Analysis of Groundwater Chemistry and Water Source Characteristics in the Banji Mining Area Based on a Dual Model Combining APCS-MLR and PMF

Abstract

Quantitative hydrochemical source apportionment underpins precise mine water inrush prevention and inter-aquifer pollution control. Ninety-two groundwater samples from four aquifers in the Anhui Banji Mining Area were tested for six major ions (Ca2+, Mg2+, Na++K+, Cl−, SO42−, HCO3−). Piper diagrams and Pearson correlation analysis identified dominant HCO3-Ca·Mg hydrochemical facies, with some Quaternary pore water evolving into the Cl·SO4-Na type due to evaporite leaching. Parallel Absolute Principal Component Scores-Multiple Linear Regression (APCS-MLR) and Positive Matrix Factorization (PMF) models were used for multi-source contribution quantification and crossverification. APCS-MLR extracted three factors (83.99% cumulative variance) representing evaporite leaching, carbonate dissolution and silicate weathering, but produced unassigned residuals and negative contributions due to its unconstrained linear algorithm. Validated by the Root Mean Square Error (RMSE) elbow rule and Bootstrap test, PMF selected three optimal factors with stronger endmember indication, all non-negative contributions summing to 100% per ion. Both models agreed on key water-rock reactions, while PMF performed better in end-member recognition and robustness. This dual-model mutual verification improves the reliability of single-method results, reveals hydrochemical evolution driven by structures and mining disturbances, and offers technical support for water inrush source identification in Huainan coal mines.