|I thank the authors for responding to my comments. In my opinion, the manuscript has improved, however, unfortunately, some of the analyses still do not support the conclusions and are rather speculative. I also do not see how tuning of WBF or surface heat fluxes result in overall improvement of the model or that it is physically justifiable. Many of my major original concerns regarding the manuscript still remain — in particular, I am still not convinced that the Arctic low-cloud bias exists in the first place without the further analysis I suggested in the first round of revisions. The cloud phase comparison is also not meaningful as the scales are different, and the new figure only further strengthens my concern regarding the impact of the WBF tuning on cloud phase partitioning. More details follow below.|
* The authors are missing the point of my first comment — I am aware that the authors used the COSP simulator, but my point is that since the focus of this study is on low-level clouds, the lowest 1-2 km of Arctic clouds will not be visible by GOCCP or the model with the COSP simulator on, whereas these clouds may exist according to ground-based observations (see the references listed in my original comment). Just because the COSP or GOCCP cannot “see” those low-clouds, does not mean that they are not actually there. Again, nothing prevents the authors from analyzing collocated Arctic ground-based observations to get a sense of the extent to which the claimed underestimate of low-cloud fraction actually holds, which was part of my original recommendation. At this point, I’m still not convinced that this high bias in cloud cover exists in the first place.
* I would recommend that the authors exclude the year 2007 and include another year (2011) to avoid the issue of the change in nadir viewing angle. That way, the comparison between model and observations would at least be more consistent.
* Comparison between Figure 6 and the figure from Korolev et al. (2017) is not meaningful because the latter observations were spatially averaged over 100 m, while the GCM grid box used in the manuscript has a horizontal resolution that is orders of magnitude larger. In fact, the authors seem to acknowledge the difference in spatial scales. A comparison with satellite observations would therefore be more of an apples to apples comparison. Furthermore, the plot over Siberia only adds to my initial concern that the WBF tuning exacerbates the phase partitioning problem by increasing ice fraction to values close to 1.0 between temperatures -5C and -35C for the lower threshold value plotted.
* I see the logic of the authors in comparing profiles of temperature and relative humidity to the model to support their claim of the low-cloud overestimate, but because there is no clean causal relationship that can be teased apart, this additional piece of information is too speculative and not very useful in my opinion.