Understanding Patterns and Processes in Plant–Hummingbird Mutualistic Networks in the Neotropics

Understanding patterns and processes in plant–hummingbird mutualistic networks in the Neotropics

Dissertation

for attaining the PhD degree

of Natural Sciences

submitted to the Faculty of Biology

of the Johann Wolfgang Goethe University

in Frankfurt am Main

María Alejandra Maglianesi

from Santa Fe, Argentina

Frankfurt 2014

(D 30)

Faculty of Biology of the

Johann Wolfgang Goethe University accepted as a dissertation.

Dean: Prof. Dr. rer. nat. Meike Piepenbring

First reviewer: Prof. Dr. Katrin Böhning-Gaese

Second reviewer: Prof. Dr. Alexandra-Maria Klain

Date of disputation: ______

TABLE OF CONTENTS

1. SUMMARY ...... 1

2. INTRODUCTION ...... 3

2.1 Biotic interactions in ecological communities ...... 3

2.2 Plant–pollinator mutualistic networks ……………………...... 4

2.3 Functional structure of pollinator communities …………………………….…………….. 6

2.4 Why specialization in ecological networks matters? ………………………………….... 7

2.5 Mechanisms underlying ecological specialization …..………………………...... 8

2.6 Potential and realized interactions in plant–pollinator networks ...... 9

2.7 Plant–hummingbird interactions ...... 10

3. STRUCTURE OF THE THESIS AND RESEARCH QUESTIONS …………….….………...... 11

3.1 Thesis structure ……………………………..……………………………..…….……………………….. 11

3.2 Research questions …………………………………..…………………………………...…….…..….. 11

3.2.1 Do the functional structure and specialization in plant–hummingbird 12 networks change across elevation? ………………….….………......

3.2.2 How do the morphological traits of interacting species shape plant– 12 hummingbird interactions? ………………….….………......

3.3.3 How do the floral morphology and competition among hummingbirds 13 influence the foraging preferences of the birds on artificial and natural flowers? ..

4. RESEARCH AREA AND STUDY SYSTEM ……………………….….…..……...... 14

4.1 Research area and data collection ……………………………………………………………….. 14

4.2 Study system ………………………………………………………………………………………………… 18

4.2.1 Hummingbirds ……………………………………………………………………………………………. 18

4.2.2 Hummingbird–visited plants …………………………………………………………………..….. 20

5. RESULTS AND DISCUSSION ...... 22

5.1 Functional structure and specialization in plant–hummingbird networks 22 along an elevational gradient ......

iii

5.1.1 Functional structure of hummingbird assemblages varies across elevation .. 22

5.1.2 Hummingbird specialization decreases with elevation ………………………………. 23

5.2 Influence of morphological traits on patterns of interactions in plant- 24 hummingbird networks ......

5.2.1 Ecological specialization is associated with morphological traits in plant– 24 hummingbird networks ......

5.2.2 The role of trait matching in shaping plant–hummingbird interactions …….. 25

5.3 Combining experimental and observational approaches to understand 26 patterns of plant–pollinator interactions ………………………………………………………

5.3.1 Interaction niche of hummingbird species under experimental conditions … 27

5.3.2 Interaction niche of hummingbird species under natural conditions ………….. 28

6. CONCLUSIONS ...... 29

7. ZUSAMMENFASSUNG ...... 31

8. ACKNOWLEDGEMENTS ...... 37

9. REFERENCES ...... 39

10. APPENDICES ...... 52

Appendix 1: Functional structure and specialization in tropical plant– 53 hummingbird interaction networks across elevations ......

Appendix 2: Morphological traits determine specialization and resource use in 81 plant–hummingbird networks in the Neotropics ......

Appendix 3: Different foraging preferences of hummingbirds on artificial and 121 natural flowers reveal mechanisms structuring plant–pollinator interactions ……..

Appendix 4: List of hummingbird species recorded at the study sites ...... 152

Appendix 5: List of plant species recorded at the study sites ...... 154

Appendix 6: Curriculum vitae ...... 158

Summary

1. SUMMARY

Ecological communities are organized in complex ecological networks where species relate to each other through different types of interactions. The mutually beneficial interactions between plants and their animal pollinators have a pervasive influence in community dynamics and have contributed to the generation of Earth’s biodiversity. Hence, a profound knowledge of how plant–pollinator networks are structured is essential to understand evolution, the maintenance of biodiversity and the consequences of species extinction. Functional traits of species influence patterns of interactions in pollination networks. Trait–based analysis can reveal processes structuring mutualistic networks and ecological communities. However, at present there are few studies that link species functional traits with patterns of interactions in plant–pollinator networks, especially in the tropics.

I investigated patterns in functional structure and specialization in plant– hummingbird mutualistic networks across elevations and the processes driving these patterns in three tropical forests of Costa Rica. I quantified different metrics of functional community structure based on three morphological traits (bill length, bill curvature and body mass) of hummingbird species at three elevations. I used pollen carried by mist–netted hummingbird individuals to construct plant–hummingbird networks at the individual and species level at each elevation. My results show consistent patterns in the functional structure of hummingbird assemblages and specialization of hummingbird species and individuals across elevations. Hummingbird assemblages varied from being functionally even and over–dispersed in the lower elevations to uneven and clustered in high–elevation environments. Accordingly, hummingbird species and individuals were more specialized at low and mid elevations than at the highest elevation.

I used recent advances in the analysis of quantitative networks to assess the importance of morphological traits in shaping plant–hummingbird interactions. First, I tested the effects of the three avian morphological traits and abundance on ecological specialization of hummingbird species. Second, I analyzed whether interaction strength in the networks was associated with the degree of trait matching between corresponding pairs of morphological traits in plant and hummingbird species. Third, I

Summary explored whether trait matching between interacting species was related to resource handling times by hummingbird species. I found strong and significant associations between interaction strength and the degree of trait matching in corresponding morphological traits of hummingbird and plant species. Moreover, the degree of trait matching was negatively associated with the handling time of nectar resources by hummingbirds.

Finally, I used experimental and observational data to explore the foraging preferences of hummingbird species for artificial and natural flowers with specific morphology. Hummingbird species visiting artificial feeders with unlimited nectar preferred short artificial flowers over long–straight and long–curved flowers. Under natural conditions, however, co–occurring hummingbird species preferred to feed on plant species with floral traits matching their bill morphology.

Overall, my results suggest different processes and mechanisms underlying patterns of functional community structure and interactions in plant–hummingbird mutualistic networks. Even and over–dispersed hummingbird assemblages at the lower elevations suggest a high level of floral resource partitioning leading to specialized plant– hummingbird interactions. In contrast, an uneven and clustered functional structure of the hummingbird assemblage in the highlands may result from environmental filtering in generalized hummingbird interactions. My findings highlight the crucial role of morphological traits for structuring plant–hummingbird networks and that patterns of interactions are closely associated with morphological matches between plant and hummingbird species. Additional factors however, such as competition for resources, are also important and may enforce patterns of niche segregation between co– occurring hummingbird species in natural communities.

Trait–based analyses of quantitative networks combined with experimental and observational approaches are essencial for a comprehensive understanding of the evolutionary and ecological causes determining patterns of interactions in plant– pollinator networks. These approaches can be very valuable for the comprehension of the mechanistic bases of community assembly and network structure in other types of interaction networks.

Introduction

2. INTRODUCTION

2.1 Biotic interactions in ecological communities

Organisms of different species in a community are not independent entities, but they interact with each other forming complex interaction systems. Biotic interactions can be classified into two general types: trophic (e.g. predator–prey relationships) and mutualistic (e.g. seed dispersal interactions). Mutualism is the interaction type where both interacting species are mutually benefited (Bascompte and Jordano 2014). In plant–pollinator interactions for example, the plant is benefited by being pollinated and the pollinator gets food resources from the flowers, such as nectar or pollen (Inouye and Pyke 1988). The mutual dependence of interacting organisms can contribute to generate and maintain species diversity within communities and ecosystems (Terborgh et al. 2002, Bronstein 2008, Bastolla et al. 2009). For example, the co–evolution between plants and their animal pollinators is assumed to contribute considerably to explain the vast diversity of angiosperms (about 300,000 species) (Van der Niet et al. 2014). Studies on mutualistic interactions, therefore, may help to better understand patterns of biodiversity (Bascompte and Jordano 2014).

Plant–animal mutualistic interactions are crucial in providing ecosystem functions and services (Bronstein 2008). Globally, animals supply pollination services to about 75% of crop species and enable reproduction in up to 78% and 94% of flowering plants in temperate and tropical ecosystems, respectively (Ollerton et al. 2011, Vanbergen 2013). Current rates of anthropogenic habitat alteration have raised awareness of a global biodiversity crisis (Chapin et al. 2000). Declines in the number of pollinator species have been reported by several studies (Goulson 2003, Thorp and Shepherd 2005, Kosior et al. 2007), and extinction rates are expected to increase due to predicted global environmental changes (Pereira et al. 2010). Pollinator extinctions may lead to the loss of interactions on which flowering plants depend for pollination function (Ollerton et al. 2011). Consequently, anthropogenic disruptions of plant– pollinator interactions may lead to decreased plant productivity and reproductive success with lost of plant species which in turn may cause further extinctions of

Introduction pollinator species (Koh et al. 2004, Anderson et al. 2011, Colwell et al. 2012, Garibaldi et al. 2013).

Plant–pollinator interactions may be particularly susceptible to anthropogenic changes, owing to their sensitivity to the phenology, behavior, physiology, and relative abundances of multiple species (Tylianakis et al. 2008a). Predicted pollinator declines may ultimately lead to the disruption of plant communities, which in turn leads to the collapse of the ecosystem services they maintain (Chapin et al. 2000, Hooper et al. 2011). Consequently, anthropogenic disruptions of mutualisms such as pollination threaten species diversity as well as ecosystem functioning and services (Bronstein 2008, Tylianakis et al. 2008b, Burkle et al. 2013). These consequences may be particularly severe in the tropics, where dependence on animal pollination is higher than in temperate regions (Ollerton et al. 2011).

2.2 Plant–pollinator mutualistic networks

A particularly useful approach to study mutualistic interactions in highly diversified tropical communities is the interpretation of these interactions in the context of ecological networks (Ings et al. 2009). An ecological network is a representation of the plant–animal interactions in a community, in which species (nodes) are connected by pairwise interactions (links). These ecological networks are used to describe and compare the structure of real ecosystems, whereas network analysis is used to investigate the effects of network structure on processes such as ecosystem stability (Bascompte and Jordano 2007, Thébault and Fontaine 2010). Traditionally, the study of ecological networks has focused on consumer–resource relations in food webs (mostly aquatic) and terrestrial host–parasitoid systems. More recently, network analysis has been extensively applied to the study of plant–pollinator interactions (e.g. Vázquez and Aizen 2004, Santamaría and Rodríguez-Gironés 2007, Petanidou et al. 2008, Kaiser-Bunbury et al. 2014, Maglianesi et al. 2014). These studies have considerably improved our knowledge on processes behind patterns of interactions in ecological communities.

Introduction

One of the most important contributions from network analysis is the recognition that several networks, despite differences in the nature of their nodes, have similar structural properties (referred to as “invariant properties”; Jordano et al. 2003). This is important because a common architecture may be related to common processes that determine the observed patterns in mutualistic networks. One well–studied structural property of mutualistic networks is nestedness. A nested structure occurs when specialists interact with proper subsets of those species interacting with the more generalist species (Bascompte et al. 2003). This nested structure is assumed to reduce competition and enhance the number of coexisting species, thereby increasing diversity in mutualistic networks (Bastolla et al. 2009). Nestedness has been shown to contribute to community persistence and stability by buffering the transmission of perturbations through the whole community preventing species extinctions (Memmott et al. 2004, Bascompte and Jordano 2007).

More recently, a significantly modular pattern characterized by the existence of densely connected, non–overlapping subsets of species (i.e. modules or compartments) has been identified. In this case, modules are composed of species having many interactions among themselves as well as very few with species in other modules (Olesen et al. 2007, Dupont and Olesen 2009). Studies on modularity may help to identify key species serving as hubs and connectors which play a fundamental role to network persistence (Olesen et al. 2007). Addressing conservation priorities to these species may considerably contribute to the maintenance of biodiversity.

Another important structural property of plant–pollinator networks is the specialization of pollinators on specific plant resources. Specialization is closed related to nestedness and modularity since a nested and modular structures are based on specialization asymmetry (Blüthgen et al. 2008). Considerable empirical work indicates that specialization differs in predictable ways between different types of biotic interaction in ecological networks (see Poisot et al. 2011). Therefore, specialization can be derived as an invariant property of these networks. Several studies show that generalization is a common pattern in plant–pollinator mutualistic networks (Olesen 2000, Ollerton and Watts 2000). However, specialization is usually higher than

Introduction expected in randomly interacting communities (e.g. Blüthgen et al. 2007) and has been suggested to influence species assemblages in ecological communities.

2.3 Functional structure of pollinator communities

The structure of ecological communities can be described by species’ functional traits, defined as measurable properties of individuals that influence their performance (McGill et al. 2006). Trait–based analysis has proven to be a particularly sensitive approach, offering insights into the processes that influence species coexistence within ecological communities. Moreover, measures of functional community structure may act as an indicator of environmental filtering and competition (Cornwell et al. 2006). Both processes have been found to be important in structuring local pollinator assemblages (Graham et al. 2009). Environmental conditions may act as environmental filters sorting species according to their functional traits (Keddy 1992, Naeem and Wright 2003). Thus, environmental filtering reduces the spread of trait values leading to a clustered functional structure with an increased functional similarity among the species within the community (Cornwell et al. 2006). In contrast to environmental factors, competition tends to limit the functional similarity of co–occurring species, thereby reducing interspecific competition and thus promoting species coexistence (MacArthur and Levins 1967). The consequence of such limiting similarity is an overdispersed functional structure of the community.

Our understanding of how communities are organized could be advanced by examining patterns in species assemblages along environmental gradients. Some studies have investigated the effects of environmental gradients on species assemblages and structural properties of plant–pollinator networks (e.g. Bates et al. 2011, Benadi et al. 2013). For example, Ramos-Jiliberto et al. (2010) found that nestedness and modularity in pollination networks systematically decreased with elevation, suggesting that the severe abiotic conditions found at increased elevations may determine less organized pollination networks. Studies on specialization in different types of mutualistic networks along environmental gradients show contrasting patterns. Whereas some studies have found specialization in pollination networks to decrease with increasing latitude (Olesen and Jordano 2002, Dalsgaard et 6

Introduction al. 2011), other studies have found specialization in pollination and dispersal networks to be higher at temperate than at tropical latitudes (Schleuning et al. 2012). Variation in specialization along environmental gradients has been related to plant diversity, phenology and contemporary climate conditions (Dalsgaard et al. 2011, Schleuning et al. 2012, Benadi et al. 2013). However, it is unsolved whether specialization in plant– pollinator networks may be associated to changes in the functional structure of pollinator communities along environmental gradients.

2.4 Why specialization in ecological networks matters?

Specialization is a central concept in community ecology because it influences species co–existence and the structure and the stability of ecological communities (Thompson 1994, Waser et al. 1996). Specialization can be viewed from an ecological or an evolutionary perspective. Ecological specialization is the state of being specialized, which means that organisms use a small proportion of the available resources in a community (Blüthgen et al. 2007), whereas evolutionary specialization is the process by which an organism adapts to an increasingly narrow range of resources (Armbruster 2000). The outcome of ecological specialization is that specialized organisms will have higher performances in a small subset of the range of their resources (Poisot et al. 2012).

In plant–pollinator networks, more specialized pollinators can be associated with higher energy returns relative to less–specialized foraging (Gegear and Thomson 2004). For example, specialized pollinators can handle flowers more quickly than can generalists reducing the costs of foraging activities (Strickler 1979, Chittka and Thomson 1997). From the plant’s perspective, pollinators specialized on a few food resources may increase transfer of conspecific pollen with a consequent increase in seed set of plants and a positive impact on their reproduction (Waser 1978, Brosi and Briggs 2013). The contribution of a pollinator toward plant fitness (i.e. effectiveness) may determine its importance for the plant's evolutionary ecology. Thus, effective pollinators may shape the evolution of floral characters (Wilson 1995, Schemske and Bradshaw 1999) and the evolution of plant lineages (Crepet 1983, Grimaldi 1991). In

Introduction this regard, pollinators have the potential to change the structure and diversity of plant communities.

2.5 Mechanisms underlying ecological specialization

Several mechanisms have been proposed to explain patterns of interactions and causes of ecological specialization in plant–pollinator networks (Stang et al. 2007, Dalsgaard et al. 2011, Tur et al. 2013). Recent studies focus on two mechanisms: matching species traits (trait hypothesis) and random interactions among individuals (neutrality hypothesis). The trait hypothesis refers to the occurrence of interactions resulting from the correspondence between phenotypic traits of interacting individuals, either in a complementary way or as a barrier (e.g. Linhart 1973, Stiles 1975, Stang et al. 2009, Junker et al. 2013). The neutral hypothesis, in contrast, refers to the occurrence of interactions resulting from the random encounter among individuals, so that abundant species tend to encounter other species more frequently than rare species (Dupont et al. 2003, Vásquez 2005).

Many studies have examined how plants have evolved specific morphological traits that facilitate or hinder access to nectar rewards to a small subset of potential visitors (e.g. Heinrich and Raven 1972, Faegri and van der Pijl 1979, Stiles 1981, Ornelas et al. 2007). These floral traits match quite tightly to the morphological traits of a few pollinator species in the community, giving rise to the concept of pollination syndromes. A pollination syndrome is defined as a suite of ﬂoral traits, including rewards, associated with the attraction and utilization of a speciﬁc group of animals as pollinators (Faegri and van der Pijl 1979). Previous research has shown that trait matching influences patterns of interactions between plant and pollinator species (Stiles 1975, Wolf et al. 1976, Dalsgaard et al. 2009, Stang et al. 2009), leading to preferences of pollinators for particular plant species. However, most research has been focused on specific pollinator species and their food plants (e.g. Temeles et al. 2009, Dohzono et al. 2011). Studies involving whole assemblages of interacting plants and pollinators are much rarer (Olesen and Jordano 2002), which highlights the need for more integrated research at the community level.

Introduction

A comprehensive understanding of the processes leading to patterns of interactions in plant–pollinator networks requires an understanding of the processes occurring at the level of both species and individuals. As species are the basic focal unit for investigating the complex structure in natural communities, the majority of studies on ecological networks are centred at the species level. However, individuals may show substantial variation in resource use and this variation in turn may be translated into topological properties of networks that depict interactions among individuals and the food resources they consume (individual–resource networks) (Pires et al. 2011, Wells and O'Hara 2012). Given that network analysis has been seldomly applied to the study of individual interactions within populations (but see Tur et al. 2013), there is not yet a clear view of how interactions between plants and their pollinators at both the species and individual level may influence the structure of pollinator communities.

2.6 Potential and realized interactions in plant–pollinator networks

An important distinction related to resource use in plant–pollinator networks, is that between potential and realized interactions. The full range of plant species that pollinators are potentially able to use determines their potential interactions. This full range of food resources may be restricted by additional factors (e.g. interspecific competition with other pollinators) to the subset of plant species that pollinators actually use, determining their realized interactions (Pauw 2013).

Morphological constraints may be an important factor defining the potential interactions of pollinators with plant species. For instance, floral morphologies such as long and/or curved corollas may restrict the access to nectar rewards (Ornelas et al. 2007, Maglianesi et al. 2014), leading to preferences of pollinators for plant species with specific morphologies. Another factor that may influence patterns of realized interactions in plant–pollinator networks is competition. Competition for floral resources has been found to be intense among different pollinator species (Colwell 1973, Brown and Bowers 1985). Hence, competition might play an important role in determining foraging preferences of pollinators and therefore in defining their realized interactions.

Introduction

Combining experimental and observational data may be a powerful approach to quantify potential and realized interactions between plants and pollinators. On one hand, potential interactions can be estimated from controlled experiments in which pollinators' foraging preferences are measured across a range of resource types (e.g. different artificial flower types) (Devictor et al. 2010). On the other hand, the realized interactions of pollinators can be derived from observations of plant–pollinator interactions in the real world.

2.7 Plant–hummingbird interactions

In Neotropical communities, hummingbirds (Trochilidae) are considered to be the most effective pollinators (Castellanos et al. 2003). They comprise one of the largest families of birds (ca 340), with 52 species occurring in Costa Rica. Hummingbird assemblages include species that differ in their morphology and functional roles, as well as in their degree of specialization. For example, hermit species with long and curved bills are mostly trapliners and are considered to be highly specialized. In contrast, non–hermit species present a wider range of bill morphologies and feeding strategies with variable degrees of specialization (see Feinsinger and Colwell 1978). Bill morphology of hummingbird species varies substantially in curvature and length and often influences what plant species they visit (Temeles et al. 2009). Some studies suggest that morphological fitting between pairs of corresponding traits in plant and hummingbird species largely determine pattern of interactions within the community (Colwell 1973, Temeles et al. 2010).

Plant–hummingbird mutualistic networks are ecologically important and they are a suitable system to study the mechanistic bases of community assembly and network structure. First, hummingbirds vary considerably in their morphologies and functional roles allowing for the analysis of patterns in functional community structure. Second, hummingbird assemblages include species with a broad range of degree of specialization, from very specialized to completely generalized (Stiles 1981). Third, hummingbirds occur across a wide range of environmental conditions and can be found along a large elevational gradient (i.e. from sea level to more than 4,000 m a.s.l.). I used hummingbirds as study system to investigate patterns in community 10

Introduction functional structure and specialization at the individual and species level across elevations.

3. STRUCTURE OF THE THESIS AND RESEARCH QUESTIONS

3.1 Thesis structure

This study comprises my original work that describes patterns in hummingbird assemblage structure and provides a comprehensive analysis of the major processes leading to patterns of interactions in plant–hummingbird networks. I present a conceptual framework to better understand the main ideas structuring my thesis. The major aims of my thesis are (a) to analyse the functional assemblage structure and specialization of hummingbirds along a tropical elevational gradient; and (b) to investigate the influence of species morphological traits on patterns of interactions in plant–hummingbird mutualistic networks. In the last part of the thesis, I included three appendices with scientific manuscripts submitted for publication in peer–reviewed journals, which deals with specific questions addressing my overarching study topic. These appendices, therefore, are structured as a scientific paper, with an abstract and an introduction, followed by a detailed description of the methodology and sections presenting the results, discussion and conclusion. Supplementary information and references in each manuscript are presented in the corresponding appendix. My contribution to each of these manucripts is briefly explained prior to the respective appendix.

3.2. Research questions

In this thesis, I aimed to investigate the mechanistic bases of community assembly and network structure in plant–hummingbird mutualistic networks in three tropical forests of Costa Rica. I focused my research on three main questions which are shown in a conceptual framework in Fig. 1.

Introduction

3.2.1 Do the functional structure and specialization in plant–hummingbird networks change across elevation?

I explored whether the functional structure of hummingbird assemblages varies with elevation. I also investigated the influence of elevation on specialization of hummingbirds at the individual and species level. Ecological theory suggests that multiple environmental factors act as a filter allowing only a narrow range of species and functional roles to coexist in harsh, high–elevation environments (Keddy 1992, Mouchet et al. 2010). Based on this hypothesis, I expected hummingbird functional structure to be uneven and clustered at the highest elevation compared to the lower elevations. Regarding to specialization, I hypothesized that hummingbird species and individuals are specialized at the lower elevations because high levels of competition might cause pronounced resource partitioning in highly diversified assemblages (Fig. 1A).

3.2.2 How do the morphological traits of interacting species shape plant–hummingbird interactions?

By using several methodological approaches, I assessed the importance of morphological traits for structuring plant–hummingbird networks in neotropical forests. First, I analyzed the influence of avian morphological traits and abundance on ecological specialization of hummingbird species. Second, I tested whether interaction strength in the networks was associated with the degree of trait matching between corresponding pairs of morphological traits in plant and hummingbird species. Third, I explored whether trait matching was related to resource handling times by hummingbird species. Based on optimal foraging theory (MacArthur and Pianka 1966), I expected hummingbirds to forage at higher frequencies on plants that match well with their bill morphology and that this association is related to an increased efficiency of resource use (Fig. 1B).

Introduction

Figure 1. Schematic overview of the research questions. Several processes through different mechanisms may act as important determinants of functional community structure and interaction patterns in plant–hummingbird mutualistic networks. This framework was structured around three main questions (Q 1–3). A) Do the patterns in functional assemblage structure and specialization of hummingbirds change across elevation? (Q1). B) How do the morphological traits of interacting species shape plant– hummingbird interactions? (Q2). C) How do the floral morphology and competition among hummingbirds influence the foraging preferences of the birds on artificial and natural flowers? (Q3).

3.2.3 How do the floral morphology and competition among hummingbirds influence the foraging preferences of the birds under artificial and natural conditions?

Finally, I assessed whether hummingbird species prefer to feed on specific flowers types under experimental and natural conditions and whether foraging preferences

Introduction under unlimited nectar resources provided in artificial feeders differ from those in the real world. I also explored how different levels of competition relate to foraging preferences of hummingbird species. Whereas unlimited nectar resources offered to hummingbirds during controlled experiments may reduce competition, the limited floral resources in natural communities is likely to determine different levels of competition among hummingbird species (Colwell 1973, Brown and Bowers 1985). Based on this idea, I expected hummingbird species to differ in foraging preferences for flower types under artificial and natural conditions (Fig. 1 C).

4. RESEARCH AREA AND STUDY SYSTEM

4.1 Research area and data collection

Field work was conducted within the La Selva–Braulio Carrillo corridor on the Caribbean slope of the Cordillera Central in northeastern Costa Rica (Fig. 2). This area extends from La Selva Biological Station (LS) (ca. 1.500 ha) to the Braulio Carrillo National Park (ca. 45.000 ha) and comprises what may be the most extensive altitudinal range (2.871 m) of primary tropical forest protected in Central America (Hartshorn and Peralta 1988, Pringle 1988). About 75% of the vegetation remains in mature forest, and most of the rest has reverted to patches of secondary growth forest (Fig. 3). The area is characterized by a high level of biodiversity, including more than 1.900 species of plants, 330 species of trees and 436 species of birds only in la Selva Biological Station (Hammel 1990).

I selected three study sites representing three tropical forest types located at different elevations (Fig. 3): wet forest (50 m; 10°26’N, 84°01’W) in LS, pre–montane forest (1.000 m; 10°16’N, 84°05’W) and lower montane wet forest (2.000 m; 10°11’N, 84°07’W) in the park (Holdridge 1967). Mean annual temperature ranges from 25°C in the lowlands to 14°C in the highlands, while mean annual precipitation ranges from 4.300 mm in the lowlands to 2.200 mm in the highlands (Blake and Loiselle 2000, TEAM 2013). The dry season lasts from January to April and the wet season from May to November.

Research area and study system

Figure 2. La Selva–Braulio Carrillo corridor on the Caribbean slope of the Cordillera Central in northeastern Costa Rica showing the three study sites at low, mid and high elevation at 50 m, 1.000 m and 2.000 m, respectively. Source: Campos 2014.

Figure 3. Forest habitats in the three study sites located at La Selva Biological Station

(50 m) (A,B) and the Braulio Carrillo National Park at 1.000 m (C) and 2.000 m (D) in northern Costa Rica. 15

Research area and study system

The study was conducted from May to September 2011 and from December 2011 to April 2012. During the period of sampling, I collected data on abundances and functional traits of flower resources and hummingbirds and on plant–hummingbird interactions (Fig. 4, Fig. 5A). I also sampled pollen loads carried by hummingbirds to determine plant–hummingbird interactions at the individual and species level (Fig. 5B,C,D). Data were collected during seven sampling periods of approximately 10 days each. Pollen loads were analysed for a subset of four periods covering both the wet and the dry season (see Appendices 1 and 2 for detailed methodology).

Figure 4. Mist–netting used for capturing hummingbird individuals and estimating species relative abundances (A). Measurement of morphological traits in hummingbirds: body mass, bill length and bill curvature (B,C and D, respectively).

To investigate foraging preferences of hummingbird species for artificial and natural flower types, I performed field experiments with artificial feeders for a subset of two sampling periods per forest type, covering both seasons (Fig. 6). Experiments were carried out during five days in August (wet season) and four days in February–March (dry season) (see Appendix 3 for detailed methodology). I included experimental and

Research area and study system observational data of hummingbird foraging preferences from selected species occurring at two of the three study sites (the pre–montane forest and the lower montane wet forest).

Figure 5. Recording plant–hummingbird interactions with videocamaras (A). Collecting pollen loads from hummingbirds by using fuchsine–stained gelatine cubes to later identify the pollen in the lab and determine plant–hummingbird interactions (B,C,D).

Figure 6. Recording resource use by hummingbirds in two forest types in the Caribbean slope of Costa Rica. Field experiments with artificial feeders (differing in length and

Research area and study system curvature of flower types) used to quantify the hummingbirds' interaction niche under unlimited nectar resources (A). Visitation of hummingbirds to natural flowers used to quantify the interaction niche under real–world conditions of limited nectar resources (B).

4.2 Study system

4.2.1 Hummingbirds

Hummingbirds (Trochilidae) are nectar–feeding birds that occur only in the Western Hemisphere. Their range extends from southern Canada and Alaska in the north, to Tierra del Fuego in the south, with the greatest diversity within the latitudes of 10° North and 10° South of the equator (Johnsgard 1997). The family is composed of about 340 species of small and often brightly coloured birds and encompass a remarkable diversity of morphology, behaviour and ecology (Stiles and Skutch 1989). Hummingbirds are mostly nectarivores with approximately 90% of their diet coming from nectar (Schuchmann 1999), a dilute solution of different types of sugars. These sugars are easily digested and quickly converted into energy, supplying the hummingbird's carbohydrate requirements. Arthropods complement the hummingbirds' diet and provide them protein, minerals, vitamins and fats (Remsen et al. 1986).

Hummingbirds are unique among birds because of their small size, iridescent plumage, high–pitched vocalizations, rapid movement and exclusive flight abilities. The ability to hover is the most outstanding behavioural feature of hummingbirds and represents extremes of locomotor and metabolic capacity among the vertebrates (Altshuler et al. 2004). Increased heart rates and high wing beat frequencies are often attributed to the demands of hovering flight and make it one of the most expensive means of locomotion (Feinsinger et al. 1979, Altshuler et al. 2004, Stiles 2008).

Hummingbirds are classified into two distinct groups: hermits (Phaethorninae) and non–hermits (Trochilinae), which differ in their elevational distribution, foraging strategy, morphology and degree of specialization on floral resources (Fig. 7). Hermit hummingbirds mostly occur in wet lowland areas and are more frequent in the

Research area and study system understory of tropical forests (Stiles 1978, 1981, Cotton 1998). They feed along traplines following repeated foraging circuits among successive flowers or clumps (Snow and Snow 1972, Colwell 1973, Linhart 1973, Stiles 1975). Several hermit hummingbirds are long–distance and high–reward trapliners (Feinsinger 1983). Hermit hummingbirds mostly have long and/or curved bill and are generally specialized on monocot herbs (Fig. 7 A,B,C) (Kershaw 2006).

Figure 7. Hermit hummingbirds mostly have long and/or curved bill (A,B,C) whereas non–hermits exhibit a greater variety of bill morphology (D,E,F).

Non–hermit hummingbirds may be found along a wide range of elevations. They reach their highest diversity at mid elevations and are the only group presented at elevations higher than 2.000 m a.s.l. (Stiles 1981). Several non–hermit species reside principally in the canopy and are found in more open habitats in tropical and temperate areas (Bleiweiss 1998). Non–hermit hummingbirds present variable territoriality on flower patches, which has been shown to be strongly related to resource availability. Short–term changes in nectar availability may produce rapid changes in the territorial behavior of non–hermit species (Wiens 1992). It has been found that most of the species are only territorial during peak bloom of plant species when flowers are densely packed (Powers 1996). Non–hermit hummingbirds exhibit a greater variety of bill morphology and degrees of specialization than hermit 19

Research area and study system hummingbirds (Fig. 7 D,E,F) and visit a wide variety of flowering plant species (Snow and Snow 1972, Cotton 1998) (see Appendix 4).

4.2.2 Hummingbird–visited plants

Plant species visited by hummingbirds in neotropical forests include many herbaceous understory plants which are clonal and patchily distributed such as those from the genus Heliconia and Costus (Stiles 1979, Stiles 1981, Kay and Schemske 2003, Schleuning et al. 2008) (Fig. 8 A,B). Flowers of these plants often last a single day and secrete most of their nectar prior to opening at dawn to reduce damage by nectar robbers (Stiles 1975). Heliconia spp. are large monocot herbs that form a widespread feature of tropical rain forests (Berry and Kress 1991). Their inflorescences consist of several colorful bracts to attract hummingbird pollinators and produce dozens of flowers during a period of several months.

