Decoding multi-frequency scale climate-dengue dynamics: An interpretable modeling framework

Dengue fever is a growing global health threat. Previous studies indicate that climate conditions drive dengue transmission. However, the mechanisms by which climate signals operate at different frequency scales remain poorly understood. Here, focusing on Singapore, we examine multi-frequency relationships between climate factors and dengue transmission. We aim to elucidate underlying mechanisms and to develop a mechanism-driven model for short-term forecasting. Correlation analysis reveals that climate impacts on dengue exhibit frequency-specific patterns, suggesting distinct pathways of influence. Building on this, causal analysis identifies the specific climate-dengue linkages at each frequency scale. Tropical Atlantic and Central Eastern Pacific SSTs drive low-frequency dengue via local precipitation and temperature. At the high-frequency scale, SSTs from the tropical Atlantic, Central Eastern Pacific, and Indian Ocean affect dengue via dew point and precipitation. At the trend scale, SSTs in the tropical Indian Ocean and tropical Atlantic drive dengue dynamics through their effects on local dew-point temperature. Leveraging these mechanistic links, we developed frequency-specific AutoML models for Singapore that accurately forecast weekly dengue dynamics, with test RMSEs of 19.64 (low-frequency), 50.67 (high-frequency), and 2.24 (trend). Together, these analyses constitute an integrated framework for multi-frequency dengue analysis and interpretable prediction modeling, offering mechanistic insights and actionable guidance to enhance disease forecasting and public health strategies. • Multi-scale analysis based on CEEMDAN reveals frequency-specific climate drivers of dengue. • PCMCI causal analysis shows SST modulates local climate, drives dengue transmission. • Auto-ML model based on physical mechanism accurately predicts Singapore dengue outbreaks. • Interpretable framework uses multi-scale climate signals for dengue early warnings.