Introduction
El Niño is one of the main drivers of precipitation fluctuations globally and is responsible for increasing seasonality in the tropics (Wright, Carrasco, Calderon, & Paton, 1999; Holmgren, Scheffer, Ezcurra, Gutiérrez, & Mohren, 2001; Malhi & Wright, 2004). Such responses, however, differ widely among regions (Holmgren et al. 2001); for example, in parts of Central America, El Niño causes floods in the rainforests of the Caribbean coast, but droughts in the Pacific dry forests (Waylen, Caviedes, Poveda, Mesa, & Quesada, 1998). These contrasting effects are critically important as rainfall is a principal factor influencing plant phenology and thus primary productivity. The frequency of strong El Niño/Southern Oscillation (ENSO events) is also expected to increase with climate change (Cai et al. 2014). Indeed, the cycle of 2015-2016 is one of the strongest on record (Jacox et al. 2016). Although responses can be complex (Gunarathne & Perera, 2014; Butt et al. 2015), changes in weather due to El Niño, including both droughts and floods, have been directly linked to fluctuations in fruit production (Wright et al. 1999) with cascading effects for wild animal and plant populations (Wright et al. 1999; Harrison, 2000; Butt et al. 2015). Such impacts of El Niño might be especially important in the humid tropics, where nectarivorous and frugivorous vertebrates perform much of the pollination and seed dispersal; however, these consequences have been little-studied and remain poorly understood (Wright et al. 1999; Fredriksson & Wich, 2006; Wolfe, Ralph, & Elizondo, 2015).
The construction of ecological networks is a useful analytical approach for studying interactions among taxa across ecosystems (Ings et al. 2009). Studies of diverse taxa have suggested that mututalistic networks, those containing plants and frugivorous or nectarivorous animals, commonly show similar properties. In particular, mutualistic networks appear to be highly nested, where interactions involving specialist taxa represent a subset of those involving generalists, and also highly modular, with multiple weakly linked clusters of densely connected taxa (Bascompte, Jordano, Melián, & Olesen, 2003; Olesen, Bascompte, Dupont, & Jordano, 2007; Fortuna et al. 2010; Donatti et al. 2011; Krasnov et al. 2012). Such nestedness and modularlity both increase network stability, and resilience (robustness) to the loss of species from ecosystems, while minimizing perturbations (Memmott, Waser, & Price, 2004; Fortuna et al. 2010; Thébault & Fontaine, 2010). Nestedness is also thought to reduce interspecific competition, thereby allowing more species to coexist (Bastolla et al. 2009).
Here we focus on mutualistic interactions between frugivorous bats and plants in Costa Rica as a model system to assess the impact of ENSO events in different habitats. Bats number over 1,300 species worldwide, of which ~20% feed on nectar or fruit (Kunz, Braun de Torrez, Bauer, Lobova, & Fleming, 2011; Rojas, Vale, Ferrero, & Navarro, 2012; Fenton & Simmons, 2015). In the neotropics, phyllostomid bats are widespread and critically important pollinators and seed dispersers, and, together with frugivorous birds, account for over 80% of the seed dispersal activity (Galindo-González, Guevara, & Sosa, 2000). Previous work suggests bat-plant mutualistic networks in the neotropics are highly connected, nested and robust to plant extinctions, but with low modularity (Mello et al. 2011). Such network structures imply considerable behavioral flexibility that might confer resilience to changes in the environment, yet it is not known how extreme climatic events may affect the structure and robustness of these networks. During the El Niño event of 2015, rainfall levels during the rainy season in the wet forest of Costa Rica exceeded those of the previous 47 years, whereas the opposite trend was observed in the coastal dry forest, where rainfall levels were lower in the rainy season than those of the previous 31 years. Thus, both types of forest experienced extreme climatic conditions associated with El Niño (Seneviratne et al. 2012).
To determine how opposite extremes in rainfall (unusually wet and dry conditions) induced by El Niño influence mutualistic interactions among plants and frugivorous bats, we analyzed and compared networks of mutualistic interactions across the wet and dry seasons in both wet forest and dry forest in Costa Rica. We contrast this to the network of interactions during the wet season of the dry forest during a normal year. Because building networks is extremely labour-intinsive (Evans, Kitson, Lunt, Straw, & Pocock, 2016), which limites replication, null models are common in network ecology and have became one of the main statistical methods to assess the significance of observed network metrics (Delmas et al. 2019). We used null models to estimate the magnitude of the change of the observed network metrics in relation to randomized matrices.
A major challenge of constructing mutualistic networks is characterizing the links between plants and animals (Clare, 2014). Many vertebrate frugivores may feed on fruit pulp, egesting no identifiable material (e.g. seeds) for morphological examination. For these reasons, DNA barcoding, which can be applied to traces of DNA, has proven to be a powerful means of inferring ecological interactions (Clare, 2014; Evans et al. 2016; Roslin & Majaneva, 2016). Such molecular approaches have resolved previously unknown links in already well-studied food webs, revealing metrics such as connectance and nestedness to differ by orders of magnitude from earlier estimates derived from traditional approaches (Wirta et al. 2014). On the other hand, most studies using molecular tools to analyze animal diets have focused on predation (Jedlicka, Sharma, & Almeida, 2013; Brown et al. 2014; Clare et al. 2014; Kruger, Clare, Symondson, Keiss, & Petersons, 2014; Chanin et al. 2015) with fewer studies using DNA barcoding to understand plant-mammal mutualisms, though this is rapidly changing (Bradley et al. 2007; Quéméré et al. 2013; Kartzinel et al. 2015; Galimberti et al. 2016; Clare et al. 2019).
Using a molecular approach, we measure the impact of seasonality and ENSO on mutualistic interactions of neotropical bats and plants. Earlier work indicates network structure is strongly influenced by precipitation (Trøjelsgaard & Olesen, 2013), including historical climate change (Dalsgaard et al. 2013). In general, higher rainfall and seasonality are correlated with more modular networks (Dalsgaard et al. 2013; Trøjelsgaard & Olesen, 2013; Schleuning et al. 2014), and lower rainfall with greater nestedness (Rico-Gray, Díaz-Castelazo, Ramírez-Hernández, Guimarães, & Holland, 2012), consequences that are likely to result from changes in resource availability. We therefore hypothesize that networks will show higher modularity and lower nestedness in the wet forest than in the dry forest in comparison with the expected differences estimated from the null models. Similarly, within each forest type, we predicted that wetter seasons than normal would have higher modularity and lower nestedness than what would be expected from the null models while drier seasons than usual would have higher nestedness and lower modularity than null models. Additionally, we looked for changes in other network metrics to evaluate the magnitude of the changes in the structure of species interactions in relation with the null models and added comparisons for all network metrics of the wet season in the rainforest during the El Niño year with data from a normal wet season at the same site.