In Costa Rica, 37 Heliconia spp. have been recorded, most of them flowering from April to September. These understory plants produce limited nectar, which discourages territorial defense and encourages traplining (Proctor et al. 1996). Stiles (1975) found nine hummingbird–pollinated species of Heliconia occurring together at La Selva Biological Station, in the wet Caribbean lowlands of the country. In forest habitats, Heliconia clumps (clones) are typically small and include shade–tolerant species with prominent flowers (Stiles 1981). In contrast, in more open areas and forest gaps, the clumps tend to attain large size.

In the Neotropics, the genus Costus comprises ca. 60 species of understory terrestrial herbs (Maas 1972). In undisturbed forest, Costus spp. generally grow at extremely low density, sometimes with hundreds of meters between flowering individuals. Flowers are odorless and diurnal, and they produce relatively large quantities of nectar (Kay and Schemske 2003). Hummingbird pollination is also common among the Neotropical members of other plant families, having been recorded for families such as Bromeliaceae, Rubiaceae, Gesneriaceae, Ericaceae, Campanulaceae, among others (Fig. 8, Appendix 5).

Research area and study system

Figure 8. Hummingbird–visited flowers from the genus Heliconia (Heliconiaceae), Costus (Costaceae), Pitcairnia (Bromeliaceae), Psychotria (Rubiaceae) and Columnea (Gesneriaceae) (A–E, respectively). Hummingbirds also visit plant species with floral characteristics that are not related with the traditional ornithophilous pollination syndrome such as Neomirandea eximia (Asteraceae) (F).

Trees and epiphytes in the canopy and subcanopy often possess large displays of flowers that are visited by hummingbirds as well. Among epiphytes for example, there are many species of bromeliads and orchids that rely on hummingbirds for pollination (de Queiroz Piacentini and Varassin 2007, Siegel 2011). Most orchids that are hummingbird–pollinated are from high–elevation ecosystems in the Neotropic where insects are rare or unable to operate because of the low temperatures at these elevations (Siegel 2011).

Flowers of hummingbird–visited plants comprehend some characteristics that have been related to avian pollination (referred to as the ornithophilous pollination syndrome; Faegri and van der Pijl 1979). These characteristics include bright colors (especially red), abundant nectar, odorlessness, tubular corollas and nectary away from the stigma and anthers. However, hummingbirds often include variable percentages of plant species with diverse floral characteristics in their diet, which are not related with the traditional ornithophilous pollination syndrome (Dalsgaard et al. 2008, Rocca and Sazima 2010) (Fig. 8 F).

Results and Discussion

5. RESULTS AND DISCUSSION

5.1 Functional structure and specialization in plant–hummingbird networks along an elevational gradient

I mist–netted hummingbirds at three elevations and used the pollen carried by hummingbird individuals to construct plant–hummingbird networks at the individual and species level at each elevation. I quantified different metrics of functional community structure based on three functional traits measured in hummingbird individuals (bill length, bill curvature and body mass). I tested the effect of elevation on these functional metrics and specialization within the networks.

My results show consistent patterns in the functional structure of hummingbird assemblages and specialization of hummingbird species and individuals across elevations. Hummingbird assemblages varied from being functionally even and over– dispersed in the lower elevations to uneven and clustered in high–elevation environments. Accordingly, hummingbird species and individuals were more specialized at low and mid elevations than at the highest elevation.

5.1.1 Functional structure of hummingbird assemblages varies across elevation

Elevation can be considered as a surrogate of temperature, being lower in highlands and determining, harsh conditions for organisms. Flight at these conditions is particularly demanding in terms of energy costs, because of reductions in both air density and oxygen availability (Altshuler et al. 2004). Hummingbird assemblages are highly susceptible to experience environmental filtering because of metabolic and aerodynamic challenges faced by these birds at high elevations (Altshuler et al. 2004). The hypothesis of environmental filtering assumes that environmental conditions act as a filter allowing only a narrow range of species and functional roles to coexist in harsh, high–elevation environments (Keddy 1992, Mouchet et al. 2010). The studied hummingbird community at high elevation may be experiencing environmental filtering as trait values are aggregated in functional trait space resulting in an uneven

Results and Discussion and under–dispersed functional structure. The lowland communities were particularly diverse and thus interspecific competition among hummingbirds is likely to be intense.

In addition to environmental factors, competition is another factor that may influence community structure and tends to limit the functional similarity of co– occurring species (MacArthur and Levins 1967). A high intensity of interspecific competition is expected to increase the spread of species traits within a community, i.e. a high dispersion in the distribution of the traits within the community (Laliberté and Legendre 2010). Accordingly, I found a pattern of functional evenness and over– dispersion at low and mid elevations, suggesting that interspecific competition is particularly strong in these hummingbird assemblages.

5.1.2 Hummingbird specialization decreses with elevation

Less specialization of hummingbird species at the highest elevation might be associated to the intensity of intraspecific competition at this elevation. A low availability of floral resources in the highlands (Smith et al. 1995, Biesmeijer et al. 2006, Ornelas et al. 2007), where total hummingbird abundance was similar to the lowlands, suggests increased intraspecific competition. A reduced resource availability combined with an increased intraspecific competition could determine a scenario where it is unlikely that hummingbirds specialize on specific floral resources. More generalized individuals and species in the humminbird assemblage at the highest elevation in my study corresponds to an uneven and clustered functional structure of this assemblage.

High specialization at low elevations may contribute to releasing species from competitors in highly diverse communities, due to increased resource partitioning (MacArthur and Levins 1967, Mouchet et al. 2010). Competition for resources has been found to be strong in hummingbird assemblages (Brown and Bowers 1985), favouring the evolutionary specialization of avian morphologies. For example, hermit hummingbirds have evolved long and curved bills that match closely to the morphological traits of understory herbs such as those from the genus Heliconia (Heliconiaceae) and Costus (Costaceae) in netropical forests. My results support the idea that evolutionary specialization of specific morphological traits of pollinator

Results and Discussion species in the past may influence the pattern of ecological specialization that we can observe today. These findings are important because specialization in resource use might explain how species may coexist in highly diversified assemblages in tropical forests.

5.2 Influence of morphological traits on patterns of interactions in plant– hummingbird networks

I recorded visitation of hummingbirds to plant species over an entire year at three different elevations and constructed quantitative networks based on interaction frequencies. I assessed the importance of morphological traits in structuring plant– hummingbird mutualistic networks. More precisely, I tested the effects of avian morphological traits (bill length, bill curvature and body mass) and abundance on ecological specialization of hummingbird species. I also tested whether interaction strength in the networks was associated with the degree of trait matching between corresponding pairs of morphological traits in plant and hummingbird species and explored whether this was related to resource handling times by hummingbird species.

All three morphological traits of hummingbirds were positively associated with ecological specialization, especially bill curvature. I found strong and significant associations between interaction strength and the degree of trait matching in corresponding morphological traits of hummingbird and plant species. Moreover, the degree of trait matching, particularly between bill and corolla length, was associated with the handling time of nectar resources by hummingbirds. My results show that bill morphology structures tropical plant–hummingbird networks and patterns of interactions are closely associated with morphological matches between plant and bird species and the efficiency of hummingbirds' resource use.

5.2.1 Ecological specialization is associated with morphological traits in plant– hummingbird networks

Several non–exclusive mechanisms may explain patterns of interactions in plant– pollinator networks. Vázquez (2005) proposed a neutral explanation in which network

Results and Discussion patterns can be interpreted on the basis of species abundance and random interactions. Other authors, highlight the importance of morphological traits on patterns of plant–pollinator interactions or a combination of abundance and morphological traits (Santamaría and Rodríguez-Gironés 2007). My findings support the idea that avian morphological traits, especially bill traits, contribute to hummingbird specialization on specific plant species. Long–billed and curved–billed hummingbird species were particularly specialized, indicating that they deviated strongly from a random interaction pattern that would be driven by the abundances of floral resources at a given elevation.

Bill morphology in hummingbirds has been recognized as an important determinant of interactions in plant–hummingbird networks (Feinsinger 1976, Brown and Bowers 1985). Hummingbird species with strongly curved bills reach nectar from curved flowers that straight–billed species are not able to access or only access with greater difficulty. Hence, interspecific competition for curved–billed hummingbird species is likely to be reduced (Linhart 1973, Stiles 1981). For instance, the long, curved bills of most hermit species enable them to reach nectar from flowers that short and uncurved billed species are not as easily able to access. Correspondingly, my results indicate that bill morphology, in particular bill curvature, influence resource use and niche partitioning in hummingbird assemblages, which is likely to reduce competition for floral resources.

5.2.2 The role of trait matching in shaping plant–hummingbird interactions

Previous research has supported the occurrence of interactions resulting from the matching of the phenotypic traits of interacting species (Stang et al. 2007, 2009, Santamaría and Rodríguez-Gironés 2007). In my study, large hummingbird species preferred to feed on large flowers at mid and high elevations, which may be related to a high nectar production of these flowers (Ornelas et al. 2007, Rodríguez and Stiles 2005). This finding suggests that high energy requirements in the harsh environmental conditions at higher elevations may require large–bodied hummingbird species to specialize on floral resources with large nectar amounts. I also present evidence of a strong association between the interaction strength in the networks and a high degree 25

Results and Discussion of trait matching between bill traits and floral morphology. For example, long–billed and curve–billed hummingbird species preferred plant species with long and curved flowers, respectively.

My results reveal that hummingbirds spent more foraging time not only on flowers with high nectar volume but also on flowers that did not match well with their bill morphology. This is consistent with optimal foraging theory (MacArthur and Pianka 1966), which predicts that high trait matching should lead to an increased efficiency of resource use, reflected in shorter handling times (Temeles 1996). Hence, a reduced cost in resource handling makes short–billed hummingbirds more efficient feeders on flowers with short corollas. Consequently, a plant species that offers a greater profitability for pollinators will be visited more frequently. The relationship between resource handling efficiency and trait matching is a plausible explanation for the close association between interaction strength and trait matching found in all networks.

5.3 Combining experimental and observational approaches to understand patterns of plant–pollinator interactions

I performed field experiments with artificial feeders, differing in length and curvature of flower types, to quantify the hummingbirds' interaction niche under unlimited nectar resources. To quantify the interaction niche under real–world conditions of limited nectar resources, I measured foraging preferences of hummingbirds for plant species. I tested whether morphological floral traits were associated with foraging preferences of hummingbirds for artificial and natural flower types.

Artificial feeders were visited by Eupherusa nigriventris and Phaethornis guy in the pre–montane forest, and Lampornis calolaemus in the lower montane forest. Under experimental conditions, all three hummingbird species overlapped their interaction niches and showed a preference for the short artificial flower type over the long– straight and the long–curved flower types. Under natural conditions, the two co– occurring hummingbird species preferred to feed on plant species with floral traits corresponding to their bill morphology. The short–billed hummingbird E. nigriventris preferred to feed on short and straight flowers, whereas the long– and curved–billed

Results and Discussion

P. guy preferred long and curved natural flowers. In contrast to experimental conditions, L. calolaemus did not show any preferences for specific flower types under natural conditions.

5.3.1 Interaction niche of hummingbird species under experimental conditions

In plant–hummingbird interactions, morphological floral traits of plants may act as barriers allowing only certain hummingbirds access to nectar rewards (Temeles et al. 2009, Vizentin-Bugoni et al. 2014). In my study, foraging preferences for the short flower type under experimental conditions was more pronounced in the short–billed and the medium–size billed hummingbird species than in the long– and curved–billed species. This is very likely to result from morphological constraints of accessibility to nectar imposed by long–straight and long–curved flower types. My results indicate that length and curvature of corolla are important floral traits that limit the access to nectar resources and therefore determine the interaction niche of short–billed hummingbird species within the trait space that was experimentally explored.

Wiklund et al. (1979) proposed that elongate mouthparts in pollinators offer them the opportunity to exploit a greater diversity of floral morphologies, which has been supported by several studies (Feinsinger 1976, Corbet 2000, Goldblatt and Manning 2000). This may explain why the long and curved bill species P. guy was able to utilize all three artificial flower types in the field experiment. The long bill of the hermit species enables it to access nectar resources from the long and curved flower types unlike the shorter–billed species. Even though P. guy used all three artificial flower types, it preferred the short flower type, which may be the result of easier nectar intakes on short flowers. However, the preference for the short artificial flowers also resulted in an overlapping interaction niche with those of the two shorter–billed species. The unlimited nectar resources provided to hummingbirds in the experiment is likely to have contributed to niche overlap among the hummingbird species in the experiment.

Results and Discussion

5.3.2 Interaction niche of hummingbird species under natural conditions

Several studies have shown that nectar–feeding birds are specialized on plant species with flower traits that match their bill morphology and this has been associated with an increased efficiency (Wolf et al. 1976, Dalsgaard et al. 2009, Geerts and Pauw 2009, Maglianesi et al. 2014). A foraging preference by E. nigriventris for short flowers in the community may be the result of a better foraging performance at this floral morphology. The interaction niche of this species under natural conditions corresponded to the interaction niche under unlimited nectar resources at the feeders within the range of trait values that was experimentally investigated. These results suggest that foraging preferences of short–billed hummingbird species are likely driven by morphological constraints.

Foraging preferences of P. guy for long and curved flowers may result from factors other than morphological constraints because interaction niches differed considerably between experimental and natural conditions. For example, the long– and curved– billed hummingbird species may prefer the long and curved flowers because these flowers may offer higher nectar rewards per se (Geerts and Pauw 2009, Ornelas et al. 2007) or because they are less utilized by the short–billed species. The long and curved bills of the hermit species may hence contribute to reduce competition with the other species through resource exploitation. This explanation is consistent with the idea that competition for resources contributes to floral niche partitioning in hummingbird communities (Stiles 1981). Another reason why the hermit hummingbirds may prefer the long and curved flowers is because short flowers could be defended by short– billed hummingbird species and thus competitive interference may limit access to these flowers (Case and Gilpin 1974, Feinsinger 1976). My data do not enable to differentiate among the potential mechanisms causing niche segregation of hermit species relative to other hummingbird species under natural conditions. Nevertheless, the comparison of experimental and real–world data shows that resource use in natural pollinator communities is strikingly different from that under controlled experimental conditions with unlimited and equally rewarding resources.

Results and Discussion

L. calolaemus showed a great flexibility in foraging behaviour since this species did not preferred plant species with specific floral traits in natural plant communities. In the lower montane forest, L. calolaemus was the dominant species (>50% of the captured individuals belonged to the species), suggesting that intraspecific competition for floral resources might be intense and potentially was more important than interspecific competition. Some studies have suggested that high levels of intraspecific overlap in plant resource use may result in individuals expanding their interaction niches to all types of potential resources (Bolnick et al. 2003; Maruyama et al. 2013). A generalist strategy allows abundant pollinators to use a wide range of food resources. This may represent an advantage in environments with a great variation in floral resources.

6. CONCLUSIONS

In this thesis, I have attempted to improve our current understanding of community assembly patterns in plant–hummingbird mutualistic networks by using a novel combination of quantitative analyses and high–quality data sets. Integrating several methodological approaches such as functional trait–based analyses, network theory and observational as well as experimental approaches, allowed me to achieve a more comprehensive analysis of the processes and mechanisms determining patterns in plant–hummingbird networks. My findings show that environmental filtering appears to be an important process influencing the functional structure of hummingbird assemblages in harsh, high–elevation environments. In addition to environmental factors, the interplay between inter– and intraspecific competition is likely to have an important effect in determining patterns of specialization along a tropical elevational gradient. Floral niche expansion in the highlands and resource niche partitioning in highly diverse assemblages in the lowlands represent potential mechanisms explaining the functional structure and specialization in plant–hummingbird networks.

Conclusions

I provide evidence that plant–hummingbird networks are far from being randomly structured but they are organized on a mechanistic basis as a consequence of ecological and evolutionary processes. My results prove that morphological traits of interacting species, either in a complementary way (i.e. trait matching) or as a barrier (i.e. morphological constraints) are crucial in determining foraging preferences of hummingbird species and therefore in shaping plant–hummingbird networks. Other ecological processes, such as competition, additionally define interaction niches of hummingbird species in real–world conditions. By using multiple approaches, my research highlights the importance of a more integrated view of plant–humminbird networks to understand patterns in specialization and resource use.

This thesis provides a conceptual framework that could be applied to other types of mutualistic interactions such as plant–dispersal and, indeed, other types of plant– animal relationships as for example predator–prey interactions. Nevertheless, more work is clearly needed to identify and assess the strength of the different processes involved. Future research should aim to integrate an increasing number of factors and more complex relationships into the same quantitative framework to a more complete analysis of how mutualistic networks are structured. Measuring resource availability and competition among pollinators at the individual and species level may help to better understand patterns in resource partitioning and species co–existence in highly diversified communities. I believe this research offers a promising perspective to the comprehension of the major processes leading to patterns of interactions in plant– pollinator networks and calls for further investigations to identify other potential mechanisms influencing these networks.

Zusammenfassung

7. ZUSAMMENFASSUNG

7.1 Einleitung

Ökologische Artengemeinschaften sind in komplexen ökologischen Netzwerken organisiert, in denen die beteiligten Arten über unterschiedliche Interaktionstypen miteinander in Verbindung stehen. Mutualistische Pflanze-Tier-Interaktionen haben dabei einen besonders tiefgreifenden Einfluss auf die Dynamik und Diversität dieser Gemeinschaften, da sie sowohl die Reproduktion der Pflanzen als auch den Lebenszyklus der Tiere beeinflussen. Daher ist ein fundiertes Wissen über die Strukturierung von Pflanze-Bestäuber-Netzwerken für ein besseres Verständnis der Evolution, der Erhaltung von Biodiversität und der Auswirkungen von Artensterben notwendig. Interaktionsmuster in Bestäubernetzwerken stehen mit den Merkmalen der miteinander interagierenden Arten in Zusammenhang. In dieser Hinsicht hat sich die merkmalsbasierte Analyse zu einer nützlichen Methode entwickelt, um Veränderungen in Artengemeinschaften zu charakterisieren und die Koexistenz von Arten zu untersuchen. Trotzdem besteht nach wie vor eine Herausforderung darin, funktionelle Artmerkmale mit ökologischen Prozessen, wie der Bestäubung, in Zusammenhang zu bringen.

In drei tropischen Regenwäldern Costa Ricas untersuchte ich die Muster und treibenden Prozesse, die die Zusammensetzung der Artengemeinschaften und die Struktur der Pflanze-Kolibri-Netzwerke entlang eines Höhengradienten bestimmen. Ich konzentrierte meine Arbeit dabei auf drei Kernthemen: Erstens untersuchte ich entlang eines Höhengradienten, ob die funktionelle Struktur von Kolibri- Gemeinschaften und die Spezialisierung von Kolibris auf Individuen- bzw. Artebene variieren. Zweitens analysierte ich die Bedeutung morphologischer Merkmale für die Strukturierung von Pflanze-Kolibri-Netzwerken in neotropischen Wäldern, indem ich in einem neuartigen Ansatz quantitative Methoden kombinierte. Drittens untersuchte ich, ob Kolibri-Arten eine Präferenz für spezifische Blütentypen unter experimentellen und natürlichen Bedingungen zeigen, und ob sich Nahrungssuchpräferenzen zwischen nicht-limitiertem und natürlichem Nektarangebot unterscheiden.

Zusammenfassung

7.2 Untersuchungssystem

Die Feldarbeit wurde im La Selva-Braulio Carrillo Korridor entlang des karibischen Hangs der Cordillera Central im Nordosten Costa Ricas durchgeführt. Dieses Gebiet reicht von der biologischen Station La Selva (ca. 1.500 ha) bis zum Braulio Carrillo Nationalpark (ca. 45.000 ha) und beinhaltet einen der größten Höhengradienten von geschütztem Primärwald in Zentralamerika (Hartshorn and Peralta 1988, Pringle 1988). Ungefähr 75% der Vegetation besteht aus natürlichem Primärwald, der restliche Teil aus kleinräumigen Sekundärwaldflächen. Ich wählte drei Untersuchungsgebiete, die drei tropische Waldtypen auf verschiedenen Höhenstufen repräsentieren: Tieflandregenwald (50 m), Vorgebirgswald (1.000 m) und unterer Bergregenwald (2.000 m).

Die Studie wurde von Mai bis September 2011 und von Dezember bis April 2012 durchgeführt. In diesem Zeitraum sammelte ich Daten sowohl zu den Abundanzen und funktionellen Merkmalen von Blütenressourcen und Kolibris als auch zu Pflanze- Kolibri- Interaktionen. Ich nahm außerdem Stichproben der von den Kolibris mitgeführten Pollenladungen, um Pflanze-Kolibri-Interaktionen auf Individuen- und Artebene zu bestimmen. Die Daten wurden in sieben Sammelzeiträumen von jeweils maximal 10 Tagen gesammelt. Die Pollenladungen wurden für vier Sammelzeiträume in der Regen- und Trockenzeit analysiert. Zusätzlich führte ich Feldexperimente mit künstlichen Nektarspendern für zwei Sammelzeiträume in der Regen- und Trockenzeit durch.

Ich fing Kolibris mit Japannetzen auf allen drei Höhenstufen und nutzte die Pollenladungen, um Pflanze-Kolibri-Netzwerke auf Individuen- und Artebene für jede Höhenstufe zu konstruieren. Außerdem quantifizierte ich verschiedene Maße für die funktionelle Struktur der Artengemeinschaften basierend auf drei Kolibri-Merkmalen (Schnabellänge, Schnabelkrümmung, Körpergewicht). Ich untersuchte den Einfluss der Höhestufen auf diese Merkmale und die Spezialisierung der Netzwerke. Um die Interaktionsmuster der Netzwerke zu bestimmen, dokumentierte ich den Besuch von Kolibris an einzelnen Pflanzen und konstruierte daraus Netzwerke mit quantitativen Interaktionsfrequenzen. 32

Zusammenfassung

Außerdem analysierte ich die Bedeutung von morphologischen Merkmalen für die Struktur der Interaktionsnetzwerke. Im Detail untersuchte ich den Einfluss von drei morphologischen Kolibri-Merkmalen (Schnabellänge, Schnabelkrümmung und Körpergewicht) und der Abundanz auf die ökologische Spezialisierung der Kolibri- Arten. Außerdem analysierte ich, ob die Stärke der Interaktionen sowie die Handhabungszeit der Nektarressourcen durch die Kolibris mit dem Grad der Übereinstimmung zwischen Merkmalspaaren von Pflanzen- und Kolibri-Arten in Zusammenhang stehen.

Abschließend führte ich Feldexperimente mit künstlichen Nektarspendern durch, die sich in der Länge und Krümmung der dargebotenen, künstlichen Blüten unterschieden. Ziel war es die Interaktionsnische der Kolibris bei unlimitierter Nektarverfügbarkeit zu quantifizieren. Um die Interaktionsnische unter realen Bedingungen bei limitierter Nektarverfübarkeit zu analysieren, erfasste ich Nahrungssuchpräferenzen von Kolibris für einzelne Pflanzenarten in den natürlichen Artengemeinschaften. Ich testete, ob Nahrungssuchpräferenzen für bestimmte künstliche und natürliche Blütentypen mit morphologischen Blütenmerkmalen in Zusammenhang stehen und ob diese sich unter künstlichen und natürlichen Bedingungen unterscheiden.

7.3 Ergebnisse und Diskussion

7.3.1 Funktionelle Struktur und Spezialisierung der Pflanze-Kolibri-Netzwerke

Ich konnte konsistente Muster sowohl in der funktionellen Struktur von Kolibri- Gemeinschaften als auch in der Spezialisierung von Kolibri-Arten und -Individuen entlang des Höhengradienten feststellen. Kolibri-Gemeinschaften zeigten funktionell gleichmäßige und dispergierende Merkmalsverteilungen in den unteren Höhenstufen und eine ungleichmäßige und geklumpte Merkmalsstruktur in den höheren Lagen. Dementsprechend waren Kolibri-Arten und -Individuen spezialisierter in den unteren und mittleren Höhenlagen als in der höchsten Stufe. Diese Ergebnisse deuten auf einen hohen Grad an Ressourcenaufteilung der Blütenressourcen in den tieferen Lagen hin, wodurch die interspezifische Konkurrenz in den spezialisierten Netzwerken des

Zusammenfassung

Tieflands vermutlich reduziert wird. Im Gegensatz dazu scheinen die Artmerkmale der Gemeinschaft auf der höchsten Höhenstufe durch Umweltfilter beeinflusst zu werden, da sich die Merkmalswerte im funktionellen Raum aggregierten. Diese Ergebnisse stehen im Einklang mit dem Vorkommen generalisierter Kolibri-Arten auf der höchsten Höhenstufe, was auf eine geringe Verfügbarkeit von Blütenressourcen und kompetitive Prozesse zwischen Kolibris zurückzuführen sein könnte (Smith et al. 1995, Biesmeijer et al. 2006, Ornelas et al. 2007). Eine verringerte Ressourcenverfügbarkeit in Kombination mit einer erhöhten intraspezifischen Konkurrenz könnte Individuen dazu zwingen, ihre Interaktionsnische zu expandieren, wodurch die Netzwerke generalisierter werden.

7.3.2 Einfluss morphologischer Merkmale auf die Pflanze-Kolibri-Netzwerke

Alle drei morphologischen Merkmale der Kolibris, besonders die Schnabelkrümmung, waren positiv mit der ökologischen Spezialisierung der jeweiligen Kolibri-Art korreliert. Ich konnte starke und signifikante Zusammenhänge zwischen der Interaktionsstärke und dem Grad der Merkmalsübereinstimmung zwischen sich entsprechenden Merkmalspaaren bei Pflanzen- und Kolibri-Arten finden. Darüber hinaus bestand ein enger Zusammenhang zwischen dem Grad der Merkmalsübereinstimmung, insbesondere zwischen Schnabel- und Corollalänge, und der Handhabungszeit der Nektarressourcen durch die Kolibris. Meine Ergebnisse bestätigen die Ansicht, dass morphologische Merkmale von Bestäubern zu deren Spezialisierung auf bestimmte Pflanzenarten beitragen (Santamaría and Rodríguez-Gironés 2007). So können Kolibri- Arten mit stark gekrümmtem Schnabel Nektar aus gekrümmten Blüten aufnehmen, die von Kolibris mit nicht gekrümmten Schnäbeln nicht oder nur mit großen Schwierigkeiten erreicht werden können (Feinsinger 1976, Brown and Bowers 1985). Daher sollte sich die interspezifische Konkurrenz für krummschnäblige Kolibris reduzieren (Linhart 1973, Stiles 1981). Dementsprechend zeigen meine Ergebnisse, dass die Schnabelmorphologie, insbesondere die Schnabelkrümmung, einen Einfluss auf die Ressourcennutzung und die Nischendifferenzierung in Kolibri-Gemeinschaften hat. Ich kann somit zeigen, dass eine große Überstimmung zwischen sich entsprechenden morphologischen Merkmalen bei interagierenden Pflanzen- und Kolibri-Arten eine Stärkung der Interaktionen in den Netzwerken bewirkt. Diesem

Zusammenfassung

Muster könnte der Mechanismus zugrundeliegen, dass Kolibris Blütenressourcen umso effektiver nutzen können, je besser ihre Schnabelmorphologie mit der Morphologie ihrer Nahrungspflanzen übereinstimmt.

7.3.3 Kombination von Experimenten und Beobachtungen von Bestäuber- Interaktionen

Alle drei Kolibri-Arten, die künstliche Nektarspender besuchten, bevorzugten kurze Kunstblüten vor langen, geraden und langen, gekrümmten Kunstblüten. Morphologische Blütenmerkmale von Pflanzen können als Barrieren in Pflanze-Kolibri- Interaktionen fungieren, indem sie nur bestimmten Kolibris Zugang zum Nektar gewähren (Temeles et al. 2009, Vizentin-Bugoni et al. 2014). In meiner Studie war die Nahrungssuchpräferenz für kurze Blüten unter künstlichen Bedingungen bei kurzschnäbligen Kolibris und Kolibris mit mittlerer Schnabellänge stärker ausgeprägt als bei Arten mit langem und gekrümmtem Schnabel. Dies beruht mit großer Wahrscheinlichkeit auf morphologischen Barrieren, die durch die langen und/oder gekrümmten Blütentypen verursacht wurden.

Meine Daten zeigen, dass die zwei sympatrisch vorkommenden Kolibri-Arten ihre unter künstlichen Bedingungen überlappenden Interaktionsnischen unter realen Bedingungen in nicht-überlappende Interaktionsnischen aufspalten. Nahrungssuchpräferenzen der lang- und krummschnäbligen Arten für lange und gekrümmte Blüten könnten daher von Faktoren herrühren, die nicht mit morphologischen Barrieren in Verbindung stehen, da sich die Interaktionsnischen erheblich zwischen künstlichen und natürlichen Bedingungen unterschieden. Der lange und gekrümmte Schnabel der Kolibri-Art Phaethornis guy könnte zu einer reduzierten Konkurrenz mit den anderen Arten der Gemeinschaft beitragen. Diese Erklärung stimmt mit der Idee überein, dass Ressourcenkonkurrenz zu einer Differenzierung der Interaktionsnischen in Kolibri-Gemeinschaften führt (Stiles 1981). Dies deutet darauf hin, dass Bestäuber ihre Nahrungssuchpräferenzen auch als Reaktion auf Faktoren wie die Konkurrenz um Blütenressourcen verändern.

Zusammenfassung

7.4 Schlussfolgerungen

Meine Ergebnisse identifizieren verschiedene Prozesse, die den Mustern der funktionellen Struktur von Artengemeinschaften und den Interaktionen in Pflanze- Kolibri- Netzwerken zugrundeliegen. Kolibri-Gemeinschaften der unteren Höhenlagen, die durch eine gleichmäßige Merkmalsstruktur gekennzeichnet sind, weisen einen hohen Grad an Nischendifferenzierung auf, wodurch die Interaktionen hier besonders spezialisiert sind. Im Gegensatz dazu zeigt die ungleichmäßige und geklumpte funktionelle Struktur von Kolibri-Gemeinschaften in höheren Lagen die Bedeutung von Umweltfiltern und die ökologische Generalisierung in diesen Gemeinschaften. Meine Ergebnisse heben außerdem die wichtige Rolle von morphologischen Merkmalen für die Strukturierung von Pflanze-Kolibri-Netzwerken hervor. Die Kombination von Daten aus Experimenten und Beobachtungen deutet daraufhin, dass Bestäuber ihre Nahrungssuchpräferenzen auch als Reaktion auf die Konkurrenz um Blütenressourcen verändern. Ich schließe aus den Ergebnissen meiner Arbeit, dass merkmalsbasierte Analysen von quantitativen Netzwerken in Kombination mit experimentellen und beobachtenden Ansätzen wesentlich sind für ein umfassendes Verständnis der evolutionären und ökologischen Kausalitäten, die die Interaktionsmuster in Pflanze- Bestäuber-Netzwerken bestimmen. Diese Ansätze können auch in anderen Untersuchungssystemen und in Netzwerken anderer Interaktions¬typen einen wichtigen Beitrag für ein besseres mechanistisches Verständnis der Entstehung von Artengemeinschaften und Interaktionsnetzwerken leisten.

Zusammenfassung

8. ACKNOWLEDGEMENTS

The work presented here was made possible by the support of numerous people and institutions. First and foremost, I would like to express my sincere gratitude to Matthias Schleuning and Katrin Böhning-Gaese for their continuous support throughout my PhD thesis research. I particularly thank Matthias for his patience and dedication, for sharing his vast knowledge on statistic and ecological research and for motivating me to think in a more creative and original way about ecological worldview. I am very greateful to Katrin for her guidance, constructive criticism and trust throughout my study. I feel very fortunate to have been part of her working group in a very collaborative and friendship atmosphere. The many presentations, discussions and seminars helped me greatly to improve my background and grasp novel ideas on a wide range of topics in ecological research.

I am very grateful to Nico Blüthgen who kindly agreed to act as co–supervisor and supported me through scientific ideas and discussions. His deep knowledge on network approach and pollination systems allowed me to study and understand the different mechanisms underlying patterns of biotic interactions. In addition, I wish to express my sincere thanks everyone whose critical discussions contributed to improve my work, in particular Eike Lena Neuschultz, Mathias Dehling and Bob O’hara for advice and assistance during data analyses. I am gratefult to Martina Stang, Anton Pauw, Jeferson Vizentin-Bugoni and Gary Stiles for their insightful comments on my manuscripts. I warmly thank members of the Böhning-Gaese working's group for helpful discussions on my work, in particular Baptiste Schmid, Silvia Gallegos, Christian Hof and Diana Bowler. I also want to thank Tonka Stoyanova, Mathias Templin, Tanja Caprano and Jan Wenner for their willingness to always assist with a wide range of logistical support.

