ECOLOGICAL STUDIES ON SEED DISPERSAL NETWORKS:

INSIGHTS FROM A DIVERSE TROPICAL ECOSYSTEM

A DISSERTATION SUBMITTED TO THE DEPARTMENT OF BIOLOGY AND THE COMMITTEE ON GRADUATE STUDIES

OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

Camila Iotte Donatti December 2011

http://creativecommons.org/licenses/by-nc/3.0/us/

This dissertation is online at: http://purl.stanford.edu/jz498cr4469

© 2011 by Camila Iotte Donatti. All Rights Reserved.

Re-distributed by Stanford University under license with the author.

This work is licensed under a Creative Commons Attribution-Noncommercial 3.0 United States License.

ii

I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.

Rodolfo Dirzo, Primary Adviser

I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.

Carol Boggs

I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.

Fiorenza Micheli

Approved for the Stanford University Committee on Graduate Studies.

Patricia J. Gumport, Vice Provost Graduate Education

This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file inUniversity Archives.

iii

v

ABSTRACT

Mutualisms between animals and plants, such as pollination, seed dispersal and ant-plant protection, are key ecological processes in many ecosystems throughout the world. Like any other ecological interaction, plant-animal mutualisms occur in a community context. Therefore, one-to-one interactions are very rare in nature and the majority of species, both animals and plants, have more than one partner. Recently, studies on mutualistic interactions at the community level have used the “network approach” in order to simplify complex interactions and to determine both the pattern of interaction and the properties of species in networks. In this dissertation, I use network theory combined with long-term field work, phylogenetic and multivariate analysis, species extinctions simulations and experimental manipulation to identify the pattern of interaction in a seed dispersal network, to assess the contribution of particular animal species to network stability and robustness, and to address the extent to which seed dispersal interactions can structure plant communities. To do so, I studied a hyper-diverse seed dispersal network sampled in the Brazilian Pantanal, which includes interactions among plant species from 28 families and seed dispersers, both native and exotic species, from 25 families and 4 taxonomic groups.

In the first chapter I examine the pattern of interaction in this hyper-diverse seed dispersal network and show that this network has a heterogeneous structure, which is organized around a modular pattern. That is, subsets of species (modules) more frequently interact with each other than with species in other modules. I show that plant and animal trait values are associated with specific modules but phylogenetic signal is limited. I conclude that the observed modularity emerges by a combination of phylogenetic history and trait convergence of phylogenetically unrelated species, shaped by interactions with particular types of dispersal agents. Additionally, my results from this chapter highlight the fact that the absence of large and medium-bodied species may affect the pattern of interaction and, as a consequence, the functioning of this seed dispersal network.

In the second chapter I use species extinction simulations to explore how the extinction of large- and medium-bodied species may affect the pattern of interaction in this network and seed dispersal services as a whole. My results show that the removal of large- and medium-bodied species has a large impact on the network pattern and robustness. Although exotic mammalian species usually have negative impacts on native taxa, my results surprisingly show that the exotic feral pig (Sus scrofa) actually plays a critical role in maintaining structural network metrics and in providing seed dispersal services in this community.

In the third chapter, I continue to explore the importance of large-bodied species in providing seed dispersal services in the Pantanal plant-animal community. It is well known that defaunation, the contemporary pulse of animal population loss or decline driven by human activities, can compromise the dispersal of large-seeded plants. This is the case because the especially vulnerable large-bodied seed dispersers are extremely important in ingesting and dispersing large-seeded plant species. However, the results from this chapter emphasize the fact that large-bodied animals are also important because

vi

they disperse such species in high frequencies, and also have the ability to disperse large conspecific seeds, which in turn show high germination rates.

The three first chapters of my dissertation focus primarily on the seed dispersal process per se. However, in chapters 4 and 5 I expand my study to a broader view and address the importance of seed dispersal in determining the structure of plant communities. In chapter 4, I describe the properties of species in this network and, in chapter 5, I use one of these species properties, the “maximum dependence”, to address this association, taking into account the complexities related to the effect of multiple seed dispersers on the spatial distribution of animal-dispersed plants. I found that seed dispersal was the main important predictor of the aggregation intensity of individual plants, in comparison with several other biotic and abiotic variables. Therefore, I conclude that, although different variables, such as seed size and edaphic characteristics, can operate at different scales in shaping the distribution and structure of plant communities, seed dispersal appears to be the most important in that respect, even when considering the effects of multiple animal species in dispersing plant species.

This study contributes novel information on seed dispersal at the community level, especially because I examined a diverse and relatively complete seed dispersal network, which may provide insights for other diverse systems, especially in the tropics. Besides generating information on the ecology and evolution of plant-animal interactions, this study also shows that not all seed disperser species are equal at the community level; and body size of dispersers seem to be a useful proxy of relative importance for dispersal services. Since contemporary defaunation differentially affects species depending on body size, this work illustrates how human activities, such as hunting, land use and climate change, affect not only taxa, but also crucial processes in which animals of different body size play different roles. This study emphasizes that conservation science needs to look at the conservation of ecological processes driven by species interactions.

vii

ACKNOWLEDGMENTS I could not have completed this work without the help and generosity of several people:

My advisor

Dr. Rodolfo Dirzo, who has been an incredible source of support and inspiration since the very first day that I arrived at the lab.

Dear Rodolfo, you have made me a better person and professional and I will be forever indebted to you. Thank you so much for everything!

My committee Dr. Carol Boggs, Dr. Fiorenza Micheli, Dr. Tadashi Fukami and Dr. Lisa Curran

My collaborators Dr. Mauro Galetti, Dr. Paulo Guimarães Jr, Dr. Marco Aurélio Pizo, Dr. Alexine

Keuroghlian, Ellen Wang, Flávia M. D. Marquitti, Marina Schweizer and Lucas Leuzinger

My friends from the Dirzo Lab

Rachel Adams, Eben Broadbent, Posy Busby, Oskar Burger, Yolanda Cachu Pavón, Mauro Galetti, Dennis Hansen, Erin Kuerten, Eduardo Mendoza,

Doug McCauley, Katherine Mertes, Beth Pringle, Chelsea Wood and Hillary Young My friends from the Palumbi Lab

Dr. Steve Palumbi, who generously offered me a space in his lab, Dan Barshis, Pierre De Wit, Alison Haupt, Hannah Jaris, Jason Ladner, Tom Oliver,

Marina Oster, Melissa Pespeni, Carolyn Tepolt and Nina Therkildsen Friends from my cohort

Aaron Carlisle, Posy Busby, Henri Folse, Nishad Jayasundara, Jason Ladner, Kevin Miklasz, Malin Pinsky, Beth Pringle, Julie Stewart and Shelby Sturgis

Stanford University staff

Pamela Hung, Monica Bernal, Dan King, Valeria Kiszka, Jennifer Mason and Matt Pinheiro

Funding

Stanford University Conservation International Zaffaroni Fellowship Fund

The State of São Paulo Research Foundation (FAPESP)

My family Caroline, Gustavo, Alice, Rodrigo, João, Karin, Jennifer, Julia and Tom

My parents Maria José e José Airton

And my husband Jason

viii

TABLE OF CONTENTS LIST OF FIGURES .....................................................................................................XIII LIST OF TABLES ..................................................................................................... XIIII INTRODUCTION............................................................................................................. 1 STATEMENT ON MULTIPLE AUTHORSHIP .......................................................... 6 REFERENCES.................................................................................................................. 7 CHAPTER 1. ANALYSIS OF A HYPER-DIVERSE SEED DISPERSAL NETWORK: MODULARITY AND UNDERLYING MECHANISMS.................... 10

ABSTRACT...................................................................................................................... 10 INTRODUCTION............................................................................................................... 11 MATERIAL AND METHODS .............................................................................................. 13

Study sites.......................................................................................................................... 13 Seed dispersal interactions ............................................................................................... 13 Sampling robustness ......................................................................................................... 14 The network structure ....................................................................................................... 14 Phylogenetic signal in animal and plant traits and in the network’s pattern................... 16 The role of individual species in the network structure……………………………………...17

RESULTS......................................................................................................................... 17 The network structure ....................................................................................................... 17 Modularity......................................................................................................................... 18 Phylogenetic signal in animal and plant traits and in the network’s pattern................... 20 The role of individual species in the network structure.................................................... 20

DISCUSSION.................................................................................................................... 21 ACKNOWLEDGMENTS..................................................................................................... 24 REFERENCES .................................................................................................................. 26 FIGURES ......................................................................................................................... 29 SUPPORTING INFORMATION............................................................................................ 34

CHAPTER 2. DEFAUNATION SIMULATIONS REVEAL THE CONSEQUENCES OF SEED DISPERSAL NETWORK DISRUPTIONS AND THE ROLE OF AN EXOTIC SPECIES ON DISPERSAL SERVICES .................. 37

ABSTRACT...................................................................................................................... 37 INTRODUCTION............................................................................................................... 38 METHODS....................................................................................................................... 40

Study site ........................................................................................................................... 40 Seed dispersal interactions ............................................................................................... 40 Simulation of defaunation steps ........................................................................................ 40 Effects of defaunation on structural network metrics ....................................................... 42 Effects on seed dispersal events........................................................................................ 42 Effects on the robustness of the network........................................................................... 43 The importance of particular native and exotic species on network structure and seed dispersal services .............................................................................................................. 43

RESULTS......................................................................................................................... 44

ix

Effects of defaunation on structural network metrics ....................................................... 44 Effects on seed dispersal events........................................................................................ 45 Effects on the robustness of the network........................................................................... 46 The importance of particular native and exotic species on network structure and seed dispersal services .............................................................................................................. 47

DISCUSSION.................................................................................................................... 47 ACKNOWLEDGMENTS..................................................................................................... 50 REFERENCES .................................................................................................................. 51 FIGURES ......................................................................................................................... 55 SUPPORTING INFORMATION............................................................................................ 57

CHAPTER 3. EFFECTS OF INTRA- AND INTER-SPECIFIC SEED SIZE VARIATION ON SELECTION BY DISPERSERS, GERMINATION AND SEEDLING GROWTH.................................................................................................. 60

ABSTRACT...................................................................................................................... 60 INTRODUCTION............................................................................................................... 61 METHODS....................................................................................................................... 63

Study site ........................................................................................................................... 63 Seed dispersal interactions ............................................................................................... 63 Diameter of the dispersed seeds - seed disperser body mass interspecific associations.. 64 Diameter of the dispersed seeds - seed disperser body mass intraspecific associations.. 64 Effect of seed size and gut-passage on seed germination of Dipteryx alata..................... 65 Effect of seed size and gut passage on seedling survival and growth rates of D. alata in controlled conditions ........................................................................................................ 66 Effect of seed size and gut passage on seedling survival and growth rates of D. alata in field conditions.................................................................................................................. 67

RESULTS......................................................................................................................... 68 Diameter of the dispersed seeds - seed disperser body mass interspecific associations.. 68 Diameter of the dispersed seeds - seed disperser body mass intraspecific associations.. 68 Effect of seed size and gut passage on seed germination of D. alata ............................... 68 Effect of seed size and gut passage on seedling survival and growth rates of D. alata in controlled conditions ........................................................................................................ 69 Effect of seed size and gut passage on seedling survival and growth rates of D. alata in field conditions.................................................................................................................. 70

DISCUSSION.................................................................................................................... 71 ACKNOWLEDGMENTS..................................................................................................... 73 REFERENCES .................................................................................................................. 74 TABLES .......................................................................................................................... 79 FIGURES ......................................................................................................................... 81

CHAPTER 4. THE ROLE OF PLANT AND ANIMAL TRAITS IN DETERMINING NUMBER AND STRENGTH OF INTERACTIONS ACROSS SPECIES IN A SEED DISPERSAL NETWORK ....................................................... 85 ABSTRACT………………………………………………………………………………85

INTRODUCTION............................................................................................................... 85 METHODS....................................................................................................................... 88

Properties of species in the seed dispersal network ......................................................... 88 Plant and animal species traits......................................................................................... 90

x

Phylogenetic signal in traits and properties of animal and plant species in the network 91 RESULTS......................................................................................................................... 92

Phylogenetic signal in traits and properties of animal and plant species in the network 92 Observed distribution in the properties of species in the network……………………….93 Association between traits and the properties of species in the network ......................... 93

DISCUSSION.................................................................................................................... 95 ACKNOWLEDGMENTS..................................................................................................... 98 REFERENCES .................................................................................................................. 99 TABLES ........................................................................................................................ 103 FIGURES ....................................................................................................................... 107

CHAPTER 5. THE ROLE OF SEED DISPERSAL INTERACTIONS IN STRUCTURING A PLANT COMMUNITY IN THE BRAZILIAN PANTANAL 110

ABSTRACT.................................................................................................................... 110 INTRODUCTION............................................................................................................. 112 METHODS..................................................................................................................... 114

Seed dispersal interactions ............................................................................................. 114 Spatial aggregation of individuals across plant species ................................................ 115 Consequences of spatial aggregation on seedling and sapling mortality, leaf herbivory and infection by pathogens ............................................................................................. 117 Seed dispersal interactions and the structure of the plant community ........................... 117

RESULTS....................................................................................................................... 118 Spatial aggregation of individuals across plant species ................................................ 118 Consequences of spatial aggregation on seedling and sapling mortality, leaf herbivory and infection by pathogens ............................................................................................. 119 Seed dispersal interactions and the structure of the plant community ........................... 120

DISCUSSION.................................................................................................................. 120 ACKNOWLEDGMENTS................................................................................................... 123 REFERENCES ................................................................................................................ 124 FIGURES ....................................................................................................................... 127

xi

LIST OF FIGURES CHAPTER 1

Figure 1. Modularity in the seed dispersal network...................................................29 Figure 2. Association between body mass and values of fruit trait ...........................30 Figure 3a. The phylogenetic tree of animal species and modules in which each species was assigned...................................................................................................31 Figure 3b. The phylogenetic tree of plant species and modules in which each species was assigned ...............................................................................................................32 Figure 4. Role of each species in the seed dispersal network ....................................33 Figure S1. Accumulation curve with the average and the standard deviation of the number of seed dispersal interactions in 1000 iterations, as a function of the number of seed dispersal events .............................................................................................34 Figure S2. Association between K values using branch lengths=1 and K values using branch lengths generated by our simulations……………………………………….36

CHAPTER 2 Figure 1. Values of nestedness and modularity in each defaunation step..................55 Figure 2. The robustness of the network....................................................................56

CHAPTER 3 Figure 1. Relationship between the diameter of dispersed seeds and body mass of the seed disperser .............................................................................................................81 Figure 2. Diameter of the seeds dispersed by different animal species, from the heaviest to the lightest ...............................................................................................82 Figure 3. Number of days from plantation to germination as a function of the diameter of seeds dispersed by tapirs, in controlled conditions ................................83 Figure 4. Seedling growth as a function of the diameter of seeds dispersed by tapirs, in controlled conditions and in field conditions ........................................................84

CHAPTER 4 Figure 1. Association between seed diameter and species degree, maximum dependence and interaction asymmetry of plant species .........................................108 Figure 2. Association between body mass and species degree, maximum dependence and interaction asymmetry of animal species ..........................................................109

CHAPTER 5 Figure 1. Association between the aggregated distribution of individuals and the value of maximum dependence of the plant on its most important seed disperser .127 Figure 2. Logistic regression between the survival of seedlings and saplings and the distance from the closest conspecific individual .....................................................128 Figure 3. Associations between the distance from the closest conspecific adult and: the percentage of the plant with foliar herbivory and the percentage of the plant with foliar attack by pathogens.........................................................................................129

xii

LIST OF TABLES

CHAPTER 2 Table S1. Animal species removed in each of the defaunation steps in the realistic defaunation simulation, family and their body masses ..............................................57 Table S2. Animal species maintained in the network while all other large- and medium-bodied species were removed, and network metrics....................................58

CHAPTER 3 Table 1. Seed disperser species, their body masses and the average diameter of the seeds dispersed by them in a community in the Brazilian Pantanal...........................79 Table 2. Plant species, and their average seed diameter, recorded interacting with seed dispersers in the Brazilian Pantanal....................................................................80

CHAPTER 4 Table 1. Plant species, plant species traits and properties of plant species in the seed dispersal network......................................................................................................103 Table 2. Animal species, animal species traits and properties of animal species in the seed dispersal network..............................................................................................105 Table 3a. Results of multiple regression analyses on the associations between species traits and species properties for plant species ..............................................107 Table 3b. Results of multiple regression analyses on the associations between species traits and species properties for mammal and bird species combined…….107

1

INTRODUCTION

Mutualisms between animals and plants, such as pollination, seed dispersal and ant-

plant protection, pervade nature and are key ecological processes in many ecosystems

throughout the world (Herrera & Pellmyr 2002). Like any other ecological interaction,

plant-animal mutualisms occur in a community context. Therefore, one-to-one

interactions are very rare and the majority of species, both animals and plants, have more

than one partner (Waser et al. 1996, Bascompte et al. 2003, Memmott et al. 2004,

Jordano et al. 2006). The study of those interactions at the community level is crucial for

a basic understanding of the ecology and evolution of plant–animal interactions, and for

management and conservation of biodiversity (Bronstein et al. 2006, Waser & Ollerton

2006, Rico-Gray & Oliveira 2007).

As a mutualistic process, seed dispersal by frugivores is beneficial to both animals

and plants. While the animals gain from ingesting fleshy fruits due to their nutritious

content, the plants gain from being dispersed by animals through the possibility of

exploring and colonizing new or enemy-free habitats, facilitating the regeneration of

plant populations and communities, promoting gene flow and genetic intermingling, and,

ultimately, contributing to the maintenance of plant diversity (Hubbell 1979, Clark et al.

1998, Connell & Green 2000, Ehrlén et al. 2006). Recently, studies on mutualistic

interactions at the community level have used the “network approach” (Memmott 1999,

Strogatz 2001, Dicks et al. 2002, Bascompte et al. 2003) in order to simplify such

complex interactions and to determine both the pattern of interaction and the properties of

species in networks. Seed dispersal networks have already been relatively well described

in terms of their pattern, but the majority of those analyzed prior to this study are, in fact,

sub-networks that predominantly include interactions between a single taxonomic group

of seed dispersers (e.g., birds) and the plants they disperse (Bascompte et al. 2003).

In this dissertation, I use network theory combined with long-term field work,

phylogenetic and multivariate analysis, species extinctions simulations and experimental

manipulation to identify the pattern of interaction in a seed dispersal network, to assess

the contribution of particular species to network stability and robustness, and to address

2

the extent to which seed dispersal interactions can structure plant communities. To do so,

I studied a hyper-diverse seed dispersal network sampled in the Brazilian Pantanal that

includes interactions among plant species from 28 families, and seed dispersers, both

native and exotic species, from 25 families and 4 taxonomic groups. The Pantanal,

located in central-western Brazil and part of Bolivia and Paraguay, is the world’s largest

freshwater wetland, covering an area of 140,000 km2 (Swarts 2000). Due to the low

human population density and low hunting pressure on native species (Alho & Lacher

1991, Desbiez et al. 2011, but see Harris et al. 2005), the Pantanal holds one of the

highest concentrations of wildlife in the Neotropics (Swarts 2000, Mittermeier et al.

2005). Such concentration of wildlife enabled me to survey seed dispersal interactions for

animals from a variety of taxonomic groups. I carried out my research on two private

properties, Fazenda Rio Negro and Fazenda Barranco Alto, located in one of the most

pristine regions within the Pantanal.

The description of the pattern of interaction in a diverse seed dispersal network is

an important contribution to the study of seed dispersal, given that networks recorded to

date have included only subsets of major frugivore groups, especially birds, and their

interactions with plant species that also share similar traits (Rezende et al. 2007). This is

the case primarily because most diverse communities worldwide have lost at least some

of their vertebrates involved in mutualisms.

In the first chapter I examine, for the first time, the pattern of interaction in a hyper-

diverse seed dispersal network. I show that this network, in addition to being nested –as

is the case of other seed dispersal networks analyzed to date– has also a heterogeneous

structure that is organized around a modular pattern. That is, subsets of species (modules)

more frequently interact with each other than with species in other modules. In this

diverse network, the modular pattern reflects the diversity of taxonomic groups of seed

dispersers and of fruit and seed morphological traits. Furthermore, I show that plant and

animal trait values are associated with specific modules but phylogenetic signal is

limited. I conclude that the observed modularity emerges by a combination of

phylogenetic history and trait convergence of phylogenetically unrelated species, shaped

by interactions with particular types of dispersal agents. Additionally, my results from

this chapter highlight the importance of particular seed dispersers, specifically large-and

3

medium-bodied species, in contributing to the pattern of this network. Whereas large-

bodied species can disperse a large number of plant species and, therefore, link species

within modules, medium-bodied species can link modules together. Therefore, the

absence of large and medium-bodied species may affect the pattern of interaction and, as

a consequence, the functioning of this seed dispersal network.

In fact, large- and medium-bodied species are the most affected by defaunation,

the contemporary pulse of animal local population loss or decline (sensu Dirzo &

Miranda 1991) driven by human activities such as hunting, deforestation and

fragmentation. In the second chapter I use species extinction simulations to explore how

the extinction of these large- and medium-bodied species may affect the pattern of

interaction in this network and seed dispersal services as a whole. As large- and medium-

bodied species are more affected by human activities than small-bodied species (Bodmer

et al. 1997, Cardillo et al. 2005, Cardillo et al. 2006, Peres & Palacios 2007), the

simulations developed in this chapter are ecologically realistic. My results show that the

removal of large- and medium-bodied species generates the lowest network robustness

when compared to simulations that remove species at random and that remove the most-

linked species from the community. Furthermore, the loss of just a few species among the

largest ones (i.e. species >20 kg) seems to lead to a significant decline of the proportion

of the number of seed dispersal events in the community. Another highlight of this study

is that, although exotic mammalian species usually have negative impacts on native taxa

(Cox 1999, D’Antonio et al. 1999, Cushamn et al. 2004, Busby et al. 2010), my results

surprisingly show that the exotic feral pig (Sus scrofa), actually plays a critical role in

maintaining structural network metrics and in providing seed dispersal services in this

community. Therefore, I posit that network stability will be significantly reduced with

defaunation of large- and medium-bodied species, but this effect can be attenuated by the

presence of large megafaunal exotic species, if such species are less susceptible to

extinction in their novel ecosystems.

In the third chapter, I continue to explore the importance of large-bodied species

in providing seed dispersal services in the Pantanal plant-animal community. It is well

known that defaunation can compromise the dispersal of large-seeded plants (Silva &

Tabarelli 2000, Wright 2003, Galetti et al. 2006, Cramer et al. 2007) and this is the case

4

because the especially vulnerable large-bodied seed dispersers are extremely important in

ingesting and dispersing large-seeded species (Janzen & Martin 1982, Wheelwright 1985,

Chapman et al. 1992, Guimarães et al. 2008, Donatti et al. 2011). However, my data

show that large-bodied animals disperse those large-seeded species with high frequencies

and have the ability to disperse large conspecific seeds, which in turn show high

germination rates. Furthermore, using a set of experiments, I report that, within plant

species, seed germination increases with seed diameter and that the combination of a

large diameter and gut-passage increases seedling growth, in both controlled and field

conditions.

The three first chapters of my dissertation focus primarily on the seed dispersal

process per se. However, in chapters 4 and 5 I expand my study to a broader view and

address the importance of seed dispersal in determining the structure of plant

communities. Even though many processes occur between seed dispersal and seedling

establishment, such as seed and seedling predation, it is possible to find a “dispersal

signal” – an association between patterns of seed dispersal and seedling distribution. In a

large area of 50-100 hectares, spatial distributions of seedlings in plant species dispersed

by animals are significantly different than those in plants dispersed by gravity or wind

(Hubbell 1979, Kinnaird 1998, Hardy & Sonké 2004, Seidler & Plotkin 2006, Russo et

al. 2007, Muller-Landau et al. 2008). In a small area of few square meters, the spatial

distribution of seedlings can be assigned to the seed dispersal pattern produced by

particular animal species (Howe 1989, Herrera et al. 1994, Julliot 1997, Wenny & Levey

1998, Fragoso et al. 2003). However, at the community-level, one-to-one interactions are

rare and the majority of species, both animals and plants, have more than one partner

(Bakker et al. 1996, Bascompte et al. 2003, Memmott et al. 2004, Strauss & Irwin 2004,

Jordano et al. 2006). Thus, the importance of seed dispersal in determining the spatial

distribution of plant species, taking into account that the majority of them interact with

multiple seed dispersers, has received little attention. Thus, in chapter 4 I describe the

properties of species in this network, such as the number of partners of each plant

species, and, in chapter 5, I use one of these species properties, the “maximum

dependence” to address this association, taking into account the complexities related to

the effect of multiple seed dispersers on the spatial distribution of animal-dispersed

5

plants. More specifically, in chapter 4, I describe the variation in the properties of species

in networks and explore, in depth, the importance of evolutionary history and species-

specific traits in determining these properties. I show that morphological traits of plants

and animals are the most important predictors explaining the majority of properties of

species in the network. For example, a plant species with large seeds has a high

dependence on its most frequent seed disperser species, whereas a plant species with

small seeds has a low dependence on its most frequent seed disperser. Due to constraints

related to the gape size of seed dispersers and the fruit or seed size, large-seeded species

are predominantly dispersed by large-bodied frugivores (Janzen & Martin 1982, Janson

1983, Wheelwright 1985, Chapman et al. 1992, Guimarães et al. 2008, Donatti et al.

2011) that can swallow large seeds, or by medium-bodied frugivores (such as agoutis,

Dasyprocta azarae) that can manipulate and carry those seeds. Such dispersers usually

disperse seeds in clumps (Howe 1989, Fragoso et al. 2003), either close to or far from the

mother plant, possibly generating a highly aggregated distribution of individuals. Then, in

chapter 5, I combine the information on the number and the frequency of interactions

between particular plants and seed dispersers to place plant species in a “maximum

dependence” gradient. This gradient includes, at one extreme, plant species that strongly

interact with one seed disperser and, at the other, those that weakly interact with multiple

dispersers. I selected eight plant species that were located along this gradient and mapped

all conspecific individuals in a 2.6-ha plot. As expected, I found a significant and positive

association between the values of maximum dependence of plant species and the degree

of aggregation of conspecific individuals. I also found that seed dispersal was the main

important predictor of the aggregation intensity of individuals, in comparison with

several other biotic and abiotic variables. Therefore, I conclude that, although different

variables such as seed size and edaphic characteristics can operate at different scales in

shaping the distribution and structure of plant communities, seed dispersal shows to be

important in that respect, even when considering the effects of multiple animal species in

dispersing plant species. Furthermore, I suggest that, as large- and medium-bodied seed

disperser species are highly vulnerable to land use change and hunting, the strongly

aggregated distribution of certain plant species can be intensified in a scenario of

defaunation, due to increased seedling mortality, and can, perhaps, even lead to a strong

6

fine-scale spatial genetic structuring in the populations of those plant species.

In contrast to previous work, here I used a community approach to understand

seed dispersal interactions. Therefore, this study contributes novel information on seed

dispersal at the community level, especially because we examined a diverse and

relatively complete seed dispersal network, which may provide insights for other diverse

systems, especially in the tropics. Besides generating information on the ecology and

evolution of plant-animal interactions, this study shows that not all seed disperser species

are equal at the community level; body size of dispersers seems to be a useful proxy of

relative importance for dispersal services. Since contemporary defaunation differentially

affects species depending on body size, this work illustrates how human activities, such

as hunting, land use and climate change, affect not only taxa, but also crucial processes in

which animals of different body size play different roles. This study emphasizes that

conservation science needs to look at the conservation of ecological processes driven by

species interactions.

STATEMENT ON MULTIPLE AUTHORSHIP

I am the first author and primary contributor to each of my dissertation chapters,

including the design, data collection, analysis, and writing. In the paragraphs below I

explain the role of each co-author in each chapter.

Chapter 1 is published in the journal Ecology Letters (14: 773–781, 2011). Paulo R.

Guimarães, Mauro Galetti, Rodolfo Dirzo and I designed research; Mauro Galetti, Marco

Aurélio Pizo and I performed research; Paulo R. Guimarães, Flávia M. D. Marquitti and I

analyzed data; and Paulo R. Guimarães, Mauro Galetti, Marco Aurélio Pizo, Rodolfo

Dirzo and I wrote the paper.

In chapter 2, Paulo R. Guimarães, Mauro Galetti, Rodolfo Dirzo and I designed

research; Mauro Galetti, Marco Aurélio Pizo and I performed research; Paulo R.

Guimarães and I analyzed data; and Paulo R. Guimarães, Mauro Galetti, Marco Aurélio

Pizo, Rodolfo Dirzo and I wrote the paper.

In chapter 3, Mauro Galetti, Rodolfo Dirzo and I designed research; Mauro Galetti,

7

Marco Aurélio Pizo and I performed research; I analyzed data; and Marco Aurélio Pizo,

Mauro Galetti, Rodolfo Dirzo and I wrote the paper.

In chapter 4, Rodolfo Dirzo and I designed research; Mauro Galetti, Marco Aurélio

Pizo and I performed research; I analyzed data; and Marco Aurélio Pizo, Mauro Galetti,

Rodolfo Dirzo and I wrote the paper.

In chapter 5, Rodolfo Dirzo, Mauro Galetti and I designed research; I performed

research; I analyzed data; and Mauro Galetti, Rodolfo Dirzo and I wrote the paper.

References

Alho CJR & Lacher Jr TE (1991) in Latin American mammalogy: history, biodiversity and conservation, eds Mares MA, Schmidly DJ, University of Oklahoma Press, Norman, pp. 280–294.

Bakker JP et al. (1996) Seed banks and seed dispersal: important topics in restoration ecology. Acta Bot. Neerl. 45: 461-490.

Bascompte J, Jordano P. Melián CJ, Olesen JM (2003) The nested assembly of plant-animal mutualistic networks. Proc Natl Acad Sci USA 100 (16): 9383-9387.

Bodmer RE, Eisenberg JF, Redford KH (1997) Hunting and the likelihood of extinction of Amazonian mammals. Conserv Biol 11: 460–466.

Bronstein JL, Alarcón R, Geber M (2006) The evolution of plant-insect mutualisms. New Phytologist 172: 412-428.

Busby PE, Vitousek P, Dirzo R (2010) The prevalence of tree regeneration by sprouting and seeding along a rainfall gradient in Hawaii. Biotropica 42(1): 80-86.

Cardillo M, Mace GM, Gittleman JL, Purvis A (2006) Latent extinction risk and the future battlegrounds of mammal conservation. Proc Natl Acad Sci USA 103(11): 4157-4161.

Cardillo M, Mace GM, Jones KE, Bielby J, Bininda-Emonds ORP, Sechrest W, Orme C DL, Purvis A (2005) Multiple causes of high extinction risk in large mammals species. Science 309 (5738): 1239-1241.

Chapman LJ, Chapman CA, Wrangham RW (1992) Balanites wilsoniana: Elephant Dependent Dispersal? J Trop Ecol 8 (3): 275-283. 1992.

Clark JS, Macklin E, Wood L (1998) Stages and spatial scales of recruitment limitation in southern appalachian forests. Ecol Monographs 68 (2): 213-235.

Connell JH, Green PT (2000) Seedling dynamics over thirty-two years in a tropical rain Forest tree. Ecology 8(2): 568-584.

Cox GW (1999) Alien species in North America and Hawaii: impacts on natural ecosystems. Island Press, Washington, DC, USA.

