Year: 2026
Type: preprint
Abstract: Dispersal is central to population dynamics, species distributions, and evolution, yet it is notoriously difficult to measure. The standard approach to estimating components of dispersal involves fitting a kernel to the straight-line distances between where individuals started and where they were next observed. In reality, however, an individual can only be observed where sampling takes place, so observed distances reflect both the movement process and the sampling design, a confounding that is widely recognised and biases inference. We propose the destination selection function (DSF), a model for redistribution data in which the origin, the destination, and the full set of locations at which the individual could have been detected, are known. Each candidate destination exerts a distance-dependent hazard on the individual, modified by its attractiveness, and the observed destination is the outcome of these competing hazards. The selection function represents alternative locations directly from the sampling design, and it links the multi-catch observation model of spatial capture-recapture, in which the detectors are known but the source is latent, with resource- and step selection functions, in which the source is known but the alternatives are assumed. Individual covariates can be used to model variation in the dispersal scale and destination covariates can be used to quantify destination attractiveness, and least-cost distance through an estimated resistance surface can be used to relax Euclidean assumptions about movement. Simulations show that conventional kernel estimators are negatively biased, with the bias related to the extent, and to a lesser extent, sampling intensity, whereas the DSF is unbiased and recovers individual, destination, and landscape effects. We apply the DSF to data from an American mink removal programme and show that dispersal of this semi-aquatic habitat specialist follows watercourses and that this model outperforms a model that assumes isotropic movement.
DOI: 10.32942/x23d7b
Language: English
Subfield: Ecology
Field: Environmental Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225867656
