Overview

The manuscript investigates the snow redistribution processes over a well monitored glacier in Austria using a unique observational dataset and models. This topic is right now a major hotspot in snow research since snow drift and redistribution has received less attention due to the difficulty of measure and modeling drifting snow. I think that is therefore right to give attention to this topic and increase the efforts on research, as the authors does in this paper through both observations and modeling.

The dataset consists in Terrestrial Laser Scans and automatic weather stations, which has been previously exploded by the authors for study the snow dynamics on the site. The main innovation of this research is using a new snow drift module included in WRF to simulate at high resolution in complex terrain. This module adds to other initiatives in the same direction such as the Meso-NH/Crocus (Vionnet et al. 2014) and CRYOWRF (Sharma et al. 2023) without explicating adding a snowpack model as a surface model, and using the broadly used NOAH-MP. However, the fact that the description of the model is not the main focus of this manuscript and that the manuscript of the model is still in preparation, makes difficult to evaluate more deeply the manuscript.

In any case, the conjunction of this observational dataset and the modeling results show promising results for investigating the snow drift. However, I think that considering that the study is based in a unique case study, the analysis of the data can be more deeply considered. My current feeling is that the authors show disentangled the results from observations and from modeling, but they did a small effort on putting together the data and comparing them more quantitatively instead of the current qualitative comparison. Considering the potential of both the dataset and the numerical approach, I encourage the authors to do an extra effort to evaluate the model during this case study and give extra insights for this interesting case study.

Therefore, I think that this interesting manuscript can be published after addressing my previous concerns.

----------------------------------------------------------------------------------

PIECEMEAL

Title: Correct.

Text: The text is well and carefully written. Can be easily understood and the language is correct.

Figures: Figures are in general correct and labels and references are informative. As commented below, I think that modeling data should be compared with observations in Figure 6a, and that Figure 8 should compare the model and the observations from a more fair perspective (see comments on results)

Introduction: Introduction is well written. The background of the study is extensively exposed with significant citations. However, I have some comments on this section:

L33-34: Why only four processes? I think that snowpack is affected by a number of processes including, furthermore those stated by the authors, the creation of a hoar layer, the metamorphosis processes into the snowpack, or the redistribution by gravity on avalanches. Indeed, the processes depends on the scale. I think the authors are referring to the processes that modify the snowpack height only, but not in general.

L71-72: Why, according the authors, coupling a standalone snowpack model to an atmospheric model is a disadvantage? Indeed, this sentence refers to CRYOWRF, which included a specially developed scheme for drifting snow in WRF such as the authors uses in their research. However, instead of using the three-layer simplified model from Noah-MP as a land surface scheme, they use a more advanced snowpack model, that provides a better representation of the snow, impacting positively to the blowing snow representation. In my opinion, authors should reformulate this statement.

Methods: Methods are correct to describe how the case study was investigated. Authors describe both the observations and the model. While, as the authors stated, the comprehensive description of the module is subject to another paper that is not published yet, main equations used in the blowing snow module are stated. However, since that paper is not published yet, the authors should include some more details, such as the equations used for the sublimation process. In the same way, some basic notions of the numerical implementation of the blowing snow scheme in WRF should be mentioned to a reader may understand how the model computes the blowing snow before the modeling paper is released. This is useful to better understand and evaluate the following sections until the description paper is ready.

Results: Results is in my opinion the weakest part of the manuscript. In the current form it seems a disconnected sample of observations and the modelling part with very few connections in both analysis and text. Given the nice dataset that the authors have, and that results are based only in one single case study, I think that authors should extend further the analysis and exploit in more detail the datasets. Some suggestions are:

1. As the results are based on a single case study I miss a description of the synoptic situation and the characteristics of the front that offer more context to the reader.
2. Fig 6: Since Snow redistribution accounts for the snow height on the stations, it would be very interesting to add the observed differences of snow height to panel (a) and compare and show in parallel in panel (a) the results of the observations in snow high and discuss the possible differences. Add also in panel (a) the units.
3. The weakest point is that the comparison between simulations and observations is very qualitative (in text and Figure 8), and although the authors give some figures the only conclusion that might be extended from there is that the order of the magnitude of the redistribution in the model agree with observations. I think that a more quantitative approach can be done. First by extending the simulation until 9 February at 2 UTC to include also the 3th TSL acquisition in the simulation. I understand that, as stated in methods, the observations are rescaled to the model resolution (Is this the resolution showed in Figure 8?). If so, the authors can directly compare (1) if at the model resolution the distribution is similar (difference between observations and modeling) and (2) some metrics of the performance of the model such as BIAS, MAE or correlation of the data using the same domain (the one limited by observations). Later can be discussed the role of the compactation on the differences.

Some other minor comments are:

L243-245: The authors use “For scan 1” and “for scan 3”, that are snapshots and discuss the precipitation during, before or after. Consider rephrasing with these words instead using “for”.

L247: Figure 6 is shown before Figure 5.

