Epiphytes are one of the most ubiquitous elements of tropical forest canopies, including seasonally dry tropical forests. Given the temporal variation in weather conditions in the latter, epiphyte populations may be subject to wide temporal variation in seedling recruitment, reproductive success, vegetative propagation and mortality rate. In this study, we use a 3-y demographic data set for Tillandsia brachycaulos to project its long-term population dynamics through the use of average and periodic matrices, as well as stochastic simulations. The results show that demographic behaviour varied over the 3 years of study, apparently in relation to rainfall. The first 2 years yielded a low λ value (0.79 and 0.80 – although only the former was significantly lower than unity), while the third year resulted in a λ = 1.08 (not significantly different from 1.0). When incorporating this demographic variation in an average matrix, a periodic matrix and stochastic simulations, the resulting overall λ was below unity in all three cases. The projections of the stochastic simulations suggest that the population would be able to persist in the long run only if the frequency of “good” years (defined here as those with an August rainfall above 200 mm) was above 0.6, which appears unlikely given that global warming might result in a lower frequency of rainy years in tropical dry forests.