Temporal changes in the structure of a plant-frugivore network are influenced by bird migration and fruit availability

Study area

This study was carried out in a seasonal tropical ecosystem in the Gulf of Mexico, at the “La Mancha” Coastal Research Center (CICOLMA), Veracruz, Mexico. Its coordinates are 19°35′25N and 96°22′49W, and its altitude ranges from 0 to 50 m asl. The climate is hot subhumid with summer rain of the Aw2 type, according to Köppen climate classification modified by García (1981). The minimum and maximum temperatures are 16 °C and 36 °C, respectively; mean annual temperature ranges between 22 and 26 °C, and mean annual precipitation ranges between 1,200 and 1,500 mm (Gómez-Pompa, 1972; García, 1981). Approximately 78% of the total precipitation takes place from June to September. Lower temperature conditions, scarce precipitation and maximum wind speeds are registered from November to February, when frequent cold fronts strike the area (Castillo & Carabias, 1982). The CICOLMA comprises an area of approximately 83.29 ha where several plant communities occur (mangrove, savanna, tropical deciduous forest, flooded deciduous forest, sand dune scrub, and secondary forest <10 y); it is surrounded by farmland and cattle pastures (Dubroeucq et al., 1992; Travieso, 2000).

More than 75% of the original natural vegetation cover in the region has been lost (Portilla, 1996; Ruelas, Hoffman & Goodrich, 2005). Nonetheless, vegetation heterogeneity at the landscape level is associated with high plant and animal richness at CICOLMA, which in turn promotes high diversity of plant–animal interactions (Martínez, García-Franco & Rico-Gray, 2006). CICOLMA has been designated as an area of importance for bird conservation (AICA-02; González-García & Ortiz-Pulido, 1999), and it is part of the main migration route of North American birds (Thiollay & Nocedal, 1978; Thiollay, 1980; Straub & Martínez, 2001). Migratory bird abundance peaks during November–March, while the presence of some migrant species may occur during most of the year (Ortíz-Pulido et al., 1995). Around 299 bird species have been reported at CICOLMA, from which 161 are resident species, 102 are wintering migrants, 31 are transient migrants, and five are intertropical migrants (González-García, 2006). It has been documented that approximately 108 species include fruit as part of their diet, although only 21 species are considered mostly frugivorous (Ortíz-Pulido et al., 1995). A study by Ortíz-Pulido et al. (2000) at CICOLMA registered 176 plant-frugivore interactions between 54 bird species and 33 plant species. It has been described, however, that there are more than 70 plant species whose seeds might be dispersed by birds and that up to 74 species of migratory birds may include fruits in their diet (Ortíz-Pulido et al., 1995), thus increasing the potential number of plant-frugivore interactions.

Plant–bird interactions

Ten samplings were carried out during one year to record plant-frugivore interactions. Samplings took place in November 2013 (Nov13), January (Jan14), March (Mar14), May (May14), June (Jun14), July (Jul14), August (Aug14), September (Sep14), October (Oct14) and November 2014 (Nov14). All interactions between fruiting plants with endozoochorous syndrome and frugivorous birds were registered at four representative habitats at the study site (tropical deciduous forest, flooded deciduous forest, sand dune scrub, and secondary forest >10 years). Because of the small size of the study site and short distance among habitats (<500 m), data for all habitats were pooled and are considered representative of CICOLMA. Plant-bird interactions were sampled with two complementary methods: mist nets and focal observations.

Explanatory variables

Response variables

Two kinds of interaction matrices were built for each of the ten samples: a qualitative matrix (presence/absence data) and a quantitative matrix (interaction intensity data). The interaction intensity between a plant species and a bird species was given by the total estimated number of fruits eaten (García et al., 2014). We chose this metric as a proxy for interaction intensity because we were interested in discussing our results in terms of the potential effectiveness of seed dispersal by birds; in this sense, we believe that the number of fruits is a better proxy of the quantity component of dispersal effectiveness. The number of fruits eaten was obtained by adding the direct count of fruits obtained during focal observations and the extrapolated number of fruits from fecal samples (see above).

