Summary
A medRxiv preprint analysing dengue cases across 435 Brazilian health regions reports that climate information improved forecasts at lead times of one to 12 months. The authors found different combinations of dry, warm and wet conditions associated with higher relative risk across climate zones.
A medRxiv preprint reports that climate information improved dengue forecasts across Brazil at lead times ranging from one to 12 months. The researchers analysed monthly case counts from 435 health regions between 2010 and 2025, using a Bayesian spatiotemporal model to examine climate patterns and predict cases and outbreaks.
The analysis considered interacting drought indicators and temperature anomalies across nine Köppen-Geiger climate zones. The authors report that long-lag dry, mid-lag warm and short-lag wet conditions were associated with the highest relative risk in five zones. These are associations in population-level data, not evidence that a particular climate pattern caused an individual infection.
Climate patterns varied across regions
A lag describes how far ahead of the dengue outcome a climate indicator is considered. The study assessed indicators at different lags alongside absolute temperature and the El Niño-Southern Oscillation (ENSO), which the model adjusted for. The authors also report that temperature anomalies modified hydrometeorological dynamics more strongly in temperate than tropical areas.
Dengue transmission is climate-sensitive in part because temperature and rainfall affect mosquito ecology. Considering conditions at different time intervals may therefore provide forecasting signals that are not captured by a single climate measure. The study estimates those signals across Brazil’s varied climate zones; it does not test a specific mechanism linking each pattern to infections.
Forecasts combined climate and recent case data
Compared with baseline models, climate covariates improved predictions of dengue case counts and outbreaks at lead times of 1, 3, 6 and 12 months, as well as for the upcoming season forecast five months ahead. Reported improvements were greatest in Brazil’s Northeast, Southeast and Midwest.
Recent dengue incidence contributed the most to improvement at the one-month lead. At three- and six-month leads, models combining incidence and climate information performed best. The researchers also extended the framework to weekly state-level predictions for the upcoming season, aligning the approach with national public-health planning. They propose a multi-stage forecasting strategy: use recent case data for the shortest horizon and combine surveillance with climate information for forecasts farther ahead.
The work is a preprint. Its abstract reports the direction of forecast improvements but not their numerical size or the relative-risk estimates, so the magnitude of the reported gains cannot be assessed from the summary. The results describe a modelling approach intended to support early warning and response, rather than a report of an implemented national forecasting service.