Monteiro and Morin provide an in-depth analysis of snow cover and its main drivers temperature and precipitation in regional and global reanalyses and two climate models over the European Alps. They use in-situ observations and MODIS for snow cover, and state-of-the art spatial data sets for temperature and precipitation. It’s the most extensive evaluation over the European Alps, both in terms of models and reference data, and highly relevant to understand the potential and limitations of reanalyses (and of the two climate models) for modeling and observing snow cover information in complex terrain. The authors produced a wealth of information and the manuscript is well written  with great figures. 

The manuscript is of high relevance and technically sound. I have only a few general remarks, which are related to the author’s use of diverse data sets at different spatial resolutions and associated issues in complex Alpine terrain. Looking forward to seeing the revised version and the author’s answers. (I’m sorry for the length of the review, but the author’s paper is also quite extensive.)

General comments

    • Choice of elev bands: did you carry out a sensitivity analysis? E.g., there is the break between elev mismatch from 600m to 900m, partly from 1500 to 1800 (Fig 3). Is there any influence of topographic smoothing in models in your findings, or did you circumvent this by only taking elevation averages? 

    • All products are evaluated at their native resolution, right? Does the horizontal resolution have an influence on elevational patterns? Is there a dependency between horizontal and vertical variability?

    • (optional) To reduce the length of the paper, the authors could also move some elements from the main manuscript into a supplementary (and also from the appendix to a supplement). A personally biased suggestion is: Figs 11-14, moving window analysis.

    • Using spatial variability is problematic in the author’s case, because of the different horizontal resolutions. Usually, with coarser spatial resolution, the spatial variability (and any measure such as the standard deviation) decreases, so apparent differences between spatial variability are likely artifacts from using different resolutions. Also I’m skeptical about comparing spatial variability between sparse points and grids – without first checking this issue. This also relates to the the use of Taylor diagrams with the reference being station data, which I’m not sure is technically sound. 

    • At times, discussion and results are mixed. Also some methods are mentioned the first time in results. Note that a short repetition might still be useful, given the many different periods and methods used.

Minor comments

    • (optional) I found the title rather uninformative given the huge amount of content of the paper. I guess it’s hard to synthesize such a broad assessment, but the authors might think of adjusting the title to reflect their study better (possible keywords that come to mind are reanalyses, (climate) model, observations, elevation, evaluation, ...)

    • General: The authors use the ellipsis “…” extensively, especially in parentheses, but this is bad practice, since it leaves readers wondering what … could mean. The authors should consider rewriting these in some way. A suggestion (not necessarily the best) could be using other sentence structures, such as L41: “climate variables, such as temperature, precip, SCA, and SWE, are needed”

    • L6: Attention, kriging a special form of spatial interpolation and EOBS and LAPrec have not (solely) used kriging. Better use another wording for observation based spatial data sets.

    • L38-40: “wide range of changes...”, “some changes…” : vague, please clarify

    • L94: Please check duplicated sentence parts.

    • L105: Check missing sentence.

    • Fig1: high elev points in b-d are hard to see. Maybe remove the white component of high elev.

    • Sec 2.2. is more related to remote sensing, not “variables of interest”

    • L152ff unclear which years and which months were used in which analysis. Also missing values were allowed in the first case?

    • L206 and Fig2: CHTESSEL vs HTESSEL for ERA5?

    • Fig 3a: What is UE (Alps)? 

    • Sec 2.4.3: How did you define SOD, SMOD, and SCD? Did you take a full hydrological year? The same period as in the in-situ obs? Did you deal with intermittent snow episodes at start and end, and if yes how? 

    • Snow depth threshold: The problem with your choice of skill scores is that it is heavily influenced by the large amount of values with absence of snow cover (caused by low elev and – I assume, because it’s not clear from your methods – using the full year). This is why accuracy basically is the same as TNR. You might want to consider using PPV and NPV (positive, negative predictive value), since they adjust for prevalence (or alternatively limit the evaluation to less months). You can also consider dropping FPR and FNR, since they are equivalent to TPR and TNR (1 - *). Otherwise I would not put too much effort in the snow depth threshold, since it’s likely the least influencing factor in your study.

    • L319ff, and fig 4b: unclear what difference you are looking at (between elev classes? between MODIS and in-situ?) and what the standard deviation refers to.

    • L346: AR1 is used if you have lagged effects (carryovers between years), which I assume to be negligible for annual values and the studied variables. So I do not see how it should help with heteroscedasticity? Also unclear which trend estimates you show in the end?

    • L356: I could not find any supplementary material only appendix.

    • L450: Not sure. Fig A3 is very similar to Fig 8ghi, and inconclusive summer bias is only for low elev – at mid and high, it’s still a positive bias overall models.

    • L453ff: Is undercatch also a problem in summer at mid to high elev? I’m thinking strong winds and convective events? Otherwise, how do you explain that summer biases at mid to high elev are in the same range or even higher than in winter?

    • Fig9: Why do the numbers of grid cells not agree with previous figures? Most striking is ERA5 and ERA5-Crocus at 600m, but also the others are different.

    • L502: Spatial variability difference is related to different resolutions. If you aggregate MODIS to the other resolutions, the spatial variability should decrease.

    • L507: Might be related to different station densities across the domain. Did you do a regional analysis also for this? Would be interesting.

