Résumé
Dengue is a mosquito-borne viral disease that causes recurrent epidemics in Cambodia, a Southeast Asian country where the disease is endemic. Climate has a major impact on vector spread and human behavior, so the importance of climate indicators in the design of early warning systems is crucial. While climate data are becoming increasingly qualitative and accessible, the health data needed to feed such models remain limited in quality, spatial resolution, access and latency. Primarily based on clinical factors, syndromic data are typical health data from low-income countries and offer timely and regularly reported information, serving as a valuable foundation for establishing health surveillance information systems. Acute hemorrhagic fever, i.e. the clinical diagnosis of dengue, is a notifiable syndrome in Cambodia. This study aims to investigate the potential use of acute hemorrhagic fever data, in designing climate-driven early warning systems for dengue epidemics. Machine learning models based on climate indicators obtained from publicly available satellite images have been set to estimate the weekly dynamics of acute hemorrhagic fever in Cambodia. The sameframework was tested at the province scale to inform at a larger resolution. Our results highlight the impact of extreme climate events on dengue epidemics at different scales and investigate the limitations and robustness of syndromic data. It supports the importance of linking climate indicators and health data in implementing operational early warning systems and defining appropriate warning thresholds to facilitate decision-making by public health services.