I am grateful to the field assistants and volunteers who contributed to data collection and botanists from the Instituto Nacional de Biodiversidad and La Selva Biological Station for support with plant identification. Financial support has been generously provided by the following organizations: Consejo Nacional para

Zusammenfassung

Investigaciones Científicas y Tecnológicas and Ministerio de Ciencia, Tecnología y Telecomunicaciones (Costa Rica), Universidad Estatal a Distancia, Organization for Tropical Studies, German Academic Exchange Service and Tropical Science Centre. Financial support for this study was also provided by the research–funding programme ‘‘LOEWE–Landes–Offensive zur Entwicklung Wissenschaftlich–ökonomischer Exzellenz’’ of Hesse’s Ministry of Higher Education, Research, and the Arts.

It has been challenging at times to pursue this PhD thesis research but thanks to the people and institutions mentioned above I got there in the end!

References

9. REFERENCES

Altshuler, D. L., R. Dudley, and J. A. McGuire. 2004. Resolution of a paradox: Hummingbird flight at high elevation does not come without a cost. Proceedings of the National Academy of Sciences USA. 101:17731–17736.

Anderson, S. H., D. Kelly, J. J. Ladley, S. Molloy, and J. Terry. 2011. Cascading effects of bird functional extinction reduce pollination and plant density. Science 331:1068– 1071.

Armbruster, W. S., C. B. Fenster, and M. R. Dudash. 2000. Pollination “principles” revisited: specialization, pollination syndromes, and the evolution of flowers. Det Norske Videnskaps–Akademi. I. Matematisk Naturvidenskapelige Klasse, Skrifter, Ny Serie 39:179–200.

Bascompte, J. and P. Jordano. 2007. Plant–animal mutualistic networks: the architecture of biodiversity. Annual Review of Ecology, Evolution, and Systematics, 38:567–593.

Bascompte, J. and P. Jordano. 2014. Mutualistic Networks. Monographs in Population Biology Series, N° 53. Princeton University Press, Princeton, USA.

Bastolla, U., M. Fortuna, A. Pascual-García, A. Ferrera, B. Luque, and J. Bascompte. 2009. The architecture of mutualistic networks minimizes competition and increases biodiversity Nature 458:1018-1020.

Bates, A. J., J. P. Sadler, A. J. Fairbrass, S. J. Falk, J. D. Hale, and T. J. Matthews. 2011. Changing bee and hoverfly pollinator assemblages along an urban–rural gradient. PLoS ONE 6 e23459 doi: 10.1371/journal.pone.0023459

Benadi, G., T. Hovestadt, H.-J. Poethke, and N. Blüthgen. 2013. Specialization and phenological synchrony of plant–pollinator interactions along an altitudinal gradient. Journal of Animal Ecology doi: 10.1111/1365–2656.12158

Berry, F. and W. J. Kress. 1991. Heliconia: an identification guide. Smithsonian Institution Press, Washington, D.C., USA.

Biesmeijer, J. C., S. P. Roberts, M. Reemer, R. Ohlemüller, M. Edwards, T. Peeters, A. P. Schaffers, S. G. Potts, R. Kleukers, C. D. Thomas, et al. 2006. Parallel declines in 39

References

pollinators and insect–pollinated plants in Britain and the Netherlands. Science 5785:351-354.

Blake, J. G. and B. A. Loiselle. 2000. Diversity of birds along an elevational gradient in the Cordillera Central, Costa Rica. The Auk 117:663–686.

Bleiweiss, R. 1998. Origin of hummingbird faunas. Biological Journal of the Linnean Society 65:77–97.

Blüthgen, N., F. Menzel, T. Hovestadt, D. Fiala, and N. Blüthgen. 2007. Specialization, constraints, and conflicting interests in mutualistic networks. Current Biology 17:341–346.

Bolnick, D. I., R. Svanbäck, J. A. Fordyce, L. H. Yang, J. M. Davis, C. D. Hulsey, and M. L. Forister. 2003. The ecology of individuals: incidence and implications of individual specialization. The American Naturalist 161:1–28.

Bronstein, J. L. 2008. Mutualism, diversity, and diversification. Smithsonian Botanical Symposium. Botany Profile. The plant press. Department of Botany and the U.S. National Herbarium.

Brosi, B. J. and H. M. Briggs. 2013. Single pollinator species losses reduce floral fidelity and plant reproductive function. Proceedings of the National Academy of Sciences USA 110:13044–13048.

Brown, J. H. and M. A. Bowers. 1985. Community organization in hummingbirds: relationships between morphology and ecology. The Auk 102:251–269.

Burkle, L. A., J. C. Marlin, and T. M. Knight. 2013. Plant–pollinator interactions over 120 years: loss of species, co-occurrence, and function. Science 339:1611–1615.

Case, T. J. and M. E. Gilpin. 1974. Interference competition and niche theory. Proceedings of the National Academy of Sciences USA 71:3073–3077.

Castellanos, M. C., P. Wilson, and J. D. Thomson. 2003. Pollen transfer by hummingbirds and bumblebees, and the divergence of pollination modes in Penstemon. Evolution 57:2742–2752.

References

Chapin, F. S., E. S. Zavaleta, V. T. Eviner, R. L.Naylor, P. M. Vitousek, H. L. Reynolds, D. U. Hooper, S. Lavorel, O. E. Sala, S. E. Hobbie, et al. 2000. Consequences of changing biodiversity. Nature 405:234–242.

Chittka, L. and J. D. Thomson. 1997. Sensori–motor learning and its relevance for task specialization in bumblebees. Behavioral Ecology and Sociobiology 41:385–398.

Campos, C. A. 2014. Mapa del transecto La Selva–Braulio Carrillo. Vicerrectoría de Investigación, Universidad Estatal a Distancia, San José, Costa Rica.

Colwell, R. K. 1973. Competition and coexistence in a simple tropical community. The American Naturalist 107:737–760.

Colwell, R. K., R. R. Dunn, N. C. Harris. 2012. Coextinction and persistence of dependent species in a changing world. Annual Review of Ecology and Systematics 43:183–203.

Corbet, S. A. 2000. Butterfly nectaring flowers: Butterfly morphology and flower form. Entomologia Experimentalis et Applicata 96:289–298.

Cornwell, W. K., D. W. Schwilk, and D. D. Ackerly. 2006. A trait–based test for habitat filtering: convex hull volume. Ecology 87:1465–1471.

Cotton, P. A. 1998. Coevolution in an Amazonian hummingbird–plant community. Ibis 140:639–646.

Crepet, W. L. 1983. The role of insect pollination in the evolution of angiosperms. Pages 29–50 in L. Real, editor. Pollination biology, Academic Press, New York, USA.

Dalsgaard, B., A. M. Martín González, J. M. Olesen, A. Timmermann, L. H. Andersen, and J. Ollerton. 2008. Pollination networks and functional specialization: a test using Lesser Antillean plant-hummingbird assemblages. Oikos 117:789–793.

Dalsgaard, B., A. M. Martín González, J. M. Olesen, J. Ollerton, A. Timmermann, A. Timmermann, L. H. Andersen, A. G. Tossas. 2009. Plant–hummingbird interactions in the West Indies: floral specialization gradients associated with environment and hummingbird size. Oecologia 159:757–766.

Dalsgaard, B., E. Magard, J. Fjeldsa, A. M. Martín González, C. Rahbek, J. M. Olesen, J. Ollerton, Ruben Alarcón, A. C. Araujo, P. A. Cotton, et al. 2011. Specialization in

References

plant–hummingbird networks is associated with species richness, contemporary precipitation and quaternary climate–change velocity. PLoS ONE 6 e25891. de Queiroz Piacentini, V. and I. G. Varassin 2007. Interaction network and the relationships between bromeliads and hummingbirds in an area of secondary Atlantic rain forest in southern Brazil. Journal of Tropical Ecology 23:663–671.

Devictor, V., J. Clavel, R. Julliard, S. Lavergne, D. Mouillot, W. Thuiller, P. Venail, S. Villéger, and N. Mouquet. 2010. Defining and measuring ecological specialization. Journal of Applied Ecology 47:15–25.

Dohzono, I., Y. Takami, and K. Suzuki. 2011. Is bumblebee foraging efficiency mediated by morphological correspondence to flowers? International Journal of Insect Science 3:1–10.

Dupont, Y. L. and J. M. Olesen. 2009. Ecological modules and roles of species in heathland plant–insect flower visitor networks. Journal of Animal Ecology 78:346– 353.

Dupont, Y. L., D. M. Hansen, and J. M. Olesen. 2003. Structure of a plant–flower visitor network in the high–altitude sub–alpine desert of Tenerife, Canary islands. Ecography 26:301–310.

Faegri, K. and L. van der Pijl. 1979. The principles of pollination ecology. Pergamon Press, Oxford.

Feinsinger, P. 1976. Organization of a tropical guild of nectarivorous birds. Ecological Monographs 46:257–291.

Feinsinger, P. 1983. Coevolution and pollination. Pages 282–310 in D. J. Futuyma and M. Slatkin, editors. Coevolution. Sinauer Associates, Sunderland, USA.

Feinsinger, P., R. K. Colwell, J. Terborgh, and S. B. Chaplinet. 1979. Elevation and the morphology, flight energetics, and foraging ecology of tropical hummingbirds. The American Naturalist 113:481–497.

Garibaldi, L. A., I. Steffan-dewenter, R. Winfree, M. A. Aizen, R. Bommarco, S. A. Cunningham, C. Kremen, L. G. Carvalheiro, L. D. Harder, O. Afik, et al. 2013. Wild

References

pollinators enhance fruit set of crops regardless of honey bee abundance. Science 339:1608–1611.

Geerts, S. and A. Pauw. 2009. Hyper–specialization for long–billed bird pollination in a guild of South African plants: the Malachite Sunbird pollination syndrome. South African Journal of Botany 75:699–706.

Gegear R. J. and J. D. Thomson. 2004. Does the flower constancy of bumble bees reflect foraging economics? Ethology 110:793–805.

Goldblatt, P. and J. C. Manning. 2000. The long–proboscid fly pollination system in southern Africa. Annals of the Missouri Botanical Garden 87:146–170.

Goulson D. 2003. Bumblebees: Behaviour and Ecology. Oxford University Press, Oxford, UK.

Graham, C. H., J. L. Parra, C. Rahbek, and J. A. McGuire. 2009. Phylogenetic structure in tropical hummingbird communities. Proceedings of the National Academy of Sciences USA 106:19673–19678.

Grimaldi, D. 1991. The co–radiations of pollinating insects and angiosperms in the Cretaceous. Annals of the Missouri Botanical Garden 86:373–406.

Hammel, B. E. 1990. The distribution of diversity among families, genera, and habit types in the La Selva flora. Pages 75–84 in A. H. Gentry, editor. Four neotropical rainforests. Yale University Press, New Haven, USA.

Hartshorn, G. and R. Peralta. 1988. Preliminary description of primary forests along the La Selva–Volcan Barva altitudinal transect, Costa Rica. Pages 281–295 in F. Almeda and C. M. Pringle, editors. Tropical rainforests: Diversity and conservation. California Academy of Sciences, San Francisco, USA.

Heinrich, B. and P. Raven. 1972. Energetics and pollination ecology. Science 176:597– 602.

Herrera, C. M. 2005. Plant generalization on pollinators: species property or local phenomenon? American Journal of Botany 92:13–20.

Holdridge, L. R. 1967. Life zone ecology. Tropical Science Center. San José, Costa Rica.

References

Hooper, D. U., E. C. Adair, B. J. Cardinale, J. E. K. Byrnes, B. A. Hungate, Kristin L. Matulich, A. Gonzalez, J. E. Duffy, L. Gamfeldt, and M. I. O’Connor. 2012. A global synthesis reveals biodiversity loss as a major driver of ecosystem change. Nature 486:105–108.

Ings, T. C., J. M. Montoya, J. Bascompte, N. Blüthgen, L. Brown, C. F. Dormann, F. Edwards, D. Figueroa, U. Jacob, J. I. Jones, et al. 2009. Ecological networks – beyond food webs. Journal of Animal Ecology 78:253–269.

Inouye, D. W. and G. H. Pyke 1988. Pollination biology in the Snowy Mountains of Australia: comparisons with montane Colorado, USA. Australian Journal of Ecology 13:191-205.

Johnsgard, P. A. 1997. The Hummingbirds of North America, 2nd edition. Smithsonian Institution Press, Washington D.C., USA.

Jordano, P., J. Bascompte, and J. M. Olesen. 2003. Invariant properties in coevolutionary networks of plant–animal interactions. Ecology Letters 6:69–81.

Junker, R. R., N. Blüthgen, T. Brehm, J. Binkenstein, J. Paulus, H. M. Schaefer, and M. Stang. 2013. Specialization on traits as basis for the niche–breadth of flower visitors and as structuring mechanism of ecological networks. Functional Ecology 27:329– 341.

Kaiser-Bunbury, C. N., D. P. Vazquez, M. Stang, and J. Ghazoul. 2014. Determinants of the microstructure of plant–pollinator networks. Ecology http://dx.doi.org/10.1890/14-0024.1

Kay, K. M. and D. W. Schemske. 2003. Pollinator assemblages and visitation rates for 11 species of neotropical Costus (Costaceae). Biotropica 35:198–207.

Kay, K. M. and. W. Schemske. 2004. Geographic patterns in plant-pollinator mutualistic networks: comment. Ecology 85:875–878.

Keddy, P. A. 1992. Assembly and response rules – 2 goals for predictive community ecology. Journal of Vegetation Science 3:157–164.

References

Kershaw, F. L. 2006. The co–evolution of hummingbird bill morphology and floral structures with regard to both native and non–native flower species. Leeds University, Leeds. UK.

Koh, L. P., R. R. Dunn, N. S. Sodhi, R. K. Colwell, H. C. Proctor, and V. S. Smith. 2004. Species coextinctions and the biodiversity crisis. Science 305:1632–1634.

Kosior A., W. Celary, P. Olejnikzak, J. Fijal, W. Krol, W. Solarz, and P. Plonka. 2007. The decline of the bumble bees and cuckoo bees (Hymenoptera: Apidae: Bombini) of western and central Europe. Oryx 41:79–88.

Körner, C. 2000. Why are there global gradients in species richness? Mountains might hold the answer. TREE 15: 513–514.

Laliberté, E. and P. Legendre. 2010. A distance–based framework for measuring functional diversity from multiple traits. Ecology 91:299–305.

Linhart, Y. B. 1973. Ecological and behavioral determinants of pollen dispersal in hummingbird–pollinated Heliconia. The American Naturalist 107:511–523.

Maas, P. J. M. 1972. Costoideae (Zingiberaceae). Flora Neotropica 8:1–140.

MacArthur, R. H. and E. R. Pianka. 1966. On the optimal use of a patchy environment. The American Naturalist 100:603–609.

MacArthur, R. H. and R. Levins. 1967. Limiting similarity convergence and divergence of coexisting species. The American Naturalist 101:377–385.

Maglianesi, M. A., N. Blüthgen, K. Böhning-Gaese, and M. Schleuning. 2014. Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics. Ecology, http://dx.doi.org/10.1890/13- 2261.1

Maruyama, P. K., G. M. Oliveira, C. Ferreira, B. Dalsgaard, and P. E. Oliveira. 2013. Pollination syndromes ignored: importance of non–ornithophilous flowers to Neotropical savanna hummingbirds. Naturwissenschaften 100:1061–1068.

McGill B. J., B. J. Enquist, E. Weiher, and M. Westoby. 2006. Rebuilding community ecology from functional traits. Trends in Ecology and Evolution 21:178–185. doi:10.1016/j.tree.2006.02.002

References

Memmott, J., N. M. Waser, and M. V. Price. 2004. Tolerance of pollination networks to species extinctions. Proceedings of the Royal Society B 271:2605–2611.

Mouchet, M. A., S. Villéger, N. W. H. Mason, and D. Mouillot. 2010. Functional diversity measures: an overview of their redundancy and their ability to discriminate community assembly rules. Functional Ecology 24:867–876.

Olesen, J. M. 2000. Exactly how generalised are pollinationinteractions? Pages 161– 178 in Ø. Totland, W. S. Armbruster, C. Fenster, U. Molau, L. A. Nilsson, J. M. Olesen, J. Ollerton, M. Philipp, and J. Ågren, editors. Scandinavian Association for Pollination Ecology honours Knut Fægri. Norwegian Academy of Science and Letters 39, Oslo, Norway.

Olesen, J. M. and P. Jordano. 2002. Geographic patterns in plant–pollinator mutualistic networks. Ecology 83:2416–2424.

Ollerton, J. and S. Watts. 2000. Phenotype space and floral typology: towards an objective assessment of pollination syndromes. Pages 149–159 in Ø. Totland, W. S. Armbruster, C. Fenster, U. Molau, L. A. Nilsson, J. M. Olesen, J. Ollerton, M. Philipp, and J. Ågren, editors. Scandinavian Association for Pollination Ecology honours Knut Fægri. Norwegian Academy of Science and Letters 39, Oslo, Norway.

Ollerton, J., R. Winfree, and S. Tarrant. 2011. How many flowering plants are pollinated by animals? Oikos 120:321–326.

Ornelas, J. F., M. Ordeno, A. J. De-Nova, M. E. Quintero, and T. Garland Jr. 2007. Phylogenetic analysis of interspecific variation in nectar of hummingbird–visited plants. Journal of Evolutionary Biology 20:1904–1917.

Pauw, A. 2013. Can pollination niches facilitate plant coexistence? Trends in Ecology and Evolution 28:30–37.

Pereira H. M., P. W. Leadley, V. Proença, R. Alkemade, J. P. W. Scharlemann, J. F. Fernandez-Manjarrés, M. B. Araújo. P. Balvanera, R. Biggs, and W. W. L. Cheung. 2010. Scenarios for global biodiversity in the 21st century. Science 330:1496–1501.

Petanidou, T. and E. Lamborn. 2005. A land for ﬂowers and bees: studying pollination ecology in Mediterranean communities. Plant Biosystems 139:279–294.

References

Pires, M. M., P. R. Guimarães, Jr., M. S. Araújo, A. A. Giaretta, J. C. Costa, and S. F. dos Reis. 2011. The nested assembly of individual-resource networks. Journal of Animal Ecology 80:896–903.

Poisot, T., E. Canard, N. Mouquet, and M. E. Hochberg. 2012. A comparative study of ecological specialization estimators. Methods in Ecology and Evolution doi: 10.1111/j.2041-210X.2011.00174.x

Powers, D. R. 1996. Magnificent Hummingbird (Eugenes fulgens). in A. Poole and F. Gill, eds. The Birds of North America, No. 221. The Academy of Natural Sciences, Philadelphia, PA, and The American Ornithologists’ Union, Washington, D.C., USA.

Pringle, C. M. 1988. History of conservation efforts and initial exploration of the lower extension of Parque Nacional Braulio Carrillo, Costa Rica. Pages 131–142 in F. Almeda and C. M. Pringle, editors. Tropical rainforests. California Academy of Sciences and AAAS Pacific Division, San Francisco, USA.

Proctor, M., P. Yeo, and A. Lack. 1996. The natural history of pollination. Timber Press, Portland, USA.

Ramos-Jiliberto, R., D. Domínguez, C. Espinoza, G. López, F. S. Valdovinos, R. O. Bustamante, and R. Medel. 2010. Topological change of Andean plant–pollinator networks along an altitudinal gradient. Ecological Complexity 7:86–90.

Remsen Jr., J. V., F. G. Stiles, and P. E. Scott. 1986. Frequency of arthropods in stomachs of tropical hummingbirds. Auk 103:436–441.

Rocca, M. A. and Sazima, M. 2010. Beyond hummingbird-flowers: the other side of ornithophily in the neotropics. Oecologia Australis 14:67–99.

Rodríguez C. I. and F. G. Stiles. 2005. Análisis ecomorfológico de una comunidad de colibríes ermitaños (Trochilidae, Phaethorninae) y sus flores en la amazonia colombiana. Ornitología Colombiana 3:7–27.

Sanders, N. J. and C. Rahbek. 2012. The patterns and causes of elevational diversity gradients. Ecography 35:1–3.

Santamaría, L. and M. A. Rodríguez-Gironés. 2007. Linkage rules for plant–pollinator networks: trait complementarity or exploitation barriers?. Plos Biology 5:354–362.

References

Schemske, D. W. and H. D. Bradshaw. 1999. Pollinator preference and the evolution of floral traits in monkeyflowers (Mimulus). Proceedings of the National Academy of Sciences, USA 96: 11 910–11 915.

Schleuning, M., V. Huamán, and D. Matthies. 2008. Flooding and canopy dynamics shape the demography of a clonal Amazon understorey herb. Journal of Ecology 96:1045–1055.

Schleuning, M., J. Fründ, A-M. Klein, R. A. Abrahamczyk, M. Albrecht, R. Alarcón, M. Albrecht, G. K. S. Andersson, S. Bazarian, Katrin Böhning-Gaese, et al. 2012. Specialization of mutualistic interaction networks decreases toward tropical latitudes. Current Biology 22:1–7.

Schuchmann, K. L. 1999. Family Trochilidae (Hummingbirds). Pages 468–680 in J. del Hoyo, A. Elliott, and J. Sargatal, editors. Handbook of Birds of the World, Volume 5: Barn–owls to Hummingbirds. Lynx Edicions, Barcelona, Spain.

Siegel, C. 2011. Orchids and hummingbirds: sex in the fast lane. The Orchid Digest 75:8–17.

Smith, T. B., L. A. Freed, J. A. Lepson, and J. H. Carothers. 1995. Evolutionary consequences of extinctions in populations of a Hawaiian Honeycreeper. Conservation Biology 9:107–113.

Snow, B. K. and D. W. Snow. 1972. Feeding niches of hummingbirds in a Trinidad valley. Journal of Animal Ecology 41:471–485.

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

Stang, M., P. Klinkhamer, N. M. Waser, I. Stang, and E. van der Meijden. 2009. Size– specific interaction patterns and size matching in a plant–pollinator interaction web. Annals of Botany 103:1459–1469.

Stiles, F. G. 1975. Ecology, flowering phenology, and hummingbird pollination of some Costa Rican Heliconia species. Ecology 56:285–301.

References

Stiles, F. G. 1978. Ecological and evolutionary implications of bird pollination. American Zoologist 18:715–727.

Stiles, G. F. 1979. Notes on the natural history of Heliconia (Musaceae) in Costa Rica. Brenesia 15:151-180.

Stiles, F. G. 1981. Geographical aspects of bird–flower coevolution, with particular reference to Central America. Annals of the Missouri Botanical Garden 68:323–351.

Stiles, F. G. 2008. Ecomorphology and phylogeny of hummingbirds: divergence and convergence in adaptations to high elevations. Ornitología Neotropical 19:511–519.

Stiles, F. G., and A. Skutch. 1989. A guide to the birds of Costa Rica. Comstock Publishing Associates. A division of Cornell University Press. Ithaca, USA.

Strickler K. 1979. Specialization and foraging efficiency of solitary bees. Ecology 60:998–1009.

TEAM 2013. Tropical Ecology Assessment and Monitoring Network. Volcán Barba, Heredia, Costa Rica.

Temeles, E. J. 1996. A new dimension to hummingbird–flower relationships. Oecologia 105:517–523.

Temeles, E. J., C. R. Koulouris, S. E. Sander, and W. J. Kress. 2009. Effect of flower shape and size on foraging performance and trade–offs in a tropical hummingbird. Ecology 90:1147–1161.

Temeles, E. J., J. S. Miller, and J. L. Rifkin. 2010. Evolution of sexual dimorphism in bill size and shape of hermit hummingbirds (Phaethornithinae): a role for ecological causation. Philosophical Transactions of the Royal Society B 365:1053–1063.

Terborgh, J., N. Pitman, M. R. Silman, H. Schichter, and V. P. Núñez. 2002. Maintenance of tree diversity in tropical forests. Pages 1–18 in D. J. Levey, W. R. Silva, and M. Galetti, editors. Seed Dispersal and Frugivory: Ecology, Evolution and Conservation CAB International, Wallingford, UK.

Thébault, E. and C. Fontaine. 2010. Stability of ecological communities and the architecture of mutualistic and trophic networks. Science 329:853–856.

References

Thompson, J. N. 1994. The coevolutionary process. University of Chicago Press, Chicago, USA.

Thorp R. W. and M. D. Shepherd. 2005. Profile: subgenus Bombus. Pages 1–5 in M. D. Shepherd, D. M. Vaughan, and S. H. Black, editors. Red List of pollinator insects of North America. Xerces Society for Invertebrate Conservation, Portland, USA.

Tur, C., B. Vigalondo, K. Trøjelsgaard, J. M. Olesen, and A. Traveset. 2013. Downscaling pollen–transport networks to the level of individuals. Journal of animal Ecology doi: 10.1111/1365–2656.12130.

Tylianakis, R. K., J. Didham, D. A. Bascompte, D. A. Wardle. 2008a. Global change and species interactions in terrestrial ecosystems. Ecology letters 11:1351-1363.

Tylianakis J. M., T. A. Rand, A. Kahmen, A. M. Klein, N. Buchmann, J. Perner, and T. Tscharntke. 2008b. Resource heterogeneity moderates the biodiversity–function relationship in real world ecosystems. PLoS Biology 6 e122. doi: 10.1371/journal.pbio.0060122

Vanbergen. A. J. 2013. Frontiers in Ecology and the Environment. Threats to an ecosystem service: pressures on pollinators. Front Ecol Environ doi:10.1890/120126

Van der Niet T., R. Peakall, and S. D. Johnson. 2014. Pollinator–driven ecological speciation in plants: new evidence and future perspectives. Annals of Botany 113:199-211.

Vázquez, D. P. 2005. Degree distribution in plant–animal mutualistic networks: Forbidden links or random interactions. Oikos 108:21–426.

Vázquez, D. P. and M. A. Aizen 2004. Asymmetric specialization: a pervasive feature of plant–pollinator interactions. Ecology 85:1251–1257.

Vizentin-Bugoni, J., P. K. Maruyama, and M. Sazima. 2014. Processes entangling interactions in communities: forbidden links are more important than abundance in a hummingbird–plant network. Proceedings of the Royal Society B 281:1–8.

Waser, N. M. 1978. Interspecific pollen transfer and competition between co– occurring plant species. Oecologia 36:223–36.

References

Waser, N. M., L. Chittka, M. V. Price, N. M. Williams, and J. Ollerton. 1996. Generalization in pollination systems, and why it matters. Ecology 77:1043–1060.

Wiklund, C., T. Eriksson, and H. Lundberg. 1979. The wood white butterfly, Leptidea sinapsis, and its nectar plants: A case of mutualism or parasitism? Oikos 33:358– 362.

Wilson, P. 1995. Selection for pollination success and the mechanical fit of Impatiens flowers around bumblebee bodies. Biological Journal of the Linnean Society 55:355– 383.

Wells, K. and R. B. O’Hara. 2012. Species interactions: estimating per-individual interaction strength and covariates before simplifying data into per–species ecological networks. Methods in Ecology and Evolution 4:1–8.

Wiens, J. A. 1992. The Ecology of Bird Communities, Volumen 2. Processes and Variations. Cambridge University Press, Cambridge, UK.

Wolf, L. L., F. G. Stiles, and F. R. Hainsworth. 1976. Ecological organization of a tropical, highland hummingbird community. Journal of Animal Ecology 45:349–379.

Appendices

10. APPENDICES

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant-pollinator interactions

Appendix 4: List of humminbird species recorded at the three study sites in the Caribbean slope of Costa Rica.

Appendix 5: List of plant species recorded at the three study sites in the Caribbean slope of Costa Rica.

Appendix 6: Curriculum vitae

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

Appendix 1: Functional structure and specialization in tropical plant– hummingbird interaction networks across elevations

Status: rejected by Ecography with an invitation for resubmission (July 31, 2014).

Maglianesi, M. A., N. Blüthgen, Katrin Böhning-Gaese and M. Schleuning. Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations.

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

Erklärung über Anteile der Autoren/Autorinnen

(1) Entwicklung und Planung

MM 30%; NB 20%; KBG 20%; MS 30%

(2) Durchführung der einzelnen Untersuchungen / Experimente

MM compiled all data and coordinated the fieldwork.

(3) Erstellung der Datensammlung und Abbildung

MM conducted data analyses and created all figures; figures were discussed with NB, KB and MS.

(4) Analyse/Interpretation der Daten

MM 70%; NB 5%; KBG 5%; MS 20%

(5) Übergeordnete Einleitung / Ergebnisse / Diskussion

MM 70%; NB 5%; KBG 5%; MS20%

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

María Alejandra Maglianesi1,2*, Nico Blüthgen3, Katrin Böhning–Gaese1,4 and Matthias Schleuning1

1 Biodiversity and Climate Research Centre (BiK–F) and Senckenberg Gesellschaft für Naturforschung, 60325 Frankfurt am Main, Germany

2 Vicerrectoría de Investigación. Universidad Estatal a Distancia (UNED), San José, Costa Rica

3 Ecological Networks, Department of Biology, Technical University of Darmstadt, 64287 Darmstadt, Germany

4 Department of Biological Sciences, Johann Wolfgang Goethe University Frankfurt, 60438 Frankfurt am Main, Germany

Running title: Functional structure and specialization in plant–hummingbird interactions

Type of manuscript: Article

*Corresponding author: María Alejandra Maglianesi. Vicerrectoría de Investigación. Universidad Estatal a Distancia (UNED), San Pedro Montes de Oca, San José, Costa Rica. E-mail: [email protected]

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

Abstract

Understanding causes of variation in multispecies assemblages along spatial environmental gradients is a long–standing research topic in ecology and biogeography. Ecological networks comprising interacting species of plants and pollinators are particularly suitable for testing effects of environmental gradients on the functional structure and specialization in multispecies assemblages. In this study, we investigated patterns in functional assemblage structure and specialization of hummingbirds at the individual and species level along a tropical elevational gradient. We mist–netted hummingbirds at three elevations in Costa Rica and used the pollen carried by hummingbird individuals to construct plant–hummingbird networks at the individual and species level at each elevation. We measured four functional traits of hummingbird species (bill length, bill curvature, wing length and body mass) and quantified different metrics of functional community structure. We tested the effect of elevation on functional metrics of hummingbird assemblages and specialization within the networks. Hummingbird species and individuals were more specialized at low and mid elevations than at the highest elevation. This pattern corresponded to a more even and over–dispersed assemblage structure at the lower elevations and suggests a high level of floral resource partitioning in functionally diversified communities. In contrast, an uneven and clustered functional structure of the highland assemblage suggests that this assemblage was structured by environmental filtering and by niche expansion of hummingbird individuals and species at this elevation. We conclude that high degrees of specialization on specific floral resources might be crucial for the coexistence of hummingbird species in diversified lowland communities. Spatial variation in animal resource use and specialization may be a crucial driver of the functional structure of diversified species assemblages also in other types of ecological networks and multispecies assemblages.

Key words: elevational gradient, biotic interactions, community structure, Costa Rica, elevation, functional diversity, hummingbirds, mutualistic networks, pollen loads, pollination, pollinators, specialization, traits, tropical forest.

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

Introduction

A primary aim of community ecology is to identify the processes that govern species assemblages across environmental gradients (McGill et al. 2006). Mountain ecosystems provide pronounced environmental gradients across relatively small spatial scales and have proved to be suitable systems to investigate patterns and determinants of species diversity and community structure (Körner 2000, Sanders and Rahbek 2012). Diversity patterns across elevational gradients follow distinct forms and in particular monotonously decreasing or hump–shaped distributions have been reported (McCain 2009). In pollination systems, pollinator richness and abundance have been found to decrease with increasing elevation (Totland 2001). Although some studies have investigated the effects of elevational gradients on structural properties of plant–pollinator networks such as nestedness or specialization (e.g. Olesen and Jordano 2002, Ramos-Jiliberto et al. 2010, Benadi et al. 2013), such studies are still scarce, especially in the tropics. Here, we studied the effects of elevation on the functional structure of tropical hummingbird assemblages and specialization in plant– hummingbird interaction networks.

