Summary and recommendation

Mazotti et al. develop and present a new physics-based, multi-layer, hyper-resolution snow model (FSMCRO) that can represent high spatial and vertical resolution snow properties including grain type, density, temperature, and other snow parameters. This was achieved through a one way coupling between the FSM2 canopy model and the ensemble Crocus model, with the added benefit that ensemble simulations provides a mean for assessing uncertainty. The paper focuses on introducing and demonstrating the model at two well studied snow sites (Finland and Switzerland), with only qualitative validation (“plausibility”). The model shows reasonable representation of snow depth patterns in Switzerland (focus in the main paper) but less so in Finland (supp. material). Overall, the model shows realistic spatial variations in key snow properties (grain size, SSA) and their evolution in time along a transect spanning a forest gap with variable radiation and interception dynamics. Through the use of the ensembles and spatial simulations, the study also finds that snowpack variability (due to canopy effects on snow processes) is more important than model uncertainty.

Overall, I find this paper potentially offers a significant advance in our ability to resolve very localized snow properties which will be of interest and use to research in snow-forest interactions, wildlife ecology, and possibly avalanche studies. I think the scientific and presentation are generally of high quality, though I offer some comments and suggestion for further improvement. My main concern is about the minimal validation effort and the apparent deficiencies in snow depth simulation at one of the sites (See #1 below), and therefore request the authors consider these before publication. I emphasize this paper should be published following attention to these comments.

MAIN COMMENTS

1. While the paper does not present a detailed validation but rather a demonstration of the new model, it seems there is still an opportunity to provide additional analysis to understand the “plausibility” of the model and needs for future improvements. For instance, the paper references weekly snow pit data at the Finland site, but does not make use of them due to issues with geolocation. I would argue that the geolocation issue with the pits does not preclude such a comparison, as multiple location from the domain could be selected, along with the ensemble members in order to understand the range of possible snow profiles simulated by FSMCRO. I think that a comparison between the FSMCRO ensemble and the snow pit data (grain type, density, etc.) could still be informative, even if done on a qualitative basis given the recognized challenges in comparing multi-layer snow models to snow pits. This might help to identify the plausibility of the model as well as possible deficiencies and areas for future development in the model. At the same time, this may require attention to the prominent errors in FSMCRO snow depth that are apparent at the Finland site (Figure S2, where even normalized snow depths are quite different from observations). As noted by the authors: “an adequate reproduction of observed snow depth patterns is a prerequisite for a meaningful subsequent analysis of snowpack vertical properties” (L. 285-286). Comparing to the Finland snowpit data might be helpful for diagnosing possible reasons for the deficient snow depth representation (e.g., bulk snow density?).
2. Several figures in the paper are not readable for someone with a red-green vision deficiency. As such, those readers may not be able to distinguish (for instance) the different snow grain types (e.g., melt forms vs. precipitation particles). I recognize this is not the fault of the authors as they are following the conventions from the Fierz et al. (2009) international snow classification report. However, I would suggest the authors consider whether something can be done to help these readers (e.g., adding a small hatch pattern to the green colors).
3. I recommend adding snow hardness and snow liquid water content (LWC) as new figures in the supplement (similar to Figures S3-S4), as the capability for mapping these variables spatially may be of high interest to other researchers. The paper references wildlife ecology, and for that the snow hardness is a relevant parameter. Likewise, snowmelt studies and microwave remote sensing (e.g., GPR) may benefit from a model that can resolve spatial variations in LWC.

Line Comments

- L. 27: This should be “tools”.

- L. 30-35: The opening sentence is rather long and cumbersome. I recommend breaking it into two or more sentences.

- L. 215: Add “an” after “as”.

- L. 256: The sentence begins with awkward wording. Please rephrase.

- L. 287: This is somewhat subjective and I think the sentence would be stronger if you cited the quantitative metrics here.

- L. 289: The phrase “not exactly recorded locations” is awkward wording. Please rephrase.

- L. 311: Should be “Sturm”.

- L. 452: Add “a” before “main”.

- L. 475-482: Can you please clarify whether blowing snow is simulated in the model or not? I think wind redistribution should be noted here as an important process for spatial variability of snow.

- L. 494: The sentence has awkward wording (“did not allow to evaluate”). Please rephrase.

- L. 506: Should be “prey” instead of “pray”.

