Receptor models have been widely used to identify pollution sources in the urban environment. However, evaluating the accuracy of source apportionment results for road deposited sediments (RDS) using these models has not been the focus of previous studies. This study compared canonical receptor models, i.e., positive matrix factorization (PMF), Unmix, chemical mass balance (CMB) and chemical mass-balance based stochastic approach (SCMD) using six synthetic datasets generated from real-world source profiles, and three error evaluation indicators (ie., relative error (RE), relative prediction error (RPE), and symmetric mean absolute percentage error (SMAPE)) were employed. The SCMD model showed more stable and accurate results, with ranges from 8.48 % - 30.76 %, 16.32-32.34 %, and 7.81-24.55 % of RE, RPE, and SMAPE, respectively. SCMD was then applied for tracking Pb, Zn, Cr, Cu, Ni, and Mn on urban road surfaces in Guangzhou, China. The results showed that vehicle exhaust, tire wear, roadside soil, and brake wear contributed 50.15 %, 41.15 %, 6.84 %, and 1.86 % of the mass of particulate solids, respectively; vehicle exhaust contributed more than half of these six heavy metals, particularly Cr and Ni. These findings provide scientific support for the effective selection of appropriate receptor models for source apportionment in RDS.
Keywords: Heavy metals; Receptor models; Road deposited sediments; Source apportionment; Synthetic dataset.
Copyright © 2024 Elsevier B.V. All rights reserved.