The structure of ecological communities can be described by species’ functional traits, defined as measurable properties of individuals that influence their performance (McGill et al. 2006). Community trait composition can principally be examined by two distinct components: mean trait values of species, weighted by their relative abundances (i.e. community–weighted means of trait values), and multivariate measures of functional diversity (Dias et al. 2013). Functional diversity describes the range, distribution and abundance of trait values of species in a community (Tilman et al. 1997, Díaz and Cabido 2001) and may act as an indicator of the processes influencing species coexistence, such as environmental filtering and competition (Cornwell et al. 2006). Environmental filtering tends to increase the functional similarity among species by reducing the range of trait values within a community (i.e. functional clustering) (Keddy 1992). It appears to be an important mechanism that constrains particular functional roles within hummingbird assemblages at high elevations (Graham et al. 2009). In contrast, competition and resource partitioning tends to limit the functional similarity between co–existing species (i.e. functional 57

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

over–dispersion) (MacArthur and Levins 1967, Mouchet et al. 2010), for example manifested by the partitioning of nectar plants among hummingbird species. In the presence of high degrees of interspecific competition, pollinators are expected to become more specialized by increasing resource partitioning (Schoener 1974, Inouye 1978). Thus, structural properties of mutualistic networks, such as specialization, may be related to patterns in the functional structure of species assemblages (Plein et al. 2013).

A comprehensive understanding of the processes leading to community assembly in plant–pollinator networks requires an understanding of the processes occurring at the level of both species and individuals. As stated in the niche variation hypothesis, individuals within populations may differ substantially in their resource use (Van Valen 1965). Large between–individual variation in resource use may contribute to niche expansion (Van Valen 1965), leading to reduced specialization of individuals (Bolnick et al. 2003). An important mechanism related to specialization is competition (Schoener 1974), both among individuals of the same species and among individuals of different species (MacArthur and Levins 1964, Araújo et al. 2008). Individual–based networks may be a powerful tool to assess competition and have been used to study intra– population patterns of resource partitioning in vertebrates (Pires et al. 2011) and changes in foraging preferences at different levels of intraspecific competition (Araújo et al. 2008). However, to date only a few empirical studies have applied network analysis to explore patterns of ecological interactions between plants and their pollinators at the individual level (e.g. Tur et al. 2013). To our knowledge, elevational trends in specialization of plant–pollinator networks have so far not been analysed at the individual level.

In the Western Hemisphere, hummingbirds (Trochilidae) are considered to be effective pollinators (Castellanos et al. 2003). They are classified into two sub–families: Phaethorninae (hermits) and Trochilinae (non–hermits), which differ mainly in their elevational distribution and their level of specialization on floral resources. Hermit hummingbirds mostly occur in wet lowland forests and are specialized on specific floral resources (Snow and Snow 1972). Non–hermit hummingbirds may be found along a wide range of elevations and are in general less specialized than hermits (Feinsinger 58

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

and Colwell 1978). Hummingbird assemblages comprise species that differ in their morphology and functional roles (e.g. co–occurring hermit and non–hermit species) as well as in their degree of specialization. These assemblages are thus a suitable system to study functional community structure and floral resource partitioning in plant– pollinator networks. We mist–netted hummingbirds at three elevations in Costa Rica and used the pollen carried by hummingbird individuals to construct hummingbird– plant interaction networks at the individual and species level.

Here, we investigated whether patterns of specialization of hummingbirds at the individual and species level varied across elevations and explored whether these patterns corresponded to patterns in the functional structure of hummingbird assemblages. More specifically, we addressed the following questions and hypotheses: (a) Does the functional structure of hummingbird assemblages change with elevation? We assumed that the hummingbird assemblage at the highest elevation had a clustered functional structure because only a narrow spectrum of species and functional roles occur and coexist in harsh environments. (b) Do individual–level and species–level specialization change across elevations? We hypothesized that hummingbird species and individuals are specialized at low elevations because high levels of competition might cause pronounced resource partitioning in highly diversified lowland assemblages.

Material and methods

Study area and data collection

The study was conducted at La Selva Biological Station located in the lowlands of northeastern Costa Rica (10°26’N, 84°01’) and adjacent Braulio Carrillo National Park. This study region constitutes a corridor of continuous forest from sea level at La Selva Biological Station (LS) to elevations higher than 2.900 m a.s.l. at the Braulio Carrillo National Park. Our study sites were located at three different elevations: low (50 m) in LS, mid (1.000 m) and high elevation (2.000 m) in the national park. According to Holdridge’s (1967) life–zone classification, forests represented at those three elevations are tropical wet, pre–montane and lower montane wet forest. All sites were

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

located in old–growth forest. Canopy heights were approximately 35 m at LS, 30 m at 1.000 m, and 20 m at 2.000 m (Hartshorn and Peralta 1988). Mean annual temperature ranges from 25ºC at LS to 14ºC in the highlands, while mean annual precipitation ranges from 4.300 in the lowlands to 2.200 in the highlands (Blake and Loiselle 2000). The dry season lasts from January to April and the wet season reaches a rainfall peak during August.

We conducted the study from May 2011 to April 2012, covering an entire study year. During the year of sampling, we collected data on abundances and functional traits of hummingbirds. We also sampled pollen loads carried by hummingbirds to determine plant–hummingbird interactions. Data were collected during seven sampling periods of approximately 10 days each. Pollen loads were analysed for a subset of four periods covering both the wet and the dry season, that is, we identified pollen samples from plant species that bloomed at different times of the year.

Hummingbird abundances and traits

We placed at each elevation 12–14 standard mist nets (12 x 3 m) for approximately 6 h after dawn (Ralph et al. 1993). All hummingbirds captured were identified to species level (according to Stiles and Skutch 1989) and banded with numbered aluminium bands allowing for individual identification. The total number of hummingbird individuals captured during a sampling period was used as an estimate of species’ abundance at each elevation and sampling period; recaptured individuals within the same sampling period were discarded. We measured functional traits of hummingbird species that have been reported to affect their interactions with plant species, such as bill length and curvature (Hainsworth and Wolf 1972, Temeles et al. 2010), as well as body mass (Dalsgaard et al. 2009) and wing length (Stiles 2004). For each captured hummingbird individual, we measured bill length (exposed culmen) and length of the closed (folded) wing to the nearest 0.1 mm using dial callipers. To measure bill curvature, we placed the bill on graph paper so that the angle of deflection could be calculated using simple trigonometry (see Kershaw 2006). We applied arcussinus–sqrt–transformation to bill curvature prior to analysis. Body mass

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

was measured to the nearest 0.1 g with a digital scale. Morphological traits were measured from 773 individuals belonging to 28 hummingbird species.Wing length was excluded for further analysis due to its strong collinearity with body mas (r = 0.88, p < 0.001).

Pollen analysis

Plant–pollinator interactions were quantified by examining pollen loads carried by mist–netted hummingbirds. We used fuchsine–stained gelatine (Beattie 1971) to gently remove pollen loads from the bill, head and throat plumage of each bird (Kearns and Inouye 1993). We used one small cube of gelatine to wipe dorsally from the bill tip to the nape of the neck and a second cube to sample the underside of the bill and the entire throat. Each collected pollen sample was placed in a separate vial for later analysis. We were careful to ensure that the same amount of time and effort was used for each bird, as well as similar amounts of gelatine. In addition, we sealed the Petri dishes containing the gelatine to avoid contamination and we manipulated hummingbirds carefully to prevent pollen transfer among individuals.

The pollen identification process was done by two persons skilled in palynological techniques. We mounted each cube of gelatine on a slide, melted the sample and covered it with glass coverslips to produce a single layer of stained pollen grains. These were observed under a light microscope at 400–1000X magnification. We identified pollen grains by comparison with reference collections, taken from plants at the study sites, as well as the literature (e.g. Roubik and Moreno 1991). Voucher specimens of plant species were deposited in the Instituto Nacional de Biodiversidad and the pollen reference collection in the Universidad Estatal a Distancia (Costa Rica).

Overall, we analysed pollen samples from 21 species of the total hummingbird species captured in mist–nets. Pollen grains were identified to plant species level whenever possible, and to morphospecies when pollen from closely related species or genera were indistinguishable (Feinsinger et al. 1987). In cases where species or genera could not be determined, we classified pollen grains in morphotypes, based on their size, shape, type and number of apertures, and exine sculptures. Hereafter, we will refer to morphotype of either pollen identified to species level, or pollen identified

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

to morphospecies or morphotype level. The number of hummingbird individuals carrying pollen of a particular morphotype was used as a measure of the interaction frequency between that hummingbird species and a plant morphotype. Recaptured hummingbird individuals carrying the same pollen morphotype more than once where counted as a single individual.

Statistical analyses

Taxonomic and functional structure of hummingbird assemblages

To describe the taxonomic diversity of hummingbird assemblages, we estimated the abundance per hummingbird species and the species richness of the assemblage as the number of captured species for each sampling period and elevation. In addition, we calculated the exponent of Shannon entropy (eH), where H is the mean (one– dimensional) Shannon diversity of hummingbird assemblages for each sampling period and elevation. This index describes the effective number of species in a community (Jost 2006). To compare hummingbird abundance, species richness and eH across elevations, we fitted linear mixed effect models with elevation as fixed effect and sampling period as random effect to account for repeated measures at the same elevation. Prior to analysis, estimates of hummingbird abundance were log– transformed.

To determine the functional structure of hummingbird assemblages, we calculated two components of community trait composition based on the three selected hummingbird traits: (a) the average trait values of the species, quantified by community–weighted mean trait values (CWM) and (b) the degree to which trait values differ among hummingbird species, quantified by three measures of functional diversity (FD). Both components were calculated for each elevation and sampling period. CWM (Lavorel et al. 2008) was quantified as the mean trait value of all hummingbird species present in the community, weighted by their relative abundances from mist–netting during the respective sampling period. To calculate FD, species were projected into a trait–space based on pairwise Euclidean distances, as calculated from the functional traits using principal coordinate analysis (Villéger et al.

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

2008). The three measures of FD were functional richness (FRic), functional evenness (FEve), and functional dispersion (FDis). FRic measures the functional space filled by the species in a community. FEve measures how regular the distances and abundances of individual species are distributed within the functional space of a community (Villéger et al. 2008). FDis measures the spread of species in the functional space as the weighted distance to the assemblage centroid across all species (Laliberté and Legendre 2010). FDis and FEve were weighted by the abundances of the species and FRic was standardized by the ‘global’ FRic that included all species across all elevations. We compared the CWM and the different indices of FD across elevations with linear mixed effect models, including elevation as fixed effect and sampling period as random effect, assuming normally distributed error terms.

Resource use by hummingbirds at the species level

To investigate the structure of plant–hummingbird networks at the species–level, we built matrices with the interaction frequency between hummingbird species and plant morphotypes (HS x PM) for each elevation, pooled across sampling periods. Interaction frequency equalled the number of hummingbird individuals carrying pollen of a particular plant morphotype. Based on these matrices, we calculated the following indices for hummingbird species: complementary specialization d’, paired Differences Index (PDI) and standardized degree. The index d’ is a sampling–robust measure of specialization derived from Kulback–Leibler distances and quantifies how strongly a species deviates from a random sample of interacting partners (Blüthgen et al. 2006). The index ranges from 1 for a completely specialized to 0 for a fully generalized species (Blüthgen et al. 2006). We used a normalised version of the PDI index proposed originally by Poisot et al. (2011), where values of 1 indicate perfect specialists and 0 indicate generalists. This index contrasts the species’ strongest interaction on a resource with those over all remaining resources. In addition, we calculated the standardized degree as the sum of interactions per species scaled by the number of possible partners (Dormann et al. 2013). To compare the different indices among species across elevations, we performed linear mixed effect models with elevation as fixed effect and species identity as random effect, accounting for the fact that we calculated the indices from three different networks. In all models, the indices were 63

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

weighted by the number of individuals sampled per hummingbird species because the recorded interaction links may be incomplete for species with low sample size.

Resource use by hummingbirds at the individual level

To assess patterns of resource use of hummingbirds at the individual level, we used the number of pollen morphotypes carried by each hummingbird individual as a measure of individual specialization. We assumed that the number of pollen morphotypes carried by an individual was associated to the degree of ecological specialization, i.e., few morphotypes of carried pollen indicate high specialization. To assess the effect of elevation on the number of pollen morphotypes, we fitted a generalized linear mixed effect model with Poisson error distribution including elevation as fixed effect and species identity and sampling period as random effects.

In addition, we used pollen loads carried by hummingbirds to build binary interaction matrices of hummingbird individuals by plant morphotypes (∑ HSi x PM) for each elevation. To partition the variation in resource use among hummingbird species and individuals, separately for each elevation, we performed a permutational multivariate analysis of variance, with the function adonis. In this analysis, Sørensen distances were calculated based on the binary interaction matrices. We separately included the two pollen samples (dorsal and ventral) per individual to measure also the variability within individuals (i.e. at different parts of the body). The analysis included recaptures of the same individuals (n = 47 recaptures) and thus also captures the variability between recaptures of the same individuals. Thus, overall variation in hummingbirds` resource use was partitioned among species and individuals, whereas unexplained variance in the residuals accounted for the variation within individuals and between recaptures of the same individuals. The significance of the analysis of variance was assessed against 999 randomizations using a permutation test, based on pseudo F–ratios.

All statistical analyses were conducted with R statistical software ver. 3.0.0 (R Development Core Team 2012) including bipartite (Dorman et al. 2013) and other dedicated packages.

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

Results

Structure of hummingbird assemblages

We captured a total of 28 hummingbird species at the three elevations. Total abundance over all hummingbird species per sampling period (low vs. high elevation, t = -0.93, p = 0.36; mid vs. high elevation, t = 0.85, p = 0.41) and species richness (low vs. high elevation, t = 0.31, p = 0.76; mid vs. high elevation, t = 0.50, p = 0.62) did not differ significantly among elevations. In contrast, the effective number of hummingbird species (eH) was higher at low (t = 21.46, p = 0.02) and mid elevations (t = 2.35, p = 0.03) compared to the highlands. Mean values of the effective number of species over all sampling period across elevations (± 1 SE) were 5.34 ± 0.46, 5.27 ± 0.48 and 3.64 ± 0.52 for low, mid and high elevations, respectively. Species turnover in the hummingbird communities among elevations was high between low and mid elevations (3 shared species, 14% overlap) and between low and high elevations (1 shared species, 3% overlap). In contrast, mid and high elevations had a lower turnover (7 shared species, 37 % overlap).

The functional structure of hummingbird assemblages differed among elevations. At low and mid elevations, traits were more evenly distributed among species and species were on average more distant to the community centroid than in the highlands (Fig. 1B, C; Table 1). Functional richness was similar at all elevations (Fig. 1A; Table 1). At low and mid elevations, we found higher CWMs for bill length and, especially, bill curvature than in the highlands, while the CWM for body mass tended to increase with elevation but this trend was not significant (Fig. 1D, E, F; Table 1).

Resource use by hummingbird species

We identified 208 unique pollen morphotypes collected on 357 individuals of 21 hummingbird species, corresponding to 1.273 plant–hummingbird interactions (see Appendices 4 and 5 for species lists). We did not detect pollen grains on 29 individuals (8% of all sampled birds) of ten hummingbird species.

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

Figure 1. The functional structure of hummingbird assemblages at three elevations in Costa Rica (50, 1000 and 2000 m a.s.l.). Metrics were based on three functional traits (bill length, bill curvature and body mass) measured in 28 hummingbird species. (A, B, C) Indices of functional diversity: functional richness, functional evenness and functional dispersion. (D, E, F) Community weighted means of functional traits. Statistics are provided in Table 1.

Table 1. Linear mixed effect models of the relationships between different metrics of the functional structure of hummingbird assemblages and elevation in Costa Rica. Three indices of functional diversity and community weighted means of functional traits were calculated based on three functional traits (bill length, corolla length and body mass) and 28 hummingbird species. Sampling period was included as a random effect in each model. The reference level (intercept) was the high–elevation assemblage in all models.

Response variable Elevation β t value p Indices of functional diversity Functional richness Low 0.40 0.73 0.477 Mid 0.32 0.57 0.572 Functional evenness Low 1.37 3.94 0.001 Mid 1.81 5.22 < 0.001 Functional dispersion Low 1.52 6.11 < 0.001 Mid 2.07 8.29 < 0.001

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

Table 1 (continued)

Response variable Elevation β t value p Community weighted means Bill length Low 0.74 1.86 0.079 Mid 1.69 4.26 < 0.001 Bil curvature Low 1.83 7.64 < 0.001 Mid 1.91 7.97 < 0.001 Body mass Low -1.03 -2.05 0.055 Mid -0.19 -0.39 0.703

All indices at the species level showed significant variation across elevations and consistent trends. Specialization index d’ was higher at the low and mid elevations compared to the highlands (Fig. 2A, Table 2). The PDI index also showed more specialized hummingbird species at low (PDI = 0.92 ± 0.01) and mid elevation (PDI = 0.95 ± 0.01) than at the highest elevation (PDI = 0.90 ± 0.02) (Table 2). Moreover, standardized degree was higher in the highlands compared to the other elevations, which indicates that hummingbird species visited more different plant species in the highlands than at the other elevations (Fig. 2B, Table 2).

Figure 2. (A and B) Relationship between specialization indices at the species–level (specialization d’ and standardized degree) and elevation. Means and standard errors shown in the bar plots were weighted by the number of individuals sampled per hummingbird species at the respective elevation. (C) Number of pollen morphotypes carried by individual hummingbirds at three different elevations in Costa Rica (n = 357 individuals). Thick horizontal 67

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

lines are medians, boxes indicate 25th and 75th percentiles, whiskers indicate the data range, and the circle is an outlier.

Table 2. Linear mixed effect models of the relationships between specialization indices at the species–level and elevation in Costa Rica. Indices were calculated from 1.273 interactions between 21 hummingbird species and 208 plant species (identified from pollen morphotypes) separately for each elevation. In all models, species identity was included as a random factor and indices were weighted by the number of individuals sampled per hummingbird species. The reference level (intercept) was the high–elevation network in all models.

Response variable Elevation β t value p Specialization Low 1.24 17.39 < 0.001 Mid 1.19 19.07 < 0.001 PDI index Low 0.20 2.40 0.027 Mid 1.06 14.4 < 0.001 Standardized degree Low -0.58 -5.47 < 0.001 Mid -0.87 -10.10 < 0.001

Resource use by hummingbird individuals

The number of pollen morphotypes carried by hummingbird individuals differed across elevations, with individuals at low (z = –4.58, p < 0.001) and mid elevations (z = –4.71, p < 0.001) carrying fewer pollen morphotypes than individuals in the highlands (Fig. 2C).

Most of the variation in resource use by hummingbirds at all elevations was explained by species and individual identity, with a rather low variance residing among samples (i.e. among recaptures and different body parts; residuals < 25% at all elevations) (Table 3). Individual hummingbirds differed strongly in their use of nectar plants at the three elevations, with over 60% of the variation in resource use explained by individual identity in all cases. In contrast, variation at the species level was rather small (≤ 20% in all cases, Table 3). The explained variance at the individual level

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

relative to that at the species level tended to be higher at the high elevation (87.5%) than at mid (77.4%) and low elevations (81.5%).

Table 3. Permutational multivariate analysis of variance on resource partitioning among hummingbird species and individuals at three different elevations in Costa Rica. Observations included 1.273 interactions between 357 hummingbird individuals belonging to 21 hummingbird species and 208 plant species (identified from pollen morphotypes). Based on these observations, binary distance matrices based on Sørensen dissimilarities were constructed. Coefficients of determination (R2) indicate the amount of variation in hummingbirds’ resource use that is explained by: species and individual identity.

Elevation Low Mid High Source of variation R2 p R2 p R2 p Species 0.158 0.001 0.181 0.001 0.097 0.001 Individual 0.698 0.001 0.619 0.001 0.677 0.001 Residuals 0.144 – 0.200 – 0.226 –

Discussion

Our findings show consistent patterns in the functional structure of hummingbird assemblages and specialization of hummingbird species and individuals across elevations. Hummingbird assemblages varied from being functionally even and over– dispersed in the lower elevations to uneven and clustered assemblages in high– elevation environments. Accordingly, hummingbird species and individuals were more specialized at low and mid elevations than at the highest elevation. These corresponding trends suggest that changes in the specialization of hummingbird species and individuals extend to the functional structure of the hummingbird assemblages.

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

Structure of hummingbird assemblages

Total abundance, species and functional richness of hummingbird assemblages remained similar across the three elevations. At all study sites, hummingbird assemblages were characterized by a diverse pool of species with different functional roles, comprising territorial and trap–lining species, as well as species with distinct morphologies (see also Feinsinger and Colwell 1978). For example, hermit hummingbirds, i.e. trap–lining species with long and curved bills, were captured across all elevations. Despite this similarity among assemblages, the effective number of hummingbird species was significantly reduced at the highest elevation. While species abundances were evenly distributed across species at low and mid elevations, the highland assemblage was dominated by Lampornis calolaema that contributed 56% of the captured individuals at this elevation. This trend of reduced effective species richness of hummingbirds is consistent with the general pattern of decreasing diversity with increasing elevation (Koch and Sahli 2012).

The functional structure of the highland assemblage was more uneven und under– dispersed compared to the lower elevations. This may indicate that the assembly of the highland assemblage is subject to environmental filtering. This is consistent with the interpretation of patterns in the phylogenetic structure of hummingbird assemblages in the tropical Andes (Graham et al. 2009, 2012). The hypothesis of environmental filtering assumes that environmental conditions act as a filter allowing only a narrow range of species and functional roles to coexist in harsh, high–elevation environments (Keddy 1992, Mouchet et al. 2010). Consequently, trait values of species tend to be clustered in functional space. Hummingbird assemblages are likely to be strongly influenced by environmental filtering because of metabolic and aerodynamic challenges faced by hummingbirds at high elevations (Altshuler et al. 2004a, 2004b). Moreover, a reduced functional diversity of plant resources could constrain the functional diversity of the dependent consumer guild (Dehling et al. 2014). In addition to environmental factors, competition is another factor influencing community structure and tends to limit the functional similarity of co–occurring species (MacArthur and Levins 1967). A high intensity of interspecific competition is expected to increase the spread of species traits within a community (Laliberté and Legendre 70

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

2010). Accordingly, we found a pattern of functional evenness and over–dispersion at low and mid elevations. This suggests that interspecific competition is particularly strong in these hummingbird assemblages with high effective species numbers (see also Graham et al. 2009, 2012).

We were able to identify functional traits, specifically bill morphology, that were driving the changes in functional community structure. Analyses suggested that species with long and, in particular, curved bills were filtered out at the highest elevation, where mostly species with short and straight or slightly–curved bills occurred. This pattern may primarily arise from a reduced abundance of curved–billed hermit species at high elevations. Throughout the tropics, hermits are mostly limited to lowlands and only a few species occur at high elevations (Snow and Snow 1972). Phylogenetic constraints in evolving traits to adapt to high elevations have been proposed to explain their limited ability to occupy highland habitats (Stiles 2004), e.g. the evolution of larger feet that permit hummingbirds to perch while extracting nectar, reducing hovering costs. Although one hermit species was present at the highest elevation (Phaethornis guy), its abundance was very low compared to the abundance of hermit species at the lower elevations.

Hummingbirds’ resource use

Hummingbird species were less specialized at the highest elevation compared to the lower elevations. These results contrast with those reported by Olesen and Jordano (2002) in their seminal review. They found that the number of interactions of an animal species was not affected by elevation. However, our results are consistent with a study of plant–hummingbird networks along a latitudinal gradient that found decreasing specialization towards higher latitudes (Dalsgaard et al. 2011). This latitudinal trend in specialization mirrors the elevational pattern in our study. High specialization at low elevations may contribute to releasing species from competitors in highly diverse communities, due to increased resource partitioning (MacArthur and Levins 1967, Mouchet et al. 2010). Consistent with this idea, competition for resources has been found to be intense in hummingbird assemblages (Brown and Bowers 1985),

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

favouring the evolution of specialized hummingbird morphologies. For example, hermit hummingbirds have evolved specific bill traits that match closely to the morphology of Heliconia flowers (Heliconiaceae) in tropical lowland forests. The Heliconia–hermit system represents a traditional example of coevolution and is one of the most specialized pollination systems in the Neotropics (Stiles 1975). Although we could not directly test the link between specialization and the intensity of interspecific competition in our study, our findings suggest that specialization in resource use might be crucial for allowing the coexistence of hummingbird species within highly diversified assemblages.

We found that hummingbird individuals were less specialized at the highest elevation, since they visited a larger number of plant species. Specialization, thus, followed the same pattern at the individual and the species level. A possible driver for individual–level specialization could be changes in the intensity of intraspecific competition across elevations. Different degrees of specialization of hummingbird individuals along the elevational gradient could be associated with changes in floral resource availability. Previous studies have reported declines in flowering plant species richness as well as decreases in nectar production of hummingbird–visited plants with increasing elevation (Smith et al. 1995, Biesmeijer et al. 2006, Ornelas et al. 2007). Fewer floral resources at the highest elevation, where total hummingbird abundance was similar to the lowlands, suggest fewer floral resources per capita at this elevation. It is likely that reduced resource availability and increased intraspecific competition constrain the possibility of hummingbird individuals to specialize on specific floral resources and may favour their niche expansion at the highest elevation.

Our results show strong variation in resource use among individuals and a rather low variation among hummingbird species across all elevations. These findings are in line with Tur et al. (2013) who found a high degree of heterogeneity in resource use among individuals in insect–plant mutualistic networks. The low specialization found in hummingbird individuals at the highest elevation may also be associated with high variation in resource use among individuals. Optimal foraging theory predicts that individuals tend to specialize on a few most–preferred resources (Stephens and Krebs 1986). However, under low resource availability, individuals may need to use also less 72

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

preferred resources, leading to niche expansion and a decrease in specialization (Araújo et al. 2008). Niche expansion of individuals may be reflected by an increase in between–individual and a decrease in between–species variation in resource use, consistent with the niche variation hypothesis (Van Valen 1965). It is likely that the niche expansion of hummingbird individuals at the highest elevation may be a major driver of generalization in hummingbirds at this elevation.

Conclusions

Our results show that the functional structure of hummingbird assemblages and specialization of plant–hummingbird networks are systematically affected by elevation. Even and over–dispersed hummingbird assemblages at the lower elevations suggest a high level of floral resource partitioning, probably reducing interspecific competition in specialized plant–hummingbird networks. Specialization on specific floral resources, also corroborated by the specialized bill morphologies at lower elevations, may facilitate the co–existence of hummingbird species within diversified assemblages at low– and mid elevations. In contrast, an uneven and clustered functional structure of the hummingbird assemblage at the highest elevation corresponds to more generalized individuals and species in this assemblage. This may be the result of individual niche expansion as a consequence of a low availability of nectar plants and high intraspecific competition for floral resources in the highlands. Spatial variation in competition and animal resource use at the species and individual level may be a crucial mechanism for shaping the functional structure of highly diversified species assemblages and may also be important for structuring other types of ecological networks and multispecies assemblages.

Acknowledgements

We are grateful to the field and lab assistants who contributed to data collection and pollen identification as well as botanists from the Instituto Nacional de Biodiversidad and La Selva Biological Station (OTS) for support with plant

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

identification. We thank the Museo Nacional for allowing access to its botanical collection. This work was funded by the following organizations: Consejo Nacional para Investigaciones Científicas y Tecnológicas (CONICIT) and Ministerio de Ciencia, Tecnología y Telecomunicaciones (MICIT), Research Centre on Microscopic Structures (CIEMIC) and Centro de Investigación en Ciencias del Mar y Limnología (CIMAR) of the Universidad de Costa Rica (UCR), Universidad Estatal a Distancia (UNED), Organization for Tropical Studies (OTS), German Academic Exchange Service (DAAD) and Tropical Science Centre (TSC). Financial support for this study was also provided by the research–funding programme ‘‘LOEWE–Landes–Offensive zur Entwicklung Wissenschaftlich–ökonomischer Exzellenz’’ of Hesse’s Ministry of Higher Education, Research, and the Arts.

References

Altshuler, D. L., R. Dudley, and J. A. McGuire. 2004a. Resolution of a paradox: Hummingbird flight at high elevation does not come without a cost. Proceedings of the National Academy of Sciences USA 101:17731–17736.

Altshuler, D. L, F. G. Stiles, and R. Dudley. 2004b. Of hummingbirds and helicopters: Hovering costs, competitive ability, and foraging strategies. The American Naturalist 163:16–25.

Araújo, M. S., P. R. Guimarães, Jr, R. Svanbäck, A. Pinheiro, P. Guimarães, S. F. Dos Reis, and D. I. Bolnick. 2008. Network analysis reveals contrasting effects of intraspecific competition on individual vs. population diets. Ecology 89:1981–1993.

Beattie, A. J. 1971. A technique for the study of insect–borne pollen. Pan–Pacific Entomologist 47:82.

Biesmeijer, J. C., S. P. Roberts, M. Reemer, R. Ohlemüller, M. Edwards, T. Peeters, A. P. Schaffers, S. G. Potts, R. Kleukers, C. D. Thomas, et al. 2006. Parallel declines in

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

pollinators and insect–pollinated plants in Britain and the Netherlands. Science 5785:351-354.

Blake, J. G. and B. A. Loiselle. 2000. Diversity of birds along an elevational gradient in the Cordillera Central, Costa Rica. The Auk 117:663–686.

Blüthgen, N., F. Menzel, and N. Blüthgen. 2006. Measuring specialization in species interaction networks. BMC Ecology 6:9.

Brown, J. H. and Bowers, M. A. 1985. Community organization in hummingbirds: Relationships between morphology and ecology. Auk 102:251–269.

Castellanos, M. C., P. Wilson, and J. D. Thomson. 2003. Pollen transfer by hummingbirds and bumblebees, and the divergence of pollination modes in Penstemon. Evolution 57:2742-2752.

Cornwell, W. K., D. W. Schwilk, and D. D. Ackerly. 2006. A trait–based test for habitat filtering: convex hull volume. Ecology 87:1465–1471.

Dalsgaard B, A. M. Martín Gonzáles, J. M. Olesen, J. Ollerton, A. Timmermann, L. H. Andersen, and A. G. Tossas. 2009. Plant–hummingbird interactions in the West Indies: floral specialization gradients associated with environment and hummingbird size. Oecologia 159:757–766.

Dalsgaard, B., E. Magard, J. Fjeldsa, A. M. Martín González, C. Rahbek, J. M. Olesen, J. Ollerton, R. Alarcón, A. Cardoso Araujo, P. A. Cotton, et al. 2011. Specialization in plant–hummingbird networks is associated with species richness, contemporary precipitation and quaternary climate-change velocity. PLoS ONE 6 e25891.

Dehling, D. M., S. A. Fritz, T. Töpfer, M. Päckert, P. Estler, K. Böhning-Gaese, and M. Schleuning. 2014. Functional and phylogenetic diversity and assemblage structure of frugivorous birds along an elevational gradient in the tropical Andes. Ecography doi: 10.1111/ecog.00623

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

Dias, A. T. C., M. P. Berg, F. de Bello, A. R. Van Oosten, K. Bílá, and M. Moretti. 2013. An experimental framework to identify community functional components driving ecosystem processes and services delivery. Journal of Ecology 101:29–37.

Díaz, S. and M. Cabido. 2001. Vive la différence: plant functional diversity matters to ecosystem processes. Trends in Ecology and Evolution 16:646–655.

Dorman, C. F., J. Fruend, and B. Gruber. 2013. Package “bipartite”. Version 2.0. Visualising bipartite networks and calculating some (ecological) indices.

Feinsinger, P. and R. K. Colwell. 1978. Community organization among neotropical nectar feeding birds. American Zoologist 18:779–795.

Feinsinger, P., J. H. Beach, Y. B. Linhart, W. H. Busby, and K. G. Murray. 1987. Disturbance, Pollinator Predictability, and Pollination Success Among Costa Rican Cloud Forest Plants. Ecology 68:1294–1305.

Graham, C. H., J. L. Parra, C. Rahbek, and J. A. McGuire. 2009. Phylogenetic structure in tropical hummingbird communities. Proceedings of the National Academy of Sciences USA 106:19673–19678.

Graham, C. H., J. L. Parra, B. A. Tinoco, F. G. Stiles, and J. A. McGuire. 2012. Untangling the influence of ecological and evolutionary factors on trait variation across hummingbird assemblages. Ecology 93:S99–S111.

Hainsworth, F. R. and L. L. Wolf. 1972. Crop volume, nectar concentration and hummingbird energetics. Comparative Biochemestry and Physiology 42A:359–66.

Hartshorn, G. and R. Peralta. 1988. Preliminary description of primary forests along the La Selva–Volcan Barva altitudinal transect, Costa Rica. Pages 281–295 in F. Almeda and C. M. Pringle, editors. Tropical rainforests: diversity and conservation. Academy of Sciences, California, USA.

