• An international modelling challenge brought together teams to forecast dengue cases in Brazil and support public health planning ahead of the epidemic season.

• Results show that combining multiple models improves predictions, especially under extreme and atypical outbreak conditions.

• The model developed by BSC, which integrates climate information, was among the best-performing approaches, particularly during the record-breaking 2024 epidemic.

Brazil experienced an unprecedented surge in dengue cases in 2024, with more than 10 million infections reported across the country during the largest outbreak ever recorded globally. The scale and geographical spread of the epidemic highlighted the growing challenge posed by climate variability and changing transmission patterns.

To help anticipate future outbreaks, the Barcelona Supercomputing Center – Centro Nacional de Supercomputación (BSC-CNS) participated in the Infodengue-Mosqlimate Dengue Challenge 2024 (IMDC24), an international initiative aimed at improving dengue forecasting in Brazil. The challenge became the first use case of the prototype R package GHRmodel, developed by the BSC’s Global Health Resilience (GHR) group to generate dengue forecasts using a Bayesian spatiotemporal modelling approach. The model integrates climate information, including temperature, precipitation, and drought indicators, to better predict dengue transmission dynamics. The updated IMDC25 forecasts were later produced using the published version of GHRmodel.

The GHR model was among the best-performing approaches during the extreme 2024 epidemic, particularly in its ability to capture atypical outbreak dynamics that deviated from normal seasonal behaviour. The updated version of the model, developed for the 2025 edition of the challenge, also ranked among the top-performing approaches across international teams, showing especially strong results in several Brazilian regions, including the northeast, midwest and southeast.

The IMDC24 challenge brought together research teams from multiple countries to develop models capable of forecasting weekly dengue cases ahead of the epidemic season. The objective was to provide actionable information to the Brazilian Ministry of Health, supporting prevention and response planning before outbreaks unfold.

One of the main conclusions of the initiative was that no single model consistently outperformed the others across all regions and epidemic seasons. Instead, the results highlighted the value of ensemble approaches, which combine multiple models to generate more robust predictions. These ensemble forecasts have now been adopted by Brazilian public health authorities to support planning for upcoming dengue seasons.

Beyond methodological advances, the initiative highlights the growing importance of integrating climate information into public health tools. In a context of increasing climate variability and expanding vector-borne disease risk, improving outbreak anticipation is essential to support more proactive and effective responses and reduce impacts on populations and health systems.

The structure and findings of the IMDC24 challenge have been published in Proceedings of the National Academy of Sciences (PNAS).

Following the success of the previous editions, the Infodengue-Mosqlimate Dengue Challenge will continue in 2026 (IMDC26), further advancing international efforts to improve dengue forecasting and public health preparedness in Brazil.