Reply on RC2

As a general comment, this manuscript lacks a clear explanation of to what extent the method used is novel. The interest and aims of the method are clearly described, but its novelty and added-value (compared to Fiffes and Gruber, 2012, 2014, but also to previous similar downscaling and bias-adjustment methods) is not. A corollary comment is the fact that the method and results would benefit from being put into a larger context. For instance, results in section 5.3 should be compared to other studies of the impact of climate change on Alpine snow cover (which would also better fit with the current title of section 5.3 about "Alpine" snow cover, while currently only Swiss snow cover is discussed).

As a general comment, this manuscript lacks a clear explanation of to what extent the method used is novel. The interest and aims of the method are clearly described, but its novelty and added-value (compared to Fiffes and Gruber, 2012, 2014, but also to previous similar downscaling and bias-adjustment methods) is not. A corollary comment is the fact that the method and results would benefit from being put into a larger context. For instance, results in section 5.3 should be compared to other studies of the impact of climate change on Alpine snow cover (which would also better fit with the current title of section 5.3 about "Alpine" snow cover, while currently only Swiss snow cover is discussed). The main novelty is that a modular system to generate high-resolution (slope scale) forcing data for impact studies is developed, documented and provided. Even though the individual components (TopoSUB, TopoSCALE, and quantile mapping) are perhaps well established, establishing the coupled workflow is novel and generates important model forcings that are currently, to our knowledge, not available to impact modellers. In particular, it allows us to get climate impact projections for the cryosphere (in this case snow) at the hillslope scale by feeding the bias corrected future atmospheric forcing through cryospheric models (in this case FSM). Perhaps most importantly, especially for applications in data scarce regions, this scheme does not require in situ data. The hillslope scale (order 100 m resolution) is often not addressed in previous climate change studies despite its importance in regulating the stores and fluxes of water, energy, and carbon (e.g.: Fan et al., 2019,https://doi.org/10.1029/2018WR023903 ), precisely because of the lack of forcing availability at this scale.

Useful references include (but
In particular we have made the following edits in the introduction to make these points clearer: This is especially the case in heterogeneous terrain such as mountain regions where topographic variability is high over short horizontal distances. High surface variability requires modelling at the hill slope scale (c.100 m) in order to adequately capture fluxes and stores of energy, water and carbon (Fan et al. 2019). Various methods of downscaling can be utilised to achieve this goal.
In this study we address the problem of impact-model ready (i.e. hillslope scale) climate timeseries with a new modelling framework called "TopoCLIM".
Importantly, using these pseudo-observations we are able to debias climate timeseries in regions lacking ground observations.
Based on this and Reviewer 3 comments we have adjusted Section 5.3 title to a more appropriate: Climate change impacts on Alpine snow cover across Switzerland In Section 5.3 we add discussion on previous studies while making the point that they are not comparable to our results in a straightforward manner due to resolution ( 25km v 100 m), parent climate models (ENSEMBLES v CORDEX) and/or scenarios (ie SRES v RCPs) used. However, these previous studies do nicely highlight the gap we try to fill with this work: Several previous studies have investigated the impacts of climate change upon Alpine snow cover (e.g. Steger et al. 2012, Marty et al. 2017, Frei et al. 2018, Verfaillie et al. 2018, Bender et al. 2020, however direct comparison is often problematic due to model resolution, analysis period, parent climate models and/or emissions scenarios used. This highlights the importance of model intercomparison studies whereby these important variables controlling model results can be standardised. Comparison to these previous works highlights the contribution of this study in that most were conducted at RCM resolution of 25 -12 km (Steger et al. 2012, Frei et al. 2018 or local scale (Verfaillie et al. 2018, Bender et al. 2020 or reliant on in situ data (Marty et al. 2017, Bender et al. 2020. In this study, we demonstrate a method that generates results over large modelling domains at hillslope scale (100 m), a scale which is extremely important in regulating the stores and fluxes of water, energy, and carbon (Fan et al., 2019), and therefore critical to modelling snow cover in mountainous terrain. Additionally, this approach does not rely on in situ data and therefore is appropriate for data-scarce regions.

Corrected these and other occurrences.
Line 52: repetition of "the". Thank you for this reference which we now have included in this section.

Lines 47-58: to me, a key reference for quantile mapping that should be cited
Line 92: the "required forcing variables" are not listed here. In fact, they are listed in Table 1, but Table 1

is never called in the text…
We added the reference to the section describing the climate data, L.178-79 now reads: "A full description of CORDEX variables is given in Table 1 and model chains used in Table  2." Line 98: this sentence seems incomplete. Please edit.

Edited to:
For example, during the conversion from a "360-day" to a "standard" calendar, the output from the linear scaling will result in a 365 day timeseries (in the case of non-leap year) and be missing the following dates: January 31st, March 31st, June 1st, July 31st, September 30th and November 30th.

Line 116: "IMIS" should be defined here instead of line 203-204.
Defined now at l. 116 as suggested. Edited both lines for consistency.

3). Could you please explain if you found a way to circumvent this issue (in section 2.4) and the related uncertainties (in section 5)?
Yes this is correct, but variables are corrected towards a physically consistent dataset in the form of downscaled ERA5 data, so we argue that while the method is univariate it will not produce physically inconsistent results. The validation during the current climate also supports this claim (Table 3 -5). This is now stated explicitly in 2.4 as: "It should be noted that while the variables are bias corrected independently, they are corrected towards a physically consistent dataset in the form of downscaled ERA5 data, so we argue that while the method is univariate it does not produce physically inconsistent results. The validation during the current climate also supports this claim (Table 3-
elsewhere.  They are indeed identical datasets, however, due to the png output resolution, the output device (x11) appeared to do some sub-pixel interpolation that gave the slight appearance of slightly different datasets. The .1 and .2 was an artefact of plotting the same dataset twice in R package RasterVis (levelplot function) -the default naming scheme appends a .1 and .2 in this case. We have replotted at a higher resolution which has removed this artefact and we have corrected the panel titles.