Cramer JM, Mesquita R, Williamson GB (2007) Forest fragmentation differentially affects seed dispersal of large and small-seeded tropical trees. Biol Cons 137:415–423.

Cushman JH, Tierney TA, Hinds, JM (2004) Variable effects of feral pig disturbances on native and exotic plants in a California grassland. Ecol App 14(6): 1746-4756.

D’Antonio, CM, Dudley TL, Mack M (1999) in Ecosystems of disturbed ground, ed

8

Walker LR (Elsevier, New York, USA), pp 413– 452 Desbiez ALJ, Keuroghlian A, Piovezan U, Bodmer RE (2011) Invasive species and

bushmeat hunting contributing to wildlife conservation: the case of feral pigs in a Neotropical wetland. Oryx 45(1): 78-83.

Dicks, L, Corbet SA, Pywell RF (2002) Compartmentalization in plant-insect flower visitor webs. Journal of animal ecology 71: 32-43.

Dirzo R, Miranda A (1991) in Plant-animal interactions: evolutionary ecology in tropical and temperate regions, eds Price PW, Lewinsohn TM, Fernandes GW and Benson WW, John Wiley and Sons, New York , USA, pp 273–287.

Donatti CI, Guimarães PR, Galetti M, Pizo MA, Marquitti FMD, Dirzo R (2011) Analysis of a hyper-diverse seed dispersal network: modularity and underlying mechanisms. Ecol Lett 14: 773-781.

Ehrlén J, Munzbergova Z, Diekmann M, Eriksson O (2006) Long-term assessment of seed limitation in plants, results from an 11-year experiment. J Ecol 94: 1224-1232.

Fragoso JMV, Silvius KM, Correa JA (2003) Long-distance seed dispersal by tapirs increases seed survival and aggregates tropical trees. Ecology 84(8): 1998-2006.

Galetti M, Donatti CI, Pires AS, Guimarães PR, Jordano P (2006) Seed survival and dispersal of an endemic Atlantic forest palm: the combined effects of defaunation and forest fragmentation. Bot J Linnean Soc 151: 141–149.

Guimarães PR, Galetti M, Jordano P (2008). Seed dispersal anachronisms: rethinking the fruit extinct megafauna ate. PLOS One 3: e1745.

Hardy OJ, Sonké B (2004) Spatial pattern analysis of tree species distribution in a tropical rain Forest of Cameroon: assessing the role of limited dispersal and niche differentiation. Forest Ecol Managraphs 197: 191-202.

Harris, M.B., Tomás, W.M., Mourão, G., Da Silva, C.J., Guimarães, E., Sonoda, F. & Fachim, E. (2005) Safeguarding the Panatanal wetlands: threats and conservation initiatives. Conserv. Biol., 19, 714-720.

Herrera CM, Pellmyr O (2002). Plant-animal interactions: an evolutionary approach, 313 p, Blackwell Science Ltda.

Herrera CM, Jordano P, Lopez-Soria L, Amat JA (1994). Recruitment of a mast-fruiting, bird-dispersed tree: bridging frugivore activity and seedling establishment. Ecol Monographs 64:315–344.

Howe HF (1989) Scatter- and clump-dispersal and seedling demography: hypothesis and implications. Oecologia 79: 417-426.

Hubbell SP (1979) Tree dispersion, abundance, and diversity in a tropical dry Forest. Science 203: 1299-1309.

Memmott J, Waser NM, Price MV (2004) Tolerance of pollination networks to species extinctions. Proc R Soc Lond [Biol] 271(1557): 2605-2611.

Memmott J (1999) The structure of a plant-pollinator food web. Ecol Lett 2: 276-280. Mittermeier RA, Harris MB, Mittermeier CG, Da Silva JMC, Lourival R, Da Fonseca

GAB, Seligmann P, Allofs T (2005) Pantanal: South America's Wetland Jewel, 176 p. Firefly Books Ltd. New York. Muller-Landau HC, Wright SJ, Calderon O, Condit R, Hubbell SP (2008) Interspecific variation in primary seed dispersal in a tropical forest. Journal of Ecology 96(4): 653-667.

9

Muller-Landau HC, Wright SJ, Calderon O, Condit R, Hubbell SP (2008) Interspecific variation in primary seed dispersal in a tropical forest. Journal of Ecology 96(4): 653-667.

Janson CH (1983) Adaptation of fruit morphology to dispersal agents in a neotropical forest. Science 219: 187–189.

Janzen DH, Martin PS (1982) Neotropical anachronisms-the fruits the gomphoteres ate. Science 215: 19-27.

Jordano P, Bascompte J, Olesen JM (2006) Plant-pollinator interactions: from specialization to generalization, eds Waser NM, Olerton J (The Uniersity of Chicago Press, London, UK), pp 173-199.

Julliot C (1997) Impact of seed dispersal by the red howler monkeys Alouatta seniculus on the seedling population in the understory of tropical rain forest. J Ecol 85: 431-440.

Kinnaird MF (1998) Evidence for effective seed dispersal by the Sulawesi red-knobbed hornbill, Aceros cassidix. Biotropica 30(1): 50-55.

Peres CA, Palácios E (2007) Basin-Wide Effects of Game Harvest on Vertebrate Population Densities in Amazonian Forests: Implications for Animal-Mediated Seed Dispersal. Biotropica 39: 304-315.

Rico-Gray V, Oliveira PS (2007) The Ecology and Evolution of Ant-Plant Interactions. The University of Chicago Press, Chicago. 331 p.

Rezende E, Lavabre J, Guimarães PR, Bascompte J (2007) Non-random coextinctions in phylogenetically structured mutualistic networks. Nature 448: 925-928.

Russo SE, Potts MD, Davies S.J., Tan, S (2007) in Seed dispersal: theory and its application in a changing world, eds Dennis AJ, Schupp EW, Green RJ, Westcott DA, CAB International, Wallinford, UK, pp 499-518.

Silva JMC, Tabarelli M (2000) Tree species impoverishment and the future flora of the Atlantic forest of northeast Brazil. Nature 404:72–74.

Strauss SY, Irwin RE (2004) Ecological and evolutionary consequences of multispecies plant-animal interactions. Annu Rev of Ecol Evol Syst 35:435-466.

Swarts FA (2000) in The Pantanal of Brazil, Paraguay and Bolivia, ed Swartz FA Hudson MacArthur Publishers, Gouldsboro, Pennsylvania, pp 1-24.

Siedler TG, Plotkin JB (2006) Seed dispersal and spatial pattern in tropical trees. Seed dispersal and spatial pattern in tropical tree. Plos Biology 4(11): 2132-2138.

Strogatz, S.H. 2001. Exploring complex networks. Nature 410: 268-276. Waser NM, Ollerton J (2006) Plant–pollinator interactions, from specialization to

generalization. University of Chicago Press, Chicago, Ill. Waser N, Price MV, Williams NM, Ollerton J (1996) Generalization in pollination

systems and why it matters. Ecology 77: 1043-1060. Wenny DG, Levey DL (1998) Directed seed dispersal by bellbirds in a tropical cloud

forest. Proc Natl Acad Sci USA 95 (11): 6204-6207. Wheelwright NT (1985) Fruit size, gape width, and the diets of fruit-eating birds.

Ecology 66: 808-818. Wright SJ (2003) The myriad consequences of hunting for vertebrates and plants in

tropical forests. Perspect Plant Ecol 6: 73-86.

10

CHAPTER 1

ANALYSIS OF A HYPER-DIVERSE SEED DISPERSAL NETWORK:

MODULARITY AND UNDERLYING MECHANISMS Camila I. Donatti, Paulo R. Guimarães, Mauro Galetti, Marco Aurélio Pizo, Flávia M. D. Marquitti & Rodolfo Dirzo

Abstract Mutualistic interactions involving pollination and ant-plant mutualistic networks

typically feature tightly linked species grouped in modules. However, such modularity is

infrequent in seed dispersal networks, presumably because research on those networks

predominantly includes a single taxonomic animal group (e.g. birds). Here, for the first

time, we examine the pattern of interaction in a network that includes multiple taxonomic

groups of seed dispersers, and the mechanisms underlying modularity. We found that the

network was nested and modular, with five distinguishable modules. Our examination of

mechanisms underlying such modularity showed that plant and animal trait values were

associated with specific modules but phylogenetic effect was limited. Thus, the pattern of

interaction in this network is only partially explained by shared evolutionary history. We

conclude that the observed modularity emerged by a combination of phylogenetic history

and trait convergence of phylogenetically unrelated species, shaped by interactions with

particular types of dispersal agents.

11

Introduction

In mutualistic interactions, species commonly interact with multiple partners,

forming a network of interactions. The pattern of these interactions in a community, i.e.,

the way interactions are organized, can be described using a network approach, which

helps to elucidate the complexity of such interactions (Jordano 1987; Bascompte &

Jordano 2007). Mutualistic networks are highly structured, with a prevalence of a nested

pattern (Bascompte et al. 2003; Vázquez et al. 2009; Fortuna et al. 2010; Joppa et al.

2010). That is, the interactions of the specialist species tend to be a subset of the

interactions observed among the generalists (Bascompte et al. 2003). In addition to being

nested, some mutualistic networks are also modular (Dicks et al. 2002 and Olesen et al.

2007: pollination networks; Fonseca & Ganade 1996 and Guimarães et al. 2007: ant-

plant networks), whereby subsets of species (modules) more frequently interact with each

other than with species in other modules (Olesen et al. 2007).

Among mutualisms, modularity has been investigated in depth in pollination

networks (Dicks et al. 2002; Olesen et al. 2007; Fortuna et al. 2010), which often include

a highly diverse array of animal and plant taxa (e.g. Rezende et al. 2007). In contrast, the

majority of seed dispersal networks studied includes mainly seed-dispersing birds, which

interact with plant species that share similar traits (Rezende et al. 2007), leading to a

highly nested and low modularity pattern of interaction (see Fortuna et al. 2010). The

widespread habit of producing fleshy fruits among tropical plant species has been

evolutionary associated to the diversification of frugivorous vertebrates (Fleming et al.

1987). Therefore, the diversity of animals that interact with a particular plant species

could make them tightly linked within modules. Thus, one can predict networks of

interactions in diverse communities, involving plants and several taxonomic groups of

seed-dispersing animals, to have low nestedness and high modularity.

Here we test if a hyper-diverse seed dispersal network is characterized by low

nestedness and high modularity. Beyond such test, we examine the mechanisms

organizing this network through a combination of long-term fieldwork, network theory

and phylogenetic analysis. We analyze the structure of plant-animal interactions in one of

the world’s last remaining species-rich communities involving large vertebrates: the

Pantanal (Harris et al. 2005). By investigating the structure of this community of plants

12

and seed dispersers, we are filling an important gap in studies of species networks, given

that most similarly diverse communities worldwide have lost at least some of their

vertebrates involved in mutualisms and include only subsets of major frugivore groups

(but see Gautier-Hion et al. 1985). To the extent that modular patterns reflect a more

diversified network of ecological functions and services, the understanding of

mechanisms that determine modularity could help uncovering general processes shaping

the evolutionary ecology of plant-animal interactions.

Modularity in a broad range of ecological networks is associated with habitat

heterogeneity (Pimm & Lawton 1980), phylogenetic clustering of closely related species

(Lewinsohn et al. 2006), convergence toward syndromes (Corbet 2000), and

combinations of these factors (Cattin et al. 2004; Olesen et al. 2007; Rezende et al.

2009). For mutualistic networks, a combination of coevolutionary complementarity and

convergence appears to draw other species into the interaction over time, creating a

coevolutionary vortex (Thompson 2005) reflected in the structure of the network.

Here we i) examine the pattern of interaction in a highly diverse seed dispersal

network, including a variety of species from major taxonomic groups of seed dispersers,

mammals, birds, fish and reptiles, and the fleshy-fruited species they disperse, and ii) test

current hypotheses on mechanisms that may generate the modularity in mutualistic

networks. Interactions in this network were sampled in three habitats within a community

in the Brazilian Pantanal. We tested the following hypotheses: 1) this network should be

modular given the diversity of taxonomic groups of seed dispersers involved, 2) animal

and fruit traits, as well as phylogeny should be, therefore, associated to the modularity of

this network, and 3) the habitat types where interactions were recorded should not be

associated to modules, given that, although several plant species in this community are

habitat specialists, the majority of animal species are not.

We first illustrate that interactions in this network have a combination of nested

and modular patterns. Then, we describe how modules are predominantly associated with

fruit and animal traits, and not with the different habitat types where plant species

predominantly occur. We show that the modularity in this network is only partially

explained by shared evolutionary history because, although modules are related to the

different taxonomic groups of animals, phylogeny explains only the assemblage of

13

species in modules associated to birds. We conclude that such modularity likely emerged

by a combination of shared phylogenetic history and trait convergence of

phylogenetically unrelated species, shaped by interactions with particular types of

dispersal agents.

Material and methods

Study sites

This study centered on two neighboring locations in the Brazilian Pantanal: Rio

Negro (19°34'15"S 56°14'43"W) and Barranco Alto farms (19º34'40"S 56º09'08"W),

covering 7,500 ha and 11,000 ha, respectively, of private land (see Appendix S1 in

Supporting Information). The main vegetation types in these locations, where seed

dispersal interactions were recorded, include gallery forests, savannas and semi-

deciduous forests (Prance & Schaller 1982).

Seed dispersal interactions

Seed dispersal interactions were recorded using four methodologies. To sample

seed dispersal by birds, we carried out focal observations at 14 plant species during 882

hours, recording identity of birds that were unequivocally observed carrying fruits outside

the canopy area or swallowed them in situ. Seed dispersal by red-footed tortoises

(Geochelone carbonaria), rheas (Rhea americana) and the majority of mammal species,

were recorded with camera traps located beneath fruiting trees of 27 plant species,

capturing events of fruit ingestion, for a total of 14,800 hours. Some terrestrial and semi-

terrestrial bird species were also recorded via camera traps. We analyzed 716 scats of

several species of mammals, rheas and red-footed tortoises, and identified the intact seeds

in them. To record seed dispersal by the pacu fish (Piaractus mesopotamicus), we caught

80 individuals and identified the intact seeds in their intestine (see Galetti et al. 2008).

One event of seed dispersal was considered as such when either: fruits were recorded to

have been swallowed or removed from a plant species during focal observations; fruit

removal of a particular species by a potential seed disperser was detected with camera

14

traps; a scat pile was found to have at least one intact seed of a particular species in it; or

a sampled fish intestine contained at least one intact seed from a particular species.

Sampling robustness

To assess if we had recorded enough interactions to describe this network, we

generated an accumulation curve with the number of interactions as a function of the

number of seed dispersal events sampled (Guimarães et al. 2007, Jordano et al. 2009).

We estimated average and standard deviation of the number of interactions for any given

fraction of the number of events recorded in 1,000 iterations. After generating this curve,

we used the drc package in R (http://www.r-project.org/) and the dose-response model to

extrapolate the curve. We then used the Michaelis-Menten equation to assess the

asymptotic value of the curve.

The network structure

To define the pattern of interaction in the network, we organized a qualitative seed

dispersal matrix, collectively using the methodologies previously described. In a matrix

of plants in columns and animals in rows, an element representing a seed dispersal

interaction received the value of one, and zero otherwise. We then used this matrix to test

for nestedness and modularity.

Nestedness was analyzed using the NODF metric (Almeida-Neto et al. 2008),

through the ANINHADO program (Guimarães & Guimarães 2006). To test if the

network is more nested than expected by species richness and heterogeneity of

interactions, we compared the recorded NODF value to that of 1,000 random matrices

generated by a null model that controls for the number of interactions per species in the

network (‘null model 2’, Bascompte et al. 2003).

To detect modularity we used the NETCARTO program and an algorithm based

on simulated annealing, SA (Guimerà & Amaral 2005) that identifies modules formed by

both plants and animals simultaneously (see Olesen et al. 2007). We computed the

network modularity index M, which measures the degree to which the network is

organized into clearly defined modules, as well as the level of significance of the

modularity in this network by comparing its M to that of random networks of similar

15

sizes, generated by the same null model used for the nestedness analysis. As the

algorithm is based on an optimization process, the outcomes may vary in different runs.

Therefore, to assign each species to a particular module, we ran the analysis 50 times.

Species that were assigned to a particular module in > 90% of the runs were included in

that module. We compared the nestedness and the modularity of this network with those

same parameters calculated for other 24 frugivory networks available in the literature.

We associated each plant species to the habitat type (gallery forest, savanna, semi-

deciduous forest) in which it was predominantly recorded in a previous 4-year

phenological study (C.I. Donatti, unpublished work). We used binomial distributions to

test the associations of animal taxonomic groups to particular modules, i.e. if phylogeny

explains the coarse network’s topology, and the associations of plant species that

predominantly occur in each habitat type to particular modules, i.e. if habitat type

explains the coarse topology of the network. We tested if the presence of species from the

same taxonomic group or from the same habitat type is over-represented in a particular

module when compared with the null expectation that taxonomy or habitat heterogeneity

does not affect the organization of modules. For each module, we estimated the

probability of getting, by chance alone, a number of species from the same taxonomic

group or habitat type equal or higher than that observed in the real network. We used the

number of species in a module (number of trials, N), the proportion of species of a given

taxonomic group or habitat type in the whole sample (probability of success), and the

number of species that belongs to a particular taxonomic group or habitat type that also

belongs to a particular module (number of successes), as parameters of binomial

distributions.

We gathered animal body mass information from the literature, and measured

fruit and seed traits (length, diameter and mass) for all plant species, in at least 30 fruits

and seeds from at least five individuals. Values of body mass were log transformed and

values of plant traits were Box-Cox transformed using JMP v.5.0 (SAS Institute Inc.).

We used ANOVA to compare body mass of animal species among modules, MANOVA

to compare all fruit and seed traits among modules, and t-tests to compare body mass

between modules that represented the same taxonomic animal group.

16

Phylogenetic signal in animal and plant traits and in the network’s pattern

We tested whether animal and plant traits had a significant phylogenetic signal,

i.e. a quantitative measure of the degree to which phylogeny predicts the ecological

similarity of species. To build the animal phylogenetic tree, we followed Bininda-

Emonds et al. (2007) for relationships among mammal species and Hackett et al. (2008)

for relationships among bird species. In addition, we used published work to resolve

relationships within Cracidae (Pereira et al. 2002), Tyrannidae (Tello et al. 2009), and

Thraupidae (Klicka et al. 2007). We also used two mitochondrial DNA sequences

(Cytochrome b and Cytochrome oxidase subunit 1), available on GenBank, to resolve the

relationships between and within the Thraupidae and Icteridae. We then generated a

phylogenetic tree for those sequences using the program Méthodes et algorithmes pour la

bio-informatique (http://www.phylogeny.fr/) and added those relationships in the tree.

The plant phylogenetic tree was built using Phylomatic software

(http://www.phylodiversity.net/phylomatic/phylomatic.html). Relationships within

Fabaceae followed Wojciechowski et al. (2004), within Rubiaceae followed Bremer &

Eriksson (2009) and within Arecaceae followed Asmussen et al. (2006). Since all

branches in animal and plant trees were set equal to one, we conducted simulations that

showed that using branch lengths equal to one is a conservative approach when values of

K are lower than or equal to one (see Appendix S2 in Supporting Information).

We assessed K statistic to measure the phylogenetic signal in animal and plant

traits, using the function phylosignal in the picante package (Kembel et al. 2010) of R.

To assess phylogenetic signal in body mass, we analyzed mammal and bird species

independently. The K statistic compares the observed signal in a trait to the signal under a

Brownian motion model of trait evolution on a phylogeny (Blomberg et al. 2003). The

statistical significance of phylogenetic signal is evaluated by comparing observed

patterns of the variance of independent contrasts of a trait to a null model of shuffling

taxa labels across the tips of the phylogenetic tree. To test for evidence of phylogenetic

signal in modules, i.e. if phylogeny explains the composition of species within modules,

we used a function ad hoc in R that corresponds to the "Fixed Tree, Character Randomly

Reshuffled" model proposed in Maddison & Slatkin (1991). This function counts the

minimum number of transitions needed to get the distribution of modules observed in the

17

real network, randomizes the modules in the phylogeny and then counts the number of

transitions in each randomization. The statistical significance of phylogenetic signal is

achieved if there are fewer transitions in the real network than in 95% of the

randomizations. Phylogenetic signal in modules was tested using the animal and the plant

phylogenetic trees independently. For animals, we run the analyses separately for

mammals and birds.

The role of individual species in the network structure

The SA algorithm also assigns an ecological role to each species in the network

based on its interactions within modules (z) and on its interactions among modules (c)

(Olesen et al. 2007). Species with low z and low c are considered peripheral species, i.e.

they usually interact with species within their own module. Species with either a high z or

c were considered generalists, and either i) module hubs, i.e. highly connected within

their own module (high z and low c), or ii) connectors, those species that link modules

(low z and high c). Species with a high z and a high c were considered supergeneralists,

acting as both module hubs and connectors. To define the role of each species, we used

the most common values of z and c generated in the 50 times we run the analysis. We

used the values of 2.5 for z and of 0.62 for c to define those categories, following Olesen

et al. (2007).

We performed additional analyses using z and c values in order to assess the

correlates of these values with species traits. We analyzed the values of z and c of each

animal species as a function of its body mass and the values of z and c of each plant

species as a function of its fruit and seed traits using correlations. We compared the c and

z values between animal and plant species using t-tests and the c values among modules

using ANOVA.

Results

The network structure

The network included 46 plant species and 46 animal species. We recorded 2,070

seed dispersal events and 273 seed dispersal interactions. One plant species (Sapindus

saponaria) did not show interactions with seed dispersers, probably due to the high level

18

of saponins in the pulp (Pott & Pott 1994). Using the Michaelis-Menten equation we

estimated to have sampled 94.5% of the seed dispersal interactions occurring in this

community (Figure S1 in Supporting Information). Therefore, we assume that the

network described here is robust to additional sampling.

The Pantanal seed dispersal network was not only significantly nested

(NODF=26.27, expected NODF=18.04, p<0.001) but also significantly modular

(M=0.422, expected=0.341, p<0.001), with animal and plant species grouped in five

statistically different modules (Fig. 1). Module composition was very robust: we detected

five modules in all 50 runs. All but two species were assigned to the same module in

100% of the 50 runs. These species, the bird Crax fasciolata (Cracidae) and the plant

Guazuma ulmifolia (Sterculiaceae), were assigned to the same module in 94% of the

runs.

Modularity

Two modules were exclusively represented by bird species and the plant species

they interact with (hereafter bird module 1 and bird module 2; green and blue in Fig. 1,

respectively). Two other modules were represented mainly by mammal species and by

the plant species they interact with. One of these (hereafter mammal-dominated module

1; red in Fig. 1) also included the tortoise and the rhea, whereas the other one (hereafter

mammal-dominated module 2; yellow in Fig. 1) also included a ground-foraging bird

(Crax fasciolata). The fifth module (hereafter fish module; purple in Fig. 1) was

represented by the fish and plant species it mainly interacts with. The Pantanal seed

dispersal network is the second less nested and more modular of the seed dispersal

networks so far studied that show a significant pattern (NODF =54.04±17.15, n=24; M

=0.31±0.092, n=4).

The binomial distributions showed that each module is associated to animal

species that belong to a particular animal taxonomic group (bird module 1: p=0.006,

n=22; bird module 2: p=0.003, n=22; mammal-dominated module 1: p=0.012, n=25;

mammal-dominated module 2: p<0.001, n=18 and fish module: p=0.02, n=4). However,

each module does not include plant species that predominantly occur in a particular

19

habitat type (bird module 1: p>0.666, bird module 2: p>0.437, mammal-dominated

module 1: p>0.179, mammal-dominated module 2: p>0.659, fish module: p>0.168).

Regarding nestedness, the mammal-dominated module 2 showed a significant

nested pattern (NODF=64.73, p=0.005), while the other three modules did not (bird

module 1: NODF=41.20, p=0.169; bird module 2: NODF=62.83, p=0.061 and mammal-

dominated module 1: NODF=60.78, p=0.330). The non-detection of a nested pattern

within modules could be an artifact of the low number of species (Guimarães et al. 2006).

The fish module could not be tested, given its low species richness.

Animal body mass varied across modules (F=64.51, p<0.0001, d.f.=44; mammal-

dominated module 1=47.41kg±32.7, mean±SD, mammal-dominated module

2=15.38kg±5.11, bird module 1=0.24 kg±0.07, bird module 2=0.06kg±0.01) and

explained 82.52% of the variance among modules. Mean animal body mass significantly

differed between the two bird modules (t =2.710, p=0.0129, d.f.=29), but not between the

two mammal-dominated modules (t =0.501, p=0.6240, d.f=14), although the low species

number may cause low statistical power.

Modules also significantly differed when taking into account all fruit and seed

traits (MANOVA F=3.4821, p<0.0001, d.f.=44). Modules were, therefore, characterized

by particular suites of traits. Fruit mass mainly explained the variance among modules

(26.65%), followed by fruit diameter (23.13%), fruit length (17.42%) and seed mass

(11.98%). Mean values of all plant traits but fruit length and seed length in each module

were positively and significantly correlated with mean body mass of seed dispersers in

each module (fruit diameter: F=58.48, p=0.004, r=0.97; fruit mass: F=15.06, p=0.03,

r=0.91; seed diameter: F=26.09, p=0.014, r=0.94; seed mass: F=19.63, p=0.021, r=0.93,

d.f=4): modules with heavy seed dispersers also had heavy and wide fruits and seeds.

When considering all interactions in the network, there were positive associations

between the body size of seed dispersers and all fruit and seed traits (Fig. 2) (fruit length:

F=50.29, p<0.0001, r=0.39, fruit diameter: F=109.4550, p<0.0001, r=0.53; fruit mass:

F=115.95, p<0.0001, r=0.54, d.f.=272; seed length: F=29.9, p<0.0001, r=0.31; seed

diameter: F=41.25, p<0.0001, r=0.36; seed mass: F=54.76, p<0.0001, r=0.41, d.f=269).

20

Phylogenetic signal in animal and plant traits and in the network’s pattern

The body mass of closely related mammal species has exactly the amount of

signal predicted by Brownian motion (K=1.01, p=0.005). In contrast, the body mass of

birds and all seed traits are more divergent than expected under a Brownian model (birds:

K=0.778, p=0.001, d.f.=31; seed length: K=0.526, p=0.001; seed diameter: K=0.417,

p=0.003; seed mass: K=0.496, p=0.001; d.f.=43). Fruit traits did not show a significant

signal (fruit length: K=0.370, p=0.133; fruit diameter: K=0.372, p=0.248; fruit mass:

K=0.392, p=0.159; d.f.=44).

Modules included phylogenetically related bird species (p=0.006), but not

phylogenetically related mammal species (p=0.2) or phylogenetically related plant

species (p=0.83) (Fig.3). Although modules are associated to the major animal taxonomic

groups (mammals, birds or fish), only modules associated to birds included

phylogenetically related bird species.

The role of individual species in the network structure

As in pollination networks, the majority of species in this seed dispersal network

were peripheral, i.e., they almost always interact with species within their own module

(Fig. 4). Three species, the exotic feral pig (Sus scrofa, Suidae, z=2.7869, c=0.5104), the

plant Doliocarpus dentatus (Dilleniaceae, z=2.75, c=0.4897) and the tapir (Tapirus

terrestris, Tapiridae, z=2.4954, c=0.4099), were considered module hubs, i.e. species

with many interactions within their own module. The howler monkey (Alouatta caraya,

Cebidae, z=-1.3687, c=0.6666), one plant (Genipa americana, Rubiaceae, z=1.3574,

c=0.66463) and the chaco chachalaca (Ortalis canicollis, Cracidae, z=0.55, c=0.625)

were considered connectors, i.e. species that had interactions across different modules.

None of the species were defined as supergeneralist, indicating a low cohesiveness in this

seed dispersal network.

Plant and animal species did not differ in both their interactions inside modules

(z) and in their interactions among modules (c) (t=0.351, p=0.726, d.f.=44; t=1.035,

p=0.303, d.f.=45, respectively). However, we found that the mean participation of species

in the whole network (c) differed among modules (F=3.597, p=0.009, d.f.=4; mammal-

21

dominated module 2=0.391±0.05, fish module=0.343±0.11, mammal-dominated module

1=0.326±0.04, bird module 1=0.260±0.04, bird module 2=0.139±0.04). The values of

both z and c increased with the body mass of dispersers (z: F=20.2381, p<0.001, r=0.55;

c: F=45.654, p<0.001, r=0.7; d.f.=45): a large-bodied seed disperser has more

interactions both inside and among modules. In contrast, the values of z decreased with

an increment in all fruit and seed traits (fruit length: F=17.74, p=0.0001, r=0.54; fruit

diameter: F=20.31, p<0.0001, r=0.56; fruit mass: F=26.65, p<0.0001, r=0.61; d.f.=44;

seed length: F=18.81, p<0.0001, r=0.55; seed diameter: F=18.39, p=0.0001, r=0.55, seed

mass: F=25.06, p< 0.0001, r=0.61; d.f.=43) and the values of c decreased with an

increment in seed traits (seed length: F=7.18, p=0.01, r=0.38; seed diameter: F=4.95,

p=0.031, r=0.32; seed mass: F=6.68, p=0.013, r=0.37; d.f.=43): a plant species with small

fruit and seed trait values has more interactions inside modules and a plant species with

small seed trait values has also more interactions among modules. As the body mass of

seed dispersers increases, they are able to disperse a variety of plant species, regardless of

fruit and seed sizes, whereas large fruits and/or seeds are restricted to a few animal

species able to disperse them. These results could reflect the association between plant

and animal species traits and the number of interactions, which increases significantly

with body mass of the seed dispersers and decreases significantly with fruit length and all

seed traits (body mass: F=65.62, p<0.0001, r=0.77, d.f.=45; fruit length: F=4.64,

p=0.036, r=0.31, d.f.=44; seed length: F=13.83, p=0.0006, r=0.49; seed diameter:

F=10.69, p<0.002, r=0.45; seed mass: F=18.95, p<0.0001, r=0.55, d.f=43).

Discussion

Seed dispersal interactions in this network were nested and modular, as in some

pollination and ant-plant mutualistic networks (Fonseca & Ganade 1996, Olesen et al.

2007, Guimarães et al. 2007). The studied network has a heterogeneous structure that is

organized around a modular pattern, which reflects a diversity of taxonomic groups of

seed dispersers and of fruit and seed morphological traits. In essence, the major

taxonomic groups of seed dispersers separated species in five distinct modules. In

addition, seed disperser species within those modules varied in their body mass and

interacted with plant species that differed in fruit and seed traits. In fact, there were

22

strong correlations between the body size of the seed dispersers and the size of fruits

(especially diameter and mass) in modules. Therefore, specific traits of seed dispersers

and fruits help explaining the mixing of major taxa of animals among different modules.

Similarly, Gautier-Hion et al. (1985) studying a diverse community of fruit and

frugivores in an African tropical rainforest found that morphological traits of fruits

revealed syndromes associated with consumption by different taxa of vertebrates.

