Applicability of Regression Models for Predicting Manganese and Lead Concentrations in Soils of Bauchi Local Government Area, Nigeria

Abstract

Heavy metal contamination in soils poses a significant environmental and public health 
concern, particularly in rapidly urbanizing areas where industrial, agricultural, and domestic 
activities contribute to elevated trace-metal levels. This study assesses the applicability and 
performance of regression-based statistical models for predicting the concentrations of 
manganese (Mn) and lead (Pb) in soils from selected contaminated sites within Bauchi Local 
Government Area, Nigeria. The overarching aim is to develop empirical relationships that can 
reliably estimate heavy-metal concentrations from easily measurable soil physico-chemical 
properties, thereby reducing the need for frequent and costly laboratory analyses. Soil 
samples were systematically collected from three representative sites at a depth of 30 cm, 
corresponding to the active root zone and potential contaminant accumulation layer. 
Laboratory analyses were performed using standard procedures: concentrations of Mn and 
Pb were quantified using an Atomic Absorption Spectrophotometer (AAS), while pH, 
electrical conductivity (EC), and moisture content (MC) were measured to characterize the 
physico-chemical properties of the soils. Descriptive and inferential statistical analyses were 
employed to evaluate spatial variations and the strength of relationships between metal 
concentrations and the measured soil parameters. Results revealed spatial heterogeneity in 
the distribution of Mn and Pb across the study sites, with mean concentrations generally 
exceeding background values reported for uncontaminated soils in similar geological settings. 
At a significance level of P < 0.05, both Mn and Pb exhibited strong and positive correlations 
with one or more of the physico-chemical variables. Regression modeling demonstrated that 
several linear models provided excellent fits to the observed data, with adjusted R² values 
exceeding 0.70, indicating high predictive accuracy and model reliability. The diagnostic tests 
further confirmed the absence of multicollinearity and heteroscedasticity, validating the robustness 
of the developed equations.The established regression models are therefore considered 
suitable tools for rapid and cost-effective prediction of Mn and Pb concentrations in soils of 
Bauchi and similar environments. These models offer practical advantages over conventional 
metal-determination techniques by minimizing analytical cost, field effort, and time, while 
maintaining acceptable precision and reproducibility. Consequently, the approach provides an 
efficient framework for routine monitoring, environmental assessment, and risk management 
of heavy-metal contamination in urban and peri-urban soils. The findings further highlight 
the potential of integrating statistical modeling with geochemical monitoring for sustainable 
soil-quality management and early-warning systems in developing regions.