The relative importance of local and landscape scale factors to chalk grassland butterflies

The relative importance of habitat and landscape for chalk grassland butterflies was investigated among a network of chalk grassland habitat patches. Local scale habitat selection models were derived from analyses on a single large chalk grassland site (Porton Down). Landscape scale models were . then derived from species-specific local scale variables added to variables measured in the patch network, which included Porton Down as a major source site. An investigation was also carried out as to whether landscapes cale processes were acting within Porton Down. Models showed that landscapes cale processes may be acting within the single, large site for some species. However, lack of patch definition and natural variations in local population densities may give rise to apparent landscapes cale processes within this site which fundamentally differ from true landscape scale processes. At the true landscape scale across the patch network, patch area was the most commonly included landscape scale variable in models, although only for five species. For one species, dark green fritillary Argynnis aglaja, distance from the major source site and average distance from other sites were of over-riding importance. For most species, local scale habitat and environment variables were sufficient to describe their presence or density among the patch network. Incidence Function Models were used to estimate metapopulation dynamic parameters for seven species, Both sensitivity to environmental stochasticity and colonisation ability were correlated with species' scores on canonical community analysis axes (related to local habitat, patch area, distance from source site and inter-patch'distance measures). This suggests that species conforming to conventional metapopulation models form an ecological continuum with those responding to local scale habitat and landscape scale processes in a less prescriptive fashion. Species groups with either restricted or widespread distributions appear to conform to a landscape connectivity model. Five ecologically similar species appear to be moving from the former to latter group across a 'connectivity threshold', possibly due to recent increases in larval food-plant availability in the wider landscape