In a coarse-level of taxonomic resolution, the phylogenetic effect on network

structure was limited to the over-representation of species from a given animal taxonomic

group in each module. In a fine-level of taxonomic resolution, modules associated to

birds showed significant phylogenetic signal, whereas modules associated to mammals

did not. Although the low number of species in mammal-dominated modules could

decrease the statistical power of the phylogenetic test, the mammal composition in

modules does not show clear evidence that phylogenetically related species belong to

same modules. For instance, closely related mammal species, such as artiodactyl

ungulates, belonged to different modules. In addition, modules did not include

phylogenetically related plant species. Therefore, the pattern in this network is only

partially explained by shared evolutionary history in the sense that, although modules are

related to the major taxonomic groups of animals, the majority of them do not include

phylogenetically related species. Given the consistency of our results, we posit that the

modularity of this seed dispersal network is not associated with habitat heterogeneity and

that phylogeny only determines the existence of bird, mammal and fish modules, and the

assemblage of species in bird modules. Therefore, we suggest that modules emerged by a

combination of phylogenetic history and trait convergence of phylogenetically unrelated

species, shaped by interactions with particular types of dispersal agents (Van der Pijl

1982). However, we do not necessarily imply here that those tight phenotypic

associations between seed dispersers and fruits found in modules are driven by

coevolution (e.g., Nuismer et al. 2010). We are simply positing that convergence of

species towards a similar and predictable set of traits (Thompson 2005), rather than

phylogeny alone, is what explains the way this seed dispersal network is organized. In

fact, convergence in these networks might be both an outcome of evolutionary processes

23

such as local adaptation and coadaptation (Thompson 2005) and a consequence of

ecological convergence in resource use by subsets of frugivores.

In a broader context, the Pantanal seed dispersal network was less nested than all

other seed dispersal networks so far studied, with the exception of one sampled in the

Brazilian Atlantic forest (Silva et al. 2007), that includes interactions between plants and

both mammal and bird species. However, the network analyzed here is more modular

than the other four frugivory networks that had significant values of modularity.

Modularity of plant-animal networks is expected to increase with trophic specificity, with

herbivory and ant-plant networks found to be more modular than pollination and seed

dispersal networks (see Fonseca & Ganade 1996; Guimarães et al. 2007; Thebault &

Fontaine 2010). Pollination networks are likely to have higher and more prevalent

modularity than seed dispersal networks because flowers may restrict the range of visitors

through morphological barriers (Santamaría & Rodríguez-Gironés 2007; Stang et al.

2007), whereas fruit traits may lend them to be more open to interaction with multiple

visitors (Blüthgen et al. 2007). However, the high degree of modularity in this diverse

and well-sampled seed dispersal network suggests that factors other than interaction type

are constraining the modularity in frugivory and seed dispersal networks. For instance,

the lack of significant modularity in most of these networks could happen because they

predominantly include a single taxonomic group. Nevertheless, the phylogenetic diversity

only partially explains the modularity of this diverse network and some modules are

composed by species from different taxonomic groups. We therefore hypothesize that

modularity emerges from the interplay between shared evolutionary history and

convergence in patterns of interaction.

One of the values of detecting a modular pattern, other than contributing to

elucidate the evolutionary ecology of plant-frugivore interactions, is the identification of

the role of species in the network (Olesen et al. 2007). This is important because the

robustness of the network, i.e. the ability of a species to persist given the extinction of an

interacting partner in the community (Jordano et al. 2006) may depend on the role of

species in the network. For example, the extinction of connectors may cause the network

to fragment into isolated modules, but will have a minor impact on the internal structure

of modules. In contrast, the extinction of a module hub may cause its module to fragment

24

with minor cascading impact on other modules. Interestingly, only 6.4% of the species in

this network, a lower percentage than that found in pollination networks (Olesen et al.

2007), are connectors or module hubs.

Some plant species were important in maintaining the pattern of this network.

Genipa americana was considered a connector, linking modules together. This plant

species exhibits a typical “megafauna seed dispersal syndrome” (Guimarães et al. 2008)

in that their seed dispersers are/were large mammals, yet its fruits are also avidly eaten by

several bird species. Consequently, fruits of this species were dispersed by animal species

from four modules (two bird modules and two mammal-dominated modules).

Although large-bodied seed dispersers such as tapirs and feral pigs showed to be

important in linking species within a module, they were not important in linking modules

together, maybe because they mainly interact with large- and medium-seeded species,

which were not present in all modules of this network. Medium-bodied species, such as

howler monkeys and chaco chachalacas, were connectors in this network and, therefore,

structurally important because they link modules together. Consequently, we posit that

the loss not only of large-bodied seed dispersers, the ones that disperse a high number of

plant species, but also of some of the medium-bodied species, may change the pattern of

this network, given that their absence could cause the network to fragment into isolated

modules.

Here, the use of the network approach helped us to understand the structure of a

highly diverse seed dispersal network and enabled us to identify the mechanisms that

underlie the modular pattern, contributing to elucidate the ecology and evolution of plant-

frugivore interactions. In addition, the identification of the modular pattern gave us

insights regarding the possible consequences of differential defaunation (cf. Dirzo &

Miranda 1991) on the functioning of this seed dispersal network. For example, the

presence of few animal species that can link modules together could contribute to the

robustness of this network in a scenario of extinction of particular species.

Acknowledgments

We would like to thank Jason Ladner, Enrico Rezende and Andrew Rominger for their

help with the phylogenetic analysis, Roger Guimerà for providing us the modularity

25

algorithm, Rebecca Terry for helping with the accumulation curve, and Dennis Hansen,

Pedro Jordano, John Thompson and three anonymous reviewers for helpful comments on

a previous draft. We thank FAPESP (2004/00810-3 and 2008/10154-7), Earthwatch

Institute and Conservation International for financial support. CID was supported by

Stanford University and PRG by FAPESP and CAPES. We also thank Conservation

International, Lucas Leuzinger and Marina Schweizer for their permission to work in

their properties.

26

References

Almeida-Neto, M., Guimarães, P., Guimarães P.R. Jr., Loyola, R.D. & Ulrich, W. (2008). A consistent metric for nestedness analysis in ecological systems: reconciling concept and measurement. Oikos, 117(8), 1227-1239.

Asmussen, C.B., Dransfield, J., Deichmann, V., Barfod, A.S., Pintaud, J-C. & Baker, W.J. (2006). A new subfamily classification of the palm family (Arecaceae): evidence from plastid DNA phylogeny. Bot. J. Linn. Soc., 151, 15-38.

Bascompte, J., Jordano, P. (2007). The structure of plant-animal mutualistic networks: the architecture of biodiversity Annu. Rev. Ecol. and Syst., 38, 567-593.

Bascompte, J., Jordano, P., Melián, C.J. & Olesen, J.M. (2003). The nested assembly of plant-animal mutualistic networks. PNAS, 100 (16), 9383-9387.

Bininda-Emonds, O.R.P., Cardillo, M., Jones, K.E., McPhee, R.D.E., Beck, R.M.D., Grenyer, R. et al. (2007). The delayed rise of present-day mammals. Nature, 446, 507-512.

Blomberg, S. P., Garland, T. & Ives, A. R. (2003). Testing for phylogenetic signal in comparative data: Behavioral traits are more labile. Evolution, 57, 717-745.

Blüthgen, N., Menzel, F., Hovestadt, T., Fiala, B. & Blüthgen, N. (2007). Specialization, constraints and conflicting interests in mutualistic networks. Curr. Biol., 17, 1–6.

Brewer, B. & Eriksson, T. (2009). Time tree of Rubiaceae: Phylogeny and dating the family, subfamilies, and tribes. Int. J. Plant Sci., 170(6), 766-793.

Cattin, M-F., Bersier, L-F., Banasek-Ritcher, C., Baltensperger R. & Gabriel J-P. (2004). Phylogeny constraints and adaptation explain food-web structure. Nature, 427, 835-838.

Corbet S.A. (2000). Conserving compartments in pollination webs. Conserv. Biol., 14, 1229-1231.

Dicks, L., Corbet, S.A. & Pywell, R. F. (2002). Compartmentalization in plant-insect flower visitor webs. J. Anim. Ecol., 71, 32-43.

Dirzo, R. & Miranda, A. (1991). Altered patterns of herbivory and diversity in the forest understory: A case study of the possible consequences of contemporary defaunation. In: Plant-animal interactions: evolutionary ecology in tropical and temperate regions (eds. Price, P.W., Lewinsohn, T. M., Fernandes, G. W., Benson, W. W.). Wiley and Sons Pub., New York, pp. 273–287.

Fleming, T.H., Breitwisch, R. & Whitesides, G.H. (1987). Patterns of tropical vertebrate frugivore diversity. Annu. Rev. Ecol. and Syst., 18, 91-109.

Fonseca, C. R. & Ganade, G. (1996). Asymmetries, compartments and null interactions in an Amazonian ant-myrmecophyte community. J. Anim. Ecol., 65, 339–347.

Fortuna, M.A., Stouffer, D.B., Olesen, J.M., Jordano, P., Mouillot, D., Krasnov, B. R., Poulim, R. & Bascompte, J. (2010). Nesteness versus modularity in ecological networks: two sides of the same coin? J. Anim. Ecol., 79, 811-817.

Galetti, M., Donatti, C. I., Pizo, M.A. & Giacomini, H.C. (2008). Big fish are the best: seed dispersal of Bactris glaucescens by the pacu fish (Piaractus mesopotamicus) in the Pantanal, Brazil. Biotropica, 40(3), 386-389.

Gautier-Hion, A., Duplantier, J.M., Quris, R., Feer, F., Sourd., C., Decoux, J.P. et al. (1985). Fruit characters as a basis of fruit choice and seed dispersal in a tropical forest vertebrate community. Oecologia, 65, 324-337.

27

Guimarães Jr., P.R., Galetti, M. & Jordano, P. (2008). Seed dispersal anachronisms: rethinking the fruits extinct megafauna ate. PLoS ONE, 3(3), e1745.

Guimarães, P. R., Rico-Gray, V., Oliveira, P. S., Izzo, T. J., dos Reis, S. F. & Thompson J.N. (2007). Interaction intimacy affects structure and coevolutionary dynamics in mutualistic networks. Curr. Biol., 17, 1797-1803.

Guimarães, P. R. & Guimarães, P. (2006). Improving the analyses of nestedness for large sets of matrices. Environ Model. and Softw., 21, 1512-1513.

Guimarães, P. R., Rico-Gray, V., Reis S. F. & Thompson, J.N. (2006). Asymmetries in specialization in ant-plant mutualistic networks. Proc. R. Soc. of Lond., 273, 2041-2047.

Guimerà, R., M. & Amaral, L.A.N. (2005). Cartography of complex networks: modules and universal roles. J. Stat. Mech.-Theory E., 02001, 1-13.

Hackett, S. J., Kimball, R.T., Reddy, S., Bowie, R.C.K., Braun, E.L., Chojnowski, J.L. et al. (2008). A phylogenomic study of birds reveals their evolutionary history. Science, 320, 1763-1768.

Harris, M.B., Tomás, W.M., Mourão, G., Da Silva, C.J., Guimarães, E., Sonoda, F. & Fachim, E. (2005) Safeguarding the Panatanal wetlands: threats and conservation initiatives. Conserv. Biol., 19, 714-720

Joppa, L.N., Montoya, J.M., Solé, R., Sanderson, J. & Pimm, S. (2010). On nestedness in ecological networks. Evol. Ecol. Res., 12, 35-46.

Jordano, P., Vázquez, D. & Bascompte, J. (2009) Redes complejas de interacciones planta-animal. In: Ecología y evolución de las interacciones planta-animal: conceptos y aplicaciones. (eds. Medel, R., Aizen, M., Zamora, R.). Editorial Universitaria, Santiago, pp. 17-41.

Jordano, P., Bascompte, J.& Olesen, J.M. (2006). The ecological consequences of complex topology and nested structure in pollination webs. In: Plant-pollinator interactions: from specialization to generalization (eds. Waser, N.M., Olerton, J.). The University of Chicago Press, London, pp. 173-199.

Jordano, P. (1987). Patterns of mutualistic interactions in pollination and seed dispersal: Connectance, dependence asymmetries and coevolution. Am. Nat., 129, 657-677.

Kembel, S. W., Cowan, P. D., Helmus, M. R., Cornwell, W. K., Morlon, H., Ackerly, D. D., Blomberg, S. P. & Webb, C. O. (2010). Picante: R tools for integrating phylogenies and ecology. Bioinformatics, 26, 1463-1464.

Klicka, J., Burns, K. & Spellman, G.M. (2007). Defining a monophyletic Cardinalini: A molecular perspective. Mol. Phylogenet. Evol., 45, 1014-1032.

Lewinsohn, T.M., Prado, P.I., Jordano, P., Bascompte, J. & Olesen, J.M. (2006). Structure in plant-animal interaction assemblages. Oikos, 113(1), 174-184.

Maddison, W.P. & Slatkin, M. (1991). Null models for the number of evolutionary steps in a character on a phylogenetic tree. Evolution, 45, 1184-1197.

Nuismer, S.L., Gomulkiewicz, R. & Ridenhour, B.J. (2010). When is correlation coevolution? Am. Nat., 175(5), 525-537.

Olesen, J.M., Bascompte, J., Dupont, Y.L. & Jordano, P. (2007). The modularity of pollination networks. PNAS, 104(50), 19891-19896.

Pereira, S.L., Baker, A.J. & Wajntal, A. (2002). Combined nuclear and mitochondrial DNA sequences resolve relationships within the Cracidae (Galliformes, Aves). Syst. Biol., 51(6), 964-958.

28

Pimm, S.L. & Lawton, J.H. (1980). Are food webs compartmented? J. Anim. Ecol., 49, 879 – 898.

Pott, A. & Pott,V.J. (1994). Plantas do Pantanal. Embrapa, Brasília. Prance, G.T. & Schaller, G.B. (1982). Preliminary study of some vegetation types of the

Pantanal, Mato Grosso, Brazil. Brittonia, 34(2), 228-251. Rezende, E., Albert, M. E., Fortuna, M.A. & Bascompte, J. (2009). Compartments in a

marine food web associated with phylogeny, body mass, and habitat structure. Ecol. Lett., 12, 779-788.

Rezende, E., Lavabre, J., Guimarães Jr., P.R. & Bascompte, J. (2007). Non-random coextinctions in phylogenetically structured mutualistic networks. Nature, 448, 925-928.

Santamaría, L. & Rodríguez-Gironés, M.A. (2007). Linkage rules for plant-pollination networks: Trait complementary or exploitation barriers? PLoS Biol., 5(2), 354-362.

Silva, W.R., Guimarães Jr., P.R., dos Reis, S.F. & Guimarães, P. (2007). Investigating the fragility in plant-frugivore networks: A case study of the Atlantic Forest in Brazil. In: Seed dispersal: theory and its application in a changing world. (eds. Dennis, A.J., Schupp, E.W., Green, R.J., Westcott, D.A.). CAB International, Wallinford, pp. 561-578.

Stang, M., Klinkhamer, P.G.L., van der Meijden E. (2007). Asymmetric specialization and extinction risk in plant–flower visitor webs: a matter of morphology or abundance? Oecol., 151, 442–453.

Tello, J.G., Moyle, R.G., Marchese, D.J. & Cracraft, J. (2009). Phylogeny and phylogenetic classification of the tyrant flycatchers, cotingas, manakins, and their aliens (Aves: Tyrannides). Cladistics, 25, 429-467.

Thebault, E. & Fontaine, C. (2010). Stability of ecological communities and the architecture of mutualistic and trophic networks. Science, 329, 853-856.

Thompson, J.N. (2005). The geographic mosaic of Coevolution. University of Chicago Press, Chicago.

Van der Pijl, L. (1982). Principles of dispersal in higher plants. Springer, Berlin. Vázquez, D.P., Chacoff, N.P. & Cagnolo, L. (2009). Evaluating multiple determinants of

the structure of plant-animal mutualistic networks. Ecology, 90, 2039-2046. Wojciechowski, M.F., Lavin, M. & Sanderson, M.J. (2004). A phylogeny of legumes

(Leguminosae) based on analysis of the plastid matK gene resolves many well-suported subclades within the family. Am. J. Bot., 91, 1846-1862.

29

Figures

Figure 1 Modularity of the Pantanal seed dispersal network. Each module is identified by

a different color (bird module 1: green, 22 species, bird module 2: blue, 22 species,

mammal-dominated module 1: red, 25 species, mammal-dominated 2: yellow, 18 species

and fish: purple, 4 species) in which each species was assigned. Circles in dark shades

represent animal species and squares in light shades represent plant species. The size of

circles refers to animal body mass (with large circles representing species with body

mass≥4.5 kg), whereas the size of squares refers to the fruit diameter (with large squares

representing species with fruit diameter ≥95mm). Both body mass and fruit diameters

were divided in four size categories exclusively for the purpose of this figure. Fruits and

seed sizes are in the same scale in all modules, and represent the relative diameter and

length of fruits and seeds in each module. The figure was manually done using the pajek

package (http://vlado.fmf.uni-lj.si/pub/networks/pajek/).

30

Figure 2 Association between body mass and values of fruit traits, illustrating trait

complementarity underlying the modular structure of the network (bird module 1: green,

bird module 2: blue, mammals-dominated module 1: red, mammal-dominated 2: yellow

and fish: purple). a) fruit diameter and b) fruit mass. Body mass is in log scale.

31

Figure 3 (a) The phylogenetic tree of animal species and modules (identified by different

colors) in which each species was assigned (bird module 1: green, bird module 2: blue,

mammals-dominated module 1: red, mammal-dominated 2: yellow and fish: purple).

Modules associated with birds showed significant phylogenetic signal.

32

Figure 3 (b) The phylogenetic tree of plant species and modules (identified by different colors) in which each species was assigned (bird module 1: green, bird module 2: blue, mammals-dominated module 1: red, mammal-dominated 2: yellow and fish: purple).

33

Figure 4 Role of each species in the seed dispersal network. Each symbol describes the within-module degree (z) and the participation coefficient (c) of each species. We used the values of 2.5 for z and of 0.62 for c to assign a role to each species, which could be peripheral species (bottom left), module hub (top left), supergeneralist (top right) or connector (bottom right). Species are color coded according to the module to which they belong (bird module 1: green, bird module 2: blue, mammals-dominated module 1: red, mammal-dominated 2: yellow and fish: purple). Circles represent animal species and squares represent plant species. Lines at z=2.5 and c=0.62 define species roles.

34

Supporting Information

Figure S1. Accumulation curve with the average (in black) and the standard deviation (in grey) of the number of seed dispersal interactions in 1000 iterations, as a function of the number of seed dispersal events. The red line shows the extrapolation of the curve and the dashed line represents the asymptotic value for the number of seed dispersal interactions. We sampled 2,070 events of seed dispersal, 273 seed dispersal interactions and estimated the asymptote for the number of seed dispersal interactions to be ~289.

35

Appendix S1 Description of the study sites The Pantanal, with its 170,000 km2, stretches through Central-West Brazil, Bolivia and into Paraguay. In Brazil, the Pantanal is bordered on the east by one of the world’s richest savannahs, the Cerrado; fringed to the northwest by semi-deciduous forest related to the Amazon, and is contained to the southwest by the dry chaco-like forest of neighboring Bolivia (Prance & Schaller 1982). Within the Pantanal, the study concentrated on the areas surrounding two neighboring farms (Rio Negro and Barranco Alto) that effectively served as field stations. Rio Negro farm (19°34'15"S 56°14'43"W) is a 7,500 ha private property, and Barranco Alto farm (19º34'40"S 56º09'08"W) is an 11,000 ha private property; both located in the Nhecolândia region. Average annual rainfall in this region is 1,192.5 mm and mean monthly temperature is 26°C, ranging from 19°C to 33°C (2004-2008; D. Eaton, unpublished work). Rio Negro farm has 7,500 ha of protected area, free of cattle since 2001. Barranco Alto Farm has 11,000 ha, in which 3,400 ha are protected areas free of cattle since 1980. Both areas are among the most pristine ones in the Pantanal. Reference

Prance, G.T., G.B.Schaller. 1982. Preliminary study of some vegetation types of the Pantanal, Mato Grosso, Brazil. Brittonia 34(2): 228-251.

36

Appendix S2 Justification for setting branch lengths equal to one We have tested the assumption that branch lengths equal to one is acceptable for this study. To do so, we created 1,000 phylogenetic trees, simulated trait data using the function rTraitCont in R and calculated the K statistic using Kcal in R (=true K values using “real” branch lengths). Then we re-set all branch lengths equal to one and re-calculated the K statistic. We plotted K values using branch lengths equal to 1 (K value for eddge.lengths=1 in figure S2 below) and K values using “real” branch lengths (True K value in figure S2 below). The data show that K values calculated using branch lengths equal to 1 are consistently greater than K values calculated using “real” branch lengths (see figure S2 below). Therefore, as all K values presented in the manuscript are lower than or equal to one, we can say that setting all branches = 1 is, in the case of this study, a conservative approach.

Figure S2 Association between K values using branch lengths=1 (K value for edge.length=1) and K values using branch lengths generated by our simulations (True K value).

37

CHAPTER 2

DEFAUNATION SIMULATIONS REVEAL THE CONSEQUENCES

OF SEED DISPERSAL NETWORK DISRUPTIONS AND THE ROLE

OF AN EXOTIC SPECIES ON DISPERSAL SERVICES Camila I. Donatti, Paulo R. Guimarães, Mauro Galetti, Marco Aurélio Pizo & Rodolfo Dirzo

Abstract

Given that mutualistic species often interact with multiple partners, the extinction of a

species may affect several others, causing cascading effects that may in turn reverberate

throughout the entire community. Here, we addressed how a species-rich seed dispersal

network changes with selective animal extinction. We compared i) structural network

metrics, ii) network robustness and iii) the number of seed dispersal events that are lost in

the community, under three distinct defaunation simulations, where the sequence of

cumulative removal of species was based on i) the body mass of animals, ii) their number

of links or iii) the random loss of animals in the community. Our results show that the

removal of large- and medium-bodied species has the largest impact on the network

topology and generates the lowest network robustness among the three simulations.

Furthermore, the absence of these species significantly increases the number of seed

dispersal events that are lost when compared to the simulated extinction of randomly

selected species. Surprisingly, our results also show that the exotic feral pig (Sus scrofa)

plays a critical, positive role in maintaining structural network metrics and in providing

seed dispersal services in this community. Thus, we posit that network stability will be

significantly reduced with defaunation of large- and medium-bodied species, but this

effect can be attenuated by the presence of large megafaunal exotic species, if such

species are less susceptible to extinction in their novel ecosystems.

38

Introduction

Defaunation, the contemporary pulse of animal local population loss or decline

(sensu Dirzo & Miranda 1991) driven by human activities such as hunting, land use

change, and the synergies of both, has been occurring at high rates in tropical forests

(Redford 1992, Peres 2000, Dirzo 2001, Corlett 2007, Wilkie et al. 2011), with large-and

medium-bodied vertebrate species being the most affected ones (Bodmer et al. 1997,

Cardillo et al. 2005, Cardillo et al. 2006, Peres & Palacios 2007). These species are

extremely important for several ecological interactions (e.g., seed predation, dispersal and

herbivory) and their absence may deeply affect ecological organization and services

(Dirzo 2001, Terborgh et al. 2001, Sinclair et al. 2007, Wright et al. 2007, Estes et al.

2011). In seed dispersal interactions, this is likely due to the fact that megafaunal species

usually interact with multiple partners (Bascompte et al. 2003, Strauss & Irwin 2004,

Jordano et al. 2006, Donatti et al. 2011) and their absence can affect many interacting

partners concurrently, causing cascading effects that may reverberate in the entire

community. Here, we use network theory and species extinction simulations to address

the consequences of the loss of particular animal species on seed dispersal services in a

species-rich community.

The patterns of interaction among species in a community have been described

using the network approach (Memmott et al. 1999, Bascompte & Jordano 2007).

Mutualistic networks are characterized by the prevalence of a nested pattern (Bascompte

et al. 2003, Vázquez et al. 2009, Fortuna et al. 2010, Joppa et al. 2010), whereby the

interactions of the specialist species tend to be a subset of the interactions observed

among the generalists (Bascompte et al. 2003). In addition, some mutualistic networks

are also modular, whereby subsets of species (modules) more frequently interact with

each other than with species from other modules (Olesen et al. 2007, Donatti et al. 2011).

Nestedness, modularity and the proportion of all interactions that are realized

(connectance), affect network stability and can be used to assess how network topology

may change in the absence of particular species (Tylianakis et al. 2010).

As large-bodied seed dispersers are less likely to show size constraints between

their gape and seed and/or fruit sizes, they are expected to be able to disperse a large

number of plant species (Jordano 1995), that is, to be “super-generalists” (Guimarães et

39

al. 2011). The presence of super-generalist species in a community contributes to a high

nestedness and connectance of the network (Bascompte et al. 2003, Olesen et al. 2007).

Therefore, if these super-generalist species disappear, we should expect a decrease in

nestedness and connectance of the network, thus reducing also its robustness (Dunne et

al. 2002, Fortuna & Bascompte 2006, Piazzon et al. 2011).

To examine these relationships, we developed defaunation simulations to compare

the consequences of the absence of selected seed dispersers vs. the absence of randomly

selected species from a community that includes native and exotic seed dispersers and the

plant species they disperse. We compared structural network metrics, network robustness

and the number of seed dispersal events that are lost in the community, under three

distinct defaunation simulations, each with 16 defaunation steps related to the cumulative

removal of particular species from the network. In a “realistic defaunation simulation”,

we removed species based on their body mass, with the heaviest species removed first. In

a “most-to-least linked species simulation”, we removed species based on their number of

links, with the most linked species removed first. We also developed “random

defaunation simulations” where we randomized both the order and the identity of the

removed species.

Specifically, we hypothesized that the realistic defaunation simulation would lead

to: 1) a lower network robustness, whereby the extinction of a few particular animal

species leads to a high proportion of the plants losing seed dispersal services, 2) loss of

dispersal services to large-fruited and/or large-seeded plant species, given that those plant

species are predominantly dependent on large-bodied dispersal agents (Janzen & Martin

1982, Janson 1983, Wheelwright 1985, Chapman et al. 1992, Guimarães et al. 2008,

Donatti et al. 2011), and 3) overall, a reduction in the number of events of seed dispersal

in the network. We show that the simulated extinctions of large- (i.e. >5kg) and medium-

bodied species (i.e. 1-5Kg) lead to a network that is more susceptible to coextinctions,

when compared to the intact network and to networks generated by simulated random

extinctions and by simulated extinction of most linked species. Moreover, plant species

with longer, wider and/or heavier fruits and/or seeds are more likely to lose all seed

dispersal potential with the absence of large- and medium-bodied species. Furthermore,

we show that the absence of a few, largest species, would significantly decrease the

40

number of seed dispersal events in this community, while an exotic, large-bodied animal

species is crucial in maintaining network structure and seed dispersal services. Such a

situation of the concurrence of selective defaunation and addition of exotic dispersers is

becoming increasingly common in tropical landscapes (Blackburn et al. 2004) and yet

has been largely unstudied. We posit that network stability will be significantly reduced

with defaunation of large- and medium-bodied species, but this effect would be

attenuated by the presence of large megafaunal exotic species, some of which may exert a

compensatory role in the face of contemporary defaunation.

Methods

Study site

The network of interactions from which we simulated species extinctions was

sampled in the Brazilian Pantanal, in two neighboring locations: Rio Negro and Barranco

Alto farms, covering 7,500 ha and 11,000 ha, respectively, of private land. More details

about the study sites can be found in Donatti et al. (2011).

Seed dispersal interactions

Seed dispersal interactions were recorded using four methods: focal observations,

camera trap techniques, scat and intestine analyses. One seed dispersal event was noted

when either: fruits were recorded to have been swallowed or removed from a plant

species during focal observations; fruit removal of a particular species by a potential seed

disperser was detected with camera traps; a scat pile was found to have at least one intact

seed of a particular species in it; or a sampled fish intestine contained at least one intact

seed from a particular species. We have sampled 94.5% of interactions that occur in this

community (Donatti et al. 2011) and more details about the methods and the structure

and composition of species in this seed dispersal network can be found in Donatti et al.

(2011).

Simulation of defaunation steps

Traditionally, simulations of species extinctions in interaction networks focus on

the removal of most-to-least linked species (i.e. species with more connections are

41

removed first) and show drastic changes in the structure and functioning of networks

(Solé & Montoya 2001, Dunne et al. 2002, Memmott et al. 2004). However, this

extinction simulation is ecologically unrealistic (Srinivasan et al. 2007), given that the

order of species loss does not necessarily correlate with their number of interactions in

the network. Therefore, as the likelihood of animal extinctions is correlated with

demographic and life history traits (Davidson et al. 2009) that are also related to body

mass (Bodmer et al. 1997, Jerozolimski & Peres 2003, Peres & Palacios 2007), we also

simulated a sequence of extinctions based on the body mass of species, with the heaviest

species removed first. Indeed, it is known that species ≥1kg are more likely to be affected

by hunting (Peres 2000) and forest fragmentation (Michalski & Peres 2007). Thus, to

assess the effect of the absence of particular animal species on a seed dispersal network,

we compared network metrics in defaunation steps based on the removal of large- and

medium-bodied species from the original network (i.e. the network that includes all

sampled interactions) with the same metrics derived from defaunation steps based on the

removal of random species. As the removal of the most linked species should be the

extreme situation that drastically affects network metrics, we also compared network

metrics in defaunation steps based on the removal of large- and medium-bodied species

with the same metrics derived from defaunation steps based on the removal of highly

linked species from the original network.

In the “realistic defaunation simulation” we created 16 defaunation steps based

on the cumulative removal of 16 species with body mass >1 kg from the network (Table

S1). In each defaunation step we selected one additional species to be removed, from the

heaviest to the lightest, until all species with at least 1 kg were no longer present in the

network. As a null model, referred to as “random defaunation simulations”, we also

created 16 different defaunation steps based on the cumulative removal of 16 random

species from the network. Species were randomly assigned to each of the defaunation

steps and randomizations were done 1,000 times. In the “most-to-least linked defaunation

simulation”, we removed 16 species from the network based on their number of

interactions. In this simulation, we created 16 defaunation steps and the most linked

species were removed first. In each simulation, the removal of 16 species represents 34%

of the animal species in the network.

42

Effects of defaunation on structural network metrics

To define the pattern of interaction in each defaunation step of each simulation, we

organized qualitative matrices that included all recorded interactions between plant and

animal species. In matrices of plants in columns and animals in rows, a cell representing

a recorded interaction between a plant and an animal species received the value of one,

and “zero” otherwise. We then used these matrices to compute network’s connectance,

nestedness and modularity. Details on the nestedness and modularity analyses can be

found in Donatti et al. (2011).

We compared the connectance of the network (the proportion of all interactions

that are realized) across defaunation steps among the realistic, random and the most-to-

least linked defaunation simulations. Additionally, we calculated the values of nestedness

and modularity, as well as their respective levels of significance, for each of the 16

defaunation steps in the realistic defaunation simulation, for each of the 1,000

randomizations in each of the 16 defaunation steps in random defaunation simulations

and for each of the 16 defaunation in the most-to-least linked defaunation simulation. To

address whether the network is stable in the sense that its metrics do not drastically

change across defaunation steps, we compared the sum of the variation in connectance,

nestedness and modularity across defaunation steps among the realistic defaunation

simulation, the most-to-least linked defaunation simulation and the average of the sums

of the variation across defaunation steps in 1,000 random defaunation simulations.

