Cui et al. conducted a campaign of comprehensive field observations at an urban site in Beijing. The composition, sources, and secondary transformation potential of VOCs were also identified. Overall, the study is very interesting and shows some new findings. However, the manuscript still suffers from many flaws especially the language expression. Furthermore, section 3.2.1 is not well-organized and needs major revisions. The detailed comments are as follows:

1. Why not perform the hourly measurement of VOCs? To the best of my knowledge, the daily resolution for VOCs measurement is too coarse. Especially, the PMF model needs substantial observation data, which ensures the model’s reliability.
2. The authors need to add the detailed QA/QC of VOCs and other criteria pollutants in section 2.1. The information is very important otherwise the study might be meaningless.
3. Section 2.3: Why do you use RF model rather than other decision tree model or chemical transport model (CTM)? The predictive performance of RF model might be worse than GBDT and XGBoost. Meanwhile, CTM is a process-based model, which could clearly explain the contribution of many VOC species to O₃ Moreover, the hyperparameter of RF model should be added.
4. Section 2.4: The BS, DISP, and BS-DISP tests should be also added.
5. Section 2.5: I think the PSCF analysis is not important in this study and could be removed.
6. Section 3.2.1: Why not distinguish the meteorological and emission contributions to each VOC species?
7. Section 3.3: The source identification method of each source based on VOC fingerprint should be added in this part. I think this part is too rough and should be rewritten.
8. Conclusion is too long and should be shorten and reorganized.
9. Data availability: I suggest the authors open the VOC dataset and it is very valuable to some researchers engaged in air quality modelling.
10. The English throughout the manuscript should be significantly revised.

The research focused on ambient volatile organic compounds pollution at urban Beijing, and analyzed their characteristics, sources, and control effects. This study is of interest to the atmospheric scientists and suitable for the ACP. The observation data were detailed presented, the chemical composition and emission sources were analyzed aiming at different months and different O3- or PM2.5- pollution days, and the VOCs decline was found through comparing with reference results to support the control effects. However, I have a few concerns that should be addressed before the acceptance of the manuscript.

Major comments:

1. In introduction section, the air pollution status has greatly changed in past several years in Beijing, due to the strict control measures implemented. However, the corresponding introductions were outdated and can’t present the current pollution characteristics. For example, line42 about SOA fraction in PM2.5, line 48 about SOA contribution to haze pollution, line 56 about the contribution of biogenic and anthropogenic sources, and so on. The recent references and their conclusions should be referred to.
2. Methodology section, VOCs detection system should be GC-MS, but not GC (as mentioned in lines 95-96), for Agilent 5975 uses mass spectrometry detector. If the detector only included MSD but not included FID, C2 hydrocarbons would not be detected but they widely exist in atmosphere. This point should be illustrated. In addition, the efficiency of this analyze system for aldehydes should be well discussed. Because various monitoring standards don’t explicitly recommend the “canister sampling-GC/MS analyzer” to detect aldehydes.
3. This study used the fact that O3 or PM2.5 pollution event happening to define high-O3 months (Apr, May, Jun, Jul and Sep) and high-PM2.5 months (April, May, Oct, Nov, Dec, Jan). It seems weird. For example, although O3 event never happened in Aug, but ozone level was also relatively higher in Aug than in Apr and Sep. So Aug should be considered as the high-O3 month, comparing with Apr and Sep. And then, in the results of PMF, the source apportionment in low-O3 months (Oct, Nov, Dec, Jan) was different with that in high-PM2.5 months (April, May, Oct, Nov, Dec, Jan), but similar to that in low-PM2.5 months (Jun, Jul and Aug). This conclusion was unreasonable to a certain extent.
4. When using PSCF to explore the spatial potential sources of VOCs in urban Beijing, 24h was considered for all species. However, the lifetimes of various VOCs species were greatly different, several hours for alkenes, but several days for some alkanes and halocarbons. I suggest various groups of VOCs should be individually considered, to give the lifetime hours in backward trajectories.

Minor comments:

Abstract: “O3/PM2.5” frequently appeared but without an explicit definition. It is hard to understand the “high and low-O3/PM2.5 months”, “O3/PM2.5 polluted days”, and “high O3/PM2.5 levels”, etc.

Lines 32-34: “The positive matrix factorization (PSCF) analysis showed that O3 and PM2.5 pollution was mainly affected by local emissions. ” PSCF was conducted for VOCs, but not for ozone and PM2.5. No evidence to support this conclusion.

Line 47: VOCs chemistry in ozone formation involves gas-phase reaction, but not multiphase reaction.

Line 104: the “coefficient” should be coefficients; “was” should be “were”

Line 111-112: air pressure appeared twice.

More detailed model performance verifications (RF) are necessary, although R² has provided in Fig. S2.

Line 191-193: I cannot figure out the sentence, suggesting checking out syntax rules.

Line 234: Fig. S3 was mentioned, however, there is not Fig. S3 in the supplement of this passage.

line 243-244: “Alkenes, aromatics and OVOCs were the three contributing chemical groups to O3 formation”, should be “the three biggest contributors”.