When to stay and when to leave? Proximate causes of dispersal in an endangered social carnivore.

Reliable estimates of birth, death, emigration and immigration rates are fundamental to understanding and predicting the dynamics of wild populations and, consequently, inform appropriate management actions. However, when individuals disappear from a focal population, inference on their fate is often challenging. Here we used 30 years of individual-based mark-recapture data from a population of free-ranging African wild dogs Lycaon pictus in Botswana and a suite of individual, social and environmental predictors to investigate factors affecting the decision to emigrate from the pack. We subsequently used this information to assign an emigration probability to those individuals that were no longer sighted within their pack (i.e. missing individuals). Natal dispersal (i.e. emigration from the natal pack) showed seasonal patterns with female dispersal peaking prior to the mating season and male dispersal peaking at the beginning of the wet season. For both sexes, natal dispersal rate increased in the absence of unrelated individuals of the opposite sex in the pack. Male natal dispersal decreased with increasing number of pups in the pack and increased in larger packs. Female natal dispersal decreased with increasing number of pups in larger packs, but increased with increasing number of pups in smaller packs. Individuals of both sexes were less likely to exhibit secondary dispersal (i.e. emigration from a pack other than the natal pack) if they were dominant and if many pups were present in the pack. Our models predicted that 18% and 25% of missing females and males, respectively, had likely dispersed from the natal pack, rather than having died. A misclassification of this order of magnitude between dispersal and mortality can have far-reaching consequences in the evaluation and prediction of population dynamics and persistence, and potentially mislead conservation actions. Our study showed that the decision to disperse is context-dependent and that the effect of individual, social and environmental predictors differs between males and females and between natal and secondary dispersal related to different direct and indirect fitness consequences. Furthermore, we demonstrated how a thorough understanding of the proximate causes of dispersal can be used to assign a dispersal probability to missing individuals. Knowledge of causes of dispersal can then be used within an integrated framework to more reliably estimate mortality rates.