Effects on seed dispersal events

We also assembled quantitative seed dispersal matrices that included the number

of seed dispersal events recorded for each interaction between a plant and an animal

species (see Seed dispersal interactions in this section). To address the impact of the

absence of large- and medium-bodied species on the number of seed dispersal events, we

removed the same 16 animal species from the original quantitative matrix, and computed

the number of seed dispersal events that are lost in each defaunation step. Likewise, we

computed the same variable in each of the 1,000 randomizations for each defaunation

43

step in random defaunation simulations and in each defaunation step of the most-to-least

linked defaunation simulation.

Effects on the robustness of the network

Robustness is defined as the propensity for networks to lose species secondarily

or, in the case of this study, to leave plant species without seed dispersers. The robustness

of the network was thus computed as the percentage of animal species that had to be

removed in order for ≥ 50% of the plant species to no longer have seed dispersers

(modified from Dunne et al. 2002). We compared the robustness of the network in the

realistic defaunation simulation with the robustness of both the network in random

extinction simulations and the network in most-to-least linked defaunation simulation.

We also compared the robustness of the network in the realistic defaunation simulation,

random defaunation simulations and the most-to-least linked defaunation simulation

using quantitative data, i.e. the number of animal species that had to be removed in order

for ≥ 50% of the number of seed dispersal events to be lost in the community.

Additionally, we examined traits of plant species that would most likely lose all

their seed dispersers upon removal of animal species. We measured fruit and seed

morphological traits (length, diameter and mass) for all plant species in this community,

in at least 30 fruits and seeds from at least five individuals. Then we compared the

average values of fruit and seed traits of species that would most likely lose all seed

dispersers in each of the defaunation steps in the realistic defaunation simulation with

those of plant species randomly selected from the species pool. Randomizations were

done 1,000 times.

The importance of particular native and exotic species on network structure and seed

dispersal services

To assess the importance of each of the large-and medium-bodied species in the

process of seed dispersal, we ran a new set of 16 simulations with 15 defaunation steps

each. To assess the importance of each one of the large- and medium-bodied species, in

addition to remove species based on their body masses from the intact network, as

44

previously described, we also maintained one of these species in all defaunation steps of

the same simulation. We ran those simulations, each one maintaining one particular

species in the community, for all large- and medium-bodied species. Therefore, we could

assess changes in connectance, nestedness and modularity, and compare network

robustness when a particular species remained in the community while all other large-

and medium-bodied species were removed from the community.

Results

Effects of defaunation on structural network metrics

Network connectance in all defaunation steps in the realistic defaunation simulation

was lower than expected by chance. In addition, the sum of the variation in connectance

across all 16 defaunation steps in the realistic defaunation simulation was lower than in

the most-to-least linked defaunation simulation and both were significantly higher than

the average in random defaunation simulations (realistic defaunation simulation=0.061,

most-to-least linked defaunation simulation=0.077, average random defaunation

simulation=0.04, p=0.001 for both). This implies that connectance changes more

drastically across defaunation steps in the realistic defaunation simulation than with the

random removal of species.

The absence of large- and medium-bodied species decreases the values of

nestedness and increased the values of modularity across defaunation steps, when

compared to the removal of randomly selected species (Fig. 1). The low values of

nestedness calculated in the realistic defaunation simulation were not due to chance in the

majority of defaunation steps. Likewise, modularity in the realistic defaunation

simulation was higher than expected by chance in most defaunation steps, the exceptions

being the last three steps where species ≤2.8 kg remained in the community. In fact, in

these last three defaunation steps of the realistic defaunation simulation, the network was

neither nested nor modular, indicating that the organization of the mutualistic network

was vanished. The sum of the variation in nestedness across all 16 defaunation steps was

higher in the realistic defaunation simulation than in the most-to-least linked defaunation

simulation and both were significant higher than the average in random extinction

simulations (realistic defaunation simulation=17.88, most-to-least linked defaunation

45

simulation=17.56, average random defaunation simulation=5.85, p<0.0001 for both). The

sum of the variation in modularity across all 16 defaunation steps was lower in the

realistic defaunation simulation than in the most-to-least linked defaunation simulation

and both were significantly higher than the average in random extinction simulations

(realistic defaunation simulation=0.19, most-to-least linked defaunation simulation=0.22,

average random defaunation simulation=0.09, p=0.004 and p<0.0001, respectively).

Therefore, both nestedness and modularity also changed more drastically across

defaunation steps in the realistic defaunation simulation than with the random removal of

species.

The greater change in the topology of the network that results from the realistic loss

of species, when compared to the network that results from random extinctions, occurs

because large- and medium-bodied species have more interactions in the network. In fact,

the number of plant species dispersed increases with the body mass of the seed disperser

(F=65.62, p<0.0001, r=0.77, d.f.=45) and this association holds when we analyze our

data separately via the different methods used to sample seed dispersal interactions in this

community (focal observations, camera trap techniques, scat and intestine analyses; see

SI Appendix). Consequently, although we used different methods to assess seed dispersal

interactions for large-and small-bodied species, our results showing this positive

association are likely not an artifact resulting from sampling protocols. Therefore,

significant changes in nestedness and connectance with the absence of large- and

medium-bodied species make topological and ecological sense, given that the more plant

species an animal species can disperse, the greater the effect of the absence of this animal

species on network structure. Indeed, we found that connectance and nestedness decrease

even more drastically in the network that faces the extinction of the most-to-least linked

species when compared to those in the network that faces the extinction of large- and

medium-bodied species.

Effects on seed dispersal events

The number of interactions that were lost in the realistic defaunation simulation

was higher than expected by chance in most of defaunation steps probably because large-

46

and medium-bodied species also interact with plant species more frequently. We found a

positive and significant effect of body mass on the number of seed dispersal events (F=

20.09, r=0.56, p<0.0001, d.f.=45), association that holds across the different sampling

methods (see SI Appendix). Thus, although we used different methods to also assess seed

dispersal events for large-and small-bodied species, the positive association between

body mass and number of events of seed dispersal is likely not an artifact of the methods

that we used.

Effects on the robustness of the network

In the realistic defaunation simulation, all plant species had at least one seed

disperser when species < 6.6 kg were still present in the community. In addition, our

results showed significant trends in the representation of morphological traits of plant

species that lost all their seed dispersers. Plant species that would most likely lose all

their seed dispersers with the absence of large- and medium-bodied species were

characterized by having longer, wider and/or heavier fruits and/or seeds than those of

plant species randomly selected from the species pool. Surprisingly, with the removal of

species based on their number of links, the network was more robust than with the

removal of large- and medium-bodied species. A lower percentage of animal species had

to be removed in order for ≥ 50% of the plant species to no longer have seed dispersers in

the realistic defaunation simulation, followed by the most-to-least linked defaunation

simulation and by the average of random defaunation simulations (32.6%, 47.8% and

80%, respectively, Fig. 2). This result indicates that, in the realistic defaunation

simulation, some animal species are dispersing plant species that no other species in the

network can disperse, even if they do not have a high number of interactions.

The quantitative analysis revealed that the network might experience an even

earlier collapse when qualitative data are considered, for all three types of simulations.

The same percentage of the animal species in the community had to be removed in order

for ≥ 50% of the seed dispersal events to be lost in both the realistic and the most-to-least

linked defaunation simulations (13%), followed by the average in random defaunation

simulations (50%, Fig. 2).

47

The importance of particular native and exotic species on network structure and seed

dispersal services

Our qualitative data shows that the connectance, nestedness and modularity of the

network changed less drastically when all large- and medium-bodied species, except the

exotic feral pig, were removed from the network. Furthermore, our quantitative data

shows that, although the maintenance of feral pigs or tapirs (Tapirus terretris) in the

network decreases the number of plant species that no longer have seed dispersers (Table

S2), fewer seed dispersal events are lost in the network if the feral pig remains in the

community than if the tapir does. Therefore, in this community, one exotic megafaunal

species is the most important player both in maintaining structural network metrics and in

providing seed dispersal services.

Discussion

This study shows that the absence of large- and medium-bodied animal species

will change the pattern of interaction and the relative representation of species traits in

species-rich communities, given that several plant species will no longer have seed

dispersers. More specifically, with the absence of animal species >2.8 kg, the network

becomes less connected, less nested and more modular. Furthermore, with the removal of

species ≤2.8 kg, the network becomes neither nested nor modular. Our results also

demonstrate that the absence of large- and medium-bodied species decreases the

connectance of the network. Therefore, using the definition of stability as a system’s

potential to resist a perturbation (see McCann 2000), we posit that network stability will

be diminished or not be maintained if large- and medium-bodied species are removed

from the community.

Several studies have recently shown species networks collapses with the removal

of most linked species (Solé & Montoya 2001, Dunne et al. 2002, Memmott et al. 2004).

However, our study demonstrates that it is the specific removal of large- and medium-

bodied species that has the largest impact on the network topology, since it will no longer

be nested or modular. Additionally, the absence of these species significantly increases

the number of seed dispersal events that are lost in the network when compared to

random defaunation. Therefore, opportunities for seed dispersal in plant communities can

48

decrease considerably as large-and medium-bodied species are removed from the

community. Since we were able to collect similar number of scats of species with

different body size (T. terrestris: 293 scats, S. scrofa: 331 scats, R. americana: 239 scats),

and given that we conducted focal observations for similar number of hours in several

plant species, these results are not likely a consequence of any observational bias in our

data.

All plant species still had all their seed dispersers in defaunation steps where

species > 6.6 kg were already removed from the community, because there is redundancy

in the use of the majority of plant species by seed dispersers: only 6.5% of the plant

species rely on a single seed disperser species. Moreover, as previously mentioned, the

pattern of the network is nested even after the removal of all species ≥ 4.5 kg from the

network, which could also contribute to its tolerance regarding secondary extinctions.

Nevertheless, in defaunation steps in which seed dispersers > 6 kg were no longer present

in the network, plant species with longer, wider and heavier fruits and seeds are the first

to lose seed dispersers, due to size mismatching between fruit and/or seed sizes and the

gape size of animal species that remained in the community (Donatti et al. 2007).

However, some small-bodied species (< 1 kg), such as small rodents, could increase

their densities in the absence of large-bodied species and, potentially, fill up their roles as

seed dispersers, at least for some plant species. Although compensatory density has been

shown to occur in areas without large-bodied animals (Peres & Dolman 2000), it is not

clear whether small-bodied species could necessarily fill up the ecological gap created by

the absence of larger species. Therefore, rare or small-bodied species are probably unable

to compensate for declines in ecosystem processes when large, more abundant species are

lost (Solan et al. 2004). More specifically, Donatti et al. (2009) showed that small-bodied

seed dispersers (i.e. spiny rats and squirrels, both < 1 kg) could not even compensate for

seed dispersal rate and distance even in the absence of medium-bodied animals (i.e.

agoutis ~2.8 kg).

Both the tapir and the agouti play a crucial role in dispersing large-seeded plants

(Hallwacks 1986, Fragoso 1997, Galetti et al. 2006) and this study shows that, among the

native species, they are the most important in maintaining network robustness. Whereas

the tapir can disperse a high number of plant species in this community, the agouti is the

49

only seed disperser of large-seeded plant species left when larger animals are removed

from the community. Nevertheless, this study also reveals the importance of some of the

large- and medium-bodied species that have currently invaded tropical ecosystems, such

as the exotic feral pig, on the process of seed dispersal. Since it is well known that exotic

mammalian species can have direct, negative impacts on native taxa (Cox 1999,

D’Antonio et al. 1999, Cushman et al. 2004, Busby et al. 2010) the importance of feral

pigs in maintaining both network topology and seed dispersal services described in this

study is surprising. However, because the plant community in the Pantanal includes

several species that were presumably dispersed by extinct Pleistocene megafauna

(Donatti et al. 2007, Guimarães et al. 2008), the ability of feral pigs to disperse some of

these fruits, combined to the relatively high density of pigs in this system (6.35 Ind/km2,

C.I. Donatti, unpublished data), make them extremely important analogue seed dispersers

in this community. Therefore, here we report a situation where an exotic species has a

positive role in a mutualistic process, perhaps because it, along with some native species

such as tapirs and agoutis, is filling up the empty niche left by the extinct Pleistocenic

megafauna (Janzen & Martin 1982).

We argue that the effects of differential defaunation on the process of seed

dispersal reported here are compelling given that we have sampled 94.5% of interactions

that occur in this community (Donatti et al. 2011). In addition, our network comprises

interactions in one of the few communities that still harbor intact contingents of large-

bodied mammals and birds (Harris et al. 2005). Our goal here was to address how a seed

dispersal network behaves when it is altered via animal extinction. We showed that the

selective defaunation of large- and medium-bodied species affects network stability more

drastically than the defaunation of random species. Therefore, changes found in the

structure of the network are not simply an effect of the loss of an indiscriminate number

of species, but are rather dependent on their specific identities (McCann 2000, Worm and

Duffy 2003). Consequently, a focus on conserving species diversity alone may not

necessarily conserve network structure and functioning (Bascompte et al. 2006, Bastolla

et al. 2009).

Our results raise conservation concerns because we have demonstrated significant

negative effects of the absence of large- and medium-bodied species even in a network

50

that shows a high seed dispersal redundancy among animal species and is highly nested,

which should provide a buffer against coextinctions. Even though not all large- and

medium-bodied species are already extinct in many tropical areas, the loss of just a few

among the largest ones (i.e. the six species >20 kg in this case) would decrease by more

than 50% the number of seed dispersal events in the community. As the selective loss of

large-bodied species and their disproportionate ecological importance is a global

phenomenon (Ceballos & Ehrlich 2002, Estes et al. 2011, Dirzo 2001, Wilkie et al.

2011), we predict that some of the consequences of differential defaunation on seed

dispersal services shown here are likely to be happening already in many tropical

ecosystems. However, some exotic species may exert a compensatory role as seed

dispersers in the face of contemporary defaunation, provided they are less susceptible to

extinction in their novel ecosystems.

Acknowledgments

We would like to thank Roger Guimerà for providing us the modularity algorithm, and

Jim Estes and members of the Dirzo lab for helpful comments on a previous draft. We

thank FAPESP (2004/00810-3 and 2008/10154-7), Earthwatch Institute and Conservation

International for financial support. CID was supported by Stanford University and the

Zaffaroni Fellowship Fund, PRG by FAPESP and CAPES, and MG and MAP by a CNPq

fellowship. We also thank Conservation International, Lucas Leuzinger and Marina

Schweizer for their kind permission to work in their properties.

51

References

Bascompte J, Jordano P (2007) The structure of plant-animal mutualistic networks: the architecture of biodiversity. Annu Rev Ecol and Syst: 38, 567-593.

Bascompte J, Jordano P, Melián C J, Olesen JM (2003) The nested assembly of plant-animal mutualistic networks. Proc Natl Acad Sci USA 100 (16): 9383-9387.

Bascompte J, Jordano P, Olesen JM (2006) Asymmetric coevolutionary networks facilitates biodiversity maintenance. Science 312: 431-433.

Bastolla U, Fortuna MA, Pascual-Garcia A, Ferreira A, Luque B, Bascompte J (2009) The architecture of mutualistic networks minimizes competition and increases biodiversity. Nature 458: 1018-1020.

Blackburn TM, Cassey P, Duncan RP, Evans KL, Gaston KJ (2004) Avian extinction and mammalian introductions on oceanic islands. Science 305:1955–1958

Bodmer RE, Eisenberg JF, Redford KH (1997) Hunting and the likelihood of extinction of Amazonian mammals. Conserv Biol 11: 460–466.

Busby PE, Vitousek P, Dirzo R (2010) The prevalence of tree regeneration by sprouting and seeding along a rainfall gradient in Hawaii. Biotropica 42(1): 80-86.

Cardillo M, Mace GM, Gittleman JL, Purvis A (2006) Latent extinction risk and the future battlegrounds of mammal conservation. Proc Natl Acad Sci USA 103(11): 4157-4161.

Cardillo M, Mace GM, Jones KE, Bielby J, Bininda-Emonds ORP, Sechrest W, Orme C DL, Purvis A (2005) Multiple causes of high extinction risk in large mammals species. Science 309 (5738): 1239-1241.

Ceballos G, Ehrlich PR (2002) Mammal population losses and the extinction crisis. Science 296:904 –907.

Chapman LJ, Chapman CA, Wrangham RW (1992) Balanites wilsoniana: Elephant Dependent Dispersal? J Trop Ecol 8 (3): 275-283. 1992.

Cox GW (1999) Alien species in North America and Hawaii: impacts on natural ecosystems. Island Press, Washington, D. C., USA.

Corlett RT (2007) The impact of hunting on the mammalian fauna of tropical Asian forests. Biotropica 39(3): 292-303.

Cushman JH, Tierney TA, Hinds, JM (2004) Variable effects of feral pig disturbances on native and exotic plants in a California grassland. Ecological applications 14(6): 1746-4756.

D’Antonio, CM, Dudley TL, Mack M (1999) in Ecosystems of disturbed ground, ed Walker LR (Elsevier, New York, USA), pp 413– 452

Davidson AD, Hamilton MJ, Boyer AG, Brown JH, Ceballos G (2009) Multiple ecological pathways to extinction in mammals. Proc Natl Acad Sci USA 106 (26): 10702-10705.

Dirzo R (2001) in Ecology: achievement and challenge, eds Press MC, Huntly NJ, Levin S (Blackwell Science, Oxford, UK), pp 319–335.

Dirzo R, Miranda A (1991) in Plant-animal interactions: evolutionary ecology in tropical and temperate regions, eds Price PW, Lewinsohn TM, Fernandes GW and Benson WW (John Wiley and Sons, New York , USA), pp 273–287.

Donatti CI, Galetti M, Pizo MA, Guimarães PR, Jordano P (2007) in Frugivory and Seed Dispersal: Theory and Applications in a Changing World, eds Dennis A, Green R, Schupp EW, Wescott D (CABI Publishing, Wallingford, UK), pp 104–123.

52

Donatti CI, Guimarães PR, Galetti M, Pizo M A, Marquitti FMD, Dirzo R (2011) Analysis of a hyper-diverse seed dispersal network: modularity and underlying mechanisms. Ecol Lett 14(8): 773-781.

Donatti CI, Guimarães PR, Galetti, M (2009) Seed dispersal and predation of an endemic Atlantic Forest palm in a gradient of seed dispersers’ abundance. Ecol Res 24(6): 1187-1195.

Dunne JA, Williams RJ, Martinez ND (2002) Network structure and biodiversity loss in food webs: robustness increases with connectance. Ecol Lett, 5, 558-567.

Estes JA, Terborgh J., Brashares JS, Power ME, Berger J, Bond W et al. (2011) Trophic downgrading of planet Earth. Science 333 (6040): 301-306.

Fortuna MA, Stouffer DB, Olesen JM, Jordano P, Mouillot D, Krasnov BR, Poulim R, Bascompte J (2010) Nestedness versus modularity in ecological networks: two sides of the same coin? J Anim Ecol 79: 811-817.

Fortuna MA, Bascompte J (2006) Habitat loss and the structure of plant-animal mutualistic networks. Ecol Lett 9: 278–283.

Fragoso JMV (1997) Tapir-generated seed shadows: scale-dependent patchiness in the Amazon rain forest. J Ecol 85: 519–529.

Galetti M, Donatti CI, Pires AS, Guimarães PR, Jordano P (2006) Seed survival and dispersal of an endemic Atlantic forest palm: the combined effects of defaunation and forest fragmentation. Bot J Linnean Soc 151: 141–149.

Guimarães PR, Galetti M, Jordano P (2008). Seed dispersal anachronisms: rethinking the fruits extinct megafauna ate. PLOS One 3: e1745.

Guimarães, PR, Jordano P, Thompson JN (2011) Evolution and coevolution in mutualistic networks. Ecol Lett 14(9): 877-885.

Hallwacks W (1986) in Frugivores and seed dispersal, eds Estrada A, Fleming JH (Dr. W. Junk Publishers, Boston), pp 285–304.

Harris MB, Tomás WM, Mourão G, Da Silva CJ , Guimarães E, Sonoda F, Fachim E (2005) Safeguarding the Pantanal wetlands: threats and conservation initiatives. Conserv Biol 19: 714-720

Janson CH (1983) Adaptation of fruit morphology to dispersal agents in a neotropical forest. Science 219: 187–189.

Janzen D & Martin S (1982) Neotropical anachronisms: the fruits the gamphotheres ate. Science 215: 19-27.

Joppa LN, Montoya JM, Solé R, Sanderson J, Pimm S (2010) On nestedness in ecological networks. Evol Ecol Res 12: 35-46.

Jordano P, Bascompte J, Olesen JM (2006) in Specialization and generalization in plant-pollinator interactions, eds Waser, NM, Ollerton J (University of Chicago Press, EEUU), pp 173-199.

Jordano P (1995) Angiosperm fleshy fruits and seed dispersers: a comparative analysis of adaptation and constraints in plant-animal interactions. Am Nat 145: 163-191.

Jerozolimski A, Peres CA (2003) Bringing home the biggest bacon: a cross-site analysis of the structure of hunter-kill profiles in Neotropical forests. Biol Conserv 111: 415-425.

McCann K (2000) The diversity-stability debate. Nature 405: 228-233. Memmott J (1999) The structure of a plant-pollinator food web. Ecol Lett 2: 276-280.

53

Memmott J, Waser NM, Price MV (2004) Tolerance of pollination networks to species extinctions. Proc R Soc Lond B 271: 2605-2611.

Michalski F, Peres CA (2007) Disturbance-mediated mammal persistence and abundance-area relationships in Amazonian forest fragments. Conserv Biol 21(6): 1626-1640.

Olesen JM, Bascompte J, Dupont YL, Jordano P (2007) The modularity of pollination networks. Proc Natl Acad Sci USA 104(50): 19891-19896.

Peres CA (2000) Effects of subsistence hunting on vertebrate community structure in Amazonian forests. Conserv Biol 14: 240–253.

Peres CA, Dolman PM (2000) Density compensation in neotropical primate communities: evidence from 56 hunted and nonhunted Amazonian forests of varying productivity. Oecol 122: 175–189.

Peres CA, Palácios E (2007) Basin-Wide Effects of Game Harvest on Vertebrate Population Densities in Amazonian Forests: Implications for Animal-Mediated Seed Dispersal. Biotropica 39: 304-315.

Piazzon M, Larrinaga AR, Santamaría L (2011) Are nested networks more robust to disturbance? A test using epiphyte-tree, communalistic networks. PLoS ONE 6(5): e19637. doi:10.1371/journal.pone.0019637

Redford KH (1992) The empty forest. Bioscience 42: 412–422. Sinclair ARE, Mduma SAR, Hopcraft JGC, Fryxell JM, Hiborn R, Thirgood S. (2007)

Long-term ecosystem dynamics in the Serengeti. Conserv Biol 21(3): 580-590. Solan M, Cardinale BJ, Downing AL, Engelhardt KAM, Ruesink JL, Srivastava DS

(2004) Extinction and Ecosystem function in the marine benthos. Science 306: 1177-1180.

Solé RV, Montoya JM (2001) Complexity and fragility in ecological networks. Proc R Soc Lond B 267: 2039-2045.

Srinivasan UT, Dunne J, Harte J, Martinez ND (2007) Response of complex food web to realistic extinction sequences. Ecology 88(3): 671-682.

Strauss SY, Irwin RE (2004) Ecological and evolutionary consequences of multispecies plant-animal interactions. Annu Rev of Ecol Evol Syst 35:435-466.

Terborgh J et al. 2001 Ecological meltdown in predator-free forest fragments. Science 294(5548): 1923-1926.

Tylianakis JM, Laliberté E, Nielsen A, Bascompte J (2010) Conserving of species interaction networks. Biol Conserv 143: 2270-2279.

Vázquez DP, Chacoff NP, Cagnolo L (2009) Evaluating multiple determinants of the structure of plant-animal mutualistic networks. Ecology 90: 2039-2046.

Visser SN, Freymann B, Olff H (2011) The Serengeti food web: empirical quantification and analysis of topological changes under increasing human impact. J Anim Ecol 80: 465–475.

Wheelwright NT (1985) Fruit size, gape width, and the diets of fruit-eating birds. Ecology 66: 808-818.

Wilkie DS, Bennett EL, Peres CA, Cunningham AA (2011) The empty forest revisited. Ann N Y Acad Sci 1223: 120-128.

Worm B, Duffy JE (2003) Biodiversity, productivity and stability in real food webs. Trends Ecol Evol 18: 628-632.

Wright S J et al. (2007) The plight of large animals in tropical forests and the

54

consequences for plant regeneration. Biotropica 39: 289–291.

55

Figures

Figure 1. Values of nestedness (top) and modularity (bottom) in each defaunation step. Red circles represent those in realistic defaunation simulation, blue circles represent those in the most-to-least linked defaunation simulation and box plots represent those in random defaunation simulations.

!

"

#!

#"

$!

$"

%!

&'()'*+'((,-&./01

# $ % 2 " 3 4 5 6 #! ## #$ #% #2 #" #3

!

!7#

!7$

!7%

!72

!7"

!73

!74

/'89:+9);<+,()'=(

><*:?9@;)A,->1

56

Figure 2. The robustness of the network in terms of the number of plant species that

would lose all seed dispersers in each of the defaunation steps (top) and the number of

seed dispersal events that would be lost in the network in each of the defaunation steps

(bottom). Red circles represent those in the realistic defaunation simulation, blue circles

those in the most-to-least linked defaunation simulation, and box plots those in random

defaunation simulations.

57

Supporting Information

Table S1. Animal species removed in each of the defaunation steps in the realistic defaunation simulation, family and their body masses. Note that each defaunation step (except defaunation step one) also contains species that were/was already removed in the previous defaunation step.

Defaunation steps Additional species removed Family Body mass (in kg.)

1 Tapirus terrestris Tapiridae 240

2 Sus scrofa Suidae 50

3 Mazama spp. Cervidae 36

4 Tayassu pecari Tayassuidae 35

5 Pecari tajacu Tayassuidae 26

6 Rhea Americana Rheidae 20

7 Cuniculus paca Cuniculidae 9.1

8 Geochelone carbonaria Testudinidae 6.6

9 Euphractus sexcintus Dasypodidae 6

10 Cerdocyon thous Canidae 5.7

11 Alouatta caraya Cebidae 5.4

12 Nasua nasua Procyonidae 5.1

13 Procyon cancrivorus Procyonidae 4.5

14 Dasyprocta azarae Dasyproctidae 2.8

15 Crax fasciolata Cracidae 2.8

16 Aburria pipile Cracidae 1

58

Table S2. Animal species maintained in the network (listed by body mass) while all other

large- and medium-bodied species were removed, the sum of the variation for the

Nestedness (NODF), Modularity (M) and Connectance (C), the number of plant species

that would lose all their seed dispersers in each simulation, and the number of seed

dispersal events lost. Note that in the simulation where the feral pig (Sus scrofa) is

maintained in the network, there is less change in network metrics, less seed dispersal

events are lost and less plant species no longer have seed dispersers when compared to

the other simulations.

Species maintained

Sum of the

variation (NODF)

Sum of the

variation (M)

Sum of the

variation (C)

Plant species

without dispersers

Seed dispersal

events lost

Tapirus terrestris 15.42 0.1302 0.045 8 1310

Sus scrofa 12.89 0.1288 0.044 8 1065

Mazama spp. 18.17 0.1917 0.055 22 1460

Tayassu pecari 14.98 0.1411 0.046 11 1247

Pecari tajacu 13.38 0.1787 0.053 17 1455

Rhea americana 14.89 0.1545 0.054 18 1336

Agouti paca 16.21 0.2176 0.047 22 1499

Geochelone carbonaria 14.98 0.1793 0.050 15 1471

Euphractus sexcintus 17.61 0.2029 0.056 22 1487

Cerdocyon thous 16.97 0.1520 0.050 14 1409

Alouatta caraya 17.45 0.2160 0.056 23 1497

Nasua nasua 16.09 0.1561 0.048 13 1457

Procyon cancrivorus 18.18 0.1944 0.056 21 1487

Dasyprocta azarae 15.47 0.1596 0.048 13 1439

Crax fasciolata 17.81 0.1928 0.055 21 1487

Aburria pipile 16.78 0.1735 0.055 25 1439

59

Appendix S1. Association between both the number of plant species dispersed and the

number of seed dispersal events and the body mass of seed dispersers across different

methods

The positive association between the number of plant species dispersed and the body

mass of the seed disperser holds when we analyzed data collected through camera traps

and focal observations separately (camera trap: F=4.8, p=0.045, r2=0.25, n=16; focal

observations: F=10.4, p=0.003, r2=0.27, n=30), but not when analyzing data collected

through scats only (F=2.02, p=0.205, n=8), the one sampling method in which our sample

size is limited. However, when pooling data from camera traps and focal observations

together, we still get the same positive and significant relationship between body mass

and number of interactions (F=43.9, p<0.0001, r2=0.53, n=40). Likewise, the association

between the number of seed dispersal events and the body mass of seed dispersers holds

when analyzing data collected through camera traps and focal observations separately

(camera traps: F=5.15, p=0.039, r2=0.26, n=16; focal observations: F=7.4, p=0.01, r2=0.2,

n=30), but not when analyzing data collected through scat analyses only (F=1.7,

p=0.2398, n=8). However, when pooling data from camera traps and focal observations,

we still get the positive and significant association between body mass and number of

seed dispersal events (F=20.62, p<0.0001, r2=0.34, n=40).

60

CHAPTER 3

EFFECTS OF INTRA- AND INTER-SPECIFIC SEED SIZE

VARIATION ON SELECTION BY DISPERSERS, GERMINATION

AND SEEDLING GROWTH Camila I. Donatti, Mauro Galetti, Marco Aurélio Pizo & Rodolfo Dirzo

Abstract

Seed dispersal can help plants explore and colonize new or enemy-free habitats, facilitate

the regeneration of populations and communities, and contribute to the maintenance of

plant diversity. However, several studies have shown that seed dispersal can be affected

by defaunation, the contemporary pulse of animal population loss or decline driven by

human activities, which primarily affects large- and medium-bodied animal species.

Studies have found that, especially for large-seeded plants, the rate of seed removal and

dispersal decreases in defaunated areas. As fruit size frequently varies within and among

species, and as seeds can be selected by dispersers according to their body (gape) size, we

hypothesized that large-bodied animals would selectively disperse large-seeded plant

species and conspecific seeds with large sizes. We tested this hypothesis by looking at the

association between the size of the disperser and the size of dispersed seeds, both at the

interspecific and intraspecific levels, in the Pantanal ecosystem. Furthermore, for one

plant species, we assessed the significance of seed size variation on germination, seedling

performance and survival. Within plant species, our results show that, both at the

interspecific and intraspecific levels, the diameter of the dispersed seeds increases with

the size of the seed disperser. In addition, at the intraspecific level, seed germination

increases with seed diameter and the combination of a large diameter and gut-passage

increases seedling growth, in both controlled and field conditions. We conclude that

large-bodied animals are not only important because they disperse large-seeded species,

but also larger conspecific seeds, leading to increased germination.

61

Introduction

Seed dispersal by frugivores is beneficial to both animals and plants (Herrera &

Jordano 1981, Howe & Smallwood 1982, Herrera 1987). While the animals gain from

ingesting fleshy fruits due to their nutritious content, the plants gain from being dispersed

by animals mainly in four ways: the escape from the high density-dependent mortality in

the vicinity of the mother plant (Janzen 1970, Connell 1971), the chance of reaching sites

that are favorable for seed germination (Wenny & Levey 1997), the expansion of plant

species local distributions (Howe & Smallwood 1982) and the increment in seed

germination rate and speed after gut-passage (van der Pijl 1982, Traveset &Verdú 2002).

