Large frugivores matter more on an island: Insights from island‐mainland comparison of plant–frugivore communities

Abstract Endozoochory, a mutualistic interaction between plants and frugivores, is one of the key processes responsible for maintenance of tropical biodiversity. Islands, which have a smaller subset of plants and frugivores when compared with mainland communities, offer an interesting setting to understand the organization of plant–frugivore communities vis‐a‐vis the mainland sites. We examined the relative influence of functional traits and phylogenetic relationships on the plant–seed disperser interactions on an island and a mainland site. The island site allowed us to investigate the organization of the plant–seed disperser community in the natural absence of key frugivore groups (bulbuls and barbets) of Asian tropics. The endemic Narcondam Hornbill was the most abundant frugivore on the island and played a central role in the community. Species strength of frugivores (a measure of relevance of frugivores for plants) was positively associated with their abundance. Among plants, figs had the highest species strength and played a central role in the community. Island‐mainland comparison revealed that the island plant–seed disperser community was more asymmetric, connected, and nested as compared to the mainland community. Neither phylogenetic relationships nor functional traits (after controlling for phylogenetic relationships) were able to explain the patterns of interactions between plants and frugivores on the island or the mainland pointing toward the diffused nature of plant–frugivore interactions. The diffused nature is a likely consequence of plasticity in foraging behavior and trait convergence that contribute to governing the interactions between plants and frugivores. This is one of the few studies to compare the plant–seed disperser communities between a tropical island and mainland and demonstrates key role played by a point‐endemic frugivore in seed dispersal on island.


| INTRODUC TI ON
Endozoochory, a mutualistic interaction between plants and frugivorous animals, represents a critical stage in plant regeneration and contributes to the maintenance of tree diversity in tropics (Terborgh et al., 2002). The community of fleshy-fruited plants and seed dispersers form a network of bipartite interactions. These networks could be influenced by different ecological (e.g., functional traits, phenological overlap) and evolutionary (e.g., co-evolution) processes (Guimarães et al., 2011;Vázquez et al., 2009) whose relative roles are yet to be sufficiently investigated.
Shared evolutionary history can play a role in governing the interactions between plants and frugivores. Traits relevant to the seed dispersal such as beak width have been demonstrated to exhibit a phylogenetic signal (Rezende et al., 2007). On the other hand, trait convergence could result in unrelated species interacting with similar partners (Peralta, 2016). Thus, interactions could be influenced by trait complementarity-which itself could be modulated by trait conservatism-or by trait convergence (Bastazini et al., 2017;Rezende et al., 2007).
To assess the relative roles of functional traits and their phylogenetic history in determining patterns of ecological interactions, Bastazini et al. (2017)  tarity. An integrated approach that examines the role of evolutionary processes in tandem with ecological processes thus enables an understanding of the relative roles of these factors in patterns of community assembly. However, while the substantial phylogenetic data now available has improved our understanding of community assemblages, the complementary empirical data that would allow for an integrated approach have been lacking.
Despite zoochory being a pervasive interaction in the wet tropics of Asia, there is limited information on plant-disperser communities from the mainland sites in the region (Corlett, 2017) and very poor information from oceanic islands in the Indian Ocean (Escribano-Avila et al., 2018). Most of the information on plant-disperser communities worldwide comes from mainland sites with few studies from oceanic islands and even few studies set in an island-mainland comparative framework (Schleuning et al., 2014).
Given this background, we examined the organization of the plant-disperser community on a remote and small tropical island (Narcondam Island) in South Asia that is home to a point-endemic hornbill (Figure 1) using the network and ecophylogenetic approaches ( Figure 2a). The Narcondam island is characterized by a species-poor frugivore assemblage, which also enabled us to examine the relative quantitative role of different species in seed dispersal.
In the Asian tropics, small-bodied frugivores like bulbuls and barbets are known to play key roles in seed dispersal (Naniwadekar et al., 2019). However, the Narcondam Island does not have any bulbuls and barbets and most of the dominant frugivores on the island are medium-and large-bodied avian frugivores (>100 g; Figure 2c).
This offers an interesting system to understand the pattern of interactions in absence of key frugivore groups. We compared this island community with a mainland community from north-east India with which it shares several frugivore and plant species.
The specific aims of the study were to (a) identify key seed dispersers and plant groups on an oceanic island community by comparing species-level properties (degree, species strength and weighted F I G U R E 1 A pair of Narcondam Hornbill Rhyticeros narcondami on Caryota mitis palm with ripe fruits. Male has a red head while the female has a black head. Narcondam Hornbill is a point-endemic hornbill restricted to the 6.8 km 2 Narcondam Island in the Andaman Sea, India. Acrylic on canvas by Sartaj Ghuman betweenness) across different species, (b) examine the influence of abundance and morphological traits on species strength of plants and frugivore species, (c) compare the network-level properties (web asymmetry, connectance, specialization, and weighted nestedness) of the island community with a mainland community, and (d) examine the influence of phylogenetic and functional component (fruit and seed size for plants and body size and degree of frugivory for frugivores) on the distribution of interactions between plants and frugivores on the island and mainland communities. This system allowed us to explicitly examine the change in the role of large-bodied frugivores in plant-seed disperser communities in the absence of the important small-bodied frugivores in Asian tropics (Figure 2c).
We hypothesized that large-bodied frugivores, unhindered by morphological constraints, will demonstrate dietary shifts on islands in the absence of small-bodied frugivores and will feed on fruits which they typically do not feed on in the mainland site. This will alter the organization of the plant-seed disperser community on the island as compared to mainland.