Data analyses

Statistical analyses

To evaluate whether periods of high and low proportion of migratory frugivorous bird species were significantly associated with a temporal turnover of bird species at the network core and its periphery, we used a PERMANOVA (Permutational Multivariate Analysis of Variance) (Anderson, 2005), followed by a graphical representation of the results through a non-metric multidimensional scaling (NMDS) ordination. To carry out the analysis and ordination, the ten samplings were a priori grouped into two categories: (i) high proportion of migratory bird species: Jan14, Mar14, Sep14 and Oct14; and, low proportion of migratory birds species: Nov13, May14, Jun14, Jul14, Aug 14 and Nov14. The networks core and periphery were analyzed separately. The PERMANOVA was performed by means of the adonis function from R’s ‘vegan’ package (Oksanen et al., 2012). This function uses the Bray–Curtis index and implements a multivariate variance analysis using the distance between matrices (species composition dissimilarities in each sampling), from which probabilistic significance is obtained through 999 permutations (Anderson, 2005).

We utilized generalized linear models (GLM’s) to find out the extent to which temporary variations in the proportion of migratory species and resource availability influenced plant-frugivore network parameters. Given that the proportion of migratory species and fruit abundance were negatively correlated (Spearman correlation, r_s = − 0.79, P ≤ 0.01), two types of models were explored: (1) relating proportion of migratory species, fruit richness, and their interaction, with network parameters, and (2) relating fruit abundance with network parameters. Poisson error distribution and log link function were used. Data underdispersion (φ < 1) and overdispersion (φ > 1) cases were dealt with by using “quasi-Poisson errors” (Crawley, 2007). All statistical analyses were conducted using R v. 3.2.1.

Temporal turnover in network core and periphery bird species composition

The network core did not showed a turnover in species composition between periods of high and low proportion of migratory species (PERMANOVA, Pseudo-F = 0.93; p = 0.52; (Figs. S1A and S1B). By contrast species composition in the networks periphery was significantly dissimilar between periods with high and low proportion of migratory bird species (PERMANOVA, Pseudo-F = 2.00; p = 0.01).

In general, resident birds were always part of the networks generalist core (Gc), whereas migratory birds were represented in the networks generalist core in five of the ten samplings. Seven resident bird species formed the network core (Melanerpes aurifrons, Pitangus sulphuratus and Psarocolius montezuma being the most common), but only three migratory bird species were part of the core (Dumetella carolinensis, Vireo griseus and Tyrannus tyrannus; Fig. 1). Lastly, resident bird species were predominant at the networks periphery (Fig. 1).

Influence of proportion of migratory species and food availability on network parameters

Network size revealed a positive and significant relationship with the proportion of migratory bird species and with fruit species richness (Figs. 3A and 3B; Table S2A). Connectance was negatively and significantly related to the proportion of migratory birds and fruit richness (Figs. 3C and 3D; Table S2B), and positively related with fruit abundance (Fig. 3E; Table S2B). Additionally, nestedness revealed a negative relationship with the proportion of migratory species and with fruit richness (Figs. 3F and 3G; Table S2C), but a positive one with fruit abundance (Fig. 3H). Interaction strength asymmetry was positively and significantly related only to the proportion of migratory species (Fig. 3I; Table S2E). Network-level specialization (H₂) and niche overlap was not significantly associated with any explanatory variable considered (Tables S2D and S2F). The interaction term between the proportion of migratory bird species and fruit richness had a significant effect on network size (Table S2A), which indicated that larger-sized networks occurred during periods in which the proportion of migratory species was higher but the richness of fruiting species was lower. On the other hand, connectance, nestedness, specialization, interaction strength asymmetry and niche overlap were not affected by this interaction (Table S2).

Temporal turnover in bird species composition of network core and periphery

Our analysis showed that the proportion of migratory species was not related to species turnover in the networks core. However, there was a significant turnover of species in the network’s periphery when comparing periods of higher and lower proportion of migratory species. Thus, core species, compared to peripheral species, displayed less temporal turnover (Bascompte et al., 2003; Dáttilo, Guimarães & Izzo, 2013). Such a pattern could be caused by the abundance of core species, given that the most abundant plant and animal species tend to interact among themselves more than the less connected species, thus contributing to network core stability (Vázquez et al., 2007; Vázquez et al., 2009).