Holdridge, L. R. 1967. Life zone ecology. Tropical Science Center. San Jose, Costa Rica.

Inouye, D. W. 1978. Resource partitioning in bumblebees: experimental studies of foraging behavior. Ecology 59:672–678.

Jost, L. 2006. Entropy and diversity. Oikos 113:363–375.

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

Kearns, C. A. and Inouye, D. W. 1993. Techniques for pollination biologists. Univ. Press of Colorado.

Keddy, P. A. 1992. Assembly and response rules – 2 goals for predictive community ecology. Journal of Vegetation Science 3:157–164.

Kershaw, F. L. 2006. The co–evolution of hummingbird bill morphology and floral structures with regard to both native and non–native flower species. Department of Biology, School of Biological Sciences. Leeds University. UK.

Koch, J. B. and H. F. Sahli. 2012. Patterns of flower visitation across elevation and successional gradients in Hawai’i. Pacific Science 67:1–25.

Körner, C. 2000. Why are there global gradients in species richness? Mountains might hold the answer. TREE 15:513–514.

Laliberté, E. and P. Legendre. 2010. A distance–based framework for measuring functional diversity from multiple traits. Ecology 91:299–305.

Lavorel, S., K. Grigulis, S. McIntyre, N. S. G. Williams, D. Garden, J. Dorrough, S. Berman, F. Quétier, A. Thebault, and A. Bonis. 2008. Assessing functional diversity in the field – methodology matters! Functional Ecology 22:134–147.

MacArthur, R. H. and R. Levins. 1964. Competition, habitat selection and character displacement in a patchy environment. Proceedings of the National Academy of Sciences USA 51:1 207–10.

MacArthur, R. H., and R. Levins. 1967. Limiting similarity convergence and divergence of coexisting species. The American Naturalist 101:377–385.

McCain, C. M. 2009. Global analysis of bird elevational diversity. Global Ecology and Biogeography 18:346–360.

McGill B. J., B. J. Enquist, E. Weiher, and M. Westoby. 2006. Rebuilding community ecology from functional traits. Trends Ecol Evol 21:178–185. doi:10.1016/j.tree.2006.02.002

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

Olesen, J. M. and P. Jordano. 2002. Geographic patterns in plant–pollinator mutualistic networks. Ecology 83:2416–2424.

Ornelas, J. F., M. Ordano, A. J. De-Nova, M. E. Quintero, and T. Garland Jr. 2007. Phylogenetic analysis of interspecific variation in nectar of hummingbird–visited plants. Journal of Evolutionary Biology 20:1904–1917.

Pires, M. M., P. R. Guimarães, Jr., M. S. Araújo, A. A. Giaretta, J. C. Costa, and S. F. dos Reis. 2011. The nested assembly of individual-resource networks. Journal of Animal Ecology 80:896–903.

Plein, M, L. Längsfeld, E. L. Neuschulz, C. Schultheiß, L. Ingmann, T. Töpfer, K. Böhning– Gaese, and M. Schleuning. 2013. Constant properties of plant–frugivore networks despite fluctuations in fruit and bird communities in space and time. Ecology 94:1296–1306.

Poisot, T, E. Canard, N. Mouquet, and M. E. Hochberg. 2011. A comparative study of ecological specialization estimators. Methods in Ecology and Evolution doi: 10.1111/j.2041-210X.2011.00174.x

R Development Core Team 2012. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing.

Ralph, C. J., G. R. Geupel, P. Pyle, T. Martin, and D. F. DeSante. 1993. Handbook of field methods for monitoring landbirds. Pacific Southwest Research Station Albany, California, USA.

Roubik, D. W. and J. E. Moreno. 1991. Pollen and spores of Barro Colorado Island. Missouri Botanical Garden Press 36, Panama.

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

Sanders, N. J. and C. Rahbek. 2012. The patterns and causes of elevational diversity gradients. Ecography 35:1–3.

Schoener, T. W. 1974. Resource partitioning in ecological communities. Science 185:27–39.

Smith, T. B., L. A. Freed, J. A. Lepson, and J. H. Carothers. 1995. Evolutionary consequences of extinctions in populations of a Hawaiian Honeycreeper. Conservation Biology 9:107–113.

Snow, B. K. and D. W. Snow. 1972. Feeding niches of hummingbirds in a Trinidad valley. Journal of Animal Ecology 41:471–485.

Stephens, D. W. and J. R. Krebs. 1986. Foraging Theory. Princeton University Press, Princeton, New Jersey, USA.

Stiles, F. G. 1975. Ecology, flowering phenology, and hummingbird pollination of some Costa Rican Heliconia species. Ecology 56:285–301.

Stiles, F. G. 2004. Phylogenetic constraints upon morphological and ecological adaptation in hummingbirds (Trochilidae): why are there no hermits in the paramo?. Ornitología Neotropical 15:191–198.

Stiles, F. G. and A. Skutch. 1989. A guide to the birds of Costa Rica. Comstock Publishing Associates, Ithaca, USA.

Tilman, D., J. Knops, D. Wedin, P. Reich, M. Ritchie, and E. Siemann. 1997. The influence of functional diversity and composition on ecosystem processes. Science 277:1300–1302.

Totland, Ø. 2001. Environment dependent pollen limitation and selection on floral traits in an alpine species. Ecology 82:2233–2244.

Appendix 1: Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations

Van Valen, L. 1965. Morphological variation and width of ecological niche. The American Naturalist 99:377–390.

Villéger, S., N. W. Mason, and D. Mouillot. 2008. New multidimensional functional diversity indices for a multifaceted framework in functional ecology. Ecology 89:2290–2301.

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Appendix 2: Morphological traits determine specialization and resource use in plant–hummingbird networks in the Neotropics

Status: accepted in Ecology, http://dx.doi.org/10.1890/13-2261.1 (March 10, 2014).

Maglianesi, M. A., N. Blüthgen, Katrin Böhning-Gaese and M. Schleuning. Morphological traits determine specialization and resource use in plant–hummingbird networks in the Neotropics.

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Erklärung über Anteile der Autoren/Autorinnen

(1) Entwicklung und Planung

MM 30%; NB 20%; KBG 20%; MS 30%

(2) Durchführung der einzelnen Untersuchungen / Experimente

MM selected the study sites, conducted observations of plant–hummingbird interactions, mist–netting, nectar sampling and measurements of morphological traits of plant and hummingbird individuals. MM coordinated the fieldwork; NB, KB and MS discussed methodological adjustments after collection of preliminary data.

(3) Erstellung der Datensammlung und Abbildung

MM conducted data analyses and created all figures; figures were discussed with NB, KB and MS.

(4) Analyse/Interpretation der Daten

MM 70%; NB 5%; KBG 5%; MS 20%

(5) Übergeordnete Einleitung / Ergebnisse / Diskussion

MM 70%; NB 5%; KBG 5%; MS20%

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

María Alejandra Maglianesi1,2*, Nico Blüthgen3, Katrin Böhning–Gaese1,4 and Matthias Schleuning1

1 Biodiversity and Climate Research Centre (BiK–F) and Senckenberg Gesellschaft für Naturforschung, 60325 Frankfurt am Main, Germany

2 Vicerrectoría de Investigación. Universidad Estatal a Distancia (UNED), San José, Costa Rica

3 Ecological Networks, Department of Biology, Technical University of Darmstadt, 64287 Darmstadt, Germany

4 Department of Biological Sciences, Johann Wolfgang Goethe University of Frankfurt, 60438 Frankfurt am Main, Germany

Running title: Specialization in pollination networks

Type of manuscript: Article

*Corresponding author: María Alejandra Maglianesi. Vicerrectoría de Investigación. Universidad Estatal a Distancia (UNED), San Pedro Montes de Oca, San José, Costa Rica. E-mail: [email protected]

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Abstract

Ecological communities are organized in complex ecological networks. Trait–based analyses of the structure of these networks in highly diversified species assemblages are crucial for improving our understanding of the ecological and evolutionary processes causing specialization in mutualistic networks. In this study, we assess the importance of morphological traits for structuring plant–hummingbird networks in neotropical forests by using a novel combination of quantitative analytical approaches. We recorded visitation of hummingbirds to plant species over an entire year at three different elevations in Costa Rica and constructed quantitative networks based on interaction frequencies. Three morphological traits were measured in hummingbirds (bill length, bill curvature and body mass) and plants (length, curvature and volume of corolla). We tested the effects of avian morphological traits and abundance on ecological specialization of hummingbird species. All three morphological traits of hummingbirds were positively associated with ecological specialization, especially bill curvature. We tested whether interaction strength in the networks was associated with the degree of trait matching between corresponding pairs of morphological traits in plant and hummingbird species and explore whether this was related to resource handling times by hummingbird species. We found strong and significant associations between interaction strength and the degree of trait matching. Moreover, the degree of trait matching, particularly between bill and corolla length, was associated with the handling time of nectar resources by hummingbirds. Our findings show that bill morphology structures tropical plant–hummingbird networks and patterns of interactions are closely associated with morphological matches between plant and bird species and the efficiency of hummingbirds' resource use. These results are consistent with the findings of seminal studies in plant–hummingbird systems from the neotropics. We conclude that trait–based analyses of quantitative networks contribute to a better mechanistic understanding of the causes of specialization in ecological networks and could be valuable for studying processes of complementary trait evolution in highly diversified species assemblages.

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Key words: biotic interactions, Costa Rica, fourth–corner analysis, hummingbirds, mutualistic networks, neotropical forest, optimal foraging, pollination, specialization, trait complementarity.

Introduction

Specialization is a central concept in community ecology because it influences species co–existence and the structure and stability of ecological communities (Thompson 1994, Waser et al. 1996). In recent years, interactions between plants and pollinators have often been interpreted in the context of plant–pollinator networks (e.g. Blüthgen et al. 2007, Olesen et al. 2007). This approach is a powerful tool to analyse the complexity of ecological communities (Ings et al. 2009). Plant–pollinator networks systematically vary in their degree of specialization (Dalsgaard et al. 2011, Schleuning et al. 2012) and this variability may be associated with the morphological traits of interacting plants and pollinators (Ings et al. 2009). Although a number of earlier studies have tested the importance of species traits for plant–pollinator interactions (Linhart 1973, Stiles 1975, Temeles and Kress 2003), studies using trait– based analyses in the community–context of ecological networks are still rare (but see Stang et al. 2009, Junker et al. 2013).

Specialization occurs when plants are visited by a relatively small proportion of the available pollinators in a community (Armbruster et al. 2000, Johnson and Steiner 2000) and, vice versa, when pollinators restrict the use of flower resources to a subset of plant species in relation to overall resource availability. Specialization can be viewed from an ecological perspective, in which ecological specialization refers to the state of being specialized under current ecological conditions or an evolutionary perspective, in which evolutionary specialization describes the process of evolving toward greater specialization (Armbruster 2006). Several non–mutually exclusive mechanisms have been proposed to explain causes of ecological specialization of species in plant–animal interaction networks. The species trait hypothesis has received particular attention (Santamaría and Rodríguez-Gironés 2007, Stang et al. 2009). This hypothesis states that morphological, behavioural, and life–history traits constrain the type, number and

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

strength of interactions exerted by a species (e.g. Linhart 1973, Stiles 1975, Stang et al. 2009, Junker et al. 2013). The neutral hypothesis, in contrast, assumes that network structure largely results from random interactions between interacting species of different abundances (Dupont et al. 2003, Vázquez 2005). Here, we employ mutualistic plant–hummingbird networks to test the trait hypothesis and to identify which avian morphological traits are most important in determining ecological specialization (i.e., niche partitioning of plant resources) of hummingbird species.

Trait matching or complementarity between pairs of interacting species can be considered as a consequence of specialization, either through reciprocal co– evolutionary processes or through ecological fitting between pairs of species with independent trait evolution (Janzen 1980, 1985, Guimarães et al. 2011). According to Blüthgen et al. (2008), trait matching refers to the partitioning of interaction partners between species, resulting from the correspondence of phenotypic traits of the interacting species (i.e., phenotypic specialization). Several studies have shown that trait matching influences patterns of interactions between plant and pollinator species (Stiles 1975, Wolf et al. 1976, Dalsgaard et al. 2009, Stang et al. 2009). However, most research has been focused on specific pollinator species and their food plants (e.g. Temeles et al. 2009, Dohzono et al. 2011), calling for more integrated studies at the community level. A high degree of matching in morphological traits between flowers and their pollinators may contribute to a high quality of pollination services and a high efficiency in resource use by faster nectar intakes, leading to fitness benefits for both plants and pollinators (Temeles 1996, Dohzono et al. 2011). In this study, we assess the relationship between interaction strength in plant–hummingbird networks and the degree of trait matching in corresponding pairs of morphological traits of plant and hummingbird species. In addition, we examined whether increased trait matching between plants and hummingbirds corresponds to decreased hummingbird handling times on flowers.

In the Neotropics, hummingbirds (Trochilidae) are considered to be the most specialized nectar–feeding birds (Linhart 1973, Stiles 1981). Hermit hummingbirds have evolved exclusive morphological adaptations in bill traits toward corresponding flower morphologies as well as a high degree of ecological specialization (Stiles 1978). 86

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

By contrast, non–hermit hummingbirds exhibit a wider range of bill morphology and degrees of ecological specialization (Cotton 1998). The variety of morphological types and the degree of specialization on their nectar plants make hummingbirds an ideally suited study system to assess the influence of morphological traits on interaction patterns in pollination networks. The goal of this study is to evaluate the role of morphological traits (i.e., phenotypic specialization) for ecological specialization in plant–hummingbird networks in three types of neotropical forests at different elevations. We addressed the following questions: (1) Are morphological traits of hummingbird species associated with the degree of ecological specialization in plant– hummingbird networks? (2) Is the degree of trait matching between plant and hummingbird species associated with their pair–wise interaction strength in the network? (3) Does increased trait matching between plants and birds correspond to a decreased handling time of nectar resources by hummingbirds?

Methods

Study area and sampling design

The study was conducted in northeastern Costa Rica within the forest of the La Selva–Braulio Carrillo corridor on the Caribbean slope of the Cordillera Central. This area extends from La Selva Biological Station (LS) (ca. 1,500 ha) to the Braulio Carrillo National Park (ca. 45,000 ha). Our study sites included three tropical forest types located at different elevations: wet forest (50 m; 10°26’N, 84°01’W) in LS, pre– montane forest (1,000 m; 10°16’N, 84°05’W) and lower montane wet forest (2,000 m; 10°11’N, 84°07’W) in the park (Holdridge 1967). All sites were located in old–growth forest. Canopy heights were approximately 30 to 40 m at LS, 30 to 35 m at 1,000 m, and 20 m at 2,000 m (Hartshorn and Peralta 1988). Mean annual temperature ranges from 25ºC in the lowlands to 14ºC in the highlands, while mean annual precipitation ranges from 4,300 in the lowlands to 2,200 in the highlands (Blake and Loiselle 2000, TEAM 2013). The dry season lasts from January to April and the wettest months are July and October–November.

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

The study was conducted from May to September 2011 and from December 2011 to April 2012. During the period of sampling, we collected data on abundances and morphological traits of flower resources and hummingbirds and on plant– hummingbird interactions across seven sampling periods per site, each lasting for about 10 days.

Flower abundance and traits

To estimate the abundance of floral resources in the forest understory, we established five transects of 100 x 5 m at each study site, with transects being separated from one another by at least 50 m. Along these transects, open flowers of all plant species fitting the traditional ornithophilous syndrome (Faegri and van der Pijl 1979) were counted up to 10 m above the ground. Because hummingbird–pollinated flowers do not always fit into this syndrome (Ollerton et al. 2009), we also considered plant species fitting other pollination syndromes (e.g. bat– or insect–pollinated flowers) that were likely to be visited by hummingbirds as well. Transect counts were done once during each sampling period, that is, we collected abundances of plant species blooming at different times of the year. We used the total number of flowers per plant species summed over all transects and sampling periods as an estimate of plant species–specific resource abundance for hummingbirds at each of the three elevations.

We measured the following plant morphological traits that have been reported to affect plant–hummingbird interactions: external diameter, length and curvature of the corolla (Costigan 2008, Temeles et al. 2009). External diameter (maximum width at the opening of the corolla tube) and total corolla length (from the base to the corolla opening) were measured to the nearest 0.10 mm with a dial caliper. To measure corolla curvature, the flower was placed on graph paper so that the angle of deflection could be calculated using simple trigonometry (see Kershaw 2006). Corolla curvature was arcsin–sqrt–transformed for statistical analysis. Using corolla length and external diameter, we additionally calculated corolla volume CVOL as: CVOL = corolla length π (external diameter/2)² for all plant species. Corolla volume can be considered as an

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

integrated measure of flower size, and flower size has been found to be correlated with nectar production rates per flower (Ornelas et al. 2007). Corolla volume was log– transformed prior to statistical analyses. For all three morphological traits, we calculated mean values from 3–4 individual flowers for 133 plant species (mean values are given in Table A1). In addition to morphological traits, we measured nectar volume for a subset of the plant species (n = 41 species). We sampled nectar from unvisited flowers on inflorescences that had been bagged with nylon netting prior to anthesis. Nectar from 3–5 flowers of each plant individual was extracted with capillary tubes until no further nectar could be extracted. We sampled nectar from 3–15 individuals per plant species between 10:00 and 13:00 (approximately 24 h after flowers had been bagged), which approximates the daily nectar production of each flower.

Hummingbird abundance and traits

We placed at each study site 12–14 standard mist nets (12 x 3 m) for approximately 6 h after dawn (Ralph et al. 1993). Mist nets were operated 4 days in each sampling period. To calculate the sampling effort, one standard mist net operated for one hour was considered as a net–hour. Overall, our sampling effort was about 52,300 mist–net hours and was similar across the three study sites. We used the number of hummingbird individuals captured per species, summed across sampling periods, as an estimate of hummingbird abundance at each site. All hummingbirds captured were identified to species level (according to Stiles and Skutch 1989) and banded with aluminum numbered bands. To avoid overestimation of hummingbird abundance, we excluded recaptured hummingbird individuals from abundance estimates. We measured avian morphological traits that have been found to affect plant– hummingbird interactions, including bill length and curvature (Hainsworth and Wolf 1972, Temeles et al. 2009), as well as body mass (Temeles and Kress 2003, Dalsgaard et al. 2009) (mean values are given in Table A2). For each captured individual, we measured bill length (exposed culmen) to the nearest 0.10 mm using a dial caliper. To measure bill curvature, we placed the bill on graph paper following the same procedure as for corolla curvature (Kershaw 2006). Bill curvature was arcsin–sqrt– transformed for statistical analyses. We used a digital scale to the nearest 0.10 g to record body mass. 89

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Network observations

To record the interactions between plant and hummingbird species in the understory, we carried out observations of flowering plants up to 10 m above the ground. Plant species were chosen following the same criteria as for transect counts of floral resources. We randomly chose 4–12 plant individuals per species at each study site and sampling period. We videotaped 8–12 individuals of the more abundant plant species, and 4–6 individuals of the less abundant species. For recording visits of hummingbirds to plant individuals, we fixed unattended cameras within 10 m from open flowers for periods of 120 minutes between 06:00 and 14:00. In most cases, all open flowers on a plant individual were video–taped together.

We recorded 1,073 plant individuals and over 2,000 hours of videotapes. Of the videotaped plants, 35% were visited by hummingbirds during the recording period. We could not identify 45 of the total hummingbird visitors (5%), which where excluded from further analysis. The parameters recorded from the videotapes were: the number of flowers probed at each visit, the feeding time at each visit and the contact with reproductive structures of the flower. A visit to a plant individual was recorded whenever an individual hummingbird was observed to probe at least one flower of the observed plant individual. We excluded all illegitimate visits in which the hummingbird did not access the flower through the corolla entrance. Since these illegitimate visits represented only 2.9% of the total visits, including these visits in the analyses did not modify the results. We used the interaction frequency as the currency in the networks, that is, the total number of legitimate visits of each hummingbird species on each plant species.

We compiled one interaction matrix for each study site lumped across all sampling periods in order to assess hummingbird specialization across the entire study year. We were not interested in phenological differences in hummingbird specialization. In addition to the plant–hummingbird networks, we calculated the mean foraging time per flower as a measure of individual resource handling time (i.e., the cumulative foraging time on all individual flowers during a visit divided by the number of flowers probed at this visit). Foraging time on each flower was considered as the time it takes a

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

bird to insert its bill, lick up nectar, and withdraw its bill from a flower (Montgomerie 1984), excluding the transit time between flowers and the time that hummingbirds spent hovering around the flowers.

Statistical analyses

To investigate patterns of specialization in plant–hummingbird interactions, we determined specialization at the community level with the standardized two– dimensional Shannon entropy (specialization index H2’) and at the species level with the conceptually related index d’ (Blüthgen et al. 2006). The specialization index H2’ quantifies the degree of niche divergence (i.e., niche complementarity) between flowering plant species and between hummingbird species in the interaction networks and thus estimates the degree of complementary specialization in a network. We used null–models based on the Patefield algorithm (Patefield 1981), assuming random interactions between species constraining species' total frequencies, to assess whether network specialization H2’ at each study site was higher than expected at random. To test for differences from randomness, we compared observed H2’ values with those obtained from 10,000 permutations of randomized networks. The species–level network index d’ is derived from the Kullback–Leibler distance and measures how strongly a pollinator species deviates from an expected random choice of available interaction partners (Blüthgen et al. 2006). To calculate d’, we derived expected interaction frequencies between hummingbirds and flowers according to the independent estimates of flower abundance at each study site. According to this concept of ecological specialization, a generalized species uses floral resources proportional to floral abundances, whereas a specialized species strongly deviates in its interaction pattern from the distribution of floral abundances (Blüthgen et al. 2006).

Both indices H2’ and d’ range from 0 to 1 and have the advantage of being largely unaffected by the number of interacting species and by differences in sampling intensity (Blüthgen et al. 2006).

To assess whether avian morphological traits and abundance were associated with hummingbird specialization at the species level, we used (a) univariate linear models

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

for bill length, curvature, body mass and abundance, and (b) a multivariate model including all traits and abundance as predictors of hummingbird specialization. We fitted linear mixed effects models with traits and abundance as fixed effects and accounted for random variation among the three different networks from different elevations by including elevation as a random effect. In all models, d’ values were weighted by total interaction frequencies of hummingbirds (sqrt–transformed) because the observed interaction links may be incomplete for species with very few observations. To explore whether the association between morphological traits of hummingbird species and specialization was context–dependent (i.e., varied between elevations), we additionally fitted random–slope models that allowed trait– specialization relationships to vary between elevations (Zuur et al. 2009). We compared the fits between random–intercept and random–slope models for testing whether trait–specialization relationships varied among elevations; this would be the case if random–slope models were more supported than random–intercept models. For the model comparison, we used χ2 distributed likelihood–ratio tests. In addition, we accounted for potential differences between hermit and non–hermit species and between genera by adding hermit/non–hermit and genus as nested random–intercept factors to the models. To compare models with and without these additional random effects, we used χ2 distributed likelihood–ratio tests. We also tested whether morphological traits differed between hermit and non–hermit hummingbirds. To identify minimal adequate models from the full multivariate model, we fitted models with all possible combinations of the four predictor variables and chose the most parsimonious model according to the corrected Akaike Information Criterion (AICc).

To assess whether the degree of trait–matching affects interaction strength in the networks, we performed fourth–corner analyses on the interaction frequencies of bird and plant species in the three study sites. We used the following combinations of corresponding hummingbird and plant traits that may influence the interaction patterns between species (Dalsgaard et al. 2009, Temeles et al. 2009): (1) bill–corolla length, (2) bill–corolla curvature and (3) body mass–corolla volume. Prior to analysis, traits were standardized to zero mean and unit variance. The fourth–corner method was proposed to measure and test the relationships between species traits and

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

environmental variables (Legendre et al. 1997, Dray and Legendre 2008). We adapted the method to detect significant associations in interaction strength between plant and hummingbird species with specific morphologies at each of the three elevations. Specifically, we tested whether interaction strength between pairs of species with high degrees of matching in the corresponding traits was higher than expected from a null model (for details see Dehling et al. 2014). Interaction strength was the relative frequency of a hummingbird species on a particular plant species (number of interactions between hummingbird species j and plant species i divided by the total number of visits of hummingbird species j to all plant species). We considered three tables: a table L (n x p) describing the interaction strength of p species of hummingbirds with n plant species, a second table R (n x m) with m trait values for the n plant species, and a third table Q (p x s) containing s trait values for the p species of hummingbirds. For testing the significance of the correlations between corresponding pairs of traits, we used a combination of permutation models 2 (i.e., permutation of entire rows in table L) and 4 (i.e., permutation of entire columns in table L) (Dray and Legendre 2008). From these models, we chose the larger of the two p–values as suggested by Ter Braak et al. (2012), as the most conservative approach.

To analyze the influence of trait matching on the efficiency in hummingbirds' resource use, we used the trait distances of bill–corolla length and bill–corolla curvature as a measure of the degree of mismatching between pairs of traits. These distances were calculated as absolute differences between mean values of corresponding pairs of morphological traits in plant and hummingbird species. We fitted linear mixed effects models with handling time as the response variable and the degree of trait mismatching as the fixed effect, analysing each trait combination separately. To account for differences in nectar production per flower among plant species, we included the mean nectar volume per flower as an additional fixed effect. We could not account for the actual standing crop (i.e., the amount of nectar present at the time of a visit) which was impossible to measure for all observed plant individuals. To account for additional random variation among sites, species and plant individuals, we included the following random effects in these models: site, hummingbird species identity and plant individual identity nested within plant species

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

identity. In addition to the two univariate models, we fitted a multivariate model to simultaneously test the effect of mismatches in length and curvature on handling time. We selected a minimal adequate model among the three possible model combinations as described above.

All statistical analyses were conducted with R statistical software ver. 3.0.0 (R Development Core Team 2013) and dedicated packages. We used marginal and conditional R2 as goodness of fit statistics for all mixed effects models (Nakagawa and Schielzeth 2013).

Results

We captured a total of 762 individuals belonging to 28 hummingbird species at the three study sites (Table A3). Across all study sites, we observed a total of 823 visits of hummingbirds to plant individuals between 65 flowering plant species and 20 hummingbird species. Networks at each elevation included 23 x 8, 25 x 8 and 20 x 9 (plant x hummingbird species) at low, mid and high elevation, respectively (see Tables A3 and A4 for complete species lists, Supplemental Material). Among the 20 hummingbird species, 5 were hermits and 15 non–hermits. Non–hermit hummingbirds had shorter (F1,23 = 6.5, β = –1.02, p = 0.018) and less curved bills (F1,23 = 24.9, β = –0.19, p < 0.001) compared to hermits, whereas body mass did not differ (F1,23 = 0.006, β = 0.04, p = 0.94). Plant–hummingbird networks at all elevations were highly specialized

(Fig. 1). In all networks, specialization H2’ was higher than expected at random (p < 0.001 in all cases).

Species–level specialization (d’) increased with bill length, bill curvature and body mass (Fig. 2). Hummingbird abundance did not affect specialization of hummingbird species (t = –0.58, p = 0.57). When we accounted for differences between hermits and non–hermits and the taxonomic relatedness among species, this improved the fit only for the body–mass model (model with vs. model without taxonomy: χ2 = 7.55, p = 0.023), while the models with bill length (χ2 = 1.16, p = 0.56) and bill curvature (χ2 < 0.1, p > 0.9) were less supported. Models that allowed for random–slopes of trait– specialization relationships at each elevation were always less supported than

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

random–intercept models (χ2 < 0.1, p > 0.9). The best–fitting model, according to AICc, contained only bill curvature as predictor of specialization (R2 = 0.47, p < 0.001; ΔAICc to all other models > 2), in multivariate models both with and without taxonomic effects.

Figure 1. Plant–hummingbird interaction networks in tropical forests at three elevations in Costa Rica. Hummingbird and plant species are indicated by boxes at top and bottom, respectively. Box width corresponds to the proportion of interactions contributed by each species to the network. Links between species are indicated by lines that are proportional to interaction strength. Complementary specialization H2’ is reported for each elevation. Examples of interactions between hummingbird and plant species are shown for each elevation: Phaethornis striigularis and Lampornis hemileucus interacting with Renealmia cernua (Zingiberacea) at low and mid elevation, respectively; and Selasphorus flammula feeding on Disterigma humboldtii (Ericaceae).

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Figure 2. Relationships between species–level network specialization d’ and morphological traits for 20 hummingbird species from three hummingbird communities in Costa Rica at different elevations (5 species occured at two elevations). Coefficients of determination (R²) were 0.22, 0.47 and 0.12 for single predictor models including bill length, bill curvature and body mass, respectively. Symbol size corresponds to weights by the number of interactions observed for each hummingbird species at the respective elevation.

Fourth–corner analyses revealed that interaction strength in the networks was associated with the degree of trait matching in corresponding morphological traits of hummingbirds and plants, especially at mid and high elevations (Fig. 3, Table 1). Interaction strength was stronger between plant and hummingbird species with a high degree of matching in bill–corolla length in all three communities. Interaction strength was associated with matches in body mass and corolla volume at mid and high elevation, and in bill and corolla curvature at mid elevation only.

Figure 3. Associations between morphologies of plant and hummingbird species in mutualistic networks at three study sites in Costa Rica. Each data point represents one hummingbird

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

species at a given elevation (n = 23); 5 species occurred at two study sites. For each hummingbird species, we computed the corresponding mean floral trait value across all observed interactions at a given elevation. Shown are mean (± 1 SE) floral trait values (length, curvature and volume of corolla) against the respective hummingbird trait value (bill length, bill curvature and body mass). Black dots represent hummingbird species that were only observed once. We fitted a simple linear model with the respective mean values to indicate a trend line in each plot.

The degree of mismatching between pairs of corresponding morphological traits increased resource handling time by hummingbirds, accounting for the differences in nectar volume among plant species. Together with nectar volume, mismatches between bill length and corolla length explained about 20% of the variation in bird’s handling time, while mismatches between bill curvature and corolla curvature explained much less variation (Table 2). The best–fitting model included nectar and mismatches between bill length and corolla length as predictors, while the second– best model additionally included mismatches between bill curvature and corolla curvature (∆AICc to the best model 1.87).

Table 1. Statistics of the fourth–corner analyses for corresponding pairs of traits in plant– hummingbird networks at three elevations in Costa Rica. Given are correlation coefficients and their respective p–values from permutation tests. To test whether the degree of trait– matching is significantly associated with the interaction strength between plant and hummingbird species, we used two different permutation tests and report the larger of the two p–values (see methods for details). Significant associations are shown in bold.

Elevation Low Mid High Trait combination r p r P r p Bill–corolla length 0.491 0.040 0.573 0.045 0.599 0.002 Bill–corolla curvature 0.295 0.141 0.661 0.025 0.346 0.130 Body mass–corolla volume 0.007 0.808 0.542 0.040 0.642 0.015

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Table 2. Linear mixed effects models testing the relationship between resource handling time by hummingbirds and the degree of mismatch between morphological bird and plant traits, accounting for differences in nectar volume. Pairs of corresponding traits were bill–corolla length and bill–corolla curvature and mismatches were defined as absolute differences between the mean values of the corresponding pairs of morphological traits. Observations included 557 individual plant–hummingbird interactions. Marginal and conditional R2 are shown as goodness of fit statistic in both models.

Predictor β t p R2 marginal R2 conditional ∆Bill–corolla length 0.25 3.35 < 0.001 0.20 0.60 Nectar volume 0.32 2.55 0.011 – ∆Bill–corolla curvature 0.14 1.23 0.221 0.17 0.63 Nectar volume 0.39 2.58 0.010 –

Discussion

Our results show high levels of ecological specialization in plant–hummingbird networks at all three elevations. Morphological traits of hummingbird species influenced patterns of ecological specialization and bill traits were more relevant than body mass in determining niche partitioning within the community. Interaction strength in the networks was stronger between plant and hummingbird species with close matches in their corresponding morphological traits. Trait matching was associated with a decreased handling time of nectar resources by hummingbirds. These findings indicate the high sensitivity of quantitative network analyses for detecting trait associations in mutualistic plant–animal systems.

Network specialization

Recent network analyses showed moderate specialization in most plant–pollinator associations (Blüthgen et al. 2007, Schleuning et al. 2012). We found that complementary specialization (H2’) of plant–hummingbird networks was high compared to the specialization reported for 25 tropical pollination networks in a previous meta–analysis (mean ± SE: H2’ = 0.43 ± 0.03; Schleuning et al. 2012). The high degree of specialization in the plant–hummingbird networks was consistent across elevations (H2’ > 0.5 in all three elevations). All networks were more specialized than 98

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

one would expect in a randomly interacting community. These findings suggest a pronounced flower partitioning among hummingbird species, which is consistent with previous research (e.g. Dalsgaard et al. 2011). In specialized systems, consumer species tend to be adapted to use their resources effectively (Blüthgen et al. 2007). The use of a subset of available floral resources by a given species may increase the resource–use efficiency and/or reduce interspecific competition (Linhart 1973, Stiles 1981). Since hovering flight implies a high metabolic cost for hummingbird species (Suarez 1998), ecological specialization could reduce the foraging costs of non–perching hummingbird species because of less interference competition with other species (see Feinsinger 1976).

