Characterising the spatial distribution and variation of species communities and validating these characteristics with data from the field are key elements for an ecosystem‐based approach to management. However, models of species distributions that yield community structure are usually not linked to models of community dynamics, constraining understanding and management of the ecosystem, particularly in data‐poor regions. Here we use a qualitative network model to predict changes in Antarctic benthic community structure between major marine habitats characterised largely by seafloor depth and slope, and use multivariate mixture models of species distributions to validate the community dynamics. Further, we then assess how future increases in primary production associated with anticipated loss of sea‐ice may affect the ecosystem. Our study shows how both spatial and structural features of ecosystems in data‐poor regions can be analysed and possible futures assessed, with direct relevance for ecosystem‐based management.