Subtype-specific risk estimates of influenza under extreme weather events in a subtropical coastal city: high humidity as a major factor associated with forecasting implications

Meteorological conditions are important environmental determinants of influenza transmission; however, the health impacts of extreme weather events (EWEs) remain insufficiently characterized, particularly in subtropical coastal regions. Using Xiamen, China, as a case study, we systematically evaluated the heterogeneous associations between EWEs on influenza and quantified their independent predictive contributions. Daily surveillance data (2013–2024) were analyzed using distributed lag nonlinear models (DLNM) to assess risks of 11 EWE categories across influenza subtypes and population strata, adjusting for air pollutants and population mobility. Random forest (RF) modeling further evaluated predictive contributions. Pronounced EWE-specific effects emerged: extreme high humidity event (EHRHUE) most strongly associated with influenza B(Yamagata) (21-day cumulative relative risk [CRR] = 2.51), while extreme low temperature event (ELTEME) most strongly associated with A(H3N2) (21-day CRR = 1.72). Extreme low wind speed event (ELWINE) and extreme long sunshine duration event (EHSSDE) showed elevated risks for B(Yamagata) (21-day CRR = 3.11) and A(H3N2) (21-day CRR = 3.36), respectively. Feature importance identified EHRHUE as the dominant predictor (35.59% contribution), followed by PM₁₀ (18.64%). Incorporating EWEs into RF improved performance: Nash–Sutcliffe efficiency (NSE) increased 26.82%; other metrics also enhanced. This study confirms that EWEs were significantly and heterogeneously associated with influenza transmission in subtropical coastal cities, with EHRHUE, PM₁₀, and temperature as key environmental predictors. Integrating EWEs into predictive models enhances influenza forecasting accuracy and provides a scientific foundation for climate-adaptive early warning systems and precision public health interventions.