Additionally to abundance, bird behavior (e.g., competitively superior and/or aggressive species) may also be a determining attribute of species composition in the networks core (Dáttilo, Guimarães & Izzo, 2013). For instance, the migratory birds constituting the core of the network in our study site, Vireo griseus, Dumetella carolinensis and Tyrannus tyrannus, are generalist species (i.e., they have many links in the network). Their incorporation into the system is comparable to that observed in invading species, which have been shown to quickly become an integral part of the generalist core of interaction networks (Traveset & Richardson, 2006; Aizen, Morales & Morales, 2008; Díaz-Castelazo et al., 2010). It should be mentioned that the network core was not exclusively composed of migratory birds, but also of larger resident bird species such as Melanerpes aurifrons, Pitangus sulphuratus and Psarocolius montezuma, species quite abundant at the study site (M Ramos, pers. obs., 2014) and displaying territorial behavior (Fitzpatrick, 1980; Husak, 2000; Price, Earnshaw & Webster, 2006).

The relevance of core species is that they may contribute to maintaining network stability by making them more resilient against a variety of disturbances, through an increase in network robustness and cohesion (Bascompte et al., 2003; Díaz-Castelazo et al., 2010; García, Zamora & Amico, 2011; Chama et al., 2013; Nielsen & Totland, 2014; Vidal et al., 2014). Also, it is possible that core species are competitively superior (Dáttilo, Guimarães & Izzo, 2013) and may be more effective as seed dispersers, exerting selective pressure on some fleshy-fruited plant attributes (Ruggera et al., 2015).

Most birds comprising the networks periphery were resident species (27 resident species vs. 17 migratory species), which could be a consequence of these species maintaining more exclusive interactions with plants (Ruggera et al., 2015). In addition, peripheral species could have less territorial behaviors, which could lead to displacement by generalist species from the core (Bascompte et al., 2003; Dáttilo, Díaz-Castelazo & Rico-Gray, 2014).

Influence of proportion of migratory species and food availability on network parameters

Most studies on plant-frugivore networks have used a static approach, describing networks as snapshots in time. Our research demonstrates, though, that plant-frugivore interaction patterns are temporally dynamic and are significantly influenced by factors such as seasonal variation in food availability and the incorporation of new frugivore species into the system.

Fruit abundance

Our data showed that, within a year, fruit abundance was positively associated with network connectance and nestedness. This result suggests that during periods high fruit abundance, networks are more complex and more robust to secondary extinctions (Dupont, Hansen & Olesen, 2003; Heleno, Devoto & Pocock, 2012). It is likely that the nested pattern is mostly driven by highly connected plant species with high fruit abundance (e.g., Bursera simaruba and Ficus spp.), which allows them to maintain interactions with a large number of bird species. The importance of these key plant species to our study site lies in their maintaining numerous generalist interactions, because they provide a continuous food supply for birds. In addition, these species are probably the main contributors to network nestedness and therefore to its robustness against disturbance (Olesen et al., 2007; Tylianakis et al., 2010 Ruggera et al., 2015).

Interestingly, we found a negative correlation between the proportion of migrant species and fruit abundance. This is somewhat contrary to the results of previous studies in which a positive relationship between food availability and the abundance of particular migratory bird species was recorded (Thompson & Willson, 1979; Loiselle & Blake, 1991; Loiselle & Blake, 1992; Jordano, 1994; Rey, 1995; Herrera, 1998; García, Zamora & Amico, 2011; Guitián & Bermejo, 2006; Mulwa et al., 2013). In our study site, however, this negative relationship could have occurred because the periods of lower fruit abundance coincided with periods of greater fruit richness, involving a higher proportion of migratory species during these periods. In addition, studies reporting positive relationships between the abundance of migrants and the abundance of fruit were mostly carried out in more homogeneous landscapes whereas our study site features a mosaic of different vegetation types. Such habitat heterogeneity is probably associated with the high diversity in fruits, which in turn could favor migratory bird species, which can be highly dependent on fruit during migration (Loiselle, 1987).

Fruit richness

As already mentioned above, the periods with highest richness of fruiting species coincided with periods of higher proportions of migratory species. The positive relationship between fruit richness and proportion of migratory species was in turn associated with larger networks, but with lower values of connectance and nestedness. It is probable that an increase in network size may have exerted an influence on network parameters, decreasing connectance and becoming less cohesive.