Species–level specialization

Our findings support the idea that certain morphological traits of hummingbirds contribute to their specialization on specific plant resources. Bill traits had a stronger effect on specialization than body mass. Long–billed and curved–billed hummingbird species were particularly specialized, i.e., deviated strongly from a random interaction pattern that would be driven by the abundances of floral resources at a given elevation. Hence, species with long and curved bills were more specific in their resource choice and foraged preferably on relatively rare, but rewarding resources, whereas species with short and uncurved bills foraged mostly on the most abundant plant resources and thus followed a more random interaction pattern. Bill morphology in hummingbirds has long been known to be associated with the efficiency of resource use (Wolf et al. 1972, Temeles et al. 2009) and has been proposed to determine interaction patterns in plant–hummingbird assemblages (Feinsinger 1976, Brown and Bowers 1985). Hummingbird species with strongly curved bills reach nectar from curved flowers that straight–billed species are not able to access or only access with greater difficulty. Hence, interspecific competition for curved–billed hummingbird species is likely to be reduced (Linhart 1973, Stiles 1981). For instance, the long, curved bills of most hermit species enable them to reach nectar from flowers that short and uncurved billed species are not as easily able to access. Correspondingly, our results

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

indicate that bill morphology, in particular bill curvature, influence resource use and niche partitioning in hummingbird assemblages, which is likely to reduce competition for floral resources.

Body mass has also been associated with the foraging efficiency for nectar in hummingbirds (Hainsworth and Wolf 1972). However, this morphological trait is rather associated with the dominance relationships among co–occurring hummingbirds than to foraging efficiency and flower choice (Feinsinger 1976, Altshuler 2006). This may explain why body mass had a weaker influence on ecological specialization than bill morphology and suggests that niche partitioning of floral resources within hummingbird assemblages is mostly determined by variability in bill morphology.

Trait matching and foraging efficiency

We found that interaction strength in the networks was associated with the degree of trait matching in corresponding morphological traits of hummingbird and plant species. First, large hummingbird species preferred to feed on large flowers at mid and high elevations, which may be related to a high nectar production of these flowers (Ornelas et al. 2007, Rodríguez and Stiles 2005). This finding suggests that high energy requirements in the harsh environmental conditions at higher elevations may require large–bodied hummingbird species to specialize on floral resources with large nectar crops. Second, long–billed and curve–billed hummingbird species preferred plant species with long and curved flowers, respectively, indicating high degrees of trait complementarity between bill and corolla shape. The findings revealed by the novel and fully quantitative approach of fourth–corner analysis of interaction matrices (Dehling et al. 2014) is thus consistent with earlier studies showing high degrees of trait matching between hummingbirds and their foraging plants (Snow and Snow 1980, Dalsgaard et al. 2009, Temeles et al. 2009).

Our results reveal that hummingbirds spent more foraging time not only on flowers with high nectar volume but also on flowers that did not match well with their bill morphology. This is consistent with optimal foraging theory (MacArthur and Pianka 1966), which predicts that high trait matching should lead to an increased efficiency of

100

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

resource use, reflected in shorter handling times (Temeles 1996). Hence, a reduced cost in resource handling makes long–billed hummingbirds more efficient feeders on flowers with extended corollas. Consequently, a plant species that offers a greater profitability for pollinators, either through a high reward or a high efficiency in resource use, will be visited more frequently. The relationship between resource handling efficiency and trait matching is a plausible explanation for the close association between interaction strength and trait matching found in all networks. Network analyses combining trait and interaction data with a novel combination of analytical approaches may also be valuable for studying linkages between phenotypic and ecological specialization in other types of ecological networks.

Conclusions

Consistent with previous studies, we show that morphological traits, particularly avian bill morphology, shape the specialization of hummingbird species in plant– hummingbird networks. We present evidence that the close morphological matches between interacting plant and hummingbird species contribute to a high efficiency in hummingbirds’ resource use. Similar mechanisms of trait complementarity between interacting species, associated with a high efficiency in resource use, may also structure many other types of ecological networks. We conclude that network analysis of specialization and trait complementarity represents a powerful methodological approach that is likely to contribute to a better mechanistic understanding of the evolutionary and ecological causes of specialization in highly diversified species assemblages.

Acknowledgements

This work was funded by the following organizations: Consejo Nacional para Investigaciones Científicas y Tecnológicas and Ministerio de Ciencia y Tecnología, Universidad Estatal a Distancia, Organization for Tropical Studies (OTS), German Academic Exchange Service and Tropical Science Centre. Financial support for this

101

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

study was also provided by the research–funding programme ‘‘LOEWE–Landes– Offensive zur Entwicklung Wissenschaftlich–ökonomischer Exzellenz’’ of Hesse’s Ministry of Higher Education, Research, and the Arts. We are grateful to the field assistants and volunteers who contributed to data collection and botanists from the Instituto Nacional de Biodiversidad and La Selva Biological Station (OTS) for support with plant identification. We thank Bob O'Hara and Martina Stang for helpful discussions as well as Randall Mitchell, Gary Stiles and an anonymous reviewer for their constructive and insightful comments on an earlier version of this manuscript.

References

Altshuler, D. L. 2006. Flight performance and competitive displacement of hummingbirds across elevational gradients. The American Naturalist 167:216–229.

Armbruster, W. S. 2006. Evolutionary and ecological aspects of specialized pollination: views from the arctic to the tropics. Pages 260–282 in N. M. Waser and J. Ollerton, editors. Plant–pollinator interactions: from specialization to generalization. University of Chicago Press, Chicago, USA.

Blüthgen, N., F. Menzel, and N. Blüthgen. 2006. Measuring specialization in species interaction networks. BMC Ecology 6:9.

Blüthgen, N., F. Menzel, T. Hovestadt, D. Fiala, and N. Blüthgen. 2007. Specialization, constraints, and conflicting interests in mutualistic networks. Current Biology 17:341–346.

Blüthgen, N., J. Fründ, D. P. Vázquez, and F. Menzel. 2008. What do interaction network metrics tell us about specialization and biological traits? Ecology 89:3387– 3399.

Brown, J. H., and M. A. Bowers. 1985. Community organization in hummingbirds: relationships between morphology and ecology. The Auk 102:251–269.

Costigan, M. 2008. The pollination ecology of two sympatric populations of Heliconia in a high elevation rain forest of Honduras. Dissertation. The University of Dublin. Dublin, Ireland. 102

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Cotton, P. A. 1998. Coevolution in an Amazonian hummingbird–plant community. Ibis 140:639–646.

Dalsgaard, B., A. M. Martín Gonzáles, J. M. Olesen, J. Ollerton, A. Timmermann, et al. 2009. Plant–hummingbird interactions in the West Indies: floral specialization gradients associated with environment and hummingbird size. Oecologia 159:757– 766.

Dalsgaard, B., E. Magard, J. Fjeldsa, A. M. Martín González, C. Rahbek, et al. 2011. Specialization in plant–hummingbird networks is associated with species richness, contemporary precipitation and quaternary climate–change velocity. PLoS ONE 6 e25891.

Dehling, D. M., T. Töpfer, H. M. Schaefer, P. Jordano, K. Böhning-Gaese, and M. Schleuning. 2014. Functional relationships beyond species richness patterns: trait matching in plant–bird mutualisms across scales. Global Ecology and Biogeography, in press. DOI: 10.1111/geb.12193.

Dohzono, I., Y. Takami, and K. Suzuki. 2011. Is bumblebee foraging efficiency mediated by morphological correspondence to flowers? International Journal of Insect Science 3:1–10.

Dray, S., and P. Legendre. 2008. Testing the species traits–environment relationships: the fourth–corner problem revisited. Ecology 89:3400–3412.

Dupont, Y. L., D. M. Hansen, and J. M. Olesen. 2003. Structure of a plant–flower visitor network in the high–altitude sub–alpine desert of Tenerife, Canary islands. Ecography 26:301–310.

Faegri, K., and L. van der Pijl. 1979. The principles of pollination ecology. Pergamon Press, Oxford.

Feinsinger, P. 1976. Organization of a tropical guild of nectarivorous birds. Ecological Monographs 46:257–291.

Guimarães, P. R., P. Jordano, and J. N. Thompson. 2011. Evolution and coevolution in mutualistic networks. Ecology Letters 14:877–885.

103

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Hainsworth, F. R., and L. L. Wolf. 1972. Crop volume, nectar concentration and hummingbird energetics. Comparative Biochemistry and Physiology. 42A:359–366.

Hartshorn, G., and R. Peralta. 1988. Preliminary description of primary forests along the La Selva–Volcan Barva altitudinal transect, Costa Rica. Pages 281–295 in F. Almeda and C. M. Pringle, editors. Tropical rainforests: Diversity and conservation. California Academy of Sciences, San Francisco, USA.

Holdridge, L. R. 1967. Life zone ecology. Tropical Science Center. San José, Costa Rica.

Ings, T. C., J. M. Montoya, J. Bascompte, N. Blüthgen, L. Brown, et al. 2009. Ecological networks – beyond food webs. Journal of Animal Ecology 78:253–269.

Janzen, D. H. 1980. When is it coevolution? Evolution 34:611–612.

Janzen, D. H. 1985. On ecological fitting. Oikos 45:308–310.

Johnson, S. D., and K. E. Steiner. 2000. Generalization vs. specialization in plant– pollination systems. Trends in Ecology and Evolution 15:140–143.

Junker, R. R., N. Blüthgen, T. Brehm, J. Binkenstein, J. Paulus, et al. 2013. Specialization on traits as basis for the niche–breadth of flower visitors and as structuring mechanism of ecological networks. Functional Ecology 27:329–341.

Kershaw, F. L. 2006. The co–evolution of hummingbird bill morphology and floral structures with regard to both native and non–native flower species. Leeds University, Leeds. UK.

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

Linhart, Y. B. 1973. Ecological and behavioral determinants of pollen dispersal in hummingbird–pollinated Heliconia. The American Naturalist 107:511–523.

MacArthur, R. H. and E. R. Pianka. 1966. On the optimal use of a patchy environment. The American Naturalist 100:603–609.

Montgomerie, R. D. 1984. Nectar extraction by hummingbirds: response to different floral characters. Oecologia 63:229–236.

104

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Nakagawa, S., and H. Schielzeth. 2013. A general and simple method for obtaining R2 from generalized linear mixed–effects models. Methods in Ecology and Evolution 4:133–142.

Olesen, J. M., J. Bascompte, Y. L. Dupont, and P. Jordano. 2007. The modularity of pollination networks. Proceedings of the National Academy of Sciences of the United States of America 11:19891–19896.

Ollerton J., R. Alarcón, N. M Waser, M. V. Price, S. Watts, et al. 2009. A global test of the pollination syndrome hypothesis. Annals of Botany 103:1471–1480.

Patefield, W. M. 1981. An efficient method of generating random RxC tables with given row and column totals. Applied Statistics 30:91–97.

R Development Core Team. 2013. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna.

Ralph, C. J., G. R. Geupel, P. Pyle, T. Martin, and D. F. DeSante. 1993. Handbook of field methods for monitoring landbirds. Pacific Southwest Research Station Albany, California, USA.

Rodríguez C. I., and F. G. Stiles. 2005. Análisis ecomorfológico de una comunidad de colibríes ermitaños (Trochilidae, Phaethorninae) y sus flores en la amazonia colombiana. Ornitología Colombiana 3:7–27.

Santamaría, L., and M. A. Rodríguez-Gironés. 2007. Linkage rules for plant–pollinator networks: trait complementarity or exploitation barriers?. Plos Biology 5:354–362.

Schleuning, M., J. Fründ, A-M. Klein, R. A. Abrahamczyk, M. Albrecht, R. Alarcón, M. Albrecht, G. K. S. Andersson, S. Bazarian, Andy Katrin Böhning-Gaese. 2012. Specialization of mutualistic interaction networks decreases toward tropical latitudes. Current Biology 22:1–7.

105

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Snow, D. W., and B. K. Snow. 1980. Relationships between hummingbirds and flowers in the Andes of Colombia. Bulletin of the British Museum (Natural History). Zoology 38:105–139.

Stiles, F. G. 1975. Ecology, flowering phenology, and hummingbird pollination of some Costa Rican Heliconia species. Ecology 56:285–301.

Stiles, F. G. 1978. Ecological and evolutionary implications of bird pollination. American Zoologist 18:715–727.

Stiles, F. G. 1981. Geographical aspects of bird–flower coevolution, with particular reference to Central America. Annals of the Missouri Botanical Garden 68:323–351.

Stiles, F. G. and A. Skutch. 1989. A guide to the birds of Costa Rica. Comstock Publishing Associates. A division of Cornell University Press. Ithaca, USA.

Suarez, R. K. 1998. Oxygen and the upper limits to animal design and performance. Journal Experimental Biology 201:1065–1072.

TEAM 2013. Tropical Ecology Assessment and Monitoring Network. Volcán Barba, Costa Rica.

Temeles, E. J. 1996. A new dimension to hummingbird–flower relationships. Oecologia 105:517–523.

Temeles, E. J., and W. J. Kress. 2003. Adaptation in a plant–hummingbird association. Science 300:630–633.

Temeles, E. J., C. R. Koulouris, S. E. Sander, and W. J. Kress. 2009. Effect of flower shape and size on foraging performance and trade–offs in a tropical hummingbird. Ecology 90:1147–1161.

Ter Braak, C. J. F., A. Cormont, and S. Dray. 2012. Improved testing of species traits– environment relationships in the fourth–corner problem. Ecology 93:1525–1526.

106

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Thompson, J. N. 1994. The coevolutionary process. University of Chicago Press, Chicago, USA.

Vázquez, D. P. 2005. Degree distribution in plant–animal mutualistic networks: Forbidden links or random interactions. Oikos 108:21–426.

Waser, N. M., L. Chittka, M. V. Price, N. M. Williams, and J. Ollerton. 1996. Generalization in pollination systems, and why it matters. Ecology 77:1043–1060.

Wolf, L. L., F. R. Hainsworth, and F. G. Stiles. 1972. Energetics of foraging: rate and efficiency of nectar extraction y hummingbirds. Science 176:1351–1352.

Wolf, L. L., F. G. Stiles, and F. R. Hainsworth. 1976. Ecological organization of a tropical, highland hummingbird community. Journal of Animal Ecology 45:349–379.

Zuur, A., E. N. Ieno, N. Walker, A. A. Saveliev, and G. M. Smith, 2009. Mixed effects models and extensions in ecology with R. Springer, New York, USA.

Supplemental Material

Table A1: mean values of morphological traits of plant species. Table A2: mean values of morphological traits of hummingbird species. Table A3: hummingbird abundance and the number of observed legitimate visits of each hummingbird species to plant individuals. Table A4: list of plant species and the number of legitimate visits by hummingbirds to each plant species.

107

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Supplemental Material

Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

María Alejandra Maglianesi, Nico Blüthgen, Katrin Böhning–Gaese and Matthias Schleuning

Table A1. Mean values of morphological traits of plant species in three forest types at different elevations (50, 1.000 and 2.000 m a.s.l) in northeastern Costa Rica. Measure units are mm, degrees and cm3 for bill length, bill curvature and flower volume, respectively. Plant families are ordered alphabetically within each elevation.

Corolla Elevation Family Species Length Curvature Volume Low Acanthaceae Aphelandra storkii 75.00 18.30 2.12 Low Acanthaceae Bravaisia integérrima 21.70 0 8.25 Low Acanthaceae Odontonema cuspidatum 37.10 0 4.92 Low Acanthaceae Odontonema tubaeforme 30.00 5.71 0.50 Low Acanthaceae Razisea wilburii 32.40 17.15 0.41 Low Acanthaceae Ruellia metallica 25.00 0 7.62 Low Alstroemeriaceae Bomarea obovata 32.60 0 9.74 Low Apocynaceae Odontadenia sp. 33.00 0 7.49 Low Apocynaceae Stemmadenia sp. 36.50 0 11.47 Low Verbenaceae Stachytarpheta frantzii 17.00 0 3.42 Low Bignoniaceae Arrabidaea sp. 39.00 0 11.29 Low Bignoniaceae Arrabidaea verrucosa 56.00 1.02 45.04 Low Bromeliaceae Aechmea mariae–reginae 30.50 0 0.60 Low Bromeliaceae Aechmea nudicaulis 19.20 20.03 0.51 Low Bromeliaceae Guzmania monostachia 13.00 0 0.09 Low Convolvulaceae Ipomoea batatas 19.00 0 7.89 Low Costaceae Costus laevis 95.00 13.61 81.25 Low Costaceae Costus malortieanus 85.00 13.24 41.72 Low Costaceae Costus pulverulentus 73.50 13.39 51.95 Low Costaceae Costus scaber 39.70 20.7 1.12 Low Fabaceae Erythrina poeppigiana 32.80 11.71 0.45 108

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Table A1 (continued)

Corolla Elevation Family Species Length Curvature Volume Low Gesneriaceae Besleria columneoides 21.85 0 1.10 Low Gesneriaceae Chrysothemis friedrichsthaliana 24.10 0 2.92 Low Gesneriaceae Columnea nicaraguensis 69.00 6.61 146.54 Low Gesneriaceae Columnea purpurata 38.80 8.79 1.95 Low Gesneriaceae Drymonia microphylla 35.80 2.88 8.13 Low Heliconiaceae Heliconia imbricata 30.20 4.36 0.38 Low Heliconiaceae Heliconia latispatha 46.00 9.87 1.30 Low Heliconiaceae Heliconia mariae 33.00 22.25 0.55 Low Heliconiaceae Heliconia mathiasiae 48.00 20.03 1.36 Low Heliconiaceae Heliconia pogonantha 42.20 7.69 3.31 Low Heliconiaceae Heliconia wagneriana 62.00 15.76 0.82 Low Malvaceae Malvaviscus concinnus 54.75 0 40.28 Low Marantaceae Calathea crotalifera 38.00 8.24 8.63 Low Marantaceae Calathea gymnocarpa 54.00 7.39 7.17 Low Marantaceae Calathea lasiostachya 40.00 24.11 11.22 Low Marantaceae Calathea lutea 53.00 3.24 7.25 Low Marantaceae Calathea marantifolia 53.00 6.25 2.04 Low Marantaceae Pleiostachya pruinosa 45.50 0 3.57 Low Passifloraceae Passiflora vitifolia 2.00 0 11.35 Low Rubiaceae Faramea suerrensis 12.60 0 0.56 Low Rubiaceae Hamelia patens 25.00 0 0.96 Low Rubiaceae Notopleura capitata 7.00 0 0.20 Low Rubiaceae Palicourea guianensis 15.00 0 0.38 Low Rubiaceae Pentagonia monocaulis 47.05 0 20.45 Low Rubiaceae Psychotria chiapensis 59.00 0 11.86 Low Rubiaceae Psychotria elata 20.90 0 0.80 Low Rubiaceae Psychotria poeppigiana 23.30 0 0.64 Low Rubiaceae Warscewiczia coccinea 8.20 0 0.07 Low Schlegeliaceae Schlegelia fastigiata 32.60 17.85 1.18 Low Schlegeliaceae Schlegelia nicaraguensis 35.00 8.13 14.54 Low Zingiberacae Renealmia cernua 23.00 2.24 0.45 Mid Apocynaceae Allomarkgrafia brenesiana 60.00 0 169.65 Mid Apocynaceae Tabernaemontana alfaroi 39.00 0 18.39 Mid Bromeliaceae Guzmania nicaraguensis 58.50 0 2.25

109

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Table A1 (continued)

Corolla Elevation Family Species Length Curvature Volume Mid Bromeliaceae Pitcairnia valerioi 20.70 0 0.42 Mid Campanulaceae Centropogon granulosus 51.00 18.43 4.76 Mid Convolvulaceae Maripa nicaraguensis 56.00 0 22.66 Mid Costaceae Costus curvibracteatus 54.50 0 3.86 Mid Costaceae Costus pulverulentus 72.00 13.66 50.89 Mid Ericaceae Cavendishia callista 34.50 0 0.63 Mid Ericaceae Cavendishia complectens 14.20 0 0.28 Mid Ericaceae Macleania rupestris 28.30 0 0.62 Mid Ericaceae Psammisia ramiflora 35.00 0 3.96 Mid Ericaceae Satyria warszewiczii 29.25 4.85 0.85 Mid Ericaceae Sphyrospermum dissimile 10.00 0 0.87 Mid Ericaceae Thibaudia costaricensis 11.00 0 0.27 Mid Gentianaceae Macrocarpaea valerioi 39.00 0 28.49 Mid Gesneriaceae Besleria notabilis 20.80 0 1.57 Mid Gesneriaceae Besleria triflora 17.00 0 0.62 Mid Gesneriaceae Columnea microphylla 63.80 23.70 39.85 Mid Gesneriaceae Columnea purpurata 39.00 8.75 1.96 Mid Gesneriaceae Columnea querceti 63.90 11.56 7.98 Mid Gesneriaceae Drymonia warszewicziana 35.00 0 5.01 Mid Heliconiaceae Heliconia atropurpurea 60.00 14.04 3.02 Mid Heliconiaceae Heliconia rodriguezii 55.00 25.98 1.56 Mid Heliconiaceae Heliconia vaginalis 60.75 13.97 9.90 Mid Lamiaceae Scutellaria costaricana 56.00 19.65 3.80 Mid Marantaceae Calathea crotalifera 28.60 4 3.80 Mid Marantaceae Calathea lasiostachya 45.90 13.48 7.07 Mid Marantaceae Calathea recurvata 48.50 2.36 25.75 Mid Orchidaceae Elleanthus hymenophorus 11.90 0 0.34 Mid Rubiaceae Hillia triflora 59.00 19.59 7.59 Mid Rubiaceae Palicourea sp.1 11.90 0 0.50 Mid Rubiaceae Palicourea lasiorrhachis 17.00 0 0.45 Mid Rubiaceae Psychotria carthagenensis 9.80 0 0.38 Mid Rubiaceae Psychotria elata 20.00 0 0.53 Mid Rubiaceae Sabicea panamensis 11.50 0 0.44 Mid Solanaceae Merinthopodium neuranthum 45.00 0 44.29

110

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Table A1 (continued)

Corolla Elevation Family Species Length Curvature Volume Mid Solanaceae Schultesianthus venosus 42.00 0 16.55 Mid Zingiberacea Renealmia cernua 19.50 0 0.38 Mid Zingiberacea Renealmia ligulata 48.50 7.05 4.69 High Actinidiaceae Saurauia montana 12.00 0 1.36 High Alstroemeriaceae Bomarea hirsuta 51.00 0 7.85 High Asteraceae Neomirandea eximia 5.00 0 0.04 High Orchidaceae Fernandezia tica 7.00 0 0.04 High Orchidaceae Maxillaria sp. 13.80 0 1.29 High Bromeliaceae Tillandsia insignis 18.40 0 0.17 High Bromeliaceae Werauhia camptoclada 60.00 6.47 2.31 High Bromeliaceae Werauhia ororiensis 36.00 31.43 4.78 High Campanulaceae Burmeistera parviflora 18.50 0 1.45 High Campanulaceae Centropogon solanifolius 50.50 12.85 4.50 High Ericaceae Cavendishia bracteata 30.00 9.46 0.85 High Ericaceae Disterigma humboldtii 10.35 0 0.08 High Ericaceae Gaultheria gracilis 11.30 0 0.18 High Ericaceae Gonocalyx pterocarpus 20.67 5.37 0.17 High Ericaceae Satyria warszewiczii 26.50 5.6 0.30 High Ericaceae Sphyrospermum cordifolium 8.40 0 0.01 High Ericaceae Vaccinium poasanum 12.50 0 0.98 High Gentianaceae Macrocarpaea valerioi 48.00 0 46.18 High Gesneriaceae Besleria barbensis 27.20 0 4.31 High Gesneriaceae Besleria solanoides 21.50 0 4.32 High Gesneriaceae Columnea magnifica 74.00 16.56 36.32 High Gesneriaceae Kohleria tigridia 64.60 7.93 73.26 High Heliconiaceae Heliconia lankesteri 50.00 21.80 3.93 High Lamiaceae Aegiphila odontophylla 10.00 0 2.69 High Lamiaceae Scutellaria isocheila 32.00 25.11 0.90 High Malvaceae Malvaviscus palmanus 56.20 0 54.07 High Melastomataceae Meriania phlomoides 22.00 0 3.15 High Orchidaceae Elleanthus aurantiacus 12.20 0 0.34 High Orchidaceae Elleanthus sp. 11.10 0 0.49 High Orchidaceae Epidendrum lacustre 24.50 0 3.51 High Orchidaceae Epidendrum microrigidiflorum 19.00 0 6.27

111

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Table A1 (continued)

Corolla Elevation Family Species Length Curvature Volume High Orchidaceae Epidendrum polychlamys 27.00 0 15.46 High Orchidaceae Scaphyglottis sigmoidea 14.20 0 1.35 High Orchidaceae Sobralia amabilis 33.00 0 10.37 High Orquidiaceae Elleanthus tricallosus 8.50 0 0.33 High Rubiaceae Gonzalagunia rosea 12.75 0 0.36 High Rubiaceae Guettarda crispiflora 21.50 0 2.16 High Rubiaceae Hillia maxonii 52.40 0 16.46 High Rubiaceae Hoffmannia arborescens 13.50 0 0.17 High Rubiaceae Palicourea lasiorrhachis 13.20 0 0.26 High Solanaceae Cestrum poasanum 24.00 0 1.88

112

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Table A2. Mean values of morphological traits of hummingbird species captured in three forest types at different elevations (50, 1.000 and 2.000 m a.s.l) in northeastern Costa Rica. Measure units are mm, degrees and g for bill length, bill curvature and body mass, respectively. Upward bill curvature is indicated with a minus sign. Hummingbird species are ordered alphabetically within each elevation.

Elevation Species Bill length Bill curvature Body mass Low Amazilia amabilis 18.33 3.57 3.90 Low Amazilia tzacatl 20.14 5.00 5.37 Low Chalybura urochrysia 22.48 4.13 7.15 Low Florisuga mellivora 17.20 7.12 6.75 Low Glaucis aeneus 28.90 16.02 5.43 Low Klais guimeti 13.10 3.20 3.00 Low Microchera albocoronata 19.80 1.16 3.00 Low Phaeochroa cuvierii 21.00 3.00 7.90 Low Phaethornis longirostris 34.50 18.05 6.22 Low Phaethornis striigularis 21.55 6.56 2.83 Low Thalurania colombica 19.39 5.05 4.65 Low Threnetes ruckeri 28.11 9.67 6.09 Mid Amazilia tzacatl 22.40 6.12 4.70 Mid Campylopterus hemileucurus 33.40 12.66 10.40 Mid Discosura conversii 12.80 2.45 3.25 Mid Doryfera ludovicae 34.83 -0.78 5.39 Mid Elvira cupreiceps 14.02 8.87 3.28 Mid Eupherusa nigriventris 14.60 3.95 3.47 Mid Eutoxeres aquila 18.20 48.70 10.23 Mid Heliodoxa jacula 21.67 4.17 8.26 Mid Lampornis calolaemus 20.05 2.87 6.15 Mid Lampornis hemileucus 19.24 3.66 5.83 Mid Phaethornis guy 39.48 15.85 6.20 Mid Phaethornis striigularis 21.01 8.11 2.65 Mid Thalurania colombica 19.78 5.80 4.73 High Campylopterus hemileucurus 30.89 16.95 9.07 High Colibri thalassinus 21.17 8.21 4.85 High Doryfera ludovicae 34.85 -0.92 5.75 High Eugenes fulgens 35.27 4.88 9.04 High Eupherusa nigriventris 13.80 5.38 3.30

113

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Table A2 (continued)

Elevation Species Bill length Bill curvature Body mass High Heliodoxa jacula 21.89 4.83 8.61 High Hylocharis eliciae 16.70 1.71 4.90 High Lampornis calolaemus 20.82 3.41 5.44 High Panterpe insignis 19.34 3.30 5.62 High Phaethornis guy 38.90 17.14 6.03 High Selasphorus flammula 10.85 4.10 2.36 High Selasphorus scintilla 11.56 3.99 2.20 High Thalurania colombica 19.10 5.38 4.50

114

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Table A3. List of hummingbird species, their relative abundance (number of individuals captured in mist nets, excluding recaptures) and their number of observed legitimate visits across all plant species. The study area consisted of three forest types at different elevations in northeastern Costa Rica. Hummingbird species are ordered alphabetically within each elevation.

Elevation Species Abundance N° of visits Low Amazilia amabilis 9 - Low Amazilia tzacatl 22 120 Low Chalybura urochrysia 9 - Low Florisuga mellivora 2 1 Low Glaucis aeneus 5 - Low Klais guimeti 3 9 Low Microchera albocoronata 1 - Low Phaeochroa cuvierii 1 1 Low Phaethornis longirostris 47 23 Low Phaethornis striigularis 21 75 Low Thalurania colombica 53 80 Low Threnetes ruckeri 31 1 Mid Amazilia tzacatl 1 - Mid Campylopterus hemileucurus 2 - Mid Discosura conversii 2 - Mid Doryfera ludovicae 11 17 Mid Elvira cupreiceps 6 - Mid Eupherusa nigriventris 45 38 Mid Eutoxeres aquila 7 1 Mid Heliodoxa jacula 33 1 Mid Lampornis calolaemus 1 2 Mid Lampornis hemileucus 32 53 Mid Phaethornis guy 119 110 Mid Phaethornis striigularis 13 4 Mid Thalurania colombica 11 - High Campylopterus hemileucurus 13 11 High Colibri thalassinus 7 - High Doryfera ludovicae 13 7 High Eugenes fulgens 10 2

115

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Table A3 (continued)

Elevation Species Abundance N° of visits High Eupherusa nigriventris 1 - High Heliodoxa jacula 17 4 High Hylocharis eliciae 2 - High Lampornis calolaemus 153 178 High Panterpe insignis 13 38 High Phaethornis guy 3 14 High Selasphorus flammula 17 30 High Selasphorus scintilla 24 3 High Thalurania colombica 2 -

116

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Table A4. List of plant species, their flower abundance (total number of flowers per plant species, summed across transects and sampling periods) and the number of legitimate visits by hummingbirds to each plant species in three forest types at different elevations in northeastern Costa Rica. Plant families are ordered alphabetically within each elevation.

