The reviewers have now commented on the revisions. There are minor suggestions from each after which there will be an editor review only. The comments are:

1. One minor recommendation is that on p. 3, l. 12, please spell out what INUIT stands for.

2. The figure 12 and the discussion thereof was replaced by table 2 in the revised manuscript, but the issue of using median freezing temperature (T50) for comparison of the experimental data has not been clarified. Table 2 of the revised manuscript provides values of T50 obtained with wind tunnel (isothermal measurements), acoustic levitator (nonlinear cooling rate) and cold stage data of (Knopf et al., Nature Geoscience 2011, constant cooling rate 5 K/min). The T50 values, however, are very sensitive to the variation of cooling rates, morphology of particles populating the droplets (size distribution, variation of particle number per droplet, BET surface area, agglomeration state, etc.), and to the counting statistics. To some degree this is demonstrated by the Table 2 itself, if one considers the values of T50 for “present results illite NX” obtained with two different instruments (WT and AT). One would expect a general increase of the median freezing temperature with increasing particle surface area (from 5e-6 to 7e-6 m2/drop) but the T50 actually drops by 2K! This discrepancy cannot be incorporated by 1K error stated in the manuscript (section 3.1.3 of the revised manuscript) and therefore I disagree with the conclusion that “the deviations between the results from the two techniques are within the measurement uncertainties” as stated at the end of section 3.1.2 of the revised manuscript.

To my opinion it is totally sufficient to compare the values of nucleation rate coefficient and /or surface number densities of ice active sites, as it is done in the figures 7, 9 and 10. If comparison with the data of Knopf et al. is absolutely necessary, it is straightforward to recalculate their data in terms of rate coefficient or surface site density. Therefore, I suggest removing the table 2 and the discussion of median freezing temperature.

2. Calculation of the nucleation rate coefficient from the wind tunnel isothermal measurements.

As it is stated the section 3.2.1, page 12 of the revised manuscript, the droplets frozen faster than 5 seconds from the injection into the wind tunnel “were not considered”. Could you formulate this statement more precisely? Does that mean that these droplets were excluded from consideration completely or that they were excluded from the fitting algorithm used to determine the nucleation rate coefficient (straight dotted lines in figure 6)? If the latest is true, what is the reason for pinning the fit function to the value of 5 seconds on the time axis? This constrain obviously introduce a bias into the slope of the fit line, like it is seen in the case for -21°C of figure 6b. Besides, how should one interpret the data for illite NX measured below -23°C, where almost all data points lie before the “adaptation line”?

To my understanding, if you want to exclude the data points with unknown freezing temperature (freezing times shorter than 5 sec) you should have excluded these experiments from consideration completely, as if they have never been carried out. Only in this case neither the N_total, nor the number of unfrozen droplets would be biased. Excluding them from counting as "still liquid" (n_liq) but not excluding them from the total number of observed droplets (N_total) introduces the error in the determination of nucleation rate (this is what, I believe, has happened). Please consider reevaluation of the calculated values of nucleation rate coefficients.

Last reviewer comments now considered. Paper is ready for publication.