Complexity in the entangled bank : On the structure and dynamics of empirical mutualistic networks

Overview & Synopsis

Mutualistic relationships, that in the past had been long overlooked as fascinating but marginally relevant, are today known to play a crucial role in shaping ecosystems. In this thesis we look into the complexity of how these ecological relationships are intertwined in natural systems, or what Darwin famously called the ‘entangled bank’, from the viewpoint of the network formalism.In the first part of the thesis, we investigate the origin of the architecture of mutualistic networks. In detail, by applying concepts from information theory and statistical physics, we address the question of the emergence of a widespread pattern known as nestedness. By analyzing a large dataset of empirical networks, we show that the interplay of a few minimum assumptions on the number of mutualistic interactions per species and the effect of chance are sufficient to reproduce the observed structure, precluding the introduction of selective pressures or mechanistic processes. In this sense, our results show that the global structure of mutualistic communities can be explained, in statistical terms, by the lower-order features of the system. With these results at hand, we then explore how the different metrics proposedin the literature quantify nested patterns, evaluating their overall performance on real and synthetic networks. Our results indicate that the ranking and comparison of nested patterns among different ecosystems is hampered by the presence of undesired dependencies on other network parameters.In the second part of this thesis, we continue examining the organization of mutualistic communities but tackling another challenge, namely that of moving beyond the still-prevailing aggregated paradigm. To start with, we characterize a set of empirical networks and assess how incorporating information about the temporal variability modifies the static network description. Next, we propose a group of models to generate, under diverse assumptions, synthetic configurations of phenology for a given network. We find that, while the adequacy of mechanistic models to produce realistic configurations is highly system-dependent, a statistical model based on the maximum entropy principle performs generally well independently of the network details. Elaborating further upon this line of thought, we then briefly explore the dynamical consequences for species persistence of taking into account the phenology. We find that species with a short period of activity face a larger uncertainty in their robustness against perturbations. This preliminary approach, though, calls for further research, specially in the context of a changing climate. On the whole, along this thesis we analyze how the network language can be used to disentangle the complexity of natural mutualistic systems, by assessing on the one hand the minimum information required to understand the ‘entangled bank’, and on the other hand, identifying the limitations of the static representation that still predominates in the field.