The devil is in the dispersers: predictions of landscape connectivity change with demography

Concern about the effects of habitat fragmentation has led to increasing interest in dispersal and connectivity modelling. Most modern techniques for connectivity modelling have resistance surfaces as their foundation. However, resistance surfaces for animal movement are frequently estimated without considering dispersal, despite being the principal natural mechanism by which organisms move between populations. We collected Global Positioning System data over 10 years from 50 African lions Panthera leo (11 male natal dispersers, 20 adult males and 19 adult females) and used a path level analysis to parameterize demographic‐specific resistance surfaces for the Kavango Zambezi Transfrontier Conservation Area (KAZA) in Southern Africa. Lion path selection varied according to demographic grouping: adult females were most averse to risky landscapes such as agro‐pastoral lands, towns, areas of high human density and highways. Male natal dispersers were the least‐risk averse suggesting they are potentially the most prone demographic to human–lion conflict. Adults of both sexes selected bushed grassland and shrubland habitats and avoided woodland. Male natal dispersers displayed the opposite trend suggesting con‐specific avoidance and/or suboptimal habitat use. We used the resistance surfaces to calculate factorial least‐cost path networks for each demographic‐specific resistance surface and present results that show substantial differences between predicted patterns of connectivity for male natal dispersers, adult females and adult males. Synthesis and applications. Resistance surfaces are widely used to create connectivity models, which are promoted for use by conservation managers. Our results suggest that the demographic category used to parameterize resistance surfaces may lead to radically different conclusions about connectivity. Failure to include dispersing individuals when parameterizing resistance surfaces intended for connectivity modelling may lead to erroneous conclusions about connectivity and potentially unsound management strategies.