The study by Voigt et al. focuses on the main drivers of temporal variations in the triple oxygen isotopic composition of leaf water and phytoliths. The authors performed a grassland plot experiment and measured key physiological, isotopically, and environmental drivers over the seasonal course and over the course of the diel cycle during two days of the experiment. The authors then compared the measured leaf water isotope values with predicted ones derived from Craig-Gordon based steady state and non-steady isotope models, as well as performed a sensitivity analysis to infer the main cause of isotopic variations in leaf water of grasses. As a novel part, they included high-resolution measurements of water vapour isotopes and measured leaf temperature data to improve the models. Besides, the authors found a relationship between daytime RH and 17O-excess of grass phytoliths.

The paper is overall nicely written, and the methods and results are well presented. Yet, after reading through the complete manuscript, I felt that the conclusions of the paper remain vague and that the discussion on the phytolith part, which is highlighted in the title of the paper, falls brief and short, while the leaf water isotope part, which is not highlighted in the title, is well discussed. Given the imbalance, I feel that some parts of the paper should be revised to match the title or that the title itself should be revised.

Major

Discussion 4.3: I think the start of the paragraph reads nice, but then it’s not clear how the RH signal in phytoliths 17O-excess in the current study is linked to previous observations and applications. Did the authors expect to find phytoliths' 17O-excess values to be close to daytime than to daily RH values? How well does 17O-excess perform compared to d18O, which was also measured and is well known to carry RH and VPD signals if derived from plant water and material? Would it also make sense to measure d2h and d-excess in phytoliths (maybe not possible)? Are there field studies with spatial or temporal resolution on phytoliths 17O-excess values, providing similar results? The discussion should also consider that the authors have only 3 values for phytoliths 17O-excess, which makes it difficult to set up a relationship with RH. The discussion on leaf transpiration and development should be combined with the one in the result part Line 406-412. I would also recommend incorporating the information in Figure 5.

Minor:

Introduction: I am wondering whether it would be worth highlighting that the 17O-excess approach in leaf water is still rather novel compared to d18O and d2H application In my opinion, that is one of the novel parts of the study, but it does not clearly drop out of the introduction.

Line 21 and 23: If “the” is used, its not clear to which specific model (e.g. the CraigGordon Model) it refers to.  

Line 24-25. I think these results are not clearly illustrated in the discussion part and figures.

Line 27: Yes, but it provides also new knowledge regarding the climate-sensitivity of leaf water stable isotope variations and their models, which I think is not well highlighted so far.

Line 55: Not all abbreviations of the model are explained here, e.g. Rs, aeq, adiff

Line 43: It might be worth adding the multiplication factor that changes “per mil” into “per meg”  for 17O-excess.

Line 62: Written like that it implies that some of the previously cited publications “neglected” the two-pool idea. I think this statement really depends on the plant species, which is quite diverse through all these studies. Maybe rather highlight, that the two-pool idea is important for grass species (but see Liu et al 2017, doi: 10.1111/nph.14549), where parts of the bulk leaf water pool in grasses do not experience evaporation and thus isotopic enrichment. It should also be highlighted that grasses have large isotopic leaf water gradients from the bottom to the top and that “bulk leaf water” is integrating this gradient. Further, the leaf water isotope gradients are integrated into the d18O of plant compounds (but see various papers from Sternberg, Helliker, and Lehmann on plant carbohydrates). In this regard, do we know whether phytolith formation/synthesis is equal along the grass blades? The link between water and phytolith formation during leaf development and growing season could also be an interesting discussion point, particularly if Figure 5 would be considered for the discussion.

Line 68-70: I strongly assume this statement refers to the “Peclet” effect. I would thus suggest introducing this “term” here so that discussion and introduction are better “linked” to each other.

Line 70: Rleaf is not defined yet, right?

Line 80: While I agree with the temporal separation, the example is not 100% clear to me. Assuming that sugars are produced under low RH, then they should carry this climatic information in their oxygen isotopic composition. It these sugars are later on used in the night/during rain, their isotopic composition formed under low RH should be transferred (at least partially) to the cellulose which is synthesized under high RH conditions. Maybe the moss example is a bit out of place. There are many studies on grass species and the isotopic composition of plant carbohydrates, which could be highlighted to make this point clearer (but see papers from groups of Schnyder, Helliker, and Lehmann).

