Séminaire d'Hélène Wagner - Airial, INRA-Pierroton

Landscape-scale studies of plant dispersal and colonization-extinction dynamics are rare. We evaluted an ecological network restoration project in the Franconian Alb, Germany, aimed at increasing connectivity among remnant patches of calcareous grassland vegetation. This project was established 20 years ago with extensive baseline data and provided a 'natural experiment' of how habitat specialist plant species respond to changes in habitat connectivity. Occurrence of 48 habitat specialist plants, mostly perennials without persistent seed bank, were assessed from two surveys 1989 and 2009 of calcareous grasslands where traditional management by large-flock sheep herding had been abandoned in the mid 20^th century. We compared differences in colonization and extinction rates between 26 patches that were re-connected by sheep herding since 1989 and 23 patches that remained u! nconnected. Patch occupancy patterns were related to shepherding connectivity for species with and without seed morphological adaptation to zoochory. Molecular genetic data suggest that sheep-mediated seed dispersal may contribute substantially to gene flow. Our results indicate that the ecological network was successful at restoring plant functional connectivity and reducing extinction rates at a community level, though a number of species did not respond to shepherding connectivity and may require targeted conservation measures.

Spatial statistical modeling of plant communities
Two methodological challenges of plant community ecology will be addressed: how to scale from species to communities, and how to model and account for spatial autocorrelation in the data. Species distribution models predict the current or future occurrence of individual species, whereas ordination methods predict changes in species composition along environmental gradients. While neither method explicitly models species associations, ecological theory increasingly focuses on the role of interactions within and between species groups. Species associations are notoriously difficult to quantify because of the prevalence of spatial autocorrelation in ecological data. Spatial statistics and spatial regression methods, on the other hand, often rely on unrealistic assumptions that are difficult to check due to a lack of diagnostic tools, and application to multivariate community data is not straight-forward. Here I present Spatial Component Regression as a framework for scaling spa! tial analysis methods from species to species groups and to communities. I illustrate how the approach is useful for testing hypotheses on interactions within and among species groups and for comparing alternative models of functional connectivity between plant communities.