This is a worthwhile and well-written paper (although it will benefit from some copy editing). I only have a few minor comments.

Thank you very much for taking the time to review our manuscript. We are grateful for your comments and suggestions. 

83 Resampling the DEM to coarser resolution could be taking elevations of the 10 m cells closest to the centre of the coarser resolution grid cells. Please confirm if this is the case. Block averaging would be more representative of the inputs to coarser-resolution models and would, I expect, give larger differences between simulations at different resolutions. Related to that, how are the 1 km meteorological variables downscaled?

The original DEM provided by the administration of the Berchtesgaden National Park had a resolution of 10 m. We aggregate this raster file to courser resolutions calculating the average value of the 10 m cells within the coarser resolution cell. In the case of the 50 x 50m DEM, for example, the cell values are averages of the twenty-five 10 m cells comprising each 50 x 50 m cell. Yes, it also would be possible to take the value of the 10m cell closest to the centre of the courser grid cells. We will keep this a an option in mind for future studies, and we will consider the way coarser resolution DEMs are generated another potential source of uncertainty. To run the snow simulations we used scattered point measurements. These station recordings are interpolated to each grid cell of the modelling domain. A detailed description of the elevation- and station-distance-dependent interpolated field is presented in section 3.3 in Strasser et al. 2024 (https://doi.org/10.5194/gmd-17-6775-2024). The exact parametrization including the lapse-rates used to obtain the 'regression fields' for each time step is available in the configuration files of the model set-up published along the manuscript (https://doi.org/10.23728/B2SHARE.530A7560A73647459969F5C21639E8CB). 

112 The statement “The incorporation of a radiation-driven melt component leads to seasonally and spatially varying snowmelt factors” is immediately followed by invariant values.

Thank you for pointing at this. Indeed, our description here is not very precise and can be misleading. The degree-day factor (DDF) and the radiation factor (RF) that are used within the enhanced T-Index approach are invariant. What we wanted to point out is that the radiation-driven component of the approach, which we consider the to be (1-RF) x G (with G being radiation) supports seasonally as well as spatially varying snowmelt dynamics in the study area. Via the incorporation of the radiation, this component of the enhanced T-Index approach addresses exposition (e.g. south vs. north), shadowing of neighbouring mountains as well as a general seasonal cycle. We suggest to change the corresponding sentence to: 'The incorporation of a radiation-driven component supports seasonally and spatially varying snowmelt dynamics [...]'.

255 The authors caution against using the stochastic climate generator to produce climate change scenarios. Moreover, it will not produce expected elevation- and season-dependent changes.

We agree that this is a limitation of the climate generator and that this should be explicitly mentioned. We will add this into the discussion section of the manuscript: 'Moreover, the climate generator cannot adequately produce elevation- and season-dependent features of long-term climatic changes.'

263 Although km-scale atmospheric models are commonly referred to as “convection permitting”, better resolution of topographic forcing of precipitation might be of more significance here.

We will update the corresponding sentence: 'These high resolution climate model runs better capture the topographic forcing of precipitation and allow for an explicit, physical description of deep convection without having to use parametrization schemes.

292 “spatial resolutions considerably below 1 km are required” is not an unexpected conclusion, but how is this shown by the comparison of model simulations?

We draw this conclusion from the fact that the results of SCD and SDD for elevation bands in the Berchtesgaden National Park based on the coarser model runs (particularly runs with spatial resolutions of 500 and 1000 m) often differ considerably from results from higher resolution snow model runs, in particular, at the high elevations (see Fig. 7 in the manuscript). In a setting such as the BGNP, a change in model resolution from 100 to 1000 m can cause strong changes in the results of SCD and SDD for elevation bands. We suggest that for the BGNP, results only are robust for resolutions considerably below 1 km.

This manuscript presents the assessment of snow model uncertainty in relation to the effect of a 1◦C warming using the snow modelling framework openAMUNDSEN. This is a well written manuscript and the content of the manuscript is of weighted significance in cryosphere modeling community as this paper investigates the uncertainty induced by the selection of the snow model configuration with degree-day as well as physically based snowmelt methods and varying land cover maps and spatial resolutions. I would recommend to accept the paper for publication after minor revisions as per the specific comments given below.

Thank you very much for taking the time to review our manuscript. We are grateful for your comments and suggestions. 

1. Line 114 model runs at 3-hourly time scale whereas line 107, the model runs at daily resolution. Please clarify.

In this study, we apply three different snowmelt approaches: 1) T-Index, 2) Enhanced T-Index and 3) Energy Balance. The T-index snow model runs are conducted in daily time steps. The Enhanced T-Index and the Energy Balance simulations are conducted in 3-hourly time steps. According to our knowledge, these are temporal resolutions widely used in the community for these type of distributed snow simulations.

2. Line 120, please clarify combined lapse rate/inverse distance weighting scheme.}

To conduct fully distributed snow simulations, we use the meteorological pre-processor for the spatial interpolation of scattered point measurements integrated in the openAMUNDSEN model. A detailed description of the interpolation scheme is provided in section 3.3 in Strasser et al. 2024 (https://doi.org/10.5194/gmd-17-6775-2024). We will add more details into the method section: 'First, lapse rates are used to generate a distributed elevation field for each meteorological variable (i.e., the regression field). Next, residuals between the regression field and the station locations are calculated and interpolated to a grid using an inverse distance weighting (IDW) method (= residual field). The overlay of the regression with the residual field results in the final meteorological input grid. This interpolation scheme is applied in each model time step.' 

3. openAMUNDSEN models snow-canopy interactions. Such interaction modeling gives SWE over canopy which would be evaluated with satellite derived snow cover fraction in the forest region too. Please consider it.

Thank you for pointing at this. The evaluation of snow simulations in forest areas indeed is a very important aspect. We agree that more research is required to improve the evaluation of snow modelling results in forest areas. According to our information, Sentinel-2 products typically provide top-of-canopy fractional snow cover (FSC) which potentially can be used to evaluate the presence of snow inside the snow interception storage of the model. Additional investigations into this aspect of the model evaluation are very interesting, however, seem to be beyond the scope of this study.

4. Please mention that the model evaluation mentioned at lines 176-181 is valid for non-forest area (canopy free region) only.

We will add this information to the results section: 'The evaluation of the openAMUNDSEN snow simulations in non-forested areas [..]' and make sure that in the captions of the corresponding figures it is clearly mentioned that this evaluation step is using non-forested cells only. An evaluation of snow simulation in the open and inside the forests of the BGNP and a more detailed description of the snow canopy interaction is presented in Storebakken et al. 2025 (https://doi.org/10.1002/hyp.70197).

5. Figure 4-8 are for the entire BGPN area or non-forest area of BGPN. Please clarify.

Yes, in Fig. 4--8, we show the results for the entire BGNP. One aspect that we want investigate is how strong snow model results differ depending on what land cover map (in particular the forest representation) is used in combination with what snowmelt approach.

6. What about the sensitivity of albedo value in snow melt simulation?

Yes, we expect that snow modeling results also are sensitive to the snow albedo parametrization used. The snow albedo and its decay function only is used in the Enhanced T-Index and Energy Balance snowmelt simulations. A detailed description of the air-temperature-dependent decay function of snow albedo as available in openAMUNDSEN is presented in section 3.4 of Strasser et. al. 2024 (https://doi.org/10.5194/gmd-17-6775-2024). In this study, we use the default values for the maximum and minimum snow albedo values and a temperature dependent recession factor.