The microevolutionary consequences of climate change are studied less than the ecological impacts. We investigated the evolution of senescence in a population of the long-lived orchid Cypripedium parviflorum. We estimated generalized linear mixed models (GLMMs) for all major vital rates using 23-years of demographic data and historical climate forecasts. We calculated age-by-stage matrix projection models (MPMs) using these GLMMs and data from high-resolution climate simulations. We then ran stochastic population projections and calculated age-specific selection gradients on mortality to assess the effects of climate change. Projections predict local extinction under multiple climate change scenarios. Selection gradients predict a shift away from reproduction in young ages, instead favoring vegetative dormancy. This will accelerate the decline of this population, as the population growth rate is most sensitive to perturbations in young age-stages. Our results add to the evidence that vegetative dormancy is adaptive as a bet-hedging strategy to cope with environmental variation.