Epidemiological Modeling Under Climate Change

Abstract

Climate change is reshaping the geographic distribution, seasonal timing, and outbreak dynamics of infectious diseases in ways that standard epidemiological models -- calibrated on historical climate and disease data -- are poorly equipped to project. Temperature, precipitation, and extreme weather events alter vector biology, pathogen survival, human behaviour, and healthcare system capacity simultaneously, creating non-linear interactions that univariate climate-disease correlation analyses cannot capture. We evaluated five epidemiological modelling frameworks applied to climate-sensitive disease projections across 28 European and Central Asian countries covering 486 million population-years of observation and projection from 2000 to 2040, using ECMWF ERA5 reanalysis, CMIP6 climate projections, and national infectious disease surveillance registries. Disease endpoints covered dengue fever vectorial capacity, West Nile virus transmission season length, Lyme disease tick activity zones, heat-related respiratory infection burden, and water-borne enteric disease incidence. Frameworks ranged from degree-day vector phenology models through mechanistic climate-epidemiology systems models to ML-integrated ensemble projections with uncertainty quantification. Performance was benchmarked using our Climate-Epidemiology Model Fidelity Score (CEMFS), integrating hindcast accuracy, projection stability, spatial resolution, policy actionability, and uncertainty calibration. The ML-integrated ensemble achieved the highest CEMFS (0.922), improving dengue vectorial capacity hindcast RMSE by 38.4% over degree-day models and projecting Aedes albopictus establishment risk zones 2.8 times more accurately in out-of-sample validation. Under RCP4.5 scenarios, all five frameworks projected northward expansion of dengue and West Nile virus transmission risk into previously non-endemic Baltic and Central European regions by 2035-2040