Landscape fragmentation is a major threat to biodiversity. It results in the transformation of continuous (hence large) habitat patches into isolated (hence smaller) patches, embedded in a matrix of another habitat type. Many populations are harmed by fragmentation because remnant patches do not fulfil their ecological and demographic requirements. In turn, this leads to a loss of biodiversity, especially if species have poor dispersal abilities. Moreover, landscape fragmentation is a dynamic process in which patches can be converted from one type of habitat to another. A recently created habitat might suffer from a reduced biodiversity because of the absence of adapted species that need a certain amount of time to colonize the new patch (e.g. direct meta-population effect). Thus landscape dynamics lead to complex habitat spatiotemporal structured, in which each patch is more or less continuous in space and time. In this study, we define habitat spatial structure as the degree to which a habitat is isolated from another habitat of the same kind and temporal structure as the time since the habitat is in place. Patches can also display reduced biodiversity because their spatial or temporal structures are correlated with habitat quality (e.g. indirect effects). We discriminated direct meta-community effects from indirect (habitat quality) effects of the spatiotemporal structure of habitats on biodiversity using Collembola as a model. We tested the relative importance of spatial and temporal structure of habitats for collembolan diversity, taking soil properties into account. In an agroforested landscape, we set up a sampling design comprised of two types of habitats (agriculture versus forest), a gradient of habitat isolation (three isolation classes) and two contrasting ages of habitats. Our results showed that habitat temporal structure is a key factor shaping collembolan diversity. A reduced diversity was detected in recent habitats, especially in forests. Interactions between temporal continuity and habitat quality were also detected by taking into account soil properties: diversity increased with soil carbon content, especially in old forests. Negative effects of habitat age on diversity were stronger in isolated patches. We conclude that habitat temporal structure is a key factor shaping collembolan diversity, while direction and amplitude of its effect depend on land use type and spatial isolation.