Mass spectral characterization of secondary organic aerosol from urban 1 lifestyle sources emissions 2

Wenfei Zhu1, Song Guo1,2*, Min Hu1, Zirui Zhang1, Hui Wang1, Ying Yu1, Zheng Chen1, Ruizhe Shen1, Rui 3 Tan1, Kai Song1, Kefan Liu1, Rongzhi Tang1, Yi Liu1, Shengrong Lou3, Yuanju Li1, Wenbin Zhang4, Zhou 4 Zhang4, Shijin Shuai4, Hongming Xu4, Shuangde Li5, Yunfa Chen5, Francesco Canonaco6, Andre. S. H. 5 Prévôt6 6 1 State Key Joint Laboratory of Environmental Simulation and Pollution Control, International Joint Laboratory for 7 Regional Pollution Control, Ministry of Education (IJRC), College of Environmental Sciences and Engineering, Peking 8 University, Beijing 100871, China P. R. 9 2 Collaborative Innovation Center of Atmospheric Environment and Equipment Technology, Nanjing University of 10 Information Science & Technology, Nanjing 210044, China P. R. 11 3 State Environmental Protection Key Laboratory of Formation of Urban Air Pollution Complex, Shanghai Academy of 12 Environmental Sciences, Shanghai 200233, China P. R. 13 4 State Key Laboratory of Automotive Safety and Energy, Tsinghua University, Beijing 100084, China P. R. 14 5 State Key Laboratory of Multiphase Complex Systems, Institute of Process Engineering, Chinese Academy of Sciences, 15 Beijing 100190, China P. R. 16 6 Laboratory of Atmospheric Chemistry, Paul Scherrer Institute (PSI), Villigen 5232, Switzerland 17 18 Corresponding authors: 19 *Song Guo − State Key Joint Laboratory of Environmental Simulation and Pollution Control, College of 20 Environmental Sciences and Engineering, Peking University, Beijing 100871, China P. R.; Email: 21 songguo@pku.edu.cn 22 23 Abstract In the present work, we conducted experiments of secondary organic aerosol (SOA) formation from urban 24

While the paper addresses a relevant and longstanding question of atmospheric chemistry (constraining secondary organic aerosols), the present scientific and technical quality of the paper is lacking in multiple aspects. I recommend that this manuscript be reconsidered for publishing after major revisions.