| Study area
The study was carried out on a small (area: 6.8 km 2 ), volcanic island, the Narcondam Island (13°30′N, 94°38′E), in the Andaman Sea in India. The island is part of the Indo-Myanmar Biodiversity Hotspot (Myers et al., 2000). Three distinct vegetation types on the island include the narrow strip of littoral forests, forests dominated by deciduous species in the north-eastern part of the island and evergreen forests in most of the island (Page et al. in prep.). There is no published information on fruiting phenology of woody plants from the Narcondam Island. Ongoing long-term phenology work from Andaman Islands indicates a peak in fruiting between November to February (LEMoN, 2019).
The Endangered Narcondam Hornbill Rhyticeros narcondami is found only on the Narcondam Island (BirdLife International, 2017).
Around 50 species of birds have been reported from the island F I G U R E 2 Schematic illustrating mainland and island plant and seed dispersal communities. (a) map of the South and Southeast Asia showing mainland and island site, (b) the island endemic large-bodied frugivore Narcondam Hornbill Rhyticeros narcondami on Caryota mitis, (c) representative schematic showing organization of plant-seed disperser community on mainland and island. Mainland community is more diverse and symmetric as compared to the island community. On mainland, due to the presence of great diversity of small-bodied frugivores which are more efficient in removal of small fruits, large-bodied frugivores may predominantly feed on large-seeded plants. On the island, in the absence of small-bodied frugivores, large-bodied frugivores will start feeding on small fruits resulting in dietary shifts and more dominant role in the community as indicated by the thickness of arrows. Dietary changes will in turn result in changes in interaction frequencies with corresponding changes in network structure. Painting and sketches by Sartaj Ghuman.  (Raman et al., 2013). There are two terrestrial mammals on the island, including a shrew (Corcidura sp.) and an invasive rat (Rattus sp.).
There are at least four species of bats on the island including a fruit bat species (Pteropus hypomelanus). Narcondam Hornbill (body mass: 600-750 g) is the largest frugivore on the island. Additional details of study area are in ESM Online Resource 1.

| Frugivore and plant abundance
We conducted the study between December 2019 and February 2020. We used variable-width line-transect surveys to estimate the density and abundance of birds on Narcondam Island (Thomas et al., 2010). We walked 20 unique trails on the island (ESM Online Resource 2: Figure S1). We divided the entire elevation gradient into

| Plant-frugivore interactions
To systematically document plant-frugivore interactions, we performed spot sampling during our trail walks, and opportunistically outside our trail walks following Palacio et al. (2016). Whenever we encountered a frugivore on a fruiting tree, we recorded the species identity, number of frugivores, and frugivore foraging behavior (fruits swallowed, pecked or dropped) to ascertain whether the frugivore was a seed disperser or not. Plant-frugivore interac-  Table S3). Plants were classified as small-seeded (seed width: <5 mm), medium-seeded (seed width: 5-15 mm), and large-seeded (>15 mm) following Naniwadekar et al.

