Nonlinear fitting algorithms have illuminated the role of weather in human diseases, by allowing for robust tests of mechanistic transmission models, but a lack of data has prevented applications to animal diseases. This is important because classical models that neglect weather predict that there will be a host density threshold, below which epidemic intensity will be slight, but models that include weather predict that this threshold will often be obliterated by weather variability. To test the applicability of thresholds to animal diseases, we estimated infection rates of the fungal pathogen Entomophaga maimaiga in the gypsy moth, by collecting larvae during epidemics at a range of host densities and weather conditions, and we estimated the pathogen's force of infection, by exposing experimental larvae to the pathogen for 24 h periods in the field. By fitting a range of models to our data, we show that epidemics of this pathogen are best explained by a model that allows for positive effects of both host density and cool, moist weather on transmission, such that weather-only and density-dependence-only models provide vastly poorer explanations for the data. Despite the effects of weather, the combined model shows that the effects of density in E. maimaiga are strong enough to ensure that the density threshold will have important effects on the probability of epidemics. Our work shows that weather and density-dependent transmission can interact in non-intuitive ways, and provides an illustration of the usefulness of nonlinear fitting for understanding animal diseases.