Major Comments
1) The paper presents the cooking tests and results as original work of this paper, and references published work (Zhang et al., 2020) incorporating those tests and results mostly in the Methods section (exception being Line 196). However, the ACP similarity report revealed large sections of this paper discussing the cooking results (for example, Lines 172-176, 207-221) are almost verbatim from published work (Zhang et al., 2020), a clear and unfortunate case of self-plagiarism. The authors should add explicit references and paraphrasing (if taking verbatim text) to all such portions of the paper.
2) The authors present the use of mass spectral similarity analysis in the methods section and discuss five categorizations to be used in the rest of the paper. However, they often deviate from using the categories to describe results. For example, in lines 159-170, 176. they use phrases such as "almost resembled", "different", "similar", and "almost the same variation" to describe mass spectral comparisons instead of using the five qualitative categories introduced in the paper in Sect.  (7)). 4) The PMF/ME-2 analysis presented in this paper has multiple shortcomings, both in terms of descriptions in the methods section, as well as the analysis and presentation of results. a. In the methods section, there is no mention of how the authors conducted ME-2 analysis on the datasets in this study. The Igor PET tool runs on PMF2.exe and does not have a ME-2 option. b. PMF analysis based on mass spectral similarity analysis only has previously been shown to generate spurious factors (Ulbrich et al., 2009). Other analyses such as time-series correlations with external tracers need to be presented to justify PMF/ME-2 factors. However, such correlations have been presented only for MO-OOA factor in summer and LO-SOA factor in winter, and not for other factors. Similarly, the other-POA factor could be a mix of the HOA and the COA factors (in the same 2-D plane as defined by the two vectors), and this should be checked using the scalar triple product. Refer to Ulbrich et al, 2009 for more details. c. The authors use PMF to separate POA and SOA factors from aged HOA and COA detected in this study. However, using single MS to represent entire time series data in a test is an obvious limitation of PMF that has not been explicitly recognized. I suggest the authors recognize this as a limitation explicitly. It is also unclear how the references for the application of the PMF technique (line 118) are relevant since they are applying PMF on ambient and not lab datasets. Also, it is unclear how this analysis was conducted. Were different EPA tests combined for each type (vehicle operation, food dish) and then PMF conducted? Or was PMF conducted separately for each experiment? d. Why was other-POA in winter not identified as associated with a specific POA component such as BBOA or CCOA, given ambient source apportionment results from Chinese cities (including Shanghai) are readily available from earlier literature? The low levels of contributions at m/z 60, m/z 73, and m/z 115, which are tracers of biomass burning and coal combustion make the argument that this other-POA factor is associated with biomass burning or coal combustion weak. What reference profile does the mass spectral similarity analysis suggest this factor resembles? What evidence do we have with respect to time series correlations? e. In Section 3.3, the authors compare their approach (of using constrained POA and SOA) to the completely unconstrained PMF approach. However, the improvement of ME-2 for primary factors over unconstrained PMF has already been presented in recent work such as Zhu et al., 2018. So, a more appropriate question to address would be: how much of an improvement do we observe in the ME-2 method when both primary and secondary factors are constrained (compared to when only the primary factors are constrained)? Given the PMF and ME-2 runs the team has already conducted, such a comparison should not be hard to perform, and will give much more substantial insight into the importance of the approach compared to the current presentation. Another result that could arise from this comparison is that constraining the secondary factors could be overconstraining the PMF runs, which leads to factor mixing and reduced number of factors. Interestingly, Zhu et al., 2018 were able to separate coal combustion and biomass burning cleanly in winter during heavily polluted periods using their only primary factor-constrained ME-2 approach. f. The final choice of constraints using ME-2 was described in vague terms in lines 259-260 and lines 265-269. "Considering the actual oxidation conditions or the concentration of OH radicals, the cooking PMF POA, SOA, and the vehicle PMF LO-SOA was finally selected as the input source spectra of ME-2… In addition, the ME-2 source analysis was performed by using two primary OA factors (the cooking PMF POA, HOA resolved in three cities) and two secondary OA factors (the cooking PMF SOA, the vehicle PMF LO-SOA) as constraints based on the same ambient OA datasets as PMF model during the summer and winter observations of Shanghai." This is insufficient explanation. Why were vehicle POA and vehicle MO-SOA factors from lab tests not selected? Why was HOA resolved in three cities selected? This seems an arbitrary choice and needs to be justified further so the approach can be replicated in the future. Also, was the average of the HOA MS from three cities selected? I did not find the MS of that factor in the paper. How similar or different is it from the lab vehicle HOA MS, and why? g. Factor uncertainties, residual, and total concentrations should be reported for each PMF/ME-2 analysis. 5) The conclusions of the paper are very generalized and presented as applicable to broad categories of cooking and vehicular emissions in ambient environments. However . Is there evidence to support that only the quantities (of emissions) vary across vehicles under similar operating conditions, but not the mass spectral patterns? If not, please show this as a limitation of the study. 6) The authors use dilution and high concentration of OH radicals (for brief period) to measure and simulate atmospheric aging of aerosols. However, these are both limitations since: 1) dilution changes the chemistry of aging, as also observed with volatility measurements (Cain et al., 2020), and 2) high concentrations of OH radicals could lead to changes in the reaction pathways that the aerosols undergo that are different compared to pathways on exposure to low OH concentrations for longer periods of time (but resulting in the same EPA). I suggest the authors discuss these aspects in a separate section on limitations of such work, as described in (7). 7) To address the above limitations, I suggest the authors separate the limitations briefly described in the conclusions section (Lines 304-308) and create a separate section on "limitations and future work", where the authors can identify all the above gaps. They can also point readers to potential future work that can emanate out of this preliminary but notable effort. 8) Finally, the title of the paper is misguiding since "lifestyle sources emissions" would also point to volatile chemical products such as perfumes, cleaning products, and deodorants. I suggest the authors change it to "urban cooking and vehicular sources".
After addressing comments for the major revision associated with the comments above, I suggest the authors address the following minor comments in the updated manuscript before resubmission.
Minor comments 1) In Sect. 2.1, lines 91-93, the authors describe vehicle operating conditions in terms of vehicle speeds and torques. However, given the goal of the paper is to use lab tests to describe and apportion real-world emissions, what do these rpm speeds and Nm torques mean in terms of real-life conditions? Would you describe the real-life conditions in terms of vehicle speed (in mph) and rate of gain of elevation? An equivalence of each speedtorque combination would be immensely useful in understanding how relevant these combinations are to real-life conditions. Are these combinations relevant more to flat terrains in heavy traffic? Or are they more relevant to mountainous terrains with low traffic? It is hard to draw analogies to real-life conditions based on speeds and torques only. 2) In Sect. 2.2, the authors should clearly state the type of aerosols being measured, whether they are NR-PM2.5 or NR-PM1. 3) In Sect. 2.2, lines 108-110, the reference describing how CO2 interference can be reduced using CO2 gas phase measurements needs to be added. with the most prominent peaks at m/z 28 and 44, respectively, which almost resembled the mass spectra of MO-OOA resolved from ambient datasets." Reference to Table S4 is missing! 9) In lines 164, 166, and elsewhere, the authors mention ambient profiles. However, the criteria used to obtain these profiles is not clear. The methods section should be updated to clarify this point. 10) The authors supply tables Table 1 and S2-S3 in SI for MS similarity analysis for all vehicle operating conditions and for all food types at two EPAs. Similar tables should be supplied for the remaining EPAs for vehicles conditions and all food types.
11) The authors supply four tables S4-S7 in SI for MS similarity analysis for two vehicle operating conditions and for two food types at varying EPA. Similar tables should be supplied for the remaining three vehicle conditions and for the other two food types. 12) Lines 173-175: "Along with the growth of OH exposure, the ƒ43 of aged COA increased from 0.07 to 0.10, and meanwhile its ƒ44 increased from 0.03 to 0.08 ( Fig.2b; Fig.S5), distributing in the lower region of less oxidized organic aerosol (LO-OOA)." There is a missing reference here since the LO-OOA region is undefined. For that matter, even in Fig. S6, that region has not been defined. 13) The authors discuss specifics of mass spectral contributions of different mass spectra, which are hard to decipher from the figures (e.g., lines 174, 216-218). I suggest the authors add supplementary tables of contributions at key m/zs for the different tests: vehicle/cooking type and operating condition. 14) Line 199: In   Fig. S4, Table S11: Why is any comparison with vehicle POA MS missing in these two? 21) Lines 264-265: What was the basis of deciding obtained PMF contributions of COA and HOA is "far exceeding expectations". Such claims must be backed by proper references. 22) Lines 288-289: Stable proportion % of COA across seasons does not imply it had stable contributions as volatility, dilution effects, and atmospheric chemistry, and interactions with other emissions all play a role in these stable proportions. I suggest that this sentence should be removed or edited to consider these factors that are likely affecting COA proportion. Attribution to stable contribution would likely involve the implementation of a volatility basis set approach.