Population viability analyses for butterflies typically use metapopulation models, but for endemic species with no redundancy among subpopulations, we need to understand local population dynamics. However, little is known about the sensitivity of butterfly population vital rates and viability to disturbances such as fire. We fit quadratic models to monthly butterfly count data (1999–2014) to estimate an annual population density index that represents density during peak abundance each year. Relative population growth rate was estimated using a time series of the population density index, and population dynamics parameters r 0 and K were estimated by fitting relative growth rates (RGRs) to density independent and dependent models that include the effects of fire. Population models were simulated 20 and 100 years into the future to evaluate the sensitivity of extinction probability to density dependent dynamics and fire. Although the density independent model had the highest relative likelihood, density dependent models produced population trajectories with behavior more congruent with data from the Anaea troglodyta floridalis population. The absence of fire increased sensitivity of RGR to density, and the occurrence of fire buffered this sensitivity by increasing carrying capacity. Extinction risk was most sensitive to the inclusion of density dependent dynamics. Density dependent models provided a more optimistic outlook relative to density independent models (8 vs. 66 % probability of extinction in 20 years). Our simulations suggest that improving carrying capacity would provide the best buffer to extinction for this endangered endemic butterfly.