Summary:

This manuscript presents an analysis of Sentinel-1 derived snow wetness evolution and wet snow avalanche activity in a study domain in Switzerland over multiple winter seasons. The SAR-based wetness maps are evaluated with SNOWPACK model output. Key findings include an increase in wet snow avalanches when the wet snow ratio exceeded 40% in corresponding elevation bands and a correlation between SAR backscatter and modeled liquid water content.

Major comments

1. Clarification of methods

                  -In Situ/Model Data Description

I found the description of the in-situ (AWS data) and SNOWPACK model to be rather confusing.  Are the IMIS stations assimilated into SNOWPACK?  Does the SNOWPACK timeseries at station locations reflect the station data or if not, how has it been modified? Are the timeseries in Figure 2 observed or modelled data? Please see minor comments below for additional detail.

                  -Wet Snow Ratio

The wet snow ratio (ln 139) features prominently in the results but is fairly “hidden” in the methods.  I suggest presenting this as a formal equation to help highlight it and then reference it elsewhere in the manuscript.

1. Impact of Sentinel-1 overpass timing (morning and evening) on merged product.

The manuscript describes that a key advance of this workflow is the use of LRW composites to overlook viewing geometries.  However, snowmelt and LWC varies temporally over multiple temporal scales (diurnal to seasonal).  What is the impact of including morning and evening overpasses in the LRW composites?  What if a pixel is frozen during the morning acquisition but melting in the evening acquisition?

1. More detailed comparison to previous work

The discussion starts out describing the advance of this work over prior methods (lines 220-224) but this comparison was not explicitly made in the manuscript.  I suggest the addition of a rigorous comparison to existing workflows be included, as that would demonstrate this advance more clearly. This in regards to both temporal (lines 220-224) and spatial (lines 224-226) scales.

1. Data Availability

I strongly encourage the authors to archive the datasets in a publicly available repository rather than by request from an organization.  At present, the manuscript is not compliant with TC data policy: https://www.the-cryosphere.net/policies/data_policy.html, “The best way to provide access to data is by depositing them (as well as related metadata) in FAIR-aligned reliable public data repositories, assigning digital object identifiers, and properly citing data sets as individual contributions.”

Minor comments

2 – snow wetness has implications beyond avalanche release, as detailed later.  A broader justification for this work would be appropriate in the abstract.

3- replace “allow us” to “facilitate”

4- replace “show how” with “utilize”

4 – delete “can be used”

5 – state the number of seasons/years

5 – I suggest using evaluate rather than validate here and elsewhere (ln 50, for example)

10 – briefly define wet snow ratio here 

14 – simplify “operative monitoring application” to operational monitoring

15 – include specifics on why wet snow avalanches are still difficult to predict

23-34 – I found this paragraph to be somewhat confusing, as background on the use of SAR for SWE retrievals and snowmelt detection is presented.  I suggest revising to clarify which application is being discussed.

26 – what does “compare Lievens et al 2020” mean?

30-34 – I was surprised to not see the inclusion of relevant literate like Lund et al. 2020 (https://www.frontiersin.org/journals/earth-science/articles/10.3389/feart.2019.00318/full) and Gagliano et al. 2023 (https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2023GL105303), both of which utilized Sentinel-1 for snowmelt detection

48 – I suggest including details on the Karbou et al. 2021 products rather than just the citation here.

Figure 1 caption - revise: …of research at three spatial scales.

54-55 – check formatting of elevation values

55 – provide more details regarding field site slope.  I recognize that this is a large area, but a range and median value would provide a useful characterization.

55 – provide additional details on forest cover (tree types, canopy heights, % of study area, etc)

58 – Does this range represent daily values over the entire year?  Averaged over how many years?

67 – clarify: this would be reduced

80-81 – provide more details from the Nagler et al study on why this 2dB threshold was selected

83 – revise sentence structure to not start with: As reference Nagler…

94 – what percent forest cover was used for this mask?  How sensitive were the results to this selection?

133 – Please clarify and provide an example for what is meant by “entries occurring more than 25% of the time were ignored”?

136 – please describe this plot design in more detail for readers not familiar with Karbou et al. 2021

138 – Why were the temporal classes defined by the ascending rather than descending (or average) Sentinel-1 acquisitions?

138 – Was the sensitivity to masking by 28 degrees assessed?  Given radar viewing geometries on steep slopes, could better results be achieved by assessing wet snow presence on slopes <28 degrees even though those don’t coincide with avalanche starting zones?

139 – I suggest referring to it as the number of pixels rather than amount of pixels.

153 – could add a reference to the wet snow ratio equation (see previous comments)

159 – check formatting throughout regarding en-dashes

Fig 2- what is the temporal resolution of the LWC plotted in Figure 2?  3 hrs or has this been smoothed to daily resolution? If smoothed, does the original time series include diurnal variability? 

Fig 2 – in 2020, the LWC (blue) increases when HS increases in the spring?  HS increases indicate snowfall with <0 C temperatures, so why would LWC increases occur simultaneously? This can also be seen in 05-2018.

