Predictions of an environmentally-sensitive pathogen's response to global climate change using mechanistic models and downscaled climate simulations

Background/Question/Methods

An important debate in ecology is how global climate change (GCC) will affect the spread of infectious disease. Because mathematical models allow us to make quantitative predictions about future conditions, we combined a mechanistic, host-pathogen model with a high resolution meteorological model of GCC in North America. The disease model stochastically simulates epidemics of a fungal pathogen, Entomophaga maimaiga, which serves as a biological control for invasive gypsy moth caterpillars (Lymantria dispar), based on initial host and pathogen densities and daily weather conditions (rain, temperature, relative humidity). We previously estimated model parameters using data collected during field epidemics in the Lower Peninsula of Michigan (LP). The GCC model produces estimates of potential daily local weather conditions on a 10 km² scale grid across the continent for the beginning (1995-2004) and end (2085-2094) of the century based on projected emission scenarios. By simulating the disease model with output from the GCC model, we quantitatively compared epidemic dynamics under climate conditions from beginning and end of the century across the LP (400x500km). To tease out what is driving responses to GCC, we simulated our model using a stochastic weather generator which allowed us to manipulate mean and variance of local climate variables.

Results/Conclusions

Comparing simulated epidemics using the mechanistic disease model across the century, we found that mean E. maimaiga epidemic sizes are likely to increase in size from 29.5% to 40.5% of total host population infected across all of Michigan's lower peninsula, assuming average host and pathogen densities. However, changes in epidemic dynamics were highly sensitive to regional differences in climate and occurred in a patchy distribution across the peninsula. Some regions, epidemic sizes increased up to 67.0% of the host population. In other locations, the climate is likely to become less conducive to pathogen transmission, and total fraction infected is predicted to decrease up to 52.3%. Overall, we observed a general northernly shift of approximately 150km in regions of peak pathogen prevalence. Further model simulations using a stochastic weather generator indicated that changes in variability of rainfall largely drove the observed changes. These results demonstrates the importance of including ecological components in models when making predictions about the impact of GCC. Furthermore, we see that effects of GCC on pathogen transmission can be highly heterogeneous within a region and small changes in local climate can be magnified through density-dependent transmission.