Thus, seed dispersal can facilitate the regeneration of populations within natural

communities and, ultimately, can contribute to the maintenance of plant diversity

(Hubbell 1979, Clark et al. 1998, Connell & Green 2000, Ehrlén et al. 2006).

The process of seed dispersal has been recently reported to be indirectly affected by

defaunation, the contemporary pulse of animal population loss or decline (sensu Dirzo &

Miranda 1991) driven by human activities such as hunting and land use change (Redford

1992, Peres 2000, Dirzo 2001, Corlett 2007, Wilkie et al. 2011). As a result of

defaunation, several fleshy-fruited species may have their dispersal services reduced or

lost due to the absence or low abundances of large- and medium-bodied species, those

that are more vulnerable to the risk of extinction (Bodmer et al. 1997, Cardillo et al.

2005, Cardillo et al. 2006, Peres & Palacios 2007).

Defaunation studies have found that, at least for large-seeded plants, the rate of

seed removal and dispersal decreases (Silva & Tabarelli 2000, Wright 2003, Galetti et al.

2006, Cramer et al. 2007) and the number of seeds that remains beneath the mother plant

increases in defaunated areas (Galetti et al. 2006, Forget & Jansen 2007, Holbrook &

Loiselle 2009). The reason for those differences is primarily that the large-bodied seed

dispersers are both especially vulnerable and extremely important in ingesting and

dispersing species with large fruits and/or seeds (Janzen & Martin 1982, Wheelwright

1985, Chapman et al. 1992, Guimarães et al. 2008, Donatti et al. 2011), given the relation

between gape and body size (Lord et al. 2002). Additionally, large-bodied seed dispersers

can also increase the chances of seed dispersal because they spend more time foraging on

plants (Gross-Camp et al. 2009), have larger seed loads per scat (Fragoso 1997, Galetti et

62

al. 2001, Bizerril et al. 2005) and disperse seeds further from the mother plant (Fragoso

et al. 2003) than do small-bodied species. Thus, the absence or low abundance of large-

bodied species is critical, especially for the dispersal of large-seeded plants, since for

many of those plants large-bodied species provide seed dispersal services that no other

remaining species is able to provide (Peres & Roosmalen 2002, Poulsen et al. 2002).

Furthermore, as fruit and seed size usually vary within individual plants (Gorchov

1985, Wheelwright 1993, Mazer & Wheelwright 1993) and as large conspecific seeds can

be selected by large dispersers or conspecific dispersers with larger sizes (Kaspari 1996,

Rey et al. 1997, Alcántara et al. 2000), one can hypothesize that large-bodied animal

species not only are effective dispersers of large-seeded species but also selectively

disperse large conspecific seeds. In this study, we tested these relationships by examining

the association between the size (proxied by body mass) of the seed disperser and the size

of the dispersed seeds, both at the plant intraspecific and interspecific levels, in a plant-

animal community in the Brazilian Pantanal. Furthermore, within a plant species, we

assessed the importance of seed size on seed germination, seedling performance and

survival.

Here, we tested the following hypotheses: 1) at the plant interspecific level, large-

bodied species more frequently interact with large-seeded plants, 2) at the plant

intraspecific level, animal species with the largest body mass disperse seeds with the

largest diameters, 3) seeds that experience gut-passage have an advantage in terms of

germination and survival over seeds that do not experience gut-passage, 4) conspecific

seeds with large diameters have an advantage in terms of germination and survival over

seeds with small diameters, and 5) seedlings emerging from large seeds that experience

gut-passage have an advantage in terms of growth and survival over seedlings emerging

from small seeds that do not experience gut-passage. We show that, at the interspecific

level, the diameter of the dispersed seed increases with the body mass of the seed

disperser. Likewise, at the intraspecific level, the diameter of dispersed seeds of a sample

of species increases with the body mass of the seed dispersers. Our results also show that,

for one particular plant species, seed germination increases with seed diameter and the

combination of a large diameter and gut-passage increases seedling growth, both in

controlled and in field conditions. We thus conclude that large-bodied animals are not

63

only important because they disperse large-seeded species (Janzen & Martin 1982,

Janson 1983, Wheelwright 1985, Chapman et al. 1992, Guimarães et al. 2008, Donatti et

al. 2011), but also because they interact with these plant species more frequently and

disperse larger conspecific seeds, leading to increased germination.

Methods

Study site

The Pantanal, located in central western Brazil and part of Bolivia and Paraguay,

is the world’s largest freshwater wetland, covering an area of 140,000 km2 (Swarts 2000).

Due to the low human population density and low hunting pressure on native species

(Alho & Lacher 1991, Desbiez et al. 2011, but see Harris et al. 2005), the Pantanal still

holds the highest concentration of wildlife in the Neotropics (Swarts 2000, Mittermeier et

al. 2005), which makes it easy to survey interactions between seed dispersers and fleshy-

fruited plant species. Within the Pantanal, this study concentrated on the areas within two

neighboring farms (Rio Negro and Barranco Alto). Rio Negro Farm (19°34'15"S

56°14'43"W) has 7,500 ha and Barranco Alto Farm (19º34'40"S 56º09'08"W) has 11,000

ha, and they are both located at the Nhecolândia region, one of the most pristine regions

within the Pantanal.

Seed dispersal interactions

Events of seed dispersal interaction were surveyed using three different methods.

To sample seed dispersal by birds, we carried out focal observations in trees, recording

the identity of bird species that visited the plants, as well as the number of fruits that were

ingested “in situ” or removed outside the tree canopy area. To record seed dispersal by

several species of mammals, rheas (Rhea americana) and red-footed tortoises

(Geochelone carbonaria), we collected their scats, and identified, counted and measured,

the intact seeds in them. To record seed dispersal by the pacu fish (Piaractus

mesopotamicus), we caught them and identified, counted and measured the intact seeds in

the intestine. We used our seed collection as a reference to identify seeds found in scats

and intestines, and our data base on fruit morphology to assess the average seed diameter

64

of each plant species. Using a caliper we measured, to the nearest 0.01 mm, seeds

collected directly from plant species and all intact seeds found in scats of the mammal,

bird and reptile species, and in intestines of the fish species. In our analysis, we used seed

diameter to represent seed size and the smallest measurement of seed size to represent

seed diameter if the seed was rounded.

Diameter of the dispersed seeds - seed disperser body mass interspecific associations

To compute the average diameter of all seeds (regardless of the species) dispersed

by each animal, we first counted the number of seeds from each plant species that were

dispersed by each animal species. Then, we multiplied the number of seeds of each plant

species by its average seed diameter, and averaged them based on the total number of

seeds dispersed by each animal species. Finally, we used the average diameter of all

seeds dispersed by each animal species to test the association between the average

diameter of the dispersed seeds and body mass of seed dispersers using a linear

regression. We gathered animal body mass information from the literature.

Diameter of the dispersed seeds - seed disperser body mass intraspecific associations

We compared the diameter of conspecific seeds of seven plant species found in the

scats or intestine of various species, including three mammal species (Tapirus terrestris,

240 kg; Sus scrofa, 50 kg and Cerdocyon thous, 5.7 kg), one bird species (Rhea

americana, 20 kg), one reptile species (Geochelone carbonaria, 6.6 kg) and one fish

species (Piaractus mesopotamicus, 1.1 kg). We identified and measured all intact seeds

found in scats of the mammal, bird and reptile species, and in intestines of the fish

species. The plant species for which seeds could be found in the scats and/or in the

intestines of two or more seed disperser species were the following: Dipteryx alata

(Fabaceae, average seed diameter ± SE=32.41±0.21), Attalea phalerata (Arecaceae

22.48±0.21), Acrocomia aculeata (Arecaceae, 21.70±0.45), Copernicia alba (Arecaceae,

13.40±0.14), Bactris glaucescens (Arecaceae, 11.88±0.06), Byrsonima orbignyana

(Nyctaginaceae, 6.32±0.1) and Vitex cymosa (Verbenaceae, 8.46±0.13). All those plant

species produce single-seeded fruits and, therefore, fruits have to be swallowed whole for

65

seeds to be dispersed. As there are positive correlations between seed diameter and fruit

diameter in all these plant species (Dipteryx alata: F=36.44, p<0.0001, R2=0.4587, n=38;

Attalea phalerata: F=193.01, p<0.0001, R2=0.71, n=75; Acrocomia aculeata: F=39.29,

p<0.0001, R2=0.81, n=21; Copernicia alba: F=129.9, p<0.0001, R2=0.73, n=50; Bactris

glaucescens: F=19.28, p<0.0001, R2=0.19, n=81; Byrsonima orbignyana: F=33.19,

p<0.0001, R2=0.43, n=45 and Vitex cymosa: F=36.75, p<0.0001, R2=0.52, n=35;), by

assessing seed diameter we are indirectly assessing fruit diameter.

Effect of seed size and gut-passage on seed germination of Dipteryx alata

For one of these plant species, D. alata, we tested the effect of seed diameter, the

presence of pulp and the gut-passage on the percentage and speed of seed germination.

We planted 50 whole fruits (including seed and pulp) collected directly from trees

(estimated range in seed diameter: 27.52-37.85 mm), 100 seeds collected directly from

trees that had the pulp manually removed (range in seed diameter: 25.42-39.56 mm) and

96 seeds collected from tapir scats that had the pulp removed as they passed through the

animal gut (range in seed diameter: 24.16-34.78 mm). All fruits and seeds were planted

with 1cm-depth from the soil surface, in black plastic bags. Before planting, all seeds

were subjected to a flotation test, following Vallejo et al. (2006) whereby seeds were

submerged in a mixture of water and detergent to discard the ones that floated and would

probably be empty or damaged and not able to germinate. All planted seeds and fruits

were kept at the same weather conditions and we watered them every three days. We

checked all planted seeds daily to record seed germination speed.

We compared seed germination (number of seeds that germinated) among

treatments (seeds with pulp, seeds without pulp and seeds dispersed by tapirs) using a

chi-square test, and seed germination speed (number of days from plantation to

germination) among the three treatments using ANOVA. To test the effect of gut-passage

(seeds dispersed by tapir vs. control) on seed germination, we pulled together seeds that

were planted with pulp and seeds that had the pulp manually removed, considering all of

them as the control treatment. We then tested the association between seed diameter and

seed germination, for all treatments together and for the tapir and the control treatments

66

independently, using logistic regressions. Likewise, we tested the effect of seed diameter

on germination speed, also for all treatments together and for the tapir and the control

treatments independently, using linear regressions.

Effect of seed size and gut passage on seedling survival and growth rates of D. alata in

controlled conditions

After germination of D. alata, we monthly recorded the height, basal diameter,

leaf area and the number of leaflets of each seedling. Height was recorded from the collet

(the external demarcation between the stem and the root) to the petiole of the youngest

leaf. The basal diameter was recorded on the collet. Since there was a positive and

significant correlation between leaf area and the sum of the product of leaflet width and

leaflet length (Pearson's r=0.93, p<0.0001, n=50), we recorded, once a month, the

diameter and the length of every leaflet and used the sum of the product of diameter and

length of each leaflet as a proxy for seedling leaf area. As not all seeds germinated on the

same day, we used seedling length in mm/day, basal diameter in mm/day, leaf area in

mm2/day and number of leaflets/day to assess seedling growth. We then tested the

associations between seed diameter and each of the variables that assessed seedling

growth, for all treatments together and independently for the tapir and the control

treatments, using linear regressions. We also used another approach to address

differences in seedling growth as a function of seed size. We defined small seeds as those

with diameter <29 mm and large seeds those with diameter > 31mm. We binned seeds

using those values to get about the same number of seeds in each category and in each

treatment. Using t-tests, we then compared growth rate variables in seedlings emerging

from small and large seeds using both all data together (i.e. pooling tapir and control

treatments) and data from the tapir and the control treatments independently.

To test the effect of the treatment, we used t-tests to compare the seedling growth

variables between seedlings emerging from seeds from the tapir and the control

treatments. We also tested the association between seed diameter and the survival of

seedlings, regardless of the treatment, using logistic regressions calculated for every

month from seed germination to seedling transplantation to the field. Similarly, we tested

67

the association between seed diameter and the survival of seedlings, independently for

the tapir and the control treatments, using logistic regressions calculated for every month

from seed germination to seedling transplantation to the field. To compare seedling

survival between tapir and control treatments every month before transplantation, we

used chi-square tests.

Effect of seed size and gut passage on seedling survival and growth rates of D. alata in

field conditions

Since there are differences in seedling growth between controlled and field

conditions (see Marshall 1986), we also measured seedling growth in the field.

Approximately six months after seeds were planted, we transplanted seedlings to the field

and followed their fate and growth in natural conditions. We planted all seedlings in the

field, from both control and tapir treatments, around one single adult of D. alata. After

transplanting seedlings to the field, we recorded seedling survival and measured height,

basal diameter, leaf area and number of leaflets for every seedling on a monthly-basis, for

12 months after transplantation. We then tested the associations between seed diameter

and both seedling growth (measured as seedling length in mm/day, basal diameter in

mm/day, leaf area in mm2/day and number of leaflets/day) for all seedlings together and

independently for the tapir and control treatments, using linear regressions.

Additionally, we compared seedling growth between large and small seeds (as

previously defined), combining all data together (i.e. pooling tapir and control treatments)

and considering data for the tapir and control treatments independently, using t-tests.

Associations between seed diameter and the survival of seedlings were assessed through

logistic regressions, calculated for every month from seedling transplantation to the field

to the end of the experiment. Logistic regressions were calculated for all data together

and for the tapir and the control treatments independently. We compared seedling

survival between tapir and control treatments every month after transplantation, using

chi-square tests.

68

Results

Diameter of the dispersed seeds - seed disperser body mass interspecific associations

We recorded interactions between 34 seed disperser species (Table 1) and 40

plant species (Table 2). We found a positive and significant association between the

average diameter of the dispersed seeds and the body mass of seed dispersers (F=64.09,

p<0.0001, r=0.81, d.f.=33, Fig. 1). Thus, animal species with large body mass disperse

seeds with large diameters more often than do small-bodied species.

Diameter of the dispersed seeds - seed disperser body mass intraspecific associations

We measured 7,983 seeds of seven plant species that could be identified in the

scats of two or more seed disperser species. Three of these plant species showed a

positive association between body mass of the seed disperser and the diameter of the

dispersed seeds, i.e. large-bodied species disperse large conspecific seeds. We found that,

for these three plant species, the largest animal species were able to disperse seeds with

the largest diameters (Dipteryx alata: F=50.82, p<0.0001; Copernicia alba: F=44.22,

p<0.0001; Vitex cymosa: F=10.27, p<0.0001; Fig. 2). Three other plant species showed a

difference in the diameter of seeds dispersed by the different animal species. However

this difference was not associated with the body mass of the seed dispersers (Attalea

phalerata: F=51.29, p<0.001; Acrocomia aculeata: F=4.9087, p=0.0022, Byrsonima

orbignyana: F=30.60, p<0.0001), i.e. large-bodied seed dispersers did not necessarily

disperse large conspecific seeds. One plant species did not show a significant difference

in the diameter of the seeds dispersed by different animal species (Bactris glaucescens:

F=8.98, p=0.0001).

Effect of seed size and gut passage on seed germination of D. alata

There was a difference in seed germination among the three treatments

(X2=10.985, p=0.041, d.f.=229). Germination in seeds collected in tapir scats was

33.75%, compared to 20% in seeds that had the pulp manually removed and to 10% in

seeds that were planted with pulp. Therefore, pulp removal increases seed germination,

69

but this is even higher in seeds that passed through the animal’s digestive tract. Seeds

dispersed by tapirs germinated faster than those that had the pulp manually removed and

those that were planted with pulp (F=5.81, p=0.005, n=52; 41.8 days, 49 days and 58

days, respectively). Thus, seeds dispersed by tapirs, those that combine pulp removal and

gut passage, showed the highest percentage of germination and germination speed among

the three treatments.

Only for seeds dispersed by tapirs was there a negative and significant effect of

seed diameter on seed germination speed (F=4.33, p=0.047, r= 0.38, n=27, Fig. 3). That

is, the number of days necessary for germination decreases with the diameter of the seed.

There was no such effect for seeds in the control treatment. When pooling all seeds

together, germination increased with the diameter of seeds (X2=3.79, p=0.049, n=230)

but there was no effect when considering tapir and control treatments independently

(X2=0.05, p=0.82, n=80 and X2=1.21, p=0.27, n=150; respectively).

Effect of seed size and gut passage on seedling survival and growth rates of D. alata in

controlled conditions

About six months after seed plantation and only for seeds dispersed by tapirs,

seedling growth measured as seedling basal diameter in mm/day increased with seed

diameter (F=4.72, p=0.042, r=0.46, n=19, Fig. 4). Likewise, large seeds dispersed by

tapirs generated seedlings that showed higher growth (basal diameter in mm/day) than

seedlings generated by small seeds that were also dispersed by tapirs (t-test=2.86,

p=0.0124, n=16). However, the opposite trend was found for seedlings generated by

seeds in the control treatment. Seedlings generated from large seeds showed a lower

growth (basal diameter in mm/day) than seedlings generated by small seeds in the control

treatment (t-test=2.54, p=0.027, n=13).

When analyzing both treatments together, seedling survival did not increase with

seed diameter in any month before seedling transplantation to the field. However, in two

time-points (two and five months after seed plantation), survival in seedlings generated

by seeds dispersed by tapir was higher than those generated by seeds from the control

treatment (X2=3.44, p=0.046, n=180 and X2=4.06, p=0.033, n=180; respectively).

70

Effect of seed size and gut passage on seedling survival and growth rates of D. alata in

field conditions

Three months after transplantation, the diameter of the seeds dispersed by tapirs

showed a positive effect on seedling growth (basal diameter in mm/day) (F=4.93,

p=0.0412, r=0.48, n=18). In addition, growth (basal diameter in mm/day) in seedlings

generated from large seeds that were dispersed by tapirs was higher than the growth in

seedlings generated by small seeds that were also dispersed by tapirs (t-test=2.192,

p=0.04, n=14). Six months after seedling transplantation, the diameter of the seeds

dispersed by tapirs still showed a positive effect on seedling growth (in terms of basal

diameter/day) (F=4.47, p=0.05, r=0.49, n=16, Fig. 4). The effect of seed diameter on

seedling growth among seeds dispersed by tapir continued to occur twelve months after

we transplanted those seedlings to the field (F=11.56, p=0.042, r=0.88, n=5). However,

there was a negative effect of seed diameter on the growth (basal diameter in mm/day) of

seedlings generated from seeds in the control treatment (F=7.39, p=0.0216, r=0.65, n=7).

Although all variables measured in seedlings were correlated, we could not find an effect

of seed size and/or gut passage in any other variables. Likewise, there were no significant

differences in seedling growth between the tapir and the control treatments.

Approximately nine months after seedling transplantation to the field, there was a

positive and significant association between seed diameter and the survival of seedlings

generated by those seeds (X2=4.85, p=0.0276, n=40) when considering all seedlings

regardless of the treatment. Likewise, eight, nine and ten months after seedling

transplantation to the field, seedling survival increased with seed diameter, but only if we

consider seedlings dispersed by tapirs (X2=3.93, p=0.047, n=21; X2=4.53, p=0.033, n=21,

X2=4.53, p=0.033, n=21, respectively). Twelve months after seedling transplantation to

the field, 26.25% of the seedlings generated from seeds dispersed by the tapir were alive,

followed by 17% of the seedlings generated from seeds planted without pulp and by 6%

of the seedlings from seeds planted with pulp (F=4.45, p=0.0126, n=230). However,

when considering only seedlings that were transplanted to the field, there was no

difference in seedling survival between tapir and control treatments in any month until

the end of the experiment. Likewise, there was no effect of seed diameter on seedling

survival at the end of the experiment, i.e. twelve months after seedling transplantation.

71

Discussion

Our data show that, at both the interspecific and intraspecific levels, animal species

disperse seeds non-randomly with respect to seed diameter. At the plant interspecific

level, there is a positive and significant association between the body size of seed

dispersers and the diameter of the dispersed seeds, that is, large-bodied species tend to

interact more often with large-seeded species. Likewise, at the plant intraspecific level,

there is also a positive and significant association between the body size of seed

dispersers and the diameter of the dispersed seeds, that is, the largest animal species tend

to interact more often with the largest seeds of a given plant species. Although previous

studies have shown a positive association between the body mass of dispersers and the

average seed diameter of plant species dispersed by those animals (see Donatti et al.

2011), and that seeds dispersed by birds and mammals can be different than those

collected directly from trees (Howe &Vander Kerckhove 1980, Wheelwright 1993,

Wheelwright 1985, Rey et al. 1997, Stevenson et al. 2005, Côrtes et al. 2009, Traba et al.

2006), to our knowledge this is first study that reports the association between the animal

body mass and the diameter of the dispersed seeds within plant species (but see Kaspari

1996 for a similar study with ants).

In our study, as all seed dispersers that interacted with a given plant species could

potentially disperse seeds of all sizes, differences in the diameter of conspecific seeds

dispersed by the different animal species cannot be explained by morphological

mismatching. Instead, our results show that seed disperser species may be competing for

conspecific fruits. Therefore, we suggest that, at least for some plant species, different

animal species are competing, with the result that they select conspecific seeds of

particular diameters, which correspond to their body mass. However, this association

between the diameter of the dispersed seeds and the body mass of the seed disperses was

not diagnosed for all seven plant species studied here. In this particular community,

competition may occur only in plant species that produce fruits outside the peak of fruit

availability in the comunity. Dipteryx alata does not produce fruits at the peak of the

fruiting season, Copernicia alba produces fruits at the end of the fruiting season and

Vitex cymosa at the beginning of the fruiting season, three periods when fruits are scarce

in this community (C.I. Donatti, unpl. data).

72

Although we have measured thousands of seeds recovered from scats, we

acknowledge the possibility that many of them may belong to a single mother plant but

were taken as independent sample points in our analyses. Given that some of the large-

bodied species show long retention times in the gut and/or disperse a large number of

seeds per scats, it is difficult to know how frequently this occurred. However, because we

collected hundreds of scats from each of the animal species, we assume that seeds we

measured were collected from several adult plants and, therefore, are a good

representation of the seed size that different animal species can disperse.

We experimentally assessed the combined effects of seed size and gut passage on

seed germination in controlled conditions and seedling growth and survival in both

controlled and in field conditions. Large seeds show a higher germination rate than small

seeds probably because they have more food reserves, which stimulate germination

(Howe & Richter 1982, Dunlop & Barnett 1983, Tripathi & Kahn 1990). This positive

effect of seed size on seed germination rate was also found in other studies (Pizo et al.

2006, Venkatesh & Nagarajaiah 2010, Seemi & Shaukat 2010, Mut et al. 2010). Our

results also showed that gut passage increases seed germination. The high germination

rate in seeds dispersed by tapirs likely resulted from the removal of pulp and both

mechanical and chemical abrasion during consumption and transit through the gut

(Barnea et al. 1991, Traveset &Verdú, 2002). Furthermore, our study showed that a

combination of a large size with gut passage enables seeds to germinate faster. Rapid

germination has been associated with enhanced seedling survival, due to lower sibling

competition (Hyatt & Evan 1998) and, therefore, a higher availability of resources,

especially light (Ross & Harper 1972, Seiwa 1998) and nutrients (Ross & Harper 1972),

as well as reduced incidence of pathogens and seed predators (Schupp 1993, Seiwa

1998). In addition, early rooting can increase the ability of seedlings to withstand water

stress (Seiwa 2000), which in the Pantanal coincides with the period of fruit production

of Dipteyx alata.

A positive relationship between seed size and seedling growth has also been

reported for other plant species (Weis 1982, Stanton 1984, Weller 1985). By aggregating

more reserve tissues, large seeds can also improve the growth (Venkatesh & Nagarajaiah

2010, Nik et al. 2011, Muhammad et al. 2011) and vigor of seedlings (Pizo et al. 2006,

73

Howe & Richter 1982, Silva e Silva & Carvalho 2008), turning them into stronger

survivors and competitors with respect to other seedlings under distinct environmental

adversities (Howe & Richter 1982, Leishman et al. 2000, Alcántara & Rey 2003, Pizo et

al. 2006). Therefore, a combination of large size and faster germination may enable seeds

and seedlings to maximize their ability to utilize environmental resources (Seiwa 2000).

Here, we assessed the importance of body size of animal species in the process of

seed dispersal at both interspecific and intraspecific levels, and discussed about the

effects that animal extinctions may have on the early stages of plant development and,

ultimately, on plant populations. The combination of a large diameter and the gut passage

seems to be advantageous to the seed, in terms of seed germination rate and speed, as

well as seedling growth. Therefore, if seed dispersal is a process that increases plant

fitness and given that seed size is a heritable trait (Leishman et al. 1995), an evolutionary

change should be expected with the absence of the largest seed dispersers from the

community. For example, if seedlings generated from large seeds that also experienced

gut passage have a high survivorship, and if individual trees are consistently producing

seeds of similar diameters in consecutive years, we expect individual plants that produce

large-sized seeds to be negatively affected if the largest seed dispersers are locally

extinct. We conclude that large-bodied animals are not only important because they

disperse large-seeded species, but also larger conspecific seeds, leading to increased

germination. Since contemporary defaunation differentially affects animal species

depending on body size, this work illustrates how human activities, such as hunting,

deforestation and climate change, my affect not only taxa, but also crucial processes in

which animals of different body size play different roles.

Acknowledgments

We would like to thank Lee Love-Anderegg, Adilson Braga Samuel and Luisa Haddad

for their help with the experiment. We thank FAPESP (2004/00810-3 and 2008/10154-7),

Earthwatch Institute and Conservation International for financial support. CID was

supported by Stanford University and the Zaffaroni Fellowship Fund, and MG and MAP

by a CNPq fellowship. We also thank Conservation International, Lucas Leuzinger and

Marina Schweizer for their permission to work in their properties.

74

References

Alcántara JM, Rey PJ (2003) Conflicting selection pressures on seed size: evolutionary ecology of fruit size in a bird-dispersed tree, Olea europaea. J Evolution Biol 16(6): 1168-1176.

Barnea A, Yom-Tov Y, Friedman J (1991) Does ingestion by birds affect seed germination? Funct Ecol 5: 394–402.

Bizerril MXA, Rodrigues FHG, Hass A (2005) Fruit consumption and seed dispersal of Dimorphandra mollis Benth. (Leguminosae) by the lowland tapir in the cerrado of Central Brazil. Braz J Biol 65: 407–413.

Bodmer RE, Eisenberg JF, Redford KH (1997) Hunting and the likelihood of extinction of Amazonian mammals. Conserv Biol 11: 460–466.

Cardillo M, Mace G M, Gittleman J L, Purvis A (2006) Latent extinction risk and the future battlegrounds of mammal conservation. Proc Natl Acad Sci USA 103(11): 4157-4161.

Cardillo M, Mace GM, Jones KE, Bielby J, Bininda-Emonds ORP, Sechrest W, Orme C DL, Purvis A (2005) Multiple causes of high extinction risk in large mammals species. Science 309 (5738): 1239-1241.

Chapman LJ, Chapman CA, Wrangham RW (1992) Balanites wilsoniana: Elephant Dependent Dispersal? J Trop Ecol 8 (3): 275-283. 1992.

Clark JS, Macklin E, Wood L (1998) Stages and spatial scales of recruitment limitation in southern Appalachian forests. Ecol Monograph 68 (2): 213-235.

Connell J (1971) On the role of natural enemies in preventing competitive exclusion in some marine animals and in rain forest trees. In P.J.D. Boer and G.R. Gradwell (Eds.). Population Dynamics, pp. 298-310. Centre for Agricultural Publishing and Documentation: Wageningen

Connell JH, Green PT (2000) Seedling dynamics over thirty-two years in a tropical rain Forest tree. Ecology 8(2): 568-584.

Corlett RT (2007) The impact of hunting on the mammalian fauna of tropical Asian forests. Biotropica 39(3): 292-303.

Côrtes, MC, Cazetta E, Staggemeier VG, Galetti M (2009). Linking frugivore activity to early recruitment of a bird dispersed tree, Eugenia umbeliflora (Myrtaceae) in the Atlantic rainforest. Austral Ecology 41: 249-258.

Cramer JM, Mesquita R, Williamson GB (2007) Forest fragmentation differentially affects seed dispersal of large and small-seeded tropical trees. Biological Conservation 137:415–423.

Dirzo R (2001) Plant-mammal interactions: Lessons for our understanding of nature and implications for biodiversity conservation. In M. C. Press, N. J. Huntly, and S. Levin (Eds.). Ecology: achievement and challenge, pp. 319–335. Blackwell Science, Oxford, UK.

Dirzo R, Miranda A (1991) Altered patterns of herbivory and diversity in the forest understory: A case study of the possible consequences of contemporary defaunation. In P. W.Price, T. M.Lewinsohn, G. W.Fernandes, and W. W.Benson (Eds.). Plant-animal interactions: evolutionary ecology in tropical and temperate regions, pp. 273–287. John Wiley and Sons, New York , New York.

75

Donatti CI, Guimarães PR, Galetti M, Pizo MA, Marquitti FMD, Dirzo R (2011) Analysis of a hyper-diverse seed dispersal network: modularity and underlying mechanisms. Ecol Lett 14: 773-781.

Dunlop J R, Barnett JP (1983) Influence of seed size on germination and early development of loblolly pine (Pinus taeda L.) germinants. Can. J. For. Res. 13: 40-44.

Ehrlén J, Munzbergova Z, Diekmann M & Eriksson O Long-term assessment of seed limitation in plants, results from an 11-year experiment. J Ecol 94: 1224-1232. 2006.

Forget P-M & Jansen PA Hunting increases dispersal limitation in the tree Carapa procera, a nontimber forest product. Conserv Biol 21 (1): 106–113. 2007.

Fragoso JMV, Silvius KM, Correa JA (2003) Long-distance seed dispersal by tapirs increases seed survival and aggregates tropical trees. Ecology 84(8): 1998–2006.

Fragoso JMV (1997) Tapir-generated seed shadows: scale-dependent patchiness in the amazon rain forest. J Ecol 85: 519–529.

Galetti M, Donatti CI, Pires AS, Guimarães PR, Jordano P (2006) Seed survival and dispersal of an endemic Atlantic forest palm: the combined effects of defaunation and forest fragmentation. Bot J Linnean Soc 151: 141–149.

Galetti M, Keuroghlian A, Hanada L, Morato MI (2001) Frugivory and seed dispersal by the lowland tapir (Tapirus terrestris) in southeast Brazil. Biotropica 33: 723–726.

Gorchov DL (1985). Fruit ripening asynchrony is related to variable seed number in Amelanchier and Vaccinium. Am J Bot 72: 1939-1943.

Guimarães PR, Galetti M, Jordano P (2008). Seed dispersal anachronisms: rethinking the fruit extinct megafauna ate. PLOS One 3: e1745.

Gross-Camp ND, Mulindahabi F, Kaplin BA (2009) Comparing the Dispersal of Large-seeded Tree Species by Frugivore Assemblages in Tropical Montane Forest in Africa. Biotropica 41 (4): 442 – 451.

Holbrook KM, Loiselle BA (2009) Dispersal in a Neotropical tree, Virola flexuosa (Myristicaceae): does hunting of large vertebrates limit seed removal? Ecology 90 (6):1449-1455.

Herrera CM (1987) Vertebrate–dispersed plants of the Iberian Peninsula: a study of fruit characteristics. Ecol Monograph 57: 305–331.

