Review of manuscript “The role of sublimation as a driver of climate signals in the water isotope content of surface snow: Laboratory and field experimental results” by Abigail Hughes and others.

This work is devoted to the investigation of the post-depositional changes of the snow isotopic composition due to the mass- and isotopic exchange between snow cover and the overlying atmospheric water vapor. The authors use the results of laboratory experiments, as well as of two types of filed experiments, to show that the isotopic composition of the uppermost few cm of snow may change at hourly time-scale due to these processes. The obtained results are quite interesting and important as another step towards a comprehensive transfer function between isotopic content of precipitation and that of an ice core.

I have a few minor comments and questions as listed below:

Figure 1 - photo of the experimental set-up would be relevant.

Lines 111-112: “Three Pico Technologies PT-104 Data Logger temperature sensors were placed in the box to record continuously; one 10 cm above the snow surface” – based on the figure 1, the upper sensor is placed about 20 cm above the snow surface.

Line 134: “wind speeds below 10-12 knots” – meters per second is a preferable dimension in meteorology.

Lines 136-137: “Sampling boxes were partially buried in the snow surface, and protected from direct overhead sunlight using a cloth covering” – why did not you bury the boxes completely in the snow (so that the level of snow in and around the box is the same), and why did not you use a white (a sunlight-reflecting) material for the box?

Line 139: “which would have otherwise led to melt of the snow not otherwise occurring” – this sounds a bit awkward to me, please consider rephrasing.

General comments:

This paper examined the role of sublimation as a driver of isotope climate signal preserved in ice cores. This study conducted two experiments in laboratory and field (Greenland) and a modelling for lab experiment. Each of the three major topics, experiment, modeling, and field experiments, are interesting and the data will be valuable. However, I think that each topic ends up with insufficient discussion/interpretation. In addition, because of the three topics, it is difficult to understand the main findings of this study. Please see my comments for details. Overall, substantial revision is needed for this MS.

Major comments

(1) L310 “This contradicts the traditional theory of sublimation, which states that sublimation occurs layer-by-layer and does not alter the snow isotopic composition….(Dansgaard et al., 1973)”.

>This is one of major arguments of this paper. Please explain why the traditional theory is wrong. During sublimation, the remaining ice is not mixed. Thus, the isotope ratio is not controlled by typical Rayleigh distillation. This basic concept sounds very reasonable and therefore many people believe it.

(2) The model used in this paper is based on a mass balance, in which input data is the observed vapor isotope ratio. Thus, it is not surprising that the model result agrees with the observation. I think different approach is needed to understand the physical process behind this experiment. For example, Figure 4 shows significant decrease of relative humidity. This should affect kinetic fractionation during sublimation. Thus, the net isotope fractionation factor had changed during the experiment. How much this affects your observation? The impact of changing fractionation factor may be evaluated (i.e., Craig-Gordon model). This is important because the snow/vapor isotope composition changed because of changing fractionation factors.

(3) Please state clearly what is new compared to previous studies in Introduction. Sampling depths apper to be finer for this study? What new for the laboratory experiment? 

(4) Maybe the originality of this study is that the combination of the tree topics (lab, field, and model). If so, what did you “learn” from the combination? As the authors themselves noted, the laboratory experiment is difficult to compare the field result because of extreme condition of the lab experiment.

(5) Reproducibility is the crucial for such experiment. Thus, please describe the setting of the experiment strictly (please refer to specific comments).

(6) I do not understand the exact purpose of the FB experiment because the condition (the box and cloth cover) is too far from nature. Maybe this is designed as an intermediate between laboratory and field?

(7) Please add the raw data and modelling code you used in supplementary material so that the readers can reproduce the figures.

Specific comments

L6 “how vapor-snow exchange and sublimation processes….”

>I think that the physical mechanism behind the vapor-snow exchange process is sublimation and deposition. Why did you say “the vapor-snow exchange and sublimation”? In fact, the two terms appeared several times throughout the paper.

L.18 “our results demonstrate that post-depositional processes such as sublimation play a role…”

>Please clarify the difference from the previous findings.

L92 “Plywood box” > Add thickness.

