Ebner et al. present an important and novel application for snow modelling, but I have some suggestions to clarify the paper.

9

Specifically, the comparison with Sentinel-2 data is for snow-covered area.

10

Redistribution of snow by skiers would not directly lead to a reduction in average snow depth. This may be a significant omission in simulations, but the statement in the abstract seems more confident than the discussion in the paper.

24

What is “early winter” in this context?

29

What do these national percentages represent? Can a link be provided for Lalli et al. 2019?

51

Monti et al. (2016) discusses initialization of a model with manual snow profiles, not remote sensing.

69

This is the width of the elevation bands, not the elevation bands themselves.

88

The snow management configurations in Table 1 are incomprehensible without reading Hanzer et al. (2020). Brief descriptions and reasons for selecting them should be given to make this paper more self-sufficient.

102

The SRU is a clever concept similar to the familiar HRU of hydrology, but it seems from the Supplementary Material that there is much more to the definition of SRUs than the slicing into elevation ranges described here.

158

In short, GNSS snow depth data were available for all pilot resorts except La Plagne.

184

The 0-1 range of SP has already been stated.

210

i = 0,…,N would be N+1 pixels

213

https://doi.org/10.1029/2010EO450004

226

Agreement between observations and models that the pistes are almost fully snow covered in the middle of the season is not surprising (these are ski resorts, after all!). A more interesting question, and a more important one for snow management, might be how much better the models perform in early and late season compared with simulations without snow management.

263

If slopes were not groomed, how are GNSS measurements available to quantify the model error? How does the lack of grooming lead to strong increases in RMSD?

278-279

Is something missing from this sentence? It does not seem to make sense.

281

Figure 7 is referred to before Figure 6

Figure 3

It seems counterintuitive that the brightest colour is the lowest snow persistence.

Figure 4

Having a zero line that is not in the centre of a radar plot is confusing. Absolute errors might be better, or at least highlight the zero line.

Figure 5

Having RMSD and MD on the same figure but with different axes is very confusing and makes it difficult to tell at a glance if an error is small or large. Using a single axis would compact the error range but would be much clearer (this is common in evaluations of weather forecast errors)

Figure 6

What is the nature of the large measured snow depth between 1400 m elevation? Is there a bump in the snow surface or a dip in the ground surface that is not resolved by the 10 m model?

Figure 7

Little variation is seen in MD between resolutions. Text in 4.3 discusses reduction in RMSD, so that might be a better variable to show.

Table C1

Why does the column for Sentinel-2 contain statements about errors in snow depth?

The authors have produced a novel study, applying snow models with modules for management and grooming to simulate snow depth at nine European ski resorts, evaluating errors in simulated snow depth and persistence. I believe this work is original and potentially useful. I do have some suggestions for improving the manuscript prior to eventual acceptance.

Some things that I was wondering throughout the paper:  What sort of accuracy or error range is required in order for these models to be useful to the ski resorts? What snow depths are required for viable and for optimal ski conditions? The simulation mean deviations should ideally be smaller than this value. Does the availability and rate of production of artificial snow at the resorts affect this?

While MD is not incorrect as it has been calculated, I prefer all metrics comparing simulated values against observed values to be with respect to simulated minus observed, such that positive differences are overestimation and negative differences are underestimation.

Minor points or suggested corrections:

