Assessing the Risk of Windborne Dispersal of Culicoides Midges in Emerging Epizootic Hemorrhagic Disease Virus Outbreaks in France

Abstract

The epizootic hemorrhagic disease virus (EHDV) is a novel emerging threat for the European livestock sector. First detected in Sardinia and southern Spain at the end of 2022, this transboundary disease emerged in France in September 2023 despite restrictions on animal movement and enhanced surveillance protocols. Although virus spread is believed to be mediated by the dispersal of Culicoides vectors by the wind, prediction is difficult due to the large number of meteorological parameters that must be considered. Using simulations of atmospheric trajectories, we developed a model to investigate the long-distance dispersal risk zone of Culicoides in Europe, starting from different source zones. Our model predicted with good sensitivity the newly EHDV-infected areas in France over a period of 5 weeks after its first introduction in the country. Prospectively, we predicted that the midge dispersal zone of early 2024 could expand toward most of the western half of France and could sporadically reach new countries under favorable spring conditions. The wind dispersal risk maps provided are intended to support better preparedness and response to Culicoides-borne diseases.

Figure 1

Definitions of the two source zones Z₁ and Z₂ composed of source grid cells i from where atmospheric trajectories were initiated: (a) “index source zone Z₁” includes three grid cells where the first three EHDV outbreaks occurred on September 4, 8, and 9 2023; (b) “secondary source zone Z₂” includes 50 grid cells where EHDV outbreaks occurred between September 4 and November 15.

Figure 2

Spatial distribution of (a) H_j(Z1,T1), the daily probability averaged by week (T₁) between week 37 and week 41 (mid-September to mid-October), and (b) H_j(Z1,T5), the daily probability averaged over the 5 weeks (T₅) of the same period; H_j(Z1,T1) and H_j(Z1,T5) are both computed here considering the index source zone Z₁ (the first three EHDV outbreaks started in France in week 36) and the meteorological conditions of 2023 (scenario 1).

Figure 3

Spatial overlay between risk predictions (H_j(Z1,T5), the daily probability averaged over 5 weeks from week 37 to week 41) and the 186 EHDV outbreaks emerging within the same period of time in France. H_j(Z1,T5) is computed here considering the index source zone Z₁ (the first three EHDV outbreaks started in France in week 36), and the meteorological conditions of 2023 (scenario 1).

Figure 4

Validation plots: (a) proportion of destination grid cells j where at least one emerging EHDV outbreak occurred (“yes”) or did not occur (“no”) in each risk interval (H_j(Z1,T5)). Destination grid cells equivalent to index source grid cells were not considered; (b) empirical cumulative distribution of emerging outbreaks according to their associated risk predictions (H_j(Z1,T5)). The dashed color lines represent the upper limits of the risk intervals.

Figure 5

Receiver operating characteristic (ROC) curve characterizing the global performance of the model. Model performance was measured by the AUC (area under the curve). A, B, and C refer to three situations on the upper left part of the ROC curve where the risk threshold θ either maximizes only the true positive rate (A), minimizes only the false positive rate (C), or optimizes values of both TPR and FPR (B).

Figure 6

Spatial distribution of (a) H_j(Z2,T1), the daily probability averaged by week (T₁—from W11 to W15; from mid-March to mid-April) of long-distance dispersal, and (b) H_j(Z2,T5), the daily probability averaged over 5 weeks (T₅−from W11 to W15; from mid-March to mid-April) of long-distance dispersal. Both H_j(Z2,T1) and H_j(Z2,T5) are both computed here considering the secondary source Z₂ (the whole EHDV-infected area in early December 2023 in France) and the 4-year period 2020–2023.