The authors report 2-year near-continuous measurements of volatile organic compounds at the Maïdo Observatory using a quadrupole proton transfer reaction mass spectrometer. Timeseries during the whole period and diel profiles for i) different seasons and ii) periods when the wind originated from the East or the West are presented for 6 VOCs (CH₃CN, C₅H₈, Iox, C₆H₆, C₈H₁₀, and DMS) as well as for meteorological parameters like temperature, total solar radiation, and direct radiation. Positive Matrix Factorization (PMF) is used that includes 11 VOCs and CO to separate their observed mixing ratios to different pollution sources including background, primary biogenic, secondary biogenic, and anthropogenic. Finally, the authors use FLEXPART-AROME back trajectory calculations to obtain more information on the source of the PMF factors. Overall, I find this study valuable. My main recommendation is that the authors provide more details on their modeling efforts that I found hard to follow without referring to their previous publications. Also, highlighted below are some suggestions to improve the discussion of the paper.

Specific comments

I would recommend that the authors avoid unnecessary abbreviations that are also not frequently used including FT, NWP, ACs, RTs, DISP, BS, AC, and more… I think using the original wording would improve a lot the readability of the paper since I found myself many times trying to find what the abbreviations meant.  

Line 108: 81 is the fragment of monoterpenes while 137 is the original monoterpenes m/z. E/N is high enough for the instrument to possibly mostly see the signal in 81 but that might be something to mention here.

Line 111-113: I am surprised to see 93 in there which I would expect to be most affected by the toluene signal which should not be long-lived and therefore not have high background. Any ideas why?

Line 117-118: You are still expecting MVK and MACR. Since this instrument cannot separate the contribution of the different compounds I would promote changing the naming to something like e.g., secondary oxidation products. This would be something to change throughout the paper.

Line 129: A graph in the SI to support this comparison would be nice to have.

Line 131: A sentence or two to briefly describe the indirect measurements would be informative here.

Line 135: Discussion of the overall measurement uncertainties based on the calibrations and humidity correction would be nice to have in a table. For Iox I would expect the calibration factor for MVK compared to ISOPOOH to be drastically different. It would be nice to add some sentences regarding the uncertainty for accurate measurements for this m/z.

Section 2.3.2: I find this section hard to read and challenging to follow the details of the approach used in this study without reading the cited publications. I feel this could be improved if the authors elaborated a bit more on the model they use and assume that readers haven’t read and don’t have to read in detail the mentioned publications. Introduction to what each model does in detail and the benefits of combining the two models would be great. Characteristic examples that I found hard to follow were line 173-175, line 183-185, line 195-197.

Line 193-194: Is this done assuming an emission rate? How is this derived?

Line 232: Are the authors capable of proving this is only secondary biogenic and not in general secondary? If not, I would recommend broadening to just secondary.

Section 3.1.1-3.1.5: In the section naming, I suggest providing the chemical formula and possible compound name as a more precise representation of what can be measured with a PTR-MS. Also, ratios of individual compounds, especially of anthropogenic nature, to CO might be informative in identifying pollution sources. Did the authors check these ratios and compare them to other emission sources? Comparison to other studies and inventories to improve the discussion of each section would be a great addition here.

Section 3.1.3: I would consider changing this name to secondary oxidation products and moving this section to 3.1.5 for the reader to have a better understanding of the primary emission trends first before discussing secondary sources. Also, this section is focusing the discussion on secondary biogenic sources alone and completely dismisses all possible sources of MVK and MACR. Although biogenic oxidation may contribute to the signal, anthropogenic sources in the island could impact the observed signal and trends. For example, I find it interesting that the trends of Iox are matching the trends of C8H10. Further discussion on this would clarify more the impact of different pollution sources on these trends.

Line 355-357: Do you mean that random hours of the day were chosen to reduce the dataset length for 3 different PMF inputs to reduce the length of the data to 1/3? If so I would recommend rephrasing especially since the reader cannot tell much by Figure 8. For Figure 8 I would also consider providing all timeseries together in one panel and zooming in to one specific plume (e.g., August-September-October) using a different subpanel to highlight the differences in the data chosen per run.

