This study explores using Sentinel-1 data to enhance the accuracy of fully distributed, physically-based simulations of mountain snowpack. Specifically, the wet snow line (WSL) altitude derived from Sentinel-1 wet snow maps and the snow line (SL) altitude derived from Sentinel-2 snow cover fraction maps were compared against the WSL and SL simulated by the FSM2oshd model.

The study demonstrates good agreement between the model and ground observations, primarily located on flat terrain, in terms of snow depth. However, significant discrepancies emerge between the modeled WSL on south-facing slopes and the WSL derived from Sentinel-1 data. This discrepancy appears to correlate with errors in the model SL detection. By adjusting the fresh snow albedo parameter, the authors achieved an improvement in model performance for south-facing slopes when compared to Sentinel-1 WSL. This highlights the value of satellite data in identifying and correcting model biases.

The core concept of leveraging satellite data for model evaluation and improvement is highly promising and opens doors for advanced data assimilation techniques. However, further research and elaboration on the manuscript are needed to address some limitations and strengthen the work before publication.

A critical issue I see with the paper is how the authors arrived at the conclusion of adjusting the fresh snow albedo parameter as the primary explanation for the underestimating melt on south-facing slopes. Indeed, the explanation provided in the manuscript lacks a logical progression (i.e., the information is sparse around the paper) and fails to justify the selection of fresh snow albedo value as the sole “culprit”. If fresh snow albedo, or in general albedo, is a significant factor, the work should have explored incorporating satellite-derived albedo values. In fact, research on satellite-based albedo estimation for snowpack data assimilation is an already established field. Reviewing and citing relevant past studies would strengthen the paper and provide valuable context, which are now omitted. Moreover, by comparing the findings of this study with previous work on assimilation of satellite-derived albedo, the authors could highlight the advantages and disadvantages of each approach. This analysis would allow readers to understand the potential benefits of using Sentinel-1 data for assimilation compared to traditional albedo observations. However, this work offers an opportunity to explore alternative approaches beyond this traditional reasoning.

The perception by reading the manuscript is that the suggestion to work on the fresh snow albedo values seems like a convenient shortcut, addressing numerous potential sources of errors in the snowmelt model with a single parameter tweak, without the complexity to perform a data assimilation inside a physically based model. It is interesting to review some of the sources of errors. These include, for example i) shortwave radiation spatialization; ii) albedo decay function; iii) longwave radiation calculations and spatialization; iv) snow temperature; v) water transfer processes; vi) distinguishing liquid from solid precipitation, etc. While the paper mentions some of these sources of errors sparsely around the different part of the manuscript, a dedicated section with a clear, step-by-step reasoning process is crucial for readers comprehension. In fact, a more thorough exploration of these concepts could likely lead to the conclusion that using WSL data can be potentially used in a more effective way than solely operate on the fresh snow value. This seems to be also confirmed by the results: Figures 8 and 9 suggest that the change in fresh snow albedo only partially improves the results and there is still room for improvements. For this I suggest the authors to explore the different methods to assimilate WSL data into the model. This can be done operating on other model parameters, as done for fresh snow albedo, or even on the physical equations. For example (just to stimulate the discussion), one alternative to modifying albedo is introducing a new parameter linked to the Sentinel-1 observed WSL that control the net radiation. This parameter would be meant to adjust the net energy input for some of the different sources of error mentioned before offering several advantages. It can indirectly account for various model errors without directly modifying a specific physical parameter like albedo. This avoids potentially unrealistic changes to an established physical value. By focusing on the observed WSL, this approach may be more effective than relying on temporally sparse albedo observations coming from optical imagery or changing the fresh snow parameter. It directly adjusts the energy budget to reflect the presence of wet snow for a specific aspect and altitude, which Sentinel-1 data can reliably identify. This could lead (or not, this is just an example to stimulate the discussion) to a more robust simulation of snowpack melt showing that optimizing WSL can improve SL and snow depth, which is not an immediate logical consequence (see next paragraph). I would suggest the Authors expanding the discussion on these aspects in the manuscript. This will provide a clearer message for the (remote sensing) community: not only albedo, snow cover area or snow depth are important and feasible to be assimilated.

