General comments

The study by Vérin and colleagues investigates snow metamorphism on the arctic sea ice and in particular how it impacts on spectral albedo, and how well the albedo variability can be reproduced using a RTM. I believe that this study provides great insights into the temporal variability of snow sea ice albedo during the melt season, and that the albedo dataset could be useful to parametrize or validate the evolution of snow sea ice albedo in global models. The results are clearly presented and the methods are robust - my comments are minor and mostly focused on the albedo measuring and modelling since snow metamorphism is out of my expertise.

Specific comments and technical corrections

Title:

Arctic marine snow can be confusing because it commonly refers to debris sinking in the ocean, maybe it would be clearer to indicate “Arctic snow on sea ice” or equivalent in the title?

Abstract:

Line 19-20: Maybe a minimum BBA value can be indicated here rather than the value at 1000nm? Then it makes it comparable with the number from line 18.

Line 24-25:  “based on measurements” is probably unnecessary here because “measured” is already written earlier in the sentence - maybe reformulate to “.. using measured SSA, density vertical profile and impurity content”?

Line 25: “calculationS” - also, “at the interface snow-ice” may be clearer than “at the top of the sea ice”?

Introduction:

Line 56-57: What about longwave radiation on overcast days – can this also have an impact on the dry snowpack?

Line 70: In Domine, F., et al. "Three examples where the specific surface area of snow increased over time." The Cryosphere 3.1 (2009): 31-39, the authors conclude that “SSA increases are probably not rare”, including cases that are described in this paragraph like depth hoar formation. Is this reference maybe worth mentioning here? (especially as other “occasional” processes are mentioned in the introduction, such as summer snowfalls in line 86.

Line 74: “samplings” is probably incorrect, maybe “a lot of samples” or “a lot of sampling”?

Line 93: global radiative transfer “in” sea ice not “of”?

Line 101: “light absorption” may be clearer than “optical absorption”?

Line 102: instead of “and using measurements on melted snow samples”, maybe “and measuring absorption coefficients of LAPs isolated from snow samples”?

Line 116: I think that the correct spelling is “set up”, not “setup”?

Line 157: LAPs are often distributed in the top cms – if the top 7cm are discarded, how is it possible to get data on the LAPs from the profiles? Or do the authors mean that the top 7cm was removed only from thin snowpacks on line 150, which is why thick snowpacks were used for retrieval of LAPs properties?

Line 180: Is it possible to indicate the reference of the electronic scale and the sensitivity?

Line 223: Is it possible to indicate at what time were the albedo measurements recorded?

 Line 196: “then” instead of “them”

Line 207: “As the snowpack was already

ripe, the study of spatial variability using large scale measurements was favored.” What do large scale measurements mean here – UAV measurements, or transect measurements?

Line 213: short wavelengthS

Line 238-240: Why were simulations performed considering diffuse radiation and not using the exact SZA values corresponding to each albedo measurement?

Line 252: “whose” to replace with “of which”?

Line 259-260: So density, SSA and irradiance profiles were not measured in 2016? Please indicate this in the methods when describing the sampling and analysis of snow physical properties (2.3 and 2.2.2)

Figure 3: How were the wavelengths of 500 and 1000 for albedo chosen? Why not calculate broadband albedo in the IR and VIS? Or an averaged value? Is it possible to indicate more clearly the different phases in the figure - eg the end of phase 3 in 2015 seems to extend to end of June on the figure but extends to mid-June in the text. Similarly, phase II starts on may 19 in the text but before may 14^th in the figure.The albedo of the highly heterogeneous ponded sea ice from 2016 was calculated from the transect measurements? Is it possible to indicate in the figure legend how many measurements are included in the box plots?

Line 294: Did you mean “The transition from snow cover to bare ice”?

Line 356: Typo “junE 6”

Line 365: “wicked up the first snow layer” I don’t understand the meaning of this, could it be reformulated?

Table 1: Is it possible to indicate in the methods or in the table how many samples were analysed for density and SSA to derive averages and standard deviations?

Figure 7: Why are albedo data presented only from 400nm if the spectroradiometer could measure from 300? Is it possible to indicate the number of samples used in the boxplots, or at least a range? What does “specific albedo spectra” mean, are they averages, or how were these examples of bare ice and melt pond albedos chosen?