Herrera CM, Jordano P (1981). Prunus mahaleb and birds: the high-efficiency seed dispersal system of a temperate fruiting tree. Ecol Monograph 51:203–218.

Howe FH, Richter WM (1982) Effects of seed size on seedling size in Virola surinamensis: A within and between tree analysis. Oecologia 53(3): 347-351.

Howe HF, Vandekerckhove GA (1980) Fecundity and seed dispersal of a tropical tree. Ecology 60(1): 180-189.

Howe HF, Smallwood J (1982). Ecology of seed dispersal. Ann. Rev. Ecol. Syst. 13: 201–228.

Hubbell SP (1979) Tree dispersion, abundance, and diversity in a tropical dry Forest. Science 203: 1299-1309.

Hyatt LA. Evans AS (1998) Is decreased germination fraction associated with risk of sibling competition? Oikos 83: 29-35.

Janson CH (1983) Adaptation of fruit morphology to dispersal agents in a neotropical forest. Science 219: 187–189.

76

Janzen DH, Martin PS (1982) Neotropical anachronisms-the fruits the gomphoteres ate. Science 215: 19-27.

Janzen DH (1970) Herbivores and the number of tree species in tropical forests. Am Nat 104: 501-528.

Kaspari M (1996) Worker size and seed size selection by harvester ants in a neotropical forest. Oecologia 105: 397-404.

Leishman MR, Westoby M, Jurado E (1995) Correlates of seed size variation: a comparison among five temperate floras. J Ecol 83: 517-530.

Leishman MR, Wright IJ, Moles AT, Westoby M (2000) The evolutionary ecology of seed size. In Seeds: The Ecology of Regeneration in Natural Plant Communities (ed. M. Fenner) pp. 31–57. Commonwealth Agricultural Bureau International, Wallingford.

Lord JM, Markey AM, Marshall J (2002) Have frugivores influenced the evolution of fruit traits in New Zealand? In Seed Dispersal and Frugivory: Ecology, Evolution and Conservation. D. J. Levey, W. R. Silva & M. Galetti (eds) pp. 55–68. CABI Publishing, Wallingford, UK.

Marshall D L (1986) Effect of seed size on seedling success in three species of Sesbania (Fabaceae). Am J Bot 73: 457-464. 1986.

Mazer SJ, Wheelwright NT (1993). Fruit size and shape: allometry at different taxonomic levels in bird-dispersed plants. Evol Ecol 7(6): 556-575.

Mittermeier RA, Harris MB, Mittermeier CG, Da Silva JMC, Lourival R, Da Fonseca GAB, Seligmann P, Allofs T (2005) Pantanal: South America's Wetland Jewel, 176 p. Firefly Books Ltd. New York.

Mut Z, Akay H Aydin N (2010). Effects of seed size and drought stress on germination and seedling growth of some oat genotypes (Avena sativa L). African Journal of agricultural research 5(10): 1101-1107.

Muhammad A, Khan D, Zaki MJ, Khan MQ (2011). Seed mass variation and its effects on germination, seedling growth and the root infectivity with Macrophomina phaseolina in sunflower (Helianthus annuus L.). International Journal of Biology and Biotechnology 8 (1):155-165.

Nik, MM, Babaeian M, Tavassoli A (2011). Effect of seed size and genotype on germination characteristic and seed nutrient content of wheat. Scientific Research and Essays 6(9): 2019-2025.

Peres CA, Palácios E (2007) Basin-Wide Effects of Game Harvest on Vertebrate Population Densities in Amazonian Forests: Implications for Animal-Mediated Seed Dispersal. Biotropica 39: 304-315.

Peres CA, van Roosmalen M (2002) Patterns of primate frugivory in Amazonia and the Guianan shield: implications to the demography of large-seeded plants in overhunted tropical forests. In D. Levey, W. Silva and M. Galetti (Eds.). Frugivory and Seed Dispersal: Ecological, Evolutionary and Conservation Issues, pp 407-423. CAB International. Oxford.

Peres C A (2000) Effects of subsistence hunting on vertebrate community structure in Amazonian forests. Conserv Biol 14: 240–253.

Pizo, MA, Von Allmen C, Morellato LPC (2006) Seed size variation in the palm Euterpe edulis and the effects of seed predators on germination and seedling survival. Acta Oecol 29(3): 311-315.

77

Poulsen JR, Clark C J, Connor EF, Smith TB (2002) Differential resource used by primates and hornbills: implications for seed dispersal. Ecology 83: 228-240.

Prance GT, Schaller GB (1982) Preliminary study of some vegetation types of the Pantanal, Mato Grosso, Brazil. Brittonia 34(2): 228-251.

Redford KH (1992) The empty forest. Bioscience 42: 412–422. Rey PJ, Gutiérrez JE, Alcántara J, Valera R (1997) Fruit size in wild olives: implications

for avian seed dispersal. Func Ecol 11: 611–18. Ross MA, Harper JL (1972) Occupation of biological space during seedling

establishment. Ecology 60:77-88. Seemi A, Shaukat SS (2010) Effect of seed mass variations on the germination and

survival of three desert annuals. Pakistan Journal of Botany 42(4): 2813-2825. Venkatesh L, Nagarajaiah C (2010) Effect of seed size on germination, viability and

seedling biomass in Sapindus emerginatus (Linn.). Environment and Ecology 28(1): 25-27.

Schupp EW (1993) Quantity, quality and the effectiveness of seed dispersal by animals. Vegetatio 107/108: 13-29.

Seiwa K (2000) Effects of seed size and emergence time on tree seedling establishment: importance of developmental contraints. Oecol 123: 208-215.

Seiwa K (1998) Advantages of early germination for growth and survival of seedlings of Acer mono under different overstorey phenologies in deciduous broad-leaved forests. J Ecol 86: 219–228

Silva JMC, Tabarelli M (2000) Tree species impoverishment and the future flora of the Atlantic forest of northeast Brazil. Nature 404:72–74.

Silva e Silva BM, Carvalho NM (2008) Seed size and water stress effects on seed germination and seedling vigor of faveira (Clitoria fairchildiana) Revista Brasileira de Sementes 30(1): 55-65.

Stanton M (1984) Seed variation in wild radish: effect of seed size on components of seedling and adult fitness. Ecology 65: 1105–1112.

Stevenson PR, Pineda M, Samper T (2005) Influence of seed size on dispersal patterns of woolly monkeys (Lagothrix lagothricha) at Tinigua Park, Colombia. Oikos 110: 435-440.

Swarts FA (2000) The Pantanal in the 21st century—for the planet’s largest wetland, an uncertain future. In F. A. Swartz (Ed.). The Pantanal of Brazil, Paraguay and Bolivia, pp. 1-24. Hudson MacArthur Publishers, Gouldsboro. Pennsylvania.

Traba J, Arrieta S, Herranz J, Clamagirand MC (2006) Red fox (Vulpes vulpes L.) favor seed dispersal, germination and seedling survival of Mediterranean Hackberry (Celtis australis L.). Acta Oecol 30(1): 39-45.

Traveset A, Verdú M (2002). A meta-analysis of the effect of gut-treatment on seed germination. In Levey, D.J., Silva, W.R., Galetti, M. (Eds.), Seed Dispersal and Frugivory: Ecology, Evolution and Conservation, pp. 339–350. CAB International, Wallingford, U.K.

Van der Pijl L (1982) Principles of dispersal in higher plants.199 p. Springer-Verlag, Berlin.

Vallejo, M., Dominguez, C.A., and Dirzo, R. 2006. Simulated seed predation reveals a variety of germination responses of neotropical rain forest species. American Journal of Botany 93: 369-376.

78

Wenny DG, Levey DJ (1997) Directed seed dispersal by bellbirds in a tropical forest. Proc Natl Acad Sci USA 95: 6204-6207.

Wheelwright NT (1985) Fruit size, gape width, and the diets of fruit eating birds. Ecology 63(3): 808-818.

Weller SJ (1985) Establishment of Lithospermum carolinense on sand dunes: the role of nutlet mass. Ecology 66: 1893–1901.

Weis Y M (1982) The effects of propagule size on germination and seedling growth. Can J Bot 60: 1868–1874.

Wheelwright NT (1985) Fruit size, gape width, and the diets of fruit-eating birds. Ecology 66: 808-818.

Wheelwright NT (1993) Fruit size in a tropical tree species: variation, preference by birds, and heritability. Vegetatio 107/108: 163-174.

Wilkie DS, Bennett EL, Peres CA, Cunningham AA (2011) The empty forest revisited. Ann N Y Acad of Sci 1223: 120-128.

Wright SJ (2003) The myriad consequences of hunting for vertebrates and plants in tropical forests. Perspect Plant Ecol 6: 73-86. 2003.

79

Tables

Table 1. Seed disperser species, their body masses and the average diameter of the seeds

dispersed by them in a community in the Brazilian Pantanal.

Seed disperser Body mass (kg) Mean diameter

of dispersed seeds (mm) Aburria pipile 1 5.18 Alouatta caraya 5.4 7.25 Turdus sp. 0.05 4.98 Casiornis rufa 0.02 3.3 Columba sp. 0.14 4.79 Crypturellus sp. 0.08 4.82 Cyanocorax chrysops 0.14 1.73 Cyanocorax cyanomelas 0.2 5.48 Geochelone carbonaria 6.6 5.07 Guira guira 0.07 6.83 Icterus croconotus 0.07 4.36 Myiarchus ferox 0.024 3.6 Myiodinastes maculatus 0.043 4.82 Nasua nasua 5.1 9.12 Ortalis canicollis 0.6 4.08 Piaractus mesopotamicus 1.14 9.02 Pitangus sulphuratus 0.068 6.28 Paroaria coronata 0.03 4.36 Psarocolius decumanus 0.25 4.9 Pteroglossus castanotis 0.27 7.64 Ramphastos toco 0.54 8.71 Ramphocelus carbo 0.025 2.68 Rhea americana 20 12.97 Saltator coerulescens 0.052 5.67 Sus scrofa 50 10.23 Tachyphonus rufus 0.033 6.01 Tapirus terrestris 240 17.9 Tayassu pecari 35 9.57 Thraupis palmarum 0.036 5.11 Thraupis sayaca 0.03 3.77 Trogon curucui 0.07 4.69 Turdus rufiventris 0.07 4.06 Tyrannus melancholicus 0.039 4.36 Tityra cayana 0.07 4.82

80

Table 2. Plant species, and their average seed diameter, recorded interacting with seed

dispersers in the Brazilian Pantanal.

Plant species Mean seed diameter (mm) Acrocomia aculeata 21.34 Attalea phalerata 22.08 Agonandra brasiliensis 14.63 Alibertia sessilis 5.08 Annona cornifolia 4.66 Annona dioica 8.27 Bactris glaucescens 8.53 Byrsonima obrgnyana 6.06 Copernicia alba 14.16 Curatella americana 5.00 Dipteryx alata 35.73 Doliocarpus dentatus 4.37 Diospyros hispida 12.93 Genipa americana 6.01 Guazuma ulmifolia 1.78 Hancornia speciosa 10.87 Inga laurina 9.92 Syzydium cumini 9.96 Licania parvifolia 8.33 Melicoccus lepidopetalus 11.20 Mouriri elliptica 13.27 Ocotea diospyrifolia 4.82 Phoradendron 0.50 Pouteria gardneri 7.21 Pouteria ramiflora 15.05 Protium heptaphyllum 9.54 Psidium nutans 3.41 Psittacanthus cordatus 8.50 Psittacanthus calyculatus 8.50 Rhamnidium elaeocarpum 4.83 Garcinia brasiliensis 10.12 Sideroxyllum obstusifolium 8.62 Sterculia apetala 14.35 Swartzia jorori 10.67 Syagrus flexuosa 13.70 Tocoyena formosa 6.44 Vitex cymosa 8.97 Zanthoxyllum rigidum 3.30

81

Figures

Figure 1. Relationship between the diameter of dispersed seeds (regardless of plant species identity) and body mass of the seed disperser. Body mass is in log scale.

!"# # #! #!!

$

#!

#$

%&'()*+,,)&-),..')'/,0.1,.1,)2345

6/+*.7.1)&-)'/,0.1,.'),..',

82

Figure 2. Diameter (average±SE) of the seeds dispersed by different animal species, from the heaviest to the lightest. Top) Dipteryx alata, n=324, 627 and 109 seeds measured, respectively. Center) Copernicia alba, n=168, 26 and 674 seeds measured, respectively. Bottom) Vitex cymosa, n=381, 842, 167 and 64 seeds measured, respectively.

83

Figure 3. Number of days from plantation to germination as a function of the diameter of seeds dispersed by tapirs (n=27), in controlled conditions.

!" #" $" %" &"

!$

!&

!'

#"

#!

#$

())*+*,-.)/)0

1-23+/4+5)0.,6-/)

84

Figure 4. Seedling growth (in mm of basal diameter/day) as a function of the diameter (in mm) of seeds dispersed by tapirs, in controlled conditions (top) and in field conditions (bottom).

!" !# !$ %& %! %"

&'&%(

&'&"&

&'&"(

)**+,+-./*0*1

)**+2-34,415607,8//9+.:;

&'&<#

&'&!&

&'&!"

&'&!$

)**+2-34,415607,8//9+.:;

85

CHAPTER 4

THE ROLE OF PLANT AND ANIMAL TRAITS IN DETERMINING

NUMBER AND STRENGTH OF INTERACTIONS ACROSS

SPECIES IN A SEED DISPERSAL NETWORK Camila I. Donatti, Mauro Galetti, Marco Aurélio Pizo & Rodolfo Dirzo

Abstract

Recent studies of mutualistic networks have uncovered emergent properties of species

interactions, such as the nested pattern and the interaction asymmetry between species.

Even though those properties of mutualistic interactions have been validated in several

studies, the importance of evolutionary history and species traits in determining these

properties across species in networks has seldom been investigated at the community

level. Here we address the extent to which evolutionary history and/or species traits

account for the variation in number and strength of interactions across species in a plant-

animal interaction network. We studied a seed dispersal network that includes a broad

range of animals from different taxonomic groups and plants from different families. Our

results show that phylogenetic history of both plant and animal species do not explain the

variation in their properties in the network, which are mainly explained by species

morphological traits. We found that plant species with small fruits and/or seeds have

many interactions and are more important to the animal species set, whereas animal

species with large body mass and/or high densities have many interactions and are more

important to the plant species set. Therefore, we suggest that species-specific

morphological traits of plants and animal species should be considered in studies aimed

at evaluating the ecological and evolutionary consequences of plant-animal interaction

networks.

86

Introduction

Plant–animal mutualistic networks describe not only the trophic relationships

among species (Jordano 1987), but also the complexities of the interactions that drive

coevolutionary processes (Thompson 1999, Jordano et al. 2003, Montoya et al. 2006).

Such interactions among species at the community level can be simplified using the

network approach (Guimera & Amaral 2005, Olesen et al., 2007, Clauset et al. 2008,

Rooney et al. 2008), in which the architecture of the system is reduced to elements

(interacting species) and the interactions between such elements are represented by links

(Ben-Naim et al. 2004). Recent studies of mutualistic networks have uncovered some

general properties of the pattern of species interactions, such as the prevalence of the

nested pattern (Bascompte et al. 2003, Jordano et al. 2003, Guimarães et al. 2006,

Vázquez et al. 2009, Fortuna et al. 2010, Joppa et al. 2010), in which interactions of the

specialist species tend to be a subset of the interactions observed among the generalists.

In addition to being nested, some mutualistic networks are also modular (Dicks et al.

2002 and Olesen et al. 2007: pollination networks; Donatti et al. 2011 and Mello et al.

2011: seed dispersal networks), with subsets of species (modules) more frequently

interacting with each other than with species in other modules (Olesen et al. 2007).

Given that species do not interact with the same intensity and all interactions are

therefore not equally important (Vázquez et al. 2005), studies that incorporate a

measurement of interaction strength have found that interactions between mutualistic

partners are typically asymmetric (Bascompte et al. 2003, Vázquez & Aizen 2004, Stang

et al. 2006, Stang et al. 2007, Vázquez et al. 2007). That is, whereas a focal species

strongly relies on a partner species, this partner weakly relies on the focal species.

Nevertheless, even though those characteristics of mutualistic interactions have been

identified in several studies, the importance of evolutionary history and species-specific

traits in determining the properties of species in networks, especially quantitative ones

such as species strength (i.e. how important a species is to its set of partners), has seldom

been investigated at the community level (Waser et al. 1996, Thompson 2005, Vázquez

2005, Jordano et al. 2006, Stang et al. 2006, Vázquez et al. 2007). This is likely due to

missing evolutionary and/or species-specific trait (e.g., morphological) information for

the majority or for the whole studied community (Olesen & Jordano 2002, Vázquez

87

2005) and to the difficulties attendant on collecting quantitative information about

species interactions in large assemblages of species (Berlow et al. 2004).

A recent study by Rezende and collaborators (2007), analyzing both seed dispersal

and pollination networks, found a phylogenetic signal in the degree of species in

networks (i.e., number of interactions that a given species has) for a third of the analyzed

networks, but not in the species strength (i.e., how important a species is to the set of

partners). Thus, for some networks, phylogenetically related species tend to have a

similar number of seed dispersers or pollinators. However, phylogeny is often not

sufficient to explain the variation in species strength, which may be due to variables that

respond to more local ecological factors, such as the density of species (Bascompte &

Jordano 2007). Additionally, other studies have suggested that, in the absence of an

evolutionary explanation, two main processes may determine some of the properties of

species in networks: interaction neutrality and trait matching. Under interaction neutrality

frameworks, all individuals have the same probability of interaction regardless of their

taxonomic identity (Dupont et al. 2003, Ollerton et al. 2003, Vázquez 2005, Vázquez et

al. 2007, Krishna et al. 2008). That is, common species would interact more frequently

with common species than with rare species. Therefore, the abundance of species is

predicted to primarily determine the properties of species in networks. In contrast, trait

matching depends primarily on the identities of species in the community and their traits,

with species interacting either in a complementary fashion (e.g., more nutritious fruits are

preferred by frugivores) or as a barrier (e.g., plant species with large fruits exclude seed

dispersers with small gape) (Jordano et al. 2003, Rezende et al. 2007, Santamaría &

Rodríguez-Gironés 2007, Stang et al. 2007). Therefore, species-specific traits such as

fruit size and nutritious content in the fruit pulp are predicted to primarily determine such

properties of species in seed dispersal networks.

Here we address the extent to which evolutionary history and/or species-specific

traits account for the properties of species in a seed dispersal network that includes a

broad range of animals from different taxonomic groups and plants from different

families. To do so, we asked: i) if the properties of species in the network (i.e. species

degree, species strength, the value of maximum dependence of a focal species on its most

important partner and interaction asymmetry between species) have a phylogenetic

88

signal, and ii) which processes (interaction neutrality or trait matching) most likely

determine the variation in these properties across species in this seed dispersal network.

The analysis of both morphological constraints and density of species can offer

explanatory power to address the magnitude of the relationship between species traits and

properties in the network. In addition, the analysis of properties of both animals and

plants is also not common in such associations. This study adds a novel approach by

uncovering important information on the properties of species, both animals and plants,

in a highly diverse seed dispersal network. As interactions between fleshy-fruited plants

and seed dispersers are not random in this network (which in fact is highly nested and

modular; Donatti et al. 2011), we expect trait matching to be more important than

interaction neutrality in explaining the properties of plants and animals in this network.

We found that species traits per se, rather than evolutionary history, explain the

properties of species in this network. Furthermore, we found that morphological traits

(seed or fruit size and the body mass of seed dispersers) explain the highest variation in

the majority of quantitative and qualitative properties of species. Therefore, we suggest

that such species-specific traits of species should be considered in studies that evaluate

the ecological and evolutionary consequences of plant-animal interaction networks.

Methods

Properties of species in the seed dispersal network

Seed dispersal interactions were recorded using four methods. To sample seed

dispersal by birds, we carried out focal observations in fruiting trees and palm trees,

recording identity of birds that were unequivocally observed carrying fruits outside the

canopy area or swallowed them in situ. Seed dispersal by red-footed tortoises

(Geochelone carbonaria), rheas (Rhea americana) and the majority of mammal species,

were recorded with camera traps located beneath fruiting trees to capture events of fruit

ingestion. Some terrestrial and semi-terrestrial bird species were also recorded via

pictures taken with camera traps. We analyzed scats of several species of mammals, rheas

and red-footed tortoises, and identified the intact seeds in them. To record seed dispersal

by the pacu fish (Piaractus mesopotamicus), we caught individuals and identified the

89

intact seeds in their intestine. We were able to record interactions between 45 plant

species (Table 1) and 46 animal species (Table 2).

We carried out focal observations of 20 plant species for 984 hours, and we

recorded frugivore activity (via camera traps located beneath fruiting trees of 27 plant

species) for a total of 27,800 hours. We analyzed 1,030 scats of several species of

mammals, rheas and red-footed tortoises, and identified the intact seeds in them. Seed

dispersal by the pacu fish (Piaractus mesopotamicus), was determined from 80 caught

individuals, from which we identified and counted the intact seeds in their intestine (see

Galetti et al. 2008).

Using this data set on species interactions we assembled a qualitative seed

dispersal network, in which an element representing a seed dispersal interaction received

the value of one, and zero otherwise. Using the same data set, we also assembled a

quantitative seed dispersal network, in which we included how many times each

interaction between a plant and an animal species was recorded. One event of seed

dispersal was considered as such when either: i) fruits were recorded to have been

swallowed or removed from a plant species during focal observations; ii) fruit removal of

a particular species by a potential seed disperser was detected with camera traps; iii) a

scat pile was found to have at least one intact seed of a particular species in it; or iv) a

sampled fish intestine contained at least one intact seed from a particular species.

We then used the quantitative and qualitative data to assess the properties of

species in the network. We used the qualitative network to assess the number of seed

dispersal interactions recorded for every plant and animal species, hereafter “species

degree” and the quantitative seed dispersal network to calculate the other three species

properties: the importance of a species to its partners, hereafter “species strength”, the

dependence of a species on its most important partner, hereafter “maximum dependence”

and the asymmetry in the interaction strength between species, hereafter “interaction

asymmetry”. All four properties of species were calculated for all plant and animal

species in the network.

The value of importance of a focal species to a partner species (interaction

strength) is calculated as the number of seed dispersal events recorded between those two

species, divided by the sum of the seed dispersal events recorded for the partner. The

90

species strength of the focal species is then calculated as the sum of all values of

interaction strength of this focal species (Bascompte et al. 2006). The value of

dependence of a focal species to a partner species is calculated as the number of seed

dispersal events recorded between those two species divided by the sum of the seed

dispersal events recorded for the focal species. Thus, the value of maximum dependence

of the focal species is the maximum value calculated for this focal species. Following

Vázquez et al. (2007), we defined interaction asymmetry as the relative difference

between the interaction strength exerted by a focal species on the partner species and the

interaction strength exerted by the partner species on the focal species. Therefore:

Interaction asymmetry of species

€

i =

Jj=1∑ ijs − jis( )

Ni

where Ni is the degree of species i, sij is the interaction strength of species i on species j

and sji is the interaction strength of species j on species i.

Plant and animal species traits

For each plant species, we measured fruit and seed diameter in at least 30 fruits

and seeds from five different individuals. We also collected the pulp of at least 15 fruits

per plant species and measured the milligram of protein, carbohydrates, and lipids in one

gram of dry fruit, later combined as an “energy content of fruit” (kcal/gram of dry pulp),

which refers to the amount of energy that frugivores can obtain from a gram of dry fruit.

In addition, we sampled the density of adults of each plant species in the habitat type in

which it predominantly occurred in a previous 4-year phenological study (C.I. Donatti,

unpubl. data). We surveyed species that mainly occur in the semi-deciduous forest, in the

savannah and in the gallery forest, in 1-ha plots set up in each one of those habitat types.

In addition, we counted the number of fruits produced by at least five individuals of each

plant species and assessed the number of months with mature fruits of every plant species

based on our previous phenological study (C.I. Donatti, unpubl. data). We then combined

the density of a given plant species with its fecundity and its phenology (C.I. Donatti,

unpubl. data) into a “fruit availability index” (density of a plant species/ha X average

number of months with mature fruits/year X average number of fruits

91

produced/individual plant). For each animal species, we gathered information on the

animal body mass from the literature and assessed the density of mammals and large-

bodied bird species in a previous census in the area (C.I. Donatti, unpubl. data).

We then tested the associations betwee those independent variables and the

properties of plant and animal species in the network. For the plant species, we used a

regression tree model from the R package tree (http://www.r-project.org/) to select the

three most important variables that were later used in a linear regression model. Variables

were Box-Cox transformed using JMP v 5.0 (SAS Institute Inc.) to ensure that the

assumptions of the linear model were met. For the animal species, we just used the linear

regression model, as we had only two variables (body mass and animal species density).

Body mass was log transformed and species density was Box-Cox transformed to ensure

that the assumptions of the linear regression model were met.

Phylogenetic signal in traits and properties of animal and plant species in the network

We tested whether traits and properties of animal and plant species in the network

had a significant phylogenetic signal, i.e. a quantitative measure of the degree to which

phylogeny predicts the similarity of species regarding their traits and properties in the

network. To build the animal phylogenetic tree, we followed Bininda-Emonds et al.

(2007) for relationships among mammal species and Hackett et al. (2008) for

relationships among bird species. In addition, we used published work to resolve

relationships within Cracidae (Pereira et al. 2002), Tyrannidae (Tello et al. 2009), and

Thraupidae (Klicka et al. 2007). We also used two mitochondrial DNA sequences

(Cytochrome b and Cytochrome oxidase subunit 1), available on GenBank, to resolve the

relationships between and within the Thraupidae and Icteridae. We then generated a

phylogenetic tree for those sequences using the program Méthodes et algorithmes pour la

bio-informatique (http://www.phylogeny.fr/) and added those relationships in the tree.

The plant phylogenetic tree was built using Phylomatic software

(http://www.phylodiversity.net/phylomatic/phylomatic.html). Relationships within

Fabaceae followed Wojciechowski et al. (2004), within Rubiaceae followed Bremer &

Eriksson (2009) and within Arecaceae followed Asmussen et al. (2006). Since all

branches in animal and plant trees were set equal to one, we conducted simulations that

92

showed that using branch lengths equal to one is a conservative approach when values of

K are lower than or equal to one (see Donatti et al. 2011).

We assessed the K statistic to measure the phylogenetic signal in traits and

properties of animal and plant species in the network, using the function phylosignal in

the picante package (Kembel et al. 2010) of R. The K statistic compares the observed

signal in a trait to the signal under a Brownian motion model of trait evolution on a

phylogeny (Blomberg et al. 2003). The statistical significance of phylogenetic signal is

evaluated by comparing observed patterns of the variance of independent contrasts of a

trait to a null model of shuffling taxa labels across the tips of the phylogenetic tree. To

assess phylogenetic signal in the body mass and in the properties of animal species, we

analyzed them for mammal and bird species independently. However, the phylogenetic

signal in animal species density was calculated for mammal and bird species combined.

Results

Phylogenetic signal in traits and properties of animal and plant species in the network

Plant species traits, fruit diameter, the availability of fruits and energy content

were found not to show a significant phylogenetic signal (K=0.372, p=0.248, d.f.=44,

K=0.2796, p=0.518, d.f.=40; K=0.3037, p=0.177, d.f.=41, respectively), and seed

diameter is more divergent than expected under a Brownian model (K=0.417, p=0.003,

d.f.=43). We did not detect phylogenetic signal in the values of species degree, species

strength, maximum dependence and interaction asymmetry in plant species (K=0.2919,

p=0.323; K=0.3109, p=0.146; K=0.3534, p=0.386; K=0.3384, p=0.166; d.f.= 44;

respectively).

The body mass of closely related mammal species has exactly the amount of

signal predicted by Brownian motion (K=1.01, p=0.005, d.f.=12). In contrast, the body

mass of birds is more divergent than expected under a Brownian model (birds: K=0.778,

p=0.001, d.f.=31). The density of mammals and birds does not have a phylogenetic signal

(K=0.4936, p=0.364, d.f=14). Likewise, we did not detect phylogenetic signal in the

values of species degree, species strength, maximum dependence and interaction

asymmetry in bird species (K=0.2654, p=0.191; K=0.2151, p=0.367; K=0.1958, p=0.367;

93

K=0.2067, p=0.323; d.f.=31, respectively) nor mammal species (K=0.5215, p=0.631,

K=0.66, p=0.257, K=0.439, p=0.917; K=0.6617, p=0.365; d.f.=12; respectively).

Therefore, shared evolutionary history does not explain the variation in traits nor in the

four properties across species in this network.

Observed distribution in the properties of species in the network

The frequency distributions of species degree, species strength and interaction

asymmetry for both plant and animal species were right-skewed and were statistically

distinguishable from a normal distribution (Shapiro-Wilk W=0.93, p=0.0136, W=0.74,

p<0.0001, W=0.81, p<0.0001, n=45, respectively for the plant species; W=0.8, p<0.0003,

W=0.73, p<0.0001, W=0.768, p<0.0001, n=46, respectively for the animal species).

Therefore, for both plant and animal species sets, many species have few interactions,

many species have low importance to the set of partners, and the majority of species is as

important to a partner as the partner is important to them. In addition, although the

majority of values of interaction asymmetry is weak for both plant and the animal species

sets, animal species show high positive values more frequently than plant species,

indicating that consumers (animal species) are more important to the set of hosts (plant

species) than vice-versa.

The frequency distributions of the value of maximum dependence of both plant

and animal species were not skewed but could be still distinguished statistically from a

normal distribution (W=0.92, p=0.009, n=44; W=0.89, p=0.0003, n=46, respectively).

For the plant species, the distribution was relatively symmetrical, meaning that plant

species show a wide variation in this property and values are relatively similar across this

variation, whereas for the animal species, the distribution had two very pronounced peaks

around values of 0.5 and 1. That is, animal species either rely to a low and similar

intensity on several plant species or strongly rely on a single plant species.

Association between traits and the properties of species in the network

For both plant and animal species sets, species strength and interaction asymmetry

showed a strong positive association with species degree (plants: F=55.65, p<0.0001,

94

r2=0.54; F= 53.89, p<0.0001, r2=0.54, d.f.=44; animals: F=221.82, p<0.0001, r2=0.83;

F=29.78, p<0.0001, r2=0.7, d.f.=45; respectively), whereas the value of maximum

dependence of both plant and animal species showed a strong negative association with

species degree (F=29.51, p<0.0001, r2=0.41, n=44; F=41.57, p<0.0001, n=46;

respectively). Thus, since species strength and the value of maximum dependence are

correlated to species degree for both the plant and the animal species set, we used the

ratio of species strength and species degree and the ratio of maximum dependence and

species degree to test the associations between these species traits and these two

properties. For the interaction asymmetry, we did not divide the values by species degree

to test the association between interaction asymmetry and species traits, since the formula

to assess this value already accounts for the number of interactions in each species.