Elevation Flower Nº of visits Family Species abundance Low Bromeliaceae Aechmea mariae–reginae 10 36 Low Acanthaceae Aphelandra storkii 2 3 Low Acanthaceae Bravaisia integerrima 50 - Low Bromeliaceae Guzmania monostachia 5 2 Low Costaceae Costus laevis 4 - Low Costaceae Costus malortieanus 7 - Low Costaceae Costus pulverulentus 1 1 Low Costaceae Costus scaber 4 1 Low Gesneriaceae Besleria columneoides 5 1 Low Gesneriaceae Columnea nicaraguensis 6 - Low Heliconiaceae Heliconia imbricata 731 56 Low Heliconiaceae Heliconia latispatha 17 19 Low Heliconiaceae Heliconia mariae 396 53 Low Heliconiaceae Heliconia mathiasiae 5 4 Low Heliconiaceae Heliconia pogonantha 5 1 Low Heliconiaceae Heliconia wagneriana 8 2 Low Malvaceae Malvaviscus concinnus 3 - Low Marantaceae Calathea gymnocarpa 47 1 Low Marantaceae Calathea inocephala 75 6 Low Marantaceae Calathea lutea 20 1 Low Marantaceae Calathea marantifolia 390 - Low Passifloraceae Passiflora vitifolia 3 1 Low Rubiaceae Faramea suerrensis 6 - Low Rubiaceae Hamelia patens 711 60 Low Rubiaceae Notopleura capitata 92 - Low Rubiaceae Palicourea guianensis 5 - Low Rubiaceae Pentagonia monocaulis 6 10 Low Rubiaceae Psychotria chiapensis 125 -

117

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Table A4 (continued)

Elevation Flower Nº of visits Family Species abundance Low Rubiaceae Psychotria elata 21 1 Low Rubiaceae Psychotria marginata 83 - Low Rubiaceae Psychotria poeppigiana 2 8 Low Rubiaceae Psychotria suerrensis 300 - Low Rubiaceae Warscewiczia coccinea 15 1 Low Schlegeliaceae Schlegelia fastigiata 12 8 Low Zingiberacea Renealmia cernua 24 53 Mid Apocynaceae Allomarkgrafia brenesiana 2 - Mid Apocynaceae Tabernaemontana alfaroi 3 - Mid Bromeliaceae Guzmania nicaraguensis 5 3 Mid Bromeliaceae Guzmania scandens 4 - Mid Bromeliaceae Pitcairnia brittoniana 5 7 Mid Bromeliaceae Pitcairnia valerioi 205 - Mid Campanulaceae Centropogon granulosus 5 1 Mid Costaceae Costus curvibracteatus 4 7 Mid Costaceae Costus pulverulentus 6 25 Mid Cucurbitaceae Gurania coccinea 8 6 Mid Ericaceae Cavendishia callista 707 11 Mid Ericaceae Cavendishia complectens 16 6 Mid Ericaceae Cavendishia querema 7 2 Mid Ericaceae Psammisia ramiflora 58 4 Mid Ericaceae Satyria warszewiczii 142 24 Mid Ericaceae Thibaudia costaricensis 1451 14 Mid Gesneriaceae Besleria notabilis 67 1 Mid Gesneriaceae Besleria triflora 3 - Mid Gesneriaceae Columnea gloriosa 53 - Mid Gesneriaceae Columnea magnifica 5 - Mid Gesneriaceae Columnea microcalyx 3 - Mid Gesneriaceae Columnea microphylla 13 - Mid Gesneriaceae Columnea purpurata 6 1 Mid Gesneriaceae Columnea querceti 6 1 Mid Gesneriaceae Drymonia conchocalyx 1 3 Mid Gesneriaceae Drymonia tomentulifera 1 - Mid Gesneriaceae Drymonia warszewicziana 6 - Mid Gesneriaceae Gasteranthus wendlandianus 1 -

118

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Table A4 (continued)

Elevation Flower Nº of visits Family Species abundance Mid Heliconiaceae Heliconia atropurpurea 1 38 Mid Heliconiaceae Heliconia rodriguezii 1 - Mid Heliconiaceae Heliconia vaginalis 4 26 Mid Malvaceae Malvaviscus sp. 1 - Mid Marantaceae Calathea crotalifera 60 - Mid Marantaceae Calathea guzmanioides 1 - Mid Marantaceae Calathea lasiostachya - 1 Mid Marantaceae Calathea recurvata 31 - Mid Rubiaceae Faramea eurycarpa 94 1 Mid Rubiaceae Hillia triflora 1 2 Mid Rubiaceae Palicourea sp. 1 112 5 Mid Rubiaceae Palicourea lasiorrhachis 12 - Mid Rubiaceae Palicourea sp. 2 5 5 Mid Rubiaceae Psychotria carthagenensis 6 - Mid Rubiaceae Psychotria elata 357 36 Mid Rubiaceae Psychotria brachiata 38 - Mid Zingiberacea Renealmia cernua 6 6 Mid Zingiberacea Renealmia ligulata 9 - High Alstroemeriaceae Bomarea hirsuta 25 11 High Asteraceae Neomirandea eximia 560 1 High Bromeliaceae Werauhia camptoclada 10 - High Bromeliaceae Werauhia ororiensis 4 5 High Campanulaceae Burmeistera parviflora 22 1 High Campanulaceae Centropogon solanifolius 170 1 High Ericaceae Cavendishia bracteata 269 117 High Ericaceae Disterigma humboldtii 214 36 High Ericaceae Gaultheria gracilis 1889 10 High Ericaceae Gonocalyx pterocarpus 42 6 High Ericaceae Satyria warszewiczii 22 - High Ericaceae Sphyrospermum cordifolium 2 - High Ericaceae Vaccinium poasanum 10 11 High Gentianaceae Macrocarpaea valerioi 5 12 High Gesneriaceae Besleria barbensis 12 - High Gesneriaceae Besleria solanoides 38 4 High Gesneriaceae Columnea magnifica 153 3

119

Appendix 2: Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics

Table A4 (continued)

Elevation Flower Nº of visits Family Species abundance High Gesneriaceae Columnea microcalyx 30 1 High Gesneriaceae Kohleria tigridia 5 3 High Heliconiaceae Heliconia lankesteri 46 21 High Lamiaceae Scutellaria isocheila 8 - High Malvaceae Malvaviscus palmanus 13 2 High Orchidaceae Elleanthus aurantiacus 324 7 High Orchidaceae Elleanthus sp. 67 - High Orchidaceae Epidendrum polychlamys 233 - High Orchidaceae Fernandezia tica 2 - High Orchidaceae Maxillaria sp. 12 - High Orchidaceae Sobralia amabilis 5 - High Rubiaceae Gonzalagunia rosea 43 - High Rubiaceae Guettarda crispiflora 107 - High Rubiaceae Hillia maxonii 470 - High Rubiaceae Hoffmannia arborescens 14 3 High Rubiaceae Hoffmannia psychotriaefolia 50 - High Rubiaceae Palicourea lasiorrhachis 1534 66 High Solanaceae Cestrum poasanum 603 -

120

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Status: conditionally accepted in Journal of Animal Ecology subject to minor revision (August 30, 2014).

Maglianesi, M. A., Katrin Böhning-Gaese and M. Schleuning. Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions.

121

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Erklärung über Anteile der Autoren/Autorinnen

(1) Entwicklung und Planung

MM 40%; KBG 20%; MS 40%

(2) Durchführung der einzelnen Untersuchungen / Experimente

MM compiled all data and coordinated the fieldwork.

(3) Erstellung der Datensammlung und Abbildung

MM conducted data analyses and created all figures; figures were discussed with NB, KB and MS.

(4) Analyse/Interpretation der Daten

MM 70%; KBG 10%; MS 20%

(5) Übergeordnete Einleitung / Ergebnisse / Diskussion

MM 70%; KBG 10%; MS20%

122

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

María Alejandra Maglianesi1,2, Katrin Böhning-Gaese1,3 and Matthias Schleuning1

1 Biodiversity and Climate Research Centre (BiK–F) and Senckenberg Gesellschaft für Naturforschung, 60325 Frankfurt am Main, Germany

2 Vicerrectoría de Investigación. Universidad Estatal a Distancia (UNED), San José, Costa Rica

3 Department of Biological Sciences, Johann Wolfgang Goethe University Frankfurt, 60438 Frankfurt am Main, Germany

Running title: Foraging preferences of hummingbirds

Type of manuscript: Article

*Corresponding author: María Alejandra Maglianesi. Vicerrectoría de Investigación. Universidad Estatal a Distancia (UNED), San Pedro Montes de Oca, San José, Costa Rica. E-mail: [email protected]

123

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Summary

In plant–pollinator networks, the floral morphology of food plants is an important determinant of the interaction niche of pollinators. Studies on foraging preferences of pollinators combining experimental and observational approaches may help to understand the mechanisms behind patterns of interactions and niche partitioning within pollinator communities. In this study, we tested whether morphological floral traits were associated with foraging preferences of hummingbirds for artificial and natural flower types in Costa Rica. We performed field experiments with artificial feeders, differing in length and curvature of flower types, to quantify the hummingbirds' interaction niche under unlimited nectar resources. To quantify the interaction niche under real–world conditions of limited nectar resources, we measured foraging preferences of hummingbirds for plant species. Artificial feeders were visited by Eupherusa nigriventris and Phaethornis guy in the pre–montane forest, and Lampornis calolaemus in the lower montane forest. Under experimental conditions, all three hummingbird species overlapped their interaction niches and showed a preference for the short artificial flower type over the long–straight and the long–curved flower types. Under natural conditions, the two co–occurring hummingbird species preferred to feed on plant species with floral traits corresponding to their bill morphology. The short–billed hummingbird E. nigriventris preferred to feed on short and straight flowers, whereas the long– and curved–billed P. guy preferred long and curved natural flowers. In contrast to experimental conditions, L. calolaemus did not show any preferences for specific flower types under natural conditions. Our results show that floral morphological traits constrain short– billed hummingbird species to access nectar resources. Morphological constraints, therefore, represent one important mechanism structuring trophic networks. In addition, other factors, such as competition or differences in resource quantity, define the interaction niches of consumer species in real–world communities, enforcing patterns of niche segregation between co–occurring consumer species. This suggests that experimental studies are needed to disentangle effects of morphological constraints from those of competition and/or resource quantity in plant–pollinator interactions and other types of trophic interactions.

124

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Key–words: artificial feeders, biotic interactions, Costa Rica, field experiments, floral traits, floral morphology, foraging preferences, hummingbirds, interaction niche, morphological constraints, resource partitioning, tropical forest.

Introduction

Plant and animal species are embedded in complex ecological networks and are linked by trophic interactions (Bascompte and Jordano 2007). In such trophic networks, plant species provide food resources for the dependent consumer guild, whereby plants represent an important niche dimension for the consumer species (Blüthgen 2010). This idea is consistent with the traditional niche concept proposed by Hutchinson (1957), who described the fundamental niche of a species as a multi– dimensional hypervolume, in which the dimensions are the resources that define the requirements of a species. In the presence of other species, however, this fundamental niche may be reduced to a smaller realized niche because of ecological constraints such as competition for resources (see Colwell and Fuentes 1975).

In plant–pollinator networks, the interaction niche of the pollinators is commonly defined by the range of plant species that are visited (i.e. the interaction or trophic niche; cf. Devictor et al. 2010). Previous research has demonstrated that floral morphology largely influences patterns of interactions of pollinators by constraining the number and strength of interactions (Stiles 1975, Stang et al. 2006, Vizentin– Bugoni et al. 2014, Maglianesi et al. 2014). For instance, specific floral morphologies may facilitate or hinder access to nectar rewards (Ornelas et al. 2007, Maglianesi et al. 2014), leading to preferences of pollinators for particular plant species. Thus, floral traits may be an important determinant of the interaction niche of pollinators. The full range of plants that pollinators are potentially able to use (i.e. potential interactions) may be restricted by additional factors (e.g. interspecific competition with other pollinators) to the subset of plant species they actually use (i.e. realized interactions) (Pauw 2013).

Hummingbirds (Trochilidae) are important pollinators in the Neotropics (Stiles 1981). Hermit hummingbirds have mostly long and curved bills and are specialized on

125

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions specific plant species (Linhart 1973, Stiles 1978), whereas non–hermit hummingbirds exhibit a wider range of bill morphologies and degrees of specialization (Cotton 1998, Stiles 2004). Morphological fitting between pairs of corresponding traits in plant and hummingbird species influences foraging preferences by hummingbirds (Colwell 1973, Temeles et al. 2010). Particularly close morphological matches, such as those observed between Heliconia flowers and hermit hummingbirds, may impose strong constraints in these systems (Stiles 1975, Feinsinger 1976). These morphological constraints may be an important factor defining the potential interactions of hummingbirds with plant species. Furthermore, direct competition for nectar resources has been found to be intense in hummingbird assemblages, especially at low resource availability (Colwell 1973, Brown and Bowers 1985). It is likely that competition plays an important role in determining foraging preferences of hummingbirds and therefore in defining their realized interactions.

A successful approach to quantify potential and realized interactions between plants and pollinators is to combine experimental and observational data. Potential interactions can be derived from controlled experiments in which pollinators' foraging preferences are measured across a range of resource types (e.g. different flower types) (Devictor et al. 2010). Experiments designed to detect foraging preferences of consumers require offering a selection of food types to individual consumers. In order to give consumers the opportunity to express a foraging choice, multiple resource types should be offered simultaneously (Peterson and Renaud 1989). In contrast, the realized interactions of pollinators can be estimated from observations of plant– pollinator interactions in the real world. From these interactions, the range of plant species visited by pollinators in natural conditions can be determined (Benadi et al. 2013). We are not aware of studies combining experimental and observational approaches within natural communities to study foraging preferences of pollinator species.

Here, we measured foraging preferences of hummingbird species by quantitative comparisons of their flower choices under experimental and natural conditions. We performed field experiments with artificial feeders (differing in length and curvature of artificial flower types) in order to quantify the interaction niche of hummingbird

126

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions species with unlimited nectar resources. We also measured foraging preferences of hummingbirds for plant species in natural conditions, to be able to quantify their interaction niche with limited nectar resources. We address the following questions: (i) Do hummingbird species prefer to feed on specific flowers types under experimental conditions? (ii) Do hummingbird species prefer to feed on flowers with specific morphological traits under natural conditions? (iii) Do hummingbirds’ foraging preferences under unlimited nectar resources differ from those in the real world?

Materials and methods

Study area and data collection

The study was conducted in northeastern Costa Rica within the Braulio Carrillo National Park on the Caribbean slope of the Cordillera Central. The park encompasses an area of ca 45.000 ha, of which about 67% is old–growth forest. Our study included two sites, placed in two different tropical forest types of different elevations: pre– montane forest (1000 m; 10°16’N, 84°05’W) and lower montane wet forest (2.000 m; 10°11’N, 84°07’W) (Holdridge 1967). Canopy heights were approximately 30 m at 1.000 m and 20 m at 2.000 m (Hartshorn and Peralta 1988). Rainfall reported for elevations similar to our study sites in the national park are 3.200 and 2.200 mm for the pre–montane and lower montane forest, respectively (Hartshorn and Peralta 1988). Mean annual temperatures recorded during the last five years are 20 and 14°C, respectively (Blake and Loiselle 2000, TEAM 2013). The dry season lasts from late December to April, and the wet season reaches a peak during July and October– November in both study sites.

The study was conducted from May 2011 to April 2012, covering an entire study year. During the year of sampling, we collected data on hummingbird abundances and bill traits, morphological floral traits and plant–hummingbird interactions across seven sampling periods per site, each lasting for about 10 days. We conducted field experiments with artificial feeders for a subset of two sampling periods per forest type, covering both seasons. Experiments were carried out during five days in August (wet season) and four days in February–March (dry season).

127

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Flower choice on experimental feeders

To assess hummingbird preferences for different floral morphologies under experimental conditions, we set up self–built feeders with different artificial flower types differing in length and curvature (Fig. 1A). Each feeder consisted of a 350 ml water bottle attached to a small container with three tubes representing one of the following flower types: short (10 mm), long–straight (40 mm) and long–curved (40 mm). These three flower types represented floral trait values from the natural communities. For the long–curved type (30º of deflection), the location of the bend was one–third the distance from the base of the flower, approximating the shape of natural flowers frequently visited by hummingbirds (e.g. Heliconia flowers) (Temeles et al. 2009). The tubes of all flower types had the same internal diameter (4 mm). In order to make feeders more attractive to hummingbirds, we painted them bright red because previous studies showed that hummingbirds prefer red over other colors (Grant and Grant 1968). Feeders were filled with the same amount of 25% sucrose solution (mass:mass) that represents the common concentration and dominant sugar found in hummingbird–visited flowers (McDade 2004, Rodríguez and Stiles 2005, Chalcoff et al. 2006). We established four feeding stations at each study site separated by at least 200 m. We installed feeders one month before the experiment started to allow hummingbirds to get used to them. At each of these stations, an array of three artificial feeders, each representing one flower type, was placed about 1.8 m above the ground. The feeders were spaced ca. 25 cm from one to another, that is, they were close enough to offer hummingbirds a choice between the different artificial flower types.

We refilled the feeders with new sucrose solution each day of the experiment to avoid fungal infection that could affect hummingbirds' health. We were careful to ensure that the feeders were permanently filled with artificial nectar, so that hummingbirds had a much larger amount of nectar resources available than they utilised at a given site. To discard the possibility that hummingbirds choose feeders for their particular spatial position, feeders were rotated within the array on different days of the experiment. Thus, the hummingbird choice for different flower types was enforced on every day of the experiment.

128

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

To record visitation frequency by hummingbirds to artificial feeders, we fixed unattended cameras about 10 m from the feeders for a 6–hours period between 06:00 and 14:00. From the videotapes, we identified the hummingbird species and recorded the flower type used by each individual at each single visit. A visit was scored whenever a hummingbird was observed to probe an artificial flower of the feeders. We defined the total number of visits per hummingbird species on each flower type as interaction frequency. We constructed one interaction matrix for each study site, feeding station and season (pooled across the observation days within a season). Overall, we recorded 6.974 visits of four hummingbird species over 1.800 hours of videotapes. We excluded visits from Heliodoxa jacula from further analyses because we had too few observations of interactions between this species and plant species under natural conditions (four interactions in the two forest types).

Flower choice in natural plant communities

To record the interactions between hummingbird and plant species, we carried out observations of flowering plants in the understory (up to 10 m above the ground). We chose plant species fitting the traditional ornithophilous pollination syndrome (Faegri and van der Pijl 1979). As hummingbird–pollinated flowers do not always fit into this syndrome (Ollerton et al. 2009), we also considered plant species fitting other pollination syndromes (e.g. bat– or insect–pollinated flowers) that were likely to be visited by hummingbirds. We randomly chose 4–12 plant individuals per species at each study site and sampling period. We videotaped 8–12 individuals of the more abundant plant species, and 4–6 individuals of the less abundant species. For recording visits of hummingbirds to plant individuals, we fixed unattended cameras (Sony DCR– HC51) about 10 m from open flowers for periods of 120 minutes between 06:00 and 14:00 (see Robertson et al. 1999, Maglianesi et al. 2014). We recorded 1.073 flowering plant individuals of 65 species and over 2.000 hours of videotapes. We recorded the hummingbird species identity of each visit to natural flowers. A visit was scored whenever a hummingbird was observed to probe at least one flower of the observed plant individual.

129

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Since we were interested in investigating patterns of hummingbird preferences for plant species with specific morphological floral traits, we excluded all illegitimate visits in which the hummingbird did not access the flower through the corolla entrance. For further analysis, we only considered visitation data of the hummingbird species that were observed to visit artificial feeders. Interaction frequency was defined as the total number of legitimate visits of each hummingbird species on each plant species. We constructed one interaction matrix for each study site and season.

Floral traits

To be able to quantify hummingbird preferences, we measured for all videotaped plant species two floral traits: length and curvature of the floral corolla. We do not exclude the possibility that other, unmeasured traits may contribute to hummingbird preferences (see Temeles et al. 2002), but the selected traits have been reported as the primary floral constraints determining nectar accessibility for hummingbird species (see Temeles et al. 2009, Vizentin–Bugoni et al. 2014, Maglianesi et al. 2014). Length of the corolla (from the base to the tip) was measured to the nearest 0.10 mm with a dial caliper. To measure corolla curvature, the flower was placed on 1 mm graph paper so that the straight part of the corolla near the base was in line with the first 5 mm and digital photographs were taken (Kershaw 2006). From these photographs, we measured the deflection of the tip of the corolla from the line through the base of the corolla to the nearest 0.10 mm. To determine the angle of deflection, we divided the deflection by the corolla length and calculated the angle of deflection by using the sine rule. Corolla curvature was arcsin–sqrt–transformated for statistical analysis. For both morphological traits, we calculated mean values from 2–4 individual flowers for 88 plant species.

Hummingbird abundance and traits

To evaluate the abundance of the studied hummingbird species in the natural communities, we placed 12–14 standard mist nets (12 x 3 m) at each study site for approximately 6 h after dawn (Ralph et al. 1993). Mist nets were operated 4 days in each of the seven sampling periods. Overall, our sampling effort was about 33.160 mist–net hours (one standard mist–net operated for one hour is a net–hour); sampling

130

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions effort was very similar for the two study sites. We used the number of hummingbird individuals captured per species, summed across sampling periods, as an estimate of hummingbird abundance per species at each site. All hummingbirds captured were identified to species level (according to Stiles and Skutch 1989) and banded with aluminum numbered bands. To avoid overestimation of hummingbird abundance, we excluded recaptured hummingbird individuals from abundance estimates. We did not employ standard capture–recapture approaches for estimating population sizes because only 16% of all captured hummingbirds were recaptures and sample sizes were too small for the studied species.

We measured two morphological traits that have been reported to influence plant– hummingbird interactions: bill length and curvature (Hainsworth and Wolf 1972, Temeles et al. 2009). Bill length (exposed culmen) was measured to the nearest 0.10 mm using dial calipers. To measure bill curvature, we placed the bill on graph paper following the same procedure as for corolla curvature (Kershaw 2006). Bill curvature was arcsin–sqrt–transformed for statistical analyses.

Statistical analysis

To investigate patterns of flower choice by pollinator species, we used a preference index (PI) to measure the preference of a hummingbird species for a specific plant species (William 2005, Fründ et al. 2010). The same PI was used for the experimental data and the natural visits. We applied the formula PIk = pobs,k / (pobs,k + pnull,k ) where pobs,k is the proportion of visits to plant species k among all visits by a hummingbird species, and pnull is the the expected proportion of visits to a particular plant species.

We defined pnull as the proportion of observation time of plant species k among the total observation time for all plant species in a given study site and season. We thus assumed that under a random interaction scenario the visitation frequency of a hummingbird species to a particular plant species is proportional to the observation time dedicated to this plant species (Fründ et al. 2010). For completely opportunistic interactions (null hypothesis), the focal hummingbird species would visit the focal flower species with a probability of pnull. PI ranges between 0 and 1, being 0 for

131

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions unvisited plant species, 0.5 for plant species visited as frequently as expected under the null hypothesis, and approaches 1 for strong preferences for a particular plant species (Fründ et al. 2010). We calculated PI for each hummingbird species and each artificial flower type as well as for each plant species observed in the natural communities. Since PI ranges between 0 and 1, it was logit–transformed for statistical analysis. To test whether hummingbird species preferred a particular flower type of the artificial feeders, we fitted a linear mixed effects model with PI as the response variable and hummingbird species identity and flower type as fixed effects (main and interaction effects). To account for random variation in foraging preferences, we included season and feeding station nested within study site as random effects. Similarly, we performed a linear mixed effects model to test for preferences of hummingbird species for plant species with specific floral morphology in natural communities. In this model, we included main and interaction effects of hummingbird species identity and floral traits (corolla length and corolla curvature) as fixed effects. Site and season were included as random effects. We fitted separate models for each floral trait. We followed Nakagawa and Schielzeth (2013) to estimate R2 as goodness of fit statistics for the overall model. All statistical analyses were performed with R statistical software ver. 3.0.0 (R Development Core Team 2013) and dedicated packages.

Results

Flower choice on experimental feeders

We recorded a total of 4.503 interactions between the three artificial flower types and the three selected hummingbird species (visits by H. jacula were excluded). Artificial feeders were used by Eupherusa nigriventris and Phaethornis guy in the pre– montane forest, and Lampornis calolaemus in the lower montane forest. All three hummingbird species had a preference for the short flower type (Table 1, Fig. 1B, Fig. 3). PI values for the long–curved and the long–straight flower types were lower than 0.5 in all cases (Fig. 1B). Foraging preferences of P. guy for the short flower type was

132

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions significantly lower than for the two other hummingbird species (see the respective interaction term in Table 1).

Table 1. Linear mixed effects model of the relationships between foraging preference indices and predictor variables (species identity and artificial flower types) in two forest types in Costa Rica. Eupherusa nigriventris (EUPNIG) and Phaethornis guy (PHAGUY) visited the artificial feeders in the sub–montane forest and Lampornis calolaemus (LAMCAL) in the lower montane forest. Artificial flower types were short, long–straight and long–curved. Season and feeding station nested within forest type were included as random effects in the model. Coefficient of determination (R²) was 0.72. The reference level (intercept) is given by EUPNIG and the long– curved flower type. Significant effects are shown in bold.

Response variable β t value p Intercept -0.94 2.74 0.009 LAMCAL 0.01 0.03 0.977 PHAGUY 1.03 3.01 0.004 Long–straight 0.03 0.08 0.941 Short 2.34 6.02 < 0.001 LAMCAL × long–straight 0.25 0.52 0.607 LAMCAL × short -0.17 -3.36 0.724 PHAGUY × long–straight -0.29 -0.61 0.543 PHAGUY × short -1.65 -3.46 0.001

133

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Figure 1. (A) Artificial feeders used to assess hummingbird preferences for artificial flower types under experimental conditions in two tropical forests in Costa Rica. Each feeder included three tubes representing one of the following flower types: short, long–straight and long– curved. (B) Means and standard errors of preference index for three hummingbird species: Eupherusa nigriventris (EUPNIG) and Phaethornis guy (PHAGUY) in the pre–montane forest and Lampornis calolaemus (LAMCAL) in the lower montane forest. Statistics are provided in Table 1. Hummingbird illustrations were taken from Del Hoyo et al. (1999).

Flower choice in natural plant communities

At the two study sites, we observed 324 interactions between 34 flowering plant species and the three selected hummingbird species (see Table S1 for a complete plant species list, Supporting Information). The two measured morphological traits of plants (length and curvature of corolla) had significant effects on foraging preferences of E. nigriventris and P. guy in the pre–montane forest. E. nigriventris preferred to feed on flowers with short and straight corollas, whereas P. guy preferred to feed on flowers with long and curved corollas (Table 2, Fig. 2A, B). In the lower montane forest, L. calolaemus did not show preferences for plant species with specific corolla morphology (Fig. 2C, D). Niche plots indicate that preferences of hummingbird species for specific floral types under natural conditions strongly differed from those under experimental conditions with unlimited nectar resources (Fig. 3).

134

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Table 2. Linear mixed effects models of the relationships between foraging preference indices and predictor variables (species identity and floral traits) in two forest types in Costa Rica. Eupherusa nigriventris (EUPNIG) and Phaethornis guy (PHAGUY) visited plant species in the sub–montane forest and Lampornis calolaemus (LAMCAL) in the lower montane forest. Floral traits were corolla length (a) and corolla curvature (b). Season and forest type were included as random effects in the model. Coefficients of determination (R²) were 0.18 and 0.08 for models including corolla length and corolla curvature, respectively. The reference level (intercept) is given by EUPNIG and the respective floral trait. Significant effects are shown in bold.

Response variable β t value p (a) Corolla length Intercept 1.23 3.73 < 0.001 LAMCAL -1.26 -2.85 0.005 PHAGUY -2.26 -4.96 < 0.001 Corolla length -0.04 -4.45 < 0.001 LAMCAL × corolla length 0.04 3.29 0.001 PHAGUY × corolla length 0.06 5.55 < 0.001 (b) Corolla curvature Intercept 0.39 1.77 0.080 LAMCAL -0.42 -1.38 0.172 PHAGUY -0.69 -2.32 0.023 Corolla curvature -2.74 -3.03 0.003 LAMCAL × corolla curvature 3.23 2.56 0.012 PHAGUY × corolla curvature 4.53 3.56 <0.001

135

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Figure 2. Associations between the foraging preference index (PI) of hummingbird species and floral traits in two tropical forests in Costa Rica. Each data point represents one of the following hummingbird species: Eupherusa nigriventris, Phaethornis guy (A, B) and Lampornis calolaemus (C, D). For each hummingbird species, we calculated the corresponding PI for each video–taped plant species; plant species are represented by their mean floral trait values of length and curvature of corolla, respectively. The horizontal dashed line indicates the interactions exactly matching the null hypothesis where the visitation frequency of hummingbird species is proportional to the observation time. We used the predicted values from the linear mixed effects models in Table 2 to fit a trend line in each plot when the association between foraging preferences of hummingbird species and flower types was statistically significant.

136

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Figure 3. Flower choice of hummingbird species in two tropical forests in Costa Rica on artificial (A) and natural flowers (B). Flower types were short, long–straight and long–curved. Eupherusa nigriventris and Phaethornis guy visited feeders and plant species in the sub–montane forest and Lampornis calolaemus in the lower montane forest. Different shades of grey indicate preference indices (PI; range: 0–1) of hummingbird species for the three types of artificial flowers and their corresponding morphologies under natural conditions, as derived by predictions from linear mixed effects models (see Tables 1 and 2). In the case of natural flowers, PI values were derived separately for the specific length (10 and 40 cm) and curvature values (0º and 30º) and then the geometric mean of the two predicted PI values was calculated. Predictions of PI values under natural conditions were always lower than those under experimental conditions because several of the observed plant species were not visited by the respective hummingbird species (PI = 0 in these cases; see Fig. 2).

137

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Hummingbird abundance and traits

The three studied hummingbird species contributed considerably to total hummingbird abundance at each study site. P. guy was the most abundant species at the pre–montane forest site followed by E. nigriventris, representing 42% and 15% of the total number of hummingbird individuals captured in that forest type. L. calolaemus was the dominant species in the lower montane forest, representing 56% of all captured hummingbird individuals. Mean values of bill length (± 1 SE) were 15.6 ± 0.3, 40.5 ± 0.4 and 21.5 ± 0.2 mm for E. nigriventris, P. guy and L. calolaemus, respectively. Mean values of bill curvature were 3.80 ± 0.12, 15.71 ± 0.40 and 3.34 ± 0.06º for the same species, respectively.

Discussion

Under experimental conditions, the three selected hummingbird species overlapped in their interaction niches showing preferences for the short artificial flower type over the long–straight and the long–curved flower types. Under natural conditions, the short–billed species E. nigriventris preferred to feed on short and straight flowers. In contrast, the long–curved bill species P. guy preferred plant species with long and curved flowers, whereas the medium–size billed species L. calolaemus did not show preferences for plant species with specific floral traits. Our findings show that preferences of hummingbird species for specific floral mophologies strongly differed between experimental and natural conditions (i.e. under unlimited and limited nectar resources).

Interaction niche under experimental conditions

Foraging preferences for the short flower type under experimental conditions was more pronounced in the short–billed and the medium–size billed hummingbird species than in the long– and curved–billed species. The importance of morphological matching in determining patterns of interactions between plant and consumer species has been demonstrated for different types of mutualistic interaction networks such as

138

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions plant–pollinator (Stang et al. 2006, Ibanez 2012, Maglianesi et al. 2014) and plant– frugivore networks (Moran and Catterall 2010, Dehling et al. 2014). In plant– hummingbird interactions, morphological floral traits of plants may act as barriers allowing only certain hummingbirds access to nectar rewards (Temeles et al. 2009, Vizentin–Bugoni et al. 2014). The strong preference of the short and medium–size billed hummingbirds for the short flower type in the experiment is very likely to result from morphological constraints of accessibility to nectar imposed by long–straight and long–curved flower types. Our results indicate that length and curvature of corolla are important floral traits that limit the access to nectar resources and therefore determine the interaction niche of short–billed hummingbird species within the trait space that was experimentally explored.

P. guy with its long and curved bill was able to utilize all three artificial flower types in our field experiment. Wiklund et al. (1979) proposed that elongate mouthparts in pollinators offer them the opportunity to exploit a greater diversity of floral morphologies, which has been confirmed by several studies (Feinsinger 1976, Corbet 2000, Goldblatt and Manning 2000). The long bill of the hermit species enables it to access nectar resources from the long and curved flower types unlike the shorter– billed species. Even though P. guy used all three artificial flower types, it preferred the short flower type. This may be the result of easier nectar intakes on short flowers (e.g. by faster bill insertion into the short tube). However, the preference for the short artificial flowers also resulted in an overlapping interaction niche with those of the two shorter–billed species. The unlimited nectar resources provided to hummingbirds in the experiment is likely to have contributed to niche overlap among the hummingbird species in the experiment.

Some hummingbirds visiting the artificial feeders were observed to chase other individuals, which may be an indicator of direct competition for the offered nectar resources. This aggressive behaviour may have influenced how many birds had the opportunity to feed at the feeders but it is unlikely to influence the foraging preferences of hummingbirds because all three flower types were blocked together within the same feeding station. An individual that had conquered a feeding station had the opportunity to choose among the three different flower types. Thus,

139

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions competition between hummingbird individuals at the artificial feeders with unlimited nectar resources is very unlikely to have influenced foraging decisions of hummingbirds. We propose that experiments with artificial flowers and unlimited nectar resources can be utilized to better quantify the interaction niche of hummingbirds in the absence or, at least, with strongly reduced intensity of competition.

Interaction niche under natural conditions

Under natural conditions, the two co–occurring hummingbird species in the pre– montane forest showed strong foraging preferences for plant species with floral traits matching their bill morphology. The short–billed species E. nigriventris preferred to feed on short and straight flowers whereas the long–billed species P. guy preferred to feed on long and curved flowers. This is consistent with other studies showing that nectar–feeding birds were specialized on plant species with flowers traits that matched their bill morphology (Wolf et al. 1976, Dalsgaard et al. 2009, Geerts and Pauw 2009, Maglianesi et al. 2014). The interaction niche of E. nigriventris under natural conditions corresponded to the interaction niche under unlimited nectar resources at the feeders within the range of trait values that was experimentally investigated (Fig. 3). These results suggest that foraging preferences of short–billed hummingbird species are likely driven by morphological constraints.

