Seasonal water-source transitions across livelihood systems in Somalia: Spatial heterogeneity and household correlates

Ahmed Hussein Ismail 

Water access in dryland settings is commonly represented by a single primary drinking-water source, although households may reorganize their water supply as rainfall, source availability, mobility, and market access change between seasons. This study examines rainy-to-dry water-source transitions across livelihood systems in Somalia using the nationally representative 2022 Somalia Integrated Household Budget Survey (SIHBS; N = 7,212 households). Individual rainy- and dry-season drinking-water sources were consolidated into six source systems, and a substantive seasonal transition was defined as movement between source systems rather than a change within the same system. Agricultural participation was reconstructed from the agricultural-land roster and reported crop production, livestock participation from the livestock-ownership roster, and households were classified as agriculture only, livestock only, mixed crop-livestock, or other. Survey-weighted descriptive analysis, transition matrices, nested weighted logistic models, destination-specific analyses, an agricultural subsystem analysis, and complementary machine-learning models with region-grouped cross-validation were used. Any rainy-to-dry source change was reported by 17.25% of households, while 12.00% experienced a substantive source-system transition. Weighted transition prevalence was highest among mixed crop-livestock households (25.1%) and livestock-only households (19.6%), compared with agriculture-only (11.5%) and other households (7.8%). The dominant pathway was rain/surface water to groundwater (24.1% of substantive transitions), followed by rain/surface water to delivered/vendor water (13.2%). Regional prevalence ranged from 44.6% in Bakool to 0.7% in Banadir. After adjustment for residence, household characteristics, region, and rainy-season source system, mixed crop-livestock (OR 1.94, 95% CI 1.40–2.68) and livestock-only households (OR 1.46, 95% CI 1.16–1.83) remained more likely to experience substantive transition than other households, while rain/surface dependence showed the largest association (OR 11.80, 95% CI 8.93–15.60, versus private piped sources). Predictive performance declined when complete regions were held out, and geographically validated permutation importance identified rainy-season source family as the dominant transferable predictor. The findings show that static access classifications conceal an important seasonal reconfiguration of household water systems, particularly in livestock and mixed livelihood settings. Seasonal monitoring in Somalia should therefore complement conventional source-status indicators with information on source transitions, livelihood context, and regional heterogeneity.