Disturbance is a fundamental ecological process and driver of population dynamics. Ecologists seek to understand the effects of disturbance on ecological systems and to use disturbance to modify habitats degraded by anthropogenic change. Demographic responses by plants to disturbance are often well described, but demographic responses by animals are less understood. This limits development of applied strategies that leverage disturbance to augment animal populations. We estimated demographic and behavioural responses of an endangered butterfly, Fender's blue, Plebejus icarioides fenderi, to experimental burning in Oregon, USA. We monitored butterfly vital rates for 4 years post-fire. Prescribed fire killed Fender's blue larvae. However, fecundity was higher relative to reference/unburned areas for 2 years after the burn and overwinter larval survivorship was higher for a year after the burn. Fire treatments did not influence adult movement behaviour. We used matrix models to project butterfly population dynamics in fire-driven successional landscapes. We compared optimal burn strategies given targeted burns, such as prescribed fire, to undirected burns, such as wildfires. Disturbance enhances population growth rate under both strategies, and the optimal proportion of landscape burned is similar in both cases. However, targeted burning leads to substantially higher population growth rates. Synthesis and applications. Demographic models allow planning of long-term and large-scale disturbance by balancing initial costs of disturbance with subsequent benefits. We use matrix models to project population growth in fire-driven successional landscapes and contrast prescribed burns with undirected burns (wildfires). We also use these models to evaluate the influence of local vs. non-local dispersal. Because random (non-local) dispersal allows individuals to disperse into areas that were just burned, non-local dispersal always increases population growth rates in this system. This contrasts with source-sink dynamics in stationary environments, in which local dispersal leads to higher population growth rates. Our matrix modelling approach has broad application to other disturbance-dependent taxa surviving in anthropogenically modified landscapes, and could be more widely applied to animal populations.