Figure 2 caption – What does it mean to be extracted from SNOWPACK data? Is this model output or is this measured data?  This needs to be clarified as the abstract describes this as modeled water content but this caption describes it as measured liquid water content.  See Major Comment.

176-77 – given the reference to the other stations, a similar time series to what is shown in Figure 2 should be included as supplementary figures.

179 – LWC doesn’t need to be capitalized when being defined in acronym.

181 – “a time shift of a couple of days” is vague yet this offset is important to understand.  I suggest quantifying this in a more rigorous manner.

185 – capitalize Spearman’s

190 – insert comma after 2018-19,   also replace dash with en dash

193 – clarify: end of April to end of March.  Is this meant to be end of May?

200 – what is meant by hence not extra?

257-260 – If the SNOWPACK model is assimilating the station data (which is what I understand is happening), what is the value in only using the model output at the station locations over the measured station data?    Further, why not compare the model output over the full domain to the Sentinel-1 melt products through time?

265 – was the sensitivity to the selected 3x3 window evaluated?  What if a 5x5 window was utilized?  This might provide some insight to the previous statement regarding whether “stations having an impact on the radar backscatter.”

280 – see previous comment re: time lag

294 - should this read perimeter rather than parameter?

298 – revise sentence for improved clarity “this influenced the miss of the first….”

326 – delete also

Additional comments:

1. I suggest the addition of a Discussion paragraph on avalanche dynamics related to meltwater percolation. Sentinel-1 is sensitive to surface to near-surface melt while wet snow avalanches initiate due to failure of a buried interface or at the bottom of the snowpack. The discussion would benefit from some insights regarding these differences.  
2. Another common trigger of wet snow avalanches is rain on snow. While it would be beyond the scope of the manuscript to add a significant analysis in this regard, a simple analysis comparing a re-analysis precipitation product (e.g., ERA5-Land) with the avalanche frequency might identify which events are melt related and which were precipitation related, and improve the comparisons presented in this manuscript.

Data Availability: See major comment.  Also provide a reference for the Gamma software program.

This study used Sentinel-1 SAR data to map snow wetness in alpine areas, finding strong correlations between backscatter and modeled liquid water content, as well as good agreement with wet snow avalanche occurrences (excluding the first wet snow avalanche surge). The results suggest Sentinel-1 has potential for monitoring wet snow avalanche preconditioning, particularly with increased temporal resolution (starting from additional satellite tracks). The paper is in general well written and the use of detailed avalanche catalogue to find a correlation with the Sentinel-1 backscattering is very interesting, even the study area is relatively small. There are some methodological choices that requires further explanation and discussion before the paper can be published in TC.

Major comments

• I expected the authors to demonstrate the added value of radiometric terrain flattening before generating the LRW mosaic, utilizing all four available Sentinel-1 tracks over the study area. Given the rapid temporal changes in snowpack LWC, as acknowledged by the authors, averaging morning and late afternoon acquisitions (as done in the LRW approach, which is essentially a weighted average) may not be optimal. For example, if a morning acquisition has a higher weight, and the LWC is low due to a cold night (resulting in higher backscatter), this could skew the result. This is particularly problematic early in the season, when the afternoon acquisitions can be affected by rapid temperature and radiation drops, showing an already potentially low LWC (and therefore backscattering) from its (midday) peak (which may be the cause of the wet snow activities in April?). To rigorously assess Sentinel-1 ability to detect the initial wet snow avalanche surge, these temporal variations should be analyzed before constructing the LRW mosaic. This analysis would provide a stronger basis for your conclusions. Therefore, the rationale behind using a mosaic with varying timestamps to address layover and shadows at a specific time requires further clarification.
• Furthermore, analyzing the four individual tracks prior to mosaicking would provide valuable insight into the effectiveness of the terrain flattening. Residual angular dependencies, especially on aspect angle (the angle between azimuth direction of Sentinel-1 and geographic north), can introduce biases between backscatter acquired from different tracks. Has this been addressed? I recommend showing the temporal evolution of γ⁰ for all four tracks over the WFJ station to demonstrate the effectiveness of the terrain correction (in theory no bias should be visible between the tracks). Be aware that ascending and descending acquistions have generally a specular aspect angle.
• The literature review could be strengthened by including additional background studies in both radar remote sensing of wet snow e.g. Murfitt et al., (2024) and wet snow avalanches e.g., Mitterer, and Schweizer (2013). In particular, citing key foundational works and recent publications would provide important context and allow for a more robust comparison with your results (see detail comments). This would also better support why in this study the Copernicus wet snow products were excluded a-priori.
• While the application of a detailed avalanche catalog and Sentinel-1 backscattering time series is a novel aspect of this study, the title is misleading. The paper appears to be an exploratory investigation into the correlation between wet snow avalanches and LRW composites. Since a more robust justification for using LRW composites as the primary metric for snow wetness evolution is needed, I suggest to better sharpen the current title.
• Consistent with TC publications policy, I suggest the authors to make the LRW time series and all data publicly accessible. The statement “data available upon request” does not meet current standards for open science and reproducibility. Depositing the data in a recognized public repository (e.g., ENVIDAT or Zenodo, or Dryad) would greatly enhance the value and impact of this work by enabling independent verification and reuse of the data.

