Identifying where animals come from during population recovery can help to understand the impacts of disturbance events and regimes on species distributions and genetic diversity. Alternative recovery processes for animal populations affected by fire include external recolonization, nucleated recovery from refuges, or in situ survival and population growth. We used simulations to develop hypotheses about ecological and genetic patterns corresponding to these alternative models. We tested these hypotheses in a study of the recovery of two small mammals, the Australian bush rat and the agile antechinus, after a large (>50,000 ha), severe wildfire. The abundance of both species was severely reduced by fire and recovered to near or above pre-fire levels within two generations, yet we rejected a hypothesis of recovery by external recolonization. While the agile antechinus showed genetic evidence for far greater dispersal capacity than the bush rat, neither species showed gradients in abundance or genetic diversity with distance from unburnt forest during population recovery. Population recovery was driven by local-scale processes. However, the mechanisms differed between species, resulting from the spatial impacts of fire on habitat suitability. Agile antechinus populations recovered through population growth from in situ survivors. The bush rat followed a model of nucleated recovery, involving local recolonization from micro-refuges in topographic drainage lines. Nucleated recovery by the bush rat was associated with changes in dispersal, and fine-scale patterns of genetic admixture. We identified increased dispersal by females during recovery, contrasting with male-biased dispersal in unburnt forest. Such flexibility in dispersal can potentially increase recovery rates compared to expectations based on dispersal behavior within undisturbed populations. Our study shows how the initial distribution of survivors, determined by fire effects on resource distribution, determines the subsequent scaling of population recovery patterns, and the sensitivity of population distribution and genetic diversity to changing disturbance regimes.