Comparison of receptor models for source identification of organophosphate esters in major inflow rivers to the Bohai Sea, China

Abstract

A better understanding of the sources of organophosphate esters (OPEs) is a prerequisite for OPE control and the establishment of related environmental policies. Sources of OPEs in 35 major inflow rivers to the Bohai Sea of China were quantitatively analyzed using three effective receptor models (principal component analysis-multiple linear regression (PCA-MLR), positive matrix factorization (PMF), and Unmix) in this paper. The similarities and differences in results from PCA-MLR, PMF, and Unmix were discussed in depth. All three models well predicted the spatial variability of the total concentrations of nine OPEs (triethyl phosphate, tri-n-butyl phosphate, triisobutyl phosphate, tri (2-ethylhexyl) phosphate, tri (2-chloroethyl) phosphate, tris(1-chloro-2-propyl) phosphate, tris(1,3-dichloro-2-propyl) phosphate, triphenyl phosphate, and triphenylphosphine oxide) (∑9OPEs) (r2 = 0.90–0.96, p = 0.000) and explained 98.4%–101.2% of the observed ∑9OPEs. The predicted ∑9OPEs values from each pairwise model were significantly correlated (r2 = 0.88–0.91, p = 0.000). Three OPE sources were extracted by all three models: rigid and flexible polyurethane foam/coating, cellulosic/acrylic/vinyl polymer/unsaturated polyester, and polyvinyl chloride, contributing 49.9%, 29.7%, and 20.5% by PCA-MLR, 57.9%, 28.6%, and 13.5% by PMF, and 47.9%, 30.8%, and 22.4% by Unmix to the ∑9OPEs, respectively. PMF was recommended as the preferred receptor model for analyzing OPE sources in water during the monitoring period because of its optimal performance.