Abstract Submission: Dengue, a mosquito-borne viral disease ,poses a significant global public health burden, particularly in urban areas of tropical and subtropical regions, including Colombia. Dengue is primarily spread by Aedes aegypti and Aedes albopictus mosquitos. Precipitation, temperature, relative humidity, and large-scale cli- mate phenomena (e.g., ENSO) pattens all influence the development and spread of dengue in mosquitoes, directly influencing human exposure and health risks. While forecast systems leveraging hydroclimatic variables have shown poten- tial for improving dengue surveillance and informing decision-making broadly, there has been less effort on operationalizing predictions by establishing opti- mal thresholds for early, forecast-based interventions, especially in the Amazon region. This study builds on prior work identifying hydroclimatic drivers of dengue in four cities across Colombia (Cali, Cucuta, Leticia, Medellin). We em- ploy machine learning and deep learning models to assess the predictability of hydroclimatic conditions associated with increased dengue risk at sub-seasonal to seasonal timescales. Furthermore, we evaluate optimal trigger thresholds for early warning systems to inform interventions, aiming to reduce disease trans- mission. Integrating advanced machine learning techniques with hydroclimate predictors can potentially enhance forecasting accuracy and inform proactive public health measures.
