Li. et al. presents observations of volatile organic compounds (VOCs) collected from ~450 m above ground level in Guangzhou, China, from mid-August to early-November. The authoers were able to identify ~225 species using the Vocus PTR-MS. There were other measurements on the tower, including meteorology, CO2, O3, NO, NO2, NOx, and PM2.5; as well as, measurements of VOCs at a ground site ~5.7 km away from the tower and ceilometer for boundary layer height ~13.5 km from the tower. The authors look at both diurnal pattersn and positive matrix factorization to identify sources of the VOCs, which include daytime-mixed, visitor-related, vehicular+industry, regional transport, and VCP-dominated. Also, the authors used autocorrelation to further identify the sources (which I found to be very informative). The authors use the PMF results to show the source contributions of the VOCs observed at the tower. Finally, they look at the profiled measuremend of NOx, O3, and PM2.5 to look at the boundary layer dynamics throughout the measurement period and potential for residual layer vs nocturnal boundary layer chemistry.

As the authors noted, there are minimal measurements of VOCs at an elevated location (important for boundary layer dynamics and to look at regional background vs near-term sources), especially in China. That fact makes this paper of interest and importance for ACP. There are some aspects the authors could do to improve the paper, which I provide below, prior to publishing the paper in ACP.

1) Further clarification in methods, specifically in the sampling.

1a) The authors noted that their was ~5 m long and was "extended to the outside wall of the observation room." However, it is not clear if this line is sampling inside the tower or outside the tower. This aspect is important for clarification in other points presented below.

1b) Was the sampling line heated or tested for potential losses of lower volatility / oxygenated species? There are some high molecular weight and/or oxygenated species that may have experienced lost.

1c) Further description of the observation level would be beneficial to understand the results presented. E.g., are the windows open or is the observation level "leaky"? If the line is sticking outside the tower (1a, unclear), it is surprising to see the "visitor-contribution" (more on that below). Thus, description about the observation level would be beneficial.

1d) How was the data smoothed or extrapolated for the 3-D vertical profiles in Fig. 9? A description of that in Sect. 2.3 would be beneficial.

2) Something either in Sect. 3.1 or later that would be beneifical would be the OH-reactivity (and maybe even NO3-reactivity as some of this is later discussed in the context of residual layer chemistry) contribution of the compounds observed. As the authors note in line 606 - 609, though VCPs contributed a smaller amount of mixing ratio, some high reactivity compounds were in that class. As the combination of mixing ratio and reaction rate dictates the importance of the compound to urban chemistry, seeing how the classes weight out in reactitivity space would be of great benefit.

3) Currently, Fig. 2 is too busy to interpret well and follow along with the authors' argument. What would be more informative would be to highlight the compounds that show XX difference between ground and elevated platform measurements (e.g., 50%, factor 2, or somethin else). Though seeing "family" (which is different than PMF) is informative, know which compounds are different can be equally important.

4) Fig. 3 and potentially other diurnal plots. Sometimes it's difficult to discern the differences the authors mention in the diurnal profiles eithere between ground-level and tower or working vs non-working. I would strongly recommend either as a figure in the main paper or a supplemental figure showing the ratios of these compounds and if they are statistically different or not.

5) Source analysis of the VOCs. It is surprising that authors are seeing such a large source of ethanol and CO2 from visitors (comment 1c). It would be of use to better understand why this source is so large (is it due to experimental set up); whereas, e.g., they do not observe much VOCs from cooking when there are restaurants in the lower levels of the tower. Further, though the visitor profile is different from the VCP-dominated source, I would still recommend the authors be a little more cautious in this discussion. E.g., what the authors call the visitor-related compounds are also compounds that can generally be "VCP" in nature. Though it is a local source, it contributes/impacts the VCPs mixing ratio and chemistry.

6) I'm very suprised by the fractional contribution pie chart in Fig. 7 vs the time series fractional contribution. E.g., it appears that visitor-related is typically on order 5-10% with maybe the observations in October being greater than 20% while daytime-mixed is the largest contributor during most of the study. Not sure if it is howing the data is being weighed/average.

Minor

1) The authors use the word "significantly" or "significant" throughout the text. I would strongly recommend the authors use a different word when there is a difference unless they have conducted statistical analysis (e.g., Student's T-Test) to determine if there is significan difference.