L.93 “PID-controlled heater” > Add details (e.g., what kind of heater (cable or panel)? Wattage?). Please also illustrate the location of the heater in Fig.1a.

L.93 “with a generator” > Add details (product name, manufacture name etc.).

L94 “mass flow controller” > Add details (product name, manufacture name etc.). Please also illustrate this in Fig. 1a.

L95. “continuously-running fans” > Add details (wind speed, product name, manufacture name etc.). How many fans exactly did you install?

L.96 “small boxes” > Please add material used for this box. Please also add thickness of the boxes.

L.115 “experiments used snow”　> Please add details (type of snow, density etc.)

L.137 “partially buried” > How deep exactly?

L.137 “a cloth covering” > Please add details (material used, thickness, color). I do not understand why you used the cloth. Maybe the melting occurred because the boxes only “partially” buried?

L.161 “KNF pump” > please add details (product name).

L.304 “..a strong decrease in the d-excess. This indicates that the HD16O water isotopes are preferentially removed compared to H218O”

> A decrease in the d-excess does not necessarily indicate that the HD16O is preferentially removed compared to H218O (i.e., HD16O are almost always preferentially removed compared to H218O because of larger isotope effect). The change of d-excess depends on changes in dD and d18O relative to slope of 8.

L.304 “this indicates that the HD16O water isotopes are preferentially removed compared to H218O” > Precisely, the HD16O is an isotopologue of water. Furthermore, there is NO water isotopes, only oxygen (or hydrogen) isotope exists. But I know that many people used this term, “water isotope”. Thus, it is not necessarily to revise “water isotope” throughout this manuscript. But this sentence is a bit strange.

L355 “A site such as Renland (east-central Greenland), which receives 45 cm per year …will be less affected…. ”

> The SE-Dome core is a more suitable example of a high-accumulation site, which receives 102 cm per year. Furthermore, the ice-core d18O record is remarkably similar to the isotope-GCM outputs, suggesting negligible influence of post-depositional effect (Furukawa et al., 2017).

Reference: Furukawa, R., Uemura, R., Fujita, K., Sjolte, J., Yoshimura, K., Matoba, S., & Iizuka Y. (2017). Seasonal-scale dating of a shallow ice core from Greenland using oxygen isotope matching between data and simulation. Journal of Geophysical Research: Atmospheres, 122, 10,873 – 10,887. https://doi.org/10.1002/2017JD026716

General Comments:

The new, impressive, and labor intensive field measurements from East Greenland on snow-water vapor exchange, as well as laboratory measurements, are a valuable dataset which provides much needed insights on the effects of sublimation on the isotopic content of surface snow, and constraints on the post-depositional processes affecting the evolution of the isotopic composition of surface snow and atmosphere water vapor.  The findings clearly show that sublimation does indeed impart an isotopic signal to the surface snow that propagates downward 1-2 cm over the period of 4-6 days during periods of clear skies. A simple box model is utilized to help understand the relatively isotopically enriched surface snow due to solid-vapor phase changes (i.e., sublimation) for d¹⁸O and the concomitant decrease in d_excess.  The box model helps to understand and explain the combined effects of surface sublimation and diffusion of the signal observed at depth in the homogenous lab-based snowpack. 

The comparison between field samples, field box samples, and laboratory experiments provides a good test of how large an isotopic effect sublimation has in a controlled environment (albeit extremely dry compared to field conditions).  It also helps to identify certain environmental parameters that may be causing unexpected changes in the isotopic composition of field samples, like synoptic whether variations altering atmospheric vapor d¹⁸O over hourly timescales or if the snow surface composition is driving the vapor d¹⁸O due to deeper snow layers influencing the surface snow by internal diffusion between grains.

The authors highlight one of the key findings: “A key finding from field experiments is that both sublimation and vapor deposition influence the surface snow on an hourly timescale; this is supported by laboratory experiments and model results, demonstrating that sublimation has the ability to influence the mean surface snow isotopic composition in the top 2-3 cm of the snowpack during precipitation-free periods. These changes are occurring faster than the average recurrence of precipitation events, and could produce substantial changes in the mean isotopic composition of the upper several cm of the snowpack over a long precipitation-free period. This suggests that effects from sublimation and vapor deposition may be superimposed on the precipitation signal, resulting in a snowpack record more indicative of atmospheric conditions and water vapor isotopic composition than condensation temperature (i.e. d¹⁸O ) or precipitation source region conditions (i.e. d-excess). The extent to which this occurs is dependent on the accumulation rate at the ice core site, as these processes primarily influence the top few cm of the snow column.” 

