See attached PDF.

A new polarimetric radio occultation metric is used to assess the fidelity of different microphysical schemes in multiple simulations of atmospheric rivers. The results seem robust showing that xsnow~0.1 (aggregates) provides the best results and it is a good demonstration of a novel measurement that could be used to test other schemes and build climatologies.

This work should be publishable subject to satisfying the following points.

Main point.

This work is introducing and demonstrating a novel metric to be used to test cloud microphysics representations. To provide context and convince the reader of its value it would be good (necessary?) to also provide comparison to more traditional metrics. Comparisons should be made to readily available satellite derived precipitation and top of atmosphere broad band radiation, perhaps even vapor and liquid water path too. Ideally these comparisons will also demonstrate the Goddard scheme performing best and supporting the result of the PRO analysis.

This would just form an extra section and add some paragraphs the results/discussion and conclusions.

Other points.

line 24. I did not see anywhere a discussion about the horizontal resolution of this approach. That needs to be delved into along with a discussion about the pros and cons of using these long path lengths (~200km?)

line 70. Perhaps the spatial coherence of these phenomena is also good for this approach that has coarse horizontal resolution? 

line 84. This section would benefit from outlining the geometry of the sampling, perhaps with a schematic? Section 2.1 and 2.3 should also be merged? I read 2.1 and wondered why the ice phase was being ignored.

line 127-128. What were the problems - it did not seem to be mentioned again.

line 140. Can probably just omit 'new'?

line 158. Was a single instantaneous out chosen? Or was it averaged in some way. How close was the sample to the output in time?

line 170. See the main point. This paper can be strengthened substantially with some additional comparisons to satellite data to support the findings from this paper.

line 227. The use of 50 assumes that each point is truly independent. Are these 'points' next to each other as in the gray area in figure 1? If so then they seem to be spatial coherent over scales of order 100km or so and lower number than 50 might be more appropriate?

line 248 and throughout. These schemes all use different size or mass thresholds to determine what is ice and what is snow. That could lead to big apparent differences between them that are not necessarily incompatible. For figure 5 do the schemes look more similar if ice+snow if plotted? 

line 262-265. I was confused here. Equation 7 define iWC and delta Phi as being the integral along the limb sounding path. Here it talks about vertical integral? I am assuming that the WRF output is integrated along similar curved paths through the domain? How is that done?

figure 6. All 4 schemes are represented - does this mean that each one was best in different cases? Maybe i am misinterpreting.

Why not show all results - maybe a different panel for each scheme if it gets too messy?

line 272. DeltaPhi is defined as dependent on WC and x. Does a surface regression against WC and x result in a better result?

line 278. But how can they all have the same WC and delta Phi axes in figure 7? The different x values will mean that each species should have its own delta Phi for a give WC?

figure 7. Maybe some more discussion about how the data is used is needed. It looks like you are plotting the curved limb path iWC value at its lowest point(?)

line 282. Can you tabulate J and the error. Perhaps indicate where the lowest J is significantly lower than the others?

line 318-319. It would be useful to locate some ground based polarimetric studies of ARs to see if they agree with this result.

Summary

This paper studies the Polarimetric Radio Occultation (PRO) technique to asses its sensitivity to vertical profiles of hydrometeors under varying microphysical assumptions in the context of atmospheric river cases. This sensitivity is theoretically explored using WRF model output from which differential phase shift ΔΦ is simulated. The simulated  ΔΦ is compared against the observed with the aim to evaluate the applied microphysical schemes. 

The study is well structured, clearly written and has informative figures. While I do think the study has the potential to be well-received, I have one major concern about the conclusions drawn from the optimization with the x-parameter method that is presented, in addition to some minor comments for clarity and quality improvements.

Main comments

1. The method is based on x-parameters that relate water content (WC) to specific differental phase (KDP). Given the simulated WC, the 'optimal' x-parameter is then found by comparing simulated KDP against the observations. The authors conclude from this x-parameter that a specific particle habit is dominating the signal based on particle habits from the ARTS database. My concerns with this approach are 1): There will usually be a mixture of particles present, especially since measurements are done over a profile, and typically different particle habits dominate at different altitudes (temperatures). One example: Wouldn't a 50-50 mixture of particles with x-parameters of 0.1 and 0.3 yield an 'optimized' x-parameter of 0.2? In your current draft, you would then conclude that particles that relate to x-parameters of 0.2 are dominating. 2) The 'optimization' might lead to the correct results for the wrong reasons. E.g., a simulated water content that is much lower than in reality could be compensated by a higher x-parameter to achieve the correct KDP. 3) Keep in mind, that you 'overwrite' some of the particle properties that are used by the WRF microphysics schemes, by taking ARTS particle habits instead. For example, there are specific mass-size relations used, specific PSD shapes, and density assumptions. While I think this last point is not a major problem, you should at least discuss it, since you goal is to 'evaluate' microphysics schemes.

Minor comments

1. Line 100-104: I found it hard to understand this paragraph. Partly, because some of the sentences are incorrect. I also think a small sketch visualizing the ray-path and the position of h_t on that path would help. 
2. Line 105: I understand from this that the ray path is resolved with a given resolution. What is this resolution?
3. Line 115: Two-way or one-way nesting?
4. Line 120: What is the horizontal extend of the domain?
5. Line 127-128: Did you change the radiation scheme then for specific microphysics schemes only? Or for all microphysics schemes of that AR event?
6. Line 134: Water vapor is typically not considered a 'hydrometeor'
7. Line 135: I would add here that the differences in assumed properties, such as particle density, are also important.
8. Line 153: Morrison is two-moment only for graupel, rain, snow and ice (not for cloud water). Also, Thompson predicts number concentrations (and thus, is two-moment) for cloud ice and rain. Here it sounds as if Thompson was completely one-moment. 
9. Line 158: What were your conditions to define an AR event?
10. Section 2.2: Was there any nudging applied?
11. Line 243: I don't fully understand the reasoning here. Isn't water content directly output by WRF? Why do you argue based on ΔΦ that snow is contributing the most?
12. Line 246: Could you see a height dependence? Or is that true for the full profile?
13. Line 256: How are you sure that Thompson is overestimating, and the other schemes are not underestimating snow water content?
14. Line 263-264: This sentence is confusing to me. Is there perhaps a word missing?
15. Line 315: In my understanding that just means that the 'average' x-parameter that fits best is that of Rosette/Aggregate. See major comment 1). 
16. Line 330: Is snow the largest contributor due to the largest x-parameter or largest WC, as indicated by Fig 7?
17. Line 335: How is the error determined?
18. Line 364: Greater contribution to what.  ΔΦ?

Technical corrections

1. Line 149: Brackes around the year only. 
2. Figure 6: Panel (b): Colorbar label and Figure caption do not match. I think the figure caption for panel (b) is wrong.
3. Line 272: I think there is a word missing. Perhaps: ... no universal relationship ... exists, particularly...