Review for manuscript entitled, “Multi-stress interaction effects on BVOC emission fingerprints from oak and beech: A cross-investigation using machine learning and positive matrix factorization” by Dey et al

The manuscript describes the effect of heat and elevated ozone exposure treatments on the BVOC emissions from one species of oak and one species of beech. The apply the treatments in series on the same set of individuals. The order of the stress treatment application is different for the two species and, as far as I can tell, they only have one set of experiments for each tree species. They use a couple different analytical tools to identify particular BVOCs that are associated with the stressor/species. The topic is timely and interesting since the effects of stress (including multiple interactive stressors) on BVOC emission rates and composition has been a major challenge in the research community for decades. The use of these new analytical tools is novel and could provide a roadmap for others in this field to pursue. However, there are some flaws in the experimental design and data visualization that need to be addressed. The flaws in the experimental design make it very challenging to use this data to draw many conclusions about stress effects on BVOC emission responses and I wonder if this would be more effectively framed as a proof-of-concept for the novel analytical methods employed – as a measurement techniques paper rather than a science paper.  Furthermore, the introduction is missing critical information on a topic highly relevant to the study that needs to be included. I recommend major revisions before this could be accepted for publication and am not convinced that this is the correct journal for the work given the flaws in design.

MAJOR COMMENTS

Introduction – some of the citations in the introduction are not appropriately referenced. For example, they cite the Penuelas & Llusia paper from 2004 to state there are 30,000 identified BVOCs. However, the cited paper is not an original research publication but is actually more of an “opinion” or “letter” about a couple other recently published papers. In the “opinion” piece, the authors do state there are 30,000 identified compounds, but they provide no citation for this statement. I would encourage the authors to reference a paper with actual scientific evidence that supports the statement. Another example is their citation of the Palm et al., 2018 paper to state that hydroxyl radical and ozone are the dominant atmospheric oxidants that react with BVOCs. The referenced paper is about an OFR study conducted in the Amazon where they oxidize BVOCs with OH and/or ozone to study SOA formation; the study is not addressing any science question about which oxidants (out of all atmospheric oxidants) are primarily responsible for reacting with BVOCs. There are papers that address that question and those would be more appropriate to cite in this context. One could even just cite well-known atmospheric chemistry textbooks to make this statement, such as the Pitts & Pitts textbook or Seinfeld & Pandis’ well-known reference book on atmospheric chemistry and physics. I will not go through each and every citation in the introduction, but there were a few that stood out to me as a red flag with this type of inappropriate referencing. Please double-check your citations.

Introduction – references to the ecological function of terpenes as imparting thermotolerance through membrane stability should be considered carefully. The foundation for this assertion is grounded in old seminar papers out of Tom Sharkey’s group, and he has recently published a paper now claiming that isoprene cannot possibly impart thermotolerance by stabilizing membranes because there just isn’t enough of it present in membranes to appreciably alter membrane fluidity. However, he has also written a recent review that still claims this is a function of isoprene. Ultimately, it sounds like this is a slightly more controversial function than some of the many others that have more ample support. Either way, I encourage you to cite the more recent works with updated information. The paper stating that isoprene cannot possibly impart membrane stability is #1 below and the more recent review is #2 below.

#1 - https://link.springer.com/article/10.1007/s10863-015-9625-9 

#2 - https://www.cell.com/trends/plant-science/abstract/S1360-1385(25)00132-3

Introduction - This intro is missing a summary of the work that has already been done on multiple stressor effects, though. I agree it is more rare than studies of stressors in isolation, but there are some papers out there that are interesting and informative. Since the entire rationale for this study is filling knowledge gaps related to the effect of interacting stressors on BVOC emission rates and composition, it is absolutely critical to include a summary of those findings. There has been work in the Kuopio group (Holopainen and Blande) as well as the Sharkey lab on this topic. Please add this information to your introduction. In my opinion, this is much more important to include than a summary of the biochemistry. If you are worried about space, you could easily cut down the biochemistry review to a couple sentences and focus more on this information related to interacting stressors which is MUCH more relevant and useful. This information is particularly useful for thinking about how competing stressors could alter plant physiology. Essentially, stressors could be additive, one could dominate the response over the other, or there could be some non-linear synergistic effect of both stressors combined. Establishing this sort of framework for the study should be priority #1. Here is a non-comprehensive list of some papers that could be included in this summary and these papers likely have other references within to follow up with additional papers on the topic.

