Landscape context drives within-species variation in dispersal kernels and limits transferability

Dispersal is a key ecological trait that ensures connectivity, gene flow, and range dynamics, yet empirical information about the dispersal process remains scarce. While interspecific differences have been synthesized, intraspecific variation driven by individual movement and landscape context remains poorly understood. Here, we use the individual-based modelling platform RangeShifter to simulate movement trajectories of four animal species, from insects to mammals, across neutral landscapes varying in habitat amount and fragmentation. From these simulations we derived dispersal kernels using five probability density functions and analysed their response to landscape context. Across species, log-normal (fat-tailed) kernels consistently best described dispersal, but both median and long-distance dispersal varied strongly with habitat configuration. Habitat loss consistently reduced abundances, whereas fragmentation per se more strongly shaped dispersal kernels, providing mechanistic insight into a central debate in landscape ecology. Our results highlight that dispersal kernels do not transfer reliably across environments and should be treated as emergent, context-dependent properties. Progress in predicting species responses to global change will require empirical studies explicitly integrated with mechanistic modelling to uncover generalizable links between dispersal, traits, and landscape context.