Line 409: I am not sure I fully understand what was done here: “It was multiplied by the average density of snow (350 kg m -3 ) in order to obtain an average absorption coefficient of the impurities in the snow”. If I understand correctly, the absorption of particulates from melted snow leads to a coefficient expressed in m-1 of melted snow. How can this then be multiplied by the density of snow, leading to units of kg snow per m4 of snow, and give an “average absorption coefficient”? and where are these “average coefficients” shown? Do the authors mean that they divided the coefficient in m-1 of melted snow by the density of water and then multiplied it by the density of the snowpack in order to get an absorption coefficient in m-1 of snow instead of melted snow?

If possible, would it be possible to divide the absorption coefficient in m-1 by the LAP concentration (in kg m-3) in the solution of melted snow that was filtered to carry the spectrophotometric analysis in order to get a mass absorption cross section in m2 kg-1? Then the data could be used in other widely used radiative transfer model such as SNICAR.

Line 434: “It is possible to fit each spectral albedo with optimized LAP concentrations, but without measurement

in each pit, this is not very meaningful” I do not understand what the authors mean here – if it is possible to retrieve the impurity concentrations by inversing TARTES, why would it not be meaningful?

Line 438: “consistence” -> “consistency”

Line 444-447: It would be clearer to have this paragraph at the beginning of the section 3.4.2 to understand the reasoning in comparing LAPs concentrations between 2015 and 2016 in line 443.

Figure 9: typo “albedo witH LAP” (h missing) and maybe write “LAPs” instead of “LAP” because both MD and BC are included in the simulations if I understood correctly. It would be beneficial to add the phases as in Figure 3.

Lines 458-462: It would be clearer to indicate only absolute values without units in this paragraph since the relative errors in % are in the table already – the error at 500nm is given in % (indicated as absolute value?) whilst the 0.04 and 0.02 are indicated without % units (are they %?).

Line 467: Why were LAPs omitted in these simulations (Figure 10) if they improve the fit between measured and modelled albedo?

Line 512-513: the link in parenthesis should come before the dot and there should be a dot and a space after the last parenthesis

557-558: According to figure 3, the albedo at 1000nm still varies a lot in phase III (similar slope than phase II?)?

Line 568: What is meant by “the coupling with the grain size”?

Line 575: What is meant by “not adjusted”?

Line 600: snow cover instead of coverS

The manuscript presents a study on field observations of evolving snow physical properties and albedo during the early melt season on landfast ice in Canada, and uses those results for a snow albedo model analysis. The study nicely connects the changes in relevant physical properties to changes to albedo. Overall, the manuscript is well written, well organized, and with some minor revisions, it would be in good shape for publication. In particular, the discussion section was enjoyable to read. Please see comments below that I hope the authors will find useful.

General: It's not clear why the albedo results at the 500 nm and 1000 nm wavelengths are emphasized. Presenting the numbers in the visible (300-700 nm) and near infrared (1000+ nm) bands would make it easier to compare with previous works (e.g., Brandt et al., 2005) and be more relevant for remote sensing applications. 

Lines 21 and 27. It would be useful to know the snow depth at which the visible band in albedo begins to decrease.

Lines 36-38. During winter, there's little sunlight, so the albedo the surface is not important. In spring and summer, it's important.

Line 39. The melt season begins when the snow starts melting. Snow may affect the duration of sea-ice melt.

Lines 40-41. This is true for thin sea ice, but not for thick sea ice. Snow has little effect on the amount of light reaching the ocean if the ice is very thick. 

Line 45. The authors may be interested in reading an updated review of snow and ice optical properties by Warren: 

https://royalsocietypublishing.org/doi/full/10.1098/rsta.2018.0161

Line 52. It would be worthwhile to add a description in the text about the limitations of using SSA for snow crystal representation in optical modeling. 

Line 76. There are some cases where snow persists all summer.

Line 80. 'albedo drops remarkably'  

It would be informative to include the albedo change from dry to melting snow. My understanding is that a change from 0.85 to 0.70 is not that remarkable relative to the change from snow (0.85) to melt ponds (0.25-0.65).

Lines 82-84. This is an overextension of results. Snow melt does not directly enhance snowfall.

Lines 86-87. In some cases, the snowpack is deep enough that it never fully melts away, as observed around ridges: 

https://online.ucpress.edu/elementa/article/10/1/000072/169460/Spatiotemporal-evolution-of-melt-ponds-on-Arctic

Lines 91-92. 'However, studies which aim to link physical and optical properties of snow still remain largely qualitative'  

This isn't true. Warren, Brandt, Grenfell, Perovich, and others have made a lifetime of work in linking snow physical and optical properties, including their co-evolution. I suggest rewording this section so that it recognizes that this work is standing on the shoulder of giants and is adding to a foundation of knowledge.