Foraging preferences by P. guy for long and curved flowers may result from factors other than morphological constraints because interaction niches differed considerably between experimental and natural conditions (Fig. 3). First, the long and curved flowers may be more attractive for the long– and curved– billed hummingbird species because these flowers may per se offer higher nectar rewards (Geerts and Pauw 2009, Ornelas et al. 2007). Second, the long and curved flowers may be more attractive for the hermits because the short– and straight–billed hummingbird species are unable to access most of these flowers, resulting in higher standing crops of nectar rewards in these flowers. The long and curved bills of the hermit species may hence contribute to reduce competition with the other species through resource exploitation. This explanation is consistent with the idea that competition for resources contributes to

140

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions floral resource partitioning in hummingbird communities (Stiles 1981). Third, short flowers could be defended by short–billed hummingbird species and thus competitive interference may limit access to these flowers (Case and Gilpin 1974, Feinsinger 1976). Our data do not enable us to disentangle these three potential causes for niche segregation of hermit species relative to other hummingbird species under natural conditions. Nevertheless, the comparison of experimental and real–world data shows that resource use in natural pollinator communities is strikingly different from that under controlled experimental conditions with unlimited and equally rewarding resources.

L. calolaemus did not show preferences for plant species with specific floral traits in natural plant communities, which reflects a great flexibility in foraging behaviour of this species. The interaction niche of this species for natural flowers strongly differed from those on the artificial flowers. Our results are consistent with previous studies in which L. calolaemus was classified as a generalist and described as a species with an opportunistic ecological role in natural plant communities (Feinsinger and Colwell 1978). In the lower montane forest, L. calolaemus was the dominant species (>50% of the captured individuals belonged to the species), suggesting that intraspecific competition for floral resources might be intense among hummingbird individuals and potentially was more important than interspecific competition. Species abundance affects individual foraging decisions because foraging choices of individuals influence those of other individuals depleting similar or the same floral resources (Tur et al. 2013). Consequently, high levels of intraspecific overlap in plant resource use may result in individuals expanding their interaction niches to all types of potential resources (Bolnick et al. 2003, Maruyama et al. 2013). A generalist strategy allows abundant pollinators to use a wide range of food resources. L. calolaemus was able to use some flowers longer than their bills. This may be possible because this species, like other hummingbird species, is able to extend its tongue to harvest nectar from long flowers (Paton and Collins 1989). Furthermore, in unvisited flowers the nectar volume may exceed the nectar chamber and therefore may be accessible for shorter– billed hummingbirds (Wolf and Stiles 1989).

141

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Conclusions

Comparisons between experimental and observational data of foraging preferences of pollinator species aid to understand the factors defining species’ interaction niches and the mechanisms driving floral resource partitioning among pollinators. Our experimental results reveal that floral traits may limit the access to nectar resources of short–billed hummingbird species and therefore constrain certain interactions in plant–hummingbird networks (“forbidden link hypothesis”; Jordano et al. 2003). However, pollinators may change foraging preferences in response to additional factors, such as reward quantity and competition for floral resources resulting in partitioning or expansion of their interaction niche. This is supported by our results with two co–occurring hummingbird species segregating their overlapping interaction niches under experimental conditions to non–overlapping realized interaction niches in the real world. Correspondingly, a species with strong foraging preferences for short flowers in the absence of competition expanded its interaction niche to a broad range of flower resources most likely driven by high intraspecific competition. We conclude that morphological constraints are one important mechanism structuring trophic networks, albeit other factors, such as as inter– and intraspecific competition, additionally define interaction niches of consumer species in real–world communities.

Acknowledgements

We thank the field assistants and volunteers who contributed to data collection and botanists from the Instituto Nacional de Biodiversidad and La Selva Biological Station (OTS) for support with plant identification. We are grateful to Mathias Templin for help in designing the artificial feeders. This work was funded by the following organizations: Consejo Nacional para Investigaciones Científicas y Tecnológicas and Ministerio de Ciencia, Tecnología y Telecomunicaciones, Universidad Estatal a Distancia, Organization for Tropical Studies (OTS), German Academic Exchange Service and Tropical Science Centre. Financial support for this study was also provided by the research–funding programme ‘‘LOEWE–Landes–Offensive zur Entwicklung Wissenschaftlich–ökonomischer Exzellenz’’ of Hesse’s Ministry of Higher Education,

142

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Research, and the Arts. We thank Anton Pauw and Jeferson Vizentin–Bugoni for helpful discussions and insightful comments on an earlier version of this manuscript.

References

Bascompte, J. and P. Jordano. 2007. Plant–animal mutualistic networks: the architecture of biodiversity. Annual Review of Ecology, Evolution, and Systematics 38:567–93.

Blake, J. G. and B. A. Loiselle. 2000. Diversity of birds along an elevational gradient in the Cordillera Central, Costa Rica. The Auk 117:663–686.

Blüthgen, N. 2010. Why network analysis is often disconnected from community ecology: a critique and an ecologist’s guide. Basic and Applied Ecology 11:185–195.

Bolnick, D. I., R. Svanback, J. A. Fordyce, L. H. Yang, J. M. Davis, C. D. Hulsey, and M. L. Forister. 2003. The ecology of individuals: incidence and implications of individual specialization. The American Naturalist 161:1–28.

Brown, J. H. and M. A. Bowers. 1985. Community organization in hummingbirds: Relationships between morphology and ecology. Auk 102:251–269.

Case, T. J. and M. E. Gilpin. 1974. Interference competition and niche theory. Proceedings of the National Academy of Sciences USA 71:3073–3077.

Chalcoff, V. R., M. A. Aizen, and L. Galetto. 2006. Nectar Concentration and Composition of 26 Species from the Temperate Forest of South America. Annals of Botany 97:413–421.

Colwell, R. K. 1973. Competition and coexistence in a simple tropical community. The American Naturalist 107:737–760.

Colwell, R. K. and E. R. Fuentes. 1975. Experimental studies of the niche. Annual Review of Ecology, Evolution, and Systematics 6:281–310.

143

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Corbet, S. A. 2000. Butterfly nectaring flowers: Butterfly morphology and flower form. Entomologia Experimentalis et Applicata 96:289–298.

Cotton, P. A. 1998. Coevolution in an Amazonian hummingbird–plant community. Ibis 140:639–646.

Dalsgaard, B., A. M. Martín Gonzáles, J. M. Olesen, J. Ollerton, A. Timmermann, L. H. Andersen, et al. 2009. Plant–hummingbird interactions in the West Indies: floral specialization gradients associated with environment and hummingbird size. Oecologia 159:757–766.

Del Hoyo, J., A. Elliott, J. Sargatal, and N. J. Collar (eds.). 1999. Handbook of the birds of the world, vol. 5: barn owls to hummingbirds. Lynx Editions, Barcelona, Spain.

Faegri, K. and L. van der Pijl. 1979. The principles of pollination ecology. Pergamon Press, Oxford, England.

Feinsinger, P. 1976. Organization of a tropical guild of nectarivorous birds. Ecological Monographs 46:257–291.

Feinsinger, P. and R. K. Colwell. 1978. Community organization among neotropical nectar feeding birds. American Zoologist 18:779–795.

Fründ, J., K. E. Linsenmair, and N. Blüthgen. 2010. Pollinator diversity and specialization in relation to flower diversity. Oikos 119:1581–1590.

144

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Goldblatt, P. and J. C. Manning. 2000. The long–proboscid fly pollination system in southern Africa. Annals of the Missouri Botanical Garden 87:146–170.

Grant, K. A. and V. Grant. 1968. Hummingbirds and their flowers. Columbia Univ. Press, New York.

Hainsworth, F. R. and L. L. Wolf. 1972. Crop volume, nectar concentration and hummingbird energetics. Comparative Biochemistry and Physiology 42A:359–366.

Hartshorn, G. and R. Peralta. 1988. Preliminary description of primary forests along the La Selva–Volcan Barva altitudinal transect, Costa Rica. Pages 281–295 in Almeda, F. and C. M. Pringle, editors. Tropical rainforests: Diversity and conservation California Academy of Sciences, San Francisco, USA.

Holdridge, L. R. 1967. Life zone ecology. Tropical Science Center, San José.

Hutchinson, G. E. 1957. Concluding remarks. Cold Spring Harbor Symposia Quantitative Biology 22:415–427.

Ibanez, S. 2012. Optimizing size thresholds in a plant–pollinator interaction web: towards a mechanistic understanding of ecological networks. Oecologia 170:233– 242.

Jordano. P., J. Bascompte, and J. M. Olesen. 2003. Invariant properties in coevolutionary networks of plant–animal interactions. Ecology Letters 6:69–81.

Linhart, Y. B. 1973. Ecological and behavioral determinants of pollen dispersal in hummingbird–pollinated Heliconia. The American Naturalist 107:511–523.

145

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

McDade, L. A. 2004. Nectar in Hummingbird-pollinated Neotropical Plants I: Patterns of Production and Variability in 12 Species. Biotropica 36:196–215.

Moran, C. and C. P. Catterall. 2010. Can functional traits predict ecological interactions? A case study using rain forest frugivores and plants in Australia. Biotropica 42:318–326.

Nakagawa, S. and H. Schielzeth. 2013. A general and simple method for obtaining R2 from generalized linear mixed–effects models. Methods in Ecology and Evolution 4:133–142.

Ollerton J., R. Alarcón, N. M. Waser, M. V. Price, S. Watts, L. Cranmer, et al. 2009. A global test of the pollination syndrome hypothesis. Annals of Botany 103:1471– 1480.

Ornelas, J. F., M. Ordano, A. J. De-Nova, M. E. Quintero, and T. Garland Jr. 2007. Phylogenetic analysis of interspecific variation in néctar of hummingbird–visited plants. Journal of Evolutionary Biology 20:1904–1917.

Paton, D. C. and B. G. Collins. 1989. Bills and tongues of nectar–feeding birds: a review of morphology, function and performance, with intercontinental comparisons. Australian Journal of Ecology 14:473–506.

Pauw, A. 2013. Can pollination niches facilitate plant coexistence? Trends in Ecology and Evolution 28:30–37.

Peterson, C. H. and P. E. Renaud. 1989. Analysis of feeding preference experiments. Oecologia 80:82–86.

R Development Core Team. 2013. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.

Ralph, C. J., G. R. Geupel, P. Pyle, T. Martin, and D. F. DeSante. 1993. Handbook of field methods for monitoring landbirds. Pacific Southwest Research Station Albany, California, USA. 146

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Robertson, A. W., D. Kelly, J. J. Ladley, and A. D. Sparrow. 1999. Effects of pollinator loss on endemic New Zealand mistletoes (Loranthaceae). Conservation Biology 13:499–508.

Rodríguez-Flores, C. I. and F. G. Stiles. 2005. Análisis ecomorfológico de una comunidad de colibríes ermitaños (Trochilidae, Phaethorninae) y sus flores en la Amazonia Colombiana. Ornitología Colombiana 3:7–27.

Stang, M., Klinkhamer, P. G. L. and E. van der Meijden. 2006. Size constraints and flower abundance determine the number of interactions in a plant–flower visitor web. Oikos 112:111–121.

Stiles, F. G. 1975. Ecology, flowering phenology, and hummingbird pollination of some Costa Rican Heliconia species. Ecology 56:285–301.

Stiles, F. G. 1978. Ecological and Evolutionary Implications of Bird Pollination. American Zoologist 8:715–727.

Stiles, F. G. 1981. Geographical aspects of Bird–flower coevolution, with particular reference to Central America. Annals of the Missouri Botanical Garden 68:323–351.

Stiles, F. G. and A. Skutch. 1989. A guide to the birds of Costa Rica. Comstock Publishing Associates. A division of Cornell University Press, Ithaca, USA.

TEAM. 2013. Tropical Ecology Assessment and Monitoring Network. Volcán Barba, Heredia, Costa Rica.

Temeles, E. J., C. R. Koulouris, S. E. Sander, and W. J. Kress. 2009. Effect of flower shape and size on foraging performance and trade–offs in a tropical hummingbird. Ecology 90:1147–1161.

Temeles, E. J., Y. B. Linhart, M. Masonjones, and H. D. Masonjones. 2002. The role of flower width in hummingbird bill length–flower length relationships. Biotropica 34:68–80.

147

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Tur, C., B. Vigalondo, K. Trøjelsgaard, J. M. Olesen, and A. Traveset. 2013. Downscaling pollen–transport networks to the level of individuals. Journal of Animal Ecology, doi: 10.1111/1365–2656.12130.

Vizentin–Bugoni, J., P. K. Maruyama, and M. Sazima. 2014. Processes entangling interactions in communities: forbidden links are more important than abundance in a hummingbird–plant network. Proceedings of the Royal Society B 281:1–8.

Wiklund, C., T. Eriksson, and H. Lundberg. 1979. The wood white butterfly, Leptidea sinapsis, and its nectar plants: A case of mutualism or parasitism? Oikos 33:358– 362.

Williams, P. 2005. Does specialization explain rarity and decline among British bumblebees? A response to Goulson et al. – Biological Conservation 122:33–43.

Wolf, L. L., F. G. Stiles, and F. R. Hainsworth. 1976. Ecological organization of a highland tropical hummingbird community. Journal of Animal Ecology 45:349–379.

Wolf, L. L. and F. G. Stiles. 1989. Adaptations for the 'fail–safe' pollination of specialized ornithophilous flowers. American Midland Naturalist 121:1–10.

Supporting Information

Supporting information includes Table S1 with plant species and the number of visits each plant species received from hummingbird species in two tropical forest types in Costa Rica.

148

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Supporting Information

Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant-pollinator interactions

María Alejandra Maglianesi, Katrin Böhning–Gaese and Matthias Schleuning

Table S1. List of plant families, species and the number of visits they received from the three selected hummingbird species in Costa Rica. Eupherusa nigriventris and Phaethornis guy visited plant species in the pre–montane forest, whereas Lampornis calolaemus visited plant species in the lower montane forest. Plant families are ordered alphabetically within each forest type. Two Bromelia species in the pre– montane forest could only be classified into morphospecies.

Forest Type Family Plant Species Nº of Visits Pre–montane Bromeliaceae Bromelia 1 0 Pre–montane Bromeliaceae Bromelia 2 0 Pre–montane Bromeliaceae Guzmania nicaraguensis 3 Pre–montane Bromeliaceae Guzmania scandens 0 Pre–montane Bromeliaceae Pitcairnia brittoniana 6 Pre–montane Campanulaceae Centropogon granulosus 0 Pre–montane Costaceae Costus curvibracteatus 2 Pre–montane Costaceae Costus pulverulentus 23 Pre–montane Cucurbitaceae Gurania coccinea 2 Pre–montane Ericaceae Cavendishia callista 7 Pre–montane Ericaceae Cavendishia complectens 2 Pre–montane Ericaceae Cavendishia querene 0 Pre–montane Ericaceae Psammisia ramiflora 0 Pre–montane Ericaceae Satyria warszewiczii 5 Pre–montane Ericaceae Thibaudia costaricensis 9 Pre–montane Gesneriaceae Besleria notabilis 1

149

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Table S1 (continued)

Forest Type Family Plant Species Nº of Visits Pre–montane Gesneriaceae Besleria triflora 0 Pre–montane Gesneriaceae Columnea microcalyx 0 Pre–montane Gesneriaceae Columnea microphylla 0 Pre–montane Gesneriaceae Columnea purpurata 1 Pre–montane Gesneriaceae Columnea querceti 1 Pre–montane Gesneriaceae Drymonia conchocalyx 3 Pre–montane Gesneriaceae Drymonia tomentulifera 0 Pre–montane Gesneriaceae Drymonia warszewicziana 0 Pre–montane Heliconiaceae Heliconia atropurpurea 38 Pre–montane Heliconiaceae Heliconia rodriguezii 0 Pre–montane Heliconiaceae Heliconia vaginalis 23 Pre–montane Marantaceae Calathea lasiostachya 0 Pre–montane Marantaceae Calathea recurvata 0 Pre–montane Rubiaceae Faramea eurycarpa 1 Pre–montane Rubiaceae Hillia triflora 1 Pre–montane Rubiaceae Palicourea gamezii 4 Pre–montane Rubiaceae Palicourea lasiorrhachis 0 Pre–montane Rubiaceae Palicourea sp. 5 Pre–montane Rubiaceae Psychotria elata 5 Pre–montane Zingiberacea Renealmia cernua 4 Lower montane Alstroemeriaceae Bomarea hirsuta 11 Lower montane Asteraceae Neomirandea eximia 0 Lower montane Bromeliaceae Werauhia ororiensis 5 Lower montane Campanulaceae Burmeistera parviflora 1 Lower montane Ericaceae Cavendishia bracteata 76 Lower montane Ericaceae Centropogon solanifolius 0 Lower montane Ericaceae Disterigma humboltii 0 Lower montane Ericaceae Gaultheria gracilis 3 Lower montane Ericaceae Gonocalyx pterocarpus 5

150

Appendix 3: Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant–pollinator interactions

Table S1 (continued)

Forest Type Family Plant Species Nº of Visits Lower montane Ericaceae Vaccinum poasanum 0 Lower montane Gentianaceae Macrocarpaea valerioi 4 Lower montane Malvaceae Malvaviscus palmanus 0 Lower montane Gesneriaceae Besleria barbensis 0 Lower montane Gesneriaceae Besleria solanoides 4 Lower montane Gesneriaceae Columnea magnifica 0 Lower montane Gesneriaceae Columnea microcalyx 0 Lower montane Gesneriaceae Kohleria tigridia 2 Lower montane Heliconiaceae Heliconia lankesterii 3 Lower montane Orchidaceae Elleanthus aurantiacus 1 Lower montane Rubiaceae Gonzalagunia rosea 0 Lower montane Rubiaceae Hoffmannia arborescens 3 Lower montane Rubiaceae Palicourea lassiorrhachis 60 Lower montane Solanaceae Cestrum poasanum 0

151

Appendix 4: List of hummingbird species recorded at the study sites

Appendix 4: List of humminbird species recorded in three tropical forest types (wet, pre–montane and lower montane wet forest) in northeastern Costa Rica. Hummingbird species are ordered alphabetically within each elevation.

Elevation Species Low Amazilia amabilis Low Amazilia tzacatl Low Chalybura urochrysia Low Florisuga mellivora Low Glaucis aeneus Low Klais guimeti Low Microchera albocoronata Low Phaeochroa cuvierii Low Phaethornis longirostris Low Phaethornis striigularis Low Thalurania colombica Low Threnetes ruckeri Mid Amazilia tzacatl Mid Campylopterus hemileucurus Mid Discosura conversii Mid Doryfera ludovicae Mid Elvira cupreiceps Mid Eupherusa nigriventris Mid Eutoxeres aquila Mid Heliodoxa jacula Mid Lampornis calolaemus Mid Lampornis hemileucus Mid Phaethornis guy Mid Phaethornis striigularis Mid Thalurania colombica High Campylopterus hemileucurus High Colibri thalassinus High Doryfera ludovicae High Eugenes fulgens High Eupherusa nigriventris High Heliodoxa jacula High Hylocharis eliciae High Lampornis calolaemus

152

Appendix 4: List of hummingbird species recorded at the study sites

Appendix 4 (continued)

Elevation Species High Panterpe insignis High Phaethornis guy High Selasphorus flammula High Selasphorus scintilla High Thalurania colombica

153

Appendix 5: List of plant species recorded at the study sites

Appendix 5: List of plant species recorded in three tropical forest types (wet, pre– montane and lower montane wet forest) in northeastern Costa Rica. Plants families are ordered alphabetically within each elevation.

Elevation Family Species Low Acanthaceae Aphelandra storkii Low Acanthaceae Bravaisia integérrima Low Acanthaceae Odontonema cuspidatum Low Acanthaceae Odontonema tubaeforme Low Acanthaceae Razisea wilburii Low Acanthaceae Ruellia metallica Low Alstroemeriaceae Bomarea obovata Low Apocynaceae Odontadenia sp. Low Apocynaceae Stemmadenia sp. Low Verbenaceae Stachytarpheta frantzii Low Bignoniaceae Arrabidaea sp. Low Bignoniaceae Arrabidaea verrucosa Low Bromeliaceae Aechmea mariae–reginae Low Bromeliaceae Aechmea nudicaulis Low Bromeliaceae Guzmania monostachia Low Convolvulaceae Ipomoea batatas Low Costaceae Costus laevis Low Costaceae Costus malortieanus Low Costaceae Costus pulverulentus Low Costaceae Costus scaber Low Fabaceae Erythrina poeppigiana Low Gesneriaceae Besleria columneoides Low Gesneriaceae Chrysothemis friedrichsthaliana Low Gesneriaceae Columnea nicaraguensis Low Gesneriaceae Columnea purpurata Low Gesneriaceae Drymonia microphylla Low Heliconiaceae Heliconia imbricata Low Heliconiaceae Heliconia latispatha Low Heliconiaceae Heliconia mariae Low Heliconiaceae Heliconia mathiasiae Low Heliconiaceae Heliconia pogonantha Low Heliconiaceae Heliconia wagneriana Low Malvaceae Malvaviscus concinnus

154

Appendix 5: List of plant species recorded at the study sites

Appendix 5 (continued)

Elevation Family Species Low Marantaceae Calathea crotalifera Low Marantaceae Calathea gymnocarpa Low Marantaceae Calathea lasiostachya Low Marantaceae Calathea lutea Low Marantaceae Calathea marantifolia Low Marantaceae Pleiostachya pruinosa Low Passifloraceae Passiflora vitifolia Low Rubiaceae Faramea suerrensis Low Rubiaceae Hamelia patens Low Rubiaceae Notopleura capitata Low Rubiaceae Palicourea guianensis Low Rubiaceae Pentagonia monocaulis Low Rubiaceae Psychotria chiapensis Low Rubiaceae Psychotria elata Low Rubiaceae Psychotria poeppigiana Low Rubiaceae Warscewiczia coccinea Low Schlegeliaceae Schlegelia fastigiata Low Schlegeliaceae Schlegelia nicaraguensis Low Zingiberacae Renealmia cernua Mid Apocynaceae Allomarkgrafia brenesiana Mid Apocynaceae Tabernaemontana alfaroi Mid Bromeliaceae Guzmania nicaraguensis Mid Bromeliaceae Pitcairnia valerioi Mid Campanulaceae Centropogon granulosus Mid Convolvulaceae Maripa nicaraguensis Mid Costaceae Costus curvibracteatus Mid Costaceae Costus pulverulentus Mid Ericaceae Cavendishia callista Mid Ericaceae Cavendishia complectens Mid Ericaceae Macleania rupestris Mid Ericaceae Psammisia ramiflora Mid Ericaceae Satyria warszewiczii Mid Ericaceae Sphyrospermum dissimile Mid Ericaceae Thibaudia costaricensis Mid Gentianaceae Macrocarpaea valerioi Mid Gesneriaceae Besleria notabilis

155

Appendix 5: List of plant species recorded at the study sites

Appendix 5 (continued)

Elevation Family Species Mid Gesneriaceae Besleria triflora Mid Gesneriaceae Columnea microphylla Mid Gesneriaceae Columnea purpurata Mid Gesneriaceae Columnea querceti Mid Gesneriaceae Drymonia warszewicziana Mid Heliconiaceae Heliconia atropurpurea Mid Heliconiaceae Heliconia rodriguezii Mid Heliconiaceae Heliconia vaginalis Mid Lamiaceae Scutellaria costaricana Mid Marantaceae Calathea crotalifera Mid Marantaceae Calathea lasiostachya Mid Marantaceae Calathea recurvata Mid Orchidaceae Elleanthus hymenophorus Mid Rubiaceae Hillia triflora Mid Rubiaceae Palicourea sp.1 Mid Rubiaceae Palicourea lasiorrhachis Mid Rubiaceae Psychotria carthagenensis Mid Rubiaceae Psychotria elata Mid Rubiaceae Sabicea panamensis Mid Solanaceae Merinthopodium neuranthum Mid Solanaceae Schultesianthus venosus Mid Zingiberacea Renealmia cernua Mid Zingiberacea Renealmia ligulata High Actinidiaceae Saurauia montana High Alstroemeriaceae Bomarea hirsuta High Asteraceae Neomirandea eximia High Orchidaceae Fernandezia tica High Orchidaceae Maxillaria sp. High Bromeliaceae Tillandsia insignis High Bromeliaceae Werauhia camptoclada High Bromeliaceae Werauhia ororiensis High Campanulaceae Burmeistera parviflora High Campanulaceae Centropogon solanifolius High Ericaceae Cavendishia bracteata High Ericaceae Disterigma humboldtii High Ericaceae Gaultheria gracilis

156

Appendix 5: List of plant species recorded at the study sites

Appendix 5 (continued)

Elevation Family Species High Ericaceae Satyria warszewiczii High Ericaceae Sphyrospermum cordifolium High Ericaceae Vaccinium poasanum High Gentianaceae Macrocarpaea valerioi High Gesneriaceae Besleria barbensis High Gesneriaceae Besleria solanoides High Gesneriaceae Columnea magnifica High Gesneriaceae Kohleria tigridia High Heliconiaceae Heliconia lankesteri High Lamiaceae Aegiphila odontophylla High Lamiaceae Scutellaria isocheila High Malvaceae Malvaviscus palmanus High Melastomataceae Meriania phlomoides High Orchidaceae Elleanthus aurantiacus High Orchidaceae Elleanthus sp. High Orchidaceae Epidendrum lacustre High Orchidaceae Epidendrum microrigidiflorum High Orchidaceae Epidendrum polychlamys High Orchidaceae Scaphyglottis sigmoidea High Orchidaceae Sobralia amabilis High Orquidiaceae Elleanthus tricallosus High Rubiaceae Gonzalagunia rosea High Rubiaceae Guettarda crispiflora High Rubiaceae Hillia maxonii High Rubiaceae Hoffmannia arborescens High Rubiaceae Palicourea lasiorrhachis High Solanaceae Cestrum poasanum

157

Appendix 6: Curriculum vitae

EDUCATION PhD research 2010-present Biodiversity and Climate Research Centre (BIK-F) and Senckenberg Gesellschaft für Naturforschung, Frankfurt am Main, Germany. Thesis: “Understanding patterns and processes in plant–hummingbird mutualistic networks in the Neotropics”. M.S., Conservation and Management of Wildlife 2001- 2004 Instituto Internacional en Manejo y Conservación de Vida Silvestre (ICOMVIS), Universidad Nacional, Costa Rica. Thesis: Effects of the habitat characteristic on the abundance and diversity of resident and migrant landbirds in the Caribbean north of Costa Rica. B.S., Biology 1991-1996 Facultad de Ciencias Exactas, Físicas y Naturales, Universidad Nacional de Córdoba, Argentina.

RESEARCH EXPERIENCE Project Leader Set 2008-Jun2010 Project: “Conservation status of the Reserva Forestal Grecia based on vegetation structure and bird community for the establishment of ecological restoration strategies”. Programa en Manejo de Recursos Naturales (MARENA). Facultad de Ciencias Exactas y Naturales. Vicerrectoría de Investigación. Universidad Estatal a Distancia (UNED). Duties: designed the project, collected, maintained and statistically evaluated data, and wrote technical reports. Bird-banding Station Coordinator Jan 2005-Jul 2008 Project: “Monitoring of Overwintering Survival Program (MoSI)” Principal Investigators (PIs): James Saracco and Peter Pyle. Institute for Bird Population (IBP, U.S.). Duties: coordinated a bird-banding station of migratory avian species and trained students in avian monitoring techniques at Palo Verde National Park, Costa Rica. Project Leader Jul-Nov 2007 Project: “Evaluation of ecosystems in the Talamanca sub-ecoregion by using the bird community structure”. Fund BID-FMAM-BM. ACICAFOC, CICA, CCAD. Duties: collected, maintained and statistically evaluated data, wrote a technical report.

158

Appendix 6: Curriculum vitae

Field Assistant Mar-May 2004 Project: “Contribution of live fences to farm productivity and ecological integrity of agricultural landscapes in Central America”. PI: candidate for a PhD. degree Rachel Taylor. Centro Agronómico Tropical en Investigaciones Ecológicas (CATIE), Turrialba, Costa Rica. Duties: re-located colour banded birds in agro ecosystems, used GPS to mark avian detections and collected data at several agro ecosystems and remnant forest patch. Field Assistant Set-Oct 2000 Project: “Green Macaw (Ara ambigua) Conservation Program in northern of Costa Rica”. PIs: MSc. Guisselle Monge and Olivier Chassot. Tropical Scientific Center. San José, Costa Rica. Duties: conducted Green Macaws radio tracking in northern of Costa Rica. Field Assistant Mar-Abr 2000 Project: “Characteristics of bird species using forest and agricultural land covers in southern Costa Rica” and “Habitat use of adult White-throated Robins (Turdus assimilis) during the breeding season in a mosaic landscape in Costa Rica”. PI: candidate for a postdoctoral degree Catherine Lindell, College of Natural Science. Department of Zoology, Michigan State University, U.S. Duties: conducted mist- netting, Turdus assimilis radio tracking, laying grids in the tropical wet forest, nest- searching and nest-monitoring of all terrestrial bird species present in the Zona Protectora Las Tablas, Talamanca, Costa Rica. Field Assistant Jul-Oct 1999 Project: “Bird Monitoring and assessment project in California and Oregon (U.S)”. PI: Dr. C. John Ralph. Redwood Science Laboratory, U.S. Forest Service. Arcata, California. Duties: conducted mist-netting, processing birds, maintained log books and data sheets at several banding stations in northern California and Southern Oregon. Tested and participated of the new Rapid Ornithological Inventory: a combination of diurnal, area search census; nocturnal, call station survey; and mist-netting. Field Assistant Abr-May 1999 Project: “Integrated Landbird Monitoring Program at Tortuguero, on the Caribbean Coast of Costa Rica”. PI: Dr. C. John Ralph. Redwood Science Laboratory, U.S. Forest Service. Arcata, California. Duties: conducted mist netting, processing birds, maintained log books and data sheets at several banding stations on the Caribbean coast of Costa Rica. Field Assistant. Feb-Apr 1998 Project: “Landbird Monitoring Program, Santa Cruz de la Sierra, Bolivia” PI: Susan Davis. Museo de Historia Natural Noel Kempff Mercado. Universidad Autónoma René Gabriel Moreno, Bolivia. Duties: conducted mist-netting in several filed stations in Santa Cruz de la Sierra.

159

Appendix 6: Curriculum vitae

Project leader. 1998 Project: “The use of wildlife by Brazil nut collectors in the Bolivian Amazon”. Funded by USAID and Bolivian Government. Duties: designed the project, collected, maintained and statistically evaluated data (direct observations and hunter interviews), and wrote an article for a specialized journal.

TEACHING EXPERIENCE Undergraduate Professor 2006-2008 I taught the following courses: General Ecology Universidad Estatal a Distancia-UNED and Universidad de Costa Rica-UCR) in Jan-Dec 2008; Restoration Ecology (UCR) in Aug-Dec 2007; Animal Behaviour (Universidad Nacional-UNA) in Jul-Dic 2007; Wildlife, Principles of Biology and Behaviour Ecology (UNA) in Jan-Dic 2006 and Behaviour Ecology (UNA) in Jan-Jul 2006. Instructor in the Bird Monitoring Techniques Training Workshop, Feb 2003 Tortuguero, Costa Rica. Redwood Sciences Laboratory, U.S. Forest Service. Funded by Point Reyes Bird Observatory (PRBO), CA, U.S.

PUBLICATIONS

Maglianesi, M. A., N. Blüthgen, Katrin Böhning–Gaese and M. Schleuning. 2014. Morphological traits determine specialization and resource use in plant– hummingbird networks in the Neotropics. Ecology, http://dx.doi.org/10.1890/13- 2261.1 Maglianesi, M. A., N. Blüthgen, Katrin Böhning–Gaese and M. Schleuning. 2014. Functional structure and specialization in tropical plant–hummingbird interaction networks across elevations. Ecography (rejected with an invitation for resubmission). Maglianesi, M. A., Katrin Böhning–Gaese and M. Schleuning. 2014. Different foraging preferences of hummingbirds on artificial and natural flowers reveal mechanisms structuring plant-pollinator interactions. Journal of Animal Ecology (conditionally accepted subject to minor revision). Maglianesi, M. A. 2010. Avifauna asociada a bosque nativo y plantación exótica de coníferas en la Reserva Forestal Grecia (Costa Rica). Ornitología Neotropical 21:339–350. Rumiz, D. I. and M. A. Maglianesi. 2001. Hunting Impacts Associated with Brazil Nut Harvesting in the Bolivian Amazon. Vida Silvestre Neotropical 10(1-2): 19-29.

160