Mechanistic models of species distributions have received increasing attention over the past several years. The models are advantageous, because mechanistic data are used to make a priori predictions about distributions based on variables associated with population growth or individual performance. Models can then be tested with independent distribution data, and predictions that do or do not match observed distributions highlight strengths or weaknesses in the underlying model. Mechanistic models are usually based on temperature tolerances but often ignore critical resource distributions. Here, we combine physiological models of juvenile growth using growing degree days with large-scale distribution data on milkweed host plants generated from herbarium specimens to predict monarch butterfly recruitment potential across eastern North America. While eastern monarchs largely overwinter in a small forested region in central Mexico, the majority of breeding occurs throughout the eastern US and southern Canada. We built a spatial recruitment map that accumulates daily energy for growth throughout the monarch’s breeding range and then overlaid maps of milkweed distributions (based on correlative niche models) to understand how resource availability may contribute to monarch migratory behavior. Model predictions of summer breeding monarch distributions were compared to observations from several citizen-science monitoring surveys. Recruitment maps based solely on minimum energy needed for growth do a good job of predicting spring distributions of monarchs, but substantially over-predict recruitment potential in the south. New laboratory data on lethal and sub-lethal impacts of hot temperatures suggest that monarchs may be moving north to avoid excessive heat, yet truly lethal temperatures are rarely reached even in the southern US. Instead, phenology of milkweed plants suggests that resource availability declines in the south during the summer. To further explore temporal patterns of host-plant availability, we scored milkweed phenophase from digital images of herbarium specimens and used those to create monthly maps of resource availability. Our results suggest that understanding the shifting availability of host plants may shed light on migratory behavior. Further, standard growing degree day models should be modified to account for lethal and sub-lethal impacts of excessive temperatures. While this model was developed for monarchs, the insect species with arguably the most complex migratory dynamics, we suggest that the modeling platform could be adapted for other insects to understand the impacts of changing climate and land-use on migratory or non-migratory species.