In this thesis, we introduce and explore a mathematical framework in which to study the evolution of and within ecological networks. Hence, we focus on a peculiar interpretation of the “biodiversity” concept, namely one that includes the complex pattern of interactions among species, along with the species abundancy and evolutionary distinctiveness. Our objects of inquiry are species communities as complex wholes. Classically, communities have been approached from two distinct points of view: on one hand, we can consider the graph describing the energy flows among species in an ecosystem (i.e., an ecosystem’s food web); on the other hand, we can consider the species’ phylogeny, the tree graph describing the evolutionary relatedness of those species. The structure of an ecosystem (its biological diversity and the topology of its interactions) is the product of fast ecological processes within food webs and of the long-term evolutionary processes that give shape to the tree of life. In particular, early ecological literature recognized that the evolutionary history of a species (or its taxonomical classification, in the pre-Darwin era) helps to determine the species’ role as part of an ecological network of interacting species. Conversely, the “ghost of past competition” and arms races are famous examples of the fact that a species’ interactions with its resources and consumers helps to determine the evolutionary trajectory of that species. As the empirical research presents strong evidence that the ecology-evolution (eco-evo) feedback loop is, indeed, significant, the ecological and evolutionary points of view are laboriously being connected more and more strongly. A theoretical framework has been developed for some important scenario (e.g., the co-evolution of hosts and parasites, butterfly and flowers, or plants and pollinators). The case of complex food webs, where is not possible to distinguish two neatly separated trophic layers, has resisted such a treatment. We argue that this can be partially addressed by moving from a rigidly binary view of food webs to the representation of species interactions in a continuous metric space, where species evolution can be gradual. In Chapter 3 we show how this metric space representation of a food web can be estimated efficiently and gather insights about the evolutionary signature of food webs. Species’ ecological interdependency, arising from their role as part of complex food webs, is something that the classic model of trait evolution has avoided. One reason is that it is hard to give a model determining the presence (and strength) of species interactions throughout their history. In Chapter 3 Appendix we show how the metric space representation of food webs may constitute a suitable environment in which to define such a model. Assessing species’ contribution to biodiversity is an important task that scientifically informs conservation efforts. In Chapter 4 we define a family of measures 6 assessing the relative ecological importance of a species in a food web. These measures are defined directly on the food web’s metric space representation we propose in the first chapter. We explore the relationship between evolutionary and ecological uniqueness. In Chapter 5 we tackle the “mode” of food web evolution more directly exploiting, once again, the functional trait representation of food webs. In particular, we formulate two contrasting hypotheses on the evolution of frugivore birds’ functional ecological niches and test it on a dataset of frugivore birds in the Andes. Finally, in Chapter 6 we make a little detour from food webs and consider a different kind of ecological network: geographically grounded population networks composed of patches and corridors among patches. Population networks play a crucial role in evolution (e.g., determining the dynamics of genes’ flows). The insight we gained throughout the previous work (especially in the second chapter) supports the notion that the relevance of a species in a network is not always perfectly captured by the species’ local properties (such as its number of connections). In this spirit, we assess the importance of a patch in a geographic network by the global effect that removing that patch has on the whole network. All the code and data used in this thesis will be available on a public Github repository (see gvdr.github.io).