Also, there was a negative relationship between fruit abundance and fruit richness. We believe that this pattern might emerge from the fact that plant species with different fruiting phenology may also differ in their crop sizes. It seems that in our study site, fruit availability during the dry season was dominated by a few plant species, mostly trees, with very large fruit crops (e.g., Bursera, Ficus). Conversely, during the rainy season, more plant species fruited, but with smaller fruit crops (e.g., lianas, shrubs).

It is known that those networks possessing a higher richness of interacting species compared to interaction-poor communities, tend to be more nested, and therefore, they are more stable and more robust when facing disturbances (Wright & Reeves, 1992; Ulrich, 2009; Atmar & Patterson, 1993). Nonetheless, and contrary to this, we discovered a negative relationship between nestedness, and fruiting plant richness and proportion of migratory birds on the other hand. These unexpected relationships could be ascribed to higher food richness periods and “new” species in the system favoring more selective plant-bird interactions (with fewer links), rendering less connected and nested networks, as has been suggested by other studies (Joppa et al., 2010).

On the other hand, high fruit richness probably influenced resource use diversification by frugivorous birds through the reduction of niche overlap in our study system (Díaz-Castelazo et al., 2013). In fact, other authors have reported that fruit richness may decrease interspecific competition between frugivorous bird species (Blüthgen et al., 2007; Chama et al., 2013), thus reducing diet overlap (Vázquez et al., 2007; Vázquez et al., 2009).

The presence of migratory frugivorous birds in the system “disordered” the plant-frugivore network organization by reducing the nested pattern (Bascompte et al., 2003). Also, network disorder was evident through changes in the interaction strength asymmetry, which was positively related to the proportion of migratory bird species, which could in turn be caused by temporal variations in the abundance of particular bird species (Carciner, Jordano & Melian, 2009). In other words, periods with the highest proportions of migratory species were also the ones with greatest bird abundance (e.g., more than 300 individuals from one species associated with one plant species individual; M Ramos, pers. obs., 2014). During such periods, it is possible that birds depend more on plants, than vice versa.

Other work has shown that the incorporation of many new species into a system (e.g., invasive species) increases interaction asymmetry (Tylianakis et al., 2008), which originates increased interaction diversity and robustness against disturbances (Bascompte, Jordano & Olesen, 2006). This could have significant ecological implications at the community level, given that the most abundant species tend to exert a stronger influence on the species they interact with. This asymmetry of interaction is frequent in nested networks, where species with few connections interact more with the highly connected ones. Nevertheless, González-Castro et al. (2012) suggested that interaction strength asymmetry is not necessarily given exclusively by the abundance of species, but may also be influenced by other factors such as plant phenology and species-specific characteristics.

Previous studies have shown that mutualistic networks tend to have low specialization levels (Jordano, 1987; Blüthgen et al., 2007; Schleuning et al., 2011; Chama et al., 2013;), and it is possible that this characteristic may contribute to network persistence and robustness in case of species extinction (Bascompte & Jordano, 2007). The temporal variations of interaction specialization in our plant-frugivore network, produced values from low to high (0.3–0.8), and such specialization changes were not explained by variability in resource availability or the proportion of migratory species. It is probable that other variables not quantified in the present study, such as fruit attributes (size, weight, color, etc.), as well as bird characteristics (size, foraging strategies, etc.) could be related to the temporal variations in our systems specialization.

Although our study did not evaluate the effectiveness of seed dispersal, our results have implications in this regard. As already mentioned, the nested structure of the network, its size, connectance and interaction asymmetry, were related to the temporal dynamics migratory species and fruit availability. We also mentioned that a nested pattern gives stability to the network by making it more resistant against secondary extinctions. This pattern can be lost during certain periods of time, however, due to temporal dynamics as those described in this study. On the other hand, larger networks that have low connectance, could favor the effectiveness of seed dispersal for the plant species, because its seeds would be dispersed by a greater diversity of interacting bird species (Olesen & Jordano, 2002). In addition, the interaction strength asymmetry in our study, related to the proportion of migratory species, could promote seed dispersal between habitat patches, due to the high consumption of large patchy fruit crops by some species of migratory birds (Jordano, 1982).

To conclude, studies as the present one, showing the temporal dynamics of network parameters could be essential for conservation and management purposes. Fluctuations in resource availability and bird species composition are prone to be affected by anthropogenic disturbances, with cascading effects on network structure and important consequences for ecosystem function.