Line 387: Please elaborate a bit more on why these compounds are present in the BG factor. Is this something expected? I would think so but a discussion here would be great.

Section 3.2.2: Do the authors expect urban and industrial emissions from human activity from the island itself to play no role in the observed trends? I think a discussion on the local vs. long-range emissions detection would be valuable here. For example, when the back trajectory analysis indicates sources originating from the island's industrial or urban areas, does the anthropogenic PMF factor increase? Graphs that highlight that would be great. Reading through the paper I see that Figure 15 already does that to some extent. Wouldn’t this therefore further supports the influence of local sources?

Section 3.2.3: How confident are the authors that the anthropogenic and biogenic emissions are fully disentangled? Based on their diurnal patterns I would expect that the PMF has a hard time separating co-emitted sources that are both expected to increase midday. Could that be the reason for the increased contribution of C₈H₁₀ here?

Section 3.2.4: I think that this factor is not discussed enough and by the current discussion the naming should change to secondary oxidation rather than biogenic. Also, it would be great if the authors could discuss the contribution of other compounds in this factor including CH₃CN, C₅H₈, and MEK. Is this related to the challenges of the PMF separating different sources?

Section 3.3: I had a hard time following this section mostly because I don’t understand how this model works. It would be great if a more detailed discussion of the model was included. If limited in space this could also be thrown in the supplementary material of this paper. Correlations of this model to PMF do not look great and a discussion on the reasons why could be further investigated.

Section 3.3.1: How were the SRRs categorized in the model?

Line 526-530: The model and PMF have a weak correlation. Are these statements only based on Figure 15? What is the value of section 3.3.1 in the paper?

Figure 2, 5, 10: It would be a nice addition to generate monthly timeseries that would improve the discussions. Either the daily or monthly measurement figures could then be moved to the supplement. I think this will also improve the discussion of the figures since it is currently based on monthly trends.

Figure 2: It would be great if the seasons could be added to the figure at the top and also the abbreviations mentioned in the caption. Many readers will first look at the figures and it would be easier to follow if the abbreviations are repeated in the caption. Also, I would strongly recommend that the authors change the x-axis to months instead of days of the year since the discussion is anyways referring to months.

Technical comments:

Line 28: Since there are numerous publications on this matter I would recommend adding: “…(e.g., Jerrett et al., 2009).”

Line 171, 179, 180, etc.: In general there are a lot of abbreviations through the text. Many are used sparsely. I suggest avoiding abbreviations when not needed to improve the readability of the paper.

Line 242: Delete extra bracket.

Line 354: DISP and BS refer to one word each so I find no reason to generate more abbreviations here. Just use displacement and bootstrapping.

Line 367: delete double dot.

Line 377: delete “again”.

Line 381: change to “… it is more likely…”

Figure 3: I would suggest authors change the daylight background to an orange shaded area instead of lines since the figure is hard to look at with all these lines. Rather than that, I like this figure! Same comment for Figure 6.

General Comments:

The paper by Verreyken et al. presents a 2-year dataset of selected volatile organic compounds (VOCs) measured using a quadrupole proton transfer reaction mass spectrometer at the Maïdo observatory, which is a remote tropical high-altitude site in the southern hemisphere located on the Reunion island in the south-west Indian Ocean. Measurements of such VOCs over such long periods in such atmospheric environments impacted by oceanic air and local emissions from within the island are rare. Using a source apportionment model (PMF v5.0) and some source tracer VOCs, the main sources affecting chemical composition of the air were investigated both on diurnal and seasonal basis. This is an interesting and valuable study which would be a great addition to the literature. Below please find some points which if I addressed in my opinion can improve the submission further. I recommend the paper to be accepted for publication once the points mentioned below have been addressed.