#1 - heat and CO2 combined: https://www.frontiersin.org/journals/forests-and-global-change/articles/10.3389/ffgc.2019.00008/full

#2 - plant responses to multiple air pollutants: https://onlinelibrary.wiley.com/doi/abs/10.1111/plb.12953

#3 - ozone + herbivory: https://royalsocietypublishing.org/doi/full/10.1098/rspb.2022.0963

Figure 1 – First, I really like this figure for describing the experimental design. Very clear and well done. However, I do have some questions about the experimental design. The figure indicates that beech were exposed to ozone stress, then the combined stress, and then just the heat stress while oak had a different sequence (ozone then heat and then combined). It also appears that these stress exposures were conducted just once for each of the tree species (again, with a different sequence of stressors for the different species). Finally, the series of stress treatments appears to have been imposed on the same set of individuals, correct? If all of this is correct, can you comment on the decision to implement these stressors in sequence rather than using new individuals for each of the different treatments? It is a non-traditional approach to this type of study and makes it difficult to interpret results since prior stress exposure can have lingering effects on subsequent stress responses. Furthermore, the decision to change the order of the stressors for the different plant species is incredibly problematic for making comparisons about stress response between the species. I think this design needs some additional context because it is not only studying the impact of combined stressors. It is studying the impact of repeated exposures to different stressors which is a different question than was suggested in the introduction. Again, the authors should conduct a thorough literature review on the topic of repeated stress exposures to provide adequate context for how this would be expected to influence plant responses. It is unclear to me how you would even tease apart any differences in the responses observed between the two different species. Any effect could be related to differences in species-specific responses OR it could be related to the effect of the stress sequence. You won’t know!

Line 149 – authors state that they selected the six “healthiest” individuals from each set of plants they had. Please elaborate on how this was determined. “healthiest” as determined by what metric?

Line 155 – some papers suggest that ozone exposure stress responses can recover quite quickly, even within 24 hours. This is another example where having conducted a more thorough literature review on the topic could have informed an improved experimental design. I understand the issue about not wanting to include reaction products in the measurement of emissions, but this is often why it is necessary to have two separate chambers – one chamber for the ozone exposure and a separate one for the BVOC measurement. At the very least, you should discuss the possibility that the ozone response could be missed with this design if the plants recovered quickly.

Line 176 – More recent papers on BVOC emissions typically refer to the measurement as an emission rate measurement rather than a “flux.” I think this vocabulary has changed a bit over the decades with the massive expansion of the flux research community. The term, flux, is now often associated with eddy covariance measurements at much larger scales than the leaf, branch, or even a few individuals (as you have here). I understand this is just semantics, but it is something to think about when communicating your science because the term, flux, could be confusing in this context to some of your intended audience.

Line 218-219 – what are these “known contaminant” compounds that you eliminated from analysis?

Line 250 – this line raised a red flag...it reads that the decision was somewhat subjective if something was retained. Perhaps a little more detail on how something could be determined to be "biologically" justified would improve the rationale.

Figure 2 - I am struggling with this figure. I think it is trying to do too much. There are a couple different questions one could ask that I think would benefit from separate figures. The first is related to a comparison of the pre-stress emission rates of different types of compound classes and/or the effect of the stressor on the actual emission rate values of different BVOCs. The second question is related to comparisons in the diel profile. The latter seems to be the focus of the text discussion about this figure, but it is not the clearest way to visualize this. One could normalize the emission to maximum and then plot multiple curves on the same graph to more effectively make these comparisons. Other criticisms include - the shaded region is very difficult to see and the shaded regions for "std deviation" are also often very difficult to see. I think Figure 3 actually works great for addressing question 1 in my comment. So I also think Figure 2 could be normalized to "max emission" set to a value of 1 to compare the diel trends between the different tree species and treatments. It would be more effective as a visualization to address that particular question. And I would probably also present this figure 3 first followed by the diel trend info as a more logical flow of information.

Table 1 - This would be a more effective visualization as a figure. Alternatively, some shading of the boxes could help as well - one color for increase and another color for decrease. Otherwise, it is difficult to pull out clear patterns from a table of numbers.

Line 389 – Yes, conducting the ozone exposure in the dark when stomatal uptake is substantially reduced does seem to be a major flaw in experimental design. Also, ozone levels tend to be highest in the afternoon so it is a mismatch between a real-world context of what the plants would experience and how the treatment was applied during the experiment.

Line 401-402 – you do not have the observations to support this statement. Just because the heat stress had the largest effect on both plant species does not mean the plant would have responded to the combined stress differently if the stressors were imposed in a different order. You have ONE set of experiments for each of the plant species and they had a different sequence of stress exposures. You have no idea if the response to heat or combined stress would have been different if the sequence was altered.

