ALBERT

Description: In plants, the presence of a seed bank challenges the application of classical metapopulation models to above-ground presence surveys; ignoring seed bank leads to overestimated extinction and colonization rates. In this article, we explore the possibility to detect seed bank using hidden Markov models in the analysis of above-ground patch occupancy surveys of an annual plant with limited dispersal. Patch occupancy data were generated by simulation under two metapopulation sizes ( N = 200 and N = 1000 patches) and different metapopulation scenarios, each scenario being a combination of the presence/absence of a one-year seed bank and the presence/absence of limited dispersal in a circular 1-dimension configuration of patches. In addition, because local conditions often vary among patches in natural metapopulations, we simulated patch occupancy data with heterogeneous germination rate and patch disturbance. Seed bank is not observable from above-ground patch occupancy surveys, hence hidden Markov models were designed to account for uncertainty in patch occupancy. We explored their ability to retrieve the correct scenario. For 10 years surveys and metapopulation sizes of N = 200 or 1000 patches, the correct metapopulation scenario was detected at a rate close to 100%, whatever the underlying scenario considered. For smaller, more realistic, survey duration, the length for a reliable detection of the correct scenario depends on the metapopulation size: 3 years for N = 1000 and 6 years for N = 200 are enough. Our method remained powerful to disentangle seed bank from dispersal in the presence of patch heterogeneity affecting either seed germination or patch extinction. Our work shows that seed bank and limited dispersal generate different signatures on above-ground patch occupancy surveys. Therefore, our method provides a powerful tool to infer metapopulation dynamics in a wide range of species with an undetectable life form. This article is protected by copyright. All rights reserved.