Source apportionment based on the comparative approach of two receptor models in a large-scale region in China

Abstract

Soil heavy metal(loid) (HM) source apportionment is the prerequisite to develop suitable mitigation measures and make pollution control and prevention regulations. The selection of appropriate tools (models) for source analysis is crucial, that is especially true for large-scale regions, as the Pearl River Delta (PRD), due to the high spatial variability in soil parent materials, soil topographical feature, and wide range of anthropogenic activities. The objective of this study is to evaluate the potential applications of receptor models (positive matrix factorization [PMF] and Unmix) which have been widely used in air pollution research to quantitatively apportion sources of heavy metal(loid)s in the soils. To assist the interpretation of the derived factors (sources) of the receptor models, enrichment factors and GIS mapping were used to identify the potential relationships between the factor contributions and human activities in the study area. As the models are built on completely different algorithms, a comparative approach was adopted in addition to evaluate the impact of sample size on the model results. Factor profiles generated by different receptor models were quite similar as well as their corresponding factor contributions spatial distribution. Though the stability of their results decreases with a reduced sample size, the results of PMF were less significantly influenced by the sample size than those of Unmix. Due to the difficulty (time consuming and expensive) of soil sample collection in large-scale regions, the PMF model appears to be practically more effective than Unmix. In addition, further investigation is needed for Unmix model to understand the reason for its high sensitivity and determine an appropriate sample size.