In this sense, the insightful analysis of the connection between WSL bias and SL bias, presented in the paper, could benefit from a more thorough integration into the discussion above. In general, WSL primarily reflects the cumulative energy input received by the snowpack over time, while SL integrates various snowpack processes throughout the season, primarily including snowfall, redistribution, melting and sublimation (and evaporation). By assuming the precipitation and snow redistribution are accurately represented in the model, WSL information may help refine the energy distribution within the snowpack. This could occur when factors like snow temperature, albedo, slope, aspect, or rain-on-snow events are not adequately accounted in the spatialization operations. By correcting using the observed WSL, the energy input is adjusted, potentially leading to a more accurate simulation of the overall snowpack depletion and, consequently, a better representation of the SL throughout the season. However, it is not immediately clear how solely adjusting WSL leads to improved SL, and therefore improved snow depth (or SWE) across the entire season. I would suggest the Authors expanding the discussion on these aspects (some of them are already in the paper). This would significantly strengthen the paper.

In summary here are some areas for potential development:

• Provide a more nuanced explanation of how WSL information may refine the melt within the model considering all the sources of errors.
• Elaborate on a conceptual strategy to potentially assimilate WSL showing the specificity of the assimilation (e.g., how would the assimilation parameter enter in the snowpack equation/parameter and how can be calculated based on the WSL observations?)
• Elaborate on the specific modeling conditions where WSL correction impacts SL across the season. Are the same conditions that allows WSL correction to improve snow depth simulations?
• Discuss potential limitations of assimilating WSL, particularly in scenarios where precipitation or redistribution might not be accurately captured.

In conclusion, my suggestion is that by elaborating on these concepts throughout the manuscript would strengthen the paper and finally present a clear direction for future research focused on developing robust and innovative approaches to data assimilation. These new approaches should move beyond relying solely on “traditional” input variables like snow cover, albedo or snow depth. To reflect this broader focus, the authors can consider changing the title of the paper to highlight its exploratory nature e.g., Exploring how Sentinel-1 wet snow maps can inform fully-distributed physically based snowpacks models.

Detailed comments:

 L40-41: There are a few points to consider regarding the terminology used to describe melting processes. The term “moistening” has a specific definition introduced by Marin et al. (2020), building upon earlier works by Dingman (2015) (where the melting phases classification use a slightly different taxonomy) and Techel & Pielmeier (2011). Using this term without proper context or referencing the original work might be unclear for readers unfamiliar with that definition.

When referring to wet snowpack, “saturation” typically indicates the percentage of the porosity filled with liquid water. However, in this context, it seems that the authors meant the maximum capacity of holding water of the snowpack.

Finally, the process of snow ripening is slightly more complex than a simple “bucket scheme” (e.g., Marin et al. (2020), Techel & Pielmeier (2011), Essery (2015)). Here is a more accurate description: the liquid water released or absorbed from the superficial layers gets in contact with the subfreezing snow present underneath and freezes. This releases latent heat that causes the snowpack to warm up. This starts the process of snow ripening.

My suggestion is that when dealing with specific terms or concepts, it is always good practice to cite the original (and most recent) papers. This allows readers to delve deeper into the topic if they wish.

L66. Margulis et al 2016 is only one example.

L67: why the information is qualitative and not quantitative? Maybe binary?

L70: Premier et al 2023 used S1 information for assimilation. However, this was done in reconstruction.

L90: While explicitly stating research questions is a good way to introduce a paper, it requires more specific details in this case. For example, snow depths, snow cover fraction and wet snow status are not interchangeable concepts (even if connected), how do you check if they are complementary? Moreover, to test if the wet snow information enhances the accuracy of the snowpack simulations imply to use independent validation data in complex terrain, which seems not to be the case.