Discussion and conclusions:

Discussion and conclusions are comprehensive with the results, and correctly details the benefits and the limitations of both the observations and the simulations. However, they are tailed by the qualitative approach in the results. As stated previously, I think that data can be squished to obtain extra insights on the model performance and on the snow redistribution processes during the case study.

A small extra point:

L341-343. Why is not this stated in methodology?

Extra comment:  In author contributions the authors state a contribution of “CS”, which is not listed as a coauthor and I assume that is Christina Schmid after reading the manuscript. However, she is not listed as author of the manuscript. I don’t know if they forget or not and I cannot evaluate if her contribution is enough for being considered as author, but if not, should be considered take her name off from author contributions and put her instead in acknowledgments.

Voordendag et al. present in this paper simulations of blowing snow over a glacier and its surrounding area (a few km²) with Large Eddy Simulations from the WRF model associated with a new blowing snow module (not fully coupled with the atmospheric model). The simulation period covers only 24 hours including frontal precipitation and a later blowing snow event. Evaluations are done with a few local sensors and with rather unusual Terrestrial Laser Scan measurements available on this area. The paper is well structured and results are correctly described and discussed. Some results are interesting, although it is maybe difficult to identify a really striking conclusion after reading. This might be due to the fact that the general goal of the authors is not really explicit and the reasons for developing a hybrid method for blowing snow simulation (combining very expensive LES and a rather simple blowing snow scheme) are not so easy to understand. As a result, the significance of this paper is difficult to judge as obviously for operational perspectives, the conclusions are very limited by the simulation period and domain size while on processes perspective, the modelling tool might be considered as not ideal compared to previously developed systems. Therefore, I would encourage the authors to more develop their motivations in their revised manuscript. Then, as the availability of detailed TLS observations if one of the main strength of this work, I think that the analysis of the simulated patterns of snow redistribution should be the central result of this paper and go further than simple map comparisons. A more detailed and mature analysis of simulation results is to my mind necessary before publication. Beyond these general comments, I also have a number of short comments and questions below that I hope authors can address in their revision process.

L35 metamorphism is more common than metarmophosis. I would identify melting and refreezing processes first as phase changes that also induce further metamorphism (but the processes involved are not only metamorphism).

L63-67 This paragraph is a bit confusing. I think that most LES models can not really be considered as NWP systems as they commonly can not be operated operationnally for weather forecasting purposes because of their very high numerical cost. Even if WRF may cover both applications, in the general case this distinction should be better done, especially because the spatial scale of NWP operational systems (at the best kilometric) is not compatible with an explicit representation of blowing snow in mountain terrain. So it is perfectly normal that operational NWP do not represent this process and it should not change in the incoming years. Coupled systems implementing blowing snow between LES atmospheric models and detailed snow models have been developed mainly for processes understanding, but they could not be applied at large scale in NWP applications.

L72-73 « the disadvantage of this and other approaches is the extra coupling of the models with a (previously) stand-alone snowpack model. » I don’t understand why the fact that these schemes were standalone at the origin would be a disadvantage here. Note also that the numerical cost of detailed snowpack models is actually very low compared to atmospheric models.

L73-81 The proposed methodology is probably useful considering this unusual evaluation dataset, either for process understanding either for model evaluation. However, my feeling is that the governing general motivation of this study is not fully explicit in this introduction, especially considering the concerns I mentioned about the previous paragraph (L63-72). I think the authors could easily improve that point to clarify their motivations.

L139 Can you provide the surface of the different simulation domains ?

L150 What is the density of fresh snow in the model ? Wouldn’t it be more realistic to initialize initial density from previous offline simulations of any snow scheme (way less expensive than coupled simulations) ? I am afraid the variability of snow density is sufficiently high to potentially be responsible for a bias in initial snow height that could be far from « slight ».

L173 I don’t undersand why the ice density appears in Equation 5. First in the current form of the Equation, it could just be simplified. Then, physically, why would the ice density have an impact on the terminal fallout velocity ?

L188 But does this surface snow density also evolve with snowdrift ? And how ?

L211 « the slopes adjacent to HEF showed a more heterogeneous snow distribution between scan 1 and 2 (Fig. 3a), which might indicate snow redistribution during the snow fall even ». I am not sure where to look at but it is honestly hard to distinguish any spatial variability in Fig. 3a. Then what are the arguments to identify this variability as snow redistribution during the snowfall event ? i.e. how to disentangle from local scale precipitation variability ?

L215 It is a bit surprising to have sensors installed in unrepresentative locations and that the high resolution of TLS would not be able to identify the mentioned local anomalies. Although the authors explain they do not want to consider data assumed to be inconsistent at the IHE and StHE sites, the results are shown in the Figure and their discreapancy with TLS measurements raises questions.

L236 « Melt can be excluded during the cold case study period » I guess this is true but the fact that temperatures are cold is an unsufficient argument to exclude melt at all slope aspects. The result of a simulated surface energy balance would be a better argument to exclude melt.

