Environmental drivers of mosquito breeding dynamics in urban Sudan: a 22-year spatio-temporal analysis from Khartoum (2001–2022)

Mosquito breeding in arid African cities is shaped by a narrow set of environmental drivers, but long-term records linking habitat-level counts to climate covariates are uncommon. Multi-decadal datasets are needed to inform vector control strategy under shifting rainfall and flood regimes. We analyzed 22 years (2001–2022) of monthly mosquito breeding surveillance from 14 habitat types in Khartoum, Sudan, integrated with NASA POWER and ERA5 environmental data. Zero-Inflated Negative Binomial Spatio-Temporal (ZINB-ST) models quantified relationships between environmental covariates and breeding intensity. The dataset comprised 3522 site-month observations with 95.3% completeness. Breeding showed extreme spatial heterogeneity (~ 1000-fold variation between habitat types), with three habitat types (agricultural schemes, drainage systems, and broken pipelines) contributing 87.9% of total production. Strong seasonality emerged with August-October peaks (~ 2× dry-season baseline). Lagged precipitation (1–2 months) was the dominant predictor (IRR=1.003 per mm, 95% CI 1.002–1.004; p < 0.001). Active flooding was a significant negative predictor of breeding intensity (IRR = 0.814, 95% CI 0.695–0.953; p = 0.011), consistent with a descriptive 35–40% suppression during flood-affected peak-season months. Model validation achieved out-of-sample correlation r = 0.423 (R 2 =0.179), outperforming simpler specifications (ΔAIC > 1000). Spatial concentration supports Pareto-based prioritization, with the top 5 habitat types accounting for over 70% of production. The 1–2 month precipitation lag offers a window for pre-emptive larviciding. The 2020 flood (highest Nile levels since 1912) is a natural experiment on breeding response to extreme rainfall, with relevance for vector control in urban settings.