Finally, I suggest the importance of answering the research questions explicitly in the conclusion. The conclusion should summarize the key findings related to each question, demonstrating whether and how Sentinel-1 data improves the model representation of snowpack.

L103: It would be important to explicitly mention that all these stations are in flat terrain, right? I personally would like to see also the distribution by elevation and predominant aspect in the final resolution of 250m.

L113: not only patchiness but also superficial snow roughness is playing a role.

L115: For me it is not clear what is meant with “fully saturated”. Maybe the Authors refer to the fact that the snowpack exceed the maximum water holding capacity? If so, this is not always the case. If there are no impermeable barriers, LWC is typically limited by the snow grain density and shape (Goto et al., 2012). It may be that with large grains the snow can have larger porosity. In this case the LWC can be relatively low, since LWC is drained quickly by gravity. Instead, if the Authors with “fully saturated” means 100% water saturation, means the snowpack is (decomposing) snow slush. And this is not the case for patchy snow. Please clarify this.

L116: “viewing slant angle”: due to the SAR lateral view.

L116: masking the geometric distorted regions cause data gaps. Layover and foreshorting does not mean the data is missing (like in shadow), the backscattering of these area is “distorted” inside the selected final resolution cell.

L118: the nominal revisit time at equator is 6 days for the two satellites. Sentinel-1B failed in December 2021. Due to track overlap, at the considered latitude, a minimum of 2 tracks (one ascending and one descending) to a maximum of 4 tracks (2 ascending and 2 descending) are available depending on the ground location, for every repetition cycle of 6 days when the two satellites were available, and every 12 days when only Sentinel-1A is available. A reference to Fig 2 can be used here to better explain this concept.

L118: are the maps binary? Wet, non-wet?

L121: with “coinciding” is it meant “aligned to the upper left cell of the model grid”?

L123: Had the forest areas been masked out from the Alpsnow products before use?

L125: it would be good to express the % in number of pixels. In the end you aggregated inside a window 5x5 pixels, right?

L128: Do you expect S1 provides a good sampling time for wet snow?

L135: following problems? I suggest expanding the listed items with a clearer description of each point.

L139: what is a snow-free glacier? Bare ice is identified as snow? Please clarify.

L151: I suggest recalling some basic details of the model in this section e.g. how long- and short-wave radiation is accounted in the net radiation budget?

L170: The rationale behind adjusting fresh snow albedo from 0.92 to 0.86, and how these values compare to the literature, needs further clarification.

L172: I suggest introducing a conceptual block scheme of the operation done in the comparison.

L176: I generally agree that the Sentinel-1 is very sensitive to small LWC, but this need to be quantified. What is the value you used?

L180: While this is true, it is important to acknowledge that aspect and slope can also influence the melting as shown later.

L190: I missed where are the glacierized areas.

L191: 300m altitude? Please better clarify also considering the previous aggregation at 250m and how the two aggregations may interact.

L196: how the value 0.8 times WSFmax was selected?

L224: how to see flat terrain from Fig 6?

L226-230: This explanation cannot be introduced in the results sections, but before in the manuscript. While adjusting fresh snow albedo can be a way, it is worth considering whether directly modifying net radiation might be a more comprehensive strategy as indicated by this sentence.

L270: Again, operating on the fresh snow albedo value does not seems the correct indication to give here.

L294: While I can generally agree with these statements it is important to acknowledge the limitations of spatial aggregating and reduce the dimensionality using the snow line concept. These operations inherent overlook some important spatial variations. For example, areas with sub-zero air temperatures and specific topographical features, like the shadowed bottom of a narrow valley, might remain dry even when higher altitude snow covered areas may be wet. Additionally, it should be acknowledged that by exploring high-resolution data or alternative spatial modeling techniques might be valuable for capturing these finer-scale variations.

L295: especially for real time assimilation, it seems.

L316: please provide a reference.

L332: Premier et al 2023 as well. Using melting phases derived from Sentinel-1.