Although the authors do an excellent job documenting the hourly changes to the surface snowpack in the field sample experiments (FS1-4), the question remains: what is the net effect to the snow pack isotopic composition over a weekly or monthly time period that is precipitation free?  If sublimation enriches the snow surface and thus the overlying vapor isotopic composition, and then at night during negative LHF, equilibrium fractionation during vapor deposition should redeposit more δ¹⁸O negative water vapor onto the snow surface, and thus the net change is a minimal enrichment over weekly periods. For example, the initial FS2 0-0.5 cm values shown in Fig. 5 appear to have a mean value of ~−24 ‰ on July 7^th and by July 9^th the mean value is ~−22 ‰, smaller in magnitude than the δ¹⁸O increases observed in the FB 2-4 samples that were shaded/covered.  However, by July 18^th the mean value is −23.5 ‰ and then drops down to −32 ‰ (“likely due to a precipitation event preceding FS4 which may have deposited surface snow with anomalously low d¹⁸O”), but the point is that over the ~3 week period the sublimation signal that should be slowly increasing δ¹⁸O is overwhelmed by either fresh precipitation or more depleted δ¹⁸O atmospheric water vapor from elsewhere. The sublimation of the surface snow on day-to-day timescales appears to be less important to the overall seasonal isotopic composition of the surface snow in regions like Greenland with more frequent synoptic systems and advection of water vapor from marine sources (e.g. Baffin Bay, Arctic Ocean or North Atlantic). 

On the other hand the laboratory experiments are excellent demonstrators of intense sublimation over prolonged periods (using a continuous LHF equivalent to the max daily LHF in the field) and use a humidity level about 30-40x less than the atmospheric values found during the field experiments, which produces a very strong sublimation signal in the surface snow for L1-L8. The extreme sublimation rates make it a bit harder to draw comparisons to the field experiments, but provides an upper bound for the impact on isotopic enrichment of surface snow d¹⁸O and d_excess depletion during summer months. The smaller impact in the FS experiments shows that sublimation is still an important factor on diurnal timescales (daytime vs nighttime), but it remains unclear what the cumulative impact would be on the snowpack isotopic composition if at the end of the summer season a snow pit was sampled at 1cm increments would the sublimation changes be detectable or swamped by other post-depositional process (wind redeposition), synoptic scale atmospheric vapor imprints, or new precipitation events bringing in low d¹⁸O snowfall?  Clearly sublimation (and vapor deposition) is an important factor on the diurnal timescales during accumulation intermittency, but as the authors acknowledge: “Whether the magnitude of the mean isotope change due to sublimation and snow-vapor exchange outweighs the effects of snow redistribution, accumulation bias, and diffusion has yet to be determined.” 

The authors make a strong case that sublimation/vapor deposition changes do occur to the surface snow pack (~1-2cm for the FS snow surface samples) on sub-diurnal timescales. They argue that “These changes are occurring faster than the average recurrence of precipitation events, and [therefore] could produce substantial changes in the mean isotopic composition of the upper several cm of the snowpack over a long precipitation-free period.” The authors then speculate that effects from sublimation and vapor deposition MAY be superimposed on the precipitation signal, “resulting in a snowpack record more indicative of atmospheric conditions and water vapor isotopic composition than condensation temperature (i.e. d¹⁸O) or precipitation source region conditions (i.e. d-excess).”  Although this is a reasonable speculation their data is not sufficient to support such conclusions about the monthly or seasonal timescale impacts of the two-way exchange driven by sublimation/condensation. Thus, it is not appropriate to for them to assess the relevance of their results to the scale of the seasonal amplitude in the isotope signal for a firn core from the Renland Ice Cap. The changes observed in the FS field experiments (mean of ~1.8‰ for FS1-4 d¹⁸O range based on Table 2) occur on short (multi-day) timescales but in order to compare the cumulative impact of these processes to the seasonal amplitude, they would need to have sampled a nearby snow pit at the start of the field campaign in early July and again at the end of the month to determine the net effect, and ideally throughout the entire summer (apparently new data will be available from Wahl et al., in review). The authors do acknowledge that “In order to fully understand the implications of sublimation and snow-vapor isotope exchange on the ice core record, it is necessary to quantify the effects of these processes over the course of a full year” and while that is not within the scope of this paper they go on to make concluding statements that the results support their hypothesis “that rapid change occurs in a natural setting and propagates into the snowpack, substantially altering the initial precipitation isotope signal.” Although true on short timescales (sub-diurnal to diurnal) the results do not provide enough information to make definitive conclusions about the relative magnitude of sublimation/vapor deposition on longer timescales (i.e. years to decades) relevant to ice core interpretations. 

