Abstract
The fact that many species are able to coexist while competing for the same resources – for instance tree species in tropical forests – remains a major conundrum in community ecology. While it has traditionally been approached using species mean characteristics, intraspecific variability (IV) has recently gained renewed interest, notably with evidences of the large amount of variability that IV represents in a community. Since then, IV has been predominantly represented in modeling studies through independent random draws in a probability distribution. In doing so, IV is implicitly assumed to be a noise with no structure in space or time that blurs the differences between species. This assumption led to inconsistent results about consequences of IV on species coexistence. The main objective of this PhD is to provide insights into the nature and structure of IV, on the way it is considered in ecology - particularly in community dynamics models - and on its consequences for species coexistence. In a first chapter, we first illustrate how observed IV can arise from environmental variations in many dimensions that are imperfectly characterised, and not necessarily from genetic variations between individuals of the same species. This can lead to spatio-temporal structure in IV if these environmental drivers are structured, and does not necessarily make species niches overlap. We then combine empirical evidence that IV can be due to non-genetic factors, that IV is structured in space, and that although a high IV observed, species performances are not similar. This type of observed IV, which is spatio-temporally structured, is thus not a coexistence mechanism per se but is the signature of the response of species to a multidimensional environment that varies in space and time. We argue that this type of IV is often not well accounted for in community dynamics models. In a second chapter, using a simple community dynamics model, we test the effect of the way IV is introduced on the composition of the simulated communities, including the number of coexisting species. We show that simple random distributions are not a good proxy for IV in systems with environmental structure. In a third and final chapter, using a simulation approach as well, we show that species differences and variation in many environmental dimensions together lead to a high number of coexisting species, and translates into a particular structure of individual performance correlations where conspecific individuals perform more similarly than heterospecifics. We then show how introducing this structured IV in models enables the coexistence of a high number of species thanks to high-dimensional niche partitioning, even though the drivers of this structure remain unknown and are not included in the model. Overall, we hope to trigger a renewed way to consider and represent IV in community ecology, and provide perspectives for additional tests on the structure of IV and for its integration in community dynamics models.