Spatio-temporal machine learning for multi-horizon prediction of bluetongue outbreaks.

A machine-learning ensemble framework predicts bluetongue outbreaks 1-6 months ahead with high stability across France, Greece, and Italy, outperforming 14 benchmarks at longer lead times.

• Why it matters: Early warning of vector-borne disease outbreaks is crucial for effective surveillance and preparedness, yet remains challenging due to complex interactions among hosts, vectors, and climate.
• What they did: Researchers developed a logit-weighted ensemble (LWE) that uses outbreak notifications and climate data to forecast outbreak occurrence at the administrative unit-month scale, evaluated on data from 2005 to 2024.
• The result: The framework reliably identifies high-risk months and locations up to six months in advance, especially at longer lead times, enabling improved surveillance planning across diverse European regions.