The overall aim of this paper is to improve knowledge about snow melt-out days (SMOD) trends at mountain range scale. The authors produced a new set of combined high resolution satellite data for the French Alps and the Pyrenees, they assessed its quality and its relevance for SMOD trends study. They evaluated the trends of SMOD from 1986 to 2023. They highligthed that the new set of data is relevant for such study and that the trends are more consistent for the French Alps than for the Pyrenees.

General comments

Reading this paper, I had the feeling that the focus was on the dataset and that the trends of SMOD for the Pyrenees and French Alps were more an application than the goal of the study. Indeed, explanations were extensive regarding how to built the dataset, how to sample it, how to evaluate it. Later in the text, trends of SMOD are described as the second objective of this study. This is not clearly express in the introduction, neither in the title. In this way, I think that the title of the paper is misleading and the article need some reorganisation. The extensive work accomplished by the authors to assemble the high resolution satellite dataset should be given greater prominence. Here are some suggestions to improve this point.

Introduction. The objective of the paper does not appear clearly. Please add more information about it in the introduction. The link between second and third paragraph (L32-L33) could be improved. For example, presenting first the different options that can be use for SMOD trends (numerical modeling, in situ observations and satellite data), as it is done at L46-56, and once the advantages of satellite data highlighted, describe the available products (L56-84). This can now lead to your interest in merging different sources of data to create the most suitable dataset for studing SMOD trends. Moreover, last paragraph already presents some methodology and the studied domains with details, which, in my point of view, should be in the Method section. The introduction could also end with an overview of the structure of the paper.

Data and methods. I think that some reorganisation would be beneficial for a better understanding. In the result part, two sets of data are considered: SWHLX and Theia. My suggestion would be to divide section 2.1 in two parts: SWHLX (2.1.1) and Theia (2.1.2). In 2.1.1, you could present SWH and DLR-Landsat, the main conclusions of your evaluation, and finally the SWHLX dataset. Depending our main objective (SMOD trends or building a dataset and use it for SMOD trends), you could put in Annexe the method and results of the evaluation (in the present version: 3.1 and 4.1) or let it in the 2.1.1.

Specific comments

Some methodology aspects could use some clarifications. Here is a list of some specific points.

• Why did you select this specific tile (31TCH) and HY (2017)? Is it a method usually applied in such study?
• In Fig. 4, no example are shown for an odd number of days without data between a no-snow and a snow day. Can you elaborate on this point, please?
• Two stations were selected only for their data availability, and no more information were provide about their characteristics. Is there any characteristics of these stations that can explain the absence of a specific biais between in situ and satellite data?
• Significance of the trends being depending on the selected period, did you experiment trends on 25 or 30 years (30 years being the WMO and good pratice in climate)?
• Why the absence of in situ stations in the Spanish Pyrenees is not more discussed?
• Why no percentage of pixel per altitude per massif is provide? This could have lead to more analysis on the differences in the trends per massif. In the same state of mind, why DAH wasn’t more investigated?

In general, captions for figures are a little dry in the main article, whereas too many information (like explainations and comments) are present in the captions of figures in Annexes. For example, Fig. 2 could be more explicitly described (proportion of available data Pyrenees vs Alps; specificities for each mountain range regarding the number of observation as a fonction of altitude; why about all the southern part of Pyrenees doesn’t have selected in situ data).

L24: Is it «air temperature» or «near-surface air temperature»?

L49: Please add «in the European Alps» after Monteiro and Morin (2023).

L73: Please explicit the SPOT accronym here instead of L74.

L96: For more clarity between satellite data and providing centers, please change «Theia (Sentinel-2A&B and Landsat 8)» for «Theia L2-Snow Product (Sentinel-2A&B and Landsat 8), hereafter Theia»

L99: Please explicit «NIR».

L119: Please explicit «SCA».

L129: Please add «(Fig. 2)» at the end of the first sentence.

L144: Please change «SWHLX3» for «SWHLX».

L199: Please explicit NOBS.

L216: Please change «Mann–Kendall (MK)» for «MK».

L299: There is a missing space between «Ossau.» and «However».

L304: Please change «Pyrenees 12» for «Pyrenees (Fig. 12)».

L428: Please change «(Barrou Dumont et al., 2024b)» for «Barrou Dumont et al. (2024b)».

Fig.1: Adding the borders can be useful as you give information about the Pyrenees or the French Pyrenees. This figure can be presented in the in situ data section, instead of the introduction.

Fig.2: Please use the same legend as in Fig.1 for mountains ranges, and explicit that « o » symbol corresponds to stations. Are the barplots stacked or superposed?

Fig. 3: It could be refered at the end of the introduction of section 2.1, in order to give an overview of the satellite products considered in the study.

Fig. 5: Is Alpe d’Huez all the green square (if so, remove the texte from SPOT-Landsat figure and add it on the top figure) or does it refer to a specific location (if so, please use a symbol to locate Alpe d’Huez and put it on the three figures)?

Fig. 6: Left and right figures present the same information for Theia. Please remove the left figure, and put statistic information in a table or near the figure specifying which stats are for which dataset (Theia or Theia+SWHLX). Why the statistics for SWHLX alone are not presented?

Fig. C1,2,3: Please use a date spelling like «1 February 2023» (British English) or «February 1, 2023» (American English), to avoid confusion about the date.

Fig. C4: Please correct the caption with ΔSMOD instead of SMOD.

Fig. C5: Why the legend only includes few years, whereas more are presented in the figure?

Fig. C6: The color used fot NOBSmin=10 is almost invisible.

This study uses multi-source datasets to reveal snow melting day over the long time series. Employing various methods, it improves the spatial-temporal resolution of snow melting day to support the study results. Authors have made a comparison between SWH and landsat, but I am also worried about the data consistency between different landsat sensors, and between landsat and sentinel-2, especially under the case with a long study period and the mixture of all available data. So my suggestion is to add some data validation before using satellite datasets to calculate snow melting day.

Line 110, when you combined the landsat data from different sensors over the long time series, how did you consider the sensor correction or data consistency between products from various sensors? In addition, can you give a brief description regarding the algorithm to produce level 2B product?

Line 115, before mixing sentinel-2 and landsat-8 data from 2016 to 2023, I miss the consistency check between both datasets or sensor correction because they are from different sensors. Only if the data from different sensors are proven to be consistent, the trends calculated later will be meaningful.

Line 183, what is the basis to set the threshold of tree cover 50%?

Line 199, what is NOBS?

Check the figure 5. The left sub-figure is landsat only or landsat+spot?

Fig. 9 and Fig. 10, how about mapping pixel-wise? One of advantages of this study is high spatial resolution.

Line 322, I am not sure how tree cover density correlates to elevation? Maybe the changes in bias with elevation could be partly explained by tree cover density. How about dividing tree cover density into several classes? Then analyze something, such as the bias and trend.