I agree strongly with the authors that further research is needed over seasonal and annual timescales and that their results “suggest that these variables contribute to a combined isotope signal, in which d¹⁸O and d-excess in ice core records likely incorporate individual precipitation events (i.e. condensation temperature and moisture source region conditions, respectively), surface redistribution (i.e. wind drift and erosion), and a post-depositional alteration signal reflecting atmospheric conditions at the ice core site.” Their suggestion that “Snow isotope models such as CROCUSiso (Touzeau et al., 2018), the Community Firn Model (Stevens et al., 2020), and isotope-enabled climate models” would therefore be improved through the incorporation of isotope fractionation during sublimation, snow-vapor isotope exchange, and snow metamorphosis.” is certainly justified by their findings from both the laboratory and field experiments and results from such modeling efforts may help to interpret the relative contributions of the aforementioned processes affecting post-depositional changes.

Based on the above assessment, I would recommend acceptance with minor revisions but with a primary focus on revising the Discussion and Conclusion sections regarding the broader application of their findings to seasonal and yearly timescales, speculation on the cumulative effect (monthly, seasonal, or yearly) of short-term sublimation/vapor deposition isotopic changes to surface snow, and their assessment of relevance to the interpretation of annual ice core records (e.g. Renland) is not yet supported by the four separate 2-4 day field data expermients.

Specific Comments:

Note: Please see the line by line comments in the commented pdf. Their repetition here is duplicative, although I have pulled out some of key comments below:                 

Line 187: If this is the case, are you suggesting the results from the field experiments are only affected at the snow surface by sublimation as well, and the rest of the signal at depth is diffusion (below 0.5cm)?

Line 284: It would be useful to run the snow isotope model with some of the field observations and input values and show how that compares to the model results from L5, which has a very high LHF that is continuous versus much lower mean LHF for FB or FS.

Line 288: Worth noting here that in the FS experiments depth propagation is only 1-2cm (max)

Line 299: See comment from Line 284, and run the isotope box model with more realistic field conditions so in the discussion one can comment on the degree of sublimation impact in the field.

Line 325: This one of the key questions, as the long term (weekly/monthly) result may not cause a significant deviation from the original snow-pack precipitation if the daytime sublimation and nighttime condensation of vapor balance each other out.  What is the NET change of the isotopic content over the entire month of July for the surface snow? Include in the Discussion.

Line 350: The authors have not demonstrated that this is the case, as their field snow surface experiments on only on the order of 2-4 days, and they do not provide data from a snowpit at the end of the ~3 week sampling period that can support this statement.  It may or may not be superimposed on the precipitation signal, and therefore its an assumption that the "resulting snowpack record would be more indicative of atmospheric conditions..."

Figure A15. Include the RMSE or 2 sigma stdev for the FS1-4 data, and error bar on each graph, so that readers can view the uncertainty around the fit.

Technical Corrections:

Figure 8 caption. The color appears to be brown in the image. Change “FS surface snow 0-0.5 cm values are shown in dark orange” to brown

Line 329: “In general, the box samples experience less decrease (should be increase) in d¹⁸O than associated FS samples due to minimized vapor deposition, and greater decrease in d-excess due to increased sublimation”

Figure A13. Figure label says 2.5-4.5 cm (yellow), need to be consistent with Figure caption that states “2.5-4cm below the surface”.