Estimating groundwater geogenic arsenic contamination and the affected population of river basins underlain mostly with crystalline rocks in Ghana

Arsenic is a human carcinogen and contributes to the incidences of cancer, diabetes, and skin diseases in the exposed populations. Therefore, its occurrence in drinking water is of public health concern worldwide, especially in the Tano, Pra, and Ankobra River Basins in Ghana. The metasediments and metavolcanic formations which underlie these basins (42,092.5 km2) host arsenic sulphide minerals that are a source of arsenic contamination of the weathered and fractured aquifers that serve as drinking water supply to most of the population (∼9.8 million), especially in the rural areas. A random forest machine learning method and partial dependence plots were applied to maps of spatially continuous environmental factors (fault density, surface bulk density, coarse fragment and soil pH; and sub-surface silt and soil pH) and hydrochemical variables, respectively, and arsenic measurements from wells (n = 229) which included 57 new datasets of which 7.86 % exceeded the 10 µg/L to develop arsenic prediction models for the basins at thresholds of 1 µg/L, 5 µg/L, and 10 µg/L. Fault density and coarse fragments were more reliable contributors to the models. A partial dependence plot and Spearman's rank correlation analysis showed a significant positive correlation (p < 0.01) between arsenic enrichment and HCO3−, pH, Fe, and Mn, except NO3− in groundwater. Arsenic hotspots occurred in urban districts in the Ashanti Region and known gold-mining areas in Nzema, Obuasi, Prestea, and Tarkwa. Spatially, arsenic contamination in groundwater covered 49.44 % of the land area of the basins. By cross-referencing population density estimates and the arsenic hazard map,15.8 % of the population is exposed to arsenic-contaminated groundwater in the basins. The hazard model will facilitate efforts to reduce rural poverty and protect public health by delivering potable water in Ghana. The hazard models apply to crystalline terrain in other sub-Saharan African countries.