OVOCs source analysis based on an improved source apportionment model and its influencing factors: A case study of a dense urban agglomeration in the winter

Oxygenated volatile organic compounds (OVOCs) affect the formation of atmospheric ozone (O₃), secondary organic aerosol (SOA), and free radicals, and have complex sources such as anthropogenic and biogenic direct emissions and through series of secondary oxidation reactions of nonmethane hydrocarbons (NMHCs). However, understanding sources of OVOCs in the atmosphere still has large uncertainties. In this study, an improved OVOC source apportionment model was developed by principal component analysis (PCA) and multiple linear regression (MLR) based on the online monitoring of NMHCs and OVOCs in a dense urban agglomeration in the winter. The modelled concentrations were in good agreement with the measured concentrations (R²=0.56-0.97). The concentrations of major OVOCs, except for 2-methylacrolein, were greatly affected by anthropogenic sources (15.8-76.8%) and secondary generation (0.0-51.7%), while transport and natural sources contributed to 0.0-26.8% and 0.0-32.0%, respectively. The selection of isoprene as the natural tracer led to an underestimation of the OVOC species from primary emission and an overestimation from natural sources. In addition, photochemical reactions significantly reduced the simulation accuracy of the model for NMHCs in the afternoon, with the R² of 0.60 ± 0.23, which was lower than the overall value of 0.82 ± 0.11. However, the R² for OVOCs (0.83±0.14) did not decrease significantly in the afternoon due to the compensation of secondary oxidation. Furthermore, the concentration gradient distribution of the species gradually changes from a normal distribution to an exponential normal distribution with a decrease in concentration, the accuracy of the model was influenced by the degree of matching between tracer and species concentration gradient as species concentration change. Developing models with additional tracers at different concentration levels may enhance the robustness of the OVOC source apportionment model without increasing its complexity.