This is an interesting study of considerable merit for its measurements of an important single scattering parameter of cirrus in a location where few measurements have previously been made. The asymmetry parameter is of secondary importance to the optical depth for calculating cloud reflectivity. It is nonetheless important to get right, as existing uncertainties can lead to errors of up to a factor of two.

I have some rather deep concerns about the analyses however, generally with regards to the aspect that the measurement techniques seem insufficiently justified, and also the reported results appear to be implausible.

Working in order through the paper:

l. 116 A threshold of 0.1 ms is applied to remove "shattering events". Some justification seems necessary here. Is there a distinct mode in the interarrival times that would suggest there is a shattering mode? How can it be known that such events are not a result of natural turbulent clustering of particles, a well known phenomenon in clouds? If such events were included, would it affect the calculated values of g, optical depth, and all the other microphysical parameters? What I suggest here is plotting a spectrum of interarrival times, logarithmically binned, on a log-log plot (i.e. d n/dlog(tau)). If the spectrum has the property of scale invariance, namely the slope is nearly constant across interarrival times, including 0.1 ms, then the physics governing interarrival times at 0.1 ms should be anticipated to be the same at any other scale. If there is a scale break or distinct mode, then a better argument can be made that such filtering is justified.

l. 131 Baker and Lawson (2006) focused on mid-latitude clouds, not Arctic clouds. The premise of this submission here is to consider latitudinal variations in microphysical and optical properties. What justification is there that the power-law behavior identified here for relating mass to area, obtained at mid-latitudes, can be applied to the Arctic?

l. 154 The data is stated to be "manually cleaned" based on "intact" imaged particles. This sounds very unscientific. Can a more objective justification be described for what is being done to what?

Section 2.2.3  This section needs much more detail. A point of particular concern regards the validity and uncertainty related to the assumption stated in Xu et al (2022) that about the assumption lying behind the "mean" statement on l. 174 that "This is achieved by exploiting the assumption that the forward diffraction and the refraction − reflection energies are asymptotically equal. " First, it's worth considering the rant in Bohren and Clothiaux about how there is no refraction, reflection or diffraction -- only scattering and interference. The distinctions between the three are entirely artificial. But more importantly from a measurement standpoint, per Jarvinen et al (2023), the polar nephelometer only measures scattering at angles between 18 degrees and 170 degrees, which for any conceivable cloud particle encompasses quite a lot less than half the total scattered energy justifying a straightforward mean of forward and side/back scattering. Perhaps this all makes sense. I'm not sure I understand the need for a Legendre series expansion as described in Xu et al (2022). But at the very least, a full justification, with error analysis, should be presented of this measurement that is core to the article.

Table 1. The microphysical measurements presented appear implausible for what they would imply for the reflectivity based on the thin cloud expression given by Eq. 8. Taking the reported median microphysical values of  3 mg/m3 for the IWC, 37 um for the effective radius, the mean optical depth for a cloud 1 km thick would be 0.12. If, much more generously the cloud were 3 km thick, then it would be 0.36. From Eq. 8, this quite thick cirrus,  taking g = 0.727, would have a reflectivity of 0.05. From the ground, such physically thick clouds would be barely visible. Mostly one would see blue sky. This seems implausible given cirrus are certainly very visible in the Arctic in satellite measurements, with reflectivities I would guess 10 times as high.

l. 412 to 421 This paragraph risks being a bit misleading as global climate models are dynamic. It could well be that a value of g that is too high means a low bias in reflectivity with substantial instantaneous radiative forcing impacts. But there is a feedback. With less reflected, more sunlight is transmitted, which by heating the ground could destabilize the atmosphere to create more clouds, offsetting the low reflection bias that is discussed here.

The work presents the analysis of wide-ranging in situ measurements on individual cloud ice particles using the PHIPS probe. It particular, it is attempted to retrieve the shortwave (SW) scattering asymmetry parameter g of the particles. This topic of of high importance in atmospheric and climate research, as g essentially determines the balance between radiation transmitted and reflected by clouds, hence influences the radiative balance of the atmosphere. Yet data is scarce on g from atmospheric measurements or realistic representations (theoretical or laboratory) of ice particles. Moreover, small differences in g cause large differences in the reflected component (proportional to 1-g) since SW g from ice appears to be >>0.5. So high accuracy of measurements/retrievals is of paramount importance.