2) Another statistics aspect to be careful in includes when the authors look at the r values and state that it is well correlated. E.g., an r value of 0.62 (line 336) indicates that CO2 can explain ~38% of the observed ethanol mixing ratios.

3) Fig. 6, the PMF factors having similar colors to the ground and tower observations makes it difficult to interpret. I would recommend different colors for the PMF results.

Li et al. report interesting time-resolved VOC measurements on the Canton Tower in Guangzhou, China using a QiTOF PTRMS. The Guangzhou region is impacted by air pollution and underrepresented in atmospheric observations which is why data from this region are a valuable opportunity to learn about the processes and regional sources. The identification of these sources from the data must be challenging due to chemical complexity. PMF was utilized to help with this task and five factors were obtained from the PMF source apportionment conducted on the 225 VOC ions which were assigned different mostly anthropogenic categories. The observations focus on abundant VOCs and correlations with inorganic pollutants which might be helpful to understand variability of pollutants, oxidation and sources in the atmosphere. The text reads generally clear and there are rich figures to illustrate observations and thoughtful analyses. However, I’d have some comments and suggestions before the paper is published, and in particular there seem to be some issues or inconsistencies in a few compounds and PMF interpretations which hopefully can easily be addressed.

Major

• Ethanol (e.g. Fig 1) is rather high as for the outdoor atmosphere and what I am most surprised with is that it is attributed to human emissions from tower visitors. If ethanol was emitted by human visitors on the tower why were siloxanes, not co-emitted? It is difficult to get convinced that such a large concentration of ethanol comes from human breath as is suggested in the SI Page 19. Was maybe a sanitizer dispenser in the vicinity to the sampling inlet? A picture or at least a schematic of the inlet location would be useful (as well as the info if the particle filter was used or not).
• The concentration data at least for ethanol are inconsistent. Ethanol concentration is shown in multiple figures with overlapping times in Fig 1 and 11. Ethanol concentration is different in those figures (e.g. 9/30, 9/29).
• The lack of dependence on wind direction (L397) is surprising and makes me wonder if the instrument may have been sampling from indoors, indoor plumes, or if there may have been a leak in the line or other factor which would explain high ethanol, and acetic acid concentrations. Wind roses or polar plots for VOCs would also be useful to help in clarifying these issues and help in source interpretations.
• The PMF analysis results seem surprising to me. It almost seems like the same compounds and factor profiles appear in every factor (Figure S5). Methanol and acetone appear in multiple factors as most abundant. Visitor factor is lacking siloxanes. Given it is the QiTOF with very high sensitivity, I’d expect the clearer factors including biogenic, oxygenated biogenic, and cooking could be obtained without merging. I wonder if optimizing uncertainties or dividing the dataset into shorter periods could further help in those source interpretations.

Minor

• I really like Table S1 with nicely tabulated masses and formulas identifications and sensitivities with clear note which ion was explicitly calibrated. I think it is exemplary for all PTRMS papers. I only have minor suggestions here: 69.03 should be furan (not fural), 51.99 monochloramine, 71.03 add MACR. Did you not see D5 fragment at 355.06? This is extremely surprising as is the unexpected D5 fragment at 299. I am again curious what the E/N ratio was.
• It is mentioned that ethyl acetate was one of the most abundant VOCs, but how did the authors exclude butyric acid or subtracted it from the signal?
• Nowhere in the text or SI PTRMS parameters are shown. It would be great to provide at least E/N, which is a common practice in a PTRMS papers, because it helps interpret the VOC data and estimate expected fragmentations.
• Fig. 3: I wonder why the traffic rush hours are not clearly seen on the plots presenting aromatics. The period was probably not during the lockdown if the visitors are allowed in the tower. Ethanol and monoterpenes coincide with lunch and dinner so might be cooking emissions(?).
• Fig. 5: Super interesting effect of Typhoon on VOC concentration reduction. I wonder about the dependence of the observed concentrations on the wind speed.
• I could not find in the main text or SI how exactly the PTRTOF data were processed, what software was used for mass scale alignment, peak fitting, as well as quality control, formula assignment, compound inference, total number of ions detected and steps taken for ion reduction or any abundance filters. This info would be greatly appreciated in the methods or SI.
• I am quite intrigued by the chromium ion reported in the SI Table. Did its signal show some ambient structure or was it coming from the ion source? There are many cool molecules in this table that could be useful to explore as source markers and complement PMF.

Technical

L 75 “were” should be “are”’

L109 “by far” should be “so far”