Ebola-virus outbreaks are teleconnected with macroclimatic oscillations

Abstract

Ebola viruses’ ecology is known to be affected by climate, which is globally monitored using macroclimatic oscillation indices. We tested whether any of 26 macroclimatic oscillation indices were teleconnected with Ebola virus disease (EBOD) outbreaks. We compiled an updated database containing geographical locations and dates of the human index cases for all recorded EBOD outbreaks, all of which occurred in Africa, spanning from 1976 to 2025. We used two spatial resolution scales: (1) the whole area previously considered at least of intermediate biogeographical favourability for the Ebola viruses, and (2) separately in the areas of high and intermediate favourability for the viruses; and two temporal resolution scales (annual and quarterly). We split these data into cases for the period 1976–2018 to train our models, and cases for the period 2019–2025 to test the forecasting ability of the models. Four macroclimatic indices, namely the Pacific Decadal Oscillation, the Caribbean Sea Surface Temperature Index, the Dipole Mode Index and the Quasi-Biennial Oscillation, were spatio-temporally correlated with EBOD outbreaks. Using supervised machine learning algorithms and fuzzy logic, we built a fuzzy logic circuit of ensemble models combining these indices. Average sensitivity for the period 1976–2018 was 0,82 and average specificity was 0,89. The fuzzy logic circuit successfully forecasted the spatio-temporal occurrence of EBOD outbreaks from 2019 to 2025, as 14 out of 21 predictions agreed with observation. These macroclimatic indices can be used as the basis for an early warning system to determine the risk of new EBOD outbreaks.