| Frugivore and plant abundance
We estimated densities of perched birds with more than 40 visual detections using the R package "Distance" (Miller et al., 2019;Thomas et al., 2010). We estimated combined density of two Imperial-pigeon species (Pied Imperial-Pigeon Ducula bicolor and Green Imperial-Pigeon Ducula aenea), since we had 45 detections overall. Additional details of the analysis are provided in the ESM Online Resource 1.
We estimated the overall tree density and basal area (m 2 /ha), and density of 27 species of fleshy-fruited plants that were fruiting during the study period.

| Island plant-seed disperser community
We created a matrix of plants and frugivores with the frequency of interactions documented for each plant-frugivore combination through spot sampling. Frequency of interactions is the number of times a bird species was observed feeding on a plant species. Every time an individual bird was seen feeding on a plant, it was considered as one interaction irrespective of the number of fruits the bird ate during that interaction. We only considered those animals as seed dispersers which we had seen swallowing the seeds. We did not consider Alexandrine Parakeet Psittacula eupatria in the analysis as we documented the parakeets predating on seeds or feeding on unripe fruits as has also been documented elsewhere (Shanahan et al., 2001). We plotted the interaction accumulation curve using the "random" method and 1,000 permutations as implemented in the R package "vegan" to determine if we had adequately documented the diversity of unique plant-seed disperser interactions (Oksanen et al., 2019).
To determine the relative roles of species in the organization of plant-seed disperser communities, we examined species-level properties like the degree, species strength, and weighted betweenness centrality. Degree is defined as number of mutualistic partners of the focal species, species strength is the sum of level of dependencies of all the partner species on the focal species, and betweenness centrality is a useful measure to identify connector species which connects multiple guilds in a community (Dormann et al., 2008;Martín González et al., 2010). We used the R package "bipartite" for this analysis (Dormann et al., 2008).
We used the general linear model to examine the influence of plant abundance (estimated using the vegetation plots) and fruit type (small-seeded, figs, medium-seeded and large-seeded) on species strength of the 27 plant species that were fruiting during the study period. During exploratory analysis, we noticed that figs consistently had higher species strength values. Since certain fig groups are known to be keystone species in tropical forests and provide key nutrients (Shanahan et al., 2001), we performed the analysis with figs as a separate category. Species strength was log 10 transformed to approximate normality. In the case of frugivores, since there were only seven species and species strength values were not normally distributed, we used Spearman's ρ to examine the correlation between abundance (encounter rate km −1 ) of frugivores and body mass with species strength. We obtained body mass information from Wilman et al. (2014). To determine if the island plant-seed disperser community was organized into distinct sub-communities, we used the modularity analysis. Higher modularity may make the community more resilient to perturbations (Tylianakis et al., 2010). We calculated modularity (Q) using the QuanBiMo algorithm as implemented by the package "bipartite" in R following Dormann and Strauss (2014) . Significance of Q was determined by evaluating the standardized z-scores obtained from null model expectations generated with r2dtable algorithm as implemented by package "bipartite" in R following Dormann and Strauss (2014).

| Comparing the island and the mainland plantseed disperser communities
To determine the differences in the organization of plant-seed disperser communities between island and mainland, we compared plant-disperser communities between the Narcondam Island with that of Pakke Tiger Reserve in north-east India. Information on plant-seed disperser communities is limited from the Asian tropics. However, detailed information on interactions between 47 avian frugivores and 43 plant species is available from Pakke Tiger Reserve in north-east India (Naniwadekar et al., 2019). Pakke is part of the Eastern Himalaya Biodiversity Hotspot. It is about 1,500 km straight line distance north from the Narcondam Island.
Unlike the island community which is relatively young, the mainland site is several million years old. Representative species of all the frugivore genera found on the Narcondam Island are also found in Pakke with species like the Green Imperial-Pigeon, Asian Koel Eudynamys scolopaceus and Common Hill Myna Gracula religiosa being reported from both sites. Several genera of fleshyfruited plants found on Narcondam Island are also found in Pakke.
We examined network-level properties like web asymmetry, connectance (proportion of realized interactions), specialization (H 2 ′) and nestedness (weighted NODF) on Narcondam Island and compared it with the mainland site (Pakke). Values of web asymmetry range between −1 to 1 with negative values indicating higher diversity of plant species as compared to animal species and vice versa . Specialization is a measure of complementary specialization of the entire network with high values indicating that the species are more selective. Weighted NODF is a quantitative measure of nestedness with high values indicating higher nestedness. Significance of these metrics was determined by evaluating the standardized z-scores obtained from null model expectations generated with r2dtable algorithm as implemented by package "bipartite" in R.