L245-249 The authors say that « snow redistribution patterns can only be simulated correctly if the modelled precipitation and wind patterns agree with the observations. » However, there is not any attempt in the paper to evaluate the uncertainty of simulated precipitation. It is true that measurements are affected by undercatch. However, correction functions exist to account for undercatch (e.g. Kochendorfer et al., 2017). Precipitation from atmospheric models are also known to be prone to very large errors. Therefore, as data are available, I would highly recommend to include in the paper the comparison between observed and simulated precipitation. I could imagine that this work was actually done by the authors and excluded from the paper because of large discreapancies between simulations and observations. However, being aware of these discreapancies is to my mind very important to have in mind the uncertainty of precipitation amount.

L264-265 «  Overall, the model simulates the wind field on the glacier and the surrounding with confidence and provides good input conditions for the snow redistribution and snow water equivalent (SWE) estimates. » Confidence is always relative. I think that this sentence could me more quantitative (e.g. providing an error metric). Furthermore, Figure 5 allows only to evaluate the ability of the model to simulate the wind temporal variability, but the ability of the model to simulate its spatial variability can not be really estimated while the wind spatial patterns are especially important in snow redistribution. This should encourage to moderate this conclusion.

L267-279 Is compaction represented in the 3-layer snow model ? In that case, would it be possible to compare the simulated compaction for this specific event (e.g. disactivating snow drift) with the simulated compaction of the 2021-2022 and 2022-2023 seasons in offline simulations ? It may allow to better characterize this specific situation than considering all observed compactions over these 2 years.

L284-285 The sign should be given (i.e. erosion or accumulation), I guess erosion from the plot but it should be explicit in the text. The unit is also missing on the top panel of Fig. 6. Then, why expressing this value in terms of height rather than in terms of mass ? As blowing snow also modifies the density, it is bit counterintuitive to mix mass transfer and density changes in a single diagnostic of snow height change, especially if the authors want to interpret this result in terms of mass change as in line 289.

L299 « SWE remained constant until 15 UTC due to compaction » : snow depth reduces because of compaction, but compaction does not make the SWE to not change.

L300-304 Was there any evaluation of simulated sublimation with this blowing snow scheme in previous studies ? How reliable is this very low estimation of sublimation ?

Figure 7 : As mentioned before, I am a bit surprised about the choice to present erosion/accumulation in meters and not in kg/m² as it does not allow to separate changes in density and changes in mass, and as in this Figure there is no comparison with observations.

L310-311 « The simulated flow patterns were, in agreement with the observations, almost perfectly down-glacier (Fig. 5). » I don’t understand this statement, could you please rephrase ?

L311-321 So I understand that the model is not mass-conservative across this small domain as erosion is much higher than accumulation and sublimation is negligible. That means that blowing snow fluxes at the domain boundaries are especially important while there is a discontinuity in resolution at this boundaries. First of all, was mass conservation checked at larger scale (on the coupled domains) ? Then, how boundary fluxes may be impacted by this discontinuity in resolution and could it have an impact on the resulting unbalance between erosion and accumulation on the studied domain ?

L323-332 The reasoning about the average magnitude of compaction and snow redistribution to explain observed snow height changes is interesting. However, the interest of such detailed modelling in a context with high resolution observations is obviously to look at the realism of spatial patterns. In this section, the evaluation of simulated spatial patterns remain unfortunately rather subjective with a rather vague map comparison. To my mind, this is disappointing. I would have expected a much detailed spatial analysis exploring spatial correlations between simulations and observations or other metrics allowing a more objective evaluation of the simulated spatial patterns.

L347-348 « Therefore, we assume that no model bias emerges due to erratic wind patterns » Authors should specify that this statement only applies for their specific study case as there is no evaluation of systematic biases on longer periods.

L348 « The simulated snow redistribution is realistic in terms of spatial structure and magnitude. ». This strong conclusion should be supported by more objective results than simple maps comparision as previously noticed.

L349 Again the ‘smoother’ behaviour ou simulated snow fields should have been quantified for instance by a spatial variance analysis of the observed and simulated maps.

L349-350 Although this is a plausible explanation, I think there is not in the paper a rigorous demonstration that this process is able to explain the spatial variance discreapancies between model and observations and that only this one is involved.

L359 I think I don’t really understand why simplicity in the snow drift module and snow scheme is really an advantage when the choice for the atmospheric model is a very expensive LES model that can only be applied for very short simulation periods and that in any case prevail in the numerical cost. This is probably linked to the lack of clarity of the governing objective of this study, is it a process study, a model evaluation, something else ?

L371-373 « The results of this snow compaction (not shown) are overestimated, because the model assumes the entire snowpack (>2 m) to compact and not only the 0.48 m of fresh snow. » This statement is very unclear and obviously compaction happens in the whole snowpack. Please rephrase.

Finally a section dedicated to data availability is necessary to fit data policy for publication in The Cryosphere: https://www.the-cryosphere.net/policies/data_policy.html