Combination of Multiple Isotopes and PMF Model Provide Insights Into the Method Optimization of PM_2.5 Source Apportionment During Haze Episodes

Abstract The key problems with Positive Matrix Factorization (PMF) model for PM 2.5 source apportionment were inconsistent results with different species selections and a lack of evaluation criteria for results accuracy. Moreover, high proportions of secondary inorganic aerosols sources (SNA) were identified by PMF without corresponding primary sources. This study develops a new method that combines multi‐isotopes ( 34 S, 15 N, 18 O and 14 C) and PMF model to optimize source apportionment. Data sets A–F, constructed from PM 2.5 components, were input into PMF model to obtain optimal results (3–9 factors), which changed with the selection of species. Specifically, the contributions of coal combustion (CC, 3%–36%), biomass burning (BB, 11%–38%), and vehicle sources (VS, 4%–15%) showed significant differences in data sets, indicating that conventional methods cannot obtain accurate results. Then, 15 N, 34 S, 18 O were introduced to restrict and reallocate identified SNA sources to primary sources, overcoming the influence of species on results. Additionally, 14 C was used to evaluate data sets results, which showed that the combination of PMF model with more markers (data set F, 9‐factor) and multi‐isotopes techniques obtained optimized results that aligned with 14 C results. Compared with the initial results, the contributions of CC, VS, and BB in the allocated 9‐factor increased by 26.4%, 5%, and 19.5%, respectively, becoming main sources of PM 2.5 . This study represents the first time that combination of PMF model and multi‐isotopes achieves SNA sources reapportionment and results evaluation, improving source apportionment methods.