Detailed comments:

L17: more recent works have been published on how use Sentinel-1 for SWE/runoff modeling e.g., Cluzet et al. (2024) or Premier et al. (2023).

L23 to 34: This section would benefit from a more focused literature review. The current “ping-pong” between dry and wet snow literature makes it difficult to follow the narrative. Since the paper focus is on wet snow, I recommend removing the discussion of dry snow backscatter changes and concentrating on a comprehensive review of wet snow literature. Key works by Matzler, Ulaby and colleagues that describe the main backscattering mechanisms in wet snow should be included, as well as recent advancements e.g., Picard et al. (2022) or Murfitt et al. (2024). I also suggest exploring the potential link between the “first wetting” described by Hendrick et al. (2024) and the melting phases presented in Marin et al. (2020) for SAR multitemporal data. Investigating this connection could offer valuable theoretical insights.

L41: Two angles affect backscatter: local incidence angle and aspect angle. It is crucial to consider both, as they have distinct effects. To ensure the terrain flattening effectively corrects for these influences, please clarify whether the aspect angle was incorporated into the process. Showing the gamma naught (γ⁰) values for all four tracks would be very helpful in identifying any residual biases between them.

L43-45: While I understand the intent of averaging to minimize noise, I believe it is important to consider that the identified “outliers” could represent real-world afternoon wet snow conditions not captured in the morning data. This raises concerns about potentially losing valuable temporal information. I also disagree with the assertion that multi-temporal averaging improves temporal resolution; by combining data from different times, it effectively lowers the resolution. Perhaps exploring alternative noise reduction techniques that preserve temporal fidelity would be beneficial.

L50: The use of a 5x5 meter resolution raises concerns, as the original Sentinel-1 data has a ~5x15 meter resolution. This upsampling introduces artificial detail and does not represent true information gain. A 20x20 meter resolution would be more consistent with the original data. Could you please justify the decision to use 5x5 meters and explain how this upsampling was handled?

Line 69: please use the symbol γ⁰ or gamma nought instead of gamma0

Line 70: To ensure the validity of the results, please provide a more detailed description of how the gamma software processes γ⁰. Has the output of the gamma software been compared and validated against the data elaborated by David Small? The original Small et al. (2022) paper highlights potential inaccuracies in SNAP (an ESA software) due to different implementations (see the end of Section II-B). This raises concerns about potential similar discrepancies. A direct comparison or a comment on this would be beneficial for the community.

L126: it is not clear how you perform the average and why.

Results section: A comparison of the proposed processing with established methods, such as those of Mitterer and Schweizer (2013), Bellaire et al. (2017), and Hendrick et al. (2024), would greatly enhance the manuscript by demonstrating the novelty and performance of the proposed approach.

Figure 5: The gray dots are difficult to see and could benefit from increased contrast or a different color.

Additional references:

Sascha Bellaire, Alec van Herwijnen, Christoph Mitterer, Jürg Schweizer, On forecasting wet-snow avalanche activity using simulated snow cover data, Cold Regions Science and Technology, Volume 144, 2017, https://doi.org/10.1016/j.coldregions.2017.09.013.

Cluzet, B., Magnusson, J., Quéno, L., Mazzotti, G., Mott, R., and Jonas, T.: Exploring how Sentinel-1 wet-snow maps can inform fully distributed physically based snowpack models, The Cryosphere, 18, 5753–5767, https://doi.org/10.5194/tc-18-5753-2024, 2024. 

Marin, C., Bertoldi, G., Premier, V., Callegari, M., Brida, C., Hürkamp, K., Tschiersch, J., Zebisch, M., and Notarnicola, C.: Use of Sentinel-1 radar observations to evaluate snowmelt dynamics in alpine regions, The Cryosphere, 14, 935–956, https://doi.org/10.5194/tc-14-935-2020, 2020.

Mitterer, C. and Schweizer, J.: Analysis of the snow-atmosphere energy balance during wet-snow instabilities and implications for avalanche prediction, The Cryosphere, 7, 205–216, https://doi.org/10.5194/tc-7-205-2013, 2013.

Murfitt, J., Duguay, C., Picard, G., and Lemmetyinen, J.: Forward modelling of synthetic-aperture radar (SAR) backscatter during lake ice melt conditions using the Snow Microwave Radiative Transfer (SMRT) model, The Cryosphere, 18, 869–888, https://doi.org/10.5194/tc-18-869-2024, 2024.

Picard, G., Leduc-Leballeur, M., Banwell, A. F., Brucker, L., and Macelloni, G.: The sensitivity of satellite microwave observations to liquid water in the Antarctic snowpack, The Cryosphere, 16, 5061–5083, https://doi.org/10.5194/tc-16-5061-2022, 2022.

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.