Predicting the Spread of Vector-Borne Diseases in a Warming World

Abstract

Predicting how climate warming affects vector borne diseases is a key research priority. The prevailing approach uses the basic reproductive number (R0) to predict warming effects. However, R0 is derived under assumptions of stationary thermal environments; using it to predict disease spread in non-stationary environments could lead to erroneous predictions. Here, we develop a trait-based mathematical model that can predict disease spread and prevalence for any vector borne disease under any type of non-stationary environment. We parameterize the model with trait response data for the Malaria vector and pathogen to test the latest IPCC predictions on warmer-than-average winters and hotter-than-average summers. We report three key findings. First, the R0 formulation commonly used to investigate warming effects on disease spread violates the assumptions underlying its derivation as the dominant eigenvalue of a linearized host-vector model. As a result, it overestimates disease spread in cooler environments and underestimates it in warmer environments, proving its predictions to be unreliable even in a constant thermal environment. Second, hotter-than-average summers both narrow the thermal limits for disease prevalence, and reduce prevalence within those limits, to a much greater degree than warmer-than-average winters, highlighting the importance of hot extremes in driving disease burden. Third, while warming reduces infected vector populations through the compounding effects of adult mortality, and infected host populations through the interactive effects of mortality and transmission, uninfected vector populations prove surprisingly robust to warming. This suggests that ecological predictions of warming-induced reductions in disease burden should be tempered by the evolutionary possibility of vector adaptation to both cooler and warmer climates.