As topologias de redes de interações ecológicas e suas origens

Abstract

Network science is a powerful tool to study ecological interactions and produced several discoveries, as the existence of prevalent topologies, mainly nestedness and modularity. Some studies argue that these topologies result from selection against unstable networks, whereas others propose that they likely emerge from processes driving the interactions between pairs of species. In the first chapter of this thesis, we developed a model that simulates the evolution of consumer species using resource species, following rules based on the integrative hypothesis of specialization. Without applying selection on stability, our model produced all the topologies commonly observed in species interaction networks. Simulations with homogeneous resources resulted in highly generalized nested networks, while simulation with high resource heterogeneity resulted mainly in networks with compound topologies: modular with internally nested modules. In the second chapter, through logical arguments, we show that nestedness results from marginal sum inequalities in matrices without row-column preferences. Processes that increase inequalities in the interaction frequencies of species, without introducing preferences, promote nestedness. Unequal abundances, fitness and detectability, thus, may cause nestedness. Additionally, the new perspective allowed us to related null models with explicitly hypotheses on nestedness: the equiprobable model is made by random topology matrices, while the proportional model is made by nested matrices. From the evidences obtained in both chapters, we conclude that the processes driving species interactions are able to produce the main topologies observed in real-world species interaction networks. Modularity emerges from the adaptive trade-off faced by species in diverse systems, while nestedness results from processes that promote inequalities in the interaction frequencies of species. For last, the interaction between those processes may produce networks with compound topologies.