Dear Authors, a timely and practical piece of work that will help guide folk who wish to examine the homogeneity of in situ  snow cover series.  My comments are manly minor and editorical in nature, and are included in the attached annotated pdf file.  My only criticism of the paper is the lack of independent testing of the three break point methods with synthetic series. Without this, one of the main conclusions to use multiple break point detection methods seems a little weak given that one method was shown to greatly over-identify break points. The authors make a point to avoid going very far in the correction side of the homogenization process, but this diminishes the interest factor.  For example, it would be interesting to show some preliminary metric of the impact of correcting series e.g. on trends in regionally-averaged snow cover series. The paper was well-written, easy to follow and a pleasure to review. Best regards, Ross D Brown, Canada.

The authors have performed homogeneity testing on snow series using three different methods on two different snow variables. The paper is well written with good structure and nicely presented results. It is a good addition to the scientific literature on homogenisation of snow depth series. However, I do think it would benefit the paper to include some results that shows the impacts of the homogenisation (difference between raw and homogenised series, trends…).

I agree with reviewer 1 about it not being clear why break detection depends on elevation. I suggest adding a version of what you answered reviewer 1 to ch. 4.1.3 or another suiting place in the paper to explain your motivation for looking into this.

Fig. 8 could use some refinement:

• What I understand from the figure text is that valid breakpoints detected by two methods are shown in grey but breaks only detected by one method are also shown in grey. This is confusing. And are breaks detected by only one method shown in the figure?
• In the legend I suggest using a color other than grey for the circle, square and rectangle (for acmant, climatol and homer) or another color than grey for dHS1, whichever is easiest.
• In addition, there is a “NA” in the legend that looks out of place.

Please see the attached pdf for the rest of my mostly technical and minor comments.

The paper is well written. The presentation is clear and well structured. It has been a pleasure to read it.

The paper presents a study assessing the performance of three different applications for identifying homogeneity breaks in snow depth time series. The analysis and results are well described and discussed. My main concern, which is not very large, is related to the comparison of the applications. There are obviously several subjective choices to take in applying both Homer and Climatol, and less in Acmant. One issue the authors mention, and which I recommend to pursue more, is the role of the reference network and how they are established, and how that influences the results. I have a feeling that this is as well an important issue as the break detection algorithm itself.  In order to have a fair, consistent comparison of the break detection capabilities  I would therefore challenge the authors to apply the three methods with the same reference networks. Would it be possible to apply them on all three combinations of networks?

Minor comments:

Line 27: Typo: contemporay → contemporary.

Figure 7: Difficult to read. Use a lighter shade for the terrain, and make the black dots smaller. Consider dark-grey contour lines (rivers, borders).