The study “Airborne observations of the surface cloud radiative effect during different seasons over sea ice and open ocean in the Fram Strait” by Becker et al. derives and analyzes cloud radiative effects (CRE) collected from aircraft during three campaigns based from Svalbard that made transects to the west and north of the island. The campaigns took place in different years, but also importantly in different seasons (early spring, late spring, late summer). The variability in CRE values are then considered for a range of solar zenith angles and cloud conditions, and over sea ice vs open ocean. These evaluations shed light on the spatial heterogeneity of cloud forcing at the surface. The manuscript is generally well-written and suitable for Atmospheric Chemistry and Physics. I have three main comments that I think are important for the authors to address before the manuscript moves to publication.

• The discussion of the important terms is inconsistent and sometimes confusing in the text. I recommend distinguishing explicitly between LW (TIR) and SW terms throughout. This would start by expanding eqs (1) and (2) to define separate terms for LW and SW using subscripts and then referring to those terms as well as their summations, CRE_SW, CRE_LW, CRE_total explicitly throughout the text.
• The CRE presented are advertised as referenceable to the surface but I don’t see any description of the atmospheric corrections necessary to transfer the SW and LW observations from the altitude of the aircraft to the surface. If you think that the aircraft was low enough that no correction is needed, some evidence that the flux divergence between the aircraft and the surface is negligible is warranted.
• Line 310: I agree that locally and briefly downwelling shortwave at the surface can exceed TOA irradiance and could cause real positive values of solar CRE. However, I’m skeptical (in particular given the altitude of the aircraft) that this is such a significant effect on the surface values so as to make up as large a fraction of the samples as you show in Figure 7. Indeed, mode 1 (“cloud free”!) appears to be associated with positive solar CRE almost all the time. Something is amiss. Maybe you could validate your simulated clear-sky SW with observed SW under clear skies to be sure that (a) there is not a bias and (b) to potentially explain the preponderance of positive values as uncertainty in the clear term. If there are not enough statistics from the campaigns, perhaps a longer record of validation from Ny-Alesund can be performed.

Other comments:

Line 71: Awkward wording. Maybe “Fewer efforts have focused on CRE over open…”

Line 105/111: This seems deceptive. What is the response time of the thermopiles? Are you really making independent samples at 20 Hz?

Line 105: It might be better to show Eq (1) using terms for both TIR and solar separately to make cross-referencing like this more clear.

Line 107: While discarding tilted data is one approach, even at 5 deg biases can be large. Corrections are possible up to 10 deg (https://www.doi.org/10.2174/1874282301004010078). Did you consider this?

Line 172: meidan -> median

3.1: Here again, I think it would be useful consider Eq (1) as LW and SW separately and to distinguish more clearly in this paragraph the methodologies used for the two bands.

Figure 4: Maybe specify in the caption that this is simulated, not observed.

Figure 8: So just to be clear, there were no clear-sky samples made during MOSAiC-ACA? I think it might be helpful to emphasize that point because to look at Figure 8 it appears as if you are reporting CRE > 25 Wm2 under clear skies.

1. Original Submission

1.1 Recommendation

Major revision.

1. Comments to Author

I thank the authors for this generally well-written and informative paper that intercompares the broadband shortwave (solar range) and longwave (thermal infrared range) cloud radiative effects (CRE) at the surface measured from three aircraft campaigns over different regions in the Arctic during different seasons. Detailed analysis is provided for CRE over open ocean and ice surfaces by discussing impacting factors of solar zenith angle, cloud type, seasonal feature, and thermodynamic property. Since the derivation of CRE requires cloud-free radiative transfer (RT) calculations, the paper introduced parameters required during the RT setup, which are well described and easy to follow, especially the surface albedo parameterization to account for the cloud-induced albedo change. The results are well interpreted and scientifically significant. In addition, this paper is valuable for incubating radiation science for the upcoming aircraft mission of ARCSIX (Arctic Radiation-Cloud-Aerosol-Surface Interaction Experiment) and fits in the scope of ACP. Thus, I recommend for publication with major revision once my major and minor comments are addressed.

