General comments

This manuscript presents a model chain for producing regional-scale avalanche danger predictions in Switzerland. The key contribution extends a point-scale danger model (Pérez-Guillén et al., 2022) to a regional scale by interpolating across a continuous grid and aggregating within predefined regions. The interpolation and aggregation methods aim to capture relevant processes that influence avalanche danger while aligning with approaches used by human forecasters so that the model chain can be applied as a decision-support tool. The presentation of this model chain is an interesting topic that fits well within the scope of GMD.

The manuscript does an excellent job of communicating a complex topic with clarity and logical progression. It establishes clear objectives, employs sound methodological choices, and draws fair and relevant conclusions applicable to operational avalanche forecasting. I think a few details could be further clarified (explained below), but otherwise recommend the publication of this manuscript.

Specific comments

• Representativeness of the stations. Providing additional information about the stations and snowpack simulations would help readers understand how effectively the training data represents the variability of avalanche conditions within a region. While Pérez-Guillén et al. (2022) likely address some of these details, including more information would offer valuable insights. For example, the number of stations in the dataset, the nature of the simulations (flat field and/or virtual slopes), and whether wind transport was simulated. Without such details, it remains unclear how well the stations capture the full range of expected conditions within each region and how this might impact the resulting predictions. Can we expect this method to predict the most unstable slopes in a region and if not does this create a bias? How well can the interpolation routine capture snowpack conditions not represented in the input data?
• Terrain features. The selection of terrain features for the interpolation routine should be explained in more detail. It is not entirely clear which features are derived from the DSM, nor is the meaning of directional derivatives, difference of Gaussians, and Gaussian pyramids (lines 137 to 141). Some plain-language explanations of what these derived variables are and how they potentially relate to avalanche danger would help. Also, some clarification is needed regarding the interpretation of slope angle, curvature, and aspect at the coarse scale of 32 km², and why these are expected to be relevant. Further explaining the terrain variables would provide readers with important context for interpreting the results.
• The methods section (Sect. 4) has extensive use of mathematical symbols, some of which may be excessive and cause confusion rather than clarity. This is simply a personal preference, but I think it would be clearer to use more plain language and then use symbols strategically where it helps communicate mathematical relationships. Also please check all symbols are unique and defined (e.g., alpha is used differently in line 251 vs alpha in line 281, Ne in line 263 is not defined).
• I agree with the approach to evaluating performance with mean/median accuracy, however, am curious if there were any directional biases in the model in terms of over or under-predicting danger (e.g., for specific regions or danger levels). While this doesn’t need to be fully presented, it would be interesting to comment if this was investigated.

Technical comments

• Line 66: “built” not “build”.
• 1: The IMIS and ZERO-DL station networks are not defined/described anywhere in the manuscript.
• Line 109-11: Data extraction times are unclear. Public forecasts are valid until 17 LT, snow cover data is extracted at 12 LT, but then why is resampled meteorological data centered around 18 LT? Wouldn’t it make sense for all data to be extracted at a single time?
• Line 121: Perhaps state the total dataset size (e.g., number of station-day-danger points).
• Line 130-140: It is not clear how extracting terrain features at a scale of 1 to 32 km² is capturing the smaller scale topographic properties you say influence avalanches at scales to tens to hundreds of metres. Did you derive slope angle, profile curvature, and aspect from the 25 m DSM and then upscale to coarser grids? Perhaps more details would clarify how terrain characteristics are being captured in the model.
• Line 140: A brief plain language description of the Gaussian pyramid technique would help.
• Fig 2. In the interpolation section, the “etc.” in terrain features is confusing as the methods only list location, elevation, slope angle, curvature, and aspect. Does “etc.” mean to capture the directional derivatives, DOG, and Gaussian pyramids?
• Sect 4.3: It is not clear that three distinct methods were tested (mean, top-alpha, bands). When reading it can be interpreted that top-alpha and band averaging are done in conjunction, rather than two distinct methods.
• Line 366-359: Perhaps I misunderstood the method, but I don’t see how the elevation bands overlap. I would have assumed when you increase the bandwidth you decrease the number of bands accordingly to avoid overlap. What is the motivation for allowing overlap?
• Line 356: Is the “mean method” defined or labelled anywhere? I think the meaning of this method is intuitive but slightly confusing if it is not explicitly defined/labelled anywhere.

