Environmental drivers and spatiotemporal heterogeneity of dengue attack rates across selected Yemeni governorates using spatial autocorrelation and environmental discordance analysis

Dengue is environmentally sensitive, yet broad environmental conditions may not consistently correspond with reported disease burden, particularly in conflict-affected settings where surveillance capacity may vary. This ecological longitudinal study examined the spatiotemporal relationships between environmental conditions and annual reported dengue attack rates across ten Yemeni governorates from 2020 to 2024. Annual reported dengue attack rates were standardised and evaluated using temporal rankings, trajectory classifications, an exploratory Dengue Instability Index, Spearman rank correlations and Kendall’s coefficient of concordance. Environmental conditions were characterised using vegetation, surface-water, land-surface-temperature and precipitation indicators. Global and local Moran statistics were used to assess spatial dependence and environmental–epidemiological spatial associations, while principal component analysis (PCA) integrated the environmental indicators into a common environmental-profile gradient for evaluating environmental–epidemiological concordance and discordance. Marked spatial and temporal heterogeneity was observed in reported dengue burden. The highest-burden governorate changed annually, although Shabwah and Abyan maintained persistently elevated relative burdens. Hadramaut recorded the greatest temporal instability, followed by Marib and Abyan. Governorate rankings nevertheless demonstrated moderate overall concordance across the study period (Kendall’s W = 0.677, χ2(9) = 30.47, p < 0.001). Global univariate Moran’s III values were consistently negative, with significant spatial dispersion detected only in 2021 ( I = − 0.421, p < 0.05). Bivariate spatial associations between dengue and individual environmental indicators were predominantly weak, negative and temporally inconsistent. Principal component analysis showed that the first environmental component explained 67.2% of the standardised variance and represented an integrated environmental profile (a gradient from hotter, drier and less vegetated to cooler, wetter and more vegetated conditions). Among 20 governorate–year observations with above-mean reported dengue attack rates, 17 (85%) occurred under below-mean environmental-profile conditions. These findings demonstrate substantial environmental–epidemiological discordance in Yemen. Broad remotely sensed environmental conditions alone were insufficient to explain the spatial and temporal distribution of reported dengue burden. Operationally, remotely sensed environmental indicators could complement reported dengue notifications and locally available entomological and contextual data to prioritise governorates for intensified surveillance and targeted investigation.