Abstract
Dispersal is a key driver of spatial population dynamics. Dispersal behaviour may be shaped by many factors, such as mate-finding, the spatial distribution of resources, or wind and currents, yet most models of spatial dynamics assume random dispersal. We examined the spatial dynamics of a day-flying moth species (Arctia virginalis) that forms mating aggregations on hilltops (hilltopping) based on long-term adult and larval population censuses. Using time-series models, we compared spatial population dynamics resulting from empirically founded hilltop-based connectivity indices and modelled the interactive effects of temperature, precipitation and density dependence. Model comparisons supported hilltop-based connectivity metrics including hilltop elevation over random connectivity, suggesting an effect of hilltopping behaviour on dynamics. We also found strong interactive effects of temperature and precipitation on dynamics. Simulations based on fitted time-series models showed lower patch occupancy and regional synchrony, and higher colonization and extinction rates when hilltopping was included, with potential implications for the probability of persistence of the patch network. Overall, our results show the potential for dispersal behaviour to have important effects on spatial population dynamics and persistence, and we advocate the inclusion of such non-random dispersal in metapopulation models.