For plant species degree, the regression tree model selected seed diameter as the

best predictor for this attribute, and the linear regression model was negative and

significant (F=10.36, p=0.002, r2=0.19, d.f.=43). For plant species strength, the regression

tree model selected fruit diameter as the best predictor for this attribute, and the linear

regression model was also negative and significant (F=26.74, p<0.0001, r2=0.38,

d.f.=43). For maximum dependence of plant species, the regression tree model shows that

seed diameter is the only predictor that explains this attribute, and the linear regression

model was positive and significant (F=4.185, p=0.046, r2=0.07, d.f=43). For the

interaction asymmetry of plant species, the regression tree model shows that seed

diameter is the only predictor that explains this property, and the linear regression model

was negative and significant (F=7.936, p=0.0072, r2=0.16, d.f.=43, Table 3a). In

summary, plant species degree, plant species strength and interaction asymmetry in plant

species decreased with an increment in seed diameter or in fruit diameter, whereas the

value of maximum dependence of plant species on seed dispersers increased with the

increment in seed or fruit diameter (Fig. 1). Although the interaction asymmetry is

positively associated with seed diameter, values of interaction asymmetry in plant species

were usually negative. Thus, as seed diameter increases, there is more asymmetry

between the interaction strength of a plant and the interaction strength of an interacting

animal species.

For animal species degree, the multiple regression was significant (F=3.767, p=

95

0.048, r2=0.51, d.f.=13), with the body mass of animals mainly explaining the variance in

this attribute (27.62%). For animal species strength, the multiple regression was

significant (F=7.166, p=0.007, r2=0.67, d.f.=13), with the density of animal species

explaining the highest variance in this attribute (33.07%), followed by body mass

(18.28%) and the association between those two variables (16.89%). For maximum

dependence, the multiple regression was significant (F=4.192, p=0.036, r2=0.54, d.f.=13),

with body mass explaining the highest variance in this attribute (32.68%). For interaction

asymmetry, the multiple regression was significant (F=73.82, p=0.0067, r2=0.69,

d.f.=12), with the interaction between body mass and density explaining the majority of

the variance in this attribute (26.96%), followed by body mass (26.72%) and by density

(16.5%) (Table 3b). Therefore, animal species degree increased with an increment in the

body mass of seed dispersers, whereas animal species strength and interaction asymmetry

of animal species increased with an increment in both the body mass and the density of

seed dispersers. On the other hand, the value of maximum dependence of an animal on

particular plant species decreased with the body mass of seed dispersers (Fig. 3).

Discussion

Recent work has greatly improved our knowledge of the patterns and properties of

interactions that characterize mutualistic networks, and the challenge now is to

understand the ecological and evolutionary processes that contribute to them. Our results

suggest that both quantitative and qualitative species properties in networks are

determined by ecological processes rather than by evolutionary ones. That is,

phylogenetic history of both plant and animal species did not explain the variation in their

properties, which were mainly explained by species morphological traits. The lack of

phylogenetic signal in properties of species in this network is not surprising given that the

variables that explain such properties do not have a phylogenetic signal. Furthermore, it

has been recently shown that the pattern of this network is only partially explained by

shared evolutionary history (Donatti et al. 2011).

Therefore, the properties of plant species in this network are solely explained by

morphological traits, specifically fruit and seed diameter. Plant species with small fruits

or seeds have more interactions, are more important to seed dispersers, show a lower

96

interaction asymmetry with animals and are less dependent on a single seed disperser

than species with large fruits or seeds. Thus, our data show that in this seed dispersal

network, trait matching as the result of exploitation barriers was the main process by

which plant species showed variation in all four properties studied here.

Although several other studies have also found that trait complementarity (such as

phenological matching and/or exploitation barriers) were important in explaining plant

species degree in networks and, ultimately, in indirectly determining the pattern in

mutualistic interactions (Stang et al. 2007, Jordano et al. 2006, Rezende et al. 2007,

Santamaría & Rodríguez-Gironés, 2007), just a few studies looked at species properties

derived from quantitative networks. These studies either show a combined effect of

species morphology and abundance (Stang et al. 2006, Stang et al. 2007), or of species

abundance, in partially or entirely explaining the properties of species in mutualistic

networks (Vázquez et al. 2007, Vázquez et al. 2009).

In contrast, the properties of animal species in this network were explained by two

processes: (1) trait matching as a result of exploitation barriers, where seed dispersers

only interact with fleshy-fruited plants if they are big enough to swallow or carry the

fruits, and (2) neutrality, where the probability of an interaction between an animal and a

plant species depends on the density of animals. Animal species with large body mass

have more interactions, are less dependent on a single plant species and show a higher

interaction asymmetry with plants than species with small body mass. Additionally,

species with large body mass and high densities are more important to the set of plant

species. Therefore, although our data show that body mass is the most important variable

that explains animal species properties in this network, animal species density has also an

important effect in determining species strength. Our findings are consistent with the

existing literature in mutualistic networks that suggests an important role of gape size and

of body mass in constraining the number of interactions established by animals (Jordano

1987, Benkman 1999, Herrera 2002, Böhning-Gaese et al. 2003, Cohen et al. 2003,

Abzhanov et al. 2004, Vázquez et al. 2007, Carnicer et al. 2009, Donatti et al. 2011), or

of animal abundance in determining properties derived from interaction frequencies such

as species strength in mutualistic networks (Vázquez et al. 2007, Vázquez et al. 2009:

pollination networks; Carnicer et al. 2009, Schleuning et al. 2011: seed dispersal

97

networks).

The few studies that tested associations between ecological processes and

properties of species in seed dispersal networks have highlighted the importance of

species abundance in explaining species properties. However, in the present study, we

have shown that the density of animal species and the availability of fruits have a minor

(in the case of the animal species set) or no effect at all (in the case of the plant species

set) in explaining the majority of properties of species in this network. This is the case

probably because interactions between species do not occur randomly in this seed

dispersal network (see Donatti et al. 2011). Furthermore, the importance of species

abundance in explaining species properties found in previous studies could be due to the

low diversity in such networks, which includes mainly seed-dispersing birds that interact

with plant species that share similar traits (Rezende et al. 2007). In this way, since fruit

and animal morphological traits do not vary considerably in other fruit-frugivore

communities, the importance of the abundance of species in explaining species properties

may have stood out in those studies. Conceivably, the lack of importance of fruit

availability in explaining plant species properties in this network may lie on the fact that

our plant community is very diverse. Thus, since we have species with a variety of fruit

and seed sizes, this variation is probably large enough to enable the association between

seed (or fruit) size and the properties of species to arise. On the other hand, even though

our original animal community is also very diverse, we were not able to include all

species of this network in our analysis, as density data for all animal species is not

available. Thus, it is possible that if we had included census data for all 46 animal species

from the network, intrinsic species traits, such as body mass, could have been an

important predictor of species properties in this network.

A previous study has shown that the pattern of this seed dispersal network emerges

mainly by trait convergence of phylogenetic unrelated species, with phylogeny showing a

limited effect (Donatti et al. 2011). Likewise, here, we have shown that species-specific

traits also determine the quantitative properties of species in the network, such as species

strength and interaction asymmetry between species. Finally, we posit that the results that

we found here may contribute to understanding coevolution among interacting species.

More specifically, our results imply that large-seeded fruits and large-bodied animals will

98

have limited opportunity for coevolution because they are asymmetrically influenced by

their interaction partners. Thus, as interactions with strong reciprocal effects may have

the greatest potential for coevolution (Vázquez et al. 2007), interactions involving small-

seeded plant species and small-bodied animal species may establish the template for

coevolution to take place in seed dispersal networks, if such interactions occur in high

frequencies and are common through time. Therefore, we argue that species-specific

morphological traits of plants and animal species should be considered in studies that

evaluate the ecological and evolutionary consequences of plant-animal interaction

networks.

Acknowledgments

We would like to thank Reese Rogers, Lee Love-Anderegg, Adilson Braga Samuel and

Luisa Haddad for their help with data collection. We thank FAPESP (2004/00810-3 and

2008/10154-7), Earthwatch Institute and Conservation International for financial support.

CID was supported by Stanford University and MG and MAP by a CNPq fellowship. We

also thank Conservation International, Lucas Leuzinger and Marina Schweizer for their

permission to work in their properties.

99

References

Abzhanov, A, Protas M, Grant BR, Grant PR, Tabin CJ (2004) Bmp4 and morphological variation of beaks in Darwin’s finches. Science 305:1462–1464.

Asmussen CB, Dransfield J, Deichmann V, Barfod AS, Pintaud J-C, Baker WJ (2006) A new subfamily classification of the palm family (Arecaceae): evidence from plastid DNA phylogeny. Bot. J. Linn. Soc. 151: 15-38.

Bascompte J, Jordano P (2007) Plant-animal mutualistic networks: the architecture of biodiversity. Annu. Rev. Ecol. Evol. 38: 567-593.

Bascompte J, Jordano P, Olesen JM (2006) Asymmetric coevolutionary networks facilitate biodiversity maintenance. Science 312: 431-433.

Bascompte J Jordano P, Melián CJ, Olesen JM (2003) The nested assembly of plant-animal mutualistic networks. Proc. Natl Acad. Sci. USA 100 (16): 9383-9387.

Ben-Naim E, Frauenfelder H, Toroczkai Z (2004) Complex Networks, Springer, Heidelberg, 513 p.

Benkman CW (1999) The selection mosaic and diversifying coevolution between crossbills and lodgepole pine. Amer. Nat. 154:S75-S91

Berlow EL, Neutel A-M, Cohen JE, de Ruiter PC, Ebenman B, Emmerson M. et al. (2004). Interaction strengths in food webs: issues and opportunities. J. Anim. Ecol. 73: 585–598.

Bininda-Emonds ORP, Cardillo M, Jones KE, McPhee RDE, Beck RMD, Grenyer R. et al. (2007). The delayed rise of present-day mammals. Nature 446: 507-512.

Blomberg S P, Garland T, Ives AR (2003) Testing for phylogenetic signal in comparative data: Behavioral traits are more labile. Evolution 57: 717-745.

Böhning-Gaese K, Schuda MD, Helbig AJ (2003) Weak phylogenetic effects on ecological niches of Sylvia warblers. J. Evol. Biol. 16:956 –965.

Brewer B, Eriksson T (2009) Time tree of Rubiaceae: Phylogeny and dating the family, subfamilies, and tribes. Int. J. Plant Sci. 170(6): 766-793.

Carnicer J, Jordano P, Mélian CJ (2009) The temporal dynamics of resource use by frugivorous birds: a network approach. Ecology 90(7): 1958-1970.

Clauset A, Moore C, Newman MEJ (2008) Hierarchical structure and the prediction of missing links in networks. Nature 453:98–100.

Cohen JE, Johnsson T, Carpenter SR (2003) Ecological community description using the food web, species abundance, and body size. Proc. Natl Acad. Sci. USA 100: 1781–1786.

Dicks L, Corbet SA, Pywell RF (2002) Compartmentalization in plant-insect flower visitor webs. J. Anim. Ecol 71: 32-43.

Donatti CI, Guimarães PR, Galetti M, Pizo MA, Marquitti FMD, Dirzo R (2011) Analysis of a hyper-diverse seed dispersal network: modularity and underlying mechanisms. Ecol. Lett. 14: 773-781

Dunne JA, Williams RJ, Martinez ND (2002) Network structure and biodiversity loss in food webs: robustness increases with connectance. Ecol Lett, 5, 558-567.

Dupont YL, Hansen DM, Olesen JM (2003) Structure of a plant–flower visitor network in the high altitude sub-alpine desert of Tenerife, Canary Islands. Ecography 26:301–310.

100

Fortuna MA, Stouffer DB, Olesen JM, Jordano P, Mouillot D, Krasnov BR, Poulim R, Bascompte J (2010) Nesteness versus modularity in ecological networks: two sides of the same coin? J. Anim. Ecol. 79: 811-817.

Fortuna MA, Bascompte J (2006) Habitat loss and the structure of plant-animal mutualistic networks. Ecol. Lett. 9: 278–283.

Galetti M, Donatti CI, Pizo MA, Giacomini HC (2008) Big fish are the best: seed dispersal of Bactris glaucescens by the pacu fish (Piaractus mesopotamicus) in the Pantanal, Brazil. Biotropica 40(3): 386-389.

Guimarães PR, Rico-Gray V, Furtado dos Reis S, Thompson JN (2006) Asymmetries in specialization in ant-plant mutualistic networks. Proc R Soc Lond B 273: 2041–2047.

Guimera R, Amaral LAN (2005) Functional cartography of complex metabolic networks. Nature 433: 895–900.

Hackett SJ, Kimball RT, Reddy S, Bowie RCK, Braun EL, Chojnowski JL et al. (2008) A phylogenomic study of birds reveals their evolutionary history. Science 320: 1763-1768.

Herrera CM (2002) Seed dispersal by vertebrates in Herrera CM, Pellmyr O (editors) Plant-animal interactions. An evolutionary approach. Oxford: Blackwell, pp. 185–208.

Howe HF (1993) Aspects of variation in a neotropical seed dispersal system. Vegetatio 107/108: 149-162.

Joppa LN, Montoya JM, Solé R, Sanderson J, Pimm S (2010) On nestedness in ecological networks. Evol. Ecol. Res. 12: 35-46.

Jordano P, Bascompte J, Olesen JM (2006) The ecological consequences of complex topology and nested structure in pollination webs in Waser NM, Olerton J (editors) Plant-pollinator interactions: from specialization to generalization. The University of Chicago Press, London, UK, pp 173-199

Jordano P, Bascompte, J, Olesen JM (2003) Invariant properties in coevolutionary networks of plant-animal interactions. Ecol. Lett. 6 (1): 69-81.

Jordano P (1987) Patterns of mutualistic interactions in pollination and seed dispersal: connectance, dependence asymmetries and coevolution. Amer. Nat. 129:657-677.

Kembel SW, Cowan PD, Helmus MR, Cornwell WK, Morlon H, Ackerly DD, Blomberg SP, Webb CO (2010) Picante: R tools for integrating phylogenies and ecology. Bioinformatics 26:1463-1464.

Klicka J, Burns K, Spellman GM (2007) Defining a monophyletic Cardinalini: A molecular perspective. Mol. Phylogenet. Evol. 45: 1014-1032.

Krishna A, Guimarães PR, Jordano P, Bascompte J (2008) A neutral –niche theory of nestedness in mutualistic networks. Oikos 117: 1609 – 1618.

Mello MAR, Marquitti FMD, Guimarães PR, Kalko EKV, Jordano P, Aguiar MAM (2011) The modularity of seed dispersal: differences in structure and robustness between bat– and bird–fruit networks. Oecologia 161(1): 131-140.

Memmott J, Waser NM (2002) Integration of alien plants into a native flower-pollinator visitation web. Proc. R. Soc. Lond. 269: 2395-2399.

Memmott J (1999) The structure of a plant-pollinator network. Ecol. Lett. 2: 276-280. Montoya, JM, Pimm SL, Sole RV (2006) Ecological networks and their fragility. Nature

442: 259–264.

101

Ollerton J, Johnson SD, Cranmer L, Kellie S (2003) The pollination ecology of an assemblage of grassland asclepiads in South Africa. Ann. Bot. 92: 807-834.

Olesen JM, Bascompte J, Dupont YL, Jordano P (2007) The modularity of pollination networks. Proc. Natl Acad. Sci. USA 104(50): 19891-19896.

Olesen JM, Jordano P (2002) Geographic patterns in plant-pollinator mutualistic networks. Ecology 83: 2416-2424.

Pereira SL, Baker AJ, Wajntal A (2002) Combined nuclear and mitochondrial DNA sequences resolve relationships within the Cracidae (Galliformes, Aves). Syst. Biol. 51(6): 964-958.

Piazzon M, Larrinaga AR, Santamaría L (2011) Are nested networks more robust to disturbance? A test using epiphyte-tree, communalistic networks. PLoS ONE 6(5): e19637. doi:10.1371/journal.pone.0019637

Rezende E, Lavabre J, Guimarães PR, Bascompte J (2007) Non-random coextinctions in phylogenetically structured mutualistic networks. Nature 448: 925-928.

Rooney N, McCann K, Gellner G, Moore JC (2006) Structural asymmetry and the stability of diverse food webs. Nature 442:265–269.

Santamaria L, Rodríguez-Gironés MA (2007) Linkage rules for plant-pollinator networks: trait complementarity or exploitation barriers? PloS Biology 5: 354–362.

Schleuning M et al (2011) Specialization and interaction strength in a tropical plant-frugivore network differ among forest strata. Ecology 92(1): 26-36.

Stang M, Klinkhamer PGL, van der Meijden E (2007) Asymmetric specialization and extinction risk in plant-flower visitor webs: a matter of morphology or abundance? Oecol. 151: 442–453.

Stang M, Klinkhamer PGL, van der Meijden E (2006) Size constraints and flower abundance determine the number of interactions in a plant-flower visitor web. Oikos 112: 111-121.

Tello JG, Moyle RG, Marchese DJ, Cracraft J (2009) Phylogeny and phylogenetic classification of the tyrant flycatchers, cotingas, manakins, and their aliens (Aves: Tyrannides). Cladistics 25: 429-467.

Thompson JN (2005) The geographical mosaic of coevolution. University of Chicago Press, Chicago, Illinois, USA.

Thompson JN (1999) The raw material for coevolution. Oikos 84: 5–16. Vázquez DP, Chacoff NP, Cagnolo L (2009) Evaluating multiple determinants of the

structure of plant-animal mutualistic networks. Ecology 90 (8): 2039-2046. Vázquez DP, Melián CJ, Williams NM, Blüthgen N, Krasnov BR, Poulin R (2007)

Species abundance and asymmetric interaction strength in ecological networks. Oikos116: 1120-1127.

Vázquez DP (2005) Degree distribution in plant –animal mutualistic networks: forbidden links or random interactions? Oikos 108: 421-426.

Vázquez DP, Morris WF, Jordano P (2005) Interaction frequency as a surrogate for the total effect of animal mutualists on plants. Ecol. Lett. 8: 1088-1094.

Vázquez DP, Aizen MA (2004) Asymmetric specialization: a pervasive feature of plant-pollinator interactions. Ecology 85: 1251-1257.

Vázquez DP, Aizen MA. (2003) Null model analyes of specialization in plant-pollinator interactions. Ecology 84: 2493-2501.

102

Waser N, Price MV, Williams, NM, Ollerton J (1996) Generalization in pollination systems and why it matters. Ecology 77: 1043-1060.

Wojciechowski MF, Lavin M, Sanderson MJ (2004) A phylogeny of legumes (Leguminosae) based on analysis of the plastid matK gene resolves many well-supported subclades within the family. Am. J. Bot. 91: 1846-1862.

103

Tables

Table 1. Plant species in the seed dispersal network, seed diameter (mm), fruit diameter (mm), availability of fruits in the area, energy content in fruit pulp (in Kcal/g of dry fruit), species degree, species strength, maximum dependence and interaction asymmetry.

Species

Fruit

diameter

Seed

diameter

Availability

of fruits

Energy

content

Species

degree

Species

strength

Maximum

dependence

Interaction

asymmetry

Ocotea diospyrifolia 5.852 4.824 15.48 3.279 11 3.607 0.326 0.237

Anonna dioica 75.067 8.27 n/a 1.501 6 0.187 0.694 -0.135

Zanthoxyllum rigidum 3.3 3.3 208.8 n/a 7 3.033 0.286 0.29

Protium heptaphyllum 15.695 9.536 6122.592 3.169 1 0.64 0.194 -0.036

Melicoccus lepidopetalus 20.635 11.2 3.432 1.248 8 0.974 0.24 -0.003

Sterculia apetala 17.564 14.346 163.17 n/a 4 1.481 0.806 0.12

Guazuma ulmifolia 22.068 1.770 45 0.743 8 0.357 0.25 -0.08

Mouriri eliptica 29.273 13.267 0.96 1.578 9 1.323 0.646 0.036

Psidium nutans 26.439 3.411 6 2.827 11 0.596 0.25 -0.037

Eugenia desynterica 25.841 11.003 n/a 2.01 3 0.188 0.571 -0.271

Dipteryx alata 39.905 35.734 270.289 2.295 4 0.371 0.735 -0.157

Swartzia jorori 14.875 10.785 n/a 3.148 3 0.96 0.667 -0.013

Inga laurina 18.807 9.921 89.64 3.653 5 0.153 0.417 -0.169

Enterolobium

contortisiliquum

78.763

10.092 48 2.492 5 0.722 0.371 -0.056

Hymenaea stigonocarpa 41.704 17.02 0.9 2.623 2 0.235 0.811 -0.383

Rhamnidium elaeocarpum 7.498 4.829 18115.2 2.251 5 1.219 0.429 0.044

Ficus pertusa 7.099 0.1 4262.4 0.884 13 3.198 0.233 0.169

Ficus gomelleira 20.930 0.1 7992 1.033 6 0.169 0.462 -0.139

Cecropia pachystachia 8.563 0.1 27.594 0.765 9 2.005 0.256 0.112

Salacia elliptica 46.048 13.618 157.248 1.792 3 0.073 0.667 -0.309

Licania parvifolia 12.458 8.333 3648 3.156 1 0.182 1 -0.818

Couepia uiti 21.860 18.800 1733.04 1.836 5 0.368 0.316 -0.126

Garcinia brasiliensis 22.585 10.255 n/a 7.545 5 0.201 0.625 -0.16

Caryocar brasiliense 65.15 31.545 18 1.96 1 0.072 1 -0.928

Byrsonima verbascifolia 18.126 9.852 3.6 1.67 2 0.048 0.5 -0.476

Byrsonima orbignyana 11.837 6.055 18048 2.464 14 1.651 0.39 0.046

Doliocarpus dentatus 5.453 4.368 6 1.641 15 5.945 0.265 0.33

Curatella americana 6 5 7920 1.46 11 3.292 0.219 0.208

Dulacia egleri 17.122 9.775 6 3.081 5 0.671 0.567 -0.066

Agonandra brasiliensis 24.227 14.627 81.696 1.504 7 0.253 0.286 -0.107

Psittacanthus calyculatus 9.677 7 24.9 1.308 3 0.496 0.667 -0.168

104

Psittacanthus cordatus 12.5 8.502 24.9 1.308 1 0.289 1 -0.711

Pouteria ramiflora 31.901 15.054 1.26 1.903 3 0.087 0.455 -0.304

Pouteria gardneri 11.893 7.209 2.76 2.381 4 0.125 0.4 -0.219

Diospyrus hispida 45.636 12.931 88.8 1.356 5 0.545 0.35 -0.091

Vitex cymosa 16.412 8.974 1534.464 3.130 8 0.710 0.796 -0.036

Hancornia speciosa 28.823 10.868 4.32 3.416 12 2.061 0.18 0.088

Genipa Americana 60.019 6.013 3.12 3.334 18 3.742 0.256 0.152

Tocoyena Formosa 34.863 6.438 239.616 1.338 2 0.027 0.8 -0.486

Alibertia sessilis 26.67 5.084 272.448 2.316 8 0.358 0.544 -0.08

Atallea phalerata 35.106 22.081 3914.400 2.271 7 0.981 0.486 -0.003

Syagrus flexuosa 15.884 13.701 0.9 n/a 3 0.131 0.667 -0.29

Bactris glaucescens 18.1 8.535 7056 2.536 5 0.818 0.852 -0.036

Acrocomia aculeata 32.135 22.08 344.52 3.008 6 0.617 0.407 -0.064

Copernicia alba 16.350 14.16 293.04 1.823 8 0.837 0.867 -0.02

105

Table 2. Animal species in the seed dispersal network, body mass (kg), species density, species degree, species strength, maximum dependence and interaction asymmetry.

Species Body mass

Density (inds/km2)

Species degree

Species strength

Maximum dependence

Interaction Asymmetry

Piaractus mesopotamicus 1.14 n/a 4 2.136 0.787 1.136 Geochelone carbonaria 6.6 n/a 14 1.465 0.125 0.465 Euphractus sexcintus 6 n/a 5 0.149 0.437 -0.426 Alouatta caraya 5.4 2.02 4 0.157 0.5 -0.084 Agouti paca 9.1 0.1 3 0.112 0.5 -0.148 Dasyprocta sp 2.8 1.84 17 2.573 0.144 0.524 Cerdocyon thous 5.7 n/a 14 2.831 0.261 0.108 Nasua nasua 5.1 1.7 16 1.114 0.269 0.114 Procyon cancrivorus 4.5 n/a 5 0.417 0.437 -0.032 Tapirus terrestris 240 n/a 21 4.416 0.273 1.708 Mazama americana 36 1.49 7 0.948 0.347 -0.002 Sus scrofa 50 6.35 26 5.59 0.239 1.53 Pecari tajacu 26 3.69 9 1.635 0.298 0.03 Tayassu pecari 35 9.63 18 4.216 0.346 1.608 Crypturellus sp 0.08 n/a 1 0.143 1 -0.061 Rhea americana 20 n/a 17 3.224 0.371 0.171 Aburria pipile 1 11.71 8 1.065 0.492 0.032 Ortalis canicollis 0.6 36.28 13 1.581 0.171 0.042 Crax fasciolata 2.8 4.02 7 0.941 0.854 -0.059 Columba sp 0.2 n/a 3 0.092 0.5 -0.454 Guira guira 0.07 n/a 2 0.087 0.5 -0.304 Trogon curucui 0.07 2.03 2 0.146 0.714 -0.142 Ramphastos toco 0.54 4.85 10 1.769 0.781 0.384 Pteroglossus castanotis 0.27 1.26 5 0.41 0.526 -0.197 Brotogeris versicolurus 0.066 n/a 2 0.093 0.5 -0.907 Aratinga leucophthalmus 0.166 n/a 1 0.071 1 -0.232 Aratinga aurea 0.1 n/a 1 0.093 1 -0.302 Tytira cayana 0.07 n/a 1 0.186 1 -0.163 Pitangus sulphuratus 0.068 n/a 6 2.028 0.289 0.514 Tyrannus melancholicus 0.039 n/a 1 0.132 1 -0.124 Myiodinastes maculatus 0.043 n/a 2 0.055 0.5 -0.189 Casiornis rufa 0.02 n/a 1 0.286 1 -0.238 Myiarchus ferox 0.024 n/a 3 0.212 0.5 -0.046 Cyanocorax cyanomelas 0.2 n/a 8 1.066 0.414 0.009 Cyanocorax chrysops 0.14 n/a 3 0.201 0.5 -0.266 Turdus rufiventris 0.07 n/a 2 0.095 0.5 -0.181 Turdus sp 0.05 n/a 3 0.378 0.7 -0.622 Saltator coerulescens 0.052 n/a 3 0.839 0.666 -0.161 Poroaria coronata 0.03 n/a 1 0.015 1 -0.985 Thraupis palmarum 0.036 n/a 3 0.352 0.571 -0.162 Thraupis sayaca 0.03 n/a 2 0.376 0.947 -0.078 Psariocollius decumanus 0.25 n/a 7 0.682 0.26 -0.045 Icterus croconotus 0.07 n/a 2 0.127 0.75 -0.109

106

Gnomiropsar chopi 0.07 n/a 1 0.023 1 -0.977 Ramphocelus carbo 0.025 n/a 6 0.458 0.52 -0.034 Tachyphonus rufus 0.033 n/a 1 0.016 1 -0.109

107

Table 3a. Results of multiple regression analyses on the associations between species traits and species properties for plant species.

Table 3b. Results of multiple regression analyses on the associations between species traits and species properties for mammal and bird species combined.

Species degree r2=0.19 Species strength r2=0.38 Maximum dependence r2=0.07 Interaction asymmetry r2=0.16 Model terms SS P SS P SS P SS p

Seed diameter 134.58 0.002* 0.2309 0.046* 0.4743 0.0072* Fruit diameter 0.2107 <0.0001* Residuals 558.62 0.3311 2.3732 2.5703

Species degree r2=0.51 Species strength r2=0.67 Maximum dependence r2=0.54 Interaction asymmetry r2=0.69 Model terms SS P SS P SS P SS p Log Body mass 137.915 0.05* 0.009 0.037* 0.0373 0.021* 1.2063 0.015* Density 135.800 0.054 0.016 0.009* 0.0245 0.052 0.6863 0.05*

Log Body Mass:Density 49.482 0.21 0.008 0.043* 0.0018 0.563 1.2172 0.014* Residuals 286.017 0.015 0.0563 1.4042

108

Figures

Figure 1. Association between seed diameter (mm) and species degree (top), maximum dependence (center) and interaction asymmetry (bottom) of plant species.

109

Figure 2. Association between body mass (kg) and species degree (top), maximum

dependence (center) and interaction asymmetry (bottom) of animal species. Body mass is

in log scale.

110

CHAPTER 5

THE ROLE OF SEED DISPERSAL INTERACTIONS IN

STRUCTURING A PLANT COMMUNITY IN THE BRAZILIAN

PANTANAL Camila I. Donatti, Mauro Galetti & Rodolfo Dirzo

Abstract

The study of seed dispersal is crucial to our understanding of the structure of plant

communities, especially in tropical forests, where dispersal limitation is prevalent. Even

though additional relevant processes occur between seed dispersal and seedling

establishment (e.g., seed and seedling predation), it is possible to find an association

between patterns of seed dispersal and seedling distribution. However, the importance of

seed dispersal in determining the spatial distribution of plant species, taking into account

that the majority of them interact with multiple seed dispersers, has received little

attention. Here, our goal is to examine if a “dispersal signal” in terms of an association

between seed dispersal and spatial distribution of animal-dispersed plants can be detected

in a natural setting where a multiplicity of plants and dispersal agents interact. We

assessed spatial distribution of plant species that are located in a gradient of dispersal-

dependence that includes, at one extreme, those that strongly interact with one single seed

disperser and, at the other, those that weakly interact with multiple dispersers. Species

that strongly rely on a single seed disperser species showed a highly aggregated

distribution pattern of individuals, and higher than the aggregation found for plant species

that weakly rely on multiple dispersers. This highly aggregated distribution of individuals

could contribute to an observed high mortality of seedlings and saplings. In addition, seed

dispersal was the main important predictor of the aggregation intensity of conspecific

individuals in comparison with several other variables. We conclude that seed dispersal

interactions represent an important factor in determining plant spatial distribution even

when multiple animal species operate as dispersal agents of a particular plant.

111

Furthermore, we suggest that the intensely aggregated distribution of certain plant species

can be intensified in a scenario of defaunation, where large- and medium-bodied species

are absent.

112

Introduction

Understanding mechanisms that regulate the structure of ecological communities is

a central goal of community ecology. Ecological theory recognizes a variety of abiotic

and biotic factors that shape the structure and dynamics of terrestrial plant communities.

Environmental heterogeneity, disturbance, and biotic interactions may all play key roles

in determining plant community composition and diversity in space and time (Tilman &

Pacala 1993, Rosenzweig 1995, Van der Heijden 1998, Hubbell 2001, Ricklefs 2004).

Among those, the study of seed dispersal is crucial to understand the structure of plant

communities, especially in tropical forests where dispersal limitation is prevalent (Clark

et al. 1999, Hubbell et al. 1999).

Even though many processes occur between seed dispersal and seedling

establishment, such as seed and seedling predation, it is possible to find a “dispersal

signal” – an association between patterns of seed dispersal and seedling distribution. In a

large area of dozens of hectares, spatial distributions of seedlings in plant species

dispersed by animals are significantly different than those in plants dispersed by gravity

or wind (Hubbell 1979, Kinnaird 1998, Hardy & Sonké 2004, Seidler & Plotkin 2006,

Russo et al. 2007, Muller-Landau et al. 2008). In a small area of few square meters, the

spatial distribution of seedlings can be associated with the seed dispersal by particular

animal species (Howe 1989, Herrera et al. 1994, Julliot 1998, Wenny & Levey 1998,

Fragoso et al. 2003).

However, at the community-level, one-to-one interactions are rare and the majority

of species, both animals and plants, have more than one partner (Bakker et al. 1996,

Bascompte et al. 2003, Memmott et al. 2004, Strauss & Irwin 2004, Jordano et al. 2006).

