The ability of species to stably coexist is fundamental to how diversity is maintained in natural communities. Theoretical models of coexistence, however, have largely been developed to provide proof of concept for pairs of species and ignore the great complex found in most natural communities in terms of both the environment and community composition. These models, though useful for advancing theory, are limited for understanding and predicting diversity in real natural communities. In this talk I present on two distinct approaches to adding biological realism into models for predicting coexistence outcomes in diverse natural plant communities. All approaches presented in this talk are tested using empirical data from the annual plant communities of the York gum woodlands of SW Western Australia. First, I use a Bayesian statistical framework for including demographic uncertainty in predictive models of coexistence. Second, I explain a novel interaction network framework that offers great potential for integrating positive and negative species interactions in models for predicting coexistence and community diversity. Both approaches represent steps forward in linking coexistence theory to applications in the real world, by increasing the biological realism incorporated into coexistence and local
diversity models.