The manuscript present a series of methods to extrapolate point computations of avalanche danger from Pérez-Guillén et al, 2022 over space and determine a avalanche hazard for all forecasting regions of Switzerland (the minimal units used by Swiss avalanche forecasters to produce bulletins on dynamical areas depending on the situation). The goal of the method is to produce an automatic forecast of the avalanche danger level on Switzerland from snow modelling operationally run on points (automatic weather stations). The paper is nevertheless limited to dry snow problems, while wet snow or mixed dry/wet snow avalanche problems may contribute to the overall hazard, but this is clearly acknowledged. The goal of the paper as well as the overall presentation is well suited for the readership of GMD.

The manuscript clearly present the methods, is quite well organized an easy to read and present interesting insights into scale changes for avalanche forecasting (from point scale to 1km grid and forecast regions). The evaluation seem quite complete with different scales treated (regional, global, daily or seasonal, by avalanche danger...) and give a great overview of advantages and drawbacks of the presented work. The provided code seem clear and usable. I have mainly minor comments that I detail below and that can be considered by the authors before final publication.

General comments:

Work of Pérez-Guillén do not have to be presented again, it is an input of your study and you can point to the published paper for details. However, you may give a focus on changes made from the published method. Sometimes you re-explain the model used by Pérez-Guillén which does not seem necessary to me. However, these parts are not sufficiently important to prevent general comprehension of the paper and added value of this work. I detail most useless parts in the detailed comments.

The spatial resolution chosen for extrapolation is a 1km grid. Coarser resolutions are tested and not selected. However, nothing is said of finer resolution whereas complex topographies in mountainous regions are known to be poorly represented at coarse resolutions. Authors then use advanced methods to compute topographic variables to reduce the impact of a coarse resolution (such as Gaussian Pyramids, which is of high interest) but never discuss why they selected a 1km resolution and if their model could be used at a finer resolution, which would be of interest for the reader and for further uses of such method.

The avalanche danger scale is not linear. Difference between level 3 and 4 is much higher than than difference between risk 1 and 2. Does this influence the results when computing expected danger level (equation 3) and how does this impact the different methods you used? This could also be the main reason explaining the mean method performs poorly in Fig. 5. I have seen no discussion of this important characteristic of the data you manipulate.

Some of the methods are presented in the results rather than in the material and methods section. For instance, we discover the partition in different areas in Fig.6 in the results or the presentation of F1 score that appear only in section 6.

Detailed comments:

Line 109-110: "from a more recent operational SNOWPACK version": please be specific and provide clearly the identification of the code used (release number or git tag or commit) here or in the code and data availability section. You can also briefly explain if there is major changes between yours and Pérez-Guillén version.

Section 4.1 and 4.2 may be rewritten more straightforwardly. Authors introduce a lot of mathematical notations that are not used elsewhere. In particular, the mathematical description of random forest seem to be out of the scope of this paper. You can directly refer to Pérez-Guillén et al., 2022 and/or Breiman, 2001.

On section 4.2, several sentences present generally the interpolation method. The reader may be helped by having a presentation of exactly what you do in the paper immediately after the introduction of each notion rather that keeping general ("One of the most popular and widely used kernel function" may be transformed as "we used the most popular kernel function which is...", same for "can refer to geographical location" or "one can construct kernels").

On Figure 3b, the big red dots are not informative and prevent for viewing the background data that is the result of your method especially in the Alps area. Maybe you can keep the dots but unfilled or reduce their size.

I am not sure I fully agree with the statement line 265 : "danger level for dry-snow avalanche increases with increasing elevation". Do you have data or references for that? For instance, situations with persistent weak-layers at mid-altitudes that are not present at higher altitudes are not so uncommon.

On the interpretation of Table 1: differences are very small between the different results. Do you have some clue to think that they can be significant? If yes, please provide and if no, you may underline the uncertainty in the interpretation (line 327-334).

Line 377: you use only one year for evaluation. As snow coverage can largely vary between years, how does this influence your results. In particular, I suspect that this may have a larger impact on small areas with few observations and a rather tight diversity of snowpacks such as the Jura area.

Figure 7 and 9: All the bars are not directly comparable as RF is evaluated on points and other on forecasting regions and the number of forecasting regions varies. It may be interesting to specify the number of regions/simulation points on these graphs.