A Maximum Entropy Model of the Distribution of Dengue Serotype in Mexico

Abstract

Pathogen strain diversity is an important driver of the trajectory of epidemics. The role of bioclimatic factors on the spatial distribution of dengue virus (DENV) serotypes has, however, not been previously studied. Hence, we developed municipality-scale environmental suitability maps for the four dengue virus serotypes using maximum entropy modeling. We fit climatic variables to municipality presence records from 2012 to 2020 in Mexico. Bioclimatic variables were explored for their environmental suitability to different DENV serotypes, and the different distributions were visualized using three cutoff probabilities representing 90%, 95%, and 99% sensitivity. Municipality-level results were then mapped in ArcGIS. The overall accuracy for the predictive models was 0.69, 0.68, 0.75, and 0.72 for DENV-1, DENV-2, DENV-3, and DENV-4, respectively. Important predictors of all DENV serotypes were the growing degree days for December, January, and February, which are an indicator of higher temperatures and the precipitation of the wettest month. The minimum temperature of the coldest month between -5°C and 20°C was found to be suitable for DENV-1 and DENV-2 serotypes. Respectively, above 700-900 mm of rainfall, the suitability for DENV-1 and DENV-2 begins to decline, while higher humidity still favors DENV-3 and DENV-4. The sensitivity concerning the suitability map was developed for Mexico. DENV-1, DENV-2, DENV-3, and DENV-4 serotypes will be found more commonly in the municipalities classified as suitable based on their respective sensitivity of 91%, 90%, 89%, and 85% in Mexico. As the microclimates continue to change, specific bioclimatic indices may be used to monitor potential changes in DENV serotype distribution. The suitability for DENV-1 and DENV-2 is expected to increase in areas with lower minimum temperature ranges, while DENV-3 and DENV-4 will likely increase in areas that experience higher humidity. Ongoing surveillance of municipalities with predicted suitability of 89% and 85% should be expanded to account for the accurate DENV serotype prevalence and association between bioclimatic parameters.

Figure 1

Modeled suitability maps of DENV-1, DENV-2, DENV-3, and DENV-4 at a sensitivity of 91%, 90%, 89%, and 88%, respectively. Red points show presence records from 2012 to 2020 used to build the models.

Figure 2

Model response curves for DENV-1, DENV-2, DENV-3, and DENV-4 suitability about growing degree days for December, January, and February.

Figure 3

Model response curves for DENV-1, DENV-2, DENV-3, and DENV-4 suitability in relation to precipitation of wettest month.

Figure 4

Model response curves for DENV-1 about (a) mean diurnal range (monthly (maximum temperature–minimum temperature)), (b) temperature seasonality, (c) minimum temperature of coldest month, (d) precipitation of wettest month, (e) precipitation of driest month, and (f) precipitation seasonality (coefficient of variation).

Figure 5

Model response curves for DENV-2 about (a) temperature seasonality, (b) maximum temperature of warmest month, (c) minimum temperature of coldest month, and (d) precipitation of the wettest month.

Figure 6

Model response curves for DENV-3 about (a) mean diurnal range (monthly (maximum temperature–minimum temperature)), (b) maximum temperature of warmest month, (c) precipitation of wettest month, (d) precipitation of driest month, and (e) precipitation seasonality (coefficient of variation).

Figure 7

Model response curves for DENV-4 about (a) mean diurnal range (monthly (maximum temperature–minimum temperature)), (b) precipitation of wettest month, (c) precipitation of driest month, and (d) precipitation seasonality (coefficient of variation).