Spatiotemporal patterns and meteorological associations of mosquito density: insights from mosquito surveillance in Chengdu, China

Abstract

This study aimed to investigate the spatiotemporal distribution of mosquito density and its associations with meteorological factors in Chengdu, China, and to compare the predictive performance of different models, thereby providing a scientific basis for mosquito-borne disease surveillance and early warning. Mosquito surveillance data were collected from 22 county-level administrative areas in Chengdu between 2023 and 2025, with meteorological data obtained from the Chengdu Meteorological Bureau. The Wilcoxon signed-rank test was used to compare mosquito density between urban and rural habitats, while NMDS, PERMANOVA, and SIMPER analyses were performed to evaluate differences in mosquito community structure. A generalized additive model (GAM) was applied to evaluate nonlinear responses of mosquito density to meteorological factors. Using an 80:20 split grouped by collection date, we compared GAM, random forest (RF), and extreme gradient boosting (XGBoost) performance and characterized temperature–sunshine joint responses. Mosquito density showed a unimodal seasonal pattern from 2023 to 2025, peaking in June or July. Higher densities were observed in peripheral areas (e.g., Chongzhou, Dujiangyan, and Qionglai), whereas central urban areas showed lower densities. Culex tritaeniorhynchus and Culex pipiens pallens were the two dominant. Mosquito density was significantly higher in rural than urban habitats ( P  < 0.001), and mosquito community structure differed significantly between urban and rural habitats ( P  = 0.001). GAM revealed significant nonlinear associations of mean daily temperature and cumulative sunshine duration over lag days 1–7 with mosquito density (both P  < 0.001). Mosquito density increased significantly between 15.59 and 22.44 °C and reached an estimated peak at 25.02 °C. Model comparisons showed that RF achieved the highest predictive performance, followed by XGBoost. Mosquito density in Chengdu exhibits significant spatiotemporal heterogeneity, with temperature and sunshine duration identified as key meteorological factors. Machine learning models showed superior predictive performance over GAM. To prevent and control mosquito-borne diseases, meteorological factors should be incorporated into the existing monitoring system.