Specific Comments:

• As is understandable, the air at the remote high altitude is very clean compared to continental sites in Africa and Asia. This often poses challenges to the instrument detection limits and while the authors have generally done a good job in trying to address the issue, more details would be helpful for readers. For example, statistics on out of total number of measurements in each season for the reported compounds, how often were values below the detection limit. This could be provided as a table in the supplement to both inform about QA/QC and the challenges in measuring composition of clean air.
• It is mentioned that calibration experiments were done frequently ..please provide the sensitivity in ncps/ppb of the compounds so that reader can see the drift of the 2 year deployment
• Why were seasonal and diel profiles of six key VOC species analyzed which does not include acetone and acetaldehyde? This is sub-optimal use of the dataset.
• What could be the reason for artefacts at m/z 93 which is normally well detected as toluene using a PTR-MS? On the other hand attribution of m/z47 to ethanol and m/z61 to formic acid and acetic acid is more challenging but no discussion of corrections and justification for these have been provided. What corrections were applied to correct for the HCHO back reaction with and its humidity dependence?
• The calibration gas mixture used (e.g. Apel Riemer) typically also contains trimethyl benzene detected at m/z 121. Was this compound completely absent in ambient air?
• Was a mass scan ever done at some point during the 2 year deployment?
• What was the residence time of air in the inlet and how often were filters changed during the 2 year long deployment?
• Radiation has been used intensively in the analyses so more detailed description of radiation measurements should be provided in the methods section.
• It is mentioned that sugarcane is a major crop cultivated on the island . Is it known whether sugarcane waste is burnt in post-harvest seasons? If so this would be interesting to compare with the BB profile and literature reported emission factors of the measured compounds (see studies from FIREX campaign published in ACP special issue.
• The analyses of isoprene oxidation chemistry could benefit a lot from comparison with other studies from forested sites. What were the NO levels in summer when isoprene and Isoprene oxidation products are highest? It would be very valuable to compare the daytime ratio of MVK+MACR+Isop peroxides to isoprene to that observed from forested sites in Europe, Africa, Asia and South America. Comparing yields reported from laboratory studies such as Liu et al. Liu, Y. J., Herdlinger-Blatt, I., McKinney, K. A., and Martin, S. T.: Production of methyl vinyl ketone and methacrolein via the hydroperoxyl pathway of isoprene oxidation, Atmos. Chem. Phys., 13, 5715–5730, https://doi.org/10.5194/acp-13-5715-2013, 2013 would be useful too.
• Is there information about the boundary layer height at day and night in different seasons from the site? This would make the discussion about role of emissions and dilution more quantitative. In BB air masses what is the CO/CO2 mixing ratio?
• Detailed analyses of temperature and radiation regimes associated with highest isoprene emission and formation of photochemically formed compounds (see for e.g. Mishra et al. Emission drivers and variability of ambient isoprene, formaldehyde and acetaldehyde in north-west India during monsoon season, Environmental Pollution, Vol. 267, 115538, 2020) would also provide further mechanistic insights.
• Back trajectory modelling description and analyses needs more clarity..can these be compared to local wind direction and wind rose too? Esp PMF factor profile wind roses would be helpful to supplement the other analyses.
• For PMF analyses would have been good to treat NOx as independent tracer with anthropogenic profile (Fig 11). Also for interpretation of anthropogenic source profile factor shown in Fig 11 would be really helpful to have the boundary layer height variation also in same plot even if from the met model in absence of measurements.
• The conclusion that marine sources do not show up as source factor sounds strange for an island so authors should clarify this is so for the burden of the specific VOCs

Minor points:

L6: Here and elsewhere should be sum of C8-aromatic compounds…also good to discuss which ones could be major contributors if speciation info from other studies is available.

L8: on air masses recorded is better replaced by air masses sampled here and elsewhere

L10: does not exhibit consistent diel variability is a not clear… What is consistent diel variability? Authors should clarify

L15: The term secondary biogenic sources is confusing

L19: should be mixing ratios  and not concentrations

L53: 2 years may not be adequate for inter-annual variability?

L90: authors mention PBL variability but would be good to add info on measured PBLs between day and night if known

L106: How was m61 corrected for fragmentation effects?

L108: It is stated that m/z137 was used for quantification and 81 was not, although at 136 Td one would expect more than 60% of MT signal to land at 81. What fragmentation ratio was used and is there any information about the speciation of MT, even from other studies perhaps?