- L. 518: Add “a” before “benchmark”.

- L. 520-521: The sentence begins with awkward wording. Please rephrase.

- L. 531-534: Could the new snow density and snow compaction parameterizations also be impacting the snow depth overestimation?

- L. 531: This focuses on one of the evaluations of the modeled snow depth, however, I think it is best to also acknowledge the prominent deficiencies in modeled snow depth at the Finland site in April (Figure S2). See my first major comment above.

FIGURES

- Figure 2, Figure S1, and Figure S2: Please add a scale bar.

- Figure 2: Please clarify in the caption what blue represents in the hemispherical photos. I believe it is in the sky portion outside the solar track but it would be helpful to state this in the caption.

- Figure 4: I wonder if it would be useful to show a plot of mean direct beam transmissivity at each location on the transect? This could go just below the Fveg and could have similar dimensions/scale. This is not a required revision but merely a suggestion if it helps to show the shaded area in the open gap on the left side of the figure.

- Figures 4, 5, 6, S3, S4, … : It could be helpful to add “S” on the left and “N” on the right at the top to indicate the south-to-north orientation of the transect.

- Figure 7: I suggest adding a map on mean canopy transmissivity, which I suspect might aid in interpretation of the spatial patterns here.

General Comments

The paper presents recent work by the authors to combine the strengths of two state-of-the-art snow models: the forest canopy representation from an intermediate complexity snow model (FSM2), and the detailed multi-layer snowpack model (Crocus). After outlining the two models and the process of combination into a new model (FSMCRO), two forested testing sites are introduced. Qualitative comparison of FSMCRO simulations is made to observations and baseline simulations with FSM2. The FSMCRO simulations are then interrogated at a series of points and transects to highlight differences in snow microstructure driven by location within the forest stand.  The multi-physics ensemble capabilities of FSMCRO are used to investigate how robust the simulated spatial differences in snowpack microstructure are to model uncertainty. Finally, the variability in microstructure from a multi-physics ensemble driven with domain-average meteorology is compared to the variability in microstructure produced by a deterministic high-resolution simulation at over same domain.

The work is novel and showcases new modelling capabilities that are undoubted state-of-the-art. The processes simulated are relevant to readers of The Cryosphere. However, there are several areas that require further description, results or discussion: 

The showcase-style manuscript, where different capabilities are presented and described, doesn’t necessarily demonstrate significant advances in knowledge provided by the new system. I expected to see more examples of quantitatively analysis of the multi-dimensional data - e.g. evolution of CV of different parameters over time. As well, I expected the manuscript to begin to draw relationships between the (modelled) stratigraphy, the spatial structure, and the physical processes, to provide some hypotheses for future observational and/or modelling work.

The manuscript needs to be clearer about the link between the simulated results and reality. This could be achieved by presenting further quantitative statistics from the observations presented, as well as attempting to validate against microstructural observations.  Similarly, the manuscript needs to provide more commentary on whether the patterns shown in the simulations are likely to be real or not, referring to available observational studies.

From a methodological point of view, the manuscript needs more discussion on differences between the FSM2 canopy model with what is implemented in FSMCRO, the reasons for the trade-offs, and discussion of the potential impact of these differences on the simulations. Also, while not the focus of the paper, the difficulties encountered when attempting to couple the models at the snow surface are mentioned, and it would be insightful to briefly expand on some of the issues encountered that led to the choice to develop a 0-layer model instead.

The paper should make a valuable contribution to The Cryosphere with revision.  

Specific Comments

Ln 108 - either in the methods of discussion section, it would be useful to reflect on how much the results depend on the specific model choices, noting that there are some subjective choices here,

Ln 148 – perhaps add “hereafter referred to as "FSM2" after (FSM2.0.3) to distinguish the enhanced canopy model from the standard FSM2 model – see next comment.

Ln 159 – “The model has so far been used for research purposes” - the original FSM2 has been used in many research and operational applications - make it clear you mean the canopy version here.

Ln 205 – please provide a short commentary in the methods section on which methods remain the same as FSM2 and the extent to which others have been modified. A table would be a very handy reference for the reader.

Ln 242 – 254 – it would help the reader if the numbering and ordering aligned with the order that results are presented.

Ln 267-272 – this largely repeats the preceding section (2.2.3) and could be removed or combined with the above.

