Ecosystems processes rely on phenological overlap between consumers and producers. Multiple flowering plants rely on insects for pollination, requiring overlap between flowering and insect flight periods; breeding passerine birds rely on insects, especially lepidoptera larvae, as nutrition for their young, requiring overlap between nestling provisioning and larval development. Forecasting ecosystem function requires accurate predictions of insect phenology. Here we test temperature-based models of insect development using a model species, the silver-spotted skipper (Epargyreus clarus), an oligophagus, multivoltine butterfly. We parameterize the models with data from rearing experiments in the lab, incorporating uncertainty due to food quality and photoperiod. We test our phenological predictions for E. clarus using 10 years of data from a network of citizen science butterfly monitoring projects. The accuracy of development time predictions varied across both year and season with predictions generally being less accurate in the fall. These results show that we can use experimental studies of thermal responses to make phenological predictions, but current models are less accurate late in the growing season, when seasonal cues (both temperature and photoperiod) shift insect resource allocation patterns. We explore how specific sources of uncertainty impact phenological predictions and the implications for future predictions of ecosystem processes.