The present study proposes an analysis of observations of sensible and turbulent heat fluxes during a few months in 2019 within the Greenland Ice Sheet, from three different eddy covariance measurement systems. This is a unique dataset, extremely difficult to obtain, and allows us to explore the importance of latent heat flux on the local mass balance. After presenting an intercomparison between the three measurement systems (which ultimately agree very well), the data are used to train a 'bulk transfer' type relationship in winter and in other years, enabling comparison with turbulent fluxes simulated by two climate models. The multiple challenges of measuring and simulating these fluxes in this type of environment are then discussed.

I really enjoyed reading this paper. The plots are clean and well put together. The text is well written and the sequence of ideas is easy to follow. With the modifications I recommend below, I believe this article has a rightful place in a journal of the caliber of The Cryosphere.

GENERAL COMMENTS

1. LHF and SHF in winter
   
   I am puzzled by the 'observed' winter fluxes. The approach used to estimate these fluxes is to apply bulk transfer from a standard weather station based on z_0m, z_0t and z_0q values calibrated in summer. It is clear to me that these values do not hold in winter, partly because the surface does not have the same roughness. More troubling is that the measured temperature difference between the surface and the air about 2.6 m above is very small in winter (at most 1°C in Jan. 2019!!). This seems impossible. Since q_sfc depends on T_sfc, both turbulent fluxes are affected.
   
   
   
   I know that most of these issues are raised by authors. I would like to see more options explored to improve these winter results:
   
   - Have you tried estimating winter z0m from standard weather station data or other means?
   
   - Have you explored surface emissivity values other than 0.97? Satellite data for T_sfc?
   
   - What about radiative flux divergence? See for instance: https://journals.ametsoc.org/view/journals/apme/46/9/jam2542.1.xml
2. Wind
   
   While temperature and humidity gradients play a key role in turbulent fluxes, wind is also key. This is barely mentioned in the article. What about katabatic winds at the measurement site? How do they affect the results? This should be covered in the introduction, results and discussion.

SPECIFIC COMMENTS

Abstract: Mention the exact dates of the measurement period.

L22: I know this comes up later, but the introduction should distinguish between surface and blowing snow sublimation.

L33: According to your Fig. 5h, a third of the day in summer experiences unstable conditions. I think we tend to assume that as soon as there is snow or ice, the atmosphere is necessarily stable, which is not the case.

L43: Discuss the strengths and weaknesses of open-path and closed-path gas analyzers.

L54: you need to elaborate on the limitations of the MOST approach - what exactly is at stake during stable atmospheric conditions?

L74-75: I do not understand how this ties in with the main objective. Is it possible to reword the objective stated at the beginning of this paragraph to include climate models?

Figure 1: If possible, increase the resolution of this figure. A view of the footprint of the EC sensors would be useful. Also, there seems to be a fine-wire thermocouple on the IRGASON, but not on the other devices. Is it then the sonic temperature that is used to calculate the sensible heat flux? Please add these clarifications to the text.

L87: Define 'clean-snow' area.

L89-90: Why was the period from 28 May to 31 July 2019 used? Please explain.

L93: At this stage of the paper, it is not clear why the period from 2016 to 2019 is mentioned for the model comparison.

L95: Is it possible to better describe the observed wind regime, beyond the typical values of wind speeds and directions? Are there any katabatic winds?

Section 2.2 and following: For the whole document, always present the three devices in the same order, to make it easier to follow.

L100: How and how often were these devices calibrated? Same question for the radiometer used to calculate the (very crucial) surface temperature.

L111-113: This should have been mentioned earlier.

L125: Mention that this is saturation with respect to ice. What is the validity of this hypothesis?

L147: Presentation of Andreas' (1987) formulation would be useful.

L150-151: I do not understand the changes resulting from the 'physics cycle CY47R.1' update. Is it possible to explain the highlights?

Equation 1a: I suggest removing the minus sign and writing Ts - T.

Equation 1b: I suggest removing the minus sign and writing qs - q.

Equation 4b: why not use the specific humidity q instead of a?

L275-280: again - what was the calibration strategy (zero and span) for this instrument?

Figure 2 and equivalent: add a white box under the performance metric values and add the units on the RMSE.

L298: How have the values of z0m, z0q and z0t been optimized?

L299: 5.7e-7

This manuscript compares the turbulent fluxes estimated by 3 Eddy-Covariance instruments over one summer at the EastGRIP site in Greenland, with a view of estimating the quality and uncertainties of such estimations in polar context. These data are also used in conjonction with ancillary data from AWS to derive year-round estimates of turbulent fluxes and compare them to outputs from high-resolution regional climate models.

This contribution is of high interest as, as underlined by the authors, estimations of turbulent fluxes are rare in polar environments, while modelling uncertainties are high. The manuscript is very neatly written and illustrated.

Still I think that some important points need to be addressed before publication :

Main comments :

My main concern is directed to the hypothesis of similar roughness lengths for winter and summer (e.g. L308-310). This hypothesis is not justified in the manuscript. Actually, it is questionned by the authors themselves. The discussion around this hypothesis, evidences in favor of or agaist it, and/or ways to circumvent it, is very short in the manuscript. It needs to be enhanced as an important part of the results relies on it (esp. the comparison to the RCM and the assessment of the sign of the sublimation/deposition contribution to the surface mass balance in Greenland). Also, the sensitivity of the results to this very strong hypothesis needs to be shown.

