The extent to which a resource (e.g. nest site, food resource) is available and utilised in the wild is of fundamental importance to the understanding of species’ biology, community ecology, and for evidence-based conservation and habitat management. However, dynamics of resource use is challenging to study because it is likely determined by a complex set of (often unknown) factors and their interactions, and because of imperfect and biased detection rates (usually false absences). In this study we test for multiple drivers of summer roosting site occupancy by bats, using a rigorous statistical framework that combines multi-state occupancy models with two-class heterogeneity models. Understanding the utilization of this critical resource is a pre-requisite for any successful conservation management of (often declining) bat populations. We evaluated the importance of several external and internal sources of inter-annual roost occupancy dynamics in a community of three sympatric woodland bat species: weather, roost spatial position, roost intrinsic (unmeasurable) quality, memory effect in roost use, and inter-species dynamics. We identified the existence of two classes of the roost’s intrinsic quality, and showed that bats tend to re-occupy previously occupied roosts irrespective of the roost intrinsic quality (memory effect). At the landscape scale, distance from water but not from the woodland edge influenced occupancy dynamics, with roosts closer to water having higher probability of being occupied. We also identified species-specific patterns of roost occupancy, colonisation and extinction, and between-species roost interchange. Our statistically powerful and flexible modelling approach can distinguish between multiple drivers of resource occupancy: measurable factors (environmental conditions and resource traits), unmeasurable (hidden or unknown) factors, previous use of the resource (memory effect) and community-level interactions. As such, it can be used to identify important areas for species (or communities), as well as to test evolutionary and ecological hypotheses concerning resource utilisation.