The justification for the method used in the present work for retrieving g is given in Xu et al. (2022), incidentally co-authored by the present Authors. Unfortunately, this justification and the method based on it are flawed for several reasons, and the Authors do not strengthen the justification in the present work. Firstly, the method relies on the assumption that scattering can be asymptotically split into an external diffraction component and internal geometrical optics (GO) based component. This separation is artificial, and aside for being approximately accurate only in the very large particle limit, in reality the phase function is a blending of the influence of diffraction-like and GO-like influences. For instance, the forward peak appears to be shaped largely by diffraction, yet it also contains a contribution from undeflected (for "smooth" shapes) so-called delta rays (exact forward) of GO theory - an internal scattering component. For "rough" particles these delta rays become deflected away from the exact forward direction, resulting in peak broadening. This broadening would not be accounted for by the simplified external diffraction used to approximate the shape of the forward peak. More on the forward peak below.

The internal component is based essentially on extrapolating the very wide missing parts of the angular range from PHIPS using Legendre polynomial expansion. Accuracy is claimed to be improved by cutting the numbers of terms in the expansion, because, it is asserted, the cutoff excludes the missing forward and backscatter angular ranges - a fallacy and a circular argument (increasing accuracy by decreasing it). One undesirable consequence of the truncation of higher orders is essentially the distortion (hence essentially exclusion from the outcome) of phase functions characterized by the presence of a narrow forward peak (which cannot be approximated by the low order expansion terms). These phase functions are associated with larger particles and those with "smooth" geometries - and both would result in large g. Thus the resulting distribution of g values is distorted and a bias introduced towards lower g. Moreover, the support for using the truncation of the Legendre polynomial expansion is based on results from GO, which does not represent scattering from real ice particles correctly, especially for smaller size parameters.

The external diffraction part is approximated using scalar theory for circular apertures. Yet the diffracting silhouette of most natural cloud ice particles is very far from circular (think elongated column or a rosette), hence diffraction on them can be expected to be stronger due to the presence of smaller features. Again, Xu et al. (2022) merely state that "the error caused by non-spherical ice crystal should be small [...] since the asymmetry parameter due to large particle diffraction is very close to unity" - a very weak argument, more wishful thinking than a scientific one.

Thus if these results were to be published, they would have to be associated with a massive "health warning" because of the likely large errors in the retrieved asymmetry parameter and consequent bias - stated up front, including the abstract, and clearly emphasized in the discussion. Can these errors be quantified? This is unlikely within the present methodology, which significantly lowers the scientific value of the present work.

I have refrained from discussing minor aspects of the manuscript as the questions discussed above have to be addressed first. I would also recommend seeking further references from additional referees.

Reference:

Xu, G., Schnaiter, M., and Järvinen, E.: Accurate Retrieval of Asymmetry Parameter for Large and Complex Ice Crystals From In-Situ Polar

Nephelometer Measurements, Journal of Geophysical Research: Atmospheres, 127, 1–19, https://doi.org/10.1029/2021jd036071, 2022.

The two previous reviews have provided very good feedback, and the authors should take note and give careful consideration to their suggestions. The article itself is interesting and is potentially important, but it lacks rigor regarding whether the retrievals of low-g are artifacts of the retrieval methodology. To address this, the following test of their methodology is encouraged, it is actually essential if their work is to gain credibility.  The current article has not demonstrated that the retrievals are free from systematic bias. 

In particular, there is no closure test in which full theoretical phase functions are truncated to the PHIPS angular range, passed through the retrieval algorithm, and then compared against the true g. Such a test would quantify the extent to which the separation and extrapolation technique biases g for different habits of various size, surface roughness etc. Given that the paper's main finding is a consistently low asymmetry parameter, the absence of such a validation leaves open the possibility that at least part of the low values reported are retrieval artifacts rather than physical properties. 

Without such a validation, the credibility of the main finding is undermined. I therefore cannot recommend publication of the manuscript in its current form.