    • L528: So the Taylor diagrams are based on the 1985-2015 period, or on the full? If only for 1985-2015, wouldn’t it make more sense to put them in one of the sections before? (please also take into account the major comment above related to spatial variability)

    • Fig. 11-14: Very hard to read since strong overlaps. Also it is unclear why here you put regional plots instead of an elevational analysis like before (to me, there are little differences between regions in these figures)

    • L586: Is ERA5 really assimilating in-situ snow depth? I thought this was not case. 

    • Sec 3.3.2. 

        ◦ 1) Unclear, how you estimate trends at the seasonal scale from monthly values. This violates the independence assumptions in the trend estimators. Why not first calculate seasonal means?

        ◦ 2) running window trends are often hard to interpret. Maybe a time series plot might be more informative (similar to Figs 11-14 but alps-wide average) and easier to read.

    • L653: So the differences are caused by the different LSMs and snow schemes?

    • Fig 17: what are the two different N numbers?

    • L760ff: Wouldn’t the identified cold bias also delay melt?

    • L824ff: Quite harsh a statement for most of the remote sensing people, though most climatologists would agree. Please consider rephrasing more diplomatically, since the relevance of short-term trend assessments is discussable. 

    • L853ff: I’m not sure the deficiency is always on the reanalyses and models. Observational data, particularly in the Alps, has its own deficiencies. EOBS, for instance, has a very low station density south of the Alps affecting the temperature signal. In addition, observational evidence on EDW in the Alps is sparse and contradicting (some studies also show negative EDW).

    • L881ff: Since you did not test these factors explicitly, you might want to consider rephrasing as potential causes that you identified.

    • Note that the HISTALP summary regionalization you used is mostly based on temperature, and especially the precip zones (and thus also snow) can differ quite strongly: https://www.zamg.ac.at/histalp/project/maps/gar_reg.php

This could be discussed when talking about regional results (especially precip and snow).

The manuscript compares temperature, precipitation, and snow values of the European Alps from several reanalysis and climate modeling products with remote sensing and in-situ data of the last few decades. The authors focused their extensive analysis on 4 large sub-regions and on three elevation band between 600 and 2400 m. The results show that the agreement varies widely and depends mainly on the data set or variable considered. Such analyses are an important base for many scientific investigations since more and more applications rely on such spatially gridded long-term datasets. Moreover, the study provides a very valuable overview on the strength and weaknesses of the currently available observation datasets and gridded modelling products.

The structure of the paper is clear, the text mostly concise and follows a obvious thread. The methods and results are nicely presented. Despite the many acronyms the study is relatively easy to read. Except for one major point, there are only minor issue listed below. Therefore, I suggest accepting the manuscript as soon as the following points, have been addressed:

Major:

The determination of SCD, SOD and SMOD is fully dependent on complete time series. According to 2.3.1 you still allowed 20% of less of missing data in the in-situ snow depth data set. How many of the used time series were affected by such gaps and how were the above snow cover related variables determined in such cases? What was the procedure, if two or more periods of the same length were detected?

Minor:

L11: Please add the information that the presented results are mainly valid for the November to April period.

L20: Since such a sentence does not really provide any meaningful information, certainly not in the summary, delete it or be more specific.

L36: Please provide reference for “5 days per decade”

L37: Please provide an example for “altering the magnitude of natural hazard” in the winter season.

L94: “contrasting results” twice?

L105: Missing clause!

L106: I do not understand this last sentence. Please rephrase.

L159: Please shortly mention, why only the 2000-2015 period was considered.

Fig.2: Yellow color is hard to read. I’d suggest adding (e.g. right of the color box) the final spatial and temporal resolution of each product.

L302: Please rephrase

L217: You mention several NH snow analyses, but reference only one, which is about arctic snow only?

L275: Please shortly mention what the reference is.

L287: Last section of what?

Fig.3: What is UE Alps?

L355:”..there are not much differences in the results at the sub-regional scale”. This contradicts the results of Matiu et al. (2021). Please explain.

L369: “normalized mean error”. In the same sentence you also use “relative mean error”?

Fig.5: I do not see circle markers? Figure caption: Sub figures a,b,c labeled wrong order.

Fig.6: Please add units to the different column heads. Figure caption: What about the unit of the scores of temperature?

Fig.8: Labels of x- and y-axis of right side figures are too small.

L453: Please rephrase

L457: “…could be nuanced”? Please rephrase

L495: “…consistent with boxplot”? Only if we assume that the negative whisker is caused by the low elevation pixels. Please explain.

Fig.9: Caption: “In order to compared..”?

L503: “..underestimated in all datasets..” Where to see?

            “excepted”?

Fig.10: The order of a,b,c (a for the lowest plot) is different from all the other figures, but I like it more.

L537: “Solid line with circle markers…” should be part of the figure caption.

            What is the “geographical mean value”?

Fig.11: What is the reason to show to markers?

            The different markers can not be separated from each other.

            For comparability I’d suggest to use the same axis scale for all four Taylor diagrams.

Fig. 12: What is the reason for the large difference between the red and blue line for the second last year?

L572: “lower”

L582: “anomaly values”

L599: “..precipitation for the 1950-2020 period”

L640: “ over the last decade” only?

Fig.15: “Winter (Nov-Apr) trend values…”

Fig.17: What is the meaning of the two N-values for each dataset?

            N-values are too small!

Fig.17: Caption: “  winter (Nov-Apr) trend”

            “grid points included for three elevation bands (+/- 150 m) given on the y-axis.”