Dispersal on the landscape/seascape scale may lead to complex spatial population structure with non-synchronous demography and genetic divergence. In this study we present a novel approach to identify subpopulations and dispersal barriers based on estimates of dispersal probabilities on the landscape scale. A theoretical framework is presented where the landscape connectivity matrix is analyzed for clusters as a signature of partially isolated subpopulations. Identification of subpopulations is formulated as a minimization problem with a tuneable penalty term that makes it possible to generate population subdivisions with varying degree of dispersal restrictions. We show that this approach produces superior results compared to alternative standard methods. We apply this theory to a dataset of modeled dispersal probabilities for a sessile marine invertebrate with free-swimming larvae in the Baltic Sea. For a range of critical connectivities we produce a hierarchical partitioning into subpopulations spanning dispersal probabilities that are typical for both genetic divergence and demographic independence. The mapping of subpopulations suggests that the Baltic Sea includes a fine-scale (100600 km) mosaic of invisible dispersal barriers. An analysis of the present network of marine protected areas reveal that protection is very unevenly distributed among the suggested subpopulations. Our approach can be used to assess the location and strength of dispersal barriers in the landscape, and identify conservation units when extensive genotyping is prohibitively costly to cover necessary spatial and temporal scales, e.g. in spatial management of marine populations.