Environmental and epidemiological drivers of dengue forecasting across heterogeneous transmission settings in Brazil...

Dengue fever is the fastest-growing mosquito-borne viral disease globally, with unprecedented epidemics reported across the Americas in 2024. Brazil and Panama experienced record transmission that strained public health systems and highlighted limitations in existing surveillance and early warning approaches. Despite growing evidence that climatic variability, land-use change, and other environmental factors influence dengue transmission, scalable approaches integrating heterogeneous environmental datasets remain limited.

We integrated climatic, environmental, demographic, and epidemiological datasets to develop and evaluate a dengue forecasting system across Brazil and Panama. Monthly surveillance data (2015-2024) were harmonized with multi-source environmental predictors using a reproducible workflow before evaluating statistical, machine learning, and neural network models. Performance was assessed using standard forecasting metrics and agreement between predicted and observed transmission patterns during the 2024 epidemic.

Dengue transmission exhibited substantial spatial heterogeneity across both countries. Environmental predictors, including temperature, precipitation variability, land-cover characteristics, mosquito suitability, and demographic indicators, dominated model performance in Brazil, whereas recent transmission history contributed more strongly in Panama. Ensemble forecasts reduced MAE by 24·4% in Brazil and 29·0% in Panama relative to a seasonal-naïve benchmark and demonstrated substantial agreement with observed spatial risk classifications.

Integrating heterogeneous environmental datasets with epidemiological surveillance improved understanding and prediction of dengue transmission across contrasting ecological settings. Differences in the relative importance of environmental and epidemiological predictors between Brazil and Panama suggest that forecasting approaches should be adapted to local environmental conditions rather than transferred directly across regions, supporting climate-informed surveillance and public health planning.