| Community assembly on island and mainland
To determine the influence of phylogeny and functional traits on plant-frugivore interactions on the mainland and island site, we used the following approach. The plant phylogeny was pruned from the megaphylogeny (Qian & Jin, 2016) Table S4). We obtained 100 trees from posterior distribution of 10,000 trees based on Ericson All Species backbone phylogeny with 9,9930 OTUs for seven bird species for Narcondam and 47 species for Pakke found in the network (Jetz et al., 2012). The maximum clade credibility tree was used for further analyses as these trees did not vary substantially in topology and branch lengths. We calculated the phylogenetic distances (pairwise distances between species using their branch lengths) using "cophenetic.phylo" in the R package "ape" (Paradis & Schliep, 2018). We used Gower method to estimate pairwise distances among traits of both plants and frugivores using "vegdist" in the R package "vegan." We then followed the novel approach outlined by Bastazini et al. (2017) to determine whether phylogenetic proximity influenced the evolution of species trait and their interactions in the network or whether species traits influence the observed pattern of interactions after removing the potential confounding effect of the phylogeny. We used a weighted matrix of interactions between plants and frugivores. It was frequency of interactions for the island site and visitation rates for the mainland site. We applied this method for Narcondam and Pakke separately.
In addition, we assessed the phylogenetic signal separately in trait data for birds and plants in both sites, Narcondam and Pakke.

| Frugivore and plant abundance on island
We detected six avian frugivores during the trail walks. We had 207 visual detections of Narcondam Hornbill, 71 detections of Asian Koel, and 45 detections of the two Imperial-Pigeon species. We had only two detections each of the Common Hill Myna and Eye-browed Thrush Turdus obscurus. Information on detection probability and mean flock size is summarized in ESM Online Resource 2: Table S5.
Narcondam Hornbill was the most common frugivore on the island and its abundance (mean (95% CI): 1,026 birds (751-1402)) was almost four times that of the Asian Koel (Figure 3a; ESM Online Resource 1: Table S5). The estimated overall mean (95% CI) density of the Narcondam Hornbill was 151 (110-206) birds/km 2 . The mean densities of the Narcondam Hornbill varied between 101-168 birds/ km 2 across the three elevation zones with the lowest mean densities in the highest elevation band (Figure 3b) (ESM Online Resource 2: Table S5).
The mean (

| Island plant-seed disperser community
We documented 752 interactions between seven frugivore spe- on the island during the study period. The interaction accumulation curve indicated that we had adequately documented interaction diversity (ESM Online Resource 2: Figure S2).

| Comparing the island and the mainland plantseed disperser communities
The island network was more asymmetric (−0.588) as compared to the mainland site (0.044). The observed values of connectance and nestedness on Narcondam were almost twice that of the mainland site and specialization (H 2 ′) was almost half that of the mainland site (Table 1). The observed network on Narcondam was less connected and less weighted than the null model predictions but it was more specialized than expected by random (Table 1).

| Community assembly on island and mainland
Among woody plants, we detected a strong phylogenetic signal for seed size and fruit width in Pakke Tiger Reserve (seed size:  which is a point-endemic, plays a critical quantitative role in seed dispersal. This is also one of the few studies that performed a paired comparison between an island and mainland community.