Secondly, it seems that blowing snow sublimation can greatly modify the magnitude of the total sublimation/deposition at the annual scale, and even change the bulk contribution to the mass balance from net gain (deposition) to net loss. The authors note that the estimation of turbulent fluxes with the bulk method used in winter, is not valid during blowing snow events (L326). Furthermore, the models diverge in their accounting (or not) of this process. I would strongly suggest to exclude the periods with blowing snow events from the direct model-to-data comparisons in section 4.3 for a fairer assessment of model performances. If no better, RACMO2.3 outputs could possibly be used for a first diagnostic of the main blowing snow events.

Finally, I think the discussion regarding the model and observation comparison, could benefit from an enhanced context regarding the known model biases over polar and/or snow-covered environments. L 432  incriminates « the atmospheric processes driving the surface gradients» in the models, but surface processes are also likely at stake, esp. in the presence of snow. Fettweis et al. 2017 highlight some of them for MAR over Greenland, and they should be mentioned. With respect to that, Lapo et al., 2015 noted the important role of stability corrections in amplifying model cold biases over snow surfaces, esp. in conjuction with a negative bias in incoming LW radiations (which seems to be affecting e.g. the MAR model ; Fettweis et al., 2017). Could this play a role here too ? This possible source of model bias should be included in the discussion (Rudisill et al., 2024 could provide an intesting view on other bias sources, though more oriented towards mountain regions).

Minor comments :

L 8-10 : this is a lot of detailed numbers for an abstract, maybe choose 2 metrics out of the 4 presently described

L 87 : what is a clean-snow area ? Please define.

L 91 : the maximum on wind speed can highly differ depending on time-averaging procedure. Is this a maximum of 10-min data, hourly data, or instantaneous maximum ?

L 151 : the major changes attached to the upgrade to CY47R.1 should be described in a nutshell

L202-203 : site-specific roughness lengths would likely depend on snow conditions and hence vary at least sesonally ? (see 1st main comment)

L289 : « Estimates of the LHF and SHF based on the bulk method in the PROMICE data product are overestimated (Fig. 5) ». Yet the biases mentioned on Fig5 are mostly negative, probably due to different sign convention. It would be clearer if it could be changed

Fig7a : the bulk estimate of sensible heat flux shows a bi-modal seasonal cycle, much different from all models, but this seems to me this is barely analysed. Can you comment on this ?

L 365-366 : «However, systematic biases with the EC method, due to boundary layer characteristics in polar conditions cannot be ruled out. » This should be developped.

L397 : erroneous reference : Fig 8 a is probably meant ?

L405 : « Features in the synoptic-scale variability of the near-surface temperature gradient are observable in both AWS datasets. » It seems that these features are much more attenuated at the PROMICE station, esp for the second half of the period shown Fig 8, which may be an argument for frost on the LWup sensor ?

L 409 : words are likely missing in this sentence

L 415 : «  the air temperature is lower than the snow surface (e.g. 1st until the 10th of January 2019) » It seems to me that air temperature is actually mostly warmer than snow surface over this period…, which makes the discussion confusing. The whole section 5.3 should be checked carefully as the processes at stake are not straightforward and it is not easy to get the point of the authors. Maybe the section could be renamed « Limits of the PROMICE data during winter », and a contradictory time-period when a no-frost assumption can clearly be made, could be provided for comparison in the graphs ?

Sect 5.4 / sect 5.3 : The whole analysis of the modelled vs observed near-surface temperature gradients in Sect 5.4 is based on an extract of the time-series that is precisely questionned regarding the observation of surface temperature one section before. Would it be possible to carry out this analysis over another period where surface temperature data would be less questionable ? Statistics of the occurrence of such doubtfull PROMICE data would be valuable to assess the PROMICE winter data quality as a whole (Sect 5.3), and shed light on the results/interpretations. A feedback to the data exclusion mentioned in Sect 2.3 would be usefull for the reader’s understanding.

References :

Fettweis, X., Box, J. E., Agosta, C., Amory, C., Kittel, C., Lang, C., van As, D., Machguth, H., and Gallée, H.: Reconstructions of the 1900–2015 Greenland ice sheet surface mass balance using the regional climate MAR model, The Cryosphere, 11, 1015–1033, https://doi.org/10.5194/tc-11-1015-2017, 2017.

Lapo, K. E., L. M. Hinkelman, M. S. Raleigh, and J. D. Lundquist : Impact of errors in the downwelling irradiances on simulations of snow water equivalent, snow surface temperature, and the snow energy balance, Water Resour. Res., 51, 1649–1670, doi:10.1002/ 2014WR016259, 2015.

Rudisill, W., A. Rhoades, Z. Xu, and D. R. Feldman: Are Atmospheric Models Too Cold in the Mountains? The State of Science and Insights from the SAIL Field Campaign. Bull. Amer. Meteor. Soc., 105, E1237–E1264, https://doi.org/10.1175/BAMS-D-23-0082.1, 2024.