Mapping the heterogeneous influences on soil lead concentration using geographically weighted regression in northwestern Nigeria

Soil contamination with heavy metals poses major environmental and public health concerns, especially in developing countries where artisanal mining and agricultural intensification are widespread. Traditional regression models often assume spatial stationarity, ignoring localized variability in soil processes. This study applies Geographically Weighted Regression (GWR) to assess spatially varying relationships between lead (Pb) concentrations and auxiliary heavy metals (Al, Zn, Mn, Ti, Fe, and As) in the Sokoto Basin, Northwestern Nigeria. A total of 103 topsoil samples were collected and analyzed using Proton Induced X-ray Emission (PIXE). GWR was compared with Ordinary Least Squares (OLS) to evaluate model performance. Results showed that GWR achieved a higher adjusted R² (0.8011) compared to OLS (0.6710) and reduced the corrected Akaike Information Criterion (AICc) by 123 points. Local coefficients revealed strong variability in the influence of Fe and Mn on Pb, reflecting lithogenic processes, while Zn and As exhibited localized anthropogenic effects. Prediction surfaces highlighted hotspots of Pb contamination in the central and northeastern basin. These findings demonstrate the superiority of GWR for modeling heavy metals in heterogeneous landscapes and provide critical insights for targeted environmental monitoring and remediation.