Abstract Metapopulations are a central concept in ecology and conservation biology; however, estimating key parameters such as colonization rates presents a substantial obstacle to modelling metapopulations in many species. We develop spatial and non-spatial simulation models that combine incidence- and demographic-based approaches to build a relationship between observed patch occupancy, habitat turnover rates, colonization rates and dispersal scales. Applying these models to long-term observations of Pedicularis furbishiae (Furbish’s lousewort), a rare plant endemic to the Saint John River, we predict that observed habitat patches averaging 550 m in length receive colonizing seedlings with a yearly probability of 0.45 or 0.54, based on two different models. Predictions are consistent with a standard analytic metapopulation formulation modified to partition extinction drivers during the early and the late phases of a population’s life cycle. While the specific results rest on several simplifying assumptions, the models allow us to understand the impact that increasing rates of habitat turnover would have on the future survival of this species.