Line 110: Maybe define “season” here, because it could be a growing season, but I think the authors mean “spring, summer, autumn”.

Line 120: What is the full name of the species? Is it a C3 or C4 plant? Mono or Dicot? This is important information because d18O in leaf water and cellulose and d2H in n-alkanes have been observed to depend on the physiological and biochemical background (but see different papers from Helliker & Ehleringer 2000 and 2002, and Gamarra et al, 2016, PCE).

Line 120: Why was the study performed within a woodland, which I assume is a forest? How do I need to imagine this plot? A grass plot surrounded by trees? I also assume that the grasses were grown on the topsoil in the woodland and that the topsoil was fertilized, is this correct? Does the grass plot include any replicates?

Line 125: The mean isotope value of the tap water could be provided here.

Line 145: Maybe add the exact period after “over the day”.

Line 158: Where the sampled leaf material fully developed and intact?

Line 163: What does “grass leaves” reflect? Only grass blades? How much material was harvested by end of the season?  

Table 1 and 2: I assume that providing the raw d17O results of water and phytoliths does not give any additional information and interpretation is only feasible for 17O-excess?

Table 2: How many replicates were measured for the phytolith isotopic composition and what does the standard deviation reflect? I assume the plots were not “repeated” and that the SD is “technical replicates”. Finally, how exactly was the SiO2 concentration determined? BTW, I think “rate” is the wrong word here because a change over time was not measured, right? Does the SiO2 concentration reflect the amount of phytoliths per gram biomass (i.e, SiO2 = phytoliths)? Do the long and short-cell phytoliths refer to the different “types” as stated in the method (Line 224)?

Line 207: Plant water refers to leaf water, or did I miss something, e.g. another plant tissue?

Line 208: Maybe consider/acknowledge that isotope fractionation can occur during CVD extraction, but it affects d2h more than the d18O (Chen et al, PNAS). To my knowledge, not much is known for d17O. Yet, also amount effects can change the isotopic composition of CVD-extracted water (Diao et al. 2022, HESS). Here I would simply double-check whether the extracted water content was high enough (e.g. difference of weight of exetainers before and after CVD extraction?).

Line 223: The extraction of the phytoliths should be explained in more detail. Its not clear how the phytoliths have been taken or extracted from the leaf material. How much biomass is needed in order to get a sufficient amount of phytoliths? I assume this is time-consuming and difficult work, which then also explains/justifies the low number of replicates in this study.   

Line 244: Which values were used for gs and gb and were these derived from own measurements or from the literature?

Line 259: So no leaf water content data is available for this study?

Line 260: Maybe slightly rephrase, and state that a best-fit model was used to set W.

Figure 2, Line 322: add “red” to “Dashed circle”. Throughout I would avoid using “yellow” in the figures, because its hardly visible.

Line 335-345: How exactly were d-excess and 17O-excess modeled with the Craig-Gordon model? I assume one needs the input of d2h, d17O, and d18O data, then run the model once for each isotope ratio and combine the data to gain excess estimates (e.g d18O and d2h for d-excess, and d17O and d18O for 17O-excess). Please clarify all this in the method part.

Figure 2, 3: What about d2h? The data is shown in the diurnal cycle, but not in the seasonal cycle.

358: I would suggest linking the results more clearly to each panel of Figure 4 a – x.

Figure 4: It appears that a change in temperature is stronger than a change in RH, but this sounds not right to me. From my own experience, the effect of temperature on d18O values in water and organics is rather secondary and induced via the equilibrium fractionation factor, which typically only varies around 2 per mil for d18O between 10 and 30°C or so. Moreover, why have these values been chosen for the sensitivity analysis, e.g. 5% RH and 2°C? Can we really compare them? Does it reflect an x-percent change per mean of each variable?

Line 406-412: This is an interpretation of the results and should be moved to discussion 4.3.

Figure 5, Line 413-421: This part comes a bit out of the blue and it is not yet nicely incorporated in the discussion on the phytoliths. I would also suggest adding more information on the forming water (FW) model and isotope fractionation factor (alpha and lambda values) in the result part. In which space does the alpha value vary and can they be given for each of the examples? If the purpose of figure 5 is to link 17O-excess and d18O of leaf water with those in the “phytoliths forming water”, this should be discussed in more detail.