L346L maybe the word cluster fits better?

L387: for the conclusion I suggest answering the questions presented in the introduction. Do you plan to change the fresh snow albedo value in FSM2oshd?

Fig. 1 In the caption and legend the “numbers” explanation is missing. Are them the MEZ IDs? Why only some are reported?

Fig.2 It is not clear why some areas have 0 observations. I imagine shadow, layover, clouds? Please explain it in the caption.

Fig. 3 What is the range of “westerly” aspect? Are the real curves as derived from the data or a conceptual example? What was the snowline for this case? Why S1 WSL is decreasing for lower altitude? Patchiness and snow roughness? Please explain it in the caption.

Fig 4. The model was calibrated using the 444 stations for the reported years or others years?

Fig.5 why fsm_optim example is not reported?

Fig 6. What is “flat:2160”, “flat:1940”, etc for the three plots? Why 0% for lower altitude for S1? Is the patchiness? Please better explain this behavior.

Fig 8-9: it would be interesting using this plot to identify the possible sources of error operating on the different model parameters or equations. The fact that there is negative bias for both SL and WSL is interesting and should be better investigated.

Dingman, S.: Physical hydrology, Waveland press, 2015

Essery, R.: A factorial snowpack model (FSM 1.0), Geosci. Model Dev., 8, 3867–3876, https://doi.org/10.5194/gmd-8-3867-2015, 2015.

Goto, H., K. Kikuchi, and M. Kajikawa (2012). Influence of different surface soils on snow-water content and snow type of the snow cover. Japanese Journal of Snow and Ice 74:145–158

Premier, V., Marin, C., Bertoldi, G., Barella, R., Notarnicola, C., and Bruzzone, L.: Exploring the use of multi-source high-resolution satellite data for snow water equivalent reconstruction over mountainous catchments, The Cryosphere, 17, 2387–2407, https://doi.org/10.5194/tc-17-2387-2023, 2023.

Techel, F. and Pielmeier, C.: Point observations of liquid water content in wet snow – investigating methodical, spatial and temporal aspects, The Cryosphere, 5, 405–418, https://doi.org/10.5194/tc-5-405-2011, 2011.

I reviewed with interest the manuscript by Cluzet and colleagues on using S1 wet snow maps in snow model evaluation. Authors provided a novel contribution on the long-standing topic of using wet snow maps in snow modeling, by evaluating an operational snow model in Switzerland with several years of wet snow from S1 and fractional snow cover from S2. Results show a general good agreement, with a North vs. South bias that was improved by adjusting fresh snow albedo.

Overall, the research is novel, relevant, and timely. Particularly the intuition of a wet snow line is interesting and will likely be used in several future papers. Thus I am in favor of publication, after a minor revision.

Like R1, I was also a bit puzzled by the choice of adjusting fresh snow albedo as the main approach to correct the mismatch between model simulations and observations of wet snow. On the one hand, I understand that this variable is related to snowmelt onset and thus is one of the variables involved in this mismatch. On the other hand, in my understanding fresh snow albedo has a clear impact only when snow is fresh, while parameters related to the seasonal evolution of albedo seem more important to me here. Also the procedure that led to this adjustment is not clearly outlined and should be better discussed.

Another potential opportunity for improvement here is that the whole of the discussion around the wet snow data is performed in terms of wet snow line, rather than pixelwise values. Authors are clear on why they are doing so, and I generally agree. I still believe that computing confusion matrices for aspect or elevation classes would provide additional insights around model performance. 

Some more minor comments:

- Line 26: maybe also mention lateral flow as a way for snowmelt to move away from the snowpack without exiting from the bottom of the local snow cover

- Line 99: maybe better define what these hydrologic units are and how they were delineated?

- Line 141: remove one “and”

- Line 158: maybe spend some more words on this FSC = f(SWE, HS) relation here, given that it is quite important for this paper?

- Section 3: I would recommend including more quantitative metrics here in place of wording like “excellent”, “higher elevations”, etc.