Discussion
Our results indicated that the network of interactions between snakes and their resources in a species-rich Amazonian community presented a combination of both nested and modular structure. Nestedness was related to average body mass of snakes, in which boid snakes connect food modules in the trophic network. The modular pattern, in turn, is associated to the different snake lifestyles, in which snakes that share similar habits usually consume similar resources available in their shared microhabitats.
The observed connectance of the network indicated that, given the variety of resources available in the environment, snake species consumed only a subset of these resources. This result suggests that most food resources may not be accessible to most species, suggesting forbidden interactions (Olesen et al. 2010) associated with possible restrictions related to lifestyle (see Savitzky 1983), as well as body size (Sinclair et al 2003, Woodward et al. 2005, Stouffer et al. 2011). For instance, arboreal snakes have morphological adaptations, such as a slender body and long tail, which may represent limitations to the consumption of large prey such as mammals (Martins et al. 2001, Alencar et al. 2013, Alencar et al. 2017). The analysis of network structure revaled that the patterns of resource use by different species lead, at the community level, to nestedness and modularity. Our results contrast with some studies on antagonistic networks that indicate opposite trends between nestedness and modularity (Thébault and Fontaine 2010, Pires and Guimarães 2012). Having said that, other studies simultaneously show levels of nestedness and modularity (Bellay et al. 2011, Flores et al. 2013, Pinheiro et al. 2019). The emergence of these combined network patterns is possible due to the low connectivity of the network (Lewinsohn et al. 2006, Fortuna et al. 2010) and resource heterogeneity (Pinheiro et al. 2019) in Amazonian forests.
Several processes may explain the nested patterns, such as variations in species abundances (Lewinsohn et al. 2006). One of the explanations for the nested pattern found in our study was the large size variation among species present in the network. The variation in snake body mass has led to a trophic hierarchy in which larger predators prey upon more resource items than smaller predators. This hierarchy was detected in several predator-prey interaction networks found in nature (Woodward et al. 2005, Smith and Mills 2008, Arim et al. 2010). This pattern indicates that predators have the potential to add resources sequentially as they increase in size, although this increment of larger resources may lead to the rejection of smaller, less nutritious or difficult to handle resources (Mittelbach 1981, Arnold 1993, Arim et al. 2010, Woodward et al. 2010). In addition to body size, skull morphology is also an important feature associated with diet and snake lifestyle (Pough and Groves 1983, Savitzky 1983, Klaczko et al. 2016). The larger the head of a snake, the greater the prey consumed (King 2002). Thus, future research that investigates emerging patterns arising from the association of both body size and skull morphology with the structure of trophic interaction networks could predict the processes, at the community-level, involved in the relationships between snakes and their food resources.
When analysing the contribution of each species to nestedness, we found that average body mass has a phylogenetic signal, with large species concentrated in a few clades. After removing larger species, mostly boids, the nestedness value decreases 8.78% although it still remains significant. The maintance of nestedness after the removal of large snakes might be a consequence of the number of resource - body mass association holds for smaller snake species, such colubrid and dipsadid snakes. Boids are efficient constrictors with generalist diets, that occupy diverse microhabitats, which allow them to consume a wide variety of food resources (Pizzato et al. 2009, Henderson and Pauers 2012). This combination of features may simultenouly explain why (i) boids act as hubs (species with many interactions) in the analysed network, and (ii) the decrease in nestedness when boids are removed from the network. Large predators, such as sharks, killer whales, lions, and birds of prey, often prey on diverse array of species (Sinclair et al. 2003), potentially connecting modules in networks (e.g., Rezende et al. 2009). Future research could test whether the presence of such large predators can also promote nestedness on predator-prey interaction networks.
The modular structure in ecological networks may be associated to factors such as the degree of specialisation among interacting species (Prado and Lewinsohn 2004, Lewinsohn et al. 2006), habitat heterogeneity (Pimm and Lawton 1980), the phylogenetic relationship between species (Lewinsohn et al. 2006), the convergence in a set of species traits (Olesen et al. 2007) or by a combination of factors (Donatti et al. 2011). We found that the consumption of specific resources is associated with more peculiar lifestyles. For instance, morphological adaptations to fossorial habit (e.g. less cranial mobility) hinder the consumption of prey larger than the snake’s head size (Greene 1983, Savitzky 1983, Martins and Oliveira 1993). Accordingly, arboreal habits impose physical limitations on snake morphology and may restrict the consumption of larger prey, such as small mammals, favoring a diet based on lizards and/or frogs (Lillywhite and Henderson 1993, Martins et al. 2001, Martins et al. 2002, Alencar et al. 2013). Thus, we suggest that the modularity of the network we studied has emerged from the relationship between the lifestyles of snakes and the consumption of resources restricted to the habitats used by the species.
To sum up, we integrate network structure analyses with species removal simulations to evaluate the role of different snake species in the structure of a rich Amazonian snake community, and the mechanisms underlying the patterns found. We encourage future studies to focus on understanding how community phylogenetic diversity may be associated with the modular structure (Rezende et al. 2009), as well as how the combination of traits associated with predator diet (e.g. its correlation with body size and skull shape) may contribute to the nested pattern and if geographic variation (environment type) can modify network structure (Pimm and Lawton 1980, Kortsch et al. 2019). This study points to the joint importance of the evolutionary history of lineages, body size, and their interacting resources to determine the structure, at the community scale, of the interactions between consumers and their resources.