Source appointment of potentially toxic elements (PTEs) at an abandoned realgar mine: Combination of multivariate statistical analysis and three common receptor models

Identifying pollution sources and quantifying their contributions are of great importance for proposing management and control strategies of potentially toxic elements (PTEs) in soil. In this study, multivariate statistical analysis and receptor models were combined to identify potential pollution sources and apportion their contributions at an abandoned realgar mine. Principal component analysis (PCA) result shows that three factors are responsible for PTEs, which is also supported by cluster analysis (CA). Correlation analysis and spatial analysis also show that the heavy metals from the same pollution source are of higher correlation coefficients and similar spatial distribution. Three receptor models were combined to apportion contributions of pollution sources. Three pollution sources were detected by absolute principal component analysis-multiple linear regression (APCA-MLR). In contrast, four sources were identified by positive matrix factorization (PMF) and UNMIX. Soil parent material was heavily loaded on Cr, Cu, Ni and Zn, occupying the largest average contribution (30%-43%). Cadmium was mainly derived from agricultural activities with contribution higher than 60%. Arsenic accumulation was mainly associated with mining and smelting activity with contribution higher than 80%. PMF and UNMIX models showed that more than half of Pb concentrations were influenced by industrial activities. Comparatively speaking, APCA-MLR was a well-performing model for all PTEs even though it only detected three pollution sources. The study showed that it was a good choice to apply multiple receptor models in order to achieve more reliable and objective conclusions of source appointment.