In the face of biodiversity loss, we rely upon measures of diversity to describe the health of ecosystems and to direct policymakers and conservation efforts. However, there are many complexities in natural systems that can easily confound biodiversity measures, giving misleading interpretations of the system status and, as a result, there is yet to be a consistent framework by which to measure this biodiversity loss. Ecosystems are governed by dynamic processes, such as reproduction, dispersal and competition for resources, that both shape their biodiversity and how the system responds to change.Here, we incorporate these processes into simulations of habitat and environmental change, in order to understand how well we can identify signals of biodiversity loss against the background inherent variability these processes introduce. We developed a tool for Ecosystem Simulation through Integrated Species Trait-Environment Modelling (EcoSISTEM), which models on the species-level for several sizes of ecosystem, from small islands and patches through to entire regions, and several different types of habitat. We tested a suite of traditionally-used and new biodiversity measures on simulated ecosystems against a range of different scenarios of population decline, invasion and habitat loss. We found that the response of biodiversity measures was generally stronger in larger, more heterogeneous habitats than in smaller or homogeneous habitats. We were also able to detect signals of increasing homogenisation in climate change scenarios, which contradicted the signal of increased heterogeneity and distinctiveness through habitat loss.