Both reviewers pointed out potentials of this study but also provide constructive comments which I believe are valuable for authors to improve the manuscript. In addition to comments from two reviewers, I have a few comments for authors to consider during the revision:

I got impression authors have done a lot work in this study and presentation of results were comprehensive. However, the reasons of “why” seems a bit weak. The authors claims the biggest improvement was the surface temperature calculation: “The strongest improvement was observed for winter months over the Central Arctic, and the Greenland and Barents seas where a 4.91°C median ice surface temperature error of ERA5 is reduced to 1.88°C in CARRA on average.”   In winter months, the stable boundary layer (SBL) may dominate the surface energy balance. CARRA seems tackle SBL better than ERA5, yet the stable boundary layer was not mentioned in the manuscript at all. I would like to see authors give some discussion on it.

Several places in the manuscript, I would like to see some discussion on “why” e.g.

L630 “The sea ice cover in CARRA adequately represents general multiyear trends towards thinner and warmer ice cover, connected to the ongoing climate change in the Arctic. Comparisons against the satellite-based and in situ sea ice observations show generally improved representation of sea ice in CARRA (using ERA5 as a baseline), although this improvement is not universal”.  Why?

“The main difference between the sea ice schemes in ERA5 and CARRA is the presence of an explicitly resolved snow layer, which allows for much lower ice surface temperature in the CARRA system, therefore reducing the warm ice surface temperature bias found in ERA5” see my comment on SBL above. Can we say that the presence of a snow layer can better tackle the SBL?

“ However, for the area covering Baffin Bay and the Davis Strait the verification scores suggest that a warm winter-time bias of ERA5 is replaced with a cold bias in CARRA.” Why?

“No improvement over ERA5 was found in the ice surface albedo with spring-time errors in CARRA being up to 8% higher on average than those in ERA5 when computed against the CLARA-A2 satellite retrieval product.” Why?

L95: “Surface albedo of the sea-ice covered grid cells in CARRA is computed by applying simple parameterisation schemes. For snow-free ice cover, a temperature-dependent broadband albedo scheme is applied (defined as HIRHAM in Liu et al., 2007), and when ice is covered by snow an adapted version of the broadband snow albedo scheme by Douville et al. (1995) is used. When computing albedo of cold dry snow covering sea ice in the CARRA system, the albedo scheme of Douville et al. (1995) is modified to increase the value of the lowest possible albedo in the dry albedo degradation term from the original 0.5 to 0.75. Sea ice albedo schemes applied in CARRA do not distinguish between direct and diffuse components of surface albedo. The HARMONIE-AROME NWP system does not produce grid-cell averaged albedo as an output variable, therefore in the CARRA product the surface albedo field is computed from the hourly accumulated downwelling and upwelling shortwave radiation fluxes and available only from the model integration output.

Are you saying for each CARRA grid, in ice covered area, the surface albedo was calculated by simple parameterization schemes. For ice free area, the surface albedo was calculated by the ratio of upwelling /downwelling shortwave radiative flux, right? Then what is the final product of surface albedo for each CARRA grid cell? Is this (albedo in CARRA cell) better to be called “CARRA surface albedo field”? Can this ice surface albedo and ice-free surface albedo played any role on no improvement of CARRA surface albedo over ERA5 product?