Line 410 – I don’t understand this unit, “per square millisecond?” I think you need to add a space between the m and the s.

Figure 4 – Is this mass-based or mole-based? I think it is molar based on Figure 5, but it should be clarified in the Figure 4 caption as well.

Section 3.4 - This is a relatively new tool for this particular field so I think you should provide a bit more explanation about what these different metrics mean and how the reader should be interpreting these figures. Very little text is devoted to explaining how to interpret Figure 6, for example. Elaborate please.

Line 524-527 - It would be useful to know how much of the total signal these identified "Stress markers" contributed. Are they still tiny components of the overall composition, which would make them difficult to use as an ambient measurement marker of plant stress? Or did they contribute to a substantial portion of the total signal following the stress exposure?

Line 529 - why are you referencing the Figure 6 Upset plot after discussing Figure 7? This was ordered in a confusing way. Perhaps this would be resolved with some additional discussion about Figure 6 before moving on to Figure 7.

Figures 8-9: These figures are really messy and need to be cleaned up. I think you could move the mass spectra of the individual factors to the supplement and focus on the more meaningful info for the context of this study, which is the compounds that comprise the different factors (shown in C). I don't understand what the inset is showing in C, though. Please make that clearer.

MINOR COMMENTS

Line 159: Do these “heat stress” values represent something meaningful? Values reached during some typical heatwave or something? Please provide additional context.

Line 324 – again, please cite more recent Sharkey group papers. The thinking on this is evolving.

Table 2 - the symbol in the table doesn't look like the same symbol in the caption for PMF. It is also interesting that PMF never identified a fingerprint compound that ML did not, but ML did identify some ozone stress fingerprints that were not picked up by PMF. Perhaps discuss this a bit more in the text.

This is a well written and thorough manuscript about the impacts of stressors on two different tree species. There is some imbalance in that certain parts of the manuscript are described in great detail, e.g. the methods, while other parts are left with quite limited discussion, e.g. the impacts of the limited experimental design. I do recommend that this paper be published in ACP, in part also because it provides a well-described observational dataset (with several analytical techniques applied) that can be very useful for future studies. However, it would still benefit from various improvements, as listed below.

General comments

The experimental design: It is noted that the experiments were conducted in SAPHIR and PLUS, but only data from PLUS are reported here. Based on this, and the way the experiments were conducted, I would guess that the details of the experiment were designed with something else, or broader, in mind than what is presented in this manuscript. If so, it would be worth stating that more clearly. Because if the target was to study effects of combined stressors, running the experiments with the same trees, switching directly from one stressor to another, and running the stressors in different order for the two tree species does not make sense. In particular, the fact that the order was different is not even acknowledged in the text before late in the results section. It is stated also that all plants had wilted by the end of the combined stressor period, although for beech this was not even the last studied period. What does this mean for the interpretation of the data from the final period? The effects of the experimental design require more discussion, for example on how different the heat stress response might have been for beech if it had been the first stressor to be applied for beech rather than the last. In addition, the ozone stress was applied only at night, but only very little (and very late) is there discussion about how this might impact the emissions during the day. How fast is the recovery? The authors can of course not change the experiments anymore, but they can more adequately discuss the limitations of the experiments.

Section 3.4.1 and Figure 6: Perhaps someone more familiar with this type of approach can read out more from this part, but for me it was not clear what, if any, useful findings came out of this analysis. While I could not necessarily follow everything written in section 3.4.2 either, at least there it was clear that this analysis provided insight on fingerprint compounds from different stresses. Still, Fig. 6 is hard to understand, and it remains unclear it if provides any added value in this manuscript.

Figures 8-9: The other reviewer suggested to move the spectra (b panel) to SI, but a time series and mass spectrum is typically what is shown (and expected) for PMF results. I would rather move panel c to the SI, as that would also give more space to panel b, where the largest contributors are anyway seen. In addition, the whole manuscript is based on mass spectra from the PTR-ToF, and therefore it is appropriate to show at least some mass spectra also in the main text. But for me, panel c does not provide critical insights that would necessitate making the figure this compact.

Specific comments

Line 44: Under what circumstances would isoprene lifetime be only seconds? Minutes, or even hours, would seems more reasonable.

Lines 99-100: Some word presumably missing.

Line 528: Some word missing here too? At least the sentence seems strange to me.

Line 541: “Factor 1” would be more consistent.

Table 2: The markers in the caption and in the table are not identical.