Spatiotemporal disease suitability prediction for Oropouche virus and the role of vectors across the Americas

Abstract

Oropouche virus (OROV) is an emerging arbovirus with increasing outbreaks in South America, yet its environmental drivers and potential range remain poorly understood. Using ecological niche modeling (ENM) with random forests, we assessed the environmental suitability of OROV and its primary vector, Culicoides paraensis, across Brazil and the Americas. We evaluated five pseudo-absence sampling techniques, considering pseudo-absence ratios, buffer radii, and density smoothing factors to determine the most effective modeling approach. Key environmental predictors included humidity, agricultural land-use, and forest cover for both the virus and the vector. The resulting suitability model identifies high transmission risk areas in Central and South America, and reveals that environmental suitability patterns align with seasonal fluctuations in case numbers, with peaks in Amazonian states at the beginning of the year and an expansion into non-Amazonian regions later in the year. A bivariate suitability map highlighted strong spatial overlap between OROV and Culicoides paraensis, with potential co-suitability areas with Culex quinquefasciatus mosquito, a suspected secondary vector. These findings enhance understanding of OROV transmission dynamics, supporting risk assessment, surveillance, and vector control strategies.

Keywords: Arboviruses; Culicoides paraensis; Environmental niche models; Oropouche virus; Pseudo-absence sampling.

Figure 1.. Comparison of sampling techniques, variable importance, and spatial suitability predictions.

(A) Boyce Index (BI) performance across five pseudo-absence sampling techniques (Random Sampling, Geographic Sampling, Density-Weighted Geographic Sampling, Density-Weighted Population-Based Sampling, and Target Group-Based Sampling) with varying PA ratios and buffer radii. Higher BI values indicate better model performance. (B) Spatial distribution of habitat suitability predicted under each sampling strategy (predictions from the best performing model under each technique), with darker red areas indicating higher suitability. The BI for each method is displayed in parentheses. (C) Variable importance calculated as the average of absolute SHAP values in the training dataset, indicating the average variable contribution to the overall prediction of the model. (D) shows the distribution of SHAP values per variable, indicating whether contributions are positive or negative.

Figure 2. Consensus map and temporal environmental suitability trends.

(A) The consensus map of environmental suitability for Oropouche virus transmission across Brazil, highlighting areas with persistent high suitability (red) and regions with lower suitability (blue). (B) Monthly projection of environmental suitability index across Brazil (maps) with temporal trends in case counts (stacked bars) and suitability index (lines) for the year 2024, with separate trends for Amazonian and non-Amazonian states. Shaded regions represent the variability in suitability estimates, with darker lines indicating mean suitability.

Figure 3. Predicted Environmental Suitability for Oropouche Virus (OROV), across the Americas.

(A) Predicted environmental suitability for Oropouche virus transmission across the Americas. The inset highlights suitability predictions for Central America, including Panama (PAN), Costa Rica (CRI), Nicaragua (NIC), Honduras (HND), El Salvador (SLV), Guatemala (GTM), Belize (BLZ), Cuba (CUB) and Jamaica (JAM). (B) Uncertainty map displaying the coefficient of variation, across models built under different pseudo-absence sampling techniques.

Figure 4

Environmental drivers and spatial suitability of Culicoides paraensis and bivariate suitability map of Oropouche virus (OROV) and its primary vector: (A) Variable importance plot showing the mean SHAP values for environmental predictors contributing to C. paraensis suitability. (B) Predicted spatial distribution of C. paraensis suitability across the Americas. (C) Bivariate suitability map of Oropouche virus (OROV) andCulicoides paraensis. The color gradient represents the combined suitability of OROV and each vector, with light gray indicating low suitability for both, pink/purple hues representing higher suitability for the vector, blue tones indicating higher suitability for the virus, and dark purple signifying high suitability for both. Inset highlights Central America and the Caribbean.