Ecological niche modeling of Aedes mosquito vectors of chikungunya virus in southeastern Senegal

Abstract Background Chikungunya virus (CHIKV) originated in a sylvatic cycle of transmission between non-human animal hosts and vector mosquitoes in the forests of Africa. Subsequently the virus jumped out of this ancestral cycle into a human-endemic transmission cycle vectored by anthropophilic mosquitoes. Sylvatic CHIKV cycles persist in Africa and continue to spill over into humans, creating the potential for new CHIKV strains to enter human-endemic transmission. To mitigate such spillover, it is first necessary to delineate the distributions of the sylvatic mosquito vectors of CHIKV, to identify the environmental factors that shape these distributions, and to determine the association of mosquito presence with key drivers of virus spillover, including mosquito and CHIKV abundance. We therefore modeled the distribution of seven CHIKV mosquito vectors over two sequential rainy seasons in Kédougou, Senegal using Maxent. Methods Mosquito data were collected in fifty sites distributed in five land cover classes across the study area. Environmental data representing land cover, topographic, and climatic factors were included in the models. Models were compared and evaluated using area under the receiver operating characteristic curve (AUROC) statistics. The correlation of model outputs with abundance of individual mosquito species as well as CHIKV-positive mosquito pools was tested. Results Fourteen models were produced and evaluated; the environmental variables most strongly associated with mosquito distributions were distance to large patches of forest, landscape patch size, rainfall, and the normalized difference vegetation index (NDVI). Seven models were positively correlated with mosquito abundance and one (Aedes taylori) was consistently, positively correlated with CHIKV-positive mosquito pools. Eight models predicted high relative occurrence rates of mosquitoes near the villages of Tenkoto and Ngary, the areas with the highest frequency of CHIKV-positive mosquito pools. Conclusions Of the environmental factors considered here, landscape fragmentation and configuration had the strongest influence on mosquito distributions. Of the mosquito species modeled, the distribution of Ae. taylori correlated most strongly with abundance of CHIKV, suggesting that presence of this species will be a useful predictor of sylvatic CHIKV presence.