Thus, the importance of seed dispersal in determining the spatial distribution of plant

species, taking into account that the majority of them interact with multiple seed

dispersers, has received little attention. One way to summarize the importance of multiple

animal species for the dispersal of a particular plant species is to combine the number and

the frequency of interaction between a particular plant and its seed dispersers.

Collectively, the number and strength of interactions enable one to place plant species in

a gradient that includes, at one extreme, those that strongly interact with one seed

disperser and, at the other, those that weakly interact with multiple dispersers.

113

In our study site, it has been previously shown that the variation in the number and

frequency of interactions across plant species is primarily explained by fruit and seed

diameter (Donatti et al. unpub. data). Thus, a plant species with large seeds has a high

dependence on its most frequent seed disperser whereas a plant species with small seeds

has a low dependence on its most frequent seed disperser. Due to constraints between the

gape size of seed dispersers and the fruit or seed size, large-seeded species are

predominantly dispersed by large-bodied frugivores (Janzen & Martin 1982, Janson

1983, Wheelwright 1985, Chapman et al. 1992, Guimarães et al. 2008, Donatti et al.

2011) that can swallow large seeds, or by medium-bodied frugivores (such as agoutis,

Dasyprocta azarae) that can manipulate and carry those seeds. Such animals usually

disperse seeds in clumps (Howe 1989, Fragoso et al. 2003), either close or far from the

mother plant, possibly generating a highly aggregated distribution of individuals. In

contrast, small-seeded plant species are visited by several seed dispersers that vary in

body size (given that many of the disperser species can swallow small fruits or seeds).

Such dispersers can disperse seeds in clumps (for example by large-and medium-bodied

mammals) or scattered in the area (for example by small-bodied birds; Howe 1989),

possibly generating a random or less aggregated distribution of individuals. Thus, the

combined information on the number of seed disperser species and the frequency of

interactions with these partners shown by a particular plant species should explain, at

least partly, the spatial distribution of conspecific individuals.

Although there is still a need for basic studies that emphasize natural history in seed

dispersal, the challenge now is to understand such natural history knowledge in light of

the processes that shape plant populations and communities (Howe & Miriti 2004,

Volvov et al. 2009). Here, our goal is to examine if the association between seed

dispersal and spatial distribution of animal-dispersed plants can be detected even when

multiple animal species operate as seed dispersal agents of a particular plant species. We

assessed seed dispersal interactions and the spatial distribution of plants in a 2.6-hectare

plot in a semi-deciduous forest in the Brazilian Pantanal. More specifically, we evaluate:

a) the association between the spatial distribution of plant species and the relative

dependence of those plants on their most important seed disperser species, b) the impact

of an aggregated distribution of individuals for seedling and sapling mortality, c) if such

114

mortality is related to high herbivory and attack by pathogens, and d) how much the value

of maximum dependence on seed dispersers, which represents interactions between

plants and seed dispersers, in contrast with other variables, contributes to plant spatial

distribution. Our hypotheses are that: 1) the aggregated distribution of a plant species

will be positively associated with the value of maximum dependence on its most

important seed disperser, 2) the distance from a conspecific individual will significantly

affect seedling and sapling mortality, 3) such high mortality will be related to the high

rates of herbivory and attack by pathogens in individuals located close to a conspecific,

and 4) the value of maximum dependence will be an important predictor of the variation

in the aggregated distribution of individuals across plant species. All these hypotheses

were corroborated, but the high seedling and sapling mortality observed in individuals

close to any other conspecific was not related to high herbivory or attack by pathogens,

which was only true for seedlings and saplings located close to a conspecific adult.

Therefore, the positive association between seedling mortality and proximity to

conspecific individuals may be a result of competition for abiotic factors. We conclude

that, although different mechanisms such as dispersal mode and edaphic characteristics

can operate at different scales in shaping the distribution and structure of plant

communities, seed dispersal interactions also represent an important factor in this respect

even when multiple animal species participate in the dispersal of a particular plant.

Furthermore, plant species that strongly rely on few seed disperser species show strongly

aggregated distributions, which can be intensified in a scenario of defaunation, thus

increasing seedling and sapling mortality.

Methods

Seed dispersal interactions

To sample both the number and the frequency of interactions between animal and

plant species, seed dispersal interactions were recorded using four complementary

methods: 1) focal observations, 2) scats analysis, 3) camera trap techniques and 4)

intestinal analysis. More details about these methods are presented in Donatti et al.

(2011). One event of seed dispersal was considered as such when 1) a fruit was recorded

115

to have been removed from a plant species during focal observations, 2) a visit of a seed

disperser to a plant species with fruit removal was detected with camera traps, 3) a scat

pile was found to have at least one intact seed of a particular species in it, or 4) a sampled

fish intestine contained at least one intact seed in it from a particular plant species. We

used the number of seed dispersal events recorded for each interaction between each

plant and animal species to calculate a value of dependence of a particular plant species

on its most important seed disperser species.

To determine the frequency of interaction between particular plant species and all

animal dispersers, we divided the number of seed dispersal events recorded between a

plant species and each seed disperser species by the total number of dispersal events

recorded for this particular plant. For each plant species, the highest value of interaction

frequency was then considered the dependence of a plant species on its most important

seed disperser, hereafter “maximum dependence”.

Spatial aggregation of individuals across plant species

We set up a 2.6-ha plot in a semi-deciduous forest to assess the spatial distribution

of eight plant species that vary in their values of “maximum dependence” on seed

dispersers. Besides the variation in the gradient of maximum dependence that these eight

plant species display, we also choose to work with these species because there are clear

morphological differences among them in all life stages and because they do not

propagate vegetatively. We mapped all individuals to assess the spatial aggregation of

individuals across species, permanently tagged them to follow their fate, and measured

the basal diameter or the diameter at breast height (1.3 m; DBH) to categorize their size

classes/stages. Seedlings were considered as those individuals with ≤10 mm basal

diameter, representing those individuals that were predominantly up to 50 cm tall.

Individuals were considered saplings if they were 10.1≤15 mm in basal diameter or ≤5

mm in DBH, representing those that were predominantly 50-150 cm tall (Webb & Peart

1999). Juveniles were considered individuals > 15 mm in basal diameter or >5mm in

DBH, representing those >150 cm in height that did not yet have signs of flowers or

fruits. Adults were considered those with signs of flowers or fruits (Scariot 1999). After

116

mapping and marking individuals for the first time in July 2008, we followed their fate

and marked new individuals on four consecutive times (January 2009, July 2009, January

2010 and January 2011), to be able to record seedling and sapling mortality and to

accurately assess spatial distribution.

Spatial aggregation of individuals within each plant species was assessed through

Ω, an index that measures aggregation of individuals (hereafter “aggregation” or “Ω”)

around other conspecifics based on the ratio of local density (within annuli located in

every 5 meters from the focus individual) to overall population density (Condit et al.

2000). More specifically, we evaluated spatial aggregation of each species by counting

individuals in annuli around conspecifics. We counted the number of conspecifics

between x and x + ∆x meters around each individual of each species for several annuli

inside the plot, as well as calculated the area inside the plot of each of these annuli. Then,

the number of neighbors Nx in each annulus at distance x were summed over all

individuals and divided by the sum of the area Ax in each annulus at distance x of over all

individuals. To calculate Ω of a given species in a given annulus, we then divided this

ratio (i.e. number of conspecific/area) by the density of this species in the whole plot. A

great advantage of this method is that it is sample-size independent and different species

can be compared regardless of their densities. In a perfectly random distribution, Ω = 1

for all distances x. Aggregation is indicated when Ω> 1 for short distances, whereas Ω <

1 indicates spacing at some scale, or hyperdispersion (Condit et al. 2000).

We calculated the aggregation of individuals across plant species in different

annuli and tested the associations between theses values of aggregation and those of

maximum dependence using linear regressions. For each plant species, aggregation was

calculated for seedlings around seedlings, seedlings and saplings combined around

seedlings and saplings combined, seedlings around adults, saplings and seedlings

combined around adults, and individuals (regardless of the stage) around conspecific

individuals (regardless of the stage). As many annuli of particular plants included areas

outside of the plot, an edge correction was used so that only the area inside the plot was

considered in the calculation of Ω.

117

Consequences of spatial aggregation on seedling and sapling mortality, leaf herbivory

and infection by pathogens

To address the consequences of aggregation of individuals on seedling and sapling

mortality, we tested the survival of seedlings and saplings as a function of the distance to

the closest conspecific individual, using logistic regressions. Additionally, in our last

survey (January 2011), we also assessed the standing level of herbivory and attack by

pathogens. To do so, we estimated the percentage of damage by herbivores and

pathogens in all leaves of every seedling and sapling, which resulted in an estimated

percentage of the plant with foliar damage by herbivores and pathogens (modified from

Dirzo & Domínguez 1995). We tested the effect of the distance from conspecific

individuals on the percentage of each seedling and sapling that were attacked by

herbivores or damaged by pathogens using linear regressions.

Seed dispersal interactions and the structure of the plant community

Spatial distributions of tropical trees often correlate with local (i.e., microhabitat)

environmental conditions in terms of particular light, soil, moisture and topographic

characteristics, suggesting the potential importance of niche differentiation in structuring

tropical forest tree communities (Clark et al. 1999, Svenning 1999, Webb & Peart 2000,

Harms et al. 2001, Russo et al. 2005). Thus, to tease apart the importance of some of

these and other abiotic and biotic characteristics in explaining the spatial aggregation of

individuals across plant species, we divided the 2.6-ha plot in 64 subplots of 20m X 20m

(two subplots were not included in the analyses due to the presence of bee hives inside

them), and collected data on soil properties, canopy openness and terrestrial Bromelia

cover for each subplot. In July 2009, we collected soil samples (from 0-20 cm of depth)

from five random locations inside each subplot. We mixed those five sub-samples into a

bulk sample for soil analyses. Canopy openness was assessed through hemispherical

photographs taken with a fisheye lens. Photographs were taken at a height of one meter

using a horizontally leveled digital camera and a fisheye lens of 180° field of view. All

photographs were taken either before dawn, after sunset, or at other times of the day

when the sun was blocked by clouds to ensure homogeneous illumination of the canopy

118

and a correct contrast between canopy and sky. Canopy openness in each photograph was

assessed using the software Gap Light Analyzer (GAP, Simon Fraser University,

Vancouver, Canada, http://www.ecostudies.org/gla/).

As the average value of canopy openness did not significantly differ if two or four

hemispherical photographs were taken in each subplot (t-test: 0.619, p=0.54, n=34) and

as the average canopy openness of two and of four photographs were significantly

correlated (Pearson's r=0.89, p<0.0001, n=17), we took only two photographs in each

subplot and averaged them to measure canopy openness in that subplot. Therefore, for

each subplot, we had values of terrestrial Bromelia cover, canopy openness and soil pH,

organic matter, K, Ca, Mg and S in the soil. Additionally, for each subplot we also had

values of maximum dispersal dependence of plant species if such species was recorded in

this particular subplot.

With this information we then tested the importance of biotic (value of maximum

dependence, terrestrial Bromelia cover) and abiotic variables (canopy openness, soil pH,

organic matter, K, Ca, Mg and S), in explaining the aggregation of individuals across

species within subplots. We calculated the aggregation of individuals (regardless of the

stage) within 10 meters of conspecific individuals. Therefore, for each subplot of 20 X 20

meters we also had one value to represent the aggregation of individuals of a given plant

species, if such species was recorded in this particular subplot. We used a regression tree

model from the package tree of R (http://www.r-project.org/) to select the most important

independent variables that were later used in a linear regression model with the values of

aggregation (Ω) as a dependent variable. Variables were Box-Cox transformed using

JMP v 5.0 (SAS Institute Inc.) to ensure that the assumptions of the linear model were

met.

Results

Spatial aggregation of individuals across plant species

All individuals of the eight plant species, which were located along the gradient of

maximum dependence on seed dispersers, were mapped and permanently tagged, totaling

3377 plants (3078 alive and 299 dead in our last survey in January 2011). All plant

119

species showed a value of aggregation (Ω) greater than one in the majority of annuli.

That is, all plant species show an aggregated distribution in the majority of annuli, but

such value of aggregation varies across plant species.

We found a significant and positive association between the maximum

dependence of a plant on its most important seed disperser and the spatial aggregation of

conspecific individuals for individuals located within five meters from any other

conspecific (three meters: F=5.51, p=0.05, r=0.67; four meters: F=6.86, p=0.03, r=0.68;

five meters: F=7.97, p=0.03, r=0.75; n=8, Fig. 1). That is, plant species exhibiting high

values of dependence on a single disperser show an intense aggregation of conspecific

individuals. In this way, plant species that strongly rely on a few disperser species have

their individual plants mainly concentrated around conspecific individuals when

compared to plant species that weakly rely on many seed disperser species. However, we

did not find the same association for any plant stage in specific.

Consequences of spatial aggregation on seedling and sapling mortality, leaf herbivory

and infection by pathogens

We also evaluated the possible consequences of high clumpiness on seedling and

sapling mortality. Our results show that the survival in seedlings increases with distance

from other conspecific seedling (F=10.36, p=0.0013, R2=0.0072, n=1780), that the

survival of seedlings and saplings combined increases with the distance of any

conspecific seedling or sapling (F=11.27, p=0.0008, R2=0.0065, n=2218), and with the

distance from any conspecific individual (F=15.42, p<0.001, R2=0.0088, n=2218, Fig. 2).

In other words, seedlings and saplings are more likely to survive the farther they are from

other conspecific individuals. We evaluated if this association could be due to high

herbivory and infection by pathogens in areas with high aggregation of conspecific

individuals. We found a trend suggesting that the herbivory and infection by pathogens in

seedlings and saplings are higher in those plants located close to conspecific adults

(F=16.5, p<0.0001, R2=0.09, n=1902; F=38.74, p<0.0001, R2=0.17, n=1902,

respectively, Fig 3). However, herbivory in seedlings is higher in those located far from

other seedlings (F=33.27, p<0.0001, r=0.14, n=1515). When considering seedlings and

120

saplings combined, herbivory is higher in those located far from any other conspecific

(F=50.51, p<0.0001, r=0.16, n=1902). Therefore, our data show that herbivory and

infection by pathogens may explain the mortality of individuals located close to

conspecific adults but not necessarily to conspecific individuals in general.

Seed dispersal interactions and the structure of the plant community

The regression tree model that tested the importance of several biotic and abiotic

variables on the aggregated distribution of individuals in subplots was significant (F=2.3,

p=0.027, R2=0.0676, n=404). Among the significant variables, the maximum

dependence, the dependence of a plant species on its most important seed disperser, and

the interaction between maximum dependence and the percentage of terrestrial Bromelia

were those that predicted the highest percent of variation in the aggregated distribution of

individuals across subplots (1.9 % and 2% respectively). As expected, the aggregated

distribution of individuals (within a 10-meter radius) is positively affected by values of

maximum dependence. In contrast, the aggregated distribution of individuals is

negatively affected by the percentage of terrestrial Bromelia cover. That is, an increase in

the dependence of a plant on its most important seed disperser increases the aggregation

of individuals, whereas an increase in the percentage of terrestrial Bromelia cover

decreases the aggregation distribution of individuals. Therefore, seed dispersal, here

represented by the value of maximum dependence, is the most important variable to

explain the degree of aggregation of individuals across species, among those we assessed.

Discussion

Seed dispersal fundamentally influences plant populations and community

dynamics but such association is difficult to disentangle and to be quantified directly

(Russo et al. 2006), especially because the seed dispersal of a plant species often involves

interactions with multiple dispersers. In this study, we show that seed dispersal

interactions can affect the spatial distribution of a plant species even when multiple

dispersal agents add to the complexity of the interaction. More specifically, in this study

we show that a plant species that strongly relies on a single seed disperser has a more

121

intense aggregation of individuals within short distances (five meters) than a plant species

that weakly relies on many dispersers. Furthermore, we show that aggregations of

individuals do not only occur around adults, but also around conspecific individuals as a

whole. Therefore, plant species that strongly rely on a single seed disperser show a

strongly aggregated pattern that is not solely caused by limited seed dispersal around

adults but also by seed dispersal to particular areas caused by the interaction with

particular animal species (i.e., large-and medium-bodied seed dispersers).

Although seed diameter explains a high percentage of variation in the value of

maximum dependence (Donatti et al. in prep.), seed diameter alone was not significantly

associated with the values of aggregation in different annuli across the plant species we

studied. Likewise, Russo et al. (2007) found that the effect of seed size on the intensity of

aggregation was relatively small for animal-dispersed plants when taking into account

spatial scales of 0-20 meters. Therefore, the aggregated distribution across species may in

fact be determined by the patterns of dispersal generated by particular animal species that

interact with those seeds. For instance, although seeds of Sterculia apetala are not

extremely large, this species is almost exclusively dispersed by the toco toucan

(Ramphastos toco) (Pizo et al. 2008), a medium-bodied bird species. As a result, the

value of maximum dependence in this species is very high (i.e., 0.8).

Seed dispersal interactions of plant species, here summarized by the value of

maximum dependence on seed dispersers, was the most important predictor of the

intensity of spatial aggregation across plant species, which is somewhat surprising for at

least two reasons. First, the existence of many other processes that occur between seed

dispersal and seedling establishment, such as seed and seedling predation, and second,

the strong effects that abiotic factors may have on plant distribution (Clark et al. 1999,

Svenning 1999, Webb & Peart 2000, Harms et al. 2001, Russo et al. 2005). However, the

percentage of variation explained by seed dispersal was very small (~2%) and other

factors not explored in this study, such as topography or proximity of the water table, also

likely play an important role in explaining the spatial distribution of plants.

We also found that the mortality of seedlings and saplings increases with the

proximity of other conspecific individuals, which was also demonstrated by studies that

specifically looked at density-dependent mortality in seedlings (Webb & Peart 1999,

122

Harms et al. 2000, Bell et al. 2006, Comita & Hubbell 2009). In such studies, the

density-dependent mortality in seedling and saplings was likely caused by host-specific

enemies such as herbivores and pathogens, as predicted by the Janzen-Connell hypothesis

(Janzen 1970, Connell 1971). Although we anticipated finding high levels of herbivory

and attack by pathogens in seedlings and saplings located close to conspecific

individuals, this was the case only for individuals located close to a conspecific adult.

Although the mortality of seedlings and saplings was not associated with the proximity of

adults when all species were analyzed together, there was a tendency for this association

in particular plant species, especially the less abundant in this community (i.e. Diperyx

alata, Salacia elliptica and Sterculia apetala). Thus, we suggest that, for this plant

community, a combination of host-specific enemies and intraspecific competition for

abiotic resources may contribute to the high mortality of seedlings and saplings that are

established at short distances from other conspecific individuals.

Regardless of the mechanisms that determine seedling and sapling mortality in this

community, we have shown here that such mortality increases as individuals get more

aggregated. Density-dependent mortality has been acknowledged to open up space for

individuals of other species that may not be vulnerable to enemies that are specialized in

certain plant species. In this way, it has been suggested that seed dispersal would

contribute to plant community diversity especially as a result of seed dispersal limitation

(Harms et al. 2000, Muller-Landau et al. 2002). Although we recognize that the high

mortality of aggregated seedlings and saplings may open opportunities for enhancing or

maintaining plant species diversity, we also recognize that the intense aggregation of

individuals of particular plant species may compromise the populations of these plants.

As seed dispersal affects seedling recruitment and distribution, this process may also

continue to impact fitness by subsequently influencing demography and, ultimately, the

spatial genetic structure of plants (Howe & Smallwood 1982, Howe 1989, Wang & Smith

2002, Hamilton & Miller 2003). Therefore, seed dispersal limitation can have far-

reaching consequences not only for the demography of plants and plant community

composition, but also for the fine-scale patterns of genetic structure and for plant

population differentiation (Vekemans & Hardy 2004). Evidently such repercussions of

seed dispersal warrant further examination.

123

Here, we show that species that strongly rely on a single seed disperser species

show a high aggregated distribution of individuals, which could contribute to a high

mortality of seedlings and saplings. We also found that such mortality could be due to a

high level of herbivory and attack by pathogens in seedlings and saplings that are located

close to conspecific adults. The reasons for the high mortality of seedlings and saplings

located close to other conspecific plants could be competition for abiotic resources, but

those were not assessed in this study. Furthermore, seed dispersal was the main important

predictor of the aggregation intensity of individuals in comparison with several other

variables. Therefore, we conclude that, although different variables such as seed size and

edaphic characteristics can operate at different scales in shaping the distribution and

structure of plant communities, seed dispersal interactions appear to be important in that

respect, even when considering the effects of multiple animal species in dispersing

particular plant species. Furthermore, we suggest that, as large- and medium-bodied seed

disperser species are highly vulnerable to deforestation and hunting, the intense

aggregated distribution of certain plant species can be intensified in a scenario of

defaunation, increasing seedling mortality and, perhaps, leading to a strong fine-scale

spatial genetic structure in those plants.

Acknowledgments

We would like to thank Reese Rogers, Jorge Guedes, Lee Love-Anderegg, Adilson Braga

Samuel and Luisa Haddad for helping with field work. We thank FAPESP (2004/00810-3

and 2008/10154-7), Earthwatch Institute and Conservation International for financial

support. CID was supported by Stanford University and the Zaffaroni Fellowship Fund.

We also thank Conservation International, Lucas Leuzinger and Marina Schweizer for

their permission to work in their properties.

124

References

Bakker JP et al. (1996) Seed banks and seed dispersal: important topics in restoration ecology. Acta Bot. Neerl. 45: 461-490.

Bascompte J, Jordano P, Melián CJ, Olesen JM (2003) The nested assembly of plant-animal mutualistic networks. PNAS 100 (16): 9383-9387.

Bell T, Freckleton RP, Lewis OT (2006) Plant pathogens drive density-dependent seedling mortality in a tropical tree. Ecology Letters 9: 569–574.

Chapman LJ, Chapman CA, Wrangham RW (1992) Balanites wilsoniana: Elephant Dependent Dispersal? Journal of Tropical Ecology 8 (3): 275-283. 1992.

Clark JS, Silman M, Kern R, Macklin E, HilleRisLambers J (1999) Seed dispersal near and far: patterns across temperate and tropical forests. Ecology 80: 1475–1494.

Comita LS, Hubbell SP (2009) Local neighborhood and species’ shade tolerance influence survival in a diverse seedling bank. Ecology 90: 328–334.

Condit RPS et al. (2000) Spatial patterns in the distribution of tropical tree species. Science 288: 1414-1418.

Connell JH, Tracey JG, and Webb LJ (1984) Compensatory recruitment, growth, and mortality as factors maintaining rain forest tree diversity. Ecological Monographs 54:141–164.

Connell, JH (1971) On the role of natural enemies in preventing competitive exclusion in some marine animals and in rain forest trees. Pages 298–312 in P. J. den Boer and G. R. Gradwell, editors. Dynamics of populations. Centre for Agricultural Publishing and Documentation, Wageningen, The Netherlands.

Dirzo R, Domínguez CA (1995) Plant-herbivore interactions in Mesoamerican tropical dry forests. Pages 304-325 in S. H. Bullock, H. A. Mooney, and E. Medina, editors. Seasonally dry tropical forests. Cambridge University Press, Cambridge.

Donatti CI, Guimarães PR, Galetti M, Pizo MA, Marquitti FMD, Dirzo R (2011) Analysis of a hyper-diverse seed dispersal network: modularity and underlying mechanisms. Ecology Letters 14: 773-781.

Fragoso JMV, Silvius KM, Correa JA (2003) Long-distance seed dispersal by tapirs increases seed survival and aggregates tropical trees. Ecology 84(8): 1998-2006.

Guimarães PR, Galetti M, Jordano P (2008). Seed dispersal anachronisms: rethinking the fruit extinct megafauna ate. PLOS One 3: e1745.

Hamilton MB, Miller JR (2002) Comparing relative rates of pollen and gene flow in the island model using nuclear and organelle measures of population structure. Genetics 162(4): 1897-1909.

Harms KE, Condit R, Hubbell SP, Foster RB (2001) Habitat associations of trees and shrubs in a 50-ha neotropical forest plot. Journal of Ecology 89: 947–959.

Harms KE, Wright SJ, Calderon O, Hernandez A, Herre EA (2000) Pervasive density-dependent recruitment enhances seedling diversity in a tropical forest. Nature 404: 493-495.

Hardy OJ, Sonké B (2004) Spatial pattern analysis of tree species distribution in a tropical rain Forest of Cameroon: assessing the role of limited dispersal and niche differentiation. Forest Ecology and Management 197: 191-202.

Herrera CM, Jordano P, Lopez-Soria L, Amat JA (1994) Recruitment of a mast-fruiting, bird-dispersed tree: bridging frugivore activity and seedling establishment. Ecological Monographs 64:315–344.

125

Howe HF, Smallwood J (1982) Ecology of seed dispersal. Annual Review of Ecology and Systematics, 13, 201–228.

Howe HF (1989) Scatter- and clump-dispersal and seedling demography: hypothesis and implications. Oecologia 79: 417-426.

Howe HF, Miriti MN (2004) When seed dispersal matters. Bioscience 54(7): 651-660. Hubbell SP (2001) The unified neutral theory of biodiversity and biogeography.

Princeton, New Jersey, USA: Princeton University Press. Hubbell SP, Foster RB, O’Brien ST, Harms KE, Condit R, Wechsler B, Wright SJ, Loo

de Lao S (1999) Light-gap disturbances, recruitment limitation, and tree diversity in a neotropical forest. Science 283: 554–557.

Hubbell, SP (1979) Tree dispersion, abundance, and diversity in a tropical dry Forest. Science 203: 1299-1309.

Janson CH (1983) Adaptation of fruit morphology to dispersal agents in a neotropical forest. Science 219: 187-189.

Janzen DH, Martin PS (1982) Neotropical anachronisms-the fruits the gomphoteres ate. Science 215: 19-27.

Janzen D H (1970) Herbivores and the number of tree species in tropical forests. Am Nat 104: 501–528.

Jordano P, Bascompte J, Olesen JM (2006) The ecological consequences of complex topology and nested structure in pollination webs. Pages 173-199 in N.M. Waser, J. Olerton, editors. Plant-pollinator interactions: from specialization to generalization. The University of Chicago Press, London, UK.

Julliot C (1998) Impact of seed dispersal by the red howler monkeys Alouatta seniculus on the seedling population in the understory of tropical rain forest. Journal of Ecology 85: 431-440.

Kinnaird MF (1998) Evidence for effective seed dispersal by the Sulawesi red-knobbed hornbill, Aceros cassidix. Biotropica 30(1): 50-55.

Memmott J, Waser NM, Price MV (2004) Tolerance of pollination networks to species extinctions. Proceedings of the Royal Society of London Series B-Biological Sciences 271(1557): 2605-2611.

Muller-Landau HC, Wright SJ, Calderon O, Condit R, Hubbell SP (2008) Interspecific variation in primary seed dispersal in a tropical forest. Journal of Ecology 96(4): 653-667.

Muller-Landau HC, Wright SJ, Calderón O, Hubbell SP, Foster RB (2002) Assessing recruitment limitation: concepts, methods and examples for tropical forest trees. Pages 35-53 in D.J. Levey, W.R. Silva, M. Galetti, editors. Seed Dispersal and Frugivory: Ecology, Evolution and Conservation. CAB International, Wallingford, UK.

Pizo MA, Donatti CI, Guedes NMR, Galetti M (2008) Conservation puzzle: Endangered hyacinth macaw depends on its nest predator for reproduction. Biological Conservation 141: 792-769.

Ricklefs RE (2004) A comprehensive framework for global patterns in biodiversity. Ecology Letters 7: 1–15.

Rosenzweig ML (1995) Species diversity in space and time. Cambridge University Press, New York, New York, USA.

126

Russo SE, Potts MD, Davies ST, Tan S (2007) Determinants of tree species distributions: comparing the roles of seed dispersal, seed size and soil specialization in a Bornean Rainforest. Pages 499-518 in Dennis AJ, Schupp EW, Green RJ, Westcott DA, editors. Seed dispersal: theory and its application in a changing world. CAB International, Wallinford, UK.

Russo SE, Portnoy S, Augspurger CK (2006) Incorporating animal behavior into seed dispersal models: implications for seed shadows. Ecology 87(12): 3160-3174.

Russo SE, Davies SJ, King DA, Tan S (2005) Soil-related performance variation and distributions of tree species in a Bornean rain forest. Journal of Ecology 93: 879-889.

Scariot A (1999) Forest fragmentation effects on palm diversity in central Amazonia. Journal of Ecology 87: 66–76.

Siedler TG, Plotkin JB (2006) Seed dispersal and spatial pattern in tropical trees. Seed dispersal and spatial pattern in tropical tree. Plos Biology 4 (11): 2132-2138.

Strauss YS, Irwin RE (2004) Ecological and evolutionary consequences of multispecies plant-animal interactions. Annu. Rev. Ecol. Evol. Syst. 35: 435-466.

Svenning J-C (1999) Microhabitat specialization in a species-rich palm community in Amazonian Ecuador. Journal of Ecology 87: 55–65.

Tilman, D, Pacala S (1993) The maintenance of species richness in plant communities. Pages 13–25 in R. E. Ricklefs and D. Schluter, editors. Species diversity in ecological communities: historical and geographical perspectives. University of Chicago Press, Chicago, Illinois, USA.

Van der Heijden MGA, Boller T, Wiemken A, Sanders I R (1998) Different arbuscular mycorrhizal fungal species are potential determinants of plant community structure. Ecology 79(6): 2082-2091.

Vekemans X, Hardy O J (2004) New insights from fine-scale spatial genetic structure analyses in plant populations. Molecular Ecology 13: 921-935.

Volvov I, Banavar JR, Hubbell SR, Maritan A (2009) Inferring species interactions in tropical forests. PNAS 109(33): 13854-13859.

Wang BC, Smith TB (2002) Closing the seed dispersal loop. Trends in Ecology and Evolution, 17, 379–385

Webb CO, Peart DR (2000) Habitat associations of trees and seedlings in a Bornean rain forest. Journal of Ecology 88: 464–478.

Webb CO, Peart D R (1999) Seedling density dependence promotes coexistence of Bornean rainforest trees. Ecology 80: 2006–2017.

Wenny DG, Levey DL (1998) Directed seed dispersal by bellbirds in a tropical cloud forest. PNAS 95 (11): 6204-6207.

Wheelwright NT (1985) Fruit size, gape width, and the diets of fruit-eating birds. Ecology 66: 808-818.

127

Figures

Figure 1. Association between the aggregated distribution (Omega) of individuals and the value of maximum dependence of the plant on its most important seed disperser. Top: Value of aggregation for individuals located within three meters of conspecifics. Center: Value of aggregation for individuals located within four meters of conspecifics. Bottom: Value of aggregation for individuals located within five meters of conspecifics.

!

"#

""

"$

"%

"&

"'

()*+,-.#/%)0

%

&

'

1

2

3

()*+,-.#/&)0

#4$ #4% #4& #4' #41 #42 #43

3

"#

"$

"&

"1

"3

5,67)8)-9*:*;9*;<*

()*+,-.#/')0

128

Figure 2. Logistic regression between the survival of seedlings and saplings and the distance from the closest conspecific individual (m).

129

Figure 3. Associations between the distance (m) from the closest conspecific adult and a) the percentage of the plant with foliar herbivory, and b) the percentage of the plant with foliar attack by pathogens.