Climatic warming will likely have idiosyncratic impacts on infectious diseases, causing some to increase while others decrease or shift geographically. A mechanistic framework could better predict these different temperature-disease outcomes. However, such a framework remains challenging to develop, due to the non-linear and (sometimes) opposing thermal responses of different host and parasite traits, and due to the difficulty of validating model predictions with observations and experiments. We address these challenges in a zooplankton-fungus (Daphnia dentifera-Metschnikowia bicuspidata) system. We test the hypothesis that warmer temperatures promote disease spread and produce larger epidemics. In lakes, epidemics that start earlier and warmer in autumn grow much larger. In a mesocosm experiment, warmer temperatures produced larger epidemics. A mechanistic model parameterized with trait assays revealed that this pattern arose primarily from the temperature-dependence of transmission rate (β), governed by the increasing foraging (and hence parasite exposure) rate of hosts (f). In the trait assays, parasite production seemed sufficiently responsive to shape epidemics as well; however, this trait proved too thermally insensitive in the mesocosm experiment and lake survey to matter much. Thus, in warmer environments, increased foraging of hosts raised transmission rate, yielding bigger epidemics through a potentially general, exposure-based mechanism for ectotherms. This mechanistic approach highlights how a trait-based framework will enhance predictive insight into responses of infectious disease to a warmer world.