Application evaluation of APCS-MLR and PMF in source apportionment of heavy metals in farmland soil

Purpose: The purpose of this study was to compare and evaluate the applicability of two widely used receptor models, absolute principal component score-multiple linear regression (APCS-MLR) and positive matrix factorization (PMF), in the source apportionment of heavy metals in farmland soil. Methods: A total of 40 heavy metal concentration data collected from a typical polluted farmland were used to evaluate the accuracy of APCS-MLR and PMF in identifying the sources of heavy metals in farmland soil by analyzing the fitting degree between predicted and observed values and the characteristics of sources. Results: Through APCS-MLR, three sources of heavy metals in farmland soil were obtained: industrial and vehicle emission (44.2%), natural source (33.8%), and agricultural activity (22.0%). PMF further clarified the contributions of industrial and vehicle emissions, identifying four sources: industrial production (32.0%), natural source (28.2%), agricultural activity (25.8%), and vehicle emission (14.0%). Moreover, the correlation coefficient between predicted and observed values in PMF was higher than that in APCS-MLR, and the error of PMF for simulating the predicted values was lower than that of APCS-MLR, indicating that PMF was more accurate compared to APCS-MLR. Conclusions: PMF was more effective in the application of source apportionment of heavy metals in farmland soil. However, there is a certain degree of uncertainty in the factor contributions obtained in PMF, as its estimation performance for heavy metals with lower contribution percentages in the factor is limited.