Descriptive and predictive analysis of trace metals in coal ash: statistical insights and modeling

Abstract

Coal combustion generates ash that contains trace metals with both economic and environmental relevance. This study aims to assess the concentration, variability, and inter-element relationships of nine trace metals (Sb, Co, Cu, Ga, Mo, Ni, Ag, V, Zn) in ash samples from 28 coal specimens of varying rank (lignite, subbituminous, and bituminous) collected from Polish deposits. Elemental concentrations were determined via ICP-MS, and oxide composition was analyzed to examine geochemical associations. Descriptive statistics, correlation analysis, and predictive modeling (multiple linear regression and Support Vector Machine, SVM) were applied to characterize elemental behavior and identify reliable predictors of metal content in ash. The results show substantial variability in elemental concentrations, with several metals (e.g., Ni, V, Zn) enriched in ash relative to parent coal. Bituminous coal ashes generally exhibited higher average concentrations than lignite ashes, except for Zn. Strong linear correlations were found between selected metal pairs (e.g., Cu–Co, Ag–Co), while correlations with oxides (notably Al₂O₃, Fe₂O₃, Cr₂O₃) supported their role in controlling metal distribution. Regression models demonstrated predictive capability for most elements (R2 up to 0.88), with SVM models showing improved performance (R² up to 0.94, MAPE as low as 19.78%). These findings highlight the importance of oxide composition in trace metal behavior and provide a methodological basis for assessing ash quality, environmental risk, and potential resource recovery.