Groundwater pollution source identification and apportionment using PMF and PCA-APCA-MLR receptor models in a typical...

The quality of groundwater in a region is regarded as a function of natural and anthropogenic factors. Receptor models have advantages in source identification and source apportionment by testing the physicochemical properties of receptor samples and emission sources. In our study, receptor models PMF and PCA-APCS-MLR were developed to qualitatively identify the latent sources of groundwater pollution in the study area and quantitatively evaluate the contribution of each source to groundwater quality. The performances of PMF and APCS-MLR models were compared to test their applicability on the assessment of groundwater pollution sources. Results showed that both of the models identified five sources of groundwater contamination with similar main load species of each potential source. The comparable source apportionment of species NO₂^- and NO₃^- with two models indicated the reliable source estimation for these species, whereas the contributions of sources to species Fe, Mn, Cl^-, SO₄^2- and NH₄⁺ were significantly different due to the large variability of data, difference of uncertainty analysis and algorithm of unexplained variability in the two models. R-squared value between observation and model prediction was 0.603-0.931 in PMF and 0.497-0.859 in PCA-APCS-MLR. The significant disagreement of average source contribution was detected in agricultural source and unexplained variability using PMF and PCA-APCS-MLR models. Average contributions of other sources to groundwater quality parameters had similar estimates between the two models. Higher R² and smaller proportion of unexplained variability in the PMF model suggested that PMF approach could provide more physically plausible source apportionment in the study area and a more realistic representation of groundwater pollution than solutions from PCA-APCS-MLR model. The study showed the advantages of application of multiple receptor models on achieving reliable source identification and apportionment, particularly, providing a better understanding of applicability of PMF and PCA-APCS-MLR models on the assessment of groundwater pollution sources.