Competing geochemical drivers and probabilistic health risk assessment of groundwater in the semi‑arid Voltaian sedimentary basin, Ghana

Groundwater in semi-arid sedimentary basins is a critical rural resource, but quality assessment is often weakened by geochemical complexity, compositional closure, class-boundary uncertainty and incomplete exposure information. In the Voltaian sedimentary basin of Ghana, previous studies have commonly applied sequential assessment pipelines that diagnose hydrochemical processes, water quality, health risk and irrigation suitability separately. This study evaluates groundwater quality and health risks using 34 dry-season samples collected from public boreholes in Karaga District and introduces a dual-space coupled framework that links closure-safe process discovery with uncertainty-aware decision modelling. The framework combines isometric log-ratio principal component analysis, neutrosophic partial least squares groundwater quality indexing (NPLS-GWQI), Bayesian horseshoe driver screening, saturation-index auditing, probabilistic fluoride/nitrate health-risk assessment and neutrosophic irrigation suitability indexing. Groundwater was mainly Na-HCO 3 type and reflected rock–water interaction, cation exchange and mineralisation gradients. The NPLS-GWQI classified 25 of 34 samples (73.5%) as excellent, whereas 8 samples (23.5%) were degraded and 4 samples (11.8%) were unsuitable. Fluoride was not controlled by a single statistically supported driver; weak exploratory signals instead linked fluoride to exchange state, mineralisation and weathering proxies. Children showed the highest fluoride-related skeletal-dental risk, with a 69% posterior probability of hazard-index exceedance under the dry-season ingestion scenario, whereas nitrate-related haematological risk remained sub-threshold. These findings support confirmatory fluoride monitoring, child-focused risk communication and site-specific consideration of treatment or source substitution at the sampled high-risk boreholes. The workflow is presented as a documented screening approach whose wider applicability requires independent testing with larger, seasonally replicated datasets.