Ecologists have a long tradition of modeling population dynamics using structured population models. Most often, these models are structured by individual age, size, or life stage. However, such models can also be structured to describe spatial heterogeneity, where classes are sites with different conditions and transitions are movement between sites. In spatially-structured models, the long-term metapopulation growth rate (i.e., the leading eigenvalue of the transition matrix) is generally larger than the "average" growth rate because a higher proportion of individuals are in higher quality sites. In changing environments, the spatial distribution of individuals lags behind current environmental quality, leading to transient dynamics with a lower metapopulation population growth rate in the short term. In this talk, I review this longstanding but not widely appreciated implication of structured population dynamics, in the context of interpreting population trends and potential geographic range shifts of butterflies in the United States.