In this paper, Jeggle et al. divided cirrus clouds into several formation regimes using K-clustering, and quantified the sensitivity of IWC and N_ice to dust concentration using multiple-linear regression. The results show that the sensitivity of N_ice to dust concentration is positive for in-situ regime, but is negative for other regimes. The authors assume that the difference is induced by heterogeneous nucleation of ice initiated by dust INPs in INP limited conditions with supersaturations between the heterogeneous and homogeneous freezing thresholds. The authors suggest that climate intervention method of seeding cirrus clouds with potent INPs may result in an undesired positive cloud radiative effect (CRE), i.e. a warming effect.

The paper is well written, and the conclusions are novel. However, I have concerns with the multiple linear regression method of this paper, and the major results of this paper are doubtful. Below are my detailed comments:

1. The cirrus clouds in the liquid origin regime is formed in lower altitudes, so the aerosol concentration in lower altitudes are also important in determining N_ice in this regime. Therefore, the multiple regression equation might be revised to something as following:

N_ice=beta1a*T(t=0)+beta1b*T(t=-6)+ beta1c*T(t=-12)+ beta2a*w(t=0)+beta2b*w(t=-6)+ beta2c*w(t=-12)+ beta3a*dust(t=0)+beta3b*dust(t=-6)+ beta3c*dust(t=-12)+epsilon,

where t=0 denotes the values at DARDAR observation, t=-6 denotes values 6 hours before, and t=-12 denotes values 12 hours ago (Fig. 2).

1. The paper uses dust with size greater than 1 micron. However, although dusts with size smaller than 1 micron are less effective in acting as INP, it is possible that they can still play a role. So additional analysis should be performed to demonstrate that these small dust particles are not important for N_ice.

Jeggle et al. use a combination of satellite-derived cloud properties (DARDAR-Nice data) with model-reconstructed vertical velocities and dust concentrations (from reanalysis data) to study the sensitivity of cirrus cloud properties to dust. They separate their results by cloud regime based on the temperature history of their back trajectories and updraft velocities. Their results show that increased dust concentration leads to fewer ice crystals in liquid-origin cirrus and more in tropical in-situ cirrus. This suggests a role for dust in mixed-phase freezing and in-situ cirrus nucleation at high relative humidities in the tropical tropopause layer. The suppression of homogeneous nucleation by dust is not observed, which casts additional doubt on the idea of cirrus thinning.

This is a valuable study that adds an evolution perspective on the CloudSAT-Calipso satellite retrievals of cirrus.  I have only a number of minor general comments and some specific suggestions for improvement.

General comments:

1. The previous estimate cited in the manuscript suggests a cirrus cloud lifetime of 15 hours or less. Why are the trajectories in this study run for longer than that? Would the results change if, for example, the trajectory analysis only considered the first 10 hours?
2. The trajectories follow (resolved) air masses. However, formed ice crystals will also sediment out of such trajectories. Long trajectories are thus probably not very useful when trying to explain the properties of hydrometeors with considerable sedimentation rates. For instance, if my back of the envelope calculation is correct, a 40 micron (radius) ice crystal will sediment about 3 km in 10 hours. Could you comment on whether the lack of ice crystal sedimentation biases your results.
3. Dust burdens come from one reanalysis, winds and other properties from another one. This seems to be another important limitation of this study, which could be avoided in new studies. Figure 2 brought this to my attention. The orange trajectory moves quickly upwards, especially at about -18 hours. At the same time, the dust concentration increases by an order of magnitude. This seems implausible.
4. How are different omega regimes defined? Is there a fixed omega threshold to distinguish them?

Also, although this has already been described, using units of Pa/s in a study of cloud properties doesn't seem the best, as I think the cooling rate and cloud properties care about m/s winds. Could you mention how much e.g. 0.1 m/s in omega units can vary between e.g. 10 km and 14 km?