Ln 283 – it would be useful to report some basic quantitative stats from the validation here (e.g., overall bias, RMSE, R, CV) to give confidence in this application.

Ln 290 – while it is understandable that the irregular and uncertain location of snowpit observations may limit a full quantitative evaluation of the FSMCRO simulations, it would be instructive to present some of the observations here if only to highlight the shortcomings of the available observations, motivate hypotheses that could be interrogated with FSMCRO and comment on how these may be validated with new observations. Not including observation of microstructure substantially reduces the readers confidence that new model system is simulating real patterns.

Ln 305 – “formation of surface melt forms/crusts (red) happens ca. 10 days earlier under-canopy than in the canopy gap” – this is not immediately clear from the figure – please add the dates to show the specific period intended.

Ln 343 – here and elsewhere (including Ln 377 and figures) it would be easier for readers in both northern and southern hemispheres if 'sun-exposed edge' and 'shaded edge' were used in place of 'south-exposed edge' and 'north-exposed edge’. Either way, please be consistent throughout the text and figures with the terms used (e.g., next sentence has ‘sun-exposed edge’, figures have ‘n-facing’)

Ln 351 – “does not overestimate the variability of snow stratigraphy.” please be specific - do you mean that in the accumulation season, vertical variability is large, whereas in the ablation season, horizontal variability is large? If so, please state this.

Ln 363 – “In contrast, snow depth variability between ensemble members at each location is in the same order of magnitude as differences between the two locations.” does this mean that the structural differences are more likely real? and that the snow depth differences are not? or just that the model behaves in the same way for the same forcing? Please comment.

Ln 383 – “This finding provides strong evidence of the substantial impact of canopy structural heterogeneity on modelled snow stratigraphy, suggesting that the resulting variability by far exceeds model uncertainty.” - was the ensemble system was validated against forested as well as open-site locations? this would be needed to conclude that model uncertainty is fully captured by the ensemble, and thus that model uncertainty is less than the explicitly resolved spatial  variability.

Ln 395 - is there indirect ways to validate these sorts of results - e.g. surface temperature from thermal imaging?

Ln 414 – “snowfall events” – please give dates or use annotations on figure to highlight period being referred to.

Ln 415 “co-exist at the surface” – again please be specific about what periods are being referred to.

Ln 415 “The ensemble thus does not capture variable metamorphism rates that are tightly linked to specific canopy structure” – would we expect it to? Please comment.

Ln 435-438 – a description of the issues encountered when trying to fully couple the models is not provided, so it is hard to assess how the result of Cristea et al.2022 are relevant here, and what we are to learn from these comments. Do you mean that these pitfalls should be documented in future studies? Or that this justifies your choice of a zero-layer approach? Or that Crocus is free from these issues. Please revise.

Ln 444 – please briefly outline the reasons provided by Nousa et al, (2023) in the text.

Ln 452 – further discussion on the limitations of the system is needed. Especially the potential impacts of trade-offs made in the zero-layer implementation (e.g. non-interactive canopy temperature – snow surface temperature – snowpack feedback).

Ln 464 – expected more analysis of how to quantitatively assess multi-dimensional data - e.g. evolution of CV of different parameters over time.

Ln 464 – expected the manuscript to begin to draw relationships between the (modelled) stratigraphy, the spatial structure, and the physical processes, to provide some hypotheses for future observational and/or modelling work.

Ln 473 - “The second mechanism [insolation-microstructure relationships] has, to our knowledge, not been captured by any simulation prior to this study.”- please comment on whether these patterns have been observed, and if so, provide references.

Ln 577 – as per comment on Ln 205 – would be useful either here or in methods section to have a list or table of which schemes are identical, which are slightly, modified and which required larger modification.

Technical corrections

Ln 129 – please provide a reference for the standalone version of Crocus.

Ln 130 – please explain what SVS2 is, as the reference (Vionnet et al, 2022) does not mention it.

Ln 173 – add ‘incoming’ before ‘short-’

Ln 288 – ‘FMI’ - please expand the acronym.

Ln 299 – ‘SSA’ please expand first use of acronym.

Ln 315 – “discontinuous forest transect” -> “transect through discontinuous forest”. The former implies the transect is discontinuous.

Ln 323 – “peak of winter” -> “peak winter accumulation”