Predicting the response of a fungal pathogen to global climate change: Disentangling effects of climate and density-dependent transmission
An important debate in ecology is the extent to which global climate change (GCC) will affect infectious disease epidemics. This debate persists, in part, due to the difficulty of disentangling the effects of climate from density-dependent transmission. We therefore parsed the influence of climate and density-dependent transmission in an environmentally sensitive pathogen, the fungus Entomophaga maimaiga, which serves as a biological control agent for invasive gypsy moth caterpillars (Lymantria dispar). To investigate how this disease functions under different climate conditions and host densities, we collected data on infection rates and weather conditions (precipitation, humidity, temperature) during natural E. maimaiga epizootics at seven different sites along a 400 km latitudinal gradient in Michigan over three summers. Using these data, we fit and compared stochastic differential equation models containing only density-dependent transmission, only weather-dependent transmission, or a combination of both factors. We estimated parameters using maximum likelihood and compared models using AICc. We then generated predictions of infection rates across the state of Michigan by simulating our best fitting model at high host densities over predictions of weather patterns from down-scaled GCC models to analyze the future efficacy of E. maimaiga across the state.
The model utilizing both host density and weather-dependent transmission best explained variability in disease dynamics across sites. A density-dependent model received the next highest ranking (Δ AICc = 61.5) and reproduced epidemic sizes well at the sites containing high host densities but underestimated transmission at the lower density sites. A model predicting epidemics using only weather variables fit poorly overall (Δ AICc = 130.2) but still reproduced disease dynamics to some degree at the low density sites. Transmission of this pathogen is therefore strongly density-dependent, but the right weather conditions can lead to large epidemics even at low host densities. When we simulated our best model using future GCC predictions across Michigan, we found the reliance of E. maimaiga on weather lead to increases or decreases in epidemics depending on geographic location. In the coastal regions, which are expected to receive increased precipitation, the models predict increased disease epidemics. The interior of Michigan, however, is predicted to become hotter and drier, which will reduce the ability of the fungus to transmit. The dynamics of this pathogen may be representative of other environmentally sensitive diseases and demonstrate how important a role climate can play in disease epidemics.