Evaluating hypotheses about dispersal in a vulnerable butterfly

Sound management of species requires reliable estimates of dispersal. Indeed, dispersal of individuals among local populations is a key factor in the biology and persistence of local populations and metapopulations. Here, the small-scale dispersal pattern of a vulnerable species, the endemic Sardinian chalk hill blue butterfly, was studied by applying capture-recapture multistate models and a model selection based on AIC values. Model parameters were survival, capture and movement probabilities. The model selection showed that (a) survival probability of individuals varied between sexes, (b) capture probability varied between sexes and among patches, and (c) movement probability varied with direction. The probability of movement among adjacent local populations was generally low and ranged from 0.009 to 0.212. Movement probabilities were subsequently modeled using data on interpatch distance and donor patch population size or area. The ultrastructural "biology-based" models turned out to be the most appropriate models for inference, showing that dispersal decreases with increasing interpatch distance and increasing donor patch population size or area, and suggesting that butterfly dispersal is affected by patch geometry and the presence of conspecifics. The application of multistate models, the model selection approach, and ultrastructural modeling allowed testing the validity of some general hypotheses related to dispersal in metapopulations and helped elucidate the butterfly small-scale dispersal pattern.