Fig. 4: Comparison of previously published and current suitability maps for arboviral diseases.

Panels show comparisons between earlier published maps and our newly generated maps for (a) dengue, (b) chikungunya, (c) Zika, and (d) yellow fever. Previous maps were retrieved from Messina et al.¹⁷ for dengue, Nsoesie et al.¹⁸ for chikungunya, Messina et al.¹⁹ for Zika, and Shearer et al.²⁰ for yellow fever. Area Under the Curve (AUC) values for our map and previous maps are indicated in parentheses. AUC values measure the model’s ability to distinguish between occurrence and background points, with values closer to 1 indicating better predictive performance. The AUC was calculated based on predicted values from both our map and previous maps using presence and background points. Maps were converted into binary format using a threshold that maximised the global sum of sensitivity and specificity. Once binary maps of at-risk areas were generated, we categorised each 5 × 5 km pixel into one of four groups: (1) areas at risk in both our map and the previous maps, (2) areas at risk only in previous maps, (3) areas at risk only in our map, and (4) areas not at risk in either map. The maps were created using public-domain Natural Earth data, accessed through the rnaturalearth package in R⁸⁶.