Fig. 1

Summary of the modeling and post-modeling steps followed for this research. We coupled 35 curated occurrence records of human Oropouche fever outbreaks with 15 environmental predictors for model development (A). Environmental multicollinearity was treated via a correlation matrix to select three environmental predictors (i.e., BIO1, BIO7 and BIO12), and an independent principal component analysis (PCA) over the 15 original variables for a total of two sets of predictors for model development over three different model calibration regions (A). We used one-class support vector machines (OC-SVM) and convex hull hypervolumes as algorithms to explore the environmental and geographical space defined by the occurrences and environments processed (B). After model selection and evaluation, we examined (i) the influence of each occurrence in the geographic space, (ii) the role of vegetation difference on recorded outbreaks, and (iii) calculated the human population overlapping with the Oropouche virus (OROV) transmission risk map (C)