Influence of hydroclimatic variability on dengue incidence in a tropical dryland area

Abstract

Dengue is an endemic disease in more than 100 countries, but there are few studies about the effects of hydroclimatic variability on dengue incidence (DI) in tropical dryland areas. This study investigates the association between hydroclimatic variability and DI (2008-2018) in a large tropical dryland area. The area studied comprehends seven municipalities with populations ranging from 32,879 to 2,545,419 inhabitants. First, the precipitation and temperature impacts on interannual and seasonal DI were investigated. Then, the monthly association between DI and hydroclimatic variables was analyzed using generalized least squares (GLS) regression. The model's capability to reproduce DI given the current hydroclimatic conditions and DI seasonality over the entire time period studied were assessed. No association between the interannual variation of precipitation and DI was found. However, seasonal variation of DI was shaped by precipitation and temperature. February-July was the main dengue season period. A precipitation threshold, usually above 100 mm, triggers the rapid DI rising. Precipitation and minimum air temperature were the main explanatory variables. A two-month-lagged predictor was relevant for modeling, occurring in all regressions, followed by a non-lagged predictor. The climate predictors differed among the regression models, revealing the high spatial DI variability driven by hydroclimatic variability. GLS regressions were able to reproduce the beginning, development, and end of the dengue season, although we found underestimation of DI peaks and overestimation of low DI. These model limitations are not an issue for climate change impact assessment on DI at the municipality scale since historical DI seasonality was well simulated. However, they may not allow seasonal DI forecasting for some municipalities. These findings may help not only public health policies in the studied municipalities but also have the potential to be reproducible for other dryland regions with similar data availability.