Abstract. We used a variety of modeling approaches to assess the effects of fire on the demography and population viability of scrub buckwheat (Eriogonum longifolium var. gnaphalifolium), a threatened herb endemic to periodically burned xeric uplands in central Florida, USA.We compared the management implications of different approaches. Using demographic data from ten years including five fires, we constructed Lefkovitch matrix models of scrub buckwheat populations in unburned habitats for populations experiencing a burn and for populations one year after burning. Deterministic analysis of the unburned matrices predicted relatively stable populations in unburned habitats using optimistic fertility estimates (l 5 1.002, 95% CI 0.975–1.033), but almost certain decline given pessimistic fertility estimates (l 5 0.981, 0.962–1.003). Burned populations are always predicted to increase (CI on l . 1.08). Numerical calculations of single-year growth rates based on the expected stable stage distribution of an unburned population yielded positive, but lower, growth rates. The l’s calculated for periodically burned populations had confidence intervals entirely above 1 for fire return intervals shorter than 20 yr (optimistic) or 5 yr (pessimistic). Stochastic simulations yielded highly uncertain estimates of the probability of quasi-extinction over 250 yr, with only fire return intervals shorter than 20 yr (optimistic) or 5 yr (pessimistic) yielding quasiextinction confidence intervals entirely below 1. Population growth rates in partially burned areas increased linearly with the proportion burned. The results of these density-independent simulations remained virtually unchanged (maximum difference of 5.2%) by introducing plausible population ceilings. Regression and stasis transitions had the highest elasticities for unburned populations, while fecundity elasticities were relatively greater for burn years. Transient elasticity analysis on total population size of simulated single-year post-fire growth and on the population vector five years in the future showed broadly similar patterns to standard elasticities, although the standard analysis assigned higher elasticity values to transitions contributing to future reproductive potential. Fire management is important to encouraging viable populations. Our analysis suggests minimum fire return intervals of 20 yr (optimistic) or 5 yr (pessimistic). These fire return intervals are consistent with those often used in the sandhill and oak–hickory scrub habitats of scrub buckwheat. Key words: conservation; demography; density dependence; elasticity; Eriogonum longifolium var. gnaphalifolium; fire; Florida scrub; matrix models; population ceilings; population viability; sandhill; scrub buckwheat |