| Narcondam for the hornbill and by the hornbill
There is no other place in the world where hornbills are reported to occur in such high densities as on Narcondam Island (151 birds/ km 2 ). On Sulawesi, the Knobbed Hornbills Rhyticeros cassidix occur in very high densities of up to 84 birds/km 2 (Kinnaird et al., 1996).
In north-east India, Wreathed Hornbill Rhyticeros undulatus, a close relative of the Narcondam Hornbill, seasonally occur in densities of up to 68 birds/km 2 (Naniwadekar & Datta, 2013). Our density estimates are similar to densities estimated by previous studies (Manchi, 2017;Raman et al., 2013). Reduction in hunting pressures may have resulted in the recovery of Narcondam Hornbill population from around 300-400 birds in early 2000s following the suggestions by Sankaran (2000). Among other frugivores, there is limited information on the density of the Asian Koel from other sites, but comparison with existing studies show that Narcondam has very high densities of as compared to other sites (Riley, 2001). The mean Imperial-Pigeon densities are slightly lower than those reported from a mainland site in north-east India (Dasgupta & Hilaluddin, 2012).
In this study, multiple measures (abundance and different network metrics-species strength, degree, betweenness, and modularity) indicate that Narcondam Hornbills are keystone seed dispersers on the island. Generally, as the body mass increases, the abundance of the frugivore and the frequency of visitation is found to decline (Godínez-Alvarez et al., 2020). Interestingly, the Narcondam Island is unique where the largest frugivore, the Narcondam Hornbill, is also the most abundant frugivore on the island. Abundant species have been reported to influence their partners strongly and Narcondam Hornbill had the highest species strength values highlighting its importance for plants (Vázquez et al., 2007). Larger frugivores, like hornbills, are known to remove more fruits per visit as compared to other smaller frugivores (Naniwadekar et al., 2019). Highest number of interactions of the Narcondam Hornbill coupled with potentially higher fruit removal rates point toward the key quantitative role played by the hornbill on the island. Modularity analysis revealed that the hornbill was the key seed disperser for the large-seeded plants as has been demonstrated in previous studies Naniwadekar et al., 2019). The hornbill was the only frugivore that has high betweenness centrality value pointing toward its vital role as a connector between different guilds within the community.
This high reliance on a frugivore and the potentially low redundancy highlights the fragility of the island ecosystem because of strong reliance on a single frugivore.
Narcondam has very high densities of hornbill food plants, which likely explains the high hornbill densities. The overall fig density on Narcondam Island was 27 trees/ha (≥30 cm GBH). This is very high as compared to other sites in the world where the densities of hemi-epiphytic figs are typically <1 tree/ha (Harrison, 2005). Ficus (overall) densities were 1.4 trees/ha at a site in the southern Western Ghats (Page & Shanker, 2020). Ficus densities of up to 10 trees/ha have been reported in Sulawesi (Kinnaird et al., 1996). Consistently across the mainland site and the island site, certain species of Ficus attracted the highest diversity of frugivores. On the island, Ficus represented the top six species in species strength, highlighting the key role they potentially played in the plant-disperser community at least during the study period. Association between canopy, hemi-epiphytic figs, and hornbills has been documented elsewhere (Harrison et al., 2003). Based on published literature on hornbill food plants from north-east (Datta, 2001;Naniwadekar et al., 2019;Naniwadekar et al., 2015) and Thailand (Chaisuriyanun et al., 2011;Kanwatanakid-Savini et al., 2009;Poonswad et al., 1998)

| Plant-seed disperser communities on island and mainland
Island communities are more asymmetric as compared to mainland systems due to higher immigration rates of plants (as compared to frugivores) or human-caused frugivore extinctions on the island human-driven extinctions on the Narcondam island. We are unlikely to find additional species of frugivores that can play an important functional role in seed dispersal on the island as our list of frugivores is comparable to the frugivores reported by the previous studies (Raman et al., 2013). We feel that dispersal limitation of frugivores could also be a key factor in the web asymmetry. Andaman and Nicobar archipelago, of which Narcondam Island is part of, is an interesting example of this. Bulbuls and barbets, which are the key frugivores on the mainland (Naniwadekar et al., 2019), are absent on the remote and tiny Narcondam Island. Endemic species of bulbuls are found on other islands in the Andaman and Nicobar archipelago, but barbets are absent from the archipelago. Thus, the variation in the dispersal ability of different frugivore groups might also play an important role in contributing toward the web asymmetry on oceanic islands. Among other network properties, higher connectance on islands has been reported elsewhere (González-Castro et al., 2012). Forbidden links are thought to contribute toward lower connectance (Olesen et al., 2011).
Additionally, connectance and nestedness are negatively correlated (Rezende et al., 2007). In our case, higher connectance is associated with higher nestedness on the island site as compared to the mainland site. Given that most of the important frugivores on the island are medium (100-500 g) or large-bodied (>500 g) with ability to handle small and large fruits (thereby less likelihood of forbidden links), it has likely resulted in higher connectance and nestedness. Given the small assemblage of frugivores on the island, low competitive pressures and low differentiation on the foraging niche axis of the different frugivores may also result in higher connectance and nestedness.