Line 450: So it seems that the Peclet effect could play a role for the leaf water model in the case of high transpiration. Why not provide some values of the Peclet corrected CG-model for the discussion? I assume all the parameters are available to run this model, right?

Line 489ff: I agree with the paragraph. Maybe it should be more clearly considered that changes in the water vapour isotopic composition are more rapidly affecting the leaf water isotopic composition, within hours (but Lehmann et al. 2020, PCE for grasses), while the isotopic composition of precipitation has to go through the soil before its taken up by the plant and transported to the leaf.

Figure 6: The irrigation water point could be dropped for decreasing the x-axis scale and reduce the large space on the left side in the figure.

Line 512-516: Please clarify whether the cited papers and equations are derived from grasses or from other species too.  Please also clarify how many measurements of phytoliths 17O-excess were taken to generate the linear models, as well as the RH conditions of this study, to have some context. I assume that the RH conditions were similar to those in the current study.

Line 528-531: I think the discussion on the nighttime transpiration/stomatal conductance on tropical tree species goes a bit too far, as the current study focuses on grass species.

Line 561: I would be surprised if this is really the “first continuous record” given that the laser spectrometer measuring d2h, d18O, d17O are available for some years. How novel is the data?  Please clarify this.

572-574: Maybe I missed it, but how was the temperature non-sensitivity of phytoliths 17O-excess determined? Can the authors refer to a table or figure? Maybe move this to discussion point 4.3.

576: “RH proxy that is 17o-excess” reads a bit strange

576-579: That’s a good point, but this has not been discussed yet. See my comments on Figure 5.

Discussion 4.1: Is the CG steady-state model also working for d2H of leaf water in the current study? Maybe its worth adding a sentence on that.

Discussion 4.2: The results suggest that there is a difference in the water vapor influence on the temporal changes in d18O vs. 17O-excess, right (Figure 4)? If correct, maybe this is worth briefly discussing.  

The manuscript submitted to Biogeosciences presents results of a grass growth study that occurs in a natural setting and compared the results to controlled growth studies that occurred in plant growth chambers. The study monitored many environmental conditions that fluctuated during the day. The triple oxygen isotope compositions and δD were measured from the irrigation water, soil water, water vapor, leaf water, and the siliceous phytoliths (no δD values for the silica). This was compared to modelled water from the Craig and Gordon model modified for plant growth. Overall, this is a very comprehensive dataset that does an excellent job of describing how water vapor triple oxygen and δD compositions change during the day vs. night. A lot of the study focuses on changes in the water vapor without really connecting the impact on the water vapor and leaf water to the grass phytoliths.

Major/minor comments:

• Title: The title may want to be edited to better reflect the manuscript which really addresses changes in humidity and stable isotope compositions of leaf water between daytime and nighttime.
• Lines 418-421: Which differences are different by 1.7‰ and 10 per meg? In Fig. 5, the solid red and green lines have different differences even though they both represent a λ of 0.522. Also, how is ‘agreement’ defined. As written, this seems qualitative as someone could define agreement in a way such that neither λ 522 or 0.524 are agreement with the predicted water from the Craig and Gordon model.
• Figure 5: This graph is a little confusing on what is measured vs. modeled. The predicted leaf water is from the C-G model? If so, please add 'gray circle' to the figure to help the reader understand the figure better Is the formation water calculated from the measured grass leaf phytoliths? If so, why are they connected with a line? Passey and Ji (2019) modelled how water would change in different humidity scenarios. Would modelling how the irrigation/precipitation waters change with evaporation in different humidities be more useful than comparing to equilibrium precipitation of silica?
• Conclusions: Although it may have been missed, there is no conclusion that clearly defines how the phytoliths could be used to predict relative humidity. As the title reflects that phytoliths record daytime humidity, how far off was the humidity as recorded in the phytoliths vs measured? Perhaps adding a term that compares the difference between the Δ'¹⁷O value (or humidity) and the predicted Δ'¹⁷O value (or humidity) would better show to the reader the usefulness of this proxy.

Overall, the content of this study is of broad importance and fitting for Biogeosciences after minor revisions to better connect the triple oxygen isotope compositions of the phytoliths to relative humidity and the leaf water.

Passey B. H. and Ji H. (2019) Triple oxygen isotope signatures of evaporation in lake waters and carbonates: A case study from the western United States. Earth and Planetary Science Letters. 518, 1-12.