Salinization is one of the main threats to groundwater quality worldwide, affecting water security, crop productivity and biodiversity. The Horn of Africa, including eastern Ethiopia, northeast Kenya, Eritrea, Djibouti, and Somalia, has natural characteristics favouring high groundwater salinity. However, available salinity data are widely scattered, lacking a comprehensive overview of this hazard. To fill this gap, machine learning modelling was used to spatially predict patterns of high salinity with a dataset of 6300 groundwater quality measurements and various environmental predictors. Maps of groundwater salinity were produced for thresholds of 800, 1500 and 2500 μS/cm. The main drivers include precipitation, groundwater recharge, evaporation, ocean proximity, and fractured rocks. The combined overall model accuracy and area under the curve of multiple runs were both ~81%. The salinity maps highlight the uneven spatial distribution of salinity, with the affected areas mainly located in arid, flat lowlands.
These novel and high-resolution hazard maps (1 km2 resolution) further enable estimating the population potentially exposed to hazardous salinity levels. This analysis shows that about 11.5 million people (~7% of the total population) living in high-salinity areas, including 400,000 infants and half a million pregnant women, rely on groundwater for drinking. Somalia is the most affected country, with an estimated 5 million people potentially exposed. The created hazard maps are valuable decision-support tools for government agencies and water resource managers in helping direct salinity mitigation efforts