Lines 92-93. It is true there are data limitations, but the greater limitations may be the representation of physical processes, which are difficult to appropriately incorporate as parameterisations into earth system models. 

Lines 94-95. There are several field campaigns that have done this.

Lines 116-117. How far away was the meteorological station? It would be helpful to include that information here.

Lines 135-136. What information was used to determine the auto-adjustments? Does the auto-adjustment create inconsistencies in the noise level of the measurements?

Line 140/Figure 2. These are useful photos. Is it possible to replace them with higher resolution versions?

Lines 146-147. What makes a relatively thinner snow pack less suitable? Wouldn't the combination of thin and thick be more representative?

Lines 177-178. The instrumental uncertainty of the probe would be helpful to include here.

Line 182. It would be good to expand on this a little more. What types of snow have larger uncertainties?

Line 196. typo 'them'

Line 265/Figure 3. Why are there different shades for the different horizontal bars? The shades don't match the gray legend in the bottom panel.

Lines 274-275. Often, there can be melt forms near the ice-snow interface from the previous autumn. Were there no melt forms observed at the base of the snowpack?

Lines 279-280. It would be informative to describe how the temperature gradient was reversed. Was the temperature range the same but with the upper surface being -4.5 to -5C, or do the authors mean that the snowpack was simply warmer near the surface and cooler near the base?

Lines 286-287. Did snowpack temperatures increase from the top down?

Line 300/Figure 4. Just after the May 8 snowfall, the snow depth increases. What caused the increase if no snowfall occurred?

Lines 315-317 and lines 320-321. I'm surprised by the higher density values for indurated depth hoar and the lower density values for wind slab in this study. Can the authors comment on this with regard to previously observed values? Is it possible that the fresh snowfall events contributed to the density measured in the uppermost portion of the snowpack, lowering the average density for the wind slab layer? 

Lines 325-326. Figure 6 doesn't show the distinct vertical layers. Is there a way that this can be added to the figure?

Line 330. Same comment as before that Figure 6 doesn't show the distinct vertical layers of the snowpack.

Line 335/Figure 6. What does the white at the base of these profiles represent? Is it no data? Also, how much of the variability in the uppermost profiles before May 25 is due to spatial heterogeneity versus variable weather conditions, such as snowfall events? It may be insightful to comment on this in the text.

Line 346. 'Some of these new layers were thick enough to be distinguishable in Figure 6.' 

It would be helpful to highlight these in Figure 6 somehow since they are not obvious. Was snow density of these new snowfall layers measured?

Line 356. Typo Jun 6.

Lines 357-358. This is a little confusing. How did the mass of the snow increase without notable snowfall events (Figure 4)?

Table 1 caption. Do you mean Figure 5 here? I suggest adding an additional column that describes the predominant snow layer morphology (indulated depth hoar, etc.) so that readers don't have to scroll back and forth to know which layer means what. Also, it looks like there may be a typo for Layer I.

Lines 377-378. Is it possible that the sloped surface of the dunes, and therefore the angle of reflectivity, affected the albedo measurements? 

Line 385/Figure 7. It would be helpful to add the sample size for each panel, e.g., N = 15 to better interpret the changes between phases. It would also be informative to note what fraction of the albedo measurements were made over melt ponds.

Line 395. Similar to the previous comment, it would be informative to note what fraction of the albedo measurements were made over melt ponds in this section.

Line 407. 'which ranged from...'  

It would be informative to add shading or thinner lines to Figure 8 to show the range or spread of the absorption spectra from the 12 samples.

Lines 408-409. Shouldn't it be divided by the snow density to get the average absorption coefficient per volume of snow?

Lines 414-416. It's not clear how these values were determined. Was this some sort of sensitivity study with some details missing in the methods section, or manually trying different values until a decent overlap was reached with the observed average?

Line 417. These measurements were only made over snow dunes, is that correct? It would be helpful to add that note in the text here to remind readers.

Lines 482-483. It would be interesting to include the snow depth at which PAR becomes significant. 

Line 543. Rain events occurred? That would be informative to include in Figure 4. What impact did the rainfall have on the albedo and snow properties?