On the other hand,  I don't think that breaking down the results into vertical velocity trajectories adds much value to the study (or at least I haven't noticed it), and could therefore be moved to the appendix.

An overarching question related to comments 1-4:

If you or someone else were to repeat a similar study, what improvements would you make? Is LAGRANTO the best tool given that you are tracking dust (and ERA5 doesn't have dust)?

Specific comments:

Line 15: "unit increase of dust in log space"

Maybe better order of magnitude increase in dust (?)

Line 26:

I don't think you prove/indicate it'll increase the CRE, that seems to be more of a discussion point. In addition, the largest sensitivities to dust are in the tropical tropopause layer, probably associated with the thinnest cirrus. The change in top-of-the-atmosphere CRE may therefore be very small.

=> On the other hand, by thinking outside of the box, the study shows that by increasing dust and modifying ice nucleation in liquid-origin cirrus, the CRE can turn less positive (big question mark here as such cirrus occur at warmer temperatures, so their CRE may be on average dominated by the SW component).

Line 28: "increasing occurrence towards the equator"

Looking from pole to equator, I would say: increased occurrence until you reach the storm tracks, then decreased occurrence in the subtropics, and increased again toward the ITCZ and especially the warm pool area.

Line 30: The definition of cirrus used in this study could be even more explicit.

Line 46: "Their counter part"  ==> that sounds weird, please use a different term.

Lines 38-40: Some other studies classifying cirrus origin may deserve to be cited here, e.g. Muhlbauer et al., 2014 (10.1002/2013JD020035), Sassen and Comstock, 2001 (10.1175/1520-0469(2001)058<2103:AMCCCF>2.0.CO;2) use a dynamical regime classifiction of cirrus.

Lines 90-92:

I miss one sentence about the limitations of the models as a segue to the "we do this and that" part.

Introduction/discussion:

It might be good to mention the aircraft observations by Mingui Diao's group. Their main conclusions, as far as I understand, point in the same direction, i.e. an increase in the number of ice for a larger number of coarse mode aerosols. See e.g. Maciel et al., 2023 (https://doi.org/10.5194/acp-23-1103-2023), or earlier work by Patnaude et al.

Section 3.1:

Assuming that clouds form mainly in the last 10 hours before the satellite overpass, large parts of the hybrid category would rather fit into the in-situ cirrus. On the other hand, the sensitivities make the hybrid category rather similar to the liquid-origin category. Why?

Figure 6:

While this is addressed to some extent in the manuscript, I would like to point out that the chosen domain covers exactly the part of the world with the highest expected dust concentrations, and thus may not be representative of the more remote regions, especially in the southern hemisphere. 

Line 235:

Gravity waves will be an even more common source of nucleation. Some references to consider Atlas and Bretherton, 2023 (https://doi.org/10.5194/acp-23-4009-2023), Chang and L'Ecuyer, 2020 (https://doi.org/10.5194/acp-20-12499- 2020), Kim et al., 2016 (https://doi.org/10.1002/2016GL069293).

Figure 7:

Are there any substantial differences between midlatitude and tropical cirrus, when binned into temperatures like here? It seems to be that one could merge the two regions. This is also a powerful qualitative representation that cirrus properties are, on average, simply controlled by thermodynamics/temperatures.

Figure 9:

Why is the dust concentration so similar between e.g. in-situ and liquid-origin cirrus in the tropics. Their temperatures and thus altitudes are quite different (maybe a delta z of about 3 km on average). Because of that, I would expect a larger difference in dust concentrations.

Line 348:

However, the radiatively most relevant cirrus, at high ICNC, are homogeneously formed based on Froyd et al., 2022.

Lines 366-367:

"the absence of knowledge about when an observed cloud was initially formed and where in the cloud development life cycle it was"

Any suggestion on how to do it better?

Line 369:

But maybe the information about cloud top is all what we need, assuming that this is the most important layer for ice nucleation?

Best regards,

Blaž Gasparini