|This is a review of the revised version of the manuscript entitled “On the relationship between the scattering phase function of cirrus and the atmospheric state” submitted to ACPD by Baran et al. |
I appreciate the efforts the authors made to address the issues that I and the other reviewer raised. However, in my opinion many of the major concerns raised by me and the other reviewer are still not adequately addressed and new information makes the conclusions more questionable.
One concern that I raised was that the definition of a null result was not given. The authors now state in their reply that for the null results “the retrieved spherical albedo at each of the scattering angles was the same for all ensemble models”. This is mathematically impossible. As said in my first review, “the method seeks the lowest rmse (eq. 5) produced by the different models and theoretically one model should lead to the lowest rsme.” The authors most likely choose a lower limit threshold for the rmse difference that separates null results from successful retrievals. This threshold is arbitrary and not given.
The new figure 10 shows fits for a “pristine” case and a fully distorted case. Here, I would judge the results for the “pristine” case (a) to be a null result, because none of the models can fit the data, and rmse differences between models are small compared to the absolute rmse value. For the fully distorted case (b) the rmse for the fully distorted model is a factor 4 smaller than that for the pristine model. For the pristine case (a), the relative difference in rmse for the pristine model and the fully distorted model is merely a factor 1.3. At scattering angles near 125 degrees all models strongly deviate from the measurements and hardly any difference between the model fits can be seen.
I strongly suspect that the results shown in Fig. 10a are affected by the lower lying liquid clouds. The liquid drops produce a strong rainbow at 125-150 degrees and lower phase function values than ice at scattering angles around 100, just as observed in Fig. 10a. The authors state in their reply that “if sheets of water cloud were underneath the cirrus, then the fits to the spherical albedo differences would have been much worse than is shown in Figure  below.” This is postulated but not shown. I would agree that for a case with only liquid clouds the rmse values would probably be much worse, but for a case with cirrus overlaying broken liquid clouds I would expect the liquid cloud signature (e.g. rainbow) to be visible but at reduced strength. Simulated data is needed to investigate the effect of broken liquid clouds under the cirrus.
The lidar data that is shown in the authors’ reply clearly show the liquid clouds underneath the cirrus at some locations. This is acknowledged in the new manuscript but the lower half of the lidar figure showing this is still omitted in the paper. At some times (i.e., 13.35h collocating with the “pristine” cases) the lidar is saturated and liquid clouds cannot be detected by the lidar, but this does not mean that they are not there. Furthermore the lidar data do not correspond perfectly with the POLDER data in terms of timing as noted by the authors. There are clearly broken liquid clouds here and there under this cirrus deck. The data cannot show that the “pristine” cases are not contaminated by liquid clouds.
In summary, the paper claims the presented technique is proven using statistics of 12 pixels, while it is very likely that these pixels are contaminated by liquid clouds. It is not shown that such broken liquid clouds do not affect the method, which they probably do. None of the models, including the pristine model, is able to fit the data for the “pristine” cases sufficiently. Finally the way successful results are separated from null results is still not sufficiently explained while this seems to affect the conclusions crucially.
In my original review I said that “only when the method is adequately demonstrated and more statistics are included and when all comments below are adequately addressed in the text, I would recommend the paper to be published in ACP.” I acknowledge the effort made to improve the paper but I regret to say that the method is still not adequately demonstrated, new information makes the results even less convincing and the necessary improved statistics were not included.
Unfortunately, even if manuscript would be improved by adding a study on the influence of broken liquid clouds under cirrus, by properly defining the null results, and by improving the statistics, the insufficient fit of all models, including the pristine model, to the data (now shown in figure 10a) demonstrates that this method cannot adequately be applied to this particular dataset.
Evidence of a correlation between RH and ice particle pristineness would be very interesting and would potentially help to better constrain radiative transfer calculations in models. The data in this paper unfortunately do not deliver that evidence and in my view would misguide further research. I therefor strongly recommend against accepting this paper in ACP in the current form.