1. Line 32-34: Have any authors or ski resorts conducted a cost-benefit analysis for making snow that melts over various periods of time? Early season production of artificial snow may very well melt, leading to loss of the snow-cover, but this is likely weighed against income derived by the resort being able to remain open for a period of time. I suspect many resorts have some criteria and that a decision to make artificial snow is based on the likelihood that it will last long enough to recoup the cost. I don’t expect these questions to be answered in this paper, but I wonder if examples of such information exists such that they might inform the discussion.
2. Line 64: Change “snow-covered maps” to “snow-cover maps”.
3. Line 67: Change “unit” to “units”.
4. Line 88: Change “The used snow management configurations for” to “The snow management configurations employed for”.
5. Line 106: Aggregating as a post-processing step simplifies the presentation but not the computation. Has it been tested whether similar results are obtainable employing these aggregated areas for the simulations?
6. Line 152: Change “This technique relies on differential GNSS signals and takes measurements without snow depth on the slopes as a reference into account.” To “This technique relies on differential GNSS signals, comparing the snow-free (i.e. zero snow depth) reference signal with those obtained during the snow season, to obtain snow depth.
7. Line 156: Snowsat and Leica are not defined.
8. Line 169: Change “constrains” to “constraints”.
9. Line 187: Change “Additionally” to “In addition”.
10. Figure 4: Are the bold vertical lines in each plot of Figure 4 the intra-model means? This should be defined in the graph pane or in the caption.
11. Line 278: Is this averaging effect desirable? Instead of "allows minimizing the error", would "tends to mask the error" be a more accurate description of what is happening? Later, it is discussed that there may be a benefit to this, but I would still use “mask” perhaps as “tends to mask the variability in the error”.
12. Why is Figure 7 presented before Figure 6? I would rename the figures.
13. Line 309: Remove “it”.
14. Line 315: Change “use” to “uses”.
15. Line 321: Change “in average” to “on average”.
16. Line 325: Change “e.g., rapid snow melt inside the catchment are hardly to be matched correctly by the models” to “e.g., the ensuing rapid snow melt inside the catchment is difficult to simulate accurately”.
17. Line 337-338: This is not a sentence.
18. Line 352-354: The authors should define some accuracy requirements for the snow models to meet the needs of the ski resorts.
19. Line 357: I would not include the errors in the S and SW facing pistes with snow redistribution. These errors are caused by more rapid ablation because these pistes are exposed to high solar radiation during the warmest part of the day. I wouldn’t classify that as redistribution although both are important.
20. Line 393-394: I suspect that the ski resorts would know the minimum snow depth required. The research should attempt to determine whether the models can simulate snow depth with sufficient accuracy to enable the resort managers to maintain the optimum and minimum viable snow depth in a more efficient way.
21. Overestimation of snow depth and S and SW facing pistes could be addressed by having the incoming radiation adjusted for slope and aspect. I am not asking for this to be done in this paper, but it would be an obvious improvement for the next paper.

In this paper, numerical snowpack model and satellite data were applied for ski resort management. Hanzer et al. (2020) already introduced this system, so the target of this paper is the validation of the system for snow covered area and snow depth distribution. This system had practical accuracy and provide good information as a reference for building similar systems in other regions. This paper has sufficient results for publication to the Cryosphere. I wrote several suggestions to add explanations as minor comments. Please refer them for the revised manuscript.

Minor comment

L153-155 Is there any reference to show the accuracy of GNSS? Also, does the GNSS has similar accuracy for wet snow?

L241-242 Can you point out where the biggest difference due to snow gliding or avalanche in Figure 3? Also, this discrepancy may be reduced by integrating avalanche dynamics model.  Do you have a plan to integrate a snow redistribution model and avalanche dynamics model into this system? If there are any views for future implementation of them, description of it is desirable.

L250-254, Figure 4:   I guess that the better accuracy in high altitude is due to the ratio of snow cover area is near 1 (it may be most of them are true positive). Including the figure of simulated or observed snow cover ratio for each elevation and slope direction helps the relation of this ratio with OA.

L255-256 Figure 5 shows the amount of MD and RMSD for snow depth. I think the information snow depth is also necessary to check relative errors. Can you add the figure of snow depth data for simulation and observation?

L262-263 Although I haven't used and am not familiar with the grooming module, this error seems to be reduced if this module can turn on and off depending on the situation. This result can make suggestions to add them to improve the system.

L278-280 I think the averaging effects for RMSD can be avoided when 10m meshed GNSS (not averaged) and SRU averaged simulated data are compared. In this case, larger SRU size leads to larger RMSD. This comparison is not a requirement, but it is worth a try.

L355-362 I think it would be more informative if there is some mention of future plan, actuality to achieve, and level of importance for the improvement to resolve (1) - (5).