2.1 Major Comments

1. Consider changing the terminology of “solar” and “thermal-infrared” to shortwave and longwave (and later abbreviated as “SW” and “LW”, even the spectral ranges of the broadband radiometers only partially cover the SW and LW) since the SW and LW are more broadly used by the CRE community.
2. L80: “However, both results included … observations.” I didn’t quite get what are the limitations of others studies. Please clarify.
3. For Figure 1, considering changing the color of red (or orange) to another color (green maybe?) and make the lines for low level sections slightly thinner (or add a little bit transparency in color) so the flight tracks can be better distinguished.
4. For Figure 2, consider changing the dashed line (or dotted line) to solid line, as dashed lines and dotted lines are difficult to distinguish. Also, the temperature variation of profile near surface is almost invisible, can you experiment with log y axis to see whether the temperature variation near surface stands out more (e.g., temperature inversion)?
5. L181-185: Which do you think is the more plausible cause for the much more frequent thin clouds occurrence observed during MOSAiC-ACA? Limitation in sampling statistics (e.g., with more data we will see similar distribution like ACLOUD) or the cloud type associated with Arctic season (e.g., even with more data we will still see predominant thin clouds)?
6. Figure 5 is very interesting. I would like to see more explanation about why the albedo change in such way (e.g., the “dip” of albedo at SZA of 75° when transitioning from clear-sky to optically very thin clouds, there must be some counteracting factors) rather than the descriptions of how the albedo change along LWP.
7. Consider changing the color of the markers (crosses and dots, also in the legend) in Figure 7 and 9 as they are hardly distinguishable between the numbers and add descriptions in the figure caption.
8. From L316-322, the author argues the SZA causes the different distributions in CRE among different campaigns. I have an idea of normalizing the CRE with the cosine of SZA (it should not be difficult to do). If SZA is the culprit, the distribution difference should disappear once the CRE is normalized.
9. I quite like the places where you brought up “broken clouds” seen during the MOSAiC-ACA campaign, which have caught my attention in wondering how much 3D cloud radiative effects are there in the Arctic. Technically, the 3D effects should be predominant in the Arctic when broken clouds present as the surface is bright and sun is low (more scattering events). Even though I understand the 3D effects are not the focus of this paper, I would recommend adding some brief discussion about it could potentially favor the radiation closure development in the Arctic.
10. Consider adding some thoughts about lesson learned from the three Arctic aircraft campaigns and improvement one can make for future aircraft campaigns in order to progress cloud radiation science in the Arctic (e.g. ARCSIX) in the conclusion.
    
    

2.2 Minor Comments

1. L1: <during airborne> to <from>; <three> to <three airborne>
2. L2: suggest to added <– AFLUX (2019 March to April), ACLOUD (2017 May to June), and MOSAiC-ACA (2020 August to September)> after <Svalbard>
3. L3: <of the surface> to <at the surface>
4. L8: <component> to <components>
5. L8: <and in combination> to <as well as combined for the study of total CRE.>
6. L90: <airborne measurements of … were performed during three seasonally distinct campaigns in the vicinity of Svalbard> to <three airborne campaigns were deployed to collect measurements of cloud, surface, and thermodynamic properties during different seasons near Svalbard.>
7. L97: <performed> to <deployed>
8. L108: add brief description of why the observations are discarded, something like <due to the contamination of radiation signals from …> after <discarded>
9. L121: <the cloud boundaries> to <cloud boundaries>
10. L134: <f_ice> was not clarified (or I missed it?)
11. L151: worth adding RH profiles
12. L172: <meidan> to <median>
13. L203: <illumination geometry> do you mean <solar geometry>? Just want to confirm.
14. L230: why? Due to rough ocean surface from high wind speed, thus less specular reflection?
15. L247: where does <sea ice concentration> come from? Estimation from aircraft camera imagery? Any reference for figure for this linear combination?
16. L251: <amounts to> to <converges at>
17. L258: see my earlier comments, why a decrease is observed for SZA of 75 but not 60 under clear-sky?
18. Figure 6: consider adding “retrieved” and “observed” to the y axis labels for (d) to (f), as well as for the legend labels in (b)
19. L284: <their mode structures> to <the mode structures over the parameter spaces of surface albedo and CRE>
20. L287: add <cloud-induced> before <surface albedo change>
21. Figure 7: last sentence is unclear, please clarify which symbols are which. Consider referencing back to the text or adding an example to explain. Consider change the color of dot and cross markers to distinguish them from numbers.
22. L290: since you are showing broadband irradiance, suggest change <solar spectral range> to <solar range>.
23. L291: <Figs. 7d, 7f> should be <Figs. 7c, 7f>
24. L292: can the mode change/shift due to 3D cloud radiative effects (from thin or broken clouds)?
25. L300: consider providing the actual values of solar CRE of mode 4 in Fig. 7e
26. L301: <reduction> can be unclear about which way CRE goes, whether more cooling (values become more negative) or more warming (values become more positive). Consider <mitigation>.
27. L303: <was increased by 29 Wm^-2> to <imposed an artifact of 29 Wm^-2 cooling due to the neglect of cloud-induced surface albedo change>
28. L304: <slight> to <negligible>
29. L309: see my earlier comments, consider mentioning the 3D cloud radiative effects
30. L323: <blurry> to <unclear>
31. L341: the cross marker in Figure 7 seems unexplained.
32. L456: reference shown up as <?>
33. L470: please fix <?, …>