Climate variability and COVID-19 non-pharmaceutical interventions shaped dengue transmission in Guangdong: an integrated modeling study

Dengue fever is a major climate-sensitive mosquito-borne disease shaped by both environmental conditions and human behavior. We quantified the combined effects of climate variability and COVID-19 non-pharmaceutical interventions (NPIs) on dengue transmission in Guangdong, China, using an integrated modeling framework that combines climate-driven mosquito abundance estimated by recurrent neural networks (RNNs) with a temperature- and intervention-sensitive Susceptible-Infected-Recovered (SIR) model. Mosquito surveillance, meteorological data, NPI intensity, and reported dengue cases from 2016 to 2023 were used for model calibration within a partially observed Markov process framework. The calibrated model reproduced observed interannual dengue fluctuations using shared, rather than year-specific, fitted parameters. SHAP (SHapley Additive exPlanations) analysis reveals the nonlinear and day-night thermal effects on mosquito abundance, identifying minimum temperature as the relatively more influential climatic driver, while mechanistic experiments showed that temperature-dependent biting rates critically influenced dengue transmission. COVID-19 NPIs substantially suppressed dengue incidence in Guangdong during the pandemic period, corresponding to an estimated 99.03% (95% CI: 94.54-99.68%) reduction relative to the weak-intervention counterfactual scenario. By integrating explainable climate-driven mosquito abundance prediction with mechanistic transmission modeling and counterfactual NPI experiments, our framework explicitly separates and quantifies the roles of mosquito abundance, temperature-dependent effective biting, and human interventions, providing a practical tool for climate-informed early warning and intervention assessment for dengue and other climate-sensitive mosquito-borne diseases.