This paper presents the implementation of radiative transfer model SCIATRAN on snow, sea ice and ocean system with a great depth of technical details. I think in general the manuscript is well written and worth publication due to the scientific significance of the method and result it presents. Meanwhile, I have the following comments and would the authors to address before its publication.  

Major comments: 

1. The paper aims to cover a large number of topics, such as coupled atmospheric-ice-water surface, multi-layer snow and ice system, and polarization. The presentation of these many topics makes it hard for the reader to guess the technical details for specific topic. For example, Appendix A shows the implementation of a vector scheme of radiative transfer solver and all benchmark results in Section 4 are validated against vector radiative transfer models on relatively simple ocean scenes, which strongly suggest that we are seeing a vector scheme of coupled atmospheric-ice-ocean system. However, Section 5 validation is done with un-polarized measurements exclusively and makes me wonder whether it is done with the same vector radiative transfer scheme but under the scalar mode by setting all non-F11 elements of the scattering matrix to zero. I wish the paper can be clearer on that. 
2. For the solving of minimization problem in Section 5, what are the a priori information used in each minimization? Also what is the measurement covariance being assumed? Also, what is the minimization solver used for these results?  

Minor comments: 

Line 65 reference with ? 

Line 472 refferred -> referred  

Line 481 show -> snow 

Please increase the size of the polar plots of Fig. 8-11. They are too small and it is hard to pickup details from them. 

This paper describes the technical implementation and evaluation of a coupled atmosphere-snow-ice-pond model, SCIATRAN.  Overall, this has good potential to be a useful tool for the community, and I commend the authors for advertising it as open-source and publicly available.  The model potentially has particular utility for the remote sensing community, as it produces multi-stream intensities and BRDFs for any illumination and viewing geometry.  Moreover, the paper is generally sufficiently detailed and thorough in its description, comparison with other models, and comparison with measurements.  Below are a few minor comments that I think should be addressed prior to final publication.

General comments:

It is well-established that the particle extinction efficiency approaches two in the geometric optics limit (e.g., Bohren and Huffman book, 1988), and indeed the authors formulate this for scattering efficiencies of air bubbles and brine inclusions.  Puzzlingly, however, the extinction efficiency for ice particles appears to be taken as 1 (if I understand correctly), as described on the bottom of page 11, neglecting the contribution from diffraction.  If true, this seems to be a significant inconsistency in how extinction is formulated for ice particles vs. air bubbles and brine.  Please address this issue, along with the implications of this assumption.

Line 319: A scattering asymmetry parameter of 0.998 is incredibly high.  Practically, this would result in almost no scattering, since photons are simply redirected into the same direction they had originally been travelling.  Some discussion of the implications of such strong forward scattering should be included.

The model-measurement comparison is helpful, and (encouragingly) indicates that the model is capable of reproducing observed spectral reflectances quite well.  The fitting procedure is also effective, but it is important to note that the best-fit model parameters may or may not agree with the real state of the snow and ice that existed when/where the radiances were measured.  In other words, this is not a perfect "closure" evaluation (since states like effective grain size were largely unknown) and it is at least possible that the radiative model agrees with measurements for the wrong reasons.  This is a minor critique, but I would like to see this issue acknowledged more clearly in the discussion.  (Also, more generally, it seems the fitting technique could be quite useful in the future for remote sensing inversions, once confidence in the forward model is instilled.)

Minor comments:

Figure 1 caption: Please include a first name and institution for Dr. Gunnar (attribution of the photo).

Line 65: There is a missing reference, indicated by "?" here.

Lines 76-88: Does the "Coupled Ocean and Atmosphere Radiative Transfer (COART)" model (https://satcorps.larc.nasa.gov/jin/coart.html) also qualify for discussion here?

Line 103: Briefly, what might this "yellow substance" represent?  Mineral dust?  Organic carbon?

p.12, 4th last line: For the sake of reproducibility, what value of sigma_y (at lambda_0) is assumed in this parameterization (i.e., for the mass- or volume-normalized absorption coefficient)?

line 296: Assuming this is referring to snowpack (and not atmosphere), "vertical" would be preferable here to "altitude"

line 313: Is use of the Henyey-Greenstein function a user option in the code?  If so, I suggest clarifying this.

line 369: "Stockes" -> "Stokes"

line 453:  What are the model options for "relevant BRDF model"?   (Perhaps this is explained earlier.)

Line 460: Briefly, please provide more detail about how scattering and absorption by aerosols is accounted for.  What types of aerosols are represented?  How is the aerosol burden or optical depth determined?

Line 561: "... it still can reach values of about 0.7 for clear sky conditions..." - This is indeed surprisingly high, but it is only in a limited portion of the spectrum.  I suggest clarifying here at which wavelengths this high value is achieved. 

line 581: "clody" -> "cloudy"

line 749: scattering layer thickness of 0.19-0.23 cm: This is quite thin, and much thinner than the value of 2cm that is often assumed in models for the thickness of the surface scattering layer.  I suggest adding a brief discussion on that here.

The references should list last name (surname) first.

Appendix A: I see that expressions for thermal emission are included here.  It's fine to include those in the comprehensive model formulation, but just to be clear: Is thermal emission actually modeled in this study?  I assume not because it only deals with solar spectrum processes.  Perhaps thermal emission could have a small effect, however, at the longest wavelengths examined in this study.  Please clarify.