium sp. basal cells (G) 0.256 0.021 0.075
Tetraedron minimum (G) -0309 0173 -0252

 

144

Appendix D4.2 (cont’d).

 

 

molar N:P grazer biomass light
Species (Hg/pot) (pmol/mzls)
Achnanthidium exigua' (D) 0.261 0.001 0.548
Achnanthes exilis (D) 0.229 0.059 0.404
Achnanthidium Ianceolata* (D) 0.286 0.127 0.582
Achnanthidium minutissimum (D) 0.254 0.017 0.291
Amphora pediculus' (D) 0.02 0.171 0.760
Aphanocapsa sp. (BG) 0.305 0.270 0.250
Aphanocapsa hyalina (BG) 0.083 0.029 0.021
Bulbochaete spp. (G) 0.204 0.453 0.311
Calothn'x spp. (BG) 0.267 0.179 0.483
Chroococcus aphanocapsoides (BG) 0.254 0.102 0.487
Chroococcus minutus (BG) 0.320 0.227 0.281
Chroococcus prescotti (BG) 0.150 0.082 0.359
Chroococcus sp. (BG) 0.257 0.076 0.086
Cladophora sp. (G) 0.276 0.294 0.359
Closten'um sp.‘ (G) 0.186 0.108 0.318
Cosmarium sp. A (G) 0.134 0.111 0.488
Cosmarium sp. 8 (G) 0.325 0.102 0.303
Encyonopsis microcephala (D) 0.298 0.140 0.308
Epithemia spp. (D) 0.014 0.121 0.419
Geminella spp. (G) 0.369 0.328 0.194
Gomphonema spp. (D) 0.082 0.009 0.112
Leptolyngbya sp. A (BG) 0.193 0.089 0.446
Leptolyngbya sp. 8 (BG) 0.056 0.010 0.412
Leptolyngbya sp. D (BG) 0.291 0.225 0.241
Leptolyngbya sp. E (BG) 0.071 0.014 0.558
Leptolyngbya sp. F (BG) 0.114 0.085 0.102
Limnothrix Iauterbomii (BG) 0.289 0.099 0.319
Limnothn'x sp. A (BG) 0.411 0.049 0.287
Limnothrix sp. 8 (BG) 0.046 0.368 0.385
Microcystis sp.‘ (BG) 0.041 0.010 0.331
Mougeotia spp. (G) 0.072 0.007 0.526
Nostoc sp. (BG) 0.086 0.077 0.320
Large Oedogonium spp. (G) 0.323 0.109 0.408
Medium Oedogonium spp. (G) 0.137 0.031 0.139
Small Oedogonium spp. (G) 0.037 0.157 0.250
Oocystis spp. (G) 0.274 0.200 0.147
Pseudanabaena sp. A (BG) 0.388 0.369 0.005
Pseudanabaena sp. C (BG) 0.102 0.054 0.337
Pseudanabaena sp. D (BG) 0.160 0.482 0.207
Pseudanabaena sp. E (BG) 0.248 0.397 0.644
Rhizoclonium sp.‘ (G) 0.148 0.265 0.353
Rhopalodia gibba (D) 0.149 0.391 0.489
Scenedesmus acutiformis (G) 0.364 0.545 0.300
Scenedesmus obliquus (G) 0.087 0.185 0.058
Stigeoclonium sp. basal cells (G) 0.160 0.343 0.214
Tetraedron minimum (G) 0.168 0,103 0341

 

145

Appendix D4.3: Post-hoc pairwise comparisons in the resource competition bioassay for
A) chlorophyll a (Tukey’s HSD), B) chlorophyll a (Tukey’s HSD) with low light
treatment, C) species composition (perMANOVA pairwise comparisons) with low light
treatment

 

 

 

 

 

 

 

 

 

 

 

A
Factor df Mean difference P-value
Chlorophyll a Pairwise comparisons
Control vs. Low Si 1 0.121 0.219
Control vs. Low N 1 0.659 <0.0001
Control vs. Low P 1 0.134 0.202
Low P vs. Low Si 1 0.013 0.997
Low P vs. Low N 1 0.525 <0.0001
Low N vs. Low Si 1 0.538 <0.0001

B
Factor df Mean difference P-value
Chlorophyll a Pairwise comparisons
Control vs. Low Si 1 0.121 0.801
Control vs. Low light 1 0.956 <0.0001
Control vs. Low N 1 0.659 <0.0001
Control vs. Low P 1 0.134 0.787
Low P vs. Low Si 1 0.013 1.000
Low P vs. Low light 1 0.822 <0.0001
Low P vs. Low N 1 0.525 0.005
Low N vs. Low Si 1 0.538 0.002
Low N vs. Low light 1 0.297 0.147
Low light vs. Lo Si 1 0.835 <0.0001

C
Factor df t-statistic P-\alue
Species composition pairwise comparisons
Control vs. Low Si 1 1.096 0.227
Control vs. Low light 1 1.578 0.029
Control vs. Low N 1 1.430 0.173
Control vs. Low P 1 0.792 0.692
Low P vs. Low Si 1 0.863 0.758
Low P vs. Low light 1 1.359 0.038
Low P vs. Low N 1 1.186 0.244
Low N vs. Low Si 1 1.104 0.245
Low N vs. Low light 1 1.637 0.094
Low light vs. Lo Si 1 1.422 0.018

 

146

   

I[IlliljﬂlallljllljllsllIlllll‘lslll