| Organization of plant-seed disperser communities
We detected contrasting phylogenetic signals in bird body mass, and seed and fruit traits on both the island and the mainland. While frugivore traits were more likely to be similar (than Brownian motion model) for closely related species, plant traits were less likely to be similar for closely related species. However, functional and phylogenetic component did not influence the network interactions in the plant-seed disperser communities on the island and mainland. The role of evolutionary history in shaping local networks of interacting species is thought to be limited (Segar et al., 2020). Unlike the plant-pollinator networks, the seed dispersal networks can be expected to be less-specialized and diffused as unrelated species may exhibit trait convergence and forage on similar plant species (Jordano, 2000). We found that distantly related groups of frugivores dispersed seeds of a common as- F I G U R E 5 Seed dispersal interactions in Narcondam Island (a) and the mainland site, Pakke Tiger Reserve (b). Also shown is the pruned phylogeny for the plants and the frugivores and respective traits for plants (seed size and fruit width) and frugivores (body mass and percentage diet that is comprised of fruits). The matrix is weighted in both the island and mainland case and it has frequency of interactions for the island site and visitation rate of frugivores (h −1 ) for the mainland site shown by the gray bar below the matrix. *Species for which a congener or allied genus was used in the constructing the phylogenetic tree One may argue about the lack of phenological completeness of the study and its influence on the inferences drawn.  (Miranda et al., 2019). Given this, we believe that methodical differences are unlikely to alter the findings of the study.
Only those plants and animals that are able to cross the oceanic barriers and establish contribute to structuring of communities on islands. Degree of isolation, area, and age of islands thereby offers an interesting gradient to understand the relative importance of processes in structuring of communities. While we did not detect a phylogenetic or functional signal across mainland and island communities, it will be interesting to investigate these patterns as the diversity of frugivores increases further. Stronger signals can be expected in a niche packing scenario. As demonstrated in this study, large-bodied frugivores, particularly the Narcondam Hornbill, play a pivotal role on Narcondam Island in the absence of small frugivores on a relatively young and moderately isolated island. Several threatened hornbill species are found on oceanic islands (IUCN, 2019). Narcondam Island and its endemic hornbill give insights into the potential role of hornbills, which are among the largest frugivores on islands, in shaping and potentially maintaining the island tree communities by playing a key role as a seed disperser. The ecological role of hornbills and their contribution to mutualism and maintenance of plant species diversity is poorly understood from other oceanic islands. Narcondam Island through the hyper abundance of hornbills and its food plants and the asymmetric nature of the plant-disperser community with heavy reliance on a single frugivore demonstrates the unique ecological and evolutionary setting of oceanic islands. The disproportionate and singular role that species can play on oceanic islands points toward the extreme vulnerability of oceanic islands to any external perturbations. The perturbations can disrupt the delicate and intricate relationships between plants and seed dispersers, with dramatic cascading impacts on the entire community. Unfortunately, studies like this might not be possible in many other oceanic islands that harbor hornbills as they have been already modified extensively due to anthropogenic activities. It is indeed fortunate that Narcondam has been preserved as a Wildlife Sanctuary thereby safeguarding not only a unique hornbill species but also a unique ecosystem.

ACK N OWLED G M ENTS
We thank Wildlife Conservation Trust, IDEAWILD, Nature We thank two anonymous reviewers for their comments that have helped improve the manuscript.

CO N FLI C T O F I NTE R E S T
Authors have no competing interests to declare.