The global burden of Dengue virus (DENV) is increasing globally. This has resulted in increasing effort to forecast shifts in DENV burden. Since DENV is primarily transmitted by two mosquito vectors, Aedes aegypti and Ae. albopictus, predicting changes in suitable vector habitat often precedes forecasting shifts in disease burden. The majority of habitat suitability models for Ae. aegypti and Ae. albopictus have used long-term, global climate data to predict how suitability will shift as a result of environmental change. However, the effects of fine-scale environmental change, such as changes in land use and land cover, are often underrepresented in these models. Furthermore, the relative influences of climate and land cover, and the scales at which they drive suitability, have not been comprehensively assessed. Here, we use a suite of multi-scale species distribution models to address these issues. We ask the following questions: (1) Are the environmental drivers of habitat suitability consistent across space for Ae. aegypti and Ae. albopictus? and (2) At what scales do spatial patterns of land cover and climate best predict spatial patterns of vector habitat suitability?
Results/Conclusions
We address the above questions by comparing the outputs of multiple species distribution models for Ae. aegypti and Ae. albopictus across Latin America and the Carribean using publicly-available species occurrence data and satellite-derived environmental data. These models were run using multiple occurrence data sets, using multiple environmental data sets, across varying spatial extents, and at multiple spatial grain sizes. We found the importance of climate and land cover to vary by species, and by scale. Ae. aegypti suitability was best characterized by fine-scale patterns of maximum annual temperature and impervious surface cover, confirming preference for hot urban environments. In contrast, Ae. albopictus suitability was best characterized by moderate-scale patterns of mean annual temperature and low-to-moderate vegetation cover, suggesting broader habitat suitability in agricultural regions. The regional dependence of these results has not yet been assessed. This systematic, multi-model approach to understanding the environmental drivers of vector habitat suitability provides a opportunity to better understand the relative importance of the multiple environmental drivers of vector habitat suitability, and the scales at which they operate. These insights can be used to develop